postharvest quality changes of leafy green...
TRANSCRIPT
Postharvest quality changes of leafy green vegetables - assessed by respiration rate, sensory analysis, multispectral imaging, and chemometrics
PhD thesis by Mette Marie Løkke March 2012
Department of Food Science Aarhus University
ii
Main supervisor
Associate professor Merete Edelenbos
Department of Food Science, Aarhus University
Co-supervisors
Post doc Helene Fast Seefeldt
Department of Food Science, Aarhus University
Head of research unit, senior scientist Anders Peter Adamsen
Department of Engineering, Aarhus University
Assessment committee
Head of research unit, senior scientist Anette K. Thybo (chairman)
Department of Food Science, Aarhus University
Associate professor Åsmund Rinnan
Department of Food Science, University of Copenhagen
Senior research fellow Pramod Mahajan
Department of Process and Chemical Engineering
College of Science, Engineering and Food Science, University College Cork, Ireland
iii
I. Preface
This PhD thesis finalizes three years’ work at the Department of Food Science, Aarhus University.
The study was part of the project, ‘Product designed packaging of fresh fruits and vegetables’, an
Innovation Consortium financed by the Danish Agency for Science, Technology and Innovation
and the Danish Council for Technology and Innovation.
First, I would like to thank my supervisors Merete Edelenbos, Helene Fast Seefeldt and Anders
Peter Adamsen for valuable discussions and help in relation to this project. I wish to thank Merete
Edelenbos for guidance throughout the project and for bringing me into the world of postharvest
biology and technology. I would like to acknowledge Anders Peter Adamsen for hosting me from
time to time in the group of Air Quality Engineering at the Department of Engineering in Foulum
and for introducing me to Anders Feilberg, who is also thanked for expert support during the PTR-
MS measurements and interpretation of the results. A very warm appreciation goes to Helene Fast
Seefeldt for always being there for me offering encouragement, discussions, and practical help. I
cannot thank you enough; you are second to none.
A number of collaborators in the Innovation Consortia generously offered material and time for
which they are greatly acknowledged: Yding grønt A/S, ScanStore Packaging A/S, Axel Månsson
A/S, Videometer A/S, and the Danish Technological Institute.
I wish to thank Jens M. Madsen for being a good and cheerful travel companion during the
postharvest course in California, it was a memorable trip. Furthermore, Jens M. Madsen and Helene
Fast Seefeldt made it fun to be in the lab in spite of smelly wild rocket and broccoli.
Thomas Skov, Department of Food Science, University of Copenhagen is acknowledged for many
fruitful discussions about chemometric issues throughout the three years.
Many persons helped me during the last intensive phase of writing this thesis, and I really
appreciate their efforts. I am especially grateful to my two reviewers, Helene Fast Seefeldt and
Morten Rahr Clausen, for their advice and constructive critique. I wish to thank Ellen Mortensen,
Anne H. Balling and Aase Sørensen for thorough and competent proofreading. I am grateful to
Grith Mortensen for her extraordinary support during these few months she has been my group
leader.
I wish to thank all my former and present colleagues at the Department of Food Science at Aarhus
University for being motivated and committed, and especially Anette Thybo for being a wonderful
leader, always with an open door facilitating an open-minded working environment. It has been a
true pleasure working with all of you in Aarslev!
Finally, I would like to express my deepest appreciation and love for family and friends. My parents
are deeply acknowledged for being a back-up in everything; my mother for practical assistance and
iv
my father for his scientific insight. A special thanks goes to my mother-in-law, Helle Ammitzbøll,
for housing me once a week for three years, I will really miss our Wednesday evenings. Last but not
least, my appreciation goes to Mikael and Laura. I am grateful for your support and patience
through the last three years.
Mette Marie Løkke, Aarhus, March 2012
v
II. Abstract
The postharvest industry is expected to supply good quality fruit and vegetables to consumers
throughout the year from various and distant fields. Postharvest products are living and respiring
materials, and from field to fork, several factors influence quality. In this thesis, focus is mainly on
leafy green vegetables. The quality of these delicate products peaks at harvest when they are green
and crisp, but during handling and storage their quality attributes may be deteriorated, resulting in
products of inferior quality when they reach the consumers. The primary aim of this thesis was to
determine the quality changes of leafy green vegetables as effect of harvest season, packaging,
and storage temperature. A novel, continuous measurement of O2 concentration was used for
determination of respiration rate, which is a key measure in the metabolism of the living product.
Furthermore, sensory analysis was used for describing quality changes, and multispectral imaging
was used for prediction of sensory attributes. The work is described in a series of papers:
Paper I deals with a novel analytical method for measurement of the respiration rate. A closed
system for respiration rate measurement was combined with a wireless sensor concomitantly
detecting O2 concentration and temperature. The respiration rate was calculated based on these
continuous values. The setup provided a unique insight into the close connection between
temperature and respiration rate. Optimization of the sensors will improve the method, especially
concerning measurements below 6.3 kPa O2 and measurements of CO2.
Paper II shows the application of the developed analytical method from Paper I for investigation of
the seasonal differences in the respiration rate of wild rocket (Diplotaxis tenuifolia L.) and seasonal
and varietal differences in the respiration rate of broccoli florets (Brassica oleracea, Italica Group)
stored at 5, 10, or 20 °C. By partial least squares regression (PLS), it was found that the most
influential factor on the respiration rate for both commodities was temperature followed by season,
variety, and storage duration. O2 concentration in the studied range of 6-20 kPa was practically
nonsignificant for respiration rate of wild rocket leaves and broccoli florets.
Paper III elucidates the effect of temperature, storage duration, and packaging on the sensory
quality of wild rocket. As freshness is considered to be the most important quality parameter for
leafy green vegetables, the study investigated how to describe freshness using sensory attributes.
Storage at 20 °C for up to six days had a high impact on freshness irrespective of other factors. The
signs of degradation varied according to the packaging material, and storage at high oxygen
transmission rate (OTR) (48.400 ml O2/m2/day/atm) led to indications of senescence with yellow
leaves being most evident. Low OTR (1900 ml O2/m2/day/atm) led to a distinct off-odor indicating
fermentative (anaerob) metabolism. At first sight, low OTR gave the highest freshness score, but
after opening the package there was no difference in freshness between the two packaging
materials. By interval PLS (iPLS), it was noted that the overall freshness was determined by the
vi
sensory attributes color, texture, and odor, and consequently could not be described by color alone,
as first anticipated by initial freshness.
Paper IV describes a rapid multispectral imaging method which was used for prediction of sensory
attributes described in Paper III. By multispectral imaging, the areas of interest – the leaves – could
be extracted from the background. The combination of wavelengths in the visual (VIS) and the near
infrared (NIR) range of the electromagnetic spectrum made it possible to predict attributes of
importance for freshness, as VIS provided information on color changes and NIR on textural
changes. This method could be used for quality determination in postharvest research as well as in
quality control.
Summarizing the results of the PhD study, multispectral imaging proved to be a promising
analytical method for measurement of color and textural quality of leafy green vegetables (Paper
IV). However, multispectral imaging did not quantify odor, which was also important for perception
of overall freshness (Paper III). Based on unpublished results, proton transfer reaction-mass
spectrometry (PTR-MS) was suggested as a potential method for rapid quantification of odor.
Temperature was the most important single factor among several factors influencing postharvest
quality. However, none of the factors should be studied individually, since they interact with each
other. By the use of continuous measurement of multiple factors, the research demonstrated that
both the improved resolution in the time domain and the inclusion of several factors resulted in a
description and understanding of the postharvest degradation of the quality of leafy green
vegetables. Continuous measures and multispectral images generate large data matrices, and data
analysis by chemometrics gives an intuitive and graphical interpretation. However, often rapid
methods cannot stand alone, and during development of new rapid methods it is necessary to
calibrate using another method, e.g. sensory analysis.
Packaging effectively prevents water loss; however it also modifies the packaging atmosphere.
Comparison of results in Paper II, Paper III, and in unpublished results of the respiration rate
revealed that it may not be possible to attain a packaging atmosphere that is beneficial for lowering
the respiration rate of wild rocket. If the storage temperature cannot be kept low, high OTR is
recommended as too low permeability causes quality loss due to anaerobic respiration.
This thesis truly substantiates the established basic recommendations to keep the temperature low
from field to fork to ensure high quality of leafy green vegetables.
vii
III. Resumé
Frugt- og grøntbranchen forventes at levere god kvalitet til forbrugerne hele året rundt fra
forskellige og fjerne steder. Frugt og grønt er levende og respirerende produkter, og på rejsen fra
jord til bord har flere forskellige faktorer indflydelse på kvaliteten. I denne afhandling er fokus
primært på grønne bladgrøntsager. Kvaliteten af disse sarte produkter topper ved høst, hvor de er
grønne og sprøde, men under håndtering og opbevaring kan kvalitetsegenskaber forringes inden
produkterne når frem til forbrugerne. Det primære mål med denne afhandling er at bestemme
kvalitetsændringer i grønne bladgrøntsager som effekt af høsttidspunkt, emballering og
opbevaringstemperatur. En ny kontinuerlig måling af O2-koncentration blev anvendt til
bestemmelse af respirationsraten, som er et vigtigt mål for produktets metabolisme. Endvidere blev
sensorisk analyse anvendt til at beskrive kvalitetsændringer, og multispektrale billeder blev brugt til
forudsigelse af de sensoriske egenskaber. Arbejdet er beskrevet i en række videnskabelige artikler:
Artikel I beskriver en ny analysemetode til måling af respirationsraten. Et lukket system til måling
af respirationsrate blev kombineret med en trådløs sensor, der detekterede O2-koncentration og
temperatur. Respirationshastigheden blev beregnet på baggrund af disse kontinuerte værdier, og
opsætningen gav et unikt indblik i den tætte sammenhæng mellem temperatur og respirationsrate.
Optimering af sensorerne vil dog forbedre målemetoden, især hvad angår målinger under 6,3 kPa
O2 og målinger af CO2.
Artikel II viser anvendelsen af den udviklede analysemetode fra Artikel I til undersøgelse af
sæsonvariation i respirationsrate af vild rucola (Diplotaxis tenuifolia L.) samt sæsonvariation og
sortsforskelle i respirationsrate af broccolibuketter (Brassica oleracea, Italica Group) opbevaret ved
5, 10 eller 20°C. Ved hjælp af partial least squares regression (PLS) blev det konstateret, at den
mest indflydelsesrige faktor på respirationsraten for begge grøntsager var temperatur efterfulgt af
høstsæson, sort og varighed af opbevaring. O2-koncentrationen i det undersøgte interval på 6-20
kPa var praktisk talt ikke væsentlig for respirationsraten af vild rucola og broccolibuketter.
Artikel III anskueliggør effekten af temperatur, opbevaringsvarighed og emballering på den
sensoriske kvalitet af vild rucola. Friskhed anses for at være den vigtigste kvalitetsparameter for
grønne bladgrøntsager, og i studiet blev det undersøgt, hvordan friskhed kan beskrives ved hjælp af
sensoriske egenskaber. Opbevaring ved 20°C i op til seks dage havde en stor indflydelse på
friskheden uafhængigt af andre faktorer. Symptomerne på nedbrydning varierede afhængig af
emballagen, og opbevaring ved høj ilt-permeabilitet (OTR) (48.400 ml O2/m2/dag/atm) førte til tegn
på aldring med gule blade som det mest tydelige tegn. Lav OTR (1900 ml O2/m2/day/atm) førte til
en udtalt ubehagelig lugt, som indikerede fermentativ (anaerob) metabolisme. Ved første øjekast
gav lav OTR den højeste friskhedsscore, men efter at have åbnet pakken var der ingen forskel i
friskhed mellem de to emballagetyper. Ved hjælp af interval PLS (iPLS) blev det konstateret, at de
underliggende sensoriske egenskaber for den samlede vurdering af friskhed består af både farve,
viii
tekstur og lugt. Friskhed kunne derfor ikke beskrives ved farve alene, som lå til grund for
bedømmelsen af friskhed ved første øjekast.
Artikel IV beskriver en hurtigmetode, hvor multispektrale billeder bruges til bestemmelse nogle af
de sensoriske egenskaber, der er beskrevet i Artikel III. Ved hjælp af multispektrale billeder kan de
interessante områder af billederne - bladene - isoleres fra baggrunden. Kombinationen af
bølgelængder i det synlige (VIS) og det nær-infrarøde (NIR) område af det elektromagnetiske
spektrum gjorde det muligt at forudsige egenskaber af betydning for friskhed, idet VIS indeholdt
oplysninger om farveændringer og NIR om teksturændringer. Denne metode vil kunne bruges til
kvalitetsbestemmelse i forskning samt i kvalitetskontrol.
Sammenfattende viste resultaterne af ph.d.-studiet, at multispektrale billeder er en lovende
analysemetode til kvalitetsmåling af farve og tekstur af grønne bladgrøntsager (Artikel IV).
Imidlertid kan multispektrale billeder ikke kvantificere lugt, som også er vigtig for den samlede
opfattelse af friskhed (Artikel III). Baseret på ikke-offentliggjorte resultater blev kemisk ionisering
med massespektrometri (PTR-MS) foreslået som en mulig metode til hurtig kvantificering af lugt.
Temperatur er den vigtigste enkeltfaktor blandt flere faktorer, der påvirker kvaliteten efter høst.
Imidlertid bør faktorerne ikke undersøges individuelt, idet de influerer på hinanden. Ved anvendelse
af kontinuerlig måling af flere faktorer viste forskningen, at den forbedrede opløsning i
tidsdomænet og optagelsen af flere faktorer samtidig resulterede i en bedre beskrivelse og forståelse
af nedbrydning af kvaliteten af bladgrøntsager efter høst. Kontinuerte målinger og multispektrale
billeder genererer store datamatricer, og dataanalyse vha. kemometri giver en intuitiv og grafisk
fortolkning. Imidlertid kan hurtige metoder ofte ikke stå alene, og under udvikling af hurtige
metoder er det nødvendigt at kalibrere mod en anden metode, fx sensorisk analyse.
Emballering er en effektiv metode til at forebygge vandtab, men det modificerer samtidig
atmosfæren i pakken. En sammenligning af resultaterne i Paper II, Papir III og i ikke-offentliggjorte
resultater af respirationsraten viste, at det formentlig ikke er muligt at opnå en gassammensætning
inde i pakningen, som fremmer reduktion af respirationsraten af vild rucola. Hvis
opbevaringstemperaturen ikke kan holdes lav, anbefales det at bruge høj OTR, da lav OTR
forårsager kvalitetforringelse som følge af anaerob respiration.
Konkluderende kan det konstateres, at denne afhandling underbygger de etablerede grundlæggende
anbefalinger om at sikre en lav temperatur fra mark til bord for at bevare en høj kvalitet af grønne
bladgrøntsager.
ix
IV. List of publications
Paper I: Løkke, M.M., Seefeldt, H.F., Edwards, G. and Green, O. (2011). Novel wireless sensor
system for monitoring oxygen, temperature and respiration rate of horticultural crops post harvest.
Sensors, 11, 8456-8468.
Paper II: Seefeldt, H.F., Løkke, M.M. and Edelenbos, M. (2012). Effect of variety and harvest time
on respiration rate of broccoli florets and wild rocket salad using a novel O2 sensor. Postharvest
Biology and Technology, 69, 7-14.
Paper III: Løkke, M.M., Seefeldt, H.F. and Edelenbos, M. Freshness and sensory quality of
packaged wild rocket. SUBMITTED to Postharvest Biology and Technology.
Paper IV: Løkke, M.M., Seefeldt, H.F., Skov, T. and Edelenbos, M. Color and textural quality of
wild rocket measured by multispectral imaging. SUBMITTED to Postharvest Biology and
Technology.
Additional peer reviewed publications not included in the thesis:
Raju, C.S., Løkke, M.M., Sutaryo, S., Ward, A.J. and Møller, H.B. (2012). NIR Monitoring of
ammonia in anaerobic digesters using a diffuse reflectance probe. Sensors, 2012, 12, 2340-2350.
Green, O., Bartzanas, T., Løkke, M.M., Jørgensen, O.J., Tortajada, V.G., Bochtis, D.D. and
Sørensen, C.G. (2012). Spatial and temporal variation of temperature and oxygen concentration
inside silage stacks. Biosystems Engineering, 111, 155-165
Kildegaard, H., Løkke, M.M. and Thybo, A.K. (2011). Effect of increased fruit and fat content in an
acidified milk product on preference, liking and wanting in children. Journal of Sensory Studies, 26,
226-236.
x
V. List of abbreviations
ATP Adenosine TriPhosphate
CO2 Carbon Dioxide
FADH2 Flavin Adenine Dinucleotide – reduced form
GC-MS Gas Chromatography-Mass Spectrometry
GC-O Gas Chromatography-Olfactometry
GC-SCD Gas Chromatography-Sulfur Chemiluminescence Detection
iPLS interval Partial Least Squares
IR Infra Red
NADH Nicotinamide Adenine Dinucleotide – reduced form
NIR Near InfraRed
O2 Oxygen
OTR Oxygen Transmission Rate
PC Principle Component
PCA Principle Component Analysis
PLS Partial Least Squares (regression)
PTR-MS Proton-Transfer-Reaction Mass Spectrometry
r Correlation Coefficient
RH Relative Humidity
RMSECV Root Mean Squared Error of Cross Validation
RMSEP Root Mean Squared Error of Prediction
RQ Respiration Quotient
SSC Soluble Solids Content
TCA Tricarboxylic Acid
VIS Visual (wavelengths)
VOC Volatile Organic Compound
xi
VI. Table of contents
I. Preface ......................................................................................................................................... iii
II. Abstract ..................................................................................................................................... v
III. Resumé .................................................................................................................................... vii
IV. List of publications................................................................................................................... ix
V. List of abbreviations.................................................................................................................. x
VI. Table of contents ...................................................................................................................... xi
1 Introduction ................................................................................................................................... 1
1.1 The concept of quality ........................................................................................................... 1
1.2 Postharvest biology and technology ...................................................................................... 1
1.3 Quality assessment ................................................................................................................ 2
1.4 Objectives .............................................................................................................................. 3
1.5 Outline of the thesis ............................................................................................................... 3
2 Postharvest quality changes in leafy green vegetables ................................................................. 4
2.1 Vegetables ............................................................................................................................. 4
2.2 The plant cell ......................................................................................................................... 5
2.3 Respiration............................................................................................................................. 8
2.4 Leaf structure ....................................................................................................................... 11
2.5 Senescence of leafy green vegetables ................................................................................. 12
2.5.1 Color changes ............................................................................................................... 15
2.5.2 Textural changes .......................................................................................................... 17
2.6 Effect of harvest season, variety, and aging on respiration rate .......................................... 19
3 Effect of postharvest technology on quality ............................................................................... 23
3.1 Processing ............................................................................................................................ 23
3.2 Temperature and moisture ................................................................................................... 25
3.3 Packaging ............................................................................................................................ 27
3.3.1 Effect of gas composition on respiration rate .............................................................. 28
3.3.2 Modified atmosphere packaging (MAP)...................................................................... 31
3.3.3 Modeling respiration rate ............................................................................................. 31
3.3.4 Effect of gas composition on produce quality ............................................................. 32
xii
4 Postharvest quality assessment ................................................................................................... 38
4.1 Measurement of respiration rate .......................................................................................... 38
4.2 Sensory descriptive analysis ................................................................................................ 42
4.3 Analysis by visual and near infrared light ........................................................................... 43
4.3.1 Visual wavelengths ...................................................................................................... 43
4.3.2 NIR wavelengths .......................................................................................................... 45
4.3.3 Imaging ........................................................................................................................ 46
4.4 Analysis of odor .................................................................................................................. 48
5 Chemometrics in postharvest quality assessment ....................................................................... 51
5.1 Data exploration: PCA ........................................................................................................ 52
5.2 Multivariate regression: PLS ............................................................................................... 55
5.3 Variable selection by iPLS .................................................................................................. 57
5.4 Validation, model accuracy, and robustness ....................................................................... 59
6 Conclusions and perspectives ..................................................................................................... 61
7 References ................................................................................................................................... 64
1
1 Introduction
Through history fruits and vegetables has been something we gathered in nature or cultivated
ourselves. Today, fruits and vegetables are mostly something we buy in the supermarket, and
regardless of season we expect nearly all kinds to be available and of good eating quality
throughout the year. This implies challenges for the fruits and vegetable industry that is expected to
supply good quality produce throughout the year from various and distant production areas.
1.1 The concept of quality
Quality is an ambiguous concept, and many definitions have been suggested. Quality has been
described as a degree of excellence, a high standard, or value (Barrett et al., 2010). Additionally,
quality is strongly item dependent. For instance, melon does not have the same quality parameters
as broccoli. Quality also has to do with both safety and originality (Martens and Martens, 2001).
Quality parameters of fruits and vegetables will vary with the commodity, its intended use, and the
preferences of the consumer (Saltveit, 2003b). Moreover, the perception of quality is learned
criterion and will therefore vary from person to person (Wills et al., 2007). Since quality is such an
ambiguous concept, it may be determined by many different characteristics or attributes, and
combinations of different attributes and their relative weight and interactions further complicate the
formulation of quality criteria (Saltveit, 2003b). In the current thesis, freshness was used as the
main parameter for quality.
Freshness is probably the most important quality parameters for fresh fruits and vegetables,
(Lappalainen et al., 1998; Peneau et al., 2006; Peneau et al., 2009; Ragaert et al., 2004). Freshness
of fruits and vegetables is not a clearly defined term; nevertheless, consumers define it as a very
important parameter for the desire to actually consume fruits and vegetables (Peneau et al., 2006).
The word ‘freshness’ originates from the Latin word priscus ‘close to the origin’ (Peneau et al.,
2009) and is related to the overall perception of product quality and not to a single sensory attribute.
1.2 Postharvest biology and technology
Postharvest is the period from harvest of the produce until consumption. Postharvest quality of high
standards is therefore often measured in terms of freshness. Peneau et al. (2009; 2007; 2006)
demonstrated that consumers’ definition of freshness of fruits and vegetables is associated with
sensory determined appearance, texture and physiological age of the product at consumption.
Consumers often buy products the first time based on appearance, but repeated purchases are driven
by expected quality factors determined by flavor compounds and texture experienced from last
purchase (Barrett et al., 2010). At the point of purchase, the consumer uses appearance factors to
provide an indication of freshness of the packaged produce. In the future, cameras in smart phones
may assist consumers in judging the quality at the point of purchase. Recently, a study was
published of the use of a mobile phone for quality monitoring in the supermarket; despite great
2
potentials, the method needs further development before it is ready for implementation for everyday
quality determination in the supermarket (Iqbal and Bjorklund, 2011).
Postharvest biology constitutes the biological changes occurring during the postharvest period,
while postharvest technology encompasses the technology that can be applied to enhance positive
changes during ripening and to postpone negative changes in quality. In postharvest science, it is
important to understand how to preserve good product quality until the products reach the point of
consumption. An optimal storage environment can be defined as those conditions that retain the
best quality product (Saltveit, 2003b). Many factors during storage influence postharvest quality.
Biological factors cause a large variation in initial quality at harvest and in the reaction to the
storage environment. Biological factors will be referred to as biotic factors and cover responses due
to genetic and physiological factors. Abiotic factors are factors which are applied during handling
and storage and which can to some extent be controlled technologically, e.g. by packaging and
refrigeration.
Although harvested and detached from the growing plant, postharvest products are living materials,
and in the preparation process from harvest to the product is ready for transport to retail, several
processing steps may induce a stress reaction in the product and thereby influence quality.
Furthermore, the storage conditions, such as temperature, relative humidity, and gas composition,
are important for the final quality. Leafy green vegetables are delicate products with a short shelf
life, and their quality is easily degraded when stored under the wrong conditions. Leafy green
vegetables were chosen as the main subject of investigation in this thesis.
1.3 Quality assessment
Perception of quality can, in its nature, only be performed by humans, although humans are costly
to pay and difficult to use consistently in everyday measurements (Martens and Martens, 2001).
Consequently, different analytical techniques are often used in an attempt to objectively quantify
quality changes experienced during storage. Postharvest products show high biological variation
and change through handling, making a preservation of the original tissue difficult (Smith et al.,
2003). Hence, rapid analytical techniques are favored, and due to the fact that numerous samples are
needed to cover the natural variation, the use of rapid analytical techniques results in increased
interest in postharvest research (Mizrach, 2008; Nicolai et al., 2007; Ruiz-Altisent et al., 2010).
Furthermore, the ability to measure larger quantities of samples quickly facilitates investigation of
additional combinations of storage conditions. This will lead to a better understanding of the effect
of biological variation and storage conditions on the quality within postharvest science.
Additionally, rapid analytical methods are interesting for the postharvest industry with a view to
quality control.
3
Often, rapid methods generate enormous amounts of data, and when many combinations of
treatments are tested (multifactorial), the generated data are immense; this implies some challenges
in interpreting the data thoroughly. Multivariate data analysis (chemometrics) gives an overview of
the underlying patterns in large data matrices in a quick and visual manner, which can facilitate the
use of the data. Postharvest science is such a multifactorial science, and it is obvious to use
chemometrics for handling the plentiful data generated.
1.4 Objectives
The overall objective of this thesis is to determine the quality changes of leafy green vegetables as
effect of harvest season, packaging and storage temperature. The feasibility of using rapid analytical
methods and chemometrics in some postharvest applications is further investigated in this thesis. It
is hypothesized that quality changes of leafy green vegetables can be quantified by rapid analytical
methods and chemometrics. Sensory analysis was used for describing the quality changes, and rapid
methods were developed to quantify the experienced changes. Four topics have been studied in this
thesis. The first aim was to develop a rapid method for determining the respiration rate (Paper I),
which is a key measure for postharvest quality. Secondly, the method was used to investigate
seasonal and varietal differences in respiration rate of wild rocket and broccoli (Paper II).
The third objective was to describe and quantify changes experienced in packaged wild rocket by
sensory descriptive analysis and investigate the underlying sensory attributes for freshness (Paper
III). The fourth aim was to apply a rapid method for prediction of the sensory perception of color
and texture changes of packaged wild rocket (Paper IV).
1.5 Outline of the thesis
The overall purpose of this thesis is to discuss the background of the quality changes and the results
obtained in the present PhD project in relation to established research.
• Chapter 2 gives an introduction to postharvest biology and the biological explanations of
quality changes occurring during postharvest aging with a focus on leafy green vegetables
• Chapter 3 discusses the effects of postharvest technology on quality with a focus on
processing (preparation), temperature and packaging
• Chapter 4 describes some methods for postharvest quality assessment of leafy vegetables,
with focus on the methods used during this PhD project
• Chapter 5 introduces some common methods in chemometrics and discusses the use of these
methods in postharvest quality assessment
• Chapter 6 concludes and puts the obtained results into perspective.
4
2 Postharvest quality changes in leafy green vegetables
Keeping the quality of fruit and vegetables is a major topic for research in postharvest science. This
chapter will introduce some of the challenges in postharvest research regarding quality aspects of
leafy green vegetables.
2.1 Vegetables
Fruit and vegetables come from different parts of the plant and are harvested at different times of
the plant´s lifecycle, and this has an impact on quality and storage life (Wood et al., 2005).
Vegetables represent no specific botanical group and exhibit a wide variety of plant structures
(Figure 1): Vegetative tissue includes leafy vegetables, stem vegetables, roots, tubers and bulbs.
Flower buds, fruit, seeds and grains are reproductive tissues (Wood et al., 2005). Vegetables can,
however, be grouped into three main categories concerning potential shelflife: seeds and pods; roots
and tubers; flowers, buds, stems and leaves, where the latter has the shortest storage life (Wills et
al., 2007).
Figure 1. Origin of some vegetables from plant tissue. The letters indicate the principal origins of representative vegetables as follows: (A) flower bud, (B) stem sprout, (C) seeds, (D) axillary bud, (E) petiole, (F) bulb (underground bud), (G) stem tuber, (H) swollen root, (I) swollen root tuber, (J) swollen hypocotyls, (K) swollen leaf base, (L) leaf blade, (M) fruit, (N) swollen inflorescence, (O) main bud. From Wills et al. (2007).
5
The main focus of this thesis is on leafy green vegetables – the so-called baby leaves represented by
wild rocket (Diplotaxis tenuifolia L.). The tissue of baby leaves is soft, as they are harvested at an
immature development state of the lifecycle of the plant (Martinez-Sanchez et al., 2012).
Furthermore, as baby leaves are in the growth state, they generally have a high metabolic rate
(Salisbury and Ross, 1992). For instance, the wild rocket leaves investigated in Paper II in the late
summer harvest were grown in the summertime and were harvested just four weeks after sowing.
The expected lifetime from sowing to maturity of seeds is approximately 4 months (Delauran et al.,
2005).
The common feature of green vegetables is the green color arising from the green pigment
chlorophyll, and the presence of the green color is a quality aspect in these types of vegetables.
Green vegetables are from different botanical groups, e.g. green asparagus, green broccoli, Brussels
sprout, leafy vegetables, lettuces, and kale; all have different chemical composition as seen from
Table 1.
Table 1. Chemical composition of selected green vegetables.
Vegetables Family Water Total sugars Fibre Protein Total Vit C
g/100g g/100g g/100g g/100g mg/100g
Spinach1 Amaranthaceae 91.7 3.4 1.9 2.6 54
Brussels sprout1 Brassicaceae 83.1 10.5 4.1 4.8 140
Iceberg lettuce1 Asteraceae 95.7 3.2 1.1 1.0 8.1
Curly kale1 Brassicaceae 81.6 10.4 6.2 4.7 169
Broccoli1 Brassicaceae 87.8 6.1 3.2 5.3 121
Green asparagus1 Asparagaceae 92.5 4.9 1.8 1.8 10
Wild rocket Brassicaceae 92.02 503
1: Fødevaredatabanken, www.foodcomp.dk
2: Average from Paper II.
3: Martinez-Sanchez et al. (2006)
In summary, baby leaves are vegetable vegetative tissue with a limited storage reserve of nutrients.
Baby leaves share the green color pigment chlorophyll as common feature with other green
vegetables and the green color is a very important quality aspect.
2.2 The plant cell
Most postharvest quality changes are due to changes at the cellular level. Therefore, a brief outline
of the structural components and a few important organelles in the plant cell will be given here
(Figure 2).
6
The plant cell is bounded by a rigid primary cell wall surrounding the cytoplasm, which comprises
a fluid matrix of proteins, other macromolecules and various solutes. Neighboring cells often have
small channels (plamodesmata) between them connecting the cytoplasma. The cell wall is
permeable to water and solutes. An important function of the cell wall is to support the cell
membrane (plasmalemma) from the hydrostatic pressure from the cytoplasm. Together, the cell
wall and cell membrane give structure to the cell and the plant tissue (Salisbury and Ross, 1992).
The cell wall can be compared with a cotton cloth bag with a water filled balloon inside (the
plasmalemma). The cell wall allows passage of water, and cannot stretch if more water is forced
into the balloon. However, if water is released from the balloon, the cell collapses. Leaves and
young stems are composed of cells that have mostly primary cell walls; consequently, they are rigid
when the fluids in their cells push against the walls. When water is lost, the hydrostatic pressure in
the cytoplasma decreases, and eventually the cells collapse and the plant part will wilt (Salisbury
and Ross, 1992). Other more mature parts of the plant develop a more rigid and thicker secondary
cell wall when the cells have stopped enlarging (Salisbury and Ross, 1992); this prevents the cell
from collapsing even when water is lost.
Figure 2. Main components of a plant cell. From Wills et al. (2007).
Within the cell, the cytoplasm and usually one or more vacuoles are found (Salisbury and Ross,
1992). The vacuole is a volume of water and solutes, such as sugars, amino acids, and salts; the
vacuole is surrounded by a semipermeable membrane, the tonoplast. The vacuoles occupy 80-90%
or more of a mature cell (Salisbury and Ross, 1992). Together, the plasmalemma and the tonoplast
are responsible for maintaining the hydrostatic pressure of the cell, allowing the passage of water,
but selectively they are responsible for restricting the movement of solutes or macromolecules, such
as proteins and nucleic acids. The resulting turgidity of the cell is responsible for the crispness of
fruit and vegetables. Dehydration of plant cells results in loss of cell turgor, which is essential for
the changes in the textural crispness of fresh vegetables (Wood et al., 2005).
7
Important processes occur in the cytoplasm including the breakdown of storage reserves of
carbohydrate by glycolysis and protein synthesis. The cytoplasm also contains several important
organelles, which are membrane-bound bodies with specialized functions; some of these are shortly
explained here. The nucleus is the control center of the cell and contains the genetic information.
The mitochondria are the energy powerhouses of the cell and contain the respiratory enzymes of
the tricarboxylic acid (TCA) cycle and the respiratory electron transport system which synthesizes
adenosine triphophate (ATP). Mitochondria utilize the product of glycolysis for energy production.
Chloroplasts are found in the green parts of the plant. Chloroplasts run the photosynthesis and are
also located in the cytoplasm. Chloroplasts contain the green pigment chlorophyll and the
photochemical apparatus for converting solar energy into chemical energy to the plant (Wills et al.,
2007). By enzymatic reactions, atmospheric carbon dioxide is synthesized into sugars and other
carbon compounds. Chlorophyll is held between layers of lipid and protein in the thylakoid
membranes within the leaf chloroplasts, along with visual light-absorbing carotenoids, i.e. carotene,
and xanthophylls molecules. The chlorophyll molecule consists of two parts: the tetrapyrrole ring
chelating magnesium and a phytol ‘tail’ (Figure 3); the chlorophylls are attached by noncovalent
bonds to protein molecules in the membrane (Salisbury and Ross, 1992). Chlorophyll a and
chlorophyll b are the major forms of the pigment existing in plants.
Figure 3. Structure of chlorophyll with the magnesium-chelating tetrapyrrole ring and the phytol ‘tail’.
In summary, the plant cell is built of several components and each is important for the living plant;
together they supply the plant with structure and energy. Water is essential for keeping the
structure of especially young and immature tissue.
8
2.3 Respiration
Harvested fruit and vegetables are still living material with metabolic activity and respiration
processes. Generally, the rate of deterioration is proportional to respiration rate (Watada and Qi,
1999), e.g. fresh produce with high respiration rates tends to have a shorter shelflife than produce
with low rates. Respiration rate is therefore an important parameter in postharvest quality.
Respiration is the oxidative breakdown of sugars and other cell components in the plant cell into
CO2 and H2O, whereby energy is produced. The overall reaction is:
C6H12O6 + 6O2 → 6CO2 + 6H2O + energy (1)
Totally, each molecule of glucose gives 38 ATP and 686 kcal heat energy (Kader and Saltveit,
2003b). Overall, the breakdown of carbohydrates during respiration is essentially the reverse of
photosynthesis. The energy and the intermediate molecules produced during the breakdown of
carbohydrates are required to sustain all the anabolic reactions essential for the maintenance of the
cellular organization and membrane integrity of the living cells. The process involves a series of
enzymatic reactions. The main processes involves the glycolysis, the tricarboxylic acid (TCA)
cycle, and the electron transport system; together they are the essential metabolic pathways of
aerobic respiration (Figure 4) (Kader and Saltveit, 2003b; Wills et al., 2007).
The glycolysis occurs in the cytoplasm and involves the production of two molecules pyruvate, two
molecules ATP, and two molecules of NADH from each molecule of glucose. Storage carbohydrate
is converted to hexose phosphate and subsequently in aerobic respiration, glycolysis produces
acetyl Coenzyme A, which enters the TCA cycle. The TCA cycle occurs in the mitochondrial
matrix and involves the breakdown of pyruvate into CO2 and water in nine sequential enzymatic
reactions. Energy is stored in ATP and in reduced cofactors NAD+ or FAD. Oxygen is required in
this process. Besides producing many precursors to cellular enzymatic reactions, the TCA cycle
also produces one molecule of FADH2 and four molecules of NADH for each molecule of pyruvate.
The electron transport system is located in the inner membranes of the mitochondria and involves
the production of ATP from FADH2 and NADH, since the energy contained in a molecule of
NADH or FADH2 exceeds what is needed for most cellular processes. The production of ATP is not
directly coupled to specific enzyme reactions but proceeds through a chemi-osmotic process across
the inner membrane of the mitochondria; the process is dependent on the chemical environment
(e.g. pH and ions) in the mitochondria (and the cell) and it requires O2 and produces water.
9
Figure 4. Simplified diagram of the biochemistry of aerobic and anaerobic respiration in plants. From Wills et al. 2007.
In the absence of O2, NADH and FADH2 accumulate, and as their oxidized forms (NAD+ and FAD)
are consumed, the TCA cycle comes to a halt and glycolysis becomes the source of ATP
production. Regeneration of NAD+ is absolutely essential for the survival of the anaerobic cell. O2
is not required for the glycolysis, and the produced pyruvate is enzymatically transformed into
acetaldehyde with production of CO2. NAD+ is regenerated by enzymatic conversion of
acetaldehyde to ethanol. Some vegetables can also transform pyruvate into lactate and thereby
regenerate NAD+. In anaerobic respiration, pyruvate is converted into lactate or acetaldehyde and
ethanol. Totally, the anaerobic respiration gives two moles of ATP and 21 kcal heat energy from
each molecule glucose (Kader and Saltveit, 2003b). From a cost-benefit point of view, aerobic
respiration is therefore optimal for the cell. Furthermore, the production of lactate lowers pH and
together with the ethanol and acetaldehyde produced in anaerobic respiration leads to a chain of
10
events ultimately leading to cell death (Salisbury and Ross, 1992). The O2 concentration necessary
for aerobic respiration varies among tissues, and therefore the gradual shift from mainly aerobic to
anaerobic respiration varies.
The energy bound as ATP is made available to the plant cell by breaking the phosphate bond:
ATP → ADP + Pi + energy (2)
The resulting energy can be used in various metabolic reactions in the plant cells. Some of the
remaining energy is lost as heat.
The respiratory quotient (RQ) is the ratio of CO2 produced and O2 consumed (respiration rate of
CO2 /respiration rate of O2). For fresh vegetables RQ range from 0.7-1.3 at aerobic respiration.
When carbohydrates are the substrate in aerobic respiration, RQ is ~1 while it is <1 for lipids and
>1 for organic acids. Much higher RQ values usually indicate anaerobic respiration (Fonseca et al.,
2002a; Kader and Saltveit, 2003b).
The produced water is incorporated into the aqueous solution of the cell. The O2 used and the CO2
produced in respiration is exchanged with the surrounding atmosphere by gas diffusion. In leaves,
this is regulated by stomata in the cell (Figure 5). One overall effect of respiration on quality is that
the metabolic reserves consumed by respiration inevitably change the features of the plant material.
If starvation becomes extensive, even proteins can be used as substrate for respiration (Salisbury
and Ross, 1992).
In summary, respiration rate is the rate of the metabolic processes in the postharvest cell
converting photosynthetic reserves to energy, and respiration is a catabolic process, where mainly
stored carbohydrates are converted into energy. Respiration is preferably aerobic, but anaerobic
respiration may also occur producing less energy and potentially toxic compounds. The aerobic
process requires O2, generates CO2, and releases heat.
11
2.4 Leaf structure
The structure of a leaf (Figure 5) can be viewed as a construction that protects the inner cells, where
most of the photosynthesis takes place.
Figure 5. Cross section of a leaf showing the surface (cuticle), stoma, internal structures, and the network of intercellular spaces. Modified from www.digitalfrog.com.
The outer layers form a ‘skin’ that protects the plant from rapid breakdown (Glenn et al., 2005).
This waxy cuticle on leaf surfaces restricts to some extent diffusion, so most water vapor and other
gases must pass between the guard cells, through the openings called stomata. Substomatal
cavities lie below the stomata and connect with the intercellular airspace network allowing for
influx and efflux of gasses. In harvested leaves, stomata normally close when the surrounding pair
of guard cells lose turgor in response to loss of water (Wills et al., 2007). Beneath the cuticle lies
the epidermis, which acts as a boundary between the inner cells and the surroundings. Epidermal
cells are usually transparent, as they often lack chloroplasts and allow the light to shine to the inner
cells (Salisbury and Ross, 1992). Directly beneath the epidermis lies a layer of tightly packed,
vertically elongated palisade parenchyma, which together with the spongy parenchyma forms the
mesophyll, where the main part of the photosynthesis supplying the living plant with energy during
daytime takes place. Water evaporates inside the leaf from the cell walls of the mesophyll into the
intercellular space, which is directly connected with the outside air when the stomatas are open
(Salisbury and Ross, 1992). If the water vapor of the surroundings is higher than in the intercellular
space, water will be lost. The relative humidity (RH) of the intercellular space is calculated to be
99.5% (Ben-Yehoshua and Rodov, 2003) meaning that transpiration will occur and be proportional
to the vapor pressure of the surroundings. The veins are the vascular tissue of the leaf; they are
Cuticle
Spongy parenchyma
Palisade parenchyma
Stoma
Intercellular space
Epidermis
Vein/vascular bundle
Guard cell
Skin
12
located in the spongy layer and transport nutrients and fluid internally in the plant. When the leaves
are harvested they are cut off from the water and nutrient supply from the plant. If no attempts are
made, water is lost due to transpiration, and loss of cell turgor will lead to wilting. Most water is
lost through the stomata, as the loss through the skin is approximately one tenth of the loss
occurring through stomata (Ben-Yehoshua and Rodov, 2003). The detached leaf only has limited
nutrient and water resources to withstand postharvest storage.
In summary, in its nature, the leaf is built to support the plant with energy via photosynthesis; when
detached, it no longer serves this purpose. However, the processes built into the leaf and in every
cell are still functioning and the metabolic processes switch to respiration mode as the main energy
generator. Transpiration will lead to water loss.
2.5 Senescence of leafy green vegetables
Plant cells, like all other living material, have a ‘date of expiration’ when the cells will age and
finally die. This process is called senescence. The senescence processes of fruit and some
vegetables are well described, but not fully understood, and the senescence process of leafy
vegetables are described in less detail (King and O'Donoghue, 1995; Toivonen and Brummell,
2008). It can therefore be advantageous to look into leaf senescence as a starting point to gain
knowledge about senescence of leafy vegetables.
Senescence is a natural change in the plant cell which causes breakdown and inevitably cell death
(King and O'Donoghue, 1995). Natural leaf senescence is the situation where the leaf changes color
on the tree and wilt; this process takes place during autumn. By definition, leaf maturity starts when
leaf expansion is over and ends with the first senescence symptoms (Figure 6) (Guiboileau et al.,
2010). During the maturity phase, the leaf is facing numerous sub-lethal events leading to many
chronic senescence syndromes and recovery events, and leaf senescence no-return syndrome is
characterized by a succession of degradation processes that will lead to death (Guiboileau et al.,
2010). Senescence contributes to the plant survival and the developmental program, as leaf
senescence aims at remobilization of leaf nutrients, such as nitrogen and carbohydrates (Guiboileau
et al., 2010; Matile et al., 1996; Thomas et al., 2003). In naturally senescing leaves, senescence
occurs in a coordinated manner at the whole-leaf level starting from the tip and margins towards the
base of the leaf (Guiboileau et al., 2010).
13
Figure 6. Schematic representation of the main phase of leaf life story, including leaf expansion, maturity and senescence. From Guiboileau et al. (2010).
Ethylene and other hormones induce senescence to the leaves on the tree, but apparently not in all
growth phases. Results indicate that in very young leaves, ethylene is not able to induce leaf
senescence, but when the leaves reach a certain level of maturity, ethylene will induce senescence,
and as senescence is progressing, the process will run independently of ethylene presence
(Guiboileau et al., 2010). Harvested leafy vegetables produce small amounts of ethylene; however,
senescence is accelerated if they are stored together with ethylene producing products (Ludford,
2003; Martinez-Romero et al., 2007; Whitaker, 2003).
Many of the changes observed during senescence of green vegetables show similarities to changes
seen during natural leaf senescence, although the senescence is induced artificially at harvest due to
removal of nutrient supplies (Hörtensteiner and Kräutler, 2011; Page et al., 2001). In vegetables
originating from vegetative tissue (Figure 1), senescence is unwanted and should be postponed as
long as possible. The sensory quality of these vegetables is optimal right after harvest as the
processes of plant senescence increase as soon as the tissue is harvested from the plant. Growth
processes, such as cell division and expansion, and protein and carbohydrate synthesis usually cease
upon harvest, and the metabolism goes into a catabolic or degradative mode (Wills et al., 2007).
The overall effects of senescence on fruit and vegetables are summarized in Table 2. The changes
that take place during senescence can be seen both at a physical level (from a consumer’s point of
view) and a physiological level (what happens in the plant). Chlorophyll degradation is an obvious
visual change during senescence, and it is accompanied by losses in membrane lipids and proteins,
eventually resulting in cell death and textural changes. In detached leaves, cell death does not occur
simultaneously across all cells in a leaf, or even within cells of a particular type; it is a gradual
process that occurs cell by cell (Wagstaff et al., 2007). The rate of senescence is linked to the rate of
the metabolic processes, i.e. the respiration rate.
14
Table 2. Physical changes (from a consumer’s point of view) and physiological changes (what happens in the plant) accompanying the senescence of fruit and vegetables. Modified after King and O’Donoghue (1995).
Physical changes:
Color Loss of green color
Texture Softening
Wilting
Drying
Loss of resistance to pathogens Development of infections
Lesions
Physiological changes:
Cellular Loss of chlorophyll, disassembly of chloroplast structure
Degradation of cell walls
Altered membrane composition
Loss of cellular compartmentation, release of vacuolar contents
Composition Altered sugar content, switch to alternative substrates for respiration
Net loss of RNA
Increased protease activity, net loss of protein
Altered amino acid content
Changes in appearance as a consequence of senescence of packed wild rocket leaves are seen in
Figure 7. The color went from an overall green color to a more heterogeneous appearance with
many yellow and rotten leaves. To the left, the wild rocket leaves had intact structure and full height
in the package, and to the left, some of the leaves had collapsed (Paper III). It is evident that both
color and textural changes occur during storage of the wild rocket leaves.
Figure 7. Senescence of wild rocket as it appeared in Papers III and IV. The wild rocket was packaged to avoid water loss, but the gas composition was almost atmospheric, hence only senescent changes were induced. Photo: Jens Madsen.
15
2.5.1 Color changes
The most visible sign of senescence in Figure 7 is the loss of green color and appearance of yellow
color due to the uncovering of the underlying carotenoid pigments. Chlorophyll is the green
pigment in plants (Figure 3), and loss of chlorophyll is a response to many biotic and abiotic
stresses (Hörtensteiner and Kräutler, 2011).
During the last decades, the pathways for chlorophyll breakdown in plants have been studied
intensively, and many similarities between plant species were found (Hörtensteiner, 2006;
Hörtensteiner and Kräutler, 2011; Matile et al., 1999). Breakdown of chlorophyll is linked to the re-
use of proteins in the plant, and the chlorophyll breakdown is closely linked to protein degradation
during senescence of living plants. At breakdown of the chlorophyll-binding proteins, chlorophyll is
released (Hörtensteiner, 2006; Matile et al., 1999). The exact starting point of the breakage of the
chlorophyll-protein complexes is not clear; it might be proteases synthesized or activated during
senescence or the breakdown of chlorophyll b with resulting destabilization of the complexes
making them accessible for proteases (Hörtensteiner, 2006). Due to its light absorbing properties,
free chlorophyll would have cell toxic effects for the cell inducing radicals and oxidative changes;
therefore, plants require a mechanism to convert chlorophyll to non-exitable, i.e. non-colored,
breakdown products. Thus, chlorophyll breakdown can be considered as a detoxification process in
the living plant during senescence – a process vitally important for plant development and survival
(Hörtensteiner, 2006; Matile et al., 1996). Moreover, recent indications reveal that the products
from the chlorophyll breakdown could have physiological roles as e.g. internal signals or
antioxidants (Hörtensteiner and Kräutler, 2011).
Figure 8 shows a generalized scheme for the initial stages of chlorophyll breakdown that eventually
lead to colorless products (Hörtensteiner and Kräutler, 2011; Matile et al., 1999). There are two
routes to the conversion of the green chlorophyll a to the olive brown pheophorbide. One is that the
bright-green chlorophyllide is formed when the enzyme chlorophyllase cleaves the phytol chain
from chlorophyll. This is converted to pheophorbide when the magnesium ion is lost. The second
route involves initial loss of the magnesium ion when the olive brown pheophytin is formed. This is
converted into pheophorbide on the cleavage of the phytol chain. This is further converted into
colorless products by cleavage of the porphyrin ring, and these products are transported from the
chloroplasts into the vacuole where they are further degraded. At acidic pH the enzymes during
chlorophyll breakdown will not have any affect, and low pH mediates the loss of magnesium from
the chlorophyll molecule creating pheophytin (Toivonen and Brummell, 2008). In that case the
olive brown color remains in the tissue as the death of the tissue inhibits further degradation.
16
Figure 8. Schematic representation of chlorophyll breakdown pathways in green plant tissue. The reactions occur in the chloroplasts. The figure is drawn on the basis of Hörtensteiner and Kräutler (2011).
However, it is not clear whether the pathway outlined in Figure 8 for the breakdown mechanism of
chlorophyll is the only mechanism of chlorophyll breakdown occurring in postharvest products. An
alternative pathway is initiation and breakdown during oxidizing activities by oxidases (Toivonen
and Brummell, 2008). However, results indicate that a chlorophyll bleaching activity of oxidases
may only be restricted to cell autolysis (Hörtensteiner, 2006). It is suggested that the pathway in
Figure 8 occurs in intact tissue postharvest and that the alternative pathway by oxidases only occurs
later in the senescence or stress response timeline (Hörtensteiner, 2006). At this point, membrane
breakdown occurs and this leads to loss of sub-cellular compartmentalization and thereby to the
intermixing of many enzyme systems and their potential substrates, again leading to possible
oxygen radical production scenarios and oxidation of chlorophyll (Toivonen and Brummell, 2008).
Leaf yellowing is considered as an important quality defect in green vegetables as well as in rocket
leaves (Koukounaras et al., 2007). The yellowing due to senescence in wild rocket is seen in Figure
7.
Chlorophyll a
(green)
Chlorophyllide a
(bright green)
Pheophorbide a
(olive brown)
Pheophytin a
(olive brown)
Chlorophyll b
Chlorophyllase
Pheophytinase
Magnesium dechatalase
(or low pH)
Magnesium dechelatase
(or low pH)
Colorless products
Transported into the
vacuole
Known and unknown
enzymes
Chlorophyl b reductase,
7-hydroxymethyl chlorophyl reductase
17
Another color defect in postharvest leafy vegetables is enzymatic browning where the tissue is
broken. The first steps of the browning oxidation of polyphenols into quinones are catalyzed by
polyphenol oxidase (PPO) in the presence of O2 (Figure 9). The quinones undergo further
polymerization into brown melanoidines. The vacuole holds 97% of the phenolic compounds of the
cells and at cell rupture they are released and mixed with the PPO leading to browning (Toivonen
and Brummell, 2008).
Figure 9. Schematic reaction of the oxidation of tyrosine (and other phenols) by polyphenol oxidase (PPO) into colored quinones in the presence of oxygen. The quinones undergo further polymerization into melanoidines.
Browning of the cut edges is a problem occurring in the soft tissue of baby leaves like lettuce and
escarole, but this, interestingly, does not affect rocket leaves to the same extent. This was linked to
a higher content of ascorbic acid in rocket leaves (Degl'Innocenti et al., 2007), as ascorbic acid will
reduce browning by reversing the reaction catalyzed by PPO (Whitaker, 1996). However, browning
of cut edges of wild rocket leaves eventually occurs at aerobic conditions as described in Paper III.
2.5.2 Textural changes
The texture of vegetable vegetative tissue is dependent on the cell wall, turgor and the structure of
the tissue as described in sections 2.2 and 2.4. A young leaf has mainly primary cell walls, and as
baby leaves are in the expansion phase (Figure 6), the cell walls are not fully developed, and the
tender tissue of baby leaves leads to faster texture loss and decay than experienced in head lettuce
(Martinez-Sanchez et al., 2012).
If no attempts are made against water loss, leafy vegetables will wilt and in case of a product like
spinach, wilting can be one of the major causes of loss of visual appearance and texture (Piagentini
et al., 2002); some technologies to prevent water loss will be discussed in section 3. Wilting
happens due to the transpiration processes explained in section 2.4, and if leafy vegetables lose
more than 3% of the original fresh weight, they become unsalable (Ben-Yehoshua and Rodov,
2003). Fruit and vegetables remain fresh only as long as they retain water. Transpiration is one of
the main processes that affect deterioration, as it induces wilting, shriveling, and loss of firmness,
crispness and succulence – all components of freshness (Ben-Yehoshua and Rodov, 2003).
Furthermore, water loss induces physiological stress, which accelerates senescence, as indicated by
18
faster rates of membrane disintegration and leakage of cellular contents (Ben-Yehoshua and Rodov,
2003).
If water loss is prevented, other textural changes will still occur. Textural changes during
senescence of fruit are widely studied as it is the path leading from immature to mature and ripe
fruit (Toivonen and Brummell, 2008). The natural cell wall modifications that cause softening in
fruit do not occur in the same way in vegetable vegetative tissues, and texture changes of these apart
from wilting is studied in less detail (Toivonen and Brummell, 2008). However, some indications in
studies of leaf senescence reveal that the cell wall in leaves may function as a storage reserve of
carbon at sugar starvation in addition to providing physical support for the plant body (Lee et al.,
2007). The senescence of vegetable vegetative tissue is exclusively a degradative process, where the
compartmentalization of the cell begins to fail, involving changes in membranes, cell walls, sub-
cellular organelles, proteins, and eventually cell death (Toivonen and Brummell, 2008). This leads
to loss of texture. Consequently, stems and leaves often senesce rapidly and thereby lose their
attractiveness as texture is often the dominant characteristic determining the quality of these
vegetables (Wills et al., 2007).
Lack of membrane integrity allows the leakage of cellular osmotic solutes into the intercellular
space (Toivonen and Brummell, 2008), and the leaf will appear darker. The loss of tissue integrity
can also be seen as a presence of dark and necrotic stains on the leaf surface (Ares et al., 2008a),
which in the sensory evaluation in Paper III was covered by the attribute ‘watersoaked leaves’.
Exudates released from collapsed cells and water produced during respiration will eventually be
transferred out of the leaf, and the leaf will appear wet. This made the leaves studied in Paper III
stick together; therefore, the leaves have a lower ability to sprinkle from each other when touched.
When the amount of water was pronounced inside the package, it is condensed at the bottom of the
tray.
When the structure of the leaf is broken, the leaf will be more vulnerable to attack from
microorganisms, and even in freshly harvested parsley, cracks in the waxy cuticle can be detected
(Wood et al., 2005). In these cracks, microorganisms can develop and lead to rot. The cell rupture
and general degradation of tissue lead to release of odorous compounds (Jacxsens et al., 2003;
Nielsen et al., 2008), and development of different off-odors was also experienced in the senescent
wild rocket in Paper III.
In summary, senescence of leafy green vegetables leads to color changes from green to yellow and
further senescence leads to cell death seen as dark areas. There is great risk of water loss and
wilting due to transpiration which should be prevented. Texture loss is experienced as a
consequence of wilting and cell death. Senescence is inevitable; it can only be postponed. The rate
of senescence is linked to respiration rate.
19
2.6 Effect of harvest season, variety, and aging on respiration rate
Quality of postharvest end products will change as the quality of the raw materials changes, and the
raw materials have a high variability not least due to the harvest season and different varieties
(Saltveit, 2003b). The effect of the harvest season and the variety on the respiration rate will be
discussed in this section.
The respiration rate of the harvested plant depends on the respiration rate in the living plant. In
section 2.3 it was stated that the metabolic reserves are consumed by respiration and that the energy
source and O2 availability will influence RQ. As fruit and vegetables have considerable biological
variation, this also gives rise to large variations in respiration rate. Respiration rate is therefore
dependent on type of commodity and maturity state or stage of development at harvest together
with chemical composition, season and variety.
Little research has focused on the impact of the season and the variety on respiration rate of fresh
produce despite the fact that low-respiring varieties are expected to have longer shelf lives than
high-respiring varieties. In three apricot varieties respiration rates were 60% lower in the least
respiring variety compared to the most respiring variety (Pretel et al., 2000). In four varieties of
kiwi fruit, varietal differences were found in respiration rate, and those with low rates had the
longest shelf life (Manolopoulou and Papadopoulou, 1998). In a case with white and violet salad
savoy, varietal differences in respiration rate were found (Kim et al., 2004). Mizuna and watercress
baby leaf showed significantly higher respiration rates in the second cutting compared to the first
cutting (Martinez-Sanchez et al., 2008). This means that for some fruit and vegetables, varietal
differences in respiration rates will affect storage stability; this may largely determine which
varieties are suited for storage. Selecting varieties with low respiration rate may be a useful tool for
longer shelf life.
In pears harvested at identical maturity, no seasonal differences were found in respiration rates
between two consecutive years (Gomes et al., 2010). For rocket and wild rocket, only minor
differences were observed in respiration rate between different harvest times (Martinez-Sanchez et
al., 2008). In Paper II, seasonal differences in respiration rate of broccoli and wild rocket were
found. However, this was not the case between all harvests of wild rocket, which corresponds to the
findings of Martinez-Sanchez et al. (2008). For broccoli, varietal differences were also found, and a
high respiration rate was linked to high dry matter content in wild rocket and broccoli.
Figure 10 shows the regression vectors from models predicting respiration rate of wild rocket (A)
and broccoli (B) from temperature, O2 concentration, storage time, and harvest season (wild rocket),
or harvest season×variety (broccoli) (Paper II). This analysis dissected the contribution from the
different factors and it is seen that temperature had a high influence on respiration rate, contributing
20
to a rise in respiration rate for both commodities (Figure 10). Storage duration had some effect on
respiration and as the storage duration progressed, the respiration rate was lowered. The respiration
rate of both wild rocket and broccoli was highly influenced by harvest season and the respiration
rate was highest at the first harvest for both commodities, which was spring for wild rocket and
early summer for broccoli. For broccoli, the effect of different varieties was tested, and the effect of
the season is pronounced in variety 2 that was tested in both seasons (Figure 10B). The effect of
season and variety together on the respiration rate had in one case an effect in the same range as
temperature, i.e. if the effect of variety 2 early summer, and variety 4 late summer, is added (Figure
10B). High respiration rate was found to be linked to high dry matter content (Paper II), and this
makes sense if the dry matter is due to high sugar content. Similarly, leaves still attached to the
plant have higher respiration rates when the sugar levels are high; furthermore, there is a fairly good
correlation between the rate of growth and respiration rate (Salisbury and Ross, 1992).
21
Figure 10. Regression vector for PLS models of respiration rate of wild rocket (A) and broccoli (B). Three harvests of wild rocket (A) and two harvests of four different variants of broccoli (B) were investigated. The x-variable with the highest positive or negative regression vector has the highest impact on respiration rate. Modified from Paper II.
Te
mp
era
ture
Sto
rag
ed
ura
tio
n O2
Sp
rin
g
Ea
rly
su
mm
er
La
tesu
mm
er-0.4
0
0.4
0.8R
eg
ressio
n v
ecto
rfo
r O
2re
sp
ira
tio
n A: Wild rocket
Te
mp
era
ture
Sto
rag
ed
ura
tio
n
O2
Va
r 1
,e
arl
ysu
mm
er
Va
r 2
, e
arl
ysu
mm
er
Va
r 3
,la
tesu
mm
er
Va
r 2
, la
tesu
mm
er
Va
r 4
,la
tesu
mm
er-0.4
-0.2
0
0.2
0.4
0.6
Re
gre
ssio
n v
ecto
rfo
r O
2re
sp
ira
tio
n B: Broccoli
22
Vegetables of vegetative tissue do not generally show a sudden change in respiration during
storage, as is seen in climacteric fruit, where a rise in respiration rate is seen at ripening; however
aging appears to have an effect of respiration rate (Uchino et al., 2004). From Figure 10A and B, it
is seen that the respiration rate decreased during storage. This means that aging of the postharvest
product during storage also had an impact on the respiration rate. Similar results of decreasing
respiration rate during storage were also found in rocket leaves during four days of storage at 8 °C
(Koukounaras et al., 2010). The effect of storage duration on respiration rate might be attributable
to depletion of carbohydrate reserves which function as substrates for the respiration during storage
(Bastrash et al., 1993; Nei et al., 2006). Furthermore, the respiration pattern might change as the
cells collapse during senescence. High respiration in the beginning of the analysis can be a result of
wounding during handling (Saltveit, 2003a). Furthermore, different varieties can easily react
differently upon stresses such as wounding and the physiological maturity will also affect the stress
reaction and thereby respiration rate (Brecht, 1995). The effect of various stresses on the respiration
rate and quality is further discussed in section 3.
In summary, a large biological variation is found in the respiration rate, which is affected by
variety and harvest season. The effect of season and variety on the respiration rate cannot be
predicted but has to be tested. Selecting varieties with low respiration rate may be a useful tool for
longer shelf life.
Overall, the metabolic resources of the postharvest plant are limited and it is of great importance to
keep respiration at a low -yet aerobic- level in order to maintain good quality for as long as
possible.
23
3 Effect of postharvest technology on quality
The postharvest life of fresh produce is limited and the highest quality of leafy vegetables is at the
point of harvest. In order to maintain freshness, it is pivotal to keep the respiration low and aerobic.
The raw material may vary in quality, and different technologies can be used to retain as much of
the initial quality as possible.
3.1 Processing
Harvested products undergo different kinds of handling from the field to the consumer. Even
though processing is performed as gently as possible, wounding will inevitably occur. Dependent
on the level of convenience of the end product, processing may include washing, cutting, trimming,
peeling, centrifuging, and packaging, and the terms fresh-cut, minimal processed, or ready-to-eat
are used almost interchangeably.
Wounding not only physically damages the membranes in the injured cells, but also disrupts
membrane function in adjacent cells, so that enzymes and substrates mix and produce unwanted and
uncontrolled reactions (as browning by PPO and wound induced chlorophyll breakdown, described
in section 2.5.1 (Saltveit, 2003a). Furthermore, the wounding causes a physiological response in the
tissue (Brecht, 1995; Saltveit, 2003a) and this increases the respiration rate (Kader and Saltveit,
2003b). In the growing plant, a stress like pathogen attack will also induce a wound signal
(Halliwell and Gutteridge, 2007), and the response to a wounding incident can be explained as a
biotic response to an abiotic stress. Some of the wound induced signals in both the growing and the
postharvest plant are thought to be jasmonic acid and ethylene (Brecht, 1995; Halliwell and
Gutteridge, 2007; Saltveit, 2003a).
As mentioned in section 2.6, the stress response may very well vary due to harvest season and
variety. The initial stress level of the product has a great influence on its response to processing
stress, and the wound response can be modified by another stress applied before harvest (Saltveit,
2003a), where the effect may last for weeks or months thereafter (Halliwell and Gutteridge, 2007).
Furthermore, Clarkson et al. (2003) found that it was possible to enhance the processability of baby
leaves by applying a mechanical stress during growth. The properties of the cell wall were found to
be important for the ability of leafy vegetables to withstand stresses experienced during processing
and handling and large epidermal cells for some leafy vegetable were related to high ability to
withstand processing (Clarkson et al., 2003).
The response to wounding is also dependent on the severity of the wounding, since the shock of
cutting affects not only those cells cut and those adjacent to the cut, but also cells far from the actual
site of injury (Saltveit, 2003a). Stresses will inevitably have an effect on the metabolism and is
hence also reflected in the respiration rate. It is common that a rise in the respiration rate may be
found in the beginning of postharvest experiments due to handling and packaging before the onset
24
of the experiment. This was experienced during the study presented in Paper II. A peak in the
respiration rate was also found just after the onset of an experiment in investigations of the
respiration rate of rocket (Koukounaras et al., 2009; Koukounaras et al., 2010). The peak in
respiration rate was coincident with a peak in ethylene production, though no effect of the degree of
cutting was found on either respiration or production of ethylene of rocket salad (Koukounaras et
al., 2010). It is difficult to set up postharvest scientific experiments without inducing some
wounding during handling, and the effect of wounding on handling may be difficult to extract from
the effect of aging. The ethylene induced by wounding is not considered a major problem relative to
postharvest industrial handling of vegetative or immature tissue, where the ethylene production is
low (Saltveit, 2003a).
Besides the respirational response to wounding, washing, and cutting of the tissue will also lead to
other quality changes. When the protective epidermal layer (Figure 5, section 2.4) is removed or
cracked, surface moisture accumulates resulting in increased water loss. Broken cell walls will lead
to water in intracellular spaces causing translucent tissue, and in leaves seen as dark stains (Ares et
al., 2008a; Saltveit, 2003a), or water soaked leaves (Paper III). Liquids on the surface of the leaves
(from either wounded tissue or washing) act as a barrier to gas diffusion, as gases diffuse very
slowly through water (Saltveit, 2003a). Baby leaves have softer tissue than mature leaves, and
hence has a lower processability, meaning lower ability to withstand the postharvest and packing
process (Clarkson et al., 2003; Martinez-Sanchez et al., 2012). Furthermore, the exposed tissue of
cut vegetables is more vulnerable to contaminants when the protective layer is removed (Saltveit,
2003a). When the plant material is wounded due to harvest and handling, it provides optimal
conditions for microorganisms, and a rise in respiration rate at the end of storage of tender tissue as
baby leaf may be experienced (Martinez-Sanchez et al., 2008). Increased respiration after some
period of time in storage may be caused by the onset of decay by microorganisms (Fonseca et al.,
2002a). Moreover, the rise in respiration rate may be a mixture of a response to wounding and to
metabolism of the microorganisms.
Often vegetables do not emit any specific odor before the tissue is wounded. The wounding starts
enzymatic reactions where free fatty acids are produced from the membrane lipids (Brecht, 1995).
These free fatty acids are toxic relative to many cellular processes and are capable of causing
organelle lysis and inactivation of proteins. Lipoxygenase catalyzes the peroxidation of certain fatty
acids to form conjugated hydroperoxides, generating radicals that can attack membranes, and thus
cause further membrane disruption (Brecht, 1995). Lipoxygenase activity is also involved in the
production of desirable and undesirable aroma volatiles. Some of the formed aldehydes and
alcohols have characteristic smells which contribute to the aroma of damaged plant tissue (Halliwell
and Gutteridge, 2007). From the headspace of pasted rocket, more than 50 compounds have been
identified to be essential volatiles for the aroma of the crushed leaves, and the origin of all of them
are not yet determined (Jirovetz et al., 2002). Interestingly, the cutting direction of lettuce appears
25
to have an influence on the emitted volatiles and this might affect the sensory perception (Deza-
Durand and Petersen, 2011).
In summary, the wounding response is a natural response to mechanical stress and it inevitably
happens during processing and handling. Hence, processing should be as gentle as possible.
Wounding leads to quality changes due to increased respiration rate and subsequently increased
senescence, ethylene production, and breakage of cell walls inducing odor emission.
3.2 Temperature and moisture
In order to keep the physiological processes and water loss low, it is pivotal to keep the temperature
in the process chain low until consumption. Cooling of the produce quickly after harvest of leafy
vegetables especially during warm weather is important for quality, and a quick reduction of the
temperature to the optimum storage temperature will result in a longer storage life (Cantwell and
Kasmire, 2002a; Paull, 1999; Wills et al., 2007). Hereafter the temperature should be kept low and
the heat produced during respiration should be removed.
There are limits to how low temperature regimes can be used during postharvest as too low
temperatures have a negative impact on plant cells and thereby induce chilling injury (Wills et al.,
2007). Generally, the optimum temperature for storage of leafy vegetables is 0 °C to 1 °C but
freezing should be avoided, due to cell disruption when ice is formed (Cantwell and Kasmire,
2002a). However, other types of vegetables, e.g. tomatoes, can experience chilling injury at sub-
optimal temperatures for a longer time as chilling injury is cumulative (Cantwell and Kasmire,
2002b).
The key abiotic factor influencing the respiration rate is temperature as the enzymes involved in
respiration have a lower activity at low temperatures rather than at high temperatures (King and
O'Donoghue, 1995). Therefore, low storage temperature, is one of the preferred means of
maintaining product quality (Fonseca et al., 2002a; Gomes et al., 2010; Iqbal et al., 2009; Jacxsens
et al., 2000; Koukounaras et al., 2007; Lammertyn et al., 2001; Uchino et al., 2004).
Wild rocket and broccoli florets are high-respiring vegetables (Cantwell and Kasmire, 2002a), and
the effect of temperature is distinct. In the continuous measurements of temperature and respiration
rate in Paper I and Paper II, the close connection between these two factors could be studied in
detail (Figure 11), and this underlines the results in Figure 10, which indicated that temperature was
the most important single factor for respiration rate.
26
Figure 11. Temperature and respiration rate of broccoli in a closed jar. At time zero, the O2 concentration was atmospheric (~21 kPa O2) and at the end it was 6.3 kPa O2. Graphical abstract from Paper I.
Packed produce designed to have optimal gas composition (see following section 3.3) at a certain
temperature can – if stored under fluctuating higher temperature – experience serious quality
defects due to fermentative respiration and extensive condensation (Jacxsens et al., 2000; Tano et
al., 2007). Storage temperature is seldom constant in the distribution chain of fresh produce
(Mahajan et al., 2007; Paull, 1999), and this hampers a tight control of the respiration rate during
the storage condition, and, as seen in Figure 11, even small changes in temperature will have an
effect on the respiration rate.
The metabolic processes in plant development need to reach certain temperature thresholds in order
to proceed. These temperature thresholds are often a combination of temperature and time and
therefore accumulated temperatures are a measure of development. The effect of temperature during
postharvest storage is dependent on the time duration at the actual temperature: Accumulated
temperature sum during storage is calculated as time×temperature. The accumulated temperature
reflects the physiological age of the plant material (Hertog et al., 2007; Uchino et al., 2004).
Physiological age is used to determine plant material age instead of chronological age. Because
temperature has such a large impact on development, it makes no sense to use chronological age in
terms of plants. Physiological age is highly dependent on respiration rate that is affected by
temperature; therefore, accumulated temperature can be used for prediction of shelf life (Hertog et
al., 2007; Uchino et al., 2004). Consequently, storage of high respiring commodities at too high
temperature as a function of time can be termed ‘temperature abuse’, as the commodities lose
quality fast under these conditions (Paull, 1999). The accumulated temperature sum was used for
designing the experimental combinations of time and temperature storage of packaged wild rocket
in Papers III and IV, and the results were reasonable. However, two days at 20 °C appeared to
27
increase the physiological age slightly more than was the case when storing for four days at 10 °C
(Papers III and IV).
The rate of moisture loss is primarily controlled by the difference in water vapor pressure between
the air in the intercellular spaces of the plant material and the air surrounding it (Thompson, 2002)
(see also section 2.4). It should be ensured that the relative humidity (RH) in the surrounding air is
high; for leafy vegetables RH should be >95% (Wang, 2003). A high RH can be maintained by
keeping the temperature low, as the vapor pressure is highly dependent on temperature (Paull, 1999;
Thompson, 2002). However, if the dimensions of the refrigeration systems at the cold storage are
not designed correctly, a high RH can be difficult to maintain (Paull, 1999; Wills et al., 2007).
In summary, it is pivotal to keep temperature low from field to fork as this is the main parameter
influencing respiration rate in postharvest products. Respiration rate of high respiring commodities
reacts instantly to changes in temperature. Furthermore, low temperature is important to keep the
humidity high and thereby prevent water loss. Accumulated temperature sum can be used as a
measure to indicate the time spent at high temperature and thereby reduce temperature abuse. RH
should be kept >95% in order to prevent loss of crispness.
3.3 Packaging
Packaging of fresh fruits and vegetables is practical from a handling point of view. Moreover,
packaging is a very effective way to prevent water loss and thereby wilted leaves (Ares et al.,
2008a; Ares et al., 2008b; Luo et al., 2004); finally, packaging serves a hygienic purpose.
By packaging, RH can be kept high in the surroundings of the produce. Water is formed during
respiration and some water loss through the packaging material is unavoidable. Cooling of the
surface of the package removes heat of respiration, and thus creates a temperature gradient from
product to package surface. This also drives a water gradient because the cold air close to the
packaging material will have a higher water activity than the relatively warm air close to the
product (Kaur et al., 2011; Saltveit, 2003a). In a package, the cooling surface is the inner surface of
the bag, and the concomitant movement of heat and water vapor leads to condensation of water on
the inner surface of the package (Kaur et al., 2011; Saltveit, 2003a). To avoid condensation, it is
therefore important that cooling of the product occurs before packaging and that the temperature is
constantly low during storage.
In unpublished results from the project the weight loss of packages of wild rocket stored at 5 °C, 10
°C and 20 °C in packaging films with low (1900 ml O2/m2/day/atm, corresponds to 0.65
pmol/s/m2/kPa in Papers III and IV), medium (7140 ml O2/m2/day/atm), and high (48400 ml
O2/m2/day/atm, corresponds to 17.4 pmol/s/m2/kPa in Papers III and IV) oxygen transmission rate
28
(OTR) was below 1% at the end of storage time, which indicates that the packaging material was
sufficient to maintain a high RH inside the package.
The gas composition inside the package will change due to respiration. Therefore, a modified
atmosphere is often applied to maintain a high quality of the product during transportation and
storage. The effect of a modified gas composition on produce quality is discussed below.
3.3.1 Effect of gas composition on respiration rate
Atmospheric composition in the air is approximately 78% N2, 21% O2, and 1% other gases
including 0.04% CO2. The gas composition in the surroundings is exchanged with the air in the
intercellular space through stomata (see Figure 5 section 2.4). In leafy vegetables and other products
with a large surface to volume ratio, gas diffusion inside the tissue may be considered to contribute
with negligible resistance (Fonseca et al., 2002a).
Respiration is assumed to be slowed down by decreasing available O2 as a consequence of
reduction of overall metabolic activity (Fonseca et al., 2002a). The respiration rate is lowered as the
O2 concentration in the surroundings decreases, as shown in Figure 12. When the O2 concentration
reaches <2%, a considerable reduction in respiration rate is typically seen (Saltveit, 2003b;
Schouten et al., 2009). Below this O2 concentration, there is a dramatic increase in injuries due to
the anaerobic conditions (see the low O2 injury curve in Figure 12). These injuries will be further
discussed in section 3.3.4.
Figure 12. Respiration of lettuce in varying O2 atmospheres. Respiration is reduced when the O2 concentration is reduced from 2 to 0.5%, but there is a concomitant increase in the percentage of heads damaged by low oxygen injury (Saltveit, 2003b).
29
The relationship between O2 concentration and respiration rate was investigated in Paper II, and it
was found that for wild rocket and broccoli, O2 concentration above 6.3 kPa did not influence
respiration rate (Figure 10, section 2.6). The relationship at O2 concentrations below 10 kPa is
shown in Figure 13A, B, and C for wild rocket stored at set-point temperatures 5, 10, and 20 °C.
Furthermore, RQ (see section 2.3) calculated from these respiration rates is shown in Figure 13D.
At 5 °C (Figure 13A), the respiration rate was not lowered considerably until O2 reached ~0 kPa. At
this O2 concentration signs of anaerobic respiration occurred as the RQ had values much higher
than 1 (red points in Figure 13D). At 10 °C there was a trend towards a lower respiration rate at O2
< 2 kPa, although this is probably not significant considering the variation in the plot (Figure 13B).
At 20 °C, very unsatisfactory number of measurements were noted (Figure 13C), as the frequency
of the discrete measurements did not perfectly correspond to the rapid O2 depletion in jars at high
temperature. Consequently, no conclusions can be drawn from those data. Furthermore, RQ varied a
lot from point to point until the O2 concentration reached ~0 kPa, hence, no conclusions can be
drawn on the substrate for respiration (Figure 13D). At ~0 kPa O2, RQ increased in some jars
indicating anaerobic respiration; in others it was still low. The reason for low RQ values at ~0 kPa
O2 may well be that the respiration is calculated as the difference between two measurements (see
section 4.1 for measurement of respiration rate) and connected to the second measurement of O2 on
the x-axis.
30
Figure 13. Wild rocket in closed jars for respiration analysis from the late summer harvest (Paper II). The O2 and CO2 respiration rates from storage at 5 °C (A), 10 °C (B), or 20 °C (C) (A) and respiration quotient (RQ) (D) were determined using a closed system and discrete measurements; the setup is described in Paper II. Only data below 10 kPa O2 is included in the plots. Unpublished results.
An optimal gas composition at storage is where the overall metabolic activity is lowered, without
inducing quality defects (Saltveit, 2003b). Therefore, enough O2 is available for maintaining
aerobic respiration and CO2 is below the tolerance for the specific commodity. High CO2 (generally
> 20% for many commodities) is not beneficial and can cause an effect similar to that caused by
anaerobic respiration (Kader and Saltveit, 2003b; Saltveit, 2003b; Wills et al., 2007). The influence
0 2 4 6 8 100
10
20
O2
CO2
0 2 4 6 8 100
20
40
Resp
iratio
n rate
(ml h
-1kg
-1)
O2
CO2
0 2 4 6 8 100
50
100
O2
CO2
0 2 4 6 8 100
1
2
kPa O2
RQ
5 oC
10 oC
20 oC
RQ>>2
kPa O2
kPa O2
kPa O2
A: 5 oC
C: 20 oC
B: 10 oC
D: RQ
Resp
iratio
n rate
(ml h
-1kg
-1)
Resp
iratio
n rate
(ml h
-1kg
-1)
31
of CO2 on lowering the respiration rate is not so clear, and it depends on type and developmental
stage of the commodity, CO2 concentrations, and time of exposure (Fonseca et al., 2002a). One
explanation is that CO2 may lower intercellular pH, which would influence the enzymatic activity
(Escalona et al., 2006). Another explanation is that CO2 might inhibit ethylene production and
thereby respiration rate, rather than having a direct effect on the respiration rate (Fonseca et al.,
2002a).
3.3.2 Modified atmosphere packaging (MAP)
At the time of packaging, the gas composition equals atmospheric composition in passive modified
atmosphere packaging (passive MAP), whereas in active MAP, the gas composition (often
concerning O2, CO2, and N2) is modified during filling of the package. Not all combinations of
concentrations of O2 and CO2 are attainable by passive MAP (Beaudry, 1999), and active MAP can
be used to induce the intended gas composition inside the package.
Passive MAP is the most commonly used packaging technology for fresh produce (Sandhya, 2010).
During respiration, the produce will use O2 and produce CO2 and thereby modify the gas
composition. If the packaging material is completely tight, the O2 concentration decreases and the
product will initiate anaerobic respiration. It is therefore important that the film is permeable to O2
and CO2. This will result in influx of O2 and efflux of CO2 through passive diffusion caused by the
concentration gradients across the film. A film with proper barrier properties should keep the steady
state gas composition inside the package at the optimal level for the specific commodity. The
barrier properties of the film are determined by the film material and perforations made by needles
or laser (Sandhya, 2010). The transmission rate through perforations is practically independent of
temperature in the range of 0-25 °C, whereas transmission through the film material of non-
perforated films increases at increasing temperatures (Mangaraj et al., 2009). The optimal
permeability is dependent on the size of the package compared to the amount of produce and the
respiration rate – and thereby all the factors that influence respiration rate (Kaur et al., 2011; Lee et
al., 1991; Mahajan et al., 2007). Respiration parameters are therefore especially important when
designing MAP, where the permeability of the packaging material matches the actual respiration
rate (Iqbal et al., 2009; Jacxsens et al., 2000; Lakakul et al., 1999; Mahajan et al., 2007).
If MAP is designed correctly, the natural interplay between the respiration of the product and the
transfer of gases through the packaging material leads to an atmosphere richer in CO2 and poorer in
O2 which potentially reduces respiration rate, ethylene sensitivity and production, decay, and
physiological and chemical changes like oxidation (Fonseca et al., 2002a).
3.3.3 Modeling respiration rate
Modeling respiration rate is central relative to the design of optimized MAP for fresh fruits and
vegetables, to determine the optimum permeability of the packaging in the given situation
concerning respiration rate and storage temperature (Fonseca et al., 2002a; Fonseca et al., 2002b;
32
Gomes et al., 2010; Hertog et al., 1999; Iqbal et al., 2009; Mahajan et al., 2007; Torrieri et al.,
2010). The respiratory models are often based on enzyme kinetics and/or gas exchange models
developed for each type of commodity as a function of controllable variables, i.e. gas composition
and temperature, and some models also take into account the storage time or elapsed time from
harvest (Fonseca et al., 2002a; Mahajan et al., 2007; Uchino et al., 2004). The success of the models
is dependent on fundamental understandings of the relationships among atmospheric composition
and physiological changes for construction of the optimal mathematical algorithms (Saltveit,
2003b).
The model should be carefully developed, as a MAP system incorrectly designed may be ineffective
or even shorten the shelf life of the product (Mahajan et al., 2007). Changing conditions (e.g. in
temperature) will affect respiration, and therefore change the rate of atmosphere modification. This
will expose the product to unfavorable gas composition leading to quality changes, as described in
following section 3.3.4. Therefore, the properly designed package, relative to both material and
size, should take into account the dynamics of respiration rate at changing conditions (Fonseca et
al., 2002a; Jacxsens et al., 2000; Tano et al., 2007).
3.3.4 Effect of gas composition on produce quality
Several studies have shown the effect of changing the atmosphere on a wide range of fruits and
vegetables; however, the effect on the different commodities have varied considerably (Saltveit,
2003b; Wills et al., 2007), and the effect on lowering the respiration rate of wild rocket proved to be
sparse or nonexistent (Figure 13, section 3.3.1). Some recommendations for gas composition of
selected vegetables are summarized in Figure 14, where the recommended O2 and CO2
concentrations for storage of different vegetables are indicated by shaded areas.
33
Figure 14. Recommended O2 and CO2 combinations for the storage of some harvested vegetables. From Saltveit (2003b).
Apart from the effect on respiration rate, there are also other beneficial results of modified
atmosphere environmental conditions that cannot be directly linked to respiration rate (Beaudry,
1999; Kader and Saltveit, 2003a; Wills et al., 2007), such as the direct antimicrobial effect of CO2
(Sandhya, 2010) and the fact that low O2 inhibits the enzymatic browning as the reaction is
dependent on O2 (Figure 9, section 2.5.1). Other quality parameters important for determination of
optimal gas composition are specifically retention of green color and volatile compound
metabolism.
In green vegetables, improved retention of green color in low oxygen atmosphere is due mainly to a
directly lowered rate of chlorophyll destruction rather than solely to a lowered respiration rate and
resultant lowered chlorophyll destruction (Kader and Saltveit, 2003a; Wills et al., 2007). Beaudry
(1999) also concluded that it was likely that the control of chlorophyll loss by low O2 and elevated
CO2 in many green tissues might be due partly to the effects of these gases on ethylene production.
However, a complete understanding of the mode of action of CO2 in the reduction of chlorophyll
loss is lacking. Schouten et al. (2009) developed a model for the color of broccoli based on
temperature and gas composition and found that the decay of chlorophyllide was affected by the gas
composition. However, the effect of low O2 and high CO2 was only pronounced at high temperature
34
(Schouten et al., 2009). Cefola et al. (2010) investigated the color retention of broccoli raab (a
broccoli-like green vegetable consisting of inflorescences and leaves) at 5 °C and varying O2 and
CO2. Some positive effect of storage at low O2 was found; there was no additional effect of high
CO2 (Cefola et al., 2010).
During the research project, the effect of packaging materials with different OTR on the quality of
wild rocket was investigated at 5, 10, and 20 °C during different seasons. Figure 15 shows the gas
composition in packages from the late summer harvest. At low OTR, the O2 concentration reached
1 kPa after seven days, and hereafter it was close to 0 kPa, leading to anaerobic respiration.
Similarly, the CO2 concentration reached 14 kPa after seven days, and hereafter it decreased
slightly, as the efflux was higher than the CO2 produced by respiration of the wild rocket, indicating
cell death due to anaerobic respiration. At medium OTR, the O2 concentration was slightly
decreased to 17 kPa, and it did no change considerably during storage. Likewise, the CO2
concentration did not change during the storage time, and it was approximately 5 kPa. At high
OTR, the O2 concentration was almost atmospheric with 20 kPa until after day 15, when it
decreased slightly to 18 kPa, which might be due to increased respiration at the end of storage as
explained in section 2.6. The CO2 concentration followed almost a mirrored path of the O2
concentration, starting at 2 kPa and ending at 3 kPa. The concentration gradient of CO2 across the
film will often be higher than that of O2, which will lead to a faster exchange of CO2 with the
surroundings. In addition, the transmission rate of CO2 is generally higher than that of O2 in non-
perforated film (Mangaraj et al., 2009).
Figure 15 only covers one temperature and one harvest season. Results were that temperature and
harvest season had a major impact on respiration rate (Figure 11, section 2.6). This also influenced
the results of the packaged wild rocket, e.g. at low OTR in the higher respiring spring harvest, the
O2 concentration reached 0 after only four days at 5 °C, and at 20 °C in late summer harvest, O2
reached 0 kPa after just two days.
35
Figure 15. Gas composition in packages of wild rocket from late summer harvest (week 32) stored at 5 °C in low OTR (1900 ml O2/m
2/day/atm), medium OTR (7140 ml O2/m2/day/atm), or high OTR (48400 ml O2/m
2/day/atm). Error bars are the standard deviation between the five replicates at each point. Unpublished data and data presented in Paper IV.
As stated above, the main risk of MAP is that anaerobic respiration is initiated, resulting in the
formation of acetaldehyde, ethanol and acids. In paper III it was noted that low permeability and
low O2, as seen in the packages of low OTR in Figure 15, lead to an olive brown color, probably
due to acid catalyzed loss of magnesium from chlorophyll creating pheopytin (Figure 8, section
2.5.1).
However, the color change is not the first sign of fermentation, as off-odor is often present at an
earlier state (Luo et al., 2004). Volatile organic compounds (VOC) other than acetaldehyde, ethanol
and acids might very well be produced. Anaerobic respiration may also occur. In Paper III, a drastic
change in the odor was detected in fermented packages, and it was described as a smoked odor.
Similar results of off-odor development in anoxic MAP of leafy vegetable have been reported
before, although the off-odor was not characterized as a smoked odor (Allende et al., 2004; Ares et
al., 2008a; Kim et al., 2004; Lonchamp et al., 2009; Luo et al., 2004; Nielsen et al., 2008). In an
investigation of the origin of off-odors in packaged rocket, research showed that dimethyl sulfide
(DMS) and dimethyl disulfide (DMDS) were the main compounds responsible for off-odor and that
they especially evolved in an anoxic environment (Nielsen et al., 2008). The compounds produced
during fermentative metabolism, together with a decrease in cytoplasmic pH (Sousa and Sodek,
2002), eventually led to cell malfunctioning and death. The fermented wild rocket salad in Paper III
furthermore experienced loss of texture due to collapse of cells. This was also seen in loss of
product height, and exudates from collapsed cells appeared at the bottom of the tray.
There were some indications that a slight modification of the atmosphere, as experienced in the
medium OTR packages (Figure 15) compared to packaging at high OTR, was beneficial for keeping
the green color in packed wild rocket. The effect of the medium OTR is further examined in Figure
36
16 where the gas composition in packages of medium OTR stored at 5, 10, and 20 °C is shown.
Temperature clearly has a large influence on the gas composition of packages: At 5 °C, the O2
concentration is approximately constant at 16-18 kPa, at 10 °C, the O2 concentration is slowly
decreasing throughout the period ending at 10 kPa, while at 20 °C, the O2 concentration is
decreasing fast ending at 1 kPa after six days; the wild rocket showed severe signs of senescence
before day six. It was noted in Paper II (Figure 11, section 2.6) and in Figure 13 (section 3.3.1) that
the respiration rate was not affected by the modifications of the atmosphere seen in Figure 16;
consequently, the retention of green color must be due to factor(s) other than lowering of the
respiration rate.
Figure 16. Gas composition in packages of middle OTR (7140 ml O2/m2/day/atm) stored at 5, 10, or 20 °C. Wild rocket
from late summer harvest (week 32). Each point is an average of five replicates, and error bars are the standard deviation between the five replicates. Unpublished data.
Kader and Saltveit (2003b) stated that in an overall relative comparison of the factors affecting
respiration rate it is obvious that temperature is the most important factor, as also found in Paper II,
followed by O2 concentration, and wounding. Additionally, Saltveit (2003b) stated that the natural
variability in the raw material and its dynamic response to processing and storage conditions make
it difficult to predict respiration rate. The study in Paper II revealed that harvest season and variety
indeed have a high influence on respiration rate, and in some cases to the same degree as
temperature (Figure 11). Furthermore, the study revealed that the influence of O2 concentration on
the respiration rate of wild rocket and broccoli above 6.3 kPa O2 was very low (Figure 11). The
study was supplied with the additional unpublished results in Figure 13A of wild rocket stored at 5
°C in O2 concentrations below 10 kPa, where the respiration rate did not seem to change until the
RQ disclosed a change to anaerobic respiration at O2 < 0.5 kPa (Figure 13D).
There were some indications in Figure 13B that the respiration rate of wild rocket stored at 10 °C
may be lowered at O2 concentrations between 0.5-2 kPa. Nevertheless, considering the temperature
37
variations during storage (Figure 11, section 3.2), the variation between different harvest times and
the data in Figure 15 and Figure 16, it is obvious that it will not be possible to reach these
concentrations throughout the storage period by MAP. The optimum gas composition in storage of
wild rocket leaves may therefore rely on color retention and not respiration rate, as long as aerobic
respiration can be maintained. One of the difficulties involved in such a model is how to measure
green color correctly. Furthermore, large variations in color degradation between the different
harvests were experienced during the experiments. However, the data need further consideration
before publishing.
In summary, packaging is important to protect leafy vegetables from water loss and to ease
handling. Packaging changes the gas composition and will have an effect on product quality.
Altered O2 concentration above 0.5 kPa did not have a marked effect on respiration rate of wild
rocket, although there were some indications that some modification of the gas composition had an
effect on retention of green color. In MAP, interplay between the respiration rate and the
permeability of the film determines the gas composition, and modeling respiration rate can be used
to find the optimum permeability for packaging in the given situation.
Overall, wounding, even during gentle processing and handling, is unavoidable and will lead to
quality changes. Temperature should be kept low from field to fork for optimal quality, and
packaging can be effective to prevent water loss. The gas composition inside the packages is
dependent on the interplay between the respiration rate and the permeability of the film. The
success of packaging to control quality is dependent on many factors including temperature, biotic
variation, and other abiotic factors that influence the respiration rate. Accumulated temperature
has a great influence on the respiration rate, and hence on shelf life.
38
4 Postharvest quality assessment
When performing postharvest experiments, many different methods can be used to quantify the
quality. The use of fast methods makes it possible to analyze large numbers of samples which is
required to cover the biological variance. Furthermore, fast methods may enable continuous
measurement and thereby an insight into the time domain of occurring quality changes. Often, fast
methods do not have the same detailed quality information as the normally used analyses, and
consequently, fast methods must be correlated to a more informative and often slower analysis.
Methods used in this thesis are presented and discussed in the following sections.
4.1 Measurement of respiration rate
Respiration rate is a key measure in postharvest science since it is an overall measure of the
metabolic activity. The determination includes not only the cellular respiration process but also the
gas exchange process and the respiration of microorganisms as well as any other plant physiological
processes that involve O2 and CO2 (synthesis of plant hormones, oxidation reactions, and
photosynthesis). Therefore it is not possible to get a ‘clean’ value for respiration rate (Fonseca et al.,
2002a).
Respiration rate is measured as the amount of produced CO2 or consumed O2 per weight unit per
time unit. It is often measured in either a closed or flowing system (Figure 17).
Figure 17. Schematic representation of measurement of respiration rate in the closed system (A) and the flowing system (B). Photo: Jens Madsen.
In the closed or static system, the product is kept in closed containers and the change in O2 and/or
CO2 concentration is measured over a period of time (Figure 17A). The initial gas composition is
often atmospheric (Escalona et al., 2006; Iqbal et al., 2009; Jacxsens et al., 2000) or the container
can be flushed before onset of the measurement (Escalona et al., 2006; Fonseca et al., 2002b; Iqbal
et al., 2009; Torrieri et al., 2010). The respiration rate is calculated from the difference between two
O2CO2
O2 CO2
Gas in Gas out
A B
39
measurements. This system is easy and fast to set up. A drawback is that the gas composition is not
constant and cannot be controlled during the measurement period. The equation for calculation of
O2 respiration rate in the closed system is:
Closed system: �2 ��������� �� ���2���� 1 � ��2���� 2� � ���� ������100 ���� �� ������� � ���� 2 � ��� 1� (3)
In the flowing or open system, the container is continuously flushed with gas (Figure 17B), and the
composition is measured at the inlet and outlet. Hence, in this system gas composition is controlled.
The respiration rate is calculated from the difference between inlet and outlet. The gas composition
can be varied during the experiment and be any mixable combination of e.g. N2, O2 and CO2. Many
have used atmospheric O2 concentration (Allende et al., 2004; Kim et al., 2004; Kim et al., 2005a;
Kim et al., 2005b; Koukounaras et al., 2007; Martinez et al., 2005; Martinez-Sanchez et al., 2008)
and some have tested different combinations of gasses (Kim et al., 2005a; Kim et al., 2005b;
Martinez-Sanchez et al., 2006). It should be ensured that the inlet gas is humidified and that the
flow is constant (Kader and Saltveit, 2003b). The system is therefore more complicated to setup
than the closed system. However, it is much more flexible concerning gas composition and can be
operated for a long time. The equation for calculation of O2 respiration rate in the flowing system is:
Flowing system: �2 ��������� �� ���2��� � ��2����� ��� ��100 ���� �� ������� (4)
The closed system is best suited for low respiring products and the flowed system for high respiring
products. The equation for calculation of respiration rate is dependent on the chosen system as seen
in equation 1 and 2. Therefore, the closed system is very sensitive to the determination of free
volume in the container and the flowed system to determination of flow rate (Fonseca et al., 2002a).
Both systems rely on accurate measurement of O2 and CO2.
The most common method to determine O2 and CO2 in respiration analysis is gas chromatography
(GC) (Allende et al., 2004; Escalona et al., 2006; Fonseca et al., 2002b; Kim et al., 2004; Kim et al.,
2005a; Kim et al., 2005b; Martinez et al., 2005; Martinez-Sanchez et al., 2006) or
electrochemical/infrared gas analyzers (Iqbal et al., 2009; Jacxsens et al., 2000; Koukounaras et al.,
2007; Martinez-Sanchez et al., 2008; Torrieri et al., 2010). Most studies rely on snapshots of
respiration rate and do not take into account the dynamic nature of respiration. As the development
of respiration rate is a continuous process and seldom linear, the discrete data points should be
chosen with great care (Fonseca et al., 2002a). In Paper I, a novel sensor type concomitantly
measuring temperature and oxygen concentration was used enabling continuous measurements
within the closed system (Figure 18). The sensors were originally developed for wireless
40
measurement of temperature and O2 in silage stacks (Green et al., 2009; Green et al., 2012) and the
study in Paper I was the first time they were tested in postharvest experiments. The use of wireless
sensors for continuous measurements enabled us to study the impact of temperature and the
immediate response in respiration rate (Figure 10, section 2.6 and Figure 11, section 3.2).
Figure 18. Measurement of respiration rate in the closed system. Manual measurement with the fast electrochemical and infrared gas analyzer (A). The experimental setup with the O2 sensor protected by a plastic housing (B). The plastic housing was placed in a glass jar of 1 L, and plant material was placed above and around the housing in the glass jar. Close-up of the sensor (C). The white circle in the middle is the oxygen sensor. Modified from Paper I, photo Jens Madsen and AU-foto, Jesper Rais.
The methods chosen for measurement of gas composition in Paper II were a fast electrochemical
and infrared gas analyzer and the wireless sensors (Figure 18). The result from the measurements
was a continuous O2 content and the respiration rate was calculated for each point by first derivative
using a local, second order polynomial Satvitzky-Golay smoothing filter (Figure 19). A positive
consequence of this was that each value of the respiration rate was connected to one point in time
and therefore also one O2 concentration. With discrete measurements of respiration rate in the
closed system, where the gas composition changes in time, and each calculated value of respiration
rate can be connected to either the first or the second measurement of O2, or more correctly all
timepoints in-between and an average value of the O2 concentration of the two points.
A CB
41
Figure 19. Changes in O2 concentration (A) and respiration rate (B) as a function of storage time. Measurements were performed in a closed system with broccoli florets harvested in early and late summer and stored at target temperature 5 °C. Measurement with the O2 sensor. Modified from Paper II.
These methods made it possible to conduct many experiments at the same time, though probably
with the price of a less accurate determination of O2 concentration compared to if GC
measurements were used. The variability of the measurements from the gas analyzer can be seen in
Figure 13 (section 3.3.1). The difference between discrete measurements and continuous
measurements concerning time resolution can be investigated by comparison of Figure 13 (section
3.3.1) and Figure 11 (section 3.2). The low time resolution of the discrete measurements in Figure
13 (section 3.3.1) made it difficult to evaluate the effect of gas composition on respiration rate.
Oppositely, the continuous measurements by the sensor in Figure 11 (section 3.2) showed the
immediate reaction of respiration rate to temperature changes.
Measurement of respiration rate is normally a time and labor intensive method (Fonseca et al.,
2002a), but the combination of the closed system with the wireless sensors eased the measurement.
Only one other study determining O2 continuously postharvest by using an O2 electrode to determine
oxygen consumption in packaged tomatoes (Cameron et al., 1989) has been found. Torrieri et al.
(2010) used a sensor too, but for continuous measurement of temperature and humidity inside each
jar during respiration measurement.
There are many possibilities for improvement of the sensor method. The sensors could not measure
O2 concentration below 6.3 kPa, and an obvious improvement would be measurement at lower
concentrations. With measurement of CO2 it would be possible to calculate RQ (section 2.3) and
thereby find the turning point from primarily aerobic to primarily anaerobic respiration. Another
possible change could be to exclude formed CO2 by addition of a CO2 absorber to the jar and
thereby only study the effect of O2 on respiration rate.
0 5 10 15 20 250
5
10
15
20
25
Storage time (h)
Oxyg
en
co
nce
ntr
atio
n (kP
a)
Early summer
Late summer
A
6
7
8
9
10
11
12
0 5 10 15 20 250
20
40
60
80
Storage time (h)
Re
sp
ira
tio
n ra
te (m
l O2kg
-1h
-1)
Early summer
Late summer
B
Temperature
6
7
8
9
10
11
12
42
In summary, the developed continuous method for measurement of respiration rate gave a unique
insight into the close correlation between temperature and respiration rate and improved the
interpretation of correlations between temperature, O2, and respiration rate.
4.2 Sensory descriptive analysis
Sensory analysis is the ultimate tool for evaluation of quality and quality changes (Martens and
Martens, 2001). Sensory descriptive analysis performed by a sensory panel can be used to
determine the underlying attributes influencing freshness of a food product (Peneau et al., 2007a).
However, to determine the level of rejection of the food product, consumers must be involved (Ares
et al., 2008a; Ares et al., 2008b; Peneau et al., 2007a).
During the training phase in sensory descriptive analysis, the sensory panel creates and defines
attributes able to describe the food product which is being evaluated (Lawless and Heymann,
2010). The attributes should preferably be orthogonal and non-redundant, meaning that they should
vary independently of each other and not explain the same variation in the food products. In the
ideal world, they should be singular rather than combinations of several terms. However, sensory
descriptive analysis has been performed successfully with mixed terms such as creaminess and
complexity (Frøst and Janhøj, 2007; Paulsen et al., 2012). Freshness which was of particular
interest in this thesis is also a mixed term (Lawless and Heymann, 2010).
In Paper III, it was found that even though freshness is a mixed term, the sensory panelists were
agreeing on the attributes that described freshness. Additionally, it was found that some packages of
wild rocket salad decreased in freshness from before opening to after opening. This means that the
freshness perceived only by appearance was not always in correspondence with the actual freshness
experienced at the point of opening the package for potential consumption. Furthermore, all
categories of tested attributes, color, odor, and texture were important for perception of freshness at
the point of consumption. The appearance or color were not enough to judge the actual freshness of
the product.
The disadvantage of sensory descriptive analysis compared to some faster sensory methods is that
the panel must be trained for the specific product category. Additionally, the results from sensory
methods are relative and not absolute, since the panelists may use different scale ranges.
In the sensory descriptive analysis in Paper III, some of the variables showed redundancy, e.g. the
texture variables turned out to describe almost the same variation. This indicates that though the
loss of texture can be seen (and sensed) using different sensory variables, they are all indications of
the same phenomenon: the degradation of the cell structure.
43
Texture and appearance change with both physiological and processing events and are thus heavily
time-dependent. It is therefore a tedious and not easy task to use sensory descriptive analysis for
postharvest quality studies as it is very labor intensive to make sensory analysis as a routine control.
Additionally, the quality changes are very dynamic and different batches behave differently which
makes it difficult to produce replicates. The approach of Paper III was to produce different degrees
of senescent and fermented packages from the same batch of wild rocket salad. This was achieved
by packaging in two different packaging materials of and by storage at elevated temperature for
varying time intervals prior to the sensory analysis. This approach meant that the control sample
was not a completely freshly harvested sample, but a sample stored at 2°C. During the experiment,
wild rocket salad from spring to late summer was used as test material. Due to the long seasonal
period, a large disparity in leave size and color degradation was experienced and it was clear that
batch variation should have been avoided in the sensory study in order to achieve significant results
with regards to experimental design of temperature and packaging.
In summary, sensory descriptive analysis could be used to describe the freshness of packaged wild
rocket salad and the underlying attributes of overall freshness were color, texture, and odor of the
wild rocket salad.
4.3 Analysis by visual and near infrared light
Analysis by visual and near infrared light is a nondestructive fast method. Color is an important
feature of foods and the sensation is caused by light and absorption of visual light (Hutchings,
1999). When the object is subjected to light, it may be reflected, transmitted and/or absorbed
(Berns, 2000). Reflectance is both specular reflectance as in a mirror and diffuse reflectance (or
scatter) caused by rough surfaces. Transmission refers to light going through the object, and
absorption is light lost inside the object. The transmitted light is therefore the light not absorbed or
reflected by the object. The absorption is dependent on the chemical composition of the object, and
different parts of the electromagnetic spectrum are absorbed by different molecules. Scatter of an
object is a physical phenomenon and the physical changes may also be due to chemical changes.
4.3.1 Visual wavelengths
The visual range of the electromagnetic spectrum goes from 380-700 nm (Berns, 2000). A color
appears when light in some part of the range is absorbed and other parts are reflected by the
material. If all light is reflected, the object will appear white, and if all is absorbed, it will appear
black. The wavelengths influence the perceived color which therefore is dependent on both the light
source and the absorption by the object.
The green color of vegetables is due to the pigment chlorophyll and the yellow/orange color to
carotenoids as described in section 2.5.1. Chlorophylls absorb light at several wavelengths (Figure
44
20), and chlorophylls appear green because light is reflected between 500 and 600 nm (Salisbury
and Ross, 1992). Carotenoids are various compounds and they mainly absorb light between 400 and
550 nm, reflect light above 550 nm, and appear yellow or orange (Figure 20) (Salisbury and Ross,
1992).
Figure 20. Absorption spectra of chlorophyll and carotenoids. From www.cfb.unh.edu
Many systems have been developed through time for reporting color. The CIELAB is a color space
defined by the CIE (the International Commission on Illumination) and the official terminology is
the CIE 1976 L*, a*, b* space (Berns, 2000). CIELAB is widely used and accepted for reporting
food color (Hutchings, 1999). The CIELAB system is a three-dimensional co-ordinate system
defined by the three axes L*, a*, and b*, where L* is the lightning, a* the redness-greenness, and
b* the yellow-blueness (Figure 21), and a color is uniquely defined by the coordinates of the three
values. Hue is the angle on a circle in the a-b coordinate system and is often used when reporting
fruit and vegetable color (McGuire, 1992).
45
Figure 21. The CIELAB color space. L* is the lightning and goes from 0 (black) to 100 (white), a* the redness-greenness, and b* is the yellow-blueness. C* is the chroma (saturation) and Hue the angle measured in degrees starting with 0 in the +a* direction. Information from Berns (2000).
Spectrophotometric determination of color is a fast measurement of quality, and the result from the
measurement is often given in CIELAB values. Typically, the measurement is performed in a
limited area, determined by the apparatus aperture, which was 8 mm in Paper IV. Therefore, in case
of leafy vegetables, every leaf has to be measured individually and many repetitions must be made
in order to cover the variation in a package (Luo et al., 2004). This imposes some challenges and
limitations for color determinations in very heterogeneous food materials, such as the wild rocket
leaves. This was investigated in Paper IV and it was found that it was difficult to cover the variation
occurring in a heterogeneous package of wild rocket leaves with the spectrophotometer. Others
have come to the same conclusion in mushrooms and spinach leaves (Lunadei et al., 2012;
Taghizadeh et al., 2010).
4.3.2 NIR wavelengths
Near-Infrared (NIR) spectroscopy is the use of light that by definition covers the wavelengths from
780-2500 nm. NIR spectroscopy is an increasingly used technology for non-destructive
measurement of the quality of fresh fruit and vegetables (Lleo et al., 2009; Nicolai et al., 2007). In
the NIR region, the absorption is due to overtones and the combination tones of the fundamental
infrared (IR) vibration bands of bonds where the electric dipole moment changes; anharmonic
bonds (Thygesen et al., 2003). These are the stretching, bending, rocking, wagging, and twisting of
the covalent bonds of dominantly C-H, O-H, N-H, C-O and C=O. Each of these vibrations has a
natural frequency depending on the environment in which the bond is present. The spectrum can
therefore be used as a fingerprint of the chemical composition of the sample. Postharvest research
often reports either multispectral or hyperspectral NIR.The designation multispectral normally
C*
Hue
-a*
-b*
a*
L* 100
-b*
L* 0
46
covers a few noncontinuous wavelengths and hyperspectral covers a continuous range of
wavelengths (Nicolai et al., 2007). The light penetration depth depends on the wavelength and the
sample characteristics and is up to 4 mm in the 700-900 nm range (Lammertyn et al., 2000; Nicolai
et al., 2007).
The emitted scatter from the object can be measured and reveals information about chemical
composition and surface structure. Scattering of fresh produce is related to the microstructure of the
tissue and absorption is related to the presence of chemical components. Therefore both
phenomenons are useful in determination of quality (Nicolai et al., 2007). Water is the most
important constituent of fresh produce and as water has high absorption of NIR light, the NIR
spectrum of fresh produce is dominated by water (Buning-Pfaue, 2003; Cen and He, 2007). Even
simple molecules show broad bands in NIR due to overlapping tone, overtones and combination
tones as well as differences in the local molecular environment, i.e. hydrogen bonding. The
information in the NIR spectrum is therefore blurred and consequently, chemometrics are important
for extraction of useful information from the spectra. The combination of NIR and chemometrics is
widely used in the food industry for process control (Cen and He, 2007), see also section 5. In order
to enhance the chemical information from NIR, mathematical scatter correction is often applied,
though many of the methods imply hyperspectra (Rinnan et al., 2009). NIR measurements are
easily automated and after proper calibration, the method is very useful.
NIR has been used to a large extent postharvest to predict soluble solids content (SSC) in fruit, a list
of references is included in the review of Nicolai et al. (2007). Scientific records of NIR
applications with green vegetables are sparse. The content of essential oils of dill and coriander was
predicted by NIR (Schulz et al., 1998). The sensory quality of chicory was predictable using vis-
NIR spectroscopy (Francois et al., 2008), and SSC and firmness were predicted in bell pepper
(Penchaiya et al., 2009). The chlorophyll content of leafy green vegetables could be predicted by
hyperspectral reflectance (Xue and Yang, 2009).
In Paper IV, a decrease in NIR reflection was observed when the tissue was degraded due to either
senescence or fermentation. The decrease in reflection may be due to both to increased water
absorption due to the release of cell water into the intercellular space and to the physical
phenomenon of reduced scatter due to the breakdown of cell walls. Light diffusion in tissue is
affected simultaneously by absorption and reflection and the two properties can therefore not be
measured independently (Nicolai et al., 2007).
4.3.3 Imaging
RGB imaging is a simplified case of multispectral imaging. In a normal image, every pixel holds
the information of three broad band wavelengths in the visual range; R, G and B (Gowen et al.,
2007; Lleo et al., 2009). In a multispectral image, every pixel holds the information of all the
47
measured wavelengths. Each image therefore includes a large amount of data. The problem can be
solved by a simple approach (Paper IV), where the image – the spatial information – first is used to
distinguish the objects of interest from the background. When the objects (in this case the wild
rocket leaves) in the image have been found, all pixels of these are averaged, and thereby each
image is represented by one spectrum (Figure 22). The price of simplicity is that the spatial
information is lost, including information of shadows and heterogeneity in the picture. In the
situation in Paper IV, the result should be correlated with the sensory overall perception of a
package of wild rocket leaves and not single leaves and the procedure was considered as an
objective and valid way of analyzing the data.
Figure 22. Average spectra from the VideometerLab of control samples, samples stored at low OTR (lowOTR20-6) and samples stored at high OTR (highOTR20-6). The bar indicates the region of the visual (VIS) and the NIR region of the spectrum. The pictures show leaves of wild rocket salad as they appeared in the packages. From Paper IV.
As soluble solids content of fruit is a popular application for NIR, and hence spectral imaging has
also been tested widely on this application (Nicolai et al., 2007), spectral imaging also has the
ability of detecting changes in firmness, bruises, and other defects where the spatial information can
be used (Abbott, 1999; Ariana and Lu, 2010; ElMasry et al., 2008; Lleo et al., 2009; Peng and Lu,
2008). Recently, VIS-NIR imaging has also been applied on different vegetables with the purpose
of testing its usability for quality control. A hyperspectral imaging system was used to predict L*-
value, an important parameter for mushrooms (Taghizadeh et al., 2010). Quality of packed spinach
was measured by an IR-R-B image system where the color change of injured leaves was detected
(Lunadei et al., 2012)
The commercial multispectral imaging device, the VideometerLab (Figure 23), is used in several
industries in Denmark. In newly published papers, the instrument has been used for determination
of astaxanthin in salmonids (Dissing et al., 2011) and for seed health testing of spinach (Olesen et
al., 2011) and in Paper IV it was for quality measurement of wild rocket salad. The instrument has
400 500 600 700 800 900 10000
20
40
60
80
100
120
nm
Mean in
tensity
Control
HighOTR20-6
Low OTR20-6
VIS NIR
48
the advantage that it is easy to use and simple image analysis is built into the software. Compared to
hyperspectral imaging instruments it can be seen as a disadvantage that the instrument covers fewer
wavelengths, on the other hand, images with fewer wavelengths facilitate data analysis (Gowen et
al., 2007).
Figure 23. Schematical setup of the VideometerLab. Optimal lightning is ensured by an integrating sphere coated with a matte white coating. In the rim of the sphere, a set of narrow band light emitting diodes ranging from 405 to 970 nm are mounted. The image acquisition is performed by a monochrome grayscale CCD camera mounted in the top of the sphere. The arrows illustrate how the light is distributed inside the sphere to uniformly illuminate the sample. Modified after Dissing et al. (2011).
The instrument was used for predicting some sensory qualities of packed wild rocket salad (Paper
IV). CIELAB values were calculated from the multispectral images and compared to results from
the spectrophotometer. The measured area was larger in the multispectral images and specified it
from the background and thereby the sampling situation was better in the multispectral images.
Especially the Hue changed during storage of packed wild rocket leaves, indicating a color change
from green to yellow which is in correspondence with other findings (Koukounaras et al., 2007;
Luo et al., 2004). The change in Hue was not as significant as the sensory-evaluated perception of
yellow color in Paper III. Furthermore, the spectra in Figure 22 was correlated to sensory attributes
and it was found that the visual band could be used to describe color changes and the NIR bands
could be used to detect textural changes occurring during aging (Paper IV).
In summary, VIS and NIR light can be used to describe color and textural changes of packed wild
rocket salad by multispectral images.
4.4 Analysis of odor
Odor from MA-packaged vegetables may come from various sources. The characteristic positive
odor of wild rocket salad is especially released at wounding as explained in section 3.1. As the cells
Sample
49
collapse due to senescence, other odors are also released (section 2.5.2). These odors can be
considered off-odors and can be described as rotten or cabbage-like (Paper III). They are originating
from mainly sulfurous breakdown products by microorganisms and of senescence (Nielsen et al.,
2008). Another odor that may be experienced from packed wild rocket salad is a strong off-odor
development as a function of anaerobic respiration in malfunctioning MAP as described in section
3.3.
In its nature, perception of odor can only be judged by humans. Even though many volatile organic
compounds (VOC) might be measurable in the headspace of an item, they are not necessarily
important for the percieved odor. This is due to several factors: An odor threshold is defined as the
smallest concentration detectable by humans, and can be used for an estimate of the perception of
off-odor based on measured concentrations. However, this assumption does not take into account
that the contribution from the individual VOC might not be additive, but could be either synergistic
or antagonistic (Hansen, 2011). Measurements of VOC should therefore be correlated to sensory
odor determination.
Measurement of off-odor and specific odorants in packed vegetables may be an early sign of
fermentative metabolism. Nielsen et al. (2008) analyzed the origin of off-odor in packed rocket
salad, but there was no clear distinction between the different sources of odor; microorganisms or
fermentative metabolism. Dimethyl sulfide (DMS) and dimethyl disulfide (DMDS) were detected
as important odorants by gas chromatography with olfactometry detection (GC-O) and quantified
by gas chromatography with mass spectrometry detection (GC-MS).
During the experiments in this project, it was found that the off-odor experienced right after
opening of a package quickly diminished, indicating that the off-odor was due to ultra-volatile
VOC. The ultra-volatiles are difficult to trap quantitatively using sampling methods for GC-MS
and, moreover, many sulfurous compounds are usually degraded during trapping and analysis
(Lestremau et al., 2004). Proton-Transfer-Reaction - Mass Spectrometry (PTR-MS) has been
identified as an alternative and fast method to GC. It has been demonstrated to provide specific and
sensitive measurement of odorants including sulfurous compounds and that these could be
correlated to measurements of odor (Feilberg et al., 2010; Hansen et al., 2012). The PTR-MS
measures compounds directly from headspace from the sample, and is in essence a true online
method. Furthermore, detection limits are extremely low (ppt level) and the method can have a fast
response time to changes in headspace in milliseconds (de Gouw and Warneke, 2007). Since each
measurement takes less than a minute, continuous measurements are possible, and the development
of volatile components can be monitored continuously. Recent research has investigated various
applications of the PTR-MS in food science and has showed its great potentials (Biasioli et al.,
2006; Biasioli et al., 2011; Gallardo-Escamilla et al., 2007; Heenan et al., 2009; Lindinger et al.,
2008; Lindinger et al., 2009; Phillips et al., 2010). In a recently published storage study of broccoli,
50
PTR-MS measurements correlated well with changes in the activity of enzymes that influence the
quality of broccoli (Raseetha et al., 2011).
A drawback of PTR-MS and other fast measurements of odor is that a reference method, e.g. GC-
MS, is needed to assign masses to compounds. In a preliminary study during this project, the
volatile profile of anaerobically packaged broccoli florets was trapped on thermal desorption tubes
and analyzed by GC-MS. The result showed very high amounts of DMDS. PTR-MS was also tested
with measurements directly of the headspace in the packages (Figure 24) and showed only
methanethiol which is known to be formed by broccoli under anaerobic conditions (Tulio et al.,
2003). Moreover, the headspace contained alkane compounds originating from the plastic material
and these compunds were very dominating in the chromatograms. Work remains in order to get
good results from the GC measurements. PTR-MS cannot measure all compounds which may be a
drawback, as e.g. ethylene and other alkene and alkane compounds are not measured. This may also
be an advantage as alkane compounds emitted from the packaging material are not disturbing in
measurement of headspace-packaged food.
Figure 24. Analysis of VOC from the headspace packaged broccoli florets. The headspace was drawn through Teflon tubes directly connected to the PTR-MS. Photo: Mette Marie Løkke.
In summary, PTR-MS is a potential candidate for fast measurement of odor of packaged wild rocket
salad and a method like GC-MS should be used for identification of compounds.
Overall, novel, fast analytical methods make it possible to analyze large numbers of samples which
is required to cover the biological variance. Simultaneous measurements are informative for
investigation of occurring correlations, and continuous methods give a unique insight into the time
domain of the reactions. Sensory analysis and other slower methods are often more informative
than fast methods and the slow methods are important for interpretation and correlation of the fast
and continuous methods.
51
5 Chemometrics in postharvest quality assessment
Chemometrics is a multivariate data analysis with origin in chemistry, where large data matrices are
generated. Hence, the name ‘chemometrics’. When quality measurements result in large data
matrices, e.g. spectral data or continuous data, it is an advantage to use chemometrics and thereby
obtain an overview of the underlying patterns in data. Chemometrics is therefore widely used and
accepted in food science and industry.
In chemometrics, the real world is under observation, and a data matrix is resolved into principal
components that can be handled by classical statistics and visualized in graphics (Munck et al.,
2010). In chemometrics, mathematical and statistical methods are used to extract information from
chemical data (or any other kind of data). The approach can be made highly visual, making the
methods available to a wider range of users than merely skilled statisticians.
In general, modeling is about separating structure from noise:
Observed data = structure + noise
(5)
The structural part constitutes the model. The purpose of the model, whether it is a traditional (hard)
model or a chemometric (soft) model, is to bring insight into the studied system. Chemometrics is a
hypothesis generating process, as it is not assumed beforehand that the data follow a special path;
instead the systematic pattern of variation in the data is studied (Martens and Martens, 2001). Often
unexpected patterns are revealed when the joint effect of all variables are taken into account
(Munck et al., 2010). However, the chemometric data analysis should be closely linked to the
scientific area of observation (Wold, 1995). The knowledge gained from chemometrics should
always be judged and compared with chemical and biological knowledge. The problems are defined
by the applications and not by the math. Furthermore, it is important to notice that models can only
indicate correlation and not causation (Kjeldahl and Bro, 2010).
All types of data have inherent noise. The noise may come from three sources (Wold, 1995):
Firstly, lack of full control of the experimental conditions. In the experimental basis of this thesis,
for instance, a set temperature of 10 °C might actually be 9.9 °C in the first experiment and 10.1 °C
in second. Sampling error due to heterogeneity can also be considered a lack of full control.
Secondly, the instrument will often not be stable. This is obvious when dealing with sensory data,
but it holds true for all instruments to some extent. Finally, there are model errors or assumption
errors, e.g. Beer’s law, in the definition only applies under ideal conditions, but is assumed to be
valid (which is often the case) in spectral analysis (Martens and Martens, 2001).
An important part of chemometric analysis is the study of the noise part, as it is used in the
judgment of the model. When plotted, the noise part should be free of structure and could be used
for detecting outliers both in objects and variables. In practice, chemometrics is an iterative process,
52
where different features from the chemometrics toolbox, e.g. pretreatment, scaling, variable
selection, number of components, sub-models are tested with the purpose of finding the
combination that explains most of the structure and leaves out the noise. At the same time, the
model should not include noise (be overfitted), and therefore proper validation of models is
important.
5.1 Data exploration: PCA
Principle component analysis (PCA) (Hotelling, 1933; Wold et al., 1987) is the workhorse of
chemometrics. It is an unsupervised method, meaning that the model is not guided in a
predetermined direction. In a PCA, the data matrix is decomposed into a structural part of a few
components and residuals:
X = TP’ +E
(6)
where X is the original data matrix that is decomposed into a score matrix (T), which is connected
to the samples, and a loading matrix (P), which is connected to the variables. This is calculated by
consecutive orthogonal subtraction of the largest variation in the data until the variation left out is
approximately unsystematic. This part is the residuals (E). E is often used as a diagnostic tool in
identifying outliers in objects or variables.
A PCA model with two components will look like this:
X = t1p1’ + t2p2’ + E2
(7)
where t is a score vector or latent variable, and p is a loading vector. The pair, tp, is a principal
component (PC). The first principal component describes the largest variation, the second largest in
the second PC, and so forth. The remaining variation will be in the residual (E).
Before modeling, the data need pretreatment for several reasons. Mean-centering is substracting the
mean of all variables and is usually applied since the purpose of the analysis is to find differences
and similarities between objects, and at the same time determine which variable is important for the
decomposition (Martens and Martens, 2001).
Scaling or weighting of the variables is necessary when the data are measured using different scales
(e.g. mg/L, pH, absorbance), which would normally be the case when reporting different quality
parameters. When spectral data are analyzed, scaling is not required as the variables are in the same
scale. Additionally, variables that have low intensities are often mostly due to noise, and scaling
would just magnify this noise (Martens and Martens, 2001).
53
In postharvest science, PCA is an emerging method for data analysis. Examples are the use of PCA
for decomposition of complex quality data of fruits (Kienzle et al., 2011; Reichel et al., 2010) or
handling of spectral data from postharvest analysis (Peirs et al., 2005).
In Papers III and IV, PCA was used to find the main patterns in sensory and spectral data of
packaged wild rocket modeling on average values (Figure 25). The score plots from the two data
sets revealed the same trend in data (Figure 25A and Figure 25C), where fresh samples (Control)
were placed to the right, and samples with increased degradation due to senescence or fermentation
were placed further to the left. Samples packaged in high OTR film were placed at the top of the
plot, whereas samples packaged in low OTR film were placed at the bottom. The aspect that the
sensory and the multispectral score plots from PCA, an unsupervised method, were so similar is a
strong indication of a correlation between these two approaches of measuring the quality of leafy
green vegetables.
The loadings of the sensory PCA (Figure 25B) showed that samples placed to the right had a high
degree of freshness, green leaves, and intact leaf structure. Samples placed to the upper left
(senescence) were characterized by yellow leaves, cabbage odor, brown cut edges, a poor texture,
and color inhomogeneity. Samples placed to the lower left (fermentation) were characterized by
olive brown leaves, smoked odor, no leaf structure, high texture, and color homogeneity.
The loadings of the spectral PCA (Figure 25D) showed that samples placed to the right had high
signals in the NIR range, and that VIS bands from 550-700 nm were important for discriminating
between yellow senescent and green or olive brown fermented leaves. For the sensory data in
Papers III and IV, autoscaling was used. The spectral data in Paper IV were not scaled, and this
gave consistency in the scores plots. By not scaling the significance of the NIR bands was increased
and the VIS bands decreased, as the NIR bands had high signals in the spectra and VIS bands lower
signals (Figure 22, section 4.3.3).
Figure 25. Scores plot (A) and loadings plot (B) from a PCA of average sensory evaluation of wild rocket packages stored at different temperatures. Scores plot (C) and loadings plot (D) from a PCA of average spectra wild rocket taken by VideometerLab. From
Moreover in Paper IV, PCA was used to visualize the variation in spectra from different packages
Figure 26. The variation within a treatment was high, as evidenced by the data in
average values projected into the plot revealed that a model based on average values would not be
able to cover the variation experienced in single packages, as
larger variation area than averages across 12 packages
54
res plot (A) and loadings plot (B) from a PCA of average sensory evaluation of wild rocket packages stored at different temperatures. Scores plot (C) and loadings plot (D) from a PCA of average spectra
From Paper IV.
PCA was used to visualize the variation in spectra from different packages
. The variation within a treatment was high, as evidenced by the data in
average values projected into the plot revealed that a model based on average values would not be
able to cover the variation experienced in single packages, as the single packages
larger variation area than averages across 12 packages (Figure 26).
res plot (A) and loadings plot (B) from a PCA of average sensory evaluation of wild rocket packages stored at different temperatures. Scores plot (C) and loadings plot (D) from a PCA of average spectra from packaged
PCA was used to visualize the variation in spectra from different packages
. The variation within a treatment was high, as evidenced by the data in Figure 26. The
average values projected into the plot revealed that a model based on average values would not be
the single packages encompassing a
55
Figure 26. Scores plot from a PCA of multispectra from single packages of wild rocket. The averages of 12 equal packages were not included in the analysis but their position was projected into the scores plot (bold ’o’s’ connected with a line). All points are colored according to treatment, dark being low, and light being high accumulated temperatures (see section XX). From Paper IV.
In summary, PCA is an unsupervised method used mainly for exploration of data.
5.2 Multivariate regression: PLS
Partial Least Squares (PLS) (Wold et al., 1983; Wold et al., 2001) is a regression method commonly
used for correlation of one data matrix (X) with another data matrix (Y) or single variable (y). In
PLS, a bilinear model is built that enables prediction of Y (for instance, obtained from a traditional,
time-consuming measurement) from X (e.g. the fast method):
y/Y = XB +F
(8)
where B is a matrix that consists of the regression coefficients, and F is the matrix of the Y-
residuals. In order to obtain a good estimation of B, the PLS model needs to be calibrated on
samples that span the variation in Y perfectly and in general are representative of future samples.
Contrary to PCA, PLS is a supervised method. PLS is based on latent variables, and the
decomposition of X during regression is guided by the variation in Y – the co-variation between X
and Y is maximized. In that way, the variation in X that is related to Y is extracted. PLS serves two
purposes: one is to predict the Y of future objects; the other is to identify information in X that is
important for Y, and thereby understand the relation between X and Y.
56
The regression coefficients are the model. They are cumulative and contain information from all the
components included in the model.
PLS has become a standard tool in chemometrics and is often used when working with spectral
data. In postharvest science, the method is therefore used in combination with spectral analysis
(Bobelyn et al., 2010; Nicolai et al., 2007; O'Gorman et al., 2010; Sanchez et al., 2011). The
method was also used for predicting sensory data from spectral data in Paper IV. For the purpose of
understanding the relation between X and Y, the method was used for characterization of core
browning in pears (Larrigaudiere et al., 2004) and in predicting sensory attributes of chicory from
VIS-NIR spectra and physico-chemical measurements (Francois et al., 2008). PLS was also
successfully used as a tool to understand the perception of freshness of strawberries, carrots, and
apples from the underlying sensory attributes (Peneau et al., 2007a; Peneau et al., 2007b). A similar
approach was used in Paper III, where PLS was used to investigate the correlation between the
sensory attributes and freshness as explained in section 4.2.
In Paper II, PLS was used to predict the respiration rate from continuous measurements of O2 and
temperature together with a broccoli variety and harvest time of broccoli and wild rocket and to
further understand the importance of the variables (Figure 10, section 2.6).
PLS may also be used to improve interpretation of the results in studies with multiple or poorly
controlled variables. In Figure 5A, discrete measurements of respiration rate of broccoli are plotted
against O2 concentration. This plot indicates that respiration rate decreases at decreasing O2
concentration. However, information of time and temperature is missing, and hence makes the
results inconclusive. Plotting data from the continuous measurement of respiration rate (Figure 5B)
shows that temperature variations covaried with respiration rate. After inclusion of continuous
respiration data supported by temperature and time from several jars in the PLS model, it was
possible to estimate that the variation was primarily caused by poor temperature control and
advancement of time (Figure 10, section 2.6). The results prove that using chemometrics for data
analysis is a powerful tool.
57
Figure 27. Changes in the respiration rate of broccoli florets harvested in late summer and stored at a target temperature of 5 °C. Measurements are taken with discrete instrumentation (A) and continuous sensor (B), the latter colored according to temperature. Modified from Paper II.
In summary, PLS is a regression method used for correlation between two data matrices. PLS is a
supervised method, where the regression is guided by the information in Y.
5.3 Variable selection by iPLS
Variable selection is another tool in the multivariate toolbox. The models described previously can
often be improved by sorting out noisy variables and variables that do not explain the variation in
Y, and only build the model of variables that actually are related to Y.
Variable selection can either be performed before the modeling or during the modeling process; the
last procedure fits perfectly into the chemometric mindset, where the variables are chosen that give
good results with the actual modeling method. There are several methods for selection of variables;
however, only interval PLS was used during this project and is therefore the only method described
in the following.
Interval PLS (iPLS) (Leardi and Nørgaard, 2004; Nørgaard et al., 2000) was developed for
optimized analysis of continuous spectral data. Instead of using the full spectrum for models, local
models based on intervals in the full spectrum are used. Research indicates that selecting the
optimum intervals in spectral data can give better precision than a model based on the full spectrum
(Raju et al., 2012; Seefeldt et al., 2009). In iPLS, local models on sub-intervals are computed and
compared with the model based on all data concerning the RMSECV, and the model giving the
lowest RMSECV is suggested. Figure 28 shows the results of an analysis of overall freshness after
O2 (kPa)
0 5 10 15 20
Re
sp
ira
tio
n r
ate
(m
l O
2 k
g-1
h-1
)
0
10
20
30
40
50
A
0 5 10 15 200
10
20
30
40
50
O2 (kPa)
Resp
iratio
n r
ate
(m
l O2
kg-1
h-1
)
B
Temperature
8
9
10
11
12
58
opening, using the underlying sensory attributes (Paper III). The objective was to find the optimum
combination of sensory attributes important for the perception of freshness, and the method was
therefore used for another kind of data than it was intended for. Since sensory data are not
continuous, the calculation was performed with one attribute in each interval and with unlimited
numbers of intervals in both forward and reverse mode. In forward mode, iterations of PLS models
are performed including more variables until the RMSECV reaches a minimum. In reverse mode,
the intervals that by exclusion induce an increase in RMSECV are found; these are subsequently
used for modeling (Leardi and Nørgaard, 2004). Only one component was used to avoid overfitting,
since the data consisted of only eight samples. From the plot appears that only the model with mold
on the leaves performed worse than the global model. The best model is marked by green bars. It
consisted of acidic odor, wild rocket odor, rotten odor, and sprinklyness.
Figure 28. Cross-validated prediction performance (RMSECV) for ‘local’ iPLS models of each underlying attribute (bars) and for the PLS model, including the underlying attributes (red line) predicting overall freshness after opening in Paper III. Italic numbers on the bars indicate the optimum number of components used in each interval model. In this case, the maximum was set to 1 in order to prevent overfitting of the models, and the calculation was performed in forward mode.
Before selection of variables, outlier detection and pretreatment of data should be performed.
Whenever a model is optimized, the risk of overfitting is increased (Kjeldahl and Bro, 2010).
Therefore, it is also important, in addition to ordinary validation, to validate the result with chemical
and biological knowledge – i.e. to assess whether it makes sense that these variables are chosen.
Co
lorho
mo
geneity
Textu
reho
mo
geneity
Pro
ductheig
ht
Tra
yliq
uid
Acid
ico
do
r
Wild
rocke
t o
do
r
Cab
bag
eo
do
r
Ro
tten o
do
r
Sm
oke
do
do
r
Gre
en leaves
Yello
wle
aves
Oliv
eb
row
nle
aves
Bro
wn c
ut
ed
ges
Ro
tten le
aves
Mo
ldo
nle
aves
Wate
r-so
ake
dle
aves
Leaf
str
uctu
re
Sp
rinkl
yness
Britt
leness
0
0.5
1
1.5
2
2.5
3
3.5
4
(1 LVs)
(# LVs)
RM
SE
CV
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
59
In summary, PLS models might be improved by using variable selection, by exclusion of noisy
variables, and by using only variables highly correlated to Y. iPLS seeks for the combination of
intervals that decreases the prediction error.
5.4 Validation, model accuracy, and robustness
In order to ensure model robustness and accuracy, the calibration results must be validated. If too
many components are included, the data will be overfitted and new data will not fit into the model.
In leave-one-out cross validation, all objects but one are used to build a calibration model;
subsequently, the object that was left out is predicted using the model. This step is repeated for all
objects, and the errors from each prediction are used for calculation of the prediction error for each
PLS component. Leave-one-out cross validation will lead to an overoptimistic prediction error if
there are replicates and too many samples (rule of thumb says 20); in those cases segmented cross
validation is more appropriate, where a segment of samples is left out at the same time. A stronger
validation method is test set validation, where the data set is split up in a calibration set and a
validation set and preferably the test set should be independent, e.g. from another harvest or
measured on the instrument on a different day; however, this is not always possible.
The model accuracy is defined as the root mean squared error of prediction (RMSEP) or cross
validation (RMSECV):
!�"#$% or !�"#( )∑ � +,- � +-�./0123 �4 (9)
where np is the number of validated objects, and ŷi and yi the predicted and measured value of the
ith observation in the test set. This value is the average uncertainty that can be expected for future
predictions and can be compared to a standard deviation.
The optimal number of components in a model is chosen at the RMSECV minimum; however, the
risk of overfitting should be considered. If too many components are used, the model is overfitted
and will be sensitive to changes in new samples. Robustness is the ability to predict new samples
irrespective of unknown changes, and in most postharvest cases this means samples from difference
batches. In Nikolai et al. (2007), an example of a prediction model of soluble solids concentration
of apples is presented where the effect of using an independent test set is visualized. The model was
initially made with an internal test set, and 14 components were decided as the optimal model
complexity. When data from a new year (independent test set) was used for validation, seven
components were optimal, though the RMSECV was higher than before. The initial RMSECV was
too optimistic and the model was overfitting the first data set. The decision upon number of
60
components is therefore a balance between as low a RMSECV as possible but still without using
too many components. Therefore, proper validation is important.
The PLS model predicting sensory perception from multispectral images in Paper IV was validated
using segmented cross validation rather than a test set. Considering the large variations in
respiration rate found in wild rocket harvested at different times, it is likely that color differences
could also be expected between different harvests. Preliminary results (not shown) indicated,
however, that the data from the other investigated harvests were fairly covered by the models. The
PLS model in Paper IV is not an attempt to build a model of all possible degradations patterns of
packed wild rocket, but it proves that it is possible to predict sensory perception of color and texture
in one batch of packed wild rocket; thus the method has the ability to be expanded to include other
harvests as well.
In summary, validation of models is important to avoid overfitting and thereby ensure robustness of
the models. Furthermore, in validation, a prediction error is calculated, and it can be used to
decide the number of components to be included.
Overall, chemometrics are useful when dealing with large data matrices, and the methodology is
becoming increasingly popular in the postharvest field with the intuitive interpretation of
multifactorial data. Chemometrics are hypothesis generating and no assumptions of the underlying
phenomenon are needed prior to the analysis, although knowledge gained from chemometrics
should always be judged and compared with chemical and biological knowledge.
61
6 Conclusions and perspectives
The postharvest research strives to generate the knowledge needed to secure consumers and
producers optimal fresh produce. In this thesis, some new analytical methods to be used in
postharvest research have been presented and discussed with a focus on postharvest quality changes
occurring in leafy green vegetables.
Newly harvested leafy green vegetables are at the peak of their quality; when detached, they no
longer serve a purpose for the plant, though the processes built into the leaf and in every cell are
still functioning. Cut off from water supply, the leaf will wilt if water loss is not prevented.
Furthermore at harvest, senescence processes are initiated. Senescence of leafy green vegetables
leads to color changes and cell death, and thereby loss of texture. The rate of senescence is linked to
respiration rate, and the energy resources of the postharvest leafy vegetable are limited. Senescence
increases with increasing respiration rate and senescence can therefore be slowed down by keeping
respiration at a low level; thereby, good quality is maintained as long as possible. Biotic factors of
the harvested plant can have a large impact on respiration rate as was reported in Paper II, and it is
suggested that new breeding strategies should select for varieties with low respiration rate to ensure
longer shelf life.
When a product is harvested, different technologies exist to lower the respiration rate, and thereby
prolong shelf life and enhance quality of leafy vegetables through storage. Six main environmental
variables influence the storage conditions: storage duration, temperature, relative humidity, the
concentrations of O2, CO2 and ethylene. The most important factors are the combined effects of
temperature and storage time and to a minor extent the gas composition. Temperature should
therefore be kept low from field to fork for optimal quality, and an unbroken cooling chain is
pivotal for keeping respiration rate and deterioration low. The gas composition can to some extent
be controlled by packaging with modified atmosphere (MAP), and in passive MAP, the interplay
between the respiration rate and the permeability of the film determines the gas composition. The
success of a package is therefore dependent on many factors including temperature and biotic
variation, and the temperature measurements in Papers I and II revealed the difficulties of
controlling temperature during storage. In storage under poorly controlled temperature conditions, it
is recommended that packaging with high OTR is used to ensure aerobic respiration of high
respiring vegetables.
In postharvest experiments, all these factors are often dealt with by keeping as many factors as
possible constant and only vary a few in order to keep control. However, preferably, when planning
postharvest experiments, all variables should be varied at the same time in order to ensure that the
covariance between the factors affecting respiration rate and quality parameters is covered. This
implies larger experiments and additional samples. None of the factors can actually be viewed
solely as single factors; there are serious indications that they all influence each other in more or
62
less unpredictable ways. In order to describe the effect of different factors on respiration rate in a
quantitative way, mathematical modeling is often the solution, and when more variables are taken
into account by the use of chemometrics, unexpected patterns might be revealed.
In Paper I a new method for continuous respiration rate measurement concomitant with temperature
was developed and enabled fine-tuned determination of the temperature dependency on respiration.
The method was used in Paper II for investigation of seasonal and varietal differences in respiration
rate of broccoli and wild rocket using chemometric models. Interestingly, it was found that season
and variety almost had the same effect on respiration rate as temperature. Under these conditions it
can be difficult to design an optimal packaging that span both biotic variations, and the variation in
temperature experienced during transport and retailing. The permeability is either too high and
thereby ineffective or too low leading to anaerobic respiration. The optimum range of O2 for
lowering the respiration rate is very narrow for some products and may not exist for wild rocket and
is therefore practically impossible to attain with MAP. Furthermore, the PLS models presented in
Paper II predicting respiration rate revealed that the O2 concentration did not influence respiration
rate in the tested range. The PLS models in Paper II is just a glimpse into what the use of
continuous methods combined with chemometrics can provide to postharvest science. When more
factors in continuous measurements are taken into account, it becomes easier to conclude on the
experiments as the time factor is covered more than in discrete measurements. Chemometrics and
proper experimental designs will not solve all the problems in postharvest data interpretation, but it
is a way of handling many factors and large variability at the same time. Furthermore, rapid and
non-invasive analytical methods are increasingly used in quality control making it possible to
measure the quality of more samples. These methods often generate large data tables, and together
with more extensive experimental designs, this implies even greater amounts of data. Chemometrics
are therefore gaining a foothold into postharvest science with the intuitive interpretation of
multifactorial large data tables.
When the hurdle of large data tables is managed, rapid and continuous methods have several
advantages to discrete methods. Moreover, the use of wireless sensors makes it possible to measure
many samples at the same time. The key measure in postharvest research, respiration rate, was
measured in a new continuous way using wireless sensors. This led to a unique insight into the
close correlation between respiration rate and temperature. Furthermore, the sensors revealed that
the O2 concentration was insignificant for respiration rate above 6.3 kPa O2. The sensors can be
optimized in several ways leading to even more sophisticated simultaneous measures. Especially
measurement below 6.3 kPa O2 and measurement of CO2 would be beneficial as this would give
insight into the turning point from aerobic to anaerobic respiration – an aspect that was not obvious
from discrete measurements.
63
However, these rapid analytical methods have the drawback that often they cannot stand alone, as
sensory analysis and other slow or non-continuous methods are important for interpretation and
calibration.
The sensory consequences of packaging wild rocket in material with too high or too low
permeability were investigated in Paper III. The research revealed that storage, as a consequence of
temperature and time, inevitably induced loss of freshness of wild rocket; however, the packaging
material has a significant influence on the sensory attributes associated with the loss. Low OTR led
to fermentative metabolism due to lack of O2, proved by the sensory attributes ‘smoked odor’,
‘olive brown leaves’, and loss of texture. High OTR led to senescence indicated by the sensory
attributes ‘yellow leaves’, ‘brown cut edges’, inhomogeneous appearance, and loss of texture.
Judged by appearance, low OTR will seem better in retaining product quality; however, after
opening, the odor and texture of these products degraded the perception of freshness. The initial
freshness assessed by appearance of packages packed in high OTR was the same as the overall
freshness assessed after opening. Research revealed that the initial freshness at first sight relied on
green color, whereas odor and texture were important for the overall impression of freshness. A
measurement of color is simply not enough to describe freshness, and thereby quality. Wild rocket
was used as a representative for leafy green vegetables, and some of the results may differ from
other leafy green vegetables. However, the main conclusion of the importance of keeping
temperature low from field to fork and to ensure enough O2 to sustain aerobic respiration will
generally hold for all leafy green vegetables.
In Paper IV prediction of sensory attributes based on multispectral imaging were investigated. The
combination of information from both VIS and NIR wavelengths appeared to hold the information
of color and texture, respectively; thereby some of the changes occurring during storage could be
quantified. The research proved that it was possible to quantify sensory quality changes concerning
appearance and texture by rapid methods and chemometrics. Next step is to further investigate
already gathered multispectral data of packed wild rocket from other harvests and the effect of
season, temperature, and packaging on color.
Odor was not quantified by a rapid method, and here lies another task to be investigated in the
future. Proton transfer resonance-Mass spectroscopy (PTR-MS) which has been used with success
to quantify odor in e.g. coffee, is a good candidate. Implementation of rapid quality determination
like multispectral imaging for online control in processing lines is already ongoing at large factories
(personal experience at a fresh cut factory in California, 2010). With more automated quality
control, better understanding of seasonal variations, and the effect of packaging and continued
temperature control during handling and storage, it is possible to supply consumers with top quality
postharvest products in the future.
64
7 References
Abbott, J.A., 1999. Quality measurement of fruits and vegetables. Postharvest Biol.Technol. 15, 207-225.
Allende, A., Luo, Y.G., Mcevoy, J.L., Artes, F., and Wang, C.Y., 2004. Microbial and quality changes in minimally processed baby spinach leaves stored under super atmospheric oxygen and modified atmosphere conditions. Postharvest Biol.Technol. 33, 51-59.
Ares, G., Lareo, C., and Lema, P., 2008a. Sensory shelf life of butterhead lettuce leaves in active and passive modified atmosphere packages. Int.J.Food Sci.Tech. 43, 1671-1677.
Ares, G., Martinez, I., Lareo, C., and Lema, P., 2008b. Failure criteria based on consumers' rejection to determine the sensory shelf life of minimally processed lettuce. Postharvest Biol.Technol. 49, 255-259.
Ariana, D.P. and Lu, R.F., 2010. Hyperspectral waveband selection for internal defect detection of pickling cucumbers and whole pickles. Comput.Electron.Agr. 74, 137-144.
Barrett, D.M., Beaulieu, J.C., and Shewfelt, R., 2010. Color, flavor, texture, and nutritional quality of fresh-cut fruits and vegetables: desirable levels, instrumental and sensory measurement, and the effects of processing. Crit.Rev.Food Sci. 50, 369-389.
Bastrash, S., Makhlouf, J., Castaigne, F., and Willemot, C., 1993. Optimal controlled-atmosphere conditions for storage of broccoli florets. J.Food Sci. 58, 338-341.
Beaudry, R.M., 1999. Effect of O2 and CO2 partial pressure on selected phenomena affecting fruit and vegetable quality. Postharvest Biol.Technol. 15, 293-303.
Ben-Yehoshua, S. and Rodov, V., 2003. Transpiration and water stress, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 111-159.
Berns, R.S., 2000. Billmeyer and Saltzman principles of color technology, Third ed. John Wiley & Sons, New York.
Biasioli, F., Gasperi, F., Aprea, E., Endrizzi, I., Framondino, V., Marini, F., Mott, D., and Mark, T.D., 2006. Correlation of PTR-MS spectral fingerprints with sensory characterisation of flavour and odour profile of "Trentingrana" cheese. Food Quality and Preference 17, 63-75.
Biasioli, F., Yeretzian, C., Gasperi, F., and Mark, T.D., 2011. PTR-MS monitoring of VOCs and BVOCs in food science and technology. TRAC-Trend Anal.Chem. 30, 968-977.
Bobelyn, E., Serban, A.S., Nicu, M., Lammertyn, J., Nicolai, B.M., and Saeys, W., 2010. Postharvest quality of apple predicted by NIR-spectroscopy: Study of the effect of biological variability on spectra and model performance. Postharvest Biol.Technol. 55, 133-143.
Brecht, J.K., 1995. Physiology of lightly processed fruits and vegetables. Hortscience 30, 18-22.
Buning-Pfaue, H., 2003. Analysis of water in food by near infrared spectroscopy. Food Chem. 82, 107-115.
65
Cameron, A.C., Boylan-Pett, W.A.L.T., and Lee, J.U.L.I., 1989. Design of modified atmosphere packaging systems: modeling oxygen concentrations within sealed packages of tomato fruits. J.Food Sci. 54, 1413-1416.
Cantwell, M. and Kasmire, R.F., 2002a. Postharvest handling systems: flower, leafy, and stem vegetables, in: Kader, A.A. (Ed.), Postharvest technology of horticultural crops. Third ed. University of California, Agriculture and natural resources, Publication 3311, Oakland, CA, pp. 423-433.
Cantwell, M. and Kasmire, R.F., 2002b. Postharvest systems: fruit vegetables, in: Kader, A.A. (Ed.), Postharvest technology of horticultural crops. Third ed. University of California, Agriculture and natural resources, Publication 3311, Oakland, CA, pp. 407-421.
Cefola, M., Amodio, M.L., Cornacchia, R., Rinaldi, R., Vanadia, S., and Colelli, G., 2010. Effect of atmosphere composition on the quality of ready-to-use broccoli raab (Brassica rapa L.). J.Sci.Food Agr. 90, 789-797.
Cen, H. and He, Y., 2007. Theory and application of near infrared reflectance spectroscopy in determination of food quality. Trends Food Sci.Tech. 18, 72-83.
Clarkson, G.J.J., O'Byrne, E.E., Rothwell, S.D., and Taylor, G., 2003. Identifying traits to improve postharvest processability in baby leaf salad. Postharvest Biol.Technol. 30, 287-298.
de Gouw, J. and Warneke, C., 2007. Measurements of volatile organic compounds in the Earth's atmosphere using proton-transfer-reaction mass spectrometry. Mass Spectrom.Rev. 26, 223-257.
Degl'Innocenti, E., Pardossi, A., Tognoni, F., and Guidi, L., 2007. Physiological basis of sensitivity to enzymatic browning in 'lettuce', 'escarole' and 'rocket salad' when stored as fresh-cut products. Food Chem. 104, 209-215.
Delauran, L.C., Gislum, R., and Boelt, B., 2005. Placement of nitrogen in spinach (Spinacea oleracea L.): a method to increase seed yield? Acta Agric.Scand.B 55, 68-75.
Deza-Durand, K.M. and Petersen, M.A., 2011. The effect of cutting direction on aroma compounds and respiration rate of fresh-cut iceberg lettuce (Lactuca sativa L.). Postharvest Biol.Technol. 61, 83-90.
Dissing, B.S., Nielsen, M.E., Ersboll, B.K., and Frosch, S., 2011. Multispectral imaging for determination of astaxanthin concentration in salmonids. Plos One 6, e19032.
ElMasry, G., Wang, N., Vigneault, C., Qiao, J., and ElSayed, A., 2008. Early detection of apple bruises on different background colors using hyperspectral imaging. LWT-Food Sci.Technol. 41, 337-345.
Escalona, V.H., Verlinden, B.E., Geysen, S., and Nicolai, B.M., 2006. Changes in respiration of fresh-cut butterhead lettuce under controlled atmospheres using low and superatmospheric oxygen conditions with different carbon dioxide levels. Postharvest Biol.Technol. 39, 48-55.
66
Feilberg, A., Liu, D.Z., Adamsen, A.P.S., Hansen, M.J., and Jonassen, K.E.N., 2010. Odorant emissions from intensive pig production measured by online proton-transfer-reaction mass spectrometry. Environ.Sci.Technol. 44, 5894-5900.
Fonseca, S.C., Oliveira, F.A.R., and Brecht, J.K., 2002a. Modelling respiration rate of fresh fruits and vegetables for modified atmosphere packages: a review. J.Food Eng. 52, 99-119.
Fonseca, S.C., Oliveira, F.A.R., Frias, J.M., Brecht, J.K., and Chau, K.V., 2002b. Modelling respiration rate of shredded Galega kale for development of modified atmosphere packaging. J.Food Eng. 54, 299-307.
Francois, I.M., Wins, H., Buysens, S., Godts, C., Van Pee, E., Nicolai, B., and De Proft, M., 2008. Predicting sensory attributes of different chicory hybrids using physico-chemical measurements and visible/near infrared spectroscopy. Postharvest Biol.Technol. 49, 366-373.
Frøst, M.B. and Janhøj, T., 2007. Understanding creaminess. Int.Dairy J. 17, 1298-1311.
Gallardo-Escamilla, F.J., Kelly, A.L., and Delahunty, C.M., 2007. Mouthfeel and flavour of fermented whey with added hydrocolloids. Int.Dairy J. 17, 308-315.
Glenn, G.M., Chiou, B.-C., Imam, S.H., Wood, D.F., and Orts, W.J., 2005. Role of cuticles in produce quality and preservation, in: Lamikanra, O., Imam, S.H., and Ukuku, D.O. (Eds.), Produce degradation: pathways and prevention. Taylor & Francis, Boca Raton, pp. 19-53.
Gomes, M.H., Beaudry, R.M., Almeida, D.P.F., and Malcata, F.X., 2010. Modelling respiration of packaged fresh-cut 'Rocha' pear as affected by oxygen concentration and temperature. J.Food Eng. 96, 74-79.
Gowen, A.A., O'Donnell, C.P., Cullen, P.J., Downey, G., and Frias, J.M., 2007. Hyperspectral imaging - an emerging process analytical tool for food quality and safety control. Trends Food Sci.Tech. 18, 590-598.
Green, O., Bartzanas, T., Løkke, M.M., Bochtis, D.D., Sørensen, C.G., Jørgensen, O.J., and Tortajada, V.G., 2012. Spatial and temporal variation of temperature and oxygen concentration inside silage stacks. Biosystems Eng. 111, 155-165.
Green, O., Nadimi, E.S., Blanes-Vidal, V., Jorgensen, R.N., Storm, I.M.L.D., and Sorensen, C.G., 2009. Monitoring and modeling temperature variations inside silage stacks using novel wireless sensor networks. Comput.Electron.Agr. 69, 149-157.
Guiboileau, A., Sormani, R., Meyer, C., and Masclaux-Daubresse, C., 2010. Senescence and death of plant organs: Nutrient recycling and developmental regulation. C.R.Biol. 333, 382-391.
Halliwell, B. and Gutteridge, J.M.C., 2007. Free radicals in biology and medicine, Fourth ed. Oxford University Press, Oxford.
Hansen, M.J., 2011. Significance of reduced sulphur compounds in relation to odor from pig production. PhD thesis. Aarhus University, Foulum,
Hansen, M.J., Adamsen, A.P.S., Pedersen, P., and Feilberg, A., 2012. Prediction of odor from pig production based on chemical odorants. J.Environ.Qual. 41, 436-443.
67
Heenan, S.P., Dufour, J.P., Hamid, N., Harvey, W., and Delahunty, C.M., 2009. Characterisation of fresh bread flavour: Relationships between sensory characteristics and volatile composition. Food Chem. 116, 249-257.
Hertog, M.L.A.T., Boerrigter, H.A.M., van den Boogaard, G.J.P.M., Tijskens, L.M.M., and van Schaik, A.C.R., 1999. Predicting keeping quality of strawberries (cv. 'Elsanta') packed under modified atmospheres: an integrated model approach. Postharvest Biol.Technol. 15, 1-12.
Hertog, M.L.A.T., Lammertyn, J., Scheerlinck, N., and Nicolai, B.M., 2007. The impact of biological variation on postharvest behaviour: The case of dynamic temperature conditions. Postharvest Biol.Technol. 43, 183-192.
Hörtensteiner, S., 2006. Chlorophyll degradation during senescence. Annu.Rev.Plant Biol. 57, 55-77.
Hörtensteiner, S. and Kräutler, B., 2011. Chlorophyll breakdown in higher plants. BBA-Bioenergetics 1807, 977-988.
Hotelling, H., 1933. Analysis of a complex of statistical variables into principal components. J.Educ.Psycol. 24, 417-441.
Hutchings, J.B., 1999. Food color and appearance, Second ed. Aspen publishers, Gaithersburg.
Iqbal, T., Rodrigues, F.A.S., Mahajan, P.V., and Kery, J.P., 2009. Mathematical modeling of the influence of temperature and gas composition on the respiration rate of shredded carrots. J.Food Eng. 91, 325-332.
Iqbal, Z. and Bjorklund, R.B., 2011. Assessment of a mobile phone for use as a spectroscopic analytical tool for foods and beverages. Int.J.Food Sci.Tech. 46, 2428-2436.
Jacxsens, L., Devlieghere, F., De Rudder, T., and Debevere, J., 2000. Designing equilibrium modified atmosphere packages for fresh-cut vegetables subjected to changes in temperature. Lebensm.Wiss.Technol. 33, 178-187.
Jacxsens, L., Devlieghere, F., Ragaert, P., Vanneste, E., and Debevere, J., 2003. Relation between microbiological quality, metabolite production and sensory quality of equilibrium modified atmosphere packaged fresh-cut produce. Int.J.Food Microbiol. 83, 263-280.
Jirovetz, L., Smith, D., and Buchbauer, G., 2002. Aroma compound analysis of Eruca sativa (Brassicaceae) SPME headspace leaf samples using GC, GC-MS, and olfactometry. J.Agric.Food Chem. 50, 4643-4646.
Kader, A.A. and Saltveit, M.E., 2003a. Atmosphere modification, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 229-246.
Kader, A.A. and Saltveit, M.E., 2003b. Respiration and gas exchange, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 7-29.
68
Kaur, P., Rai, D.R., and Paul, S., 2011. Quality Changes in Fresh-Cut Spinach (Spinacia Oleracea) Under Modified Atmospheres with Perforations. J.Food Quality 34, 10-18.
Kienzle, S., Sruamsiri, P., Carle, R., Sirisakulwat, S., Spreer, W., and Neidhart, S., 2011. Harvest maturity specification for mango fruit (Mangifera indica L. 'Chok Anan') in regard to long supply chains. Postharvest Biol.Technol. 61, 41-55.
Kim, J.G., Luo, Y.G., and Gross, K.C., 2004. Effect of package film on the quality of fresh-cut salad savoy. Postharvest Biol.Technol. 32, 99-107.
Kim, J.G., Luo, Y.G., Saftner, R.A., and Gross, K.C., 2005a. Delayed modified atmosphere packaging of fresh-cut romaine lettuce: Effects on quality maintenance and shelf-life. J.Am.Soc.Hortic.Sci. 130, 116-123.
Kim, J.G., Luo, Y.G., Tao, T., Saftner, R.A., and Gross, K.C., 2005b. Effect of initial oxygen concentration and film oxygen transmission rate on the quality of fresh-cut romaine lettuce. J.Sci.Food Agr. 85, 1622-1630.
King, G.A. and O'Donoghue, E.M., 1995. Unravelling senescence: New opportunities for delaying the inevitable in harvested fruit and vegetables. Trends Food Sci.Tech. 6, 385-389.
Kjeldahl, K. and Bro, R., 2010. Some common misunderstandings in chemometrics. J.Chemometr. 24, 558-564.
Koukounaras, A., Siomos, A.S., and Sfakiotakis, E., 2007. Postharvest CO2 and ethylene production and quality of rocket (Eruca sativa Mill.) leaves as affected by leaf age and storage temperature. Postharvest Biol.Technol. 46, 167-173.
Koukounaras, A., Siomos, A.S., and Sfakiotakis, E., 2009. Impact of heat treatment on ethylene production and yellowing of modified atmosphere packaged rocket leaves. Postharvest Biol.Technol. 54, 172-176.
Koukounaras, A., Siomos, A.S., and Sfakiotakis, E., 2010. Effects of Degree of Cutting and Storage on Atmosphere Composition, Metabolic Activity and Quality of Rocket Leaves Under Modified Atmosphere Packaging. J.Food Quality 33, 303-316.
Lakakul, R., Beaudry, R.M., and Hernandez, R.J., 1999. Modeling respiration of apple slices in modified-atmosphere packages. J.Food Sci. 64, 105-110.
Lammertyn, J., Franck, C., Verlinden, B.E., and Nicolai, B., 2001. Comparative study of the O2, CO2 and temperature effect on respiration between 'Conference' pear cell protoplasts in suspension and intact pears. J.Exp.Bot. 52, 1769-1777.
Lammertyn, J., Peirs, A., De Baerdemaeker, J., and Nicolai, B., 2000. Light penetration properties of NIR radiation in fruit with respect to non-destructive quality assessment. Postharvest Biol.Technol. 18, 121-132.
Lappalainen, R., Kearney, J., and Gibney, M., 1998. A pan EU survey of consumer attitudes to food, nutrition and health: An overview. Food Qual.Prefer. 9, 467-478.
69
Larrigaudiere, C., Lentheric, I., Puy, J., and Pinto, E., 2004. Biochemical characterisation of core browning and brown heart disorders in pear by multivariate analysis. Postharvest Biol.Technol. 31, 29-39.
Lawless, H.T. and Heymann, H., 2010. Sensory evaluation of food: principles and practices, Second ed. Springer, New York.
Leardi, R. and Nørgaard, L., 2004. Sequential application of backward interval partial least squares and genetic of relevant spectral regions. J.Chemometr. 18, 486-497.
Lee, D.S., Haggar, P.E., Lee, J., and Yam, K.L., 1991. Model for Fresh Produce Respiration in Modified Atmospheres Based on Principles of Enzyme-Kinetics. J.Food Sci. 56, 1580-1585.
Lee, E.J., Matsumura, Y., Soga, K., Hoson, T., and Koizumi, N., 2007. Glycosyl hydrolases of cell wall are induced by sugar starvation in Arabidopsis. Plant Cell Physiol. 48, 405-413.
Lestremau, F., Andersson, F.A.T., and Desauziers, V., 2004. Investigation of artefact formation during analysis of volatile sulphur compounds using solid phase microextraction (SPME). Chromatographia 59, 607-613.
Lindinger, C., Labbe, D., Pollien, P., Rytz, A., Juillerat, M.A., Yeretzian, C., and Blank, I., 2008. When machine tastes coffee: Instrumental approach to predict the sensory profile of espresso coffee. Anal.Chem. 80, 1574-1581.
Lindinger, C., Yeretzian, C., and Blank, I., 2009. When machine tastes coffee: successful prediction of coffee sensory profiles by instrumental methods based on on-line PTR-MS. Chimia 63, 292.
Lleo, L., Barreiro, P., Ruiz-Altisent, M., and Herrero, A., 2009. Multispectral images of peach related to firmness and maturity at harvest. J.Food Eng. 93, 229-235.
Lonchamp, J., Barry-Ryan, C., and Devereux, M., 2009. Identification of volatile quality markers of ready-to-use lettuce and cabbage. Food Res.Int. 42, 1077-1086.
Ludford, P.M., 2003. Hormonal changes during postharvest, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 31-77.
Lunadei, L., Diezma, B., Lleo, L., Ruiz-Garcia, L., Cantalapiedra, S., and Ruiz-Altisent, M., 2012. Monitoring of fresh-cut spinach leaves through a multispectral vision system. Postharvest Biol.Technol. 63, 74-84.
Luo, Y.G., Mcevoy, J.L., Wachtel, M.R., Kim, J.G., and Huang, Y., 2004. Package atmosphere affects postharvest biology and quality of fresh-cut Cilantro leaves. Hortscience 39, 567-570.
Mahajan, P.V., Oliveira, F.A.R., Montanez, J.C., and Frias, J., 2007. Development of user-friendly software for design of modified atmosphere packaging for fresh and fresh-cut produce. Innov.Food Sci.Emerg. 8, 84-92.
Mangaraj, S., Goswami, T., and Mahajan, P., 2009. Applications of Plastic Films for Modified Atmosphere Packaging of Fruits and Vegetables: A Review. Food Eng.Rev. 1, 133-158.
70
Manolopoulou, H. and Papadopoulou, P., 1998. A study of respiratory and physico-chemical changes of four kiwi fruit cultivars during cool-storage. Food Chem. 63, 529-534.
Martens, H. and Martens, M., 2001. Multivariate analysis of quality: an introduction, First ed. John Wiley & Sons, Chichester.
Martinez, J.A., Chiesa, A., Tovar, F., and Artes, F., 2005. Respiration rate and ethylene production of fresh cut lettuce as affected by cutting grade. Agr.Food Sci. 14, 354-361.
Martinez-Romero, D., Bailen, G., Serrano, M., Guillen, F., Valverde, J.M., Zapata, P., Castillo, S., and Valero, D., 2007. Tools to maintain postharvest fruit and vegetable quality through the inhibition of ethylene action: A review. Crit.Rev.Food Sci. 47, 543-560.
Martinez-Sanchez, A., Allende, A., Cortes-Galera, Y., and Gil, M.I., 2008. Respiration rate response of four baby leaf Brassica species to cutting at harvest and fresh-cut washing. Postharvest Biol.Technol. 47, 382-388.
Martinez-Sanchez, A., Luna, M.C., Selma, M.V., Tudela, J.A., Abad, J., and Gil, M.I., 2012. Baby-leaf and multi-leaf of green and red lettuces are suitable raw materials for the fresh-cut industry. Postharvest Biol.Technol. 63, 1-10.
Martinez-Sanchez, A., Marin, A., Llorach, R., Ferreres, F., and Gil, M.I., 2006. Controlled atmosphere preserves quality and phytonutrients in wild rocket (Diplotaxis tenuifolia). Postharvest Biol.Technol. 40, 26-33.
Matile, P., Hortensteiner, S., and Thomas, H., 1999. Chlorophyll degradation. Annu.Rev.Plant Phys. 50, 67-95.
Matile, P., Hortensteiner, S., Thomas, H., and Krautler, B., 1996. Chlorophyll breakdown in senescent leaves. Plant Physiol. 112, 1403-1409.
McGuire, R.G., 1992. Reporting of Objective Color Measurements. Hortscience 27, 1254-1255.
Mizrach, A., 2008. Ultrasonic technology for quality evaluation of fresh fruit and vegetables in pre- and postharvest processes. Postharvest Biol.Technol. 48, 315-330.
Munck, L., Jespersen, B.M., Rinnan, A., Seefeldt, H.F., Engelsen, M.M., Nørgaard, L., and Engelsen, S.B., 2010. A physiochemical theory on the applicability of soft mathematical models-experimentally interpreted. J.Chemometr. 24, 481-495.
Nei, D., Uchino, T., Sakai, N., and Tanaka, S., 2006. Prediction of sugar consumption in shredded cabbage using a respiratory model. Postharvest Biol.Technol. 41, 56-61.
Nicolai, B.M., Beullens, K., Bobelyn, E., Peirs, A., Saeys, W., Theron, K.I., and Lammertyn, J., 2007. Nondestructive measurement of fruit and vegetable quality by means of NIR spectroscopy: A review. Postharvest Biol.Technol. 46, 99-118.
Nielsen, T., Bergstrom, B., and Borch, E., 2008. The origin of off-odours in packaged rucola (Eruca sativa). Food Chem. 110, 96-105.
71
Nørgaard, L., Saudland, A., Wagner, J., Nielsen, J.P., Munck, L., and Engelsen, S.B., 2000. Interval partial least-squares regression (iPLS): A comparative chemometric study with an example from near-infrared spectroscopy. Appl.Spectrosc. 54, 413-419.
O'Gorman, A., Downey, G., Gowen, A.A., Barry-Ryan, C., and Frias, J.M., 2010. Use of fourier transform infrared spectroscopy and chemometric data analysis to evaluate damage and age in mushrooms (Agaricus bisporus) grown in Ireland. J.Agric.Food Chem. 58, 7770-7776.
Olesen, M.H., Carstensen, J.M., and Boelt, B., 2011. Multispectral imaging as a potential tool for seed health testing of spinach (Spinacia oleracea L.). Seed Sci.Technol. 39, 140-150.
Page, T., Griffiths, G., and Buchanan-Wollaston, V., 2001. Molecular and biochemical characterization of postharvest senescence in broccoli. Plant Physiol. 125, 718-727.
Paull, R.E., 1999. Effect of temperature and relative humidity on fresh commodity quality. Postharvest Biol.Technol. 15, 263-277.
Paulsen, M.T., Ueland, O., Nilsen, A.N., Ostrom, A., and Hersleth, M., 2012. Sensory perception of salmon and culinary sauces - An interdisciplinary approach. Food Qual.Prefer. 23, 99-109.
Peirs, A., Schenk, A., and Nicolai, B.M., 2005. Effect of natural variability among apples on the accuracy of VIS-NIR calibration models for optimal harvest date predictions. Postharvest Biol.Technol. 35, 1-13.
Penchaiya, P., Bobelyn, E., Verlinden, B.E., Nicolai, B.M., and Saeys, W., 2009. Non-destructive measurement of firmness and soluble solids content in bell pepper using NIR spectroscopy. J.Food Eng. 94, 267-273.
Peneau, S., Brockhoff, P.B., Escher, F., and Nuessli, J., 2007a. A comprehensive approach to evaluate the freshness of strawberries and carrots. Postharvest Biol.Technol. 45, 20-29.
Peneau, S., Brockhoff, P.B., Hoehn, E., Escher, F., and Nuessli, J., 2007b. Relating consumer evaluation of apple freshness to sensory and physico-chemical measurements. J.Sens.Stud. 22, 313-335.
Peneau, S., Hoehn, E., Roth, H.R., Escher, F., and Nuessli, J., 2006. Importance and consumer perception of freshness of apples. Food Qual.Prefer. 17, 9-19.
Peneau, S., Linke, A., Escher, F., and Nuessli, J., 2009. Freshness of fruits and vegetables: consumer language and perception. Brit.Food J. 111, 243-256.
Peng, Y. and Lu, R., 2008. Analysis of spatially resolved hyperspectral scattering images for assessing apple fruit firmness and soluble solids content. Postharvest Biol.Technol. 48, 52-62.
Phillips, K., Niimi, J., Hamid, N., Silcock, P., Delahunty, C., Barker, M., Sewell, M., and Bremer, P., 2010. Sensory and volatile analysis of sea urchin roe from different geographical regions in New Zealand. LWT-Food Sci.Technol. 43, 202-213.
Piagentini, A.M., Guemes, D.R., and Pirovani, M.E., 2002. Sensory characteristics of fresh-cut spinach preserved by combined factors methodology. J.Food Sci. 67, 1544-1549.
72
Pretel, M.T., Souty, M., and Romojaro, F., 2000. Use of passive and active modified atmosphere packaging to prolong the postharvest life of three varieties of apricot (Prunus armeniaca, L.). Eur.Food Res.Technol. 211, 191-198.
Ragaert, P., Verbeke, W., Devlieghere, F., and Debevere, J., 2004. Consumer perception and choice of minimally processed vegetables and packaged fruits. Food Qual.Prefer. 15, 259-270.
Raju, C.S., Løkke, M.M., Suraryo, S., Ward, A.J., and Møller, H.B., 2012. NIR monitoring of ammonia in anaerobic digesters using a diffuse reflectance probe. Sensors 12, 2340-2350.
Raseetha, S., Heenan, S.P., Oey, I., Burritt, D.J., and Hamid, N., 2011. A new strategy to assess the quality of broccoli (Brassica oleracea L. italica) based on enzymatic changes and volatile mass ion profile using Proton Transfer Reaction Mass Spectrometry (PTR-MS). Innov.Food Sci.Emerg. 12, 197-205.
Reichel, M., Carle, R., Sruamsiri, P., and Neidhart, S., 2010. Influence of harvest maturity on quality and shelf-life of litchi fruit (Litchi chinensis Sonn.). Postharvest Biol.Technol. 57, 162-175.
Rinnan, A., van den Berg, F., and Engelsen, S.B., 2009. Review of the most common pre-processing techniques for near-infrared spectra. TRAC-Trend Anal.Chem. 28, 1201-1222.
Ruiz-Altisent, M., Ruiz-Garzia, L., Moreda, G.P., Lu, R., Hernandez-Sanchez, N., Correa, E.C., Diezma, B., Nicolaï, B., and García-Ramos, J., 2010. Sensors for product characterization and quality of speciality crops- a review. Comput.Electron.Agr. 74, 176-194.
Salisbury, F.B. and Ross, C.W., 1992. Plant physiology, Fourth ed. Wadsworth, Belmont.
Saltveit, M.E., 2003a. Fresh-cut vegetables, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and patology of vegetables. 2 ed. Marcel Dekker, New York, pp. 691-712.
Saltveit, M.E., 2003b. Is it possible to find an optimal controlled atmosphere? Postharvest Biol.Technol. 27, 3-13.
Sanchez, M.T., De la Haba, M.J., Guerrero, J.E., Garrido-Varo, A., and Perez-Marin, D., 2011. Testing of a local approach for the prediction of quality parameters in intact nectarines using a portable NIRS instrument. Postharvest Biol.Technol. 60, 130-135.
Sandhya, 2010. Modified atmosphere packaging of fresh produce: Current status and future needs. LWT-Food Sci.Technol. 43, 381-392.
Schouten, R.E., Zhang, X.B., Verschoor, J.A., Otma, E.C., Tijskens, L.M.M., and van Kooten, O., 2009. Development of colour of broccoli heads as affected by controlled atmosphere storage and temperature. Postharvest Biol.Technol. 51, 27-35.
Schulz, H., Drews, H.-H., Quilitzsch, R., and Krüger, H., 1998. Application of near infrared spectroscopy for the quantification of quality parameters in selected vegetables and essential oil plants. J.Near Infrared Spec. 6, A125-A130.
Seefeldt, H.F., Blennow, A., Jespersen, B.M., Wollenweber, B., and Engelsen, S.B., 2009. Accumulation of mixed linkage (1 -> 3) (1 -> 4)-beta-D-glucan during grain filling in barley: A vibrational spectroscopy study. J.Cereal Sci. 49, 24-31.
73
Smith, A.C., Waldron, K.W., Maness, N., and Perkins-Veazie, P., 2003. Vegetable texture: Measurement and structural implications, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 297-329.
Sousa, C. and Sodek, L., 2002. The metabolic response of plants to oxygen deficiency. Braz.J.Plant Physiol. 14, 83-94.
Taghizadeh, M., Gowen, A., Ward, P., and O'Donnell, C.P., 2010. Use of hyperspectral imaging for evaluation of the shelf-life of fresh white button mushrooms (Agaricus bisporus) stored in different packaging films. Innov.Food Sci.Emerg. 11, 423-431.
Tano, K., Oule, M.K., Doyon, G., Lencki, R.W., and Arul, J., 2007. Comparative evaluation of the effect of storage temperature fluctuation on modified atmosphere packages of selected fruit and vegetables. Postharvest Biol.Technol. 46, 212-221.
Thomas, H., Ougham, H.J., Wagstaff, C., and Stead, A.D., 2003. Defining senescence and death. J.Exp.Bot. 54, 1127-1132.
Thompson, J.F., 2002. Psychrometrics and perishable commodities, in: Kader, A.A. (Ed.), Postharvest technology of horticultural crops. Third ed. University of California, Agriculture and natural resources, Publication 3311, Oakland, CA, pp. 129-134.
Thygesen, L.G., Løkke, M.M., Micklander, E., and Engelsen, S.B., 2003. Vibrational microspectroscopy of food. Raman vs. FT-IR. Trends Food Sci.Tech. 14, 50-57.
Toivonen, P.M.A. and Brummell, D.A., 2008. Biochemical bases of appearance and texture changes in fresh-cut fruit and vegetables. Postharvest Biol.Technol. 48, 1-14.
Torrieri, E., Perone, N., Cavella, S., and Masi, P., 2010. Modelling the respiration rate of minimally processed broccoli (Brassica rapa var. sylvestris) for modified atmosphere package design. Int.J.Food Sci.Tech. 45, 2186-2193.
Tulio, A.Z., Yamanaka, H., Ueda, Y., Imahori, Y., and Chachin, K., 2003. Reversible and irreversible emission of methanethiol and dimethyl disulfide from anaerobically stored broccoli. J.Agric.Food Chem. 51, 6774-6781.
Uchino, T., Nei, D., Hu, W.Z., and Sorour, H., 2004. Development of a mathematical model for dependence of respiration rate of fresh produce on temperature and time. Postharvest Biol.Technol. 34, 285-293.
Wagstaff, C., Clarkson, G.J.J., Rothwell, S.D., Page, A., Taylor, G., and Dixon, M.S., 2007. Characterisation of cell death in bagged baby salad leaves. Postharvest Biol.Technol. 46, 150-159.
Wang, C.Y., 2003. Leafy, floral, and succulent vegetables, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 599-623.
Watada, A.E. and Qi, L., 1999. Quality of fresh-cut produce. Postharvest Biol.Technol. 15, 201-205.
74
Whitaker, B.D., 2003. Chemical and physical changes in membranes, in: Bartz, J.A. and Brecht, J.K. (Eds.), Postharvest physiology and pathology of vegetables. Second ed. Marcel Dekker, New York, pp. 79-110.
Whitaker, J.R., 1996. Enzymes, in: Fennema, O.R. (Ed.), Food chemistry. Third ed. Marcel Dekker, New York, pp. 431-530.
Wills, R.B.H., McGlasson, W.B., Graham, D., and Joyce, D.C., 2007. Postharvest: an introduction to the physiology and handling of fruit, vegetables and ornamentals, Fifth ed. Cabi, Oxfordshire.
Wold, S., 1995. Chemometrics; What do we mean with it, and what do we want from it? Chemometr.Intell.Lab. 30, 109-115.
Wold, S., Esbensen, K., and Geladi, P., 1987. Principal Component Analysis. Chemometr.Intell.Lab. 2, 37-52.
Wold, S., Martens, H., and Wold, H., 1983. The Multivariate Calibration-Problem in Chemistry Solved by the Pls Method. Lecture Notes in Mathematics 973, 286-293.
Wold, S., Sjostrom, M., and Eriksson, L., 2001. PLS-regression: a basic tool of chemometrics. Chemometr.Intell.Lab. 58, 109-130.
Wood, D.F., Imam, S.H., Sabellano, G.P., Eyerly, P.R., Orts, W.J., and Glenn, G.M., 2005. Microstructure of produce degradation, in: Lamikanra, O., Imam, S.H., and Ukuku, D.O. (Eds.), Produce degradation: pathways and prevention. Taylor & Francis, Boca Raton, USA, pp. 529-561.
Xue, L.H. and Yang, L.Z., 2009. Deriving leaf chlorophyll content of green-leafy vegetables from hyperspectral reflectance. ISPRS J.Photogramm. 64, 97-106.
Paper I
Novel wireless sensor system for monitoring oxygen, temperature and respiration rate of horticultural crops post harvest
Løkke, M.M., Seefeldt, H.F., Edwards, G. and Green, O.
Sensors, 11, 8456-8468 (2011) (open access)
Sensors 2011, 11, 8456-8468; doi:10.3390/s110908456
sensors ISSN 1424-8220
www.mdpi.com/journal/sensors
Article
Novel Wireless Sensor System for Monitoring Oxygen, Temperature and Respiration Rate of Horticultural Crops Post Harvest
Mette Marie Løkke 1, Helene Fast Seefeldt 1, Gareth Edwards 2 and Ole Green 2,*
1 Department of Food Science, Aarhus University, Aarslev DK-5792, Denmark; E-Mails: [email protected] (M.M.L.); [email protected] (H.F.S.)
2 Department of Engineering, Aarhus University, Tjele DK-8830, Denmark; E-Mail: [email protected] (G.E.)
* Author to whom correspondence should be addressed; E-Mail: [email protected]; Tel.: +45-4082-6150; Fax: +45-8999-3100.
Received: 24 July 2011; in revised form: 24 August 2011 / Accepted: 29 August 2011 / Published: 30 August 2011
Abstract: In order to design optimal packages, it is of pivotal importance to determine the rate at which harvested fresh fruits and vegetables consume oxygen. The respiration rate of oxygen (RRO2) is determined by measuring the consumed oxygen per hour per kg plant material, and the rate is highly influenced by temperature and gas composition. Traditionally, RRO2 has been determined at discrete time intervals. In this study, wireless sensor networks (WSNs) were used to determine RRO2 continuously in plant material (fresh cut broccoli florets) at 5 °C, 10 °C and 20 °C and at modified gas compositions (decreasing oxygen and increasing carbon dioxide levels). Furthermore, the WSN enabled concomitant determination of oxygen and temperature in the very close vicinity of the plant material. This information proved a very close relationship between changes in temperature and respiration rate. The applied WSNs were unable to determine oxygen levels lower than 5% and carbon dioxide was not determined. Despite these drawbacks in relation to respiration analysis, the WSNs offer a new possibility to do continuous measurement of RRO2 in post harvest research, thereby investigating the close relation between temperature and RRO2. The conclusions are that WSNs have the potential to be used as a monitor of RRO2 of plant material after harvest, during storage and packaging, thereby leading to optimized consumer products.
OPEN ACCESS
Sensors 2011, 11
8457
Keywords: wireless sensor networks; vegetable; post harvest research; respiration rate; temperature
1. Introduction
Postharvest research and technology refer to the handling, sorting, storage, transportation and sale of plant material from harvest to consumption. Special focus is on quality changes and loss reduction in the postharvest chain of fresh fruits and vegetables. Although harvested, the fruits and vegetables are still alive; they respire, consume oxygen and emit carbon dioxide. The respiration rate is the rate at which oxygen is consumed or carbon dioxide is generated. The rate is defined either as the consumed mL of oxygen (RRO2) per kg per hour or as the generated mL carbon dioxide (RRCO2) per kg per hour and is a key measure in post harvest research [1]. Respiratory parameters are correlated to the rate of deterioration of the plant material, and these parameters are especially important when designing modified atmosphere packaging (MAP), where the permeability of the packaging material must be designed to match the respiratory parameters [2-5]. Fresh fruit and vegetable respiration is affected by temperature, wounding, gas composition, and physiological factors such as pathogen attack and varietal differences [6-9]. Respiration is measured in either closed/static systems [1,5,10], where the product is kept in closed containers or in flow-through/open systems [10-12] where the gas composite is changed with specified gas concentrations [1] or left at atmospheric conditions. The changes in concentration of O2 or CO2 are measured over a period of time in the static system as well as in the open system. In the flushed system, the changes are measured between the inlet and the outlet. The initial gas composition can be either atmospheric [13] or composed in a specific manner [5,10,12].
In the literature, the most common method to determine oxygen and carbon dioxide in respiration analysis is gas chromatography [6,7,9,12-14] or electrochemical/infrared gas analyzers [5,10,11,15]. All the methods rely on the removal of a small amount of gas, and they are not continuous measurements, thus resulting in discrete data points. As the development of respiration rate is a continuous process and seldom linear, the discrete measurement points should be chosen with great care [1]. Oxygen partial pressure and temperature have an important impact on fruit and vegetable respiration [1]. Therefore, plant material is often tested under various temperature regimes in controlled climate chambers [5,7,10,12]. However, the exact temperature in the intimate proximity of the plant material has not been determined continuously during these experiments.
Respiration rate measurements would gain from continuously analyzing the gas composition without disturbing the system. Oxygen probes could be an option enabling non-invasive determination of O2 in closed systems. In the recent decades, development of Wireless Sensor Networks (WSN) has taken place. In contrast to wired sensors, the obstacle has been to develop hardware that is capable of transmitting data under difficult circumstances, and developing low-cost, long-term energy sources for the sensor nodes [16]. The wireless sensors enable monitoring of processes non-invasively and where cabling is not possible [17]. WSN are in intimate connection with the immediate physical environment allowing each sensor to provide detailed information on environment of material that is otherwise difficult to obtain by means of traditional, wired instrumentation [16]. WSN have been used for
Sensors 2011, 11
8458
different aspects of agricultural measuring, monitoring and control [18], such as precision irrigation, environmental field data collection systems, automated fertilizer applicators, and animal behavior monitoring [19-22].
Recently, WSN have been used within agricultural post harvest research, e.g., storage monitoring [23], or measuring and modeling of processed agricultural biomass quality in storage [24]. Within horticulture, WSN have mainly been used for monitoring environmental and growing conditions in the field or greenhouse. During transport and storage postharvest, WSN are widely used for temperature and psychrometric logging [17]. However, within horticultural post harvest there is still a lack of research and development of WSN [17] for quality monitoring. Only one study determining O2 continuously postharvest by using an O2 electrode to determine oxygen consumption in packaged tomatoes [25] has been found. Sensors with interesting features for post harvest research are gas/volatile sensors detecting oxygen, carbon dioxide, ethanol or volatile organic compounds, which are emitted from the plant material, reflecting the quality status of the produce. Other interesting features include relative humidity, shock and light impacts [16] and also biosensors used for microbial detection under development [17]. As a novel approach, WSN could be applied to measure respiratory parameters in post harvest technology, such as oxygen consumption. Furthermore, the respiration parameters can be related to the exact temperature experienced by the plant material. Using such WSN enables the continuous measurements of respiration in time-dependent experiments with fluctuating temperature. Finally, the wireless systems make it possible to do measurements without disturbing the system, and thereby preventing introduced changes in gas compositions.
The objectives of this study are: (1) to test novel wireless sensors capable of measuring the temperature and oxygen changes continuously inside 1 L glass jars containing vegetables (broccoli florets) under traditional respiration analysis conditions; (2) to test the communication reliability of the sensors from within climate chambers in changing temperature and oxygen regimes; and (3) to compare the measurements with a standard respiration measurement.
2. Experimental Section
2.1. Wireless Sensors
The wireless sensors used in this experiment monitor temperature and oxygen concentration. The sensor measures the temperature and oxygen level at specific time intervals and transmits the data wirelessly to a receiver station. The sampling time interval was set to roughly every 1 min. To obtain a long effective transmission communication range with high penetration capability, 433 MHz was selected as the communication frequency for this application.
The sensor is powered by a 3.6 V lithium battery. The oxygen sensor is an O2 A3 (Alphasense, Great Notley, Essex, UK) of the galvanic type. The temperature sensor is a TMP36 (Analog Devices, Norwood, MA, USA). The sensor unit consists of a microcontroller, radio, A/D converter, antenna circuit, power unit (battery), temperature sensor, and relative humidity sensor. The nRF9E5 is a single-chip system with fully integrated RF transceiver, 8051-compatible microcontroller and a four-input, 10-bit, 80 kilo samples per seconds (ksps) AD converter. The circuit has embedded voltage regulators, which provide maximum noise immunity and allow operation on a single 1.9–3.6 V supply.
S
T(pppkftris
aa
pEinfptrsserw
Sensors 201
The transceiCRC). The
programmabprogrammedport. The nRkeying (GFSfrequency sransceiver cs its ability
In this prand the dataactively part
The sensphysical (e.gEight ventilnside a 1 L
free movemproximity toransmit sig
suffered fromsensors somextra replicarespiration awill be repor
FigureThe pland armiddle
1, 11
iver of the e RF transble interfacd as interrupRF9E5 has SK) modulaynthesizer, can be turneto set low-l
roject, each a was receivticipated in sors were pg., pressurelation holesL glass jar, ment of gaso the plant
gnals and onm some int
metimes exerates should analysis. Onrted elsewh
e 1. The exlastic housiround the hoe is the oxyg
system autosceiver is ce (SPI). Tpts to the ma radio tranation at a da power a
ed on and olevel hardwsensor nod
ved by a gathandshakin
placed in 6 e forces) ans were madand plant mses. The lot material. ne showed truding exurt fall-outs dbe made in
nly data fromhere with a f
xperimentalng was placousing in thgen sensor.
omatically haccessed th
The data-remicrocontronsceiver fordata rate of amplifier, a off under so
ware functionde acted as ateway node.ng communi
cm plasticnd chemicalde just belomaterial waogged tempOf the 36 obvious fa
udates or coduring transorder to en
m an experifocus on the
setup withced in a gla
he glass jar.
handles prehrough an eady, carrieoller or pollr the 433 Mf 100 kbps.
modulatoroftware contnality to acha transmit-o. To enhancication.
c jars (Figul stresses (ew the lid i
as placed abperature is t
sensor nodalse valuesondensed wsmission, an
nsure enoughiment with se postharves
h the O2 senass jar of 1 Close-up o
eamble, addinternal p
er-detect, aned via a ge
MHz ISM baThe transc
r, and a rectrol. An imphieve low-ponly device ce communi
ure 1) to pre.g., acidificn the plasti
bove and pathe temperades tested, (O2 > 21%
water duringnd some jarh data whensix of the 36st use.
nsor protectL, and plan
of the senso
dress, and cparallel portnd address
eneral purpoands with Geiver consi
ceiver unit. portant asp
power sleep in a single-
ication relia
revent damcation) duriic jars. Theartially arouature insideeight senso
%). These seg handling ors were foun using this6 sensors ar
ted by a plant material
or (B). The w
cyclic redunt or an in
s-match sigose input-ouGaussian frests of a fulFor power
ect of the nstates.
-hop broadcability, each
mage to the ing the store plastic jarund the sene the glass or nodes prensors haveof the glass
und to be leas kind of senre reported
astic housinwas placedwhite circle
845
ndancy checnternal serignals can butput (GPIOequency shilly integrater saving, th
nRF9E5 nod
cast networkh sensor nod
sensor fromrage periodr was place
nsor allowinjar in clos
roved not te most likels jars. As thaking gassensor nodes ihere, the re
ng (A). d above e in the
59
ck al be O) ift ed he de
k, de
m ds. ed ng se to ly he es, in
est
Sensors 2011, 11
8460
2.2. Respiration Rate Measurements
During postharvest experiments with broccoli in the summer 2010, it was possible to test the usability of the sensors in a real plant produce experimental set-up designed to test the plant respiration rate at three different storage temperatures at either 5 °C, 10 °C or 20 °C in climate chambers (400 L, −9–99øC Binder KB400, Binder, Tuttlingen, Germany). A 1 L glass jar was filled with plant material of either 160 g broccoli florets. The glass jars were closed with a metal lid using a momentum key with a G-force of 5 kg. The lid of the glass jar was equipped with fittings for gas analysis. In all cases a closed system was used, meaning that once closed the glass jars were not opened until all oxygen was consumed by the plant material. The broccoli florets were of the commercial variety ‘Ironman’, and five replicates (glass jars) at each temperature were prepared. Two of the glass jars contained, besides the plant material, also the sensor in the plastic jar, and data from these six sensors (two jars at 5 °C, two at 10 °C and two at 20 °C) are reported here to test the usability of the sensors in a real set-up. Three glass jars were without the sensors but were used for discrete measurements of oxygen and CO2 using an electrochemical O2 analyzer and infrared CO2 analyzer (Checkmate 9900, PBI Dansensor, Ringsted, Denmark). The glass jars were removed from the climate chambers for conventional, discrete respiration measurements using the PBI Dansensor at nine time intervals for the glasses at 5 °C and 10 °C, and at six time intervals for the jars at 20 °C. The time intervals were chosen in order to cover closely the initial respiration period, but also to cover the longer storage period a 5 °C and 10 °C. At all the discrete measurement points, the two sensor-replicates were also removed from the climate chambers in order to handle the five replicates in the same way. Due to the handling of the glass jars, the temperature within the glass jar changed. Furthermore, it is expected that the oxygen levels in the glass jars were equally distributed within the glass jar due to the handling. The jars were measured at room temperature in the laboratory and were outside the climate chamber for maximum 15 min for each measurement before it was returned to the desired temperature and left until next analysis. Discrete measurements are reported in averages of the three replicates.
2.3. O2 Calibration of the PBI Dansensor
Oxygen measured by the PBI Dansensor was calibrated against gas mixtures of N2, 1% O2/2% CO2 in nitrogen, 10% O2/10% CO2 in nitrogen, 15% O2/25% CO2 in nitrogen, and compressed atmospheric air (all AGA Gas AB, Sundbyberg, Sweden). The calibration of the PBI Dansensor had a root mean square error of calibration (RMSE) of 0.12% oxygen.
2.4. O2 Calibration of the Sensors
The oxygen calibration was made by combining two different data sets: oxygen calibration1 and oxygen calibration2. To obtain controlled calibration points (oxygen calibration1); all 36 sensors were placed in one 3 L jar closed with a lid with flushing fittings. The jar was filled with a gas mixture of 10% O2/10% CO2 in nitrogen (AGA Gas AB, Sundbyberg, Sweden) and placed at 5 °C for 2 h followed by 10 °C for 18 h and 20 °C for 2 h. Then the jar was filled with a gas mixture of 15% O2/25% CO2 in nitrogen (AGA Gas AB, Sundbyberg, Sweden) and placed at 20 °C for 2 h followed by 10 °C for 2 h and 5 °C for 18 h. The sensors were also tested in a gas mixture
Sensors 2011, 11
8461
of 1% O2/2% CO2 in nitrogen (AGA Gas AB, Sundbyberg, Sweden), but these tests could not be used for calibration due to the sensors incapability of measuring below 3–4% O2. Oxygen calibration1 therefore consists of two oxygen levels at three different temperatures for each sensor, which alone would result in a two-point oxygen calibration with correction for temperature.
In order to improve oxygen calibration1 with successive points of oxygen levels, sensor data from other similar post harvest experiments than the one reported in Section 2.2 was used for Oxygen calibration2. Two sensors from each of the temperature tested (5 °C, 10 °C and 20 °C) were chosen, and only sensor data points coincident with reference measurement from the same jar were used leading to six data points for each sensor. The two data sets, oxygen calibration1 and oxygen calibration2, were then joint for calibration and resulting in a total 12 data points for each sensor. Each sensor node was oxygen calibrated individually and corrected for temperature in the calibration. Each of the six sensors was calibrated with an R2 of 0.97–0.98, and the mean correction for the oxygen signal was 0.76% O2. The RMSE of the oxygen calibration was in average 0.77% O2 for the six sensors, and this number can be compared with the RMSE of 0.12% from the calibration of the PBI Dansensor oxygen analyzer. The uncertainty of the PBI oxygen analyzer is also a part of the uncertainty of the sensors.
2.5. Temperature Calibration of the Sensors
The data for calibration of temperature was made concomitantly with oxygen calibration1. For the 3 L jar to be temperature equilibrated, it took approximately four hours. Therefore, for calibration of the temperature, only steady periods of more than four hours were used, leading to four time points—two time-points at 20 °C, one at 10 °C and one at 5 °C for calibration. Means of sensor signal over 1 h were used. The mean correction for temperature was −0.07 varying from −0.14 to 0.005. As the O2 signal was found to be influenced by temperature shifts, both the O2 signal and the calibrated temperature were used in the O2 calibration.
2.6. Data Analysis and Determination of Respiration Rate
All calibrations and data analysis were made in MATLAB (MathWorks, Natick, MA, USA) using in-house written scripts. The O2 and temperature signals were read into Matlab for each logger in means for every 5 min data-logging. All zeroes were replaced with missing number (NaN) and mis-functional loggers were expelled. Missing data in short intervals during the logging for respiration rate measurements were estimated using linear interpolation. Sensors showing longer intervals of missing data were expelled. The reference data from the gas calibration (oxygen calibration1) and the PBI-Dansensor (oxygen calibration2) were joined with sensor data and linear regression analysis was used for the oxygen calibration. Estimation of the respiration rate (the oxygen consumption per kg per hour) was determined by calculating the first derivative using a local second order polynomial Satvitzky-Golay smoothing filter with a window size of 17 points corresponding to 85 min. The respiration rate is calculated per weight material per hour.
Sensors 2011, 11
8462
3. Results and Discussion
3.1. Stability and Usability of Sensors in Closed Respiration Systems
During respiration analysis in closed systems, the material often starts to deteriorate because the material is still alive and respiring. As anaerobic conditions develop, leakage of moisture and exudates is inevitable. Therefore, the sensor nodes were protected from the material and exudates by placing them in plastic jars with a lid underneath the plant material (Figure 1). The sensors used in this experiment had dimensions that suited the experimental set-up very well, and the sensor set-up did not harm the plant material by pressure or other means. The closed jars were placed in climate chambers, but the extra coverage from the climate chambers did not generally disturb the signals from the sensor nodes to the receiver unit. The receiver was placed up to 3 m from the furthest climate chamber. During the measurements, some sensors had periodically fall-outs, where no data was received, probably due to lack of connection between the sensor and the receiver. The oxygen sensors were found to be well suited for postharvest experiments with regards to the physical set-up used.
3.2. Temperature and O2 Calibration of the Sensors
The sensor nodes transmitted an uncalibrated signal to the receiver. Before exploring the data, the nodes had to be individually calibrated for both temperature and oxygen. The calibrated temperature data are shown in Figure 2. The plot is dominated by sharp peaks, especially visible at 5 °C and 10 °C, and these peaks are ascribed to the opening of the climate chamber doors and the removal of the glass jars for reference measurements. The discrete measurements were performed in an uncooled laboratory at 22–24 °C and that influenced the temperature within the glass jars. Figure 2 also reveals that the equilibration of the temperature of the plant material before the start of the experiment at 20 °C was not sufficient. It is seen that the sensors enable a very fine-tuned and continuous determination of the temperature profile in the glass jar system. This is a major advantage in using the sensors, as the temperature plays a pivotal role in metabolic systems [1].
Figure 3 shows a plot of the calibrated oxygen signal for six of the sensors placed at 5 °C, 10 °C and 20 °C respectively. When the glass jars were filled with plant material, the lid was immediately closed, and only at the very closing time atmospheric gas composition was present in the jars. The glass jars were filled in another lab with no connection to the receiver from the sensor nodes, resulting in a time lag of approximately one hour from initiating the experiment until signals from the sensors could be received. The sensors placed at 5 °C had almost atmospheric levels of oxygen (20.3%) when the sensors started to log oxygen level, whereas at 10 °C and 20 °C the initial concentration of oxygen was 19.5% and 17.6%, respectively at initiation of logging. These differences in oxygen level after one hour were caused by the very high respiration rate of the plant material at elevated temperatures. This led to consumed oxygen at very high speeds, so even a short lag-period between closing of the lid and logging of the oxygen level would lead to a marked reduction in oxygen. In order to prevent these differences in oxygen levels, it is important to use an amount of plant material that leads to moderate changes in oxygen levels and does not respire too fast. However, the very fast consumption of oxygen found at elevated temperatures further underpins the efficiency and usability of WSN in respiration analysis as it will be possible to log the initial, very fast changes in oxygen level to determine the
Sensors 2011, 11
8463
initial respiration rate. The glass jars were stopped when the plant material was unsuitable for human consumption such as the broccoli seen in Figure 1.
Figure 2. The calibrated temperature signals from six sensors placed at 5 °C (blue lines), 10 °C (green lines) and 20 °C (red lines) at hours after initiating the experiment. The grey, vertical lines represent the time-points were reference measurements were performed.
Figure 3. The calibrated oxygen signals from six sensors placed at 5 °C (blue lines), 10 °C (green lines) and 20 °C (red lines) at hours after initiating the experiment, and the corresponding discrete reference measurements made by PBI Dansensor are shown as marks.
0 5 10 15 20 25 30 35 40 450
5
10
15
20
Hours after start
Cal
ibra
ted
tem
pera
ture
sig
nal (
°C)
5°C10°C20°C
0 5 10 15 20 25 30 35 40 45 500
5
10
15
20
Hours after start
Cal
ibra
ted
oxyg
en s
igna
l (%
)
5°C10°C20°C5°C(*)10°C(*)20°C(*)
Sensors 2011, 11
8464
At low oxygen levels, the sensors returned their individual calibrated minimum value. The sensors are, as they are designed now, generally unable to measure oxygen levels below 5%, although some of them could measure O2 levels of 2.3%. The mean minimum O2 value of the 25 sensors was 4.3% and the highest minimum was 6.9% O2. The inability to detect low oxygen levels is a major disadvantage for using the sensors in respiration analysis, as it is often of interest to define exactly when the oxygen drops below 1–2%. This level is, in many cases, the threshold between preservation of the fresh plant material, and anaerobic respiration leading to decomposition of the plant material [6,26]. The used O2-chip was the best commercially available chip with low energy requirements suitable for battery operation. It is possible that a chip capable of measuring low oxygen levels with low energy requirements will be available in the future.
3.3. Respiration Rates Determined by Sensors
It is of keen interest in postharvest research to follow even subtle changes in respiration rate in order to identify the underlying reasons. Figure 4 shows the estimated respiration rates (RRO2) calculated from the continuous measurements of oxygen by the sensors in a postharvest set-up. The sensor measurements are compared to the respiration rates calculated from the discrete measurements made by PBI Dansensor. The respiration rates obtained from the continuous measuring sensors showed fluctuations that are not detectable when using PBI Dansensor (Figure 4). Especially at 5 °C, but also at 10 °C, fluctuations in the sensor respiration rates were seen. At 20 °C, the oxygen level reached levels below 5% oxygen too fast, and therefore it was impossible to get a comparable picture of the respiration reaching a steady state as seen at 5 °C and 10 °C. The fluctuations seen in Figure 4 from the sensors hold information that would never be revealed when using conventional, discrete oxygen measurements.
Figure 4. The estimated respiration rates (O2) for broccoli placed at 5 °C (blue lines), 10 °C (green lines) and 20 °C (red lines) at hours after initiating the experiment, and the corresponding discrete reference measurements made by PBI Dansensor are shown as marks.
0 10 20 30 40 50 60 70 80 90 100
0
20
40
60
80
100
120
Hours after start
Res
pira
tion
(ml O
2/(k
g*h)
5°C10°C20°C5°C(*)10°C(*)20°C(*)
Sensors 2011, 11
8465
At 20 °C the discrete respiration measurements were much higher than those found for the sensors. The reason for this might be that the respiration rate is very high at elevated temperatures and irregularities between jars are enlarged. It is known that for high respiring plant material the closed system is difficult to use [1] no matter the method for measuring the oxygen content.
A drawback of the sensors is that they cannot measure carbon dioxide concomitantly with temperature and oxygen. Carbon dioxide is emitted from the plant material during respiration, and especially when respiration changes from aerobic to anaerobic conditions a large increase in carbon dioxide is found. The ratio RRCO2/RRO2 (the respiratory quotient, RQ) is used to determine the main metabolic substrates oxidized during respiration [6,12] and to determine the point at which anaerobic conditions appear. CO2 sensors are available, but commercial available chips with low energy requirements for battery operation have not been found.
3.4. The Temperature Dependency of Respiration Rates
The close relationship between temperature and respiration is shown in Figure 5 and has not previously been shown with this high time resolution. Information about temperature and respiration rate are important parameters in software designed to optimize modified atmosphere packaging of fresh fruits and vegetables [4]. Especially at 5 °C where the fluctuations are largest, the relationship is very predominant. It was expected that the respiration rate would increase as a consequence of increasing temperatures. However, Figure 5 shows that the increase in respiration rate was almost synchronic with the increase in temperature. Respiration rates in fresh plant material are very temperature dependent [1], and in this experiment the respiration rate was shown to increase 2–3 times for each 10 °C increase in temperature. Respiration rate curves, based on discrete measurements, determined at temperatures above 10 °C are therefore often too coarse-grained. A temperature increase of 2–3 °C resulted in an increase in respiration rate (RRO2) of 10 mL/kg/h. Whether the reason for the discrepancy in the timing is due to physical factors related to the sensors, or to the experimental setup is not clear. The oxygen RMSE of the sensors was much higher than the RMSE of the reference method. By smoothing the calculation of respiration rate, the effects of momentary small irregular changes in the sensor signal were reduced. Considering the match between sensor data and discrete data, despite the high biological variation in the different jars in Figure 3, and looking at the fine-tuned correlation between temperature and respiration rate in Figure 5, the fluctuations in respiration rate following a temperature increase was considered to be meaningful. The calibration of the sensors would be improved by more calibration points; a test with 30 points reduced the uncertainty to approximately 0.5. Aging of the sensors due to the harsh environment with intruding exudates and condensed water in postharvest experiments and biomass monitoring might also disturb the calibration. The effect of the harsh environment on aging is considered to be substantial compared to the aging due to high levels of CO2 and humidity. In these kinds of experiments, the lifetime of the sensors must therefore be considered to be lower than in experiments in more gentle environments. The sensors might also show cross sensitivity towards CO2 and other gasses that change during the experiment, and this might interfere with the calibration as well. The cross sensitivity should be of focus in future research.
Sensors 2011, 11
8466
Figure 5. The temperature and respiration rate relationship for each of the three temperatures 5 °C (A), 10 °C (B) and 20 °C (C). In each subfigure the red line refers to the respiration data, whereas the black line represents the temperature data from the same sensor. The respiration rates are only shown until the sensors reached the minimum oxygen level.
4. Conclusions
The described O2/temperature sensors show major perspectives within post harvest research and determination of respiration rates, and the physical set-up of the sensors worked very well. The sensors were able to work from within climate chambers and could transmit signals to the receiver. However, due to fall-outs of some sensors extra replicates must be considered. The sensors had a higher uncertainty in measuring the oxygen content than the reference method used, this could be improved by further calibration. The drawbacks of the sensors are their lack of low-level oxygen sensitivity and lack of CO2 information. There is a need for further development of chips that will make it possible to improve the sensors on these two points. The major advantage and reason for using the sensors is the ability to measure temperature and oxygen levels concomitantly and continuously. This enables fine-tuned determination of the temperature dependency of respiration leading to a better time-resolution of respiration. In a following paper, the sensors are used to determine respiration rates in broccoli and wild rocket harvested in different seasons and with varietal differences.
0 10 20 30 405
10
15
°C
0 10 20 30 400
20
40
Hours after start
RR
O2
0 5 10 15 20 255
10
15
°C
0 5 10 15 20 2520
40
60
Hours after startR
RO
2
0 2 4 6 8 1015
20
25
°C
0 2 4 6 8 1020
40
60
80
100
Hours after start
RR
O2
A
B
C
Sensors 2011, 11
8467
Acknowledgments
The authors would like to thank Technician Jens Michael Madsen, Aarhus University, Department of Food Science for the assistance of the experimental work. The Danish Ministry of Science, Technology and Innovation and the Danish Agency for Science, Technology and Innovation are greatly acknowledged for financial support to the Innovation consortium (Innovationskonsortium—Produkttilpasset pakning af frisk frugt og grønt, J.nr. 08-034100).
References
1. Fonseca, S.C.; Oliveira, F.A.R.; Brecht, J.K. Modelling respiration rate of fresh fruits and vegetables for modified atmosphere packages: A review. J. Food Eng. 2002, 52, 99-119.
2. Lakakul, R.; Beaudry, R.M.; Hernandez, R.J. Modeling respiration of apple slices in modified-atmosphere packages. J. Food Sci. 1999, 64, 105-110.
3. Jacxsens, L.; Devlieghere, F.; De Rudder, T.; Debevere, J. Designing equilibrium modified atmosphere packages for fresh-cut vegetables subjected to changes in temperature. Lebensm.-Wiss. Technol. 2000, 33, 178-187.
4. Mahajan, P.V.; Oliveira, F.A.R.; Montanez, J.C.; Frias, J. Development of user-friendly software for design of modified atmosphere packaging for fresh and fresh-cut produce. Innov. Food Sci. Emerg. 2007, 8, 84-92.
5. Iqbal, T.; Rodrigues, F.A.S.; Mahajan, P.V.; Kery, J.P. Mathematical modeling of the influence of temperature and gas composition on the respiration rate of shredded carrots. J. Food Eng. 2009, 91, 325-332.
6. Conesa, A.; Verlinden, B.E.; Artes-Hernandez, F.; Nicolai, B.; Artes, F. Respiration rates of fresh-cut bell peppers under supertamospheric and low oxygen with or without high carbon dioxide. Postharvest Biol. Technol. 2007, 45, 81-88.
7. Gomes, M.H.; Beaudry, R.M.; Almeida, D.P.F.; Malcata, F.X. Modelling respiration of packaged fresh-cut ‘Rocha’ pear as affected by oxygen concentration and temperature. J. Food Eng. 2010, 96, 74-79.
8. Hertog, M.L.A.T.; Boerrigter, H.A.M.; van den Boogaard, G.J.P.M.; Tijskens, L.M.M.; van Schaik, A.C.R. Predicting keeping quality of strawberries (cv. ‘Elsanta’) packed under modified atmospheres: An integrated model approach. Postharvest Biol. Technol. 1999, 15, 1-12.
9. Pretel, M.T.; Souty, M.; Romojaro, F. Use of passive and active modified atmosphere packaging to prolong the postharvest life of three varieties of apricot (Prunus armeniaca L.). Eur. Food Res. Technol. 2000, 211, 191-198.
10. Torrieri, E.; Perone, N.; Cavella, S.; Masi, P. Modelling the respiration rate of minimally processed broccoli (Brassica rapa var. sylvestris) for modified atmosphere package design. Int. J. Food Sci. Technol. 2010, 45, 2186-2193.
11. Kim, J.G.; Luo, Y.G.; Gross, K.C. Effect of package film on the quality of fresh-cut salad savoy. Postharvest Biol. Technol. 2004, 32, 99-107.
Sensors 2011, 11
8468
12. Escalona, V.H.; Verlinden, B.E.; Geysen, S.; Nicolai, B.M. Changes in respiration of fresh-cut butterhead lettuce under controlled atmospheres using low and superatmospheric oxygen conditions with different carbon dioxide levels. Postharvest Biol. Technol. 2006, 39, 48-55.
13. Manolopoulou, H.; Papadopoulou, P. A study of respiratory and physico-chemical changes of four kiwi fruit cultivars during cool-storage. Food Chem. 1998, 63, 529-534.
14. Ke, D.Y.; Vangorsel, H.; Kader, A.A. Physiological and quality responses of bartlett pears to reduced O2 and enhanced CO2 levels and storage-temperature. J. Am. Soc. Hortic. Sci. 1990, 115, 435-439.
15. Torrieri, E.; Cavella, S.; Masi, P. Modelling the respiration rate of fresh-cut Annurca apples to develop modified atmosphere packaging. Int. J. Food Sci. Technol. 2009, 44, 890-899.
16. Ruiz-Garcia, L.; Lunadei, L.; Barreiro, P.; Robla, J.I. A review of wireless sensor technologies and applications in agriculture and food industry: State of the art and current trends. Sensors 2009, 9, 4728-4750.
17. Ruiz-Altisent, M.; Ruiz-Garzia, L.; Moreda, G.P.; Lu, R.; Hernandez-Sanchez, N.; Correa, E.C.; Diezma, B.; Nicolaï, B.; García-Ramos, J. Sensors for product characterization and quality of speciality crops—A review. Comput. Electron. Agric. 2010, 74, 176-194.
18. Wang, N.; Zhang, N.Q.; Wang, M.H. Wireless sensors in agriculture and food industry—Recent development and future perspective. Comput. Electron. Agric. 2006, 50, 1-14.
19. Damas, M.; Prados, A.M.; Gomez, F.; Olivares, G. HidroBus (R) system: Fieldbus for integrated management of extensive areas of irrigated land. Microprocess. Microsyst. 2001, 25, 177-184.
20. Nadimi, E.S.; Sogaard, H.T.; Bak, T. ZigBee-based wireless sensor networks for classifying the behaviour of a herd of animals using classification trees. Biosyst. Eng. 2008, 100, 167-176.
21. Schumann, A.W.; Miller, W.M.; Zaman, Q.U.; Hostler, K.H.; Buchanon, S.; Cugati, S. Variable rate granular fertilization of citrus groves: Spreader performance with single-tree prescription zones. Appl. Eng. Agric. 2006, 22, 19-24.
22. Vivoni, E.R.; Camilli, R. Real-time streaming of environmental field data. Comput. Geosci. 2003, 29, 457-468.
23. Green, O.; Nadimi, E.S.; Blanes-Vidal, V.; Jorgensen, R.N.; Storm, I.M.L.D.; Sorensen, C.G. Monitoring and modeling temperature variations inside silage stacks using novel wireless sensor networks. Comput. Electron. Agric. 2009, 69, 149-157.
24. Bochtis, D.D.; Sørensen, C.G.; Green, O.; Bartzanas, T. A diagnostic system for improving biomass quality based on a sensor network. Sensors 2011, 11, 4990-5004.
25. Cameron, A.C.; Boylan-Pett, W.A.L.T.; Lee, J.U.L.I. Design of modified atmosphere packaging systems: Modeling oxygen concentrations within sealed packages of tomato fruits. J. Food Sci. 1989, 54, 1413-1416.
26. Oms-Oliu, G.; Soliva-Fortuny, R.; Martin-Belloso, O. Physiological and microbiological changes in fresh-cut pears stored in high oxygen active packages compared with low oxygen active and passive modified atmosphere packaging. Postharvest Biol. Technol. 2008, 48, 295-301.
© 2011 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).
Paper II
Effect of variety and harvest time on respiration rate of broccoli florets and wild rocket salad using a novel O2 sensor
Seefeldt, H.F., Løkke, M.M. and Edelenbos, M.
Used with permission from Postharvest Biology and Technology, 69, 7-14 (2012)
(This is a sample cover image for this issue. The actual cover is not yet available at this time.)
This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies areencouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Postharvest Biology and Technology 69 (2012) 7–14
Contents lists available at SciVerse ScienceDirect
Postharvest Biology and Technology
jou rna l h omepa g e: www.elsev ier .com/ locate /postharvbio
Effect of variety and harvest time on respiration rate of broccoli florets and wildrocket salad using a novel O2 sensor
Helene Fast Seefeldt, Mette Marie Løkke, Merete Edelenbos ∗
Aarhus University, Dept. of Food Science, Kirstinebjergvej 10, 5792 Årslev, Denmark
a r t i c l e i n f o
Article history:Received 7 November 2011Accepted 29 January 2012
Keywords:Respiration rateBiological variabilityBrassica oleraceaItalica GroupDiplotaxis tenuifoliaDry matter contentO2 sensor
a b s t r a c t
The impact of temperature and gas composition on respiration rates of postharvest produce is well under-stood, but only a few studies have documented variation in respiration rates of different varieties and atdifferent harvest times of the growing season. Most studies rely on discrete determinations of respira-tion rates and do not depict the dynamic nature of respiration. The aim of this study was to determinethe respiration rates in broccoli florets and wild rocket salad at different harvest times during the sea-son, and of different varieties. Storage temperature and respiration rates were determined using a novelwireless sensor for continuous and non-invasive measurements of O2 concentrations and temperaturein close proximity to the plant material. Respiration rates differed between broccoli varieties. Seasonaldifferences in respiration rate were found for broccoli and wild rocket salad. The response of respirationrate to storage temperature was distinct and differed between harvest times. These differences could berelated to differences in dry matter content. Biological differences in respiration rates prompt empiricaldeterminations when used in models for product-designed MAP.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Maintaining freshness and preventing loss are important aims inpostharvest handling of fresh fruit and vegetables (fresh produce)to ensure high quality products, minimize waste and increase foodavailability. Loss of quality is often a consequence of photosyntheticreserves being consumed by respiration. In general, deteriorationof fresh produce is correlated with the rate of respiration (Watadaand Qi, 1999); for example, fresh produce with high respirationrates tends to have a shorter shelf-life than produce with low rates.Therefore, selecting varieties with low respiration rates and low-ering the respiration rates after harvest are very useful tools toprolong the shelf-life of fresh produce.
Respiration rate is the rate at which O2 is consumed, and CO2is produced. Respiration rates for O2 (RRO2) and for CO2 (RRCO2)are defined as the rates of consumed O2 or produced CO2 per kgfresh produce per hour, and these rates are key measurements inpostharvest research (Fonseca et al., 2002). Respiration rates can becalculated from the differences in O2 and CO2 concentrations mea-sured inside containers with fresh produce kept with or withoutgas flushing (Fonseca et al., 2002; Iqbal et al., 2009; Torrieri et al.,2010). In both systems, the changes in O2 or CO2 concentrationsare measured over a period of time and used in the calculation of
∗ Corresponding author. Tel.: +45 8715 8334; fax: +45 87154812.E-mail address: [email protected] (M. Edelenbos).
RRO2 and RRCO2 (Escalona et al., 2006; Iqbal et al., 2009; Torrieriet al., 2010).
Gas chromatography and measurement with electrochemicaland infrared gas analyzers are the most common methods used fordetermination of O2 and CO2 concentrations for respiration analy-sis (Ke et al., 1990; Manolopoulou and Papadopoulou, 1998; Pretelet al., 2000; Escalona et al., 2006; Conesa et al., 2007; Gomes et al.,2010). These methods rely on removal of a small amount of gas fora ‘snapshot’ analysis of the gas concentration under specific stor-age conditions. From these measurements, RRO2 and RRCO2 can becalculated. Discrete data points are obtained, but these data shouldbe treated with care as many biological processes are dynamic andnot static (Fonseca et al., 2002; Saltveit, 2003b). Recently, Løkkeet al. (2011) reported that produce respiration can be calculatedfrom dynamic measurements of O2 concentrations using a novelO2 sensor.
Respiration rate in fresh produce is affected by abioticfactors such as storage temperature, time and gas composi-tion/concentration and biotic factors such as species, variety,growing conditions and harvest time (Fonseca et al., 2002). Res-piration rates of fresh produce can be controlled by refrigerationwith or without the application of controlled (CA) or modifiedatmosphere (MA) storage (Banaras et al., 2005; Kim et al., 2005;Escalona et al., 2006; Mahajan et al., 2007; Kalio, 2008; Iqbal et al.,2009; Lertsiriyothin, 2009). Storage temperature is the most criticalparameter that influences respiration rate, and it has been demon-strated that respiration rates increase at elevated temperatures at
0925-5214/$ – see front matter © 2012 Elsevier B.V. All rights reserved.doi:10.1016/j.postharvbio.2012.01.010
Author's personal copy
8 H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14
Fig. 1. Mean air temperature at Billund Airport (broccoli), Horsens (wild rocket salad) and Aarslev (respiration measurements) weather stations. The precipitation at eachgrowing site is shown by bars. Stars indicate the time of harvest. Weather data are from www. dmi.dk/dmi/vejretIdanmark–aaret2010.
all combinations of O2 and CO2 concentrations (Lammertyn et al.,2001; Fonseca et al., 2002; Uchino et al., 2004; Koukounaras et al.,2007; Iqbal et al., 2009). Handling of fresh produce and the lag-timebetween handling and gas analysis are also reflected in the respira-tion rate (Saltveit, 2003b; Martinez et al., 2005; Barrett et al., 2010).However, dynamic measurements of the impact of temperatureand time on respiration rate are lacking in the literature. Moreover,very little research focuses on the impact of biotic factors on res-piration rates of fresh produce despite the fact that low-respiringvarieties are expected to have longer shelf-lives than high-respiringvarieties.
In three apricot varieties, varietal differences up to 60% werefound in respiration rates (Pretel et al., 2000). In pears harvested atidentical maturity, no differences were found in respiration ratesbetween two consecutive years (Gomes et al., 2010). In four vari-eties of kiwifruit, varietal differences were found in respiration rate,and those with low rates had the longest shelf-life (Manolopoulouand Papadopoulou, 1998). In white and violet salad savoy, vari-etal differences in respiration rate were also found (Kim et al.,2004). Mizuna and watercress baby-leaf salad showed significantlyhigher respiration rates in the second compared to the first cutting(Martinez-Sanchez et al., 2008). For salad and wild rocket salad,only minor differences were observed in respiration rate betweencuttings (Martinez-Sanchez et al., 2008).
The aim of this work is to study the impact of variety and har-vest time on the dynamic response of fresh produce respirationrate in a variable storage environment. High-respiring broccoli flo-rets and wild rocket salad were used as model produce for theseexperiments. A novel O2 sensor was used to measure the dynamicnature of O2 depletion concomitantly with measurements of airtemperature inside closed glass jars filled with produce. With thisinstrumentation, it was possible to measure the actual tempera-ture in close proximity to the produce and to calculate the impactof a variable storage environment on respiration rate of broccoliflorets and wild rocket salad without gas sampling. For compar-ison, gas samples were daily taken from the jars for ‘snapshot’measurements using an O2 and CO2 gas analyzer.
2. Materials and methods
2.1. Plant material
Broccoli (Brassica oleracea, Italica Group) was grown and har-vested from a commercial farm in Jutland, Denmark (55◦56′ N, 9◦10′
E) in June and August 2010. For the early summer crop, the vari-eties ‘Green Magic’ and ‘Ironman’ were planted in week 13 andharvested in June in week 24. For the late summer crop, ‘Chronos’,
‘Ironman’ and ‘Monaco’ were planted in week 24 and harvested inAugust in week 33. The broccoli heads were, at all harvest times,harvested at commercial maturity, i.e. heads were firm with a headsize of approximately 400 g. Hence, the length of the growing sea-son could vary. The temperature and the precipitation during thegrowing season are shown in Fig. 1. A total of 20 heads of eachvariety was harvested and stored at 1 ◦C and >95% RH for 1–2 duntil the 2 h transport to the Research Centre in polystyrene boxeswith bagged, crushed ice for temperature control. At the Centre,the heads were stored, according to the experimental design of thestudy, for another 2–3 d at 2 ◦C and >95% RH.
Wild rocket salad (Diplotaxis tenuifolia L.) was grown andharvested in Jutland at a commercial grower (56◦00′N, 9◦46′E)during spring, and early and late summer 2010 (Fig. 1). Thespring harvest took place in May using overwintering wild rocketsalad, sown in week 36, 2009 and harvested in week 20, 2010.Salad for the early summer crop in June was sown in week15 and harvested in week 26, while salad for the late sum-mer crop in August was sown in week 28 and harvested inweek 32. After harvest, the wild rocket salad was stored in openboxes at 1 ◦C and >95% RH in a commercial warehouse untilpackaging 2–3 d later. Samples of 100 g wild rocket salad werepackaged on a commercial packaging line in polyethylene tereph-thalate (PET) trays (185 mm × 145 mm × 70 mm) and wrapped withlaser-perforated, -oriented polypropylene film (OPP) with an O2transmission rate (OTR) of 17.4 pmol s−1 m−2 kPa−1 at 23 ◦C and50% RH. The packages were then transported in a commercial cool-ing truck at 2 ◦C to the Research Centre where they were storedat 2 ◦C and >95% RH until the experiments started the followingday.
2.2. Sample preparation
All sample preparation with broccoli and wild rocket salad wasperformed at 2 ◦C and >95% RH in a cold room.
With broccoli, the heads were cut into 5 cm long florets. The flo-rets were mixed and then sorted into weight size classes of <10 g(small size), between 10–20 g (middle size) and >20 g (large size).For respiration measurements, three 1 L glass jars were filled withtwo large florets, six–seven middle size florets and then toppedwith small size florets until a product mean weight of approxi-mately 160 g was reached in each jar.
With wild rocket salad, a total of 30 g baby-leaves were carefullyremoved from one tray into a 1 L glass jar for respiration measure-ments. Three replicates of each sample were made for each storagetreatment.
Author's personal copy
H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14 9
2.3. Storage conditions
All jars were stored in climate chambers (400 L, −9–99øC BinderKB400, Binder, Tuttlingen, Germany) at 5◦, 10◦ or 20 ◦C, respec-tively, until the termination of the experiments, which took placewhen the O2 levels inside the jars reached approximately 6.3 kPa.
2.4. Respiration measurements
For respiration measurements, the static system described byFonseca et al. (2002) was used. A calibrated sensor for O2 and tem-perature measurements was placed in each jar below the produce(Green et al., 2009; Løkke et al., 2011). The sensor was protectedfrom the produce in a plastic cup fitted with 8 holes for air cir-culation. The plant material was placed on top of the cup, andthe glass jar was closed with an airtight lid and tightened witha force of 5 N. The sensors collected signals in close proximity tothe plant material and these signals were transmitted wirelesslyinto a sensor platform placed outside the climate chambers. Thesensors gave continuously measurements every 10 min of the O2concentration and air temperature inside the jars. For comparison,gas samples were daily taken from the jars for ‘snapshot’ measure-ments using an O2 and CO2 gas analyzer equipped with a zirconiaand a dual beam infrared sensor (Checkmate 9900, PBI Dansen-sor, Ringsted, Denmark). For these measurements, the jars wereremoved from the climate chambers to the laboratory desk andback into the chambers, usually within 15 min. The novel O2 sen-sor was calibrated to O2 in the 0–20.9 kPa range using standard gasmixtures from flasks (AGA, København S, Denmark). No readingscould be obtained from the novel O2 sensor below 6.3 kPa. Withinthe 6.3–20.9 kPa range the correlation coefficient (r) between theO2 concentration and the sensor readings was 0.98. Data in this O2range is exclusively included in this article. The root mean squareerror (RMSE) of the O2 calibration curve for the sensors was onaverage 0.77 kPa for the novel sensor compared to 0.12 kPa for thePBI Dansensor zirconia sensor. The O2 signals from the novel sensorwere slightly influenced by temperature. The O2 output signal fromthe sensor was therefore corrected for variation in temperature.
The RRO2 and RRCO2 rates were determined as differ-ences between two consecutive measurements and calculated asmmol kg−1 h−1 at 1 atm. A first derivative of a local, second orderpolynomial Satvitzky-Golay smoothing filter was used for the sen-sor data with a 210 min window for broccoli and 310 min for wildrocket salad. Different window sizes were used in order to achievethe same level of smoothing in the data analysis.
2.5. Analysis of dry matter content and specific leaf area
The dry matter content was determined at each harvest timeon three replicates of each species and variety (approximately 40 gof broccoli florets and 30 g of wild rocket salad). The material wasoven-dried (Lytzen, Herlev, Denmark) for 24 h at 80 ◦C. The spe-cific leaf area (SLA) of wild rocket salad was determined on 20randomly chosen leaves taken from six randomly chosen packagesand expressed as area per g fresh weight. From each leaf, a discof 10 mm was punched out, and a total of 20 discs were weighedtogether.
2.6. Data analysis
For statistical analyses of variances, the general linear models(GLM) procedure of the Statistical Analysis System (SAS Institute,Cary, NC) was used. The data were analyzed by one-way analy-sis of variance (ANOVA) including test for normal distribution andvariance homogeneity. The sources of variances were variety andharvest time for broccoli florets, and harvest time for wild rocket
salad. Turkey’s honest significance difference (HSD) test was usedto assess the significant differences between samples at P < 0.05. Apartial least square (PLS) regression analysis was made to predictRRO2 in each time point from the continuous variables: tempera-ture, time, O2 concentration, and the dummy variables: varietiesand harvest time (broccoli) and harvest time (wild rocket salad).Data from 26 jars of broccoli florets and 9 jars of wild rocket saladwere used for the PLS models. The models were made using thePLS toolbox (Eigenvector Research Inc., Wenatchee, WA, UnitedStates) in Matlab (MathWorks, Natick, MA, United States). The res-piration data were log-transformed to obtain normal distribution,which is a prerequisite for linear chemometric analysis. All vari-ables were auto-scaled for the regression analysis and segmentedcross validation was performed with each jar in each segment.
3. Results
3.1. Instrumentation, gas composition and respiration rate
The dynamic nature of O2 depletion inside a closed glass jarfilled with broccoli florets is shown in Fig. 2A as determined by thenovel sensor. The O2 concentration dropped from initially 20.9 kPato 6.3 kPa as time developed (Fig. 2A). From these data continuousRRO2 were calculated (Fig. 2B). At the start of the experiment, whenall jars were closed, the O2 concentration had already droppedto around 20 kPa (Fig. 2A). The O2 sensor reached its minimumof detection in the early and late summer broccoli florets after15–21 h at 5 ◦C (Fig. 2A) and after 2–7 h at 20 ◦C (not shown).The dynamic nature of O2 and CO2 concentrations inside a closedjar with broccoli florets is shown in Fig. 3A using the CheckMateinstrumentation. Again the O2 concentration decreased as storagetime developed as seen with the sensor (Fig. 2A). However, at thesame time the CO2 concentrations increased (Fig. 3A). This dynamicnature of gas exchange over time in closed jars with fresh produceis caused by produce respiration. When plant material respires, O2is consumed and CO2 produced. For every 1 kPa reduction in O2,CO2 increased with 1 kPa (data not shown). The correlation coef-ficient between RRO2 and RRCO2 for all broccoli experiments wasr = 0.995 within the 6.3–20.9 kPa range (Fig. 3B). The wild rockedsamples showed similar results (data not shown).
3.2. Concomitant determination of temperature and respirationrate
The temperature inside the climate chambers was set to 5 ◦C,10 ◦C and 20 ◦C, respectively. However, the air temperature insidethe glass jars differed substantially from these target temperatures(Fig. 4). This divergence was due to elevated temperatures in thelaboratory caused by high outside air temperatures (Fig. 1). Espe-cially, the jars placed in the chamber with a target temperatureof 5 ◦C showed large fluctuations in the actual air temperaturedepending on the outside air temperatures. The temperature insidethese jars fluctuated between 5 ◦C and 14 ◦C (Fig. 4). Generally, thejars targeted at 10 ◦C and 20 ◦C also had higher air temperaturesranging from 9 ◦C to 16 ◦C and from 15 ◦C to 23 ◦C, respectively(Fig. 4). The reasons for the low temperature readings in the 20 ◦Cjars were due to improper temperature equilibration before theonset of the experiments. The large differences found betweenactual air temperature and target values reflected the fact that thejars were removed daily for gas analysis to a laboratory withoutair-conditioning, and that the air temperature in the proximity tothe fresh produce fluctuated with the laboratory and outdoor tem-peratures (Fig. 1). July was especially warm, and this influenced theair temperature inside the jars.
Author's personal copy
10 H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14
Fig. 2. Changes in O2 concentration (A) and O2 respiration rate (B) inside a glass jar filled with broccoli florets (‘Ironman’) harvested in early and late summer and stored ata target temperature of 5 ◦C. Gas measurements with the novel O2 sensor.
The temperature inside the jars with broccoli florets very clearlyinfluenced the RRO2 of the produce as seen from the respirationpeaks arising concomitantly with elevated temperatures (Fig. 2B).The impact of temperature on RRO2 and RRCO2 in relation to gascomposition is shown in Fig. 5 with the CheckMate instrumenta-tion. These figures show that storage temperature has an effect onbroccoli respiration and that the relative effect of temperature onrespiration rates increases with decreasing temperature (Fig. 5).Overall, the RRO2 decreased 13% at 10 ◦C but 50% at 5 ◦C in the6.3–18.7 kPa O2 range (Fig. 5A). Within these same jars, there wasno reduction in the RRCO2 at 10 ◦C but a 40% reduction at 5 ◦C withinthe 0–11 kPa CO2 range (Fig. 5B). These data show that temperature,rather than O2 and CO2 gas concentrations in the measured ranges,influence the respiration rates of broccoli.
3.3. Parameters influencing the respiration rate of broccoli florets
In general, a positive relationship is seen between storagetemperature and broccoli respiration rate (Fig. 4). However, thisrelationship varies with variety and harvest time. In early summer,‘Ironman’ had an almost three times higher average RRO2 than inlate summer (Table 1) with a RRO2 of 3.56 at the early comparedto 1.55 mmol kg−1 h−1 at the late harvest. Interestingly, the respi-ration rate of ‘Ironman’ was around the same level at 20 ◦C in late
summer (RRO2 = 3.14 mmol kg−1 h−1) as in early summer at 10 ◦Cwith RRO2 = 3.56 mmol kg−1 h−1 (Table 1). As there was no signifi-cant difference (P > 0.05) in the measured air temperature betweenthese two harvest times, neither at 10 ◦C nor at 20 ◦C, the observeddifferences in RRO2 must be due to biological variability. At 10 ◦Cthere was a clear separation between the varieties regarding toRRO2 at the early, but not at the late harvest (Table 1). However, at20 ◦C, this separation was even more pronounced (Table 1). At bothtemperatures, ‘Ironman’ had the highest RRO2 followed by ‘GreenMagic’. Later in the season, ‘Chronos’ at 20 ◦C had the highest RRO2,followed by ‘Ironman’ and ‘Monaco’ (Table 1), while there wereno significant differences (P > 0.05) at 10 ◦C. It is noteworthy that‘Chronos’ had the highest and ‘Monaco’ the lowest dry matter con-tent, and that ‘Ironman’ had higher dry matter content at the earlythan the late harvest time. This result also corresponded with thehigher RRO2 of ‘Ironman’ found in early summer (Table 1).
A PLS model based on the data of storage time, temperature,O2 concentration, variety and harvest time was made to predictthe impact of these factors on RRO2 (Fig. 6). The model consistedof all the respiration data for broccoli florets except for the earlysummer samples kept at 20 ◦C as these had much higher, deviatingrespiration rates than the other samples. The PLS model explained91% of the variation in the data using two components, and it had aroot mean square error of cross validation (RMSECV) of 0.34 mmol
Fig. 3. Changes in O2 and CO2 concentrations (A) inside glass jars filled with broccoli florets (‘Ironman’) harvested in late summer and stored at a target temperature of 5 ◦C.(B) Correlation between the O2 and CO2 respiration rates of broccoli florets. Gas measurements with the CheckMate instrumentation.
Author's personal copy
H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14 11
Fig. 4. O2 respiration rates of broccoli florets determined at the actual storage temperature. The boxes indicate the minimum and maximum storage temperature andrespiration rates measured at a 5 ◦C, 10 ◦C and 20 ◦C target temperatures. Gas measurements with the novel O2 sensor.
Fig. 5. Changes in the O2 (A) and CO2 (B) respiration rates of broccoli florets (‘Ironman’) harvested in late summer and stored at target temperatures of 5 and 10 ◦C. Gasmeasurements with the CheckMate instrumentation.
Table 1Dry matter content and average O2 respiration rate of florets from different broccoli varieties and harvest times. Data are average of 3 determinations at each targettemperature. Samples within each column with different letters are significant different at P < 0.05.
Variety Harvest time Dry mattercontent %
Target temperature 10 ◦C Target temperature 20 ◦C
Measured temperature Respiration ratemmol O2kg−1h−1
Measured temperature Respiration ratemmol O2kg−1h−1
Green Magic Early summer 12.78a 12.8a 2.39b 19.9a 6.33b
Ironman 12.06b 13.2a 3.56a 20.9a 9.15a
Chronos Late summer 10.52c 13.5a 1.56c 21.9a 3.79c
Ironman 9.85cd 13.6a 1.55c 20.8a 3.14cd
Monaco 9.20d 12.9a 1.23c 21.2a 2.41d
Author's personal copy
12 H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14
Fig. 6. A partial least square (PLS) regression model of the O2 respiration rate of broccoli florets (A) with two PLS compontents explaining 91% of the variation in the dataand of wild rocket salad (B) with three PLS compontents explaining 87% of the variation in the data. The x-variable with the highest positive or negative regression vectorhas the highest impact on respiration rate i.e. storage temperature and time.
O2 kg−1 h−1. Storage temperature and time were the most influ-ential parameters, i.e. at prolonged storage time, a decrease in theRRO2 was found (Fig. 6A). Harvest time and variety also influencedRRO2, whereas an O2 concentration down to 6.3 kPa had no effectas the regression vector for O2 was ∼0 (Fig. 6A).
3.4. Parameters influencing the respiration rate of wild rocketsalad
The experiments with wild rocket salad showed that there wasa positive relationship between storage temperature and salad res-piration rate. In general, the highest RRO2 was experienced inthe spring-harvested salad (RRO2 = 2.73 and 6.95 mmol kg−1 h−1 at10 ◦C and 20 ◦C, respectively) taken from overwintering wild rocketplants, and the lowest was found in the salad harvested in late sum-mer with RRO2 = 1.08 and 3.92 mmol kg−1 h−1 at 10 ◦C and 20 ◦C,respectively (Table 2). At 20 ◦C there was no significant difference(P > 0.05) in the RRO2 between rocket salad harvested in early andin late summer (Table 2). Within the same harvest time, a two- tothree-fold increase in RRO2 was observed when the temperatureincreased from 10 ◦C to 20 ◦C (Table 2).
The biological variability in the RRO2 of wild rocket salad fromdifferent harvest times could not be ascribed to the specific leafarea (SLA), as the SLAs were similar for the spring and late summercrops (Table 2). However, as seen for the broccoli florets, wild rocketsalad with a high dry matter content had higher RRO2.
A PLS model for wild rocket salad showed that storage tem-perature is the most influential parameter for RRO2 (Fig. 6B). Atincreasing temperature, RRO2 increased while it decreased at pro-longed storage time. Harvest time also influenced RRO2. Especiallyspring-harvested wild rocket salad had higher RRO2 than salad har-vested in early and late summer. As with broccoli, the regressionvector for O2 was close to zero, meaning that an O2 concentrationdown to 6.3 kPa had no impact on the RRO2 (Fig. 6B). The PLS modelfor wild rocket salad was based on fewer samples than the broccolimodel. By using three PLS components, 87% of the variance couldbe explained with a RMSECV of 0.81 mmol O2 kg−1 h−1.
4. Discussion
4.1. Effect of storage temperature on respiration rate
Broccoli florets and wild rocket salad are high-respiring vegeta-bles that react with an increased RRO2 at elevated temperatures.The temperature effect on respiration rate is well established(Jacxsens et al., 2000; Lammertyn et al., 2001; Fonseca et al., 2002;Uchino et al., 2004; Koukounaras et al., 2007; Iqbal et al., 2009;Gomes et al., 2010). Moreover, produce respiration is more sen-sitive to temperature fluctuations than gas composition and filmpermeability (Lakakul et al., 1999; Jacxsens et al., 2000). The sen-sors used in our experiments enabled for the first time real-timetemperature and O2 measurements and thus the calculation of areal-time RRO2 depicting the dynamic nature of respiration. The
Author's personal copy
H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14 13
Table 2Dry matter content, specific leaf area and average O2 respiration rate of wild rocket salad at different harvest times. Data are average of 3 determinations at each targettemperature. Samples with different letters within each column are significant different at P < 0.05.
Harvest time Dry mattercontent %
Specific leaf areamm2/mg
Target temperature 10 ◦C Target temperature 20 ◦C
Measured temperature Respiration ratemmol O2kg−1h−1 Measuredtemperature
Respiration ratemmol O2kg−1h−1
Spring 9.35a 5.31a 11.1a 2.73a 21.1a 6.95a
Early summer 7.18b 3.92b nd nd 21.7a 3.99b
Late summer 7.47b 5.67a 11.7a 1.08b 20.5a 3.92b
nd: not determined.
results revealed that even minor fluctuations in air temperatureinside the jars influenced respiration rates, and that the produceresponded promptly to temperature with altered respiration rates.A distinct and fine-tuned response in broccoli respiration rate totemperature fluctuations has also been reported by Uchino et al.(2004). When considering the large biological variability observedand the way the heads were handled, the reported RRO2 for broc-coli florets in our experiments are within the ranges reportedby Uchino et al. (2004), who found RRO2 = 0.7 mmol kg−1 h−1forbroccoli heads stored at 10 ◦C, and RRO2 = 2.12 mmol kg−1 h−1 at20 ◦C. The handling of fresh produce, e.g. removal of the jars fromthe climate chambers when using traditional gas analysis, hadlarge and long-lasting effects on the inside air temperatures whichaffected the respiration rates. This effect should be consideredwhen designing respiration experiments with fresh produce in cli-mate chambers.
4.2. Effect of wounding on respiration rate
Broccoli heads were processed into florets, and both florets andwild rocket salad leaves were transferred into glass jars for analy-sis. It is well-known that handling and preparation of fresh producemay entail wounding which causes physiological responses in planttissue (Saltveit, 2003a). The decreasing respiration rate over timeseen in Fig. 2B was either due to a transitory effect of wound-ing and/or to depletion of carbohydrate reserves to replenish thesugars lost during storage (Bastrash et al., 1993). The tendency to anegative relationship between temperature and RRO2 in the earlybroccoli experiment at 20 ◦C reflects that this experiment was initi-ated during development of transitory respiration due to wounding(Saltveit, 2003a).
4.3. Effect of gas composition on respiration rate
Many studies have focused on the effect of gas composition onrespiration rate (Fonseca et al., 2002; Kim et al., 2005; Escalonaet al., 2006; Conesa et al., 2007; Iqbal et al., 2009; Gomes et al.,2010), and it is proven that retention of quality of broccoli and wildrocket salad can be extended by controlled atmosphere storage. Inthe present study, the effects of O2 concentrations on RRO2 weremodeled, and there was no effect of levels above 6.3 kPa. This resultis in agreement with Saltveit (2003a) who finds that O2 levels above6 kPa have a minor effect on respiration rates of head lettuce. Thesensors used in the present study did not give reliable measure-ments below 6.3 kPa O2, which explains the lack of an effect of lowO2 levels on RRO2 in the PLS models.
The results with broccoli florets showed that there was no sig-nificant effect (P > 0.05) of 6–11 kPa CO2 concentrations in the closeproximity of the plant material at a 7–12 kPa O2 concentration level(Fig. 5). Also, from the close relationship between RRO2 and RRCO2(Fig. 3B) there were no indications that there was a confoundingeffect of CO2 on the RRO2.
4.4. Biological variability in respiration rate
The present study involved plant material grown at differ-ent seasons. The climate during growth will naturally influenceplant development and assimilation of carbohydrates for respira-tion (Tijskens et al., 2003b; Moretti et al., 2010). The significantlyhigher (P < 0.05) RRO2 of produce grown in the spring and earlysummer can be explained by a higher photosynthetic turnover earlyin the season and rapid growth during periods with warm weather(Fig. 1).
According to Table 1, harvest time had the most pronouncedeffect on respiration rate, but varietal differences were also found.There are indications that the differences in RRO2 between thestudied broccoli varieties can be related to structure of heads andinflorescences size. Varieties with low RRO2 had small inflores-cence gathered in a compact head, whereas those with a high RRO2had large inflorescence in loose heads.
It is noteworthy that varieties with high dry matter contentsalso had high RRO2 within the same species. The dry matter contentis dependent on air temperature and period for growth as the drymatter content decreases with increasing air temperature (Tijskenset al., 2003b). The early summer broccoli was harvested 11 weeksafter transplanting compared to 9 weeks for the later crop. Fur-ther analysis is warranted if there is a general relationship betweendry matter content and RRO2 within specific commodities of freshproduce.
The biological variability between varieties and harvest timesincreased with storage temperature, i.e. more samples were sig-nificantly different (P < 0.05) from each other at 20 ◦C than at 10 ◦C.When selecting varieties for cold storage or MA packaging, it is ben-eficial to test the quality retention at elevated temperatures as lowtemperatures will minimize and stabilize the biological variability(Tijskens et al., 2003a).
5. Conclusions
The novel O2 and temperature sensors enabled measurementof the dynamic nature of respiration in fresh produce. It was foundthat even minor fluctuations in air temperature inside the jars influ-enced respiration rates, and that the produce responded promptlyto temperature change by altered respiration rates. Hence, whendesigning respiration experiments the effect of even minor changesin temperature should be considered. A large biological variabilitywas found between RRO2 and varieties as well as between RRO2and harvest times. This variability increased with storage tempera-ture. It is beneficial to test respiration rate and quality retention offresh produce at elevated temperatures to maximize the biologicalvariability.
Acknowledgments
The authors greatly acknowledge the support from Axel Maan-son A/S and Yding Groent A/S for broccoli heads and wild rocket
Author's personal copy
14 H.F. Seefeldt et al. / Postharvest Biology and Technology 69 (2012) 7–14
salad for the experiments. The authors thank Jens Michael Madsen,Aarhus University, Department of Food Science for the assis-tance with the experimental work. Ole Green, Aarhus University,Department of Engineering is greatly thanked for providing thewireless sensors. The Danish National Meteorological Institute isacknowledged for providing the temperature and precipitationdata. The Danish Ministry of Science, Technology and Innovationand the Danish Agency for Science, Technology and Innovation areacknowledged for the financial support (J.nr. 08-034100).
References
Banaras, M., Bosland, P.W., Lownds, N.K., 2005. Effects of harvest time and growthconditions on storage and post-storage quality of fresh peppers (Capsicumannuum L.). Pak. J. Bot. 37, 337–344.
Barrett, D.M., Beaulieu, J.C., Shewfelt, R., 2010. Color, flavor, texture, and nutritionalquality of fresh-cut fruits and vegetables: desirable levels, instrumental andsensory measurement, and the effects of processing. Crit. Rev. Food Sci. 50,369–389.
Bastrash, S., Makhlouf, J., Castaigne, F., Willemot, C., 1993. Optimal controlled-atmosphere conditions for storage of broccoli florets. J. Food Sci. 58,338–341.
Conesa, A., Verlinden, B.E., Artes-Hernandez, F., Nicolai, B., Artes, F., 2007. Respira-tion rates of fresh-cut bell peppers under supertamospheric and low O2 with orwithout high CO2. Postharvest Biol. Technol. 45, 81–88.
Escalona, V.H., Verlinden, B.E., Geysen, S., Nicolai, B.M., 2006. Changes in respira-tion of fresh-cut butterhead lettuce under controlled atmospheres using lowand superatmospheric O2 conditions with different CO2 levels. Postharvest Biol.Technol. 39, 48–55.
Fonseca, S.C., Oliveira, F.A.R., Brecht, J.K., 2002. Modelling respiration rate of freshfruits and vegetables for modified atmosphere packages: a review. J. Food Eng.52, 99–119.
Gomes, M.H., Beaudry, R.M., Almeida, D.P.F., Malcata, F.X., 2010. Modelling respi-ration of packaged fresh-cut ‘Rocha’ pear as affected by O2 concentration andtemperature. J. Food Eng. 96, 74–79.
Green, O., Nadimi, E.S., Blanes-Vidal, V., Jorgensen, R.N., Storm, I.M.L.D., Sorensen,C.G., 2009. Monitoring and modeling temperature variations inside silagestacks using novel wireless sensor networks. Comput. Electron. Agr. 69,149–157.
Iqbal, T., Rodrigues, F.A.S., Mahajan, P.V., Kery, J.P., 2009. Mathematical modelingof the influence of temperature and gas composition on the respiration rate ofshredded carrots. J. Food Eng. 91, 325–332.
Jacxsens, L., Devlieghere, F., De Rudder, T., Debevere, J., 2000. Designing equilibriummodified atmosphere packages for fresh-cut vegetables subjected to changes intemperature. Lebensm. Wiss. Technol. 33, 178–187.
Kalio, H.E., 2008. The effect of film wrapping and nitrogen flushing ofproduce atmosphere on the storage quality of lettuce. Agric. J. 3,17–24.
Ke, D.Y., Vangorsel, H., Kader, A.A., 1990. Physiological and quality responses ofBartlett pears to reduced O2 and enhanced CO2 levels and storage-temperature.J. Am. Soc. Hortic. Sci. 115, 435–439.
Kim, J.G., Luo, Y.G., Gross, K.C., 2004. Effect of package film on the quality of fresh-cutsalad savoy. Postharvest Biol. Technol. 32, 99–107.
Kim, J.G., Luo, Y.G., Tao, T., Saftner, R.A., Gross, K.C., 2005. Effect of initial O2 concen-tration and film O2 transmission rate on the quality of fresh-cut romaine lettuce.J. Sci. Food Agric. 85, 1622–1630.
Koukounaras, A., Siomos, A.S., Sfakiotakis, E., 2007. Postharvest CO2 and ethyleneproduction and quality of rocket (Eruca sativa Mill.) leaves as affected by leafage and storage temperature. Postharvest Biol. Technol. 46, 167–173.
Lakakul, R., Beaudry, R.M., Hernandez, R.J., 1999. Modeling respiration of apple slicesin modified-atmosphere packages. J. Food Sci. 64, 105–110.
Lammertyn, J., Franck, C., Verlinden, B.E., Nicolai, B., 2001. Comparative study of theO2, CO2 and temperature effect on respiration between ‘Conference’ pear cellprotoplasts in suspension and intact pears. J. Exp. Bot. 52, 1769–1777.
Lertsiriyothin, W., 2009. Sensitivity of produce respiration models used in theMAP-DESIGN software on the shelf life simulation of broccoli in the modifiedatmosphere package. Maejo Int. J. Sci. Technol. 1, 85–94.
Løkke, M.M., Seefeldt, H.F., Edwards, G., Green, O., 2011. Novel wireless sensor sys-tem for monitoring O2, temperature and respiration rate of horticultural cropspost harvest. Sensors 11, 8456–8468.
Mahajan, P.V., Oliveira, F.A.R., Montanez, J.C., Frias, J., 2007. Development of user-friendly software for design of modified atmosphere packaging for fresh andfresh-cut produce. Innov. Food Sci. Emerg. 8, 84–92.
Manolopoulou, H., Papadopoulou, P., 1998. A study of respiratory and physico-chemical changes of four kiwifruit cultivars during cool-storage. Food Chem.63, 529–534.
Martinez, J.A., Chiesa, A., Tovar, F., Artes, F., 2005. Respiration rate and ethyleneproduction of fresh cut lettuce as affected by cutting grade. Agr. Food Sci. 14,354–361.
Martinez-Sanchez, A., Allende, A., Cortes-Galera, Y., Gil, M.I., 2008. Respiration rateresponse of four baby leaf Brassica species to cutting at harvest and fresh-cutwashing. Postharvest Biol. Technol. 47, 382–388.
Moretti, C.L., Mattos, L.M., Calbo, A.G., Sargent, S.A., 2010. Climate changes and poten-tial impacts on postharvest quality of fruit and vegetable crops: a review. FoodRes. Int. 43, 1824–1832.
Pretel, M.T., Souty, M., Romojaro, F., 2000. Use of passive and active modified atmo-sphere packaging to prolong the postharvest life of three varieties of apricot(Prunus armeniaca, L.). Eur. Food Res. Technol. 211, 191–198.
Saltveit, M.E., 2003a. Fresh-cut vegetables. In: Bartz, J.A., Brecht, J.K. (Eds.), Posthar-vest physiology and pathology of vegetables. , 2nd ed. Marcel Dekker, New York,pp. 691–712.
Saltveit, M.E., 2003b. Is it possible to find an optimal controlled atmosphere?Postharvest Biol. Technol. 27, 3–13.
Tijskens, L.M.M., Konopacki, P., Simcic, M., 2003a. Biological variance, burden orbenefit? Postharvest Biol. Technol. 27, 15–25.
Tijskens, L.M.M., Veltman, R.H., Heuvelink, E., Simcic, M., 2003b. Modelling posthar-vest quality behaviour as affected by preharvest conditions. Acta Hortic. 599,469–477.
Torrieri, E., Perone, N., Cavella, S., Masi, P., 2010. Modelling the respiration rate ofminimally processed broccoli (Brassica rapa var. sylvestris) for modified atmo-sphere package design. Int. J. Food Sci. Technol. 45, 2186–2193.
Uchino, T., Nei, D., Hu, W.Z., Sorour, H., 2004. Development of a mathematical modelfor dependence of respiration rate of fresh produce on temperature and time.Postharvest Biol. Technol. 34, 285–293.
Watada, A.E., Qi, L., 1999. Quality of fresh-cut produce. Postharvest Biol. Technol. 15,201–205.
Paper III
Freshness and sensory quality of packaged wild rocket
Løkke, M.M., Seefeldt, H.F. and Edelenbos, M.
Submitted to Postharvest Biology and Technology
1
Research paper for Postharvest Biology and Technology 1
Freshness and sensory quality of packaged wild rocket 2
3
Mette Marie Løkke, Helene Fast Seefeldt, Merete Edelenbos* 4
Aarhus University, Dept. of Food Science, Kirstinebjergvej 10, 5792 Årslev, Denmark 5
*Corresponding author: Merete Edelenbos, Aarhus University, Dept. of Food Science, 6
Kirstinebjergvej 10, 5792 Årslev, Denmark 7
Tel: +45 87158334, Fax: +45 87154812 8
E-mail address: [email protected] 9
10
Abstract 11
Sensory descriptive analysis was used to evaluate the underlying attributes of freshness of 12
packaged wild rocket leaves (Diplotaxis tenuifolia L.). Variation was induced by differences in 13
package atmosphere, temperature and storage time. The leaves were packaged in trays wrapped 14
in film with a low oxygen transmission rate (OTR) (low OTR; 0.65 pmol s-1 m-2 kPa-1) or high OTR 15
(high OTR; 17.4 pmol s-1 m-2 kPa-1) and stored for 2, 10 or 20 °C for up to 7 days. At low OTR, the 16
O2 concentration decreased to 0.1 kPa depending on temperature and storage time and the leaves 17
developed a fermentative odor, lost tissue integrity and texture, and turned olive brown. At high 18
OTR, the O2 concentrations inside the package remained around 10-20 kPa depending on 19
temperature and storage time and wild rocket leaves showed signs of senescence, i.e. light-green 20
to yellow leaves and a heterogeneous appearance in the tray. Freshness of the leaves was 21
evaluated by sensory analysis without prior consensus. Initial freshness was evaluated before 22
opening and overall freshness after opening of the packages. Initial freshness was highly 23
correlated to green color (r2 = 0.98), whereas overall freshness was described by color, odor and 24
texture as found by interval PLS. Low OTR had higher scores of initial freshness before opening, 25
however, after opening high and low OTR had the same scores of overall freshness. Storage at 20 26
°C should be avoided to ensure overall freshness of packaged wild rocket. 27
2
28
Highlights 29
• Freshness of wild rocket leaves can be described by appearance, odor and texture 30
attributes. 31
• Evaluation of initial freshness relies on green color while overall freshness relies on color, 32
odor and texture. 33
• Storage temperature, time and package atmosphere O2 concentrations affect postharvest 34
wild rocket freshness 35
• Low OTR deteriorated packaged wild rocket leaves by fermentative metabolism, and high 36
OTR by senescence. 37
Keywords 38
Wild rocket leaves, sensory quality, freshness, temperature, modified atmosphere packaging, 39
iPLS, chemometrics 40
41
3
1. Introduction 42
Wild rocket, spinach and other baby leaves are popular leafy vegetables in Europe (RijkZwaan, 43
personal communication) and are often used raw in salads (Martinez-Sanchez et al., 2012). In 44
Denmark, wild rocket is sold unwashed in plastic trays wrapped in polypropylene (PP) film with 45
barrier properties. Packaging of leafy vegetables in trays or bags eases the handling of the 46
produce in the distribution chain, reduces loss of humidity, prevents spoilage, and prolongs the 47
shelf life, if the package atmosphere is modified to an optimum level (Kim et al. 2004). Modified 48
atmosphere packaging (MAP) in combination with low temperature storage is an effective tool to 49
prolong the shelf-life of leafy vegetables by decreasing oxygen (O2) and increasing carbon dioxide 50
(CO2) concentrations inside the package (Escalona et al., 2006; Kalio, 2008; Mahajan et al., 2007). 51
MAP delays senescence and slows down the breakdown of green chlorophyll pigments, and hence 52
the yellowing of the green tissues (Page et al., 2001; Toivonen and Brummell, 2008). However, this 53
beneficial effect of MAP can be lost by too low O2 and too high CO2 concentrations inside the 54
packages (Martinez-Sanchez et al., 2006b). Eventually off-odors develop, which may not be 55
apparent at purchase (Nielsen et al., 2008). 56
At the point of purchase, consumers evaluate packaging, labeling and product appearance (Barrett 57
et al., 2010). Visual quality is thus important for purchase while product freshness is important for 58
consumption (Lappalainen et al., 1998; Ragaert et al., 2004). Attractive, fresh-looking wild rocket 59
leaves are dark-green, uniform in color and brittle. In contrast, yellow and rotten leaves are 60
unwanted as they are senescent and too old for consumption (Koukounaras et al., 2007). High 61
quality, fresh-looking wild rocket leaves have no defects caused by diseases or insects, or from 62
postharvest handling (Barrett et al., 2010). 63
Freshness of fruit and vegetables is often ambiguously defined, however, it is a very important 64
parameter for the desire to purchase and consume fruit and vegetables (Peneau et al., 2006, 65
2009). The word “freshness” originates from the Latin word priscus, which means ”close to the 66
origin” (Peneau et al., 2009). Freshness is thus related to how much or little a product changes 67
4
quality after harvest and not to a single sensory attribute. Negative attributes, e.g. off-odors are 68
often more important for the perception of freshness than attributes adding positively to the 69
attribute (Peneau et al., 2007). Peneau et al. (2009; 2007; 2006) demonstrated that freshness of 70
fruits and vegetables is related to product appearance (color), texture (crispness) and physiological 71
age of the product at consumption. The researchers suggested using sensory descriptive analysis 72
to determine freshness and the underlying sensory attributes of freshness. Freshness is a mixed 73
term, and freshness is therefore not usually a suitable term to use in sensory analysis as sensory 74
terms should be precise and descriptive (Lawless and Heymann, 2010). Frost and Janhøj (2007) 75
and Paulsen et al. (2012) showed that mixed terms can be used in sensory analysis; however, it is 76
also important to evaluate the underlying attributes of the term. A mixed term is a meta-descriptor, 77
i.e. it is a property that is the result of a number of other properties (Frost and Janhoj, 2007). 78
Understanding the underlying attributes of product freshness is therefore important in postharvest 79
research. 80
At purchase, consumers use product appearance as an indication of freshness (Barrett et al., 81
2010). To judge overall freshness of a product based on appearance alone is insufficient, if no 82
other clues are available. Physiological ageing of packaged wild rocket can follow two pathways; a 83
senescence or a fermentation pathway. The senescence pathway applies when packaged wild 84
rocket is stored under high O2 concentrations. Then, leaves turn yellow due to enzymatic 85
breakdown of the green chlorophyll pigments uncovering the underlying yellow carotenoids. Brown 86
cut edges also appear during senescence due to enzymatic oxidation of phenolic compounds by 87
polyphenol oxidase in the presence of O2 (Toivonen and Brummell, 2008). Fermentation takes 88
place when packaged wild rocket leaves are kept under < 2 kPa O2 concentrations for a period of 89
time (Saltveit, 2003). At low O2 concentrations, organic acids will form and these cause pH to drop. 90
Eventually, the color changes from dark-green to olive-brown due to acidic degradation of 91
chlorophylls (Toivonen and Brummell, 2008). However, this discoloration is not the first sign of 92
fermentation, as off-odors develop at an earlier state (Luo et al., 2004). 93
5
Off-odors have been reported in green vegetables such as fresh-cut broccoli (Hansen et al., 2001; 94
Jacobsson et al., 2004), spinach leaves (Allende et al., 2004), butterhead lettuce (Ares et al., 95
2008a; Lonchamp et al., 2009), green cabbage (Lonchamp et al., 2009), rocket leaves (Nielsen et 96
al., 2008), cilantro leaves (Luo et al., 2004) and salad savoy (Kim et al., 2004). Formation of off-97
odors in packaged fruit and vegetables is an indicator of anaerobic respiration (Jacxsens et al., 98
2003; Kim et al., 2005; Nielsen et al., 2008) which eventually leads to fermentation (Jacxsens et 99
al., 2003; Nielsen et al., 2008). 100
Little information is available on the underlying sensory attributes of packaged wild rocket 101
freshness. The objective is to identify the underlying sensory attributes of product freshness of 102
packaged wild rocket subjected to variation in package-atmosphere O2 and CO2 concentrations, 103
different temperatures and storage times. 104
2. Materials and methods 105
2.1. Plant material 106
Wild rocket (Diplotaxis tenuifolia L.) for training and sensory analysis was harvested at a commercial 107
grower (56o00’N, 9o46’E) in Jutland, Denmark during September 2010. The wild rocket was 108
harvested directly into open boxes and boxes were transferred to a cold room (1 °C and > 95% RH) 109
immediately after harvest. The leaves were stored in the cold room until packaging 2-3 days later. 110
Samples of 100 g wild rocket were packaged on a commercial packaging line in polyethylene 111
terephthalate (PET) trays (185 mm x 145 mm x 70 mm) and wrapped with laser perforated, oriented 112
polypropylene film (OPP) with an O2 Transmission Rate (OTR) of 0.65 (low OTR) and 17.4 (high 113
OTR) pmol s-1 m-2 kPa-1, respectively, at 23 °C and 50% RH. The packages were then transported in 114
a commercial cooling truck at 2 °C to Research Centre Aarslev and stored at 2 °C and > 95% RH 115
until storage experiments were initiated the following day. 116
2.2. Storage experiments 117
6
Samples of wild rocket packaged in low and high OTR films were stored at different temperatures 118
(2, 10 and 20 °C) and times (2, 3, 4 and 6 days) to create packages with variation in package 119
atmosphere O2 and CO2 concentrations (Table 1). The control samples were packaged in high 120
OTR film and stored for six or seven days at 2 °C as sensory evaluation took place on these two 121
consecutive days. All samples were stored at 2 °C until transfer to elevated temperatures as 122
shown in Table 1. This design enabled sensory evaluation of packaged wild rocket from the same 123
batch but with differences in sensory quality. The biological variance in-between-samples of 124
packaged wild rocket for sensory analysis were then minimized. The variation in sensory quality 125
between samples was created by the use of packaging material with different OTR’s, storage 126
temperatures and times. Samples named e.g. HighOTR10-4 and LowOTR20-6 in Table 1 were 127
packaged in a film with a 17.4 pmol s-1 m-2 kPa-1 (high OTR) and a 0.65 pmol s-1 m-2 kPa-1 (low 128
OTR), respectively, and were stored before evaluation at 10 °C for 4 days or 20 °C for 6 days, 129
respectively. 130
Table 1. Packaged samples of wild rocket for sensory analysis. The symbol color changes according 131
to accumulated sum at elevated temperature and is used in figures. 132
Color
symbol
Samples
1
Film OTR2
pmols-1
m-2
kPa-1
Elevated
temperature
°C
Days at elevated
temperature
Accumulated
temperature sum at elevated
temperature3
°C
Gas
concentration at sensory analysis
4
O2
kPa
CO2
kPa
Control
17.4 No elevation 0 0 19.9a 1.7g
HighOTR10-4
10 4 40 18.4b 3.6f
HighOTR20-3 17.4 20 3 60 13.4c 8.4e
HighOTR20-4 20 4 80 12.5d 9.3d
HighOTR20-6 20 6 120 9.9e 11.8c
LowOTR10-4 10 4 40 0.5f 15.7a
LowOTR20-2 0.65 20 2 40 0.4f 15.9a
LowOTR20-6 20 6 120 0.1f 14.9b
1All samples were stored at 2 °C and >95% RH until they were transferred to elevated temperatures. No 133
samples were stored for more than seven days after the experiment begun. 134
2OTR: Oxygen Transmission Rate at 23 °C and 50% RH. An OTR of 17.4 is named High OTR and an OTR 135
of 0.65 is named Low OTR. 136
7
3The accumulated sum is calculated as the number of days at elevated temperature. 137
4Mean separations within each column by Turkey’s HSD test at P ≤ 0.05. 138
139
2.3. Gas composition 140
The package atmosphere O2 and CO2 concentrations were measured approximately two hours 141
before the sensory analysis was performed using a gas analyzer equipped with a zirconia and a 142
dual beam infrared sensor (Checkmate 9900, PBI Dansensor, Ringsted, Denmark). A needle was 143
placed directly into the package. After removal of the needle, the hole was covered by tape to 144
avoid leakage of odor from the package. 145
146
2.4. Sensory analysis 147
2.4.1. Sampling for sensory analysis 148
Sensory analysis was conducted on two consecutive days, day six and day seven, as four 149
replicates were needed, and it was impossible to run all treatments and replicates on one day 150
(Table 1). The only difference between the two days was that the samples differed 2 °C in 151
accumulated temperature sum, i.e. trays were stored at 2 °C for an extra day until they were 152
transferred to elevated temperatures (Table 1). The design was developed during initial studies of 153
packaged wild rocket leaves. These studies showed that there were no obvious differences in 154
visual, texture and odor quality of packaged wild rocket between these two evaluation days if 155
samples were stored at 2 °C (data not shown). For each session, 16 packages of each treatment 156
were prepared. Only 11 of these packages were used for sensory analysis as the panel consisted 157
of 11 panelists. The rest of the packages were discarded. Samples for evaluation were selected on 158
the basis of package atmosphere O2 and CO2 concentrations and visual appearance. Samples that 159
were similar within each treatment were selected. Samples were placed at room temperature for 160
approximately two hours before serving to ensure room temperature at evaluation. 161
162
2.4.2. Sensory descriptive analysis 163
8
Sensory descriptive analysis was performed as a tool to quantify the degree of freshness of wild 164
rocket leaves before and after opening of the packages and to identify the underlying sensory 165
attributes of freshness. The panelists were unaware of the purpose of the experiment and the 166
underlying design of it. Sensory profiling was performed according to international standards 167
(ASTM STP 913, 1986) in a sensory evaluation laboratory. The panel consisted of 11 panelists 168
employed exclusively to work as panelists in sensory analysis at the Research Centre. All panelists 169
were tested for their ability regarding basic taste, odor detection and color vision, as well as their 170
ability to communicate sensory attributes as recommended in ISO 8586-1_1993. Prior to sensory 171
evaluation, the panel went through two training sessions, each lasting two hours. In the first 172
session, panelists developed a vocabulary of 21 sensory attributes covering product appearance, 173
odor and texture. Two of these attributes were related to initial and overall freshness. Both 174
attributes were evaluated without prior consensus among panelists, i.e. each panelist used his or 175
her own understanding of the attribute (Frost and Janhoj, 2007; Paulsen et al., 2012). In the 176
second session, the panelists agreed on the selected attributes (Table 2) and were trained on a 177
subset of samples (Control, HighOTR20-6, and LowOTR20-6, Table 1) spanning wild rocket 178
freshness and sensory quality. 179
The sensory evaluation was carried out in compliance with international standards (ASTM STP 180
913, 1986; ISO 8589, 1988). The panelists evaluated eight samples of packaged wild rocket 181
treated as shown in Table 1 at each session. ‘Initial freshness’ was first evaluated. Then samples 182
were evaluated for visual appearance at first sight (Table 2). Then one cm of the corner of the film 183
was cut off on each package and odors were allowed to penetrate to the surrounding atmosphere. 184
Some odors were only present at the opening of the packages (‘acidic odor’ and ‘wild rocket odor’) 185
while other odors were more persistent (‘cabbage odor’, ‘rotten odor’ and ‘smoked odor’) 186
(Piagentini et al., 2002). All odors were evaluated at the opening of the packaging material. 187
Hereafter, the film was removed, and the panelists evaluated the visual appearance but this time 188
more closely. First, the leaf color and the color of the cutting surface were evaluated. Then 189
panelists evaluated the degree of rot, mold and water-soaked leaves by turning the leaves using a 190
9
fork or their fingers. Following this, the panelists evaluated the texture of the leaves using their 191
fingers. Finally, ‘overall freshness’ was evaluated (Table 2). Flavor and taste were not evaluated, 192
since packaged wild rocket will be rejected before eating if the overall product freshness does not 193
meet consumer expectations (Ares et al., 2008b; Martinez-Sanchez et al., 2006a). Standards were 194
provided for the evaluation of odor attributes. Standards were taken from well-known food items 195
such as fermented white cabbage (sauerkraut), fresh, senescent and fermented wild rocket and 196
broccoli. Each standard was filled into a white, non-transparent, flexible 200 ml plastic bottle closed 197
with a lid containing a 1 mm hole for release of odors. 198
All packages were coded by a three-digit number and served in a balanced randomized order at 199
each session to account for sample order and carry-over effects. The samples were evaluated at 200
individual speed on an unstructured 15-point line scale with intensity rating ranging from low (value 201
0) to high (value 15) using a direct computerized registration system (Fizz software, 2.30C, 202
Biosystemes, Couternon, France). 203
204
10
Table 2. Description of the attributes used for sensory evaluation of packaged wild rocket. 205
Method of
evaluation
Attribute
Description
Before opening
the packages
Initial freshness
Degree of initial freshness at first sight.
Color homogeneity Degree of overall uniformity of leaf color.
Texture homogeneity Degree of overall uniformity of the leaf texture.
Product height Relative height of product in the tray.
Tray liquid Degree of liquid at the bottom of the tray.
Right after
opening the
packages
Acidic odor
Intensity of an acidic and sour odor. Standard: sauerkraut.
Wild rocket odor
Intensity of wild rocket odor. Standard: freshly chopped wild rocket.
Cabbage odor
Intensity of a sulfurous cabbage odor. Standard: freshly chopped broccoli.
Rotten odor
Intensity of a rotten odor similar to one week old water in a vase with cut
flowers. Standard: rotten wild rocket leaves.
Smoked odor
Intensity of a chemical-like, smoked odor. Standard: fermented wild rocket
leaves.
After removal of packaging material
Green leaves
Intensity of the green color of the leaves.
Yellow leaves
Intensity of the yellow color of the leaves.
Olive brown leaves
Intensity of olive-brown color of the leaves.
Brown cut edges
Intensity of brown-cut edges of leaves.
Rotten leaves
Degree of rotten leaves.
Mold on leaves
Degree of leaves with mold.
Water-soaked leaves
Degree of leaves with water in the intracellular spaces noticed as extreme
dark-green areas on the leaves.
Leaf structure
Degree of leaves with an initial leaf structure.
Sprinklyness
The ability of leaves to sprinkle from each other.
Brittleness
The ability of leaves to snap into two pieces when bent.
Overall freshness
Degree of overall freshness after a careful sensory inspection using
appearance, odor and texture attributes.
206
207
11
2.5 Data analysis 208
2.5.1. ANOVA 209
For statistical analyses of variances, the general linear models (GLM) procedure of the Statistical 210
Analysis System (SAS Institute, Cary, NC) was used. The gas concentration data were analyzed 211
by one-way analysis of variance (ANOVA) to test the significant differences between treatments. 212
The descriptive sensory data were analyzed by mixed model ANOVA to test the significant 213
differences between treatments. In this model, the treatments were tested against the 214
treatment*panelist interaction (Naes and Langsrud, 1998). ‘Initial’ and ‘overall freshness’ within 215
each treatment were also analyzed by mixed model ANOVA using time (before and after opening) 216
as main effect and panelists as random. Turkey’s honest significance difference (HSD) test was 217
used for multiple comparisons of treatment means at a significance level of P ≤ 0.05. Pearson’s 218
correlation coefficients were calculated between attributes. 219
220
2.5.2. PCA 221
Principal Component Analysis (PCA) was applied on the sensory data to visualize the main 222
variations in the data. The calculation was performed in PLS_toolbox ver. 6.2.1 (Eigenvector 223
Research Inc., Wenatchee, WA) in Matlab ver. 7.12 (MathWorks, Natick, MA) on the average 224
values across panelist and replicates. The sensory attributes were autoscaled before the 225
calculation. 226
227
2.5.3. PLS 228
Partial Least Squares regressions (PLS) were applied with ‘initial freshness’ and ‘overall freshness’ 229
as Y-variables and the underlying sensory attributes as X-variables (Table 2). All attributes were 230
auto-scaled before modeling and leave-one-out validation was applied. For the calculation, 231
average scores across panelists (n=11) and replicates (n=4) were used. Interval partial least 232
squares (iPLS) modeling was applied on the sensory data as this procedure has provided better 233
precision in prediction models on spectral data (Nørgaard et al., 2000; Raju et al., 2012; Seefeldt et 234
12
al., 2009). The calculation was performed with one attribute in each interval and with unlimited 235
numbers of intervals in both forward and reverse mode. In forward mode, iterations of PLS models 236
were performed including more attributes until the RMSECV reached a minimum (Nørgaard et al., 237
2000). In reverse mode,.the intervals, which by exclusion induce an increase in RMSECV, were 238
found and these were subsequently used for modeling (Leardi and Nørgaard, 2004). All PLS 239
modeling was performed in PLS_toolbox ver. 6.2.1 (Eigenvector Research Inc., Wenatchee, WA) 240
in Matlab ver. 7.12 (MathWorks, Natick, MA). 241
242
3. Results and discussion 243
3.1. The experimental design 244
This study was designed to minimize the initial biological variation between packages of wild 245
rocket. Therefore, the wild rocket was harvested on at the same day from the same field and 246
handled and packaged as one batch. Recent results show that harvest time has an impact on wild 247
rocket quality (Seefeldt et al., 2012). Variation in product quality and freshness was thus created by 248
using films with different OTR values, storage temperatures and times (Table 1). It is well-known 249
that low temperature storage increases the shelf life of rocket leaves as the deterioration rate 250
decreases with temperature (Koukounaras et al., 2007). Initial studies showed no changes in visual 251
or odor quality of wild rocket stored for up to seven days at 2 °C (data not shown) which led to the 252
experimental design shown in Table 1. In this experiment, the control samples were considered 253
fresh despite the six or seven days of storage at 2 °C, i.e. the quality of the leaves at evaluation 254
was close to the original quality of wild rocket at harvest (Peneau et al., 2009). At evaluation, the 255
accumulated temperature sum differed between treatments representing different physiological 256
ages of the leaves (Hertog et al., 2007). By using an accumulated temperature sum, samples could 257
be compared. The different OTR’s of the films combined with the different temperatures and 258
storage times also created packages with variation in package atmosphere O2 and CO2 259
concentrations (Table 1). In the low OTR packages, the O2 concentration decreased from 21 kPa 260
13
to ≤0.5 kPa during the experimental period. In contrast, the O2 concentration remained high around 261
10-20 kPa depending on temperature and storage time in the high OTR samples. This result is in 262
line with Kim et al. (2004), who showed that OTR has an impact on package atmosphere O2 and 263
CO2 concentrations in packaged fresh-cut salad savoy. Also, the effect of temperature on the 264
inside package atmospheres is well documented (Jacxsens et al., 2000; Tano et al., 2007). 265
266
3.2. Freshness of packaged wild rocket 267
The sensory panel developed a list of 21 attributes (Table 2) to describe the sensory quality of 268
packaged wild rocket. ‘Initial’ and ‘overall freshness’ were evaluated without prior consensus at 269
‘first sight’ before opening the packages and again after looking, smelling and touching the product 270
at the end of the session. A high freshness score at first sight and after careful inspection indicated 271
that the actual quality at evaluation was close to initial quality at harvest (Peneau et al. 2009). 272
Freshness of the packaged wild rocket decreased as the temperature sum, and thus physiological 273
ageing increased (Fig. 1). Samples with high scores were stored six days at 2 °C or four days at 10 274
°C. These packages were rated significantly fresher (P≤ 0.05) than those stored at 20 °C for two to 275
six days (Table 3). The ratings for initial and overall freshness were not significantly different (P> 276
0.05) in the high OTR samples as was the case (P≤ 0.05) in the low OTR samples regardless of 277
storage temperature (Table 3). This result indicates that overall freshness of the high OTR 278
packages could be judged by appearance alone while appearance was insufficient for the 279
evaluation of the low OTR packages for freshness. With the low OTR packages, panelists needed 280
other attributes than those related to appearance at first sight to judge overall freshness. 281
14
282
Fig. 1. Correlation between ‘initial freshness’ and ‘overall freshness’ of packaged wild rocket leaves. 283
The accumulated temperature sum is higher with a light-colored than a dark-colored symbol. See 284
Table 1 for explanation. 285
286
3.3. The underlying sensory attributes of freshness 287
The underlying attributes of freshness were evaluated in different ways (Table 2). Appearance at 288
first sight described wild rocket color homogeneity, texture homogeneity, product height and liquid 289
in tray. These attributes were evaluated by looking at the product in the tray. The odor was 290
evaluated after making a 1 cm opening in the corner of the packaging material. Subsequently, odor 291
was described by attributes such as acidic, wild rocket, cabbage, rotten and smoked. Then the 292
packaging material was removed and the color and appearance were carefully examined by 293
turning the leaves. Green color, yellow color, olive brown color and brown cut edge color was then 294
evaluated followed by rotten leaves, mold on leaves and water-soaked leaves. The texture of the 295
leaves was examined by touching, i.e. leaf structure, sprinklyness and brittleness. The appearance 296
attributes at first sight gave an overall impression of the color and texture quality aspects of the 297
packaged wild rocket in the tray. If green color, yellow color, olive brown color, brown cut edges, 298
rotten leaves, mold on leaves, and water soaked leaves are evenly distributed in the tray, these 299
15
attributes might as well be evaluated at first sight. However, if this is not the case, turning the 300
product and touching it will give additional information on the quality of the product. As appearance 301
at first sight is the only parameter consumers can use at purchase, it is very important that the 302
appearance of packaged wild rocket reflects the actual freshness of the product. Otherwise the 303
packaged product will not meet the customer requirements relative to appearance (Barrett et al., 304
2010). Few studies report the effects of temperature, time and packaging OTR on packaged 305
product quality. For instance, Martinez-Sanchez et al. (2006a) evaluate visual quality and color of 306
packaged rocket leaves using an expert panel. In another study, Martinez-Sanchez et al. (2012) 307
evaluated visual quality, cut edge browning, leaf surface browning, off-odors, flavor and texture of 308
baby leaves for fresh-cut production. In another study with MAP on butterhead leaves, off-odor, 309
wilting, dark and necrotic stains on the leaf surface and browning of midribs were reported (Ares et 310
al., 2008a). The sensory characteristics of fresh-cut spinach were described by attributes, i.e. off-311
odor, overall appearance, wilting, browning and leaf color (Piagentini et al., 2002). 312
313
16
Table 3. Sensory scores from low (0) to high (15) for packaged wild rocket. The wild rocket was 314
packaged in high or low OTR and stored for up to six days at 2, 10 or 20 °C (see Table 1 for 315
description). The sensory attributes are grouped according to type of evaluation. 316
Evaluation Attribute1
High OTR Low OTR
Control 10-4 20-3 20-4 20-6 10-4 20-2 20-6
Freshness At first sight Initial freshness 11.8aA
1 9.5bA 2.7dA 1.8deA 0.2fA 11.5aA 4.7cA 0.6efA
Using all attributes
Overall freshness 12.2aA 7.9bA 1.8cA 1.2cdA 0.2dA 7.7bB 1.3cdB 0.2dB
Underlying attributes of freshness
Appearance at first sight or after examination
Color homogeneity 13.3a 10.7b 3.1c 2.5c 2.0c 13.3a 12.6a 11.1b
Texture homogeneity 12.4a 10.6b 4.2c 3.0cd 1.8d 12.7a 11.6ab 10.6b
Product height 12.9a 12.4a 10.2b 10.0b 3.0d 12.6a 6.3c 3.3d
Tray liquid 0.3d 0.4d 1.4cd 2.5bc 9.3a 0.9cd 4.2b 7.9a
Green leaves 12.9a 10.5b 5.0d 4.2d 1.2e 12.1ab 7.0c 0.6e
Yellow leaves 0.3c 1.3c 7.5b 8.3b 11.2a 0.4c 0.8c 0.3c
Olive brown leaves 1.1c 2.0c 1.5c 1.6c 1.4c 1.7c 6.2b 13.5a
Brown cut edges 3.1d 7.1c 10.1b 10.8b 12.7a 2.4d 2.4d 2.4d
Rotten leaves 0.5d 1.6cd 5.5b 5.9b 10.3a 0.8d 2.5c 5.9b
Mold on leaves 0.3b 0.3b 0.9b 1.2b 1.0b 0.2b 0.6b 5.3a
Water-soaked leaves 0.7d 1.1cd 2.4bc 2.9b 5.7a 1.7bcd 5.4a 5.6a
Texture at touching
Leaf structure 12.7a 11.1ab 8.4cd 8.3d 4.5e 10.0bc 2.0f 0.7f
Sprinklyness 12.4a 11.4ab 6.4c 5.7c 1.9d 9.8b 1.6d 0.4d
Brittleness 11.0a 10.0a 6.9b 6.3b 2.7c 9.7a 0.6d 0.4d
Odor when smelled Acidic odor 1.0c 3.6b 4.6ab 5.6ab 6.1a 4.8ab 5.6ab 5.4ab
Wild rocket odor 4.7a 3.3a 1.7b 1.4b 1.3b 3.5a 1.6b 0.6b
Cabbage odor 3.4b 6.7a 6.9a 7.1a 7.2a 3.9b 3.3b 3.1b
Rotten odor 1.1e 4.8d 8.0bc 7.7bc 10.8a 4.3d 5.6cd 8.2b
Smoked odor 0.7c 2.7bc 3.0b 3.2b 4.3b 3.2b 10.6a 12.7a
1Small letters indicate significant difference between treatments, and capital letters denote differences within 317
treatments at P≤0.05. 318
319
320
17
3.4. The relationship between sensory attributes and storage conditions 321
The treatments were significantly different (P≤0.05) for all attributes (Table 3). The differences in 322
sensory quality between treatments were overall greater than the biological variation between the 323
packages within a treatment. The scores plot and the loadings plot from the PCA (Fig. 2) showed 324
the overall trends in the quality of the packages of wild rocket. Principal component 1 (PC1) 325
accounted for 63% of the variance in the data (Fig. 2A). This component explained the differences 326
in the accumulated temperature sum, as samples with a low temperature sum were placed to the 327
left and samples with a high sum were placed to the right in the figure. PC2 accounted for 32% of 328
the variance in the data (Fig. 2A). The PC2 explained the differences in packaging materials, as 329
low OTR packages were placed at the bottom and high OTR packages were placed at the top of 330
the figure. The difference between samples with different packaging materials was greater as 331
temperature and time increased (samples were then located further away from each other). The 332
control sample, which had an average freshness score of 12.0 (Table 3), was described as having 333
a high intensity of green leaf color (score 12.9), a high degree of leaves with initial structure (score 334
12.7), a high ability of leaves to sprinkle from each other (sprinklyness score 12.4), a high ability of 335
leaves to snap into two pieces when bent (brittleness score 11.0), a high relative height of product 336
in the tray (score 12.9), a low intensity of wild rocket odor (score 4.7), and a very low intensity of 337
rotten odor (score 1.1) (Fig. 2 and Table 3). As the control leaves were not rotten or in other ways 338
physiologically damaged by the storage conditions (Nielsen et al., 2008), the wild rocket odor in 339
these samples originated from shaking the packages at the sensory evaluation. It is well-known 340
that Brassica odor is formed immediately after wounding the tissues when sulfurous precursors 341
and other compounds are enzymatically broken down into characteristic Brassica odors 342
(Christensen et al., 2007). 343
344
345
18
346
347
Fig. 2. Scores plot (A) and loadings plot (B) from PCA of the sensory data (eight averages over 44 348
replicates). Samples and symbols in Fig. 2A are named according to Table 1. A senescence and 349
fermentation line for packaged wild rocket is shown by arrows. 350
351
19
Leaves packaged in the high OTR film eventually developed brown cut edges, lost green color, 352
turned yellow, smelled, and became rotten as the storage temperature and time increased and 353
leaves became senescent (Fig. 2; Table 2) (Barrett et al., 2010; Nielsen et al., 2008; Toivonen and 354
Brummell, 2008). Leaves packaged in the low OTR film also lost green color as the storage 355
temperature and time increased but these leaves eventually became olive brown with an unusual 356
smoked odor (Fig. 2; Table 3). At the time of sensory evaluation, all the low OTR packages except 357
for the LowOTR10-4 treatment had changed color, and the leaves had lost their initial structure due 358
to too low O2 concentrations for too long a time (Table 1). In these packages, the OTR of the 359
packaging material was too low to compensate for the need for O2 for respiration leading to anoxia 360
and eventually fermentation (Kim et al., 2004). Already after four days’ storage at 10 °C, the 361
sensory quality had changed (Table 3). Interestingly, there were significant differences between 362
the LowOTR20-2 and the LowOTR10-4 samples for ‘initial freshness’, ‘overall freshness’, ‘product 363
height’, ‘tray liquid’, ‘green leaves’, ‘olive green leaves’, ‘rotten leaves’, ‘leaf structure’, 364
‘sprinklyness’, ‘brittleness’, ‘wild rocket odor’ and ‘smoked odor’ (Table 3), even though both 365
treatments had an accumulated temperature sum of 40 °C at evaluation. In this case, the 366
accumulated temperature sum was not a good indication of physiological ageing and it could not 367
be used in the modeling of freshness. Samples packaged in high OTR exposed to 20 °C for three 368
or four days (HighOTR20-3; HighOTR20-4) were not significantly different (P>0.05) (Table 3). 369
These samples were also placed close to each others in the scores plot (Fig. 2A). In the high OTR 370
packages, individual leaves turned yellow and changed texture at different rates as the 371
temperature sum increased from ~0 (Control) to 120 °C resulting in significant differences (P≤0.05) 372
between treatments for ‘color homogeneity’ and ‘texture homogeneity’. The differences in the rate 373
of senescence were due to variations in the physiological ageing of samples and leaves in the tray 374
(Koukounaras et al., 2007). A slight reduction of the O2 concentrations as compared to 375
atmospheric air inside the HighOTR packages (Table 1) did not prevent samples from showing 376
symptoms of senescence within three days’ storage at 20 °C (Martinez-Sanchez et al., 2006b). 377
These results are consistent with other studies of packaged green vegetables (Allende et al., 2004; 378
20
Ares et al., 2008a; Hansen et al., 2001; Kim et al., 2004; Luo et al., 2004; Martinez-Sanchez et al., 379
2006b). Leaves packaged in high OTR kept the relative height in the tray for a longer period than 380
leaves stored at low OTR, which corresponds with other findings (Allende et al., 2004). 381
‘Wild rocket odor’ was linked to ‘fresh’ leaves, ‘cabbage odor’ to senescent leaves and ‘smoked 382
odor’ to fermented leaves (Table 3 and Fig.2B). ‘Wild rocket odor’ was linked to ‘green leaves’ and 383
‘initial and overall freshness’, and thus to fresh looking packages with leaves. ‘Cabbage odor’ was 384
linked to ‘brown cut edges’ and ‘yellow leaves’, and thus to packages with senescence leaves. 385
‘Smoked odor’ was linked to ‘olive brown leaves’, and thus to packages with fermented leaves. 386
‘Rotten odor’ and ‘acidic odor’ were linked to senescence and fermented leaves. As in the loadings 387
plot, these attributes were placed close to zero on PC2 (Fig. 2B). Panelists were unable to 388
differentiate between senescent and fermented leaves on ‘rotten odor’ and ‘acidic odor’. Samples 389
evaluated high in ‘rotten odor’ were also rated significantly different (P≤0.05) in ‘overall freshness’ 390
(Table 3). These samples had lower scores for overall freshness. A high degree of leaves with rot 391
and decay was found in the high and low OTR packages stored for six days at 20 °C. Mold 392
appeared in the low OTR packages stored for six days at 20 °C. Interestingly, the rotting symptoms 393
were different between the low and the high OTR packages due to differences in inside O2 394
concentrations (Table 1). At high OTR level, the leaves became slimy and rotted but kept their 395
initial leaf structure (Table 3). In contrast, leaves lost structure as they rotted in the low OTR 396
packages. Kim et al. (2004) reported decay in low and high OTR packages with fresh-cut salad 397
savoy, however, they did not differentiate between the various symptoms of decay. 398
399
3.5. Correlation between freshness and underlying attributes 400
‘Initial freshness’ was highly correlated with ‘green leaves’ (r2 of 0.98; P≤0.001). ‘Initial freshness’ 401
could be determined by sensory evaluation of ‘green leaves’ alone. ‘Overall freshness’ was highly 402
correlated with ‘wild rocket odor’ (r2 = 0.98, P ≤0.0001), and ‘sprinklyness’ (r2 = 0.91, P = 0.0015), 403
which gave the highest correlation coefficients. In order to further explore the role of the underlying 404
21
attributes for ‘overall freshness’ of packaged wild rocket leaves, different PLS models were 405
performed (Table 4). Interval PLS (iPLS) increased the prediction power of the models as the root 406
mean squared error of cross validation (RMSECV) and the explained variance increased 407
compared to a model based on all sensory attributes. Forward iPLS’s were tested, where 408
increasingly more sensory attributes were included, and reverse iPLS was applied, where 409
attributes were excluded to decide their importance. The models were based on eight samples, 410
and therefore only one PLS component was used, since the low number of samples increase the 411
risk of coincidental correlations (Kjeldahl and Bro, 2010), and furthermore, both procedures were 412
used to find the overall correlations. Via the forward iPLS model, ‘overall freshness’ could be 413
predicted with an RMSECV of 1.01 (Table 4). The attributes contributing to this model were ‘green 414
leaves’, ‘sprinklyness’, ‘wild rocket odor’, ‘acidic odor’ and ‘rotten odor’. The first three attributes 415
were related to a high degree of ‘overall freshness’ while the latter two were related to a low 416
degree of ‘overall freshness’ (Table 3; Fig 2). From the attributes, only ‘green leaves’ could be 417
evaluated at the time of purchase. The other attributes could only be evaluated by sensing of 418
emitted odors or by touching the leaves after removal of the packaging material. In reverse iPLS, 419
the appearance attributes were important, i.e. ‘color and texture homogeneity’. ‘Acidic odor’, ‘wild 420
rocket odor’, ‘cabbage odor’, ‘smoked odor’, ‘yellow leaves’ and ‘olive brown leaves’ were 421
important for attributes related to color and odor (Table 4). Reverse iPLS included more attributes 422
in the model than forward iPLS; however, the RMSECV was slightly lower (0.9), which indicates 423
that by including additional attributes, the overall freshness can be better explained. 424
In summary, the iPLS models showed that ‘overall freshness’ cannot be predicted only from 425
appearance attributes at first sight, as texture and odor attributes are equally important. 426
Furthermore, Peneau et al. (2009) found that both appearance and texture are important for the 427
perception of freshness, and Martinez-Sanchez et al. (2006a) reported that changes in texture 428
leads to unacceptable quality of wild rocket leaves. 429
430
22
Table 4. PLS models predicting ‘overall freshness’ of packaged wild rocket using the underlying 431
sensory attributes. The range of ‘overall freshness’ scores was 0.2-12.2. 432
Model PLS
Components
Explained
Y-variance
R2 (CV) RMSECV Variables
1
PLS 1 91.6% 0.77 2.28 All underlying attributes
iPLS
forward
1 96.8% 0.95 1.01
Green leaves (+)
Sprinklyness (+)
Wild rocket odor (+)
Acidic odor (-)
Rotten odor (-)
iPLS
reverse
1 97.6% 0.96 0.90 Color homogeneity +
Texture homogeneity +
Acidic odor -
Wild rocket odor +
Cabbage odor-
Smoked odor -
Yellow leaves -
Olive brown leaves -
Brown cut edges -
Leaf structure +
Sprinklyness +
Brittleness +
1Attributes with a plus affect the model positively and a minus affect the model negatively. 433
434
5. Conclusion 435
The composition of O2 and CO2 concentrations in the package atmosphere, temperature and 436
storage time had a high impact on ‘initial and overall freshness’ and sensory quality of packaged 437
wild rocket. The ‘initial freshness’ was evaluated at first sight and it was related to the intensity of 438
green leaf color. The attributes related to ‘overall freshness’ could not be evaluated until the 439
opening of the packages. The ‘overall freshness’ was related to color, odor and leaf texture. 440
Packaged wild rocket stored for six days at 2 °C in a film with a high OTR had high scores for 441
‘initial and overall freshness’, i.e. the quality of the leaves was close to the original quality at 442
23
harvest. In contrast, scores for ‘overall freshness’ were low in the high and low OTR packaged wild 443
rocket stored for up to six days at 20 °C. These packages showed signs of either senescence (high 444
OTR) or fermentation (low OTR). Judged by appearance, the low OTR packages were better in 445
retaining product quality than the high OTR packages, but after opening, these differences 446
disappeared. As consumers use appearance factors to provide an indication of freshness, wild 447
rocket stored at too high temperatures (20 °C) must be packaged in high OTR film, alternatively 448
packaged in low OTR film and stored at 2 °C to fulfill customer requirements relative to 449
appearance and brand. 450
451
Acknowledgements 452
The authors would like to thank Jens Michael Madsen, Aarhus University, Department of Food 453
Science, for assistance with the experimental work. Yding Grønt is greatly acknowledged for 454
providing packages of wild rocket and ScanStore A/S for supplying packaging materials for the 455
experiment. The Danish Ministry of Science, Technology and Innovation and the Danish Agency 456
for Science, Technology and Innovation are acknowledged for financial support provided to the 457
Innovation consortium (Innovationskonsortium – Produkttilpasset pakning af frisk frugt og grønt, 458
J.nr. 08-034100). 459
460
24
References 461
Allende, A., Luo, Y.G., Mcevoy, J.L., Artes, F., and Wang, C.Y., 2004. Microbial and quality 462
changes in minimally processed baby spinach leaves stored under super atmospheric oxygen and 463
modified atmosphere conditions. Postharvest Biol.Technol. 33, 51-59. 464
Ares, G., Lareo, C., and Lema, P., 2008a. Sensory shelf life of butterhead lettuce leaves in active 465
and passive modified atmosphere packages. Int.J.Food Sci.Tech. 43, 1671-1677. 466
Ares, G., Martinez, I., Lareo, C., and Lema, P., 2008b. Failure criteria based on consumers' 467
rejection to determine the sensory shelf life of minimally processed lettuce. Postharvest 468
Biol.Technol. 49, 255-259. 469
Barrett, D.M., Beaulieu, J.C., and Shewfelt, R., 2010. Color, flavor, texture, and nutritional quality of 470
fresh-cut fruits and vegetables: desirable levels, instrumental and sensory measurement, and the 471
effects of processing. Crit.Rev.Food Sci. 50, 369-389. 472
Christensen, L.P., Edelenbos, M., and Kreutzman, S., 2007. Fruits and vegetables of moderate 473
climate, in: Berger, R.G. (Ed.), Flavours and fragrances. Chemistry, bioprocessing and 474
sustainability. Springer-Verlag, Berlin, pp. 135-187. 475
Escalona, V.H., Verlinden, B.E., Geysen, S., and Nicolai, B.M., 2006. Changes in respiration of 476
fresh-cut butterhead lettuce under controlled atmospheres using low and superatmospheric oxygen 477
conditions with different carbon dioxide levels. Postharvest Biol.Technol. 39, 48-55. 478
Frost, M.B. and Janhoj, T., 2007. Understanding creaminess. Int.Dairy J. 17, 1298-1311. 479
Hansen, M.E., Sorensen, H., and Cantwell, M., 2001. Changes in acetaldehyde, ethanol and 480
amino acid concentrations in broccoli florets during air and controlled atmosphere storage. 481
Postharvest Biol.Technol. 22, 227-237. 482
Hertog, M.L.A.T., Lammertyn, J., Scheerlinck, N., and Nicolai, B.M., 2007. The impact of biological 483
variation on postharvest behaviour: The case of dynamic temperature conditions. Postharvest 484
Biol.Technol. 43, 183-192. 485
Jacobsson, A., Nielsen, T., and Sjoholm, I., 2004. Influence of temperature, modified atmosphere 486
packaging, and heat treatment on aroma compounds in broccoli. J.Agric.Food Chem. 52, 1607-487
1614. 488
Jacxsens, L., Devlieghere, F., De Rudder, T., and Debevere, J., 2000. Designing equilibrium 489
modified atmosphere packages for fresh-cut vegetables subjected to changes in temperature. 490
Lebensm.Wiss.Technol. 33, 178-187. 491
Jacxsens, L., Devlieghere, F., Ragaert, P., Vanneste, E., and Debevere, J., 2003. Relation 492
between microbiological quality, metabolite production and sensory quality of equilibrium modified 493
atmosphere packaged fresh-cut produce. Int.J.Food Microbiol. 83, 263-280. 494
25
Kalio, H.E., 2008. The effect of film wrapping and nitrogen flushing of produce atmosphere on the 495
storage quality of lettuce. Agric.J. 3, 17-24. 496
Kim, J.G., Luo, Y.G., and Gross, K.C., 2004. Effect of package film on the quality of fresh-cut salad 497
savoy. Postharvest Biol.Technol. 32, 99-107. 498
Kim, J.G., Luo, Y.G., Tao, T., Saftner, R.A., and Gross, K.C., 2005. Effect of initial oxygen 499
concentration and film oxygen transmission rate on the quality of fresh-cut romaine lettuce. 500
J.Sci.Food Agr. 85, 1622-1630. 501
Kjeldahl, K. and Bro, R., 2010. Some common misunderstandings in chemometrics. J.Chemometr. 502
24, 558-564. 503
Koukounaras, A., Siomos, A.S., and Sfakiotakis, E., 2007. Postharvest CO2 and ethylene 504
production and quality of rocket (Eruca sativa Mill.) leaves as affected by leaf age and storage 505
temperature. Postharvest Biol.Technol. 46, 167-173. 506
Lappalainen, R., Kearney, J., and Gibney, M., 1998. A pan EU survey of consumer attitudes to 507
food, nutrition and health: An overview. Food Qual.Prefer. 9, 467-478. 508
Lawless, H.T. and Heymann, H., 2010. Sensory evaluation of food: principles and practices, 509
second ed. Springer, New York. 510
Leardi, R. and Nørgaard, L., 2004. Sequential application of backward interval partial least squares 511
and genetic of relevant spectral regions. J.Chemometr. 18, 486-497. 512
Lonchamp, J., Barry-Ryan, C., and Devereux, M., 2009. Identification of volatile quality markers of 513
ready-to-use lettuce and cabbage. Food Res.Int. 42, 1077-1086. 514
Luo, Y.G., Mcevoy, J.L., Wachtel, M.R., Kim, J.G., and Huang, Y., 2004. Package atmosphere 515
affects postharvest biology and quality of fresh-cut Cilantro leaves. Hortscience 39, 567-570. 516
Mahajan, P.V., Oliveira, F.A.R., Montanez, J.C., and Frias, J., 2007. Development of user-friendly 517
software for design of modified atmosphere packaging for fresh and fresh-cut produce. Innov.Food 518
Sci.Emerg. 8, 84-92. 519
Martinez-Sanchez, A., Allende, A., Bennett, R.N., Ferreres, F., and Gil, M.I., 2006a. Microbial, 520
nutritional and sensory quality of rocket leaves as affected by different sanitizers. Postharvest 521
Biol.Technol. 42, 86-97. 522
Martinez-Sanchez, A., Luna, M.C., Selma, M.V., Tudela, J.A., Abad, J., and Gil, M.I., 2012. Baby-523
leaf and multi-leaf of green and red lettuces are suitable raw materials for the fresh-cut industry. 524
Postharvest Biol.Technol. 63, 1-10. 525
Martinez-Sanchez, A., Marin, A., Llorach, R., Ferreres, F., and Gil, M.I., 2006b. Controlled 526
atmosphere preserves quality and phytonutrients in wild rocket (Diplotaxis tenuifolia). Postharvest 527
Biol.Technol. 40, 26-33. 528
26
Naes, T. and Langsrud, O., 1998. Fixed or random assessors in sensory profiling? Food 529
Qual.Prefer. 9, 145-152. 530
Nielsen, T., Bergstrom, B., and Borch, E., 2008. The origin of off-odours in packaged rucola (Eruca 531
sativa). Food Chem. 110, 96-105. 532
Nørgaard, L., Saudland, A., Wagner, J., Nielsen, J.P., Munck, L., and Engelsen, S.B., 2000. 533
Interval partial least-squares regression (iPLS): A comparative chemometric study with an example 534
from near-infrared spectroscopy. Appl.Spectrosc. 54, 413-419. 535
Page, T., Griffiths, G., and Buchanan-Wollaston, V., 2001. Molecular and biochemical 536
characterization of postharvest senescence in broccoli. Plant Physiol. 125, 718-727. 537
Paulsen, M.T., Ueland, O., Nilsen, A.N., Ostrom, A., and Hersleth, M., 2012. Sensory perception of 538
salmon and culinary sauces - An interdisciplinary approach. Food Qual.Prefer. 23, 99-109. 539
Peneau, S., Brockhoff, P.B., Escher, F., and Nuessli, J., 2007. A comprehensive approach to 540
evaluate the freshness of strawberries and carrots. Postharvest Biol.Technol. 45, 20-29. 541
Peneau, S., Hoehn, E., Roth, H.R., Escher, F., and Nuessli, J., 2006. Importance and consumer 542
perception of freshness of apples. Food Qual.Prefer. 17, 9-19. 543
Peneau, S., Linke, A., Escher, F., and Nuessli, J., 2009. Freshness of fruits and vegetables: 544
consumer language and perception. Brit.Food J. 111, 243-256. 545
Piagentini, A.M., Guemes, D.R., and Pirovani, M.E., 2002. Sensory characteristics of fresh-cut 546
spinach preserved by combined factors methodology. J.Food Sci. 67, 1544-1549. 547
Ragaert, P., Verbeke, W., Devlieghere, F., and Debevere, J., 2004. Consumer perception and 548
choice of minimally processed vegetables and packaged fruits. Food Qual.Prefer. 15, 259-270. 549
Raju, C.S., Løkke, M.M., Suraryo, S., Ward, A.J., and Møller, H.B., 2012. NIR monitoring of 550
ammonia in anaerobic digesters using a diffuse reflectance probe. Sensors 12, 2340-2350. 551
Saltveit, M.E., 2003. Is it possible to find an optimal controlled atmosphere? Postharvest 552
Biol.Technol. 27, 3-13. 553
Seefeldt, H.F., Blennow, A., Jespersen, B.M., Wollenweber, B., and Engelsen, S.B., 2009. 554
Accumulation of mixed linkage (1 -> 3) (1 -> 4)-beta-D-glucan during grain filling in barley: A 555
vibrational spectroscopy study. J.Cereal Sci. 49, 24-31. 556
Seefeldt, H.F., Løkke, M.M., and Edelenbos, M., 2012. Effect of variety and harvest time on 557
respiration rate of broccoli florets and wild rocket salad using a novel O2 sensor. Postharvest 558
Biol.Technol. 69, 7-14. 559
Tano, K., Oule, M.K., Doyon, G., Lencki, R.W., and Arul, J., 2007. Comparative evaluation of the 560
effect of storage temperature fluctuation on modified atmosphere packages of selected fruit and 561
vegetables. Postharvest Biol.Technol. 46, 212-221. 562
27
Toivonen, P.M.A. and Brummell, D.A., 2008. Biochemical bases of appearance and texture 563
changes in fresh-cut fruit and vegetables. Postharvest Biol.Technol. 48, 1-14. 564
565
Paper IV
Color and textural quality of wild rocket measured by multispectral imaging
Løkke, M.M., Seefeldt, H.F., Skov, T. and Edelenbos, M.
Submitted to Postharvest Biology and Technology
1
Research paper for Postharvest Biology and Technology 1
Color and textural quality of packaged wild rocket 2
measured by multispectral imaging 3
4
Mette Marie Løkke1, Helene Fast Seefeldt1, Thomas Skov2, Merete Edelenbos1* 5
1Dept. of Food Science, Aarhus University, Kirstinebjergvej 10, 5792 Årslev, Denmark 6
2Dept. of Food Science, University of Copenhagen, Rolighedsvej 30, 1958 Frederiksberg C, 7
Denmark 8
*Corresponding author: Merete Edelenbos, Aarhus University, Dept. of Food Science, 9
Kirstinebjergvej 10, 5792 Årslev, Denmark 10
Tel: +45 87158334, Fax: +45 87154812 11
E-mail address: [email protected] 12
13
Abstract 14
Green color and texture are important attributes for the perception of freshness of wild rocket. 15
Packaging of green leafy vegetables can postpone senescence and yellowing, but a drawback is 16
the risk of anaerobic fermentation leading to loss of tissue integrity. In this study multispectral 17
imaging was correlated to sensory perception of packaged wild rocket. CIELAB values derived 18
from the images changed during storage, but the values were not sufficient to describe the sensory 19
perception of quality. Furthermore, the combination of wavelengths in the NIR range describing 20
textural changes and in the visual range describing color changes was optimal for explaining the 21
2
impression of freshness. Measurement with a spectrophotometer with an 8 mm aperture was not 22
sufficient to cover the variation in a 100 g package of wild rocket leaves. However, applying 23
multispectral images enabled subtraction of background information and proved to be a more 24
precise method for prediction of color of wild rocket leaves. The sensory perception of initial 25
freshness could be predicted with a RMSECV of 1.5 on a 0-12 scale by multispectral imaging. 26
27
Highlights 28
• Multispectral imaging covers a higher degree of variation than a spectrophotometer with an 29
8 mm aperture 30
• Multispectral images can be used to predict sensory perceived freshness before opening of 31
packaged wild rocket 32
• Aging and loss of tissue integrity due to senescence or anaerobic fermentation can be 33
detected by NIR wavelengths 34
• Yellowing due to senescence can be detected by wavelengths in the visual region 35
36
Keywords: Rucola, wild rocket, multispectral imaging, NIR, quality, green leafy vegetable 37
38
3
1. Introduction 39
Green color is an important quality parameter of packaged wild rocket and is indicative of 40
freshness (Løkke et al. 2012). The actual odor and texture experienced at consumption, along with 41
the color, warrent the repurchase of the product (Barrett et al., 2010). It is therefore important that 42
the quality and freshness experienced at purchase comply with the perceived quality and 43
freshness experienced at consumption. 44
The color of heterogeneous food products can be determined by the human eye or by instrumental 45
color analysis. Green color of wild rocket and other leafy green vegetables is determined by 46
chlorophyll fluorescence (Ferrante and Maggiore, 2007; Lu, 2007), chemical analysis of individual 47
and total chlorophyll pigments (Koukounaras et al., 2006; Martinez-Sanchez et al., 2006a; 48
Piagentini et al., 2002), spectrophotometrical analysis of color using the CIELAB color system 49
(Able et al., 2005; Koukounaras et al., 2006, 2007; Luo et al., 2004), and by sensory analysis 50
(Martinez-Sanchez et al., 2006a, 2006b). Sensory and consumer analyses are optimal methods for 51
evaluation of quality, but the methods are expensive and time-consuming and cannot be used for 52
routine analysis in postharvest research (Brookfield et al., 2011). 53
Spectrophotometric determination of color is a fast and non-destructive method. Often, the color is 54
determined using a hand-held apparatus with a measure aperture of 3-11 mm (Barrett et al., 2010). 55
In the case of green leafy vegetables, many measurements on each leaf have to be taken to cover 56
the variation in color in a package (Koukounaras et al., 2006, 2007; Luo et al., 2004). Therefore, 57
the use of spectrophotometers imposes challenges and limitations for color determinations in 58
heterogeneous food materials, such as senescence of wild rocket. 59
Variation in color of green baby leaves after harvest is a result of high biological variance and 60
heterogeneity of the product at harvest. In packages of wild rocket, an uneven distribution in the 61
loss of green color is found (Løkke et al., 2012). During postharvest senescence, the green 62
chlorophyll pigments are degraded into colorless substances and the yellow carotenoids appear. 63
4
Senescence takes place in all plant cells at different rates after harvest. In a tray of wild rocket, 64
leaves may have different ages at harvest (Koukounaras et al., 2007). Secondly, leaves may have 65
different contents of chlorophylls (Xue and Yang, 2009), and thirdly, the respiration rate of the 66
leaves might also vary at different harvest times depending on growth conditions. Therefore, the 67
large biological variations must be considered not only with regard to optimizing packaging, but 68
also in experimental design, the choice of analytical methods, and during data modeling for 69
predicting quality of packed fruits and vegetables (De Ketelaere et al., 2006; Hertog et al., 2007; 70
Tijskens et al., 2003). 71
Wild rocket is primarily packaged to prevent water loss and wilting. The living cells of the wild 72
rocket leaves respire and thereby use O2 and produce CO2 resulting in modified atmospheres (MA) 73
inside the package. Usually, film materials used for packaging of wild rocket will permit O2 influx 74
and CO2 efflux. The interplay between the O2 transmission rate (OTR) of the film and the 75
respiration rate of the product will determine the final gas composition inside the package. 76
Investigations have shown that optimally designed MAP can delay loss of green color of wild rocket 77
leaves (Martinez-Sanchez et al., 2006b) and other leafy green vegetables like bok choy (Lu, 2007) 78
and cilantro leaves (Luo et al., 2004). MAP, not correctly designed to match the respiration rate of 79
the product, will either lead to anaerobic conditions (too low transmission rate of gasses over the 80
packaging material) or an internal atmosphere composition close to atmospheric (too high 81
transmission rate). Anaerobic conditions will lead to development off-odor and tissue degradation 82
due to respirational fermentation (Allende et al., 2004; Kim et al., 2004), and finally, fermentation 83
will result in acidic degradation of chlorophyll leading to an olive brown color (Toivonen and 84
Brummell, 2008). Atmospheric conditions will accelerate the senescence of the product, initiating 85
enzymatic breakdown of chlorophyll and thereby revealing the underlying yellow carotenoids 86
(Martinez-Sanchez et al., 2006b; Toivonen and Brummell, 2008). Hence, the biological variance in 87
respiration rate leads to different requirements for optimal transmission rate of the packaging 88
material (Martinez-Sanchez et al., 2008; Seefeldt et al., 2012). 89
5
Another important quality parameter of leafy vegetables is the changes in textural characteristics 90
(Løkke et al., 2012; Martinez-Sanchez et al., 2012). Senescence of vegetables is a degradative 91
process, where the cell walls are broken down leading to cell death, and water is released into the 92
intercellular space (Toivonen and Brummell, 2008). Cell collapse may furthermore be induced by 93
fermentation (Allende et al., 2004; Luo et al., 2004). The collapse of cells results in loss of texture. 94
Near infrared (NIR) spectroscopy is an increasingly used technology for non-destructive 95
measurement of quality of fresh fruit and vegetables. The NIR range covers the wavelengths from 96
780-2500 nm and visual (VIS) range covers 380-750 nm. NIR has been used especially for 97
detection of soluble solids content (SSC) in fruit (Nicolai et al., 2007). Scientific records of NIR 98
applications with green vegetables are sparse. The content of essential oils of green plant parts of 99
dill and coriander was predicted by NIR (Schulz et al., 1998) and sensory eating quality of chicory 100
was predicted using a combination of VIS-NIR spectroscopy (Francois et al., 2008). SSC and 101
firmness were predicted in bell pepper by NIR (Penchaiya et al., 2009), and NIR was found to 102
correlate with the chlorophyll content of leafy green vegetables (Xue and Yang, 2009). The 103
conventional NIR apparatuses have drawbacks relative to the fact that the measuring probe only 104
enables measurement without providing spatial information (Gowen et al., 2007). Recently, 105
multispectral and hyperspectral VIS-NIR imaging methods have been used to determine the quality 106
of various fruit (Gowen et al., 2007; Lleo et al., 2009; Lu, 2004; Lunadei et al., 2011; Peng and Lu, 107
2008), and vegetables such as mushrooms (Taghizadeh et al., 2010) and spinach leaves (Lunadei 108
et al., 2012). A novel multispectral imaging device, VideometerLab, enables the determination of 109
both VIS and NIR wavelengths of items with a diameter of 11 cm, and it has been tested for 110
determination of astaxanthin in salmonids (Dissing et al., 2011) and for seed health testing of 111
spinach (Olesen et al., 2011). An advantage of the multispectral imaging method is the size of the 112
sampling area; furthermore, the background of the sample can be subtracted before data analysis 113
leading to relevant information of interest to be considered for further data treatment. 114
6
The aim of this study was to compare CIELAB values from a spectrophotometric analysis of green 115
leafy vegetables with those from a multispectral imaging analysis, and to compare the multispectral 116
analysis with sensory analysis of packaged wild rocket. The hypothesis is that color and textural 117
quality of packaged wild rocket leaves can be predicted by multispectral imaging. Variation in color, 118
appearance and other sensory attributes was induced in the same batch of wild rocket by 119
packaging in films with low or high permeability relative to O2 and storage at different 120
temperatures. 121
122
2. Materials and methods 123
2.1. General 124
Two experiments were carried out with packaged wild rocket. In the first experiment, conventional 125
color determination analysis (spectrophotometer) was compared with multispectral imaging. In the 126
second experiment, sensory evaluation and multispectral imaging were compared. 127
128
2.2. Plant material 129
The wild rocket (Diplotaxis tenuifolia L.) was grown at a commercial grower in Jutland, Denmark 130
(56o00’N, 9o46’E). Two batches of wild rocket were used. Wild rocket for the first experiment was 131
harvested in week 32, 2010 and wild rocket for the second experiment was harvested in week 38, 132
2010. The batches of wild rocket were treated in the same way after harvest. The wild rocket was 133
harvested directly into open boxes. Then the boxes were transferred to air storage at 1 °C and > 134
95% RH in a commercial warehouse until packaging 2 days later. The wild rocket was packaged 135
(100±5 g) on a commercial packing line in polyethylene terephthalate (PET) trays (185×145×70 136
mm) and wrapped with oriented polypropylene film. The OTR was 0.65 pmols-1m-2kPa-1 (LowOTR) 137
and 17.4 pmols-1m-2kPa-1 (HighOTR), respectively, at 23 °C and 50% RH. The packages were 138
7
transported in a commercial cooling truck at 2 °C to the Research Centre and stored at 2 °C and > 139
95% RH until the experiment started the following day. 140
141
2.3. Storage experiments 142
Two storage experiments with packaged wild rocket were carried out. Samples from the first 143
experiment (harvest week 32, 2010) were used for spectrophotometer and multispectral imaging of 144
color. Samples from the second experiment (harvest week 38, 2010) were used for sensory 145
evaluation and multispectral imaging. 146
147
2.3.1. Sampling for spectrophotometer and multispectral imaging 148
A total of 20 packages of wild rocket were used for the color measurements distributed with 10 149
LowOTR and 10 HighOTR packages. The packages were stored at 20 °C in climate chambers 150
(400 L, −9–99øC, Binder KB400, Binder, Tuttlingen, Germany) and measured after 0, 1, 2, 4, and 151
6 days of storage. At each day of measurement, the O2 and CO2 concentrations were measured 152
using an O2 and CO2 gas analyzer equipped with a zirconia and a dual beam infrared sensor 153
(Checkmate 9900, PBI Dansensor, Ringsted, Denmark). After these measurements, the packages 154
were opened and a representative sample of approximately 15 g wild rocket leaves was sampled 155
from the bottom, middle and top layers in each package and placed randomly on a 13 cm petri dish 156
for color measurements. One petri dish of leaves was analyzed per package giving a total of 30 157
samples per OTR treatment. 158
159
2.3.2. Sampling for sensory analysis and multispectral imaging 160
8
Sensory analysis and multispectral imaging were carried out on packages of wild rocket treated in 161
accordance with the experimental design shown in Table 1. The same batch of wild rocket leaves 162
was used, and samples of wild rocket packaged in low and high OTR film were stored at different 163
temperatures (2, 10 and 20 °C) and times (2, 3, 4 and 6 days) to create packages with variation in 164
package atmosphere O2 and CO2 concentrations. The control samples were packaged in high 165
OTR film and stored for six or seven days at 2 °C as sensory evaluation took place on day six or 166
day seven after initiation of the experiment. All samples were then stored at 2 °C until they were 167
transferred to elevated temperatures as shown in Table 1. This design enabled sensory evaluation 168
of wild rocket from the same batch and thus minimized the biological variance of samples for post-169
evaluation treatments. The control sample was a sample with the lowest degree of accumulated 170
temperature sum at evaluation. Although not completely freshly harvested, the control sample was 171
as close to a fresh sample as possible. With this design, it was possible to create color and textural 172
differences in the packaged wild rocket by the use of packaging film OTR, storage temperature and 173
time. A total of 352 packages (8 treatments, 4 sessions, 11 panelists) of wild rocket were used for 174
the sensory analysis. Prior to this analysis, the O2 and CO2 concentrations inside the packages 175
were measured using an O2 and CO2 gas analyzer, as described above. After each sensory 176
session, wild rocket from three panelists were analyzed by multispectral imaging to make a direct 177
link between the results of the sensory analysis and the multispectral imaging data; in total, 96 178
packages (8 treatments, 4 sessions, 3 panelists) were analyzed. For multispectral imaging, the 179
leaves were sampled as described in section 2.3.1. 180
181
9
Table 1. The experimental design for sensory and instrumental analysis of packaged wild rocket 182
color and textural quality. The symbol color changes according to accumulated sum at elevated 183
temperature and is used in figures. 184
Sample Symbol
color
OTR
(pmol s-1
m-2
kPa-1
)
Elevated temperature
°C
Days at elevated
temperature °C
Accumulated sum at
elevated temperature
°C
Control 17.4 - 0 0
LowOTR10-4 0.65 10 4 40
LowOTR20-2 0.65 20 2 40
LowOTR20-6 0.65 20 6 120
HighOTR10-4 17.4 10 4 40
HighOTR20-3 17.4 20 3 60
HighOTR20-4 17.4 20 4 80
HighOTR20-6 17.4 20 6 120
185
2.4. Spectrophotometric analysis 186
A Minolta Spectrophotometer (CM-600d) (Minolta, Osaka, Japan) with an aperture of 8 mm was 187
used for spectrophotometric analysis. The spectrophotometer collected spectra in the range 360-188
740 nm with 10 nm intervals with a 10° standard observer and D65 standard illuminant. The 189
spectrophotometer was calibrated using a standard white plate from the manufacturer. The color 190
was recorded in the L*, a* and b* color space. Samples with a low L-value are darker than samples 191
with a high L-value. The Hue [H° = 180 + tan-1(b*/a*)] was calculated from the a* and b* values 192
(Mcguire, 1992). Samples with a high Hue are greener than samples with a low Hue. The color of 193
the leaves in the petri dish was measured at five points spread on the petri dish. Average values of 194
the readings in these five points were used for the data analysis. Measurements from the 195
spectrophotometer day 1 were removed since they were defective. 196
2.5. Multispectral imaging 197
The VideometerLab instrument (Videometer A/S, Denmark) was used for the multispectral image 198
analysis. The instrument obtained multispectral images at 18 different wavelengths in the VIS and 199
10
the NIR range of the electromagnetic spectrum: 405, 435, 450, 470, 505, 525, 570, 590, 630, 645, 200
660, 700, 780, 850, 870, 890, 940 and 970 nm. The instrument illuminates the object with diffuse 201
light and an integrating sphere ensures that the light is scattered evenly with a uniform, diffuse light 202
at illumination. Each diode emits light in a specific wavelength ensuring that only light of one 203
wavelength is present at a time. The setup of the instrument is further described in Dissing et al. 204
(2011). Each multispectral image consists of 18 separate images, one from of each of the 18 205
wavelengths. For measuring, a petri dish with wild rocket leaves was placed under the integrated 206
sphere. From the multispectral image, L*a*b* values were calculated in each pixel with a standard 207
illuminant of D50 and with a 2° standard observer. 208
209
2.6. Sensory analysis 210
The sensory analysis was performed with a trained panel according to international standards 211
(ASTM STP 913, 1986) as described by Løkke et al. (2012). In brief, the panel was tested for their 212
ability to describe and evaluate color vision, taste, odor and texture. The panel developed a 213
vocabulary of 19 sensory attributes covering product appearance, odor and texture which the 214
panelists were trained in during several training sessions. In addition, the descriptor ‘freshness’ 215
was evaluated before (initial freshness) and after opening (overall freshness) without prior 216
consensus among the panelists, i.e. each panelist used his or her own concept of ‘freshness’. In 217
total, 21 attributes were included. The attributes specifically used for PLS modeling in this study 218
were: ‘initial freshness’ (by individual concept, freshness 1), ‘green leaves’ (overall intensity of the 219
green color of the leaves), ‘yellow leaves’ (overall intensity of the yellow and yellow-greenish color 220
of the leaves), ‘olive-brown leaves’ (overall intensity of the olive-brown color of the leaves) and 221
sprinklyness (degree of the ability of the leaves to be sprinkled, assessed by touching). For full 222
description of all attributes are referred to Løkke et al. (2012). The samples of wild rocket were 223
assessed on two consecutive days with two sessions each day. The samples were assessed at 224
11
individual speed on an unstructured 15 point line scale with intensity rating ranging from low (value 225
0) to high (value 15) using a direct computerized registration system (Fizz software, 2.30C, 226
Biosystemes, Couternon, France). 227
2.7. Data analysis 228
Segmenting images into distinct regions is an important preprocessing step in image analysis 229
before further analysis. Image segmentation was performed using the VideometerLab software 230
version 2.10 (Videometer A/S, Hørsholm, Denmark). To exclude the background from the images, 231
all items but the wild rocket leaves, were subtracted by a Canonical Discriminant Analysis (CDA) 232
(Cruzcastillo et al., 1994) and segmented using a simple threshold. The average spectra from the 233
multispectral imaging analysis and the average L*a*b* values from the spectrophotometer and the 234
imaging analysis were calculated and imported into MATLAB. From these data, the Hue was 235
calculated as described in section 2.4. The average spectrum of the leaves in the multispectral 236
images was used for the data analysis after subtraction of background. This was a simplified, but 237
objective, way of analyzing multispectral images. This was considered to be a valid approach since 238
the data were correlated to sensory analysis that was an overall perception of the wild rocket 239
leaves in each package and not of individual leaves. 240
Tukey’s honest significance difference (HSD) test was used for multiple comparisons of treatment 241
means of the multispectral CIELAB values at a significance level of P ≤ 0.05. Pearsons correlation 242
coefficients were calculated between L*, Hue and five sensory attributes. The calculations were 243
made in MATLAB ver. 7.12 (MathWorks, Natick, MA, USA). The sensory data was analyzed as 244
described in Løkke et al. (2012). 245
A principal component analysis (PCA) model of the sensory data was calculated in order to find the 246
main variations in data (Wold et al., 1987). Data were averaged across panelists and replicates 247
(11×4) and autoscaled before the analysis. A PCA model was also calculated in order to find the 248
main variations in the multispectral image data. The calculation was done using averages across 249
12
panelists and replicates (3×4). In order to visualize the variation of single packages compared to 250
the average of 12 packages a PCA of the spectra from single packages was also performed and 251
the loadings were used to predict the score values of the average spectra in the model. 252
A Partial Least Squares (PLS) regression model (Wold et al., 2001) was computed in order to test 253
the correlation between the sensory and the spectral data. The calculation was performed both 254
with the average of spectra from all replicates and with spectra from single samples (packages). 255
Sensory data were autoscaled and average values were used. 256
Individual PLS models were used to estimate the possibility of predicting sensory attributes from 257
spectral data of single packages. Average sensory values across panelists and replicates were 258
used in the PLS models in order to overcome uncertainty in sensory evaluation (Lawless and 259
Heymann, 2010). It was not possible to let all assessors test the same package since the ultra-260
volatile odors can only be assessed instantly after opening of a package. Therefore, each single 261
spectrum was paired with the average value of the sensory evaluation of the particular treatment. 262
Segmented cross validation was performed to find the optimal number of components providing a 263
prediction error that can be expected if future packages are predicted by the model (each session 264
was a segment). Determination of the prediction error is possible when testing the model in this 265
way, hereby revealing the suitability of the prediction model for wild rocket quality from a future 266
session. PCA models of the replicate spectra were computed for each treatment for detection of 267
outliers and one HighOTR20-3 package was considered an outlier. Overall, the packages were 268
considered as representatives for (future) packages of wild rocket with differences in color and 269
textural quality. 270
13
3. Results and discussion 271
3.1. Comparison of color measurements from spectrophotometer and multispectral imaging 272
Variation in storage temperature, time and O2 transmission rate (OTR) of the packaging film was 273
used to create wild rocket samples with variations in color and texture. The O2 and CO2 274
concentrations within the packages changed during storage (Figs. 1A and 1B) which had an impact 275
on product color (Figs. 1C, 1D, 1E and 1F). At low OTR, the O2 concentration decreased to <0.2 276
kPa after 2 days at 20 °C while the CO2 concentration increased to 16 kPa (Fig. 1A). In packages 277
with high OTR, the O2 and CO2 concentrations also changed, but more slowly. The O2 278
concentration decreased to 10 kPa and CO2 increased to 10 kPa at day 6 (Fig. 1B). The 279
differences in L* and Hue during storage of the high and low OTR packages were greater 280
measured by multispectral images (Figs. 1D and 1F) than by the spectrophotometer (Figs. 1C and 281
1E). The packaged wild rocket had a Hue of 115 at the start of the experiment (Fig. 1F), as 282
measured by multispectral image analysis. During storage, the Hue of the leaves decreased to 101 283
in the high OTR and to 112 in the low OTR film indicating that the green color decreased. This 284
result corresponds with a study of rocket leaves, where the color of mature leaves decreased from 285
a Hue of 130 to a Hue of 117 during 10 days’ storage at 10 °C in the presence of high O2 286
concentrations (Koukounaras et al., 2007). The numbers from the two instruments cannot be 287
compared directly, since the calculation of a* and b* values, and thus Hue in the color space, 288
differed slightly between instruments in the standard illuminant and observer. Overall, the 289
multispectral image analysis (Fig. 1F) reported greater variations in Hue during storage than the 290
spectrophotometer analysis (Fig. 1E). Also L* differed between the two instruments, and likewise 291
Hue, the L* of the images reported greater variations (Fig. 1D) than those of the spectrophotometer 292
(Fig. 1C). At high OTR L* increased from 37 to 44 during storage showing that the leaves lost 293
greenness and obtained a lighter color (Fig. 1D). At low OTR, L* first increased reaching a value of 294
40 and then decreased to 36, indicating that the color of the leaves first became lighter (value 40) 295
and then darker (value 36) as the O2 concentrations decreased (Fig. 1A). This result could indicate 296
14
that the leaves eventually became water-soaked from the development of low O2 atmosphere 297
concentrations inside the packages (Lunadei et al., 2012). 298
The difference in CIELAB values of the wild rocket leaves packaged in the two materials was more 299
pronounced when determined from multispectral images (Figs. 1C and 1E) than from 300
spectrophotometer (Figs. 1E and 1F). The multispectral image analysis covers larger amounts of 301
leaves than the spectrophotometer analysis leading to more representative values from the 302
multispectral image analysis. This method covers approximately 80 cm2 of leaves in a sample 303
compared to approximately 0.5 cm2 by the spectrophotometer analysis. It is possible to repeat the 304
number of measurements with the spectrophotometer compared to the multispectral image 305
analysis, but this is time consuming and a non-efficient approach for measurements of color in 306
green leafy vegetables (Lunadei et al., 2012). Based on these results, multispectral image analysis 307
was chosen for further investigations on quality aspects of packaged wild rocket. 308
15
309
Fig. 1. Packaged wild rocket stored for up to 6 days at 20 °C, harvest week 32, 2010. Bars represent 310
the standard deviation between two replicate packages. L*a*b*-values were measured with a 311
spectrophotometer or by multispectral image analysis and from a*b*-values Hue were calculated. 312
313
3.2. The relationship between CIELAB values and sensory analysis 314
In the previous section, research results showed that multispectral image analysis could 315
discriminate between samples of wild rocket with varied internal packaging atmospheres. In this 316
experiment, we studied the relationship between L* and Hue and sensory quality of packaged wild 317
16
rocket with differences in O2 and CO2 concentrations. The results of the gas measurements 318
showed that the inside package atmospheres in this experiment (harvest week 38, year 2010) 319
were almost similar to those of an earlier experiment (not shown) in which the O2 and the CO2 320
concentrations varied with temperature, time and OTR (Table 2). The samples from week 38 were 321
thus considered as representative for packaged wild rocket. The gas composition influenced the 322
quality of the packaged wild rocket (Fig. 1A and B). It was found previously that the perception of 323
overall freshness was dependent on color, odor and texture quality of the wild rocket (Løkke et al 324
2012). Therefore, it is of interest to be able to measure the quality changes instrumentally, since 325
sensory analysis is time consuming and costly. Samples from the sensory descriptive analysis 326
were also subjected to multispectral image analysis, in order to test whether the instrument could 327
be used to predict some of the quality changes experienced during storage of packed wild rocket. 328
The CIELAB values of the wild rocket in each multispectral image were calculated from the 329
spectral images. Only the Hue showed significant difference (P≤0.05) between samples (Table 2). 330
Also the sensory data showed significant differences (P≤0.05) between samples for green color, 331
yellow color and olive brown color (Table 2). The Control, HighOTR10-4 and LowOTR10-4 332
packages were for example significantly different (P≤0.05) from the HighOTR20-6 packages on 333
Hue (Table 2). L* was not significantly different (P>0.05) between samples (Table 2). Overall, the 334
sensory panel could distinguish better than the multispectral imaging analysis between different 335
treatments of packaged wild rocket, which is seen on the Tukey HSD letters (Table 2). One can 336
argue that the results of the sensory evaluation is a result of an average of 44 packages (11 337
panelists in 4 sessions), while the multispectral imaging is a result of an average of 12 packages 338
(samples from 3 panelists in 4 sessions). Given the differences among individual humans in 339
sensory functions, repeated measurements are necessary in sensory analysis (Lawless and 340
Heymann, 2010). 341
342
17
Table 2. Mean L* and Hue values from the CIELAB color system, calculated from the multispectral 343
images. O2 and CO2 measured before the sensory analysis and sensory determination of color is also 344
reported in Løkke et al. (2012). 345
Sample L* Hue °
O2 kPa
CO2 kPa
Green leaves
Yellow leaves
Olive brown leaves
Sprinklyness
Control 39a 110.7ab 19.9a 1.7g 12.9a 0.3c 1.1c 12.4a LowOTR10_4 38a 110.1ab 0.5f 15.7a 12.1ab 0.4c 1.7c 9.8b LowOTR20_2 38a 108.3abc 0.4f 15.9a 7.0c 0.8c 6.2b 1.6d LowOTR20_6 39a 106.6abc 0.1f 14.9b 0.6e 0.3c 13.5a 0.4d HighOTR10_4 39a 111.9a 18.4b 3.6f 10.5b 1.3c 2.0c 11.4ab HighOTR20_3 42a 107.0abc 13.4c 8.4e 5.0d 7.5b 1.5c 6.4c HighOTR20_4 40a 104.8bc 12.5d 9.3d 4.2d 8.3b 1.6c 5.7c HighOTR20_6 41a 101.8c 9.9e 11.8c 1.2e 11.2a 1.4c 1.9d Averages within columns with different letters are significant different (P≤0.05). 346
347
The Hue of the high OTR samples was negatively correlated to ‘yellow leaves’ (r=-0.96). When all 348
treatments were included, the correlation coefficient decreased to r= -0.83 (Fig. 2). ‘Initial 349
freshness’ was positively correlated to ‘green leaves’ (r=0.98), and green color’ was correlated to 350
Hue (r=0.85). Løkke et al. (2012) showed that ‘overall freshness’ of packages was influenced by 351
texture attributes here represented by sensory ‘sprinklyness’. Hue was slightly correlated to 352
‘sprinklyness’ (r=0.72), but this result probably reflects that there was a correlation between 353
‘sprinklyness’ and ‘green color’ (r=0.88). In summary, CIELAB values can report simple color 354
measurements of packaged wild rocket but not textural attributes. 355
18
356
Fig. 2. Correlation plots of average L* and Hue and sensory scores of ‘initial freshness’, ‘green 357
leaves’, ‘yellow leaves’, ‘olive brown leaves’ and ‘sprinklyness’. L* and Hue are calculated from the 358
multispectral images. The correlation coefficient (r) is shown in each plot. Plots with black-filled 359
symbols are described in the text. 360
3.3. Relationship between multispectra and sensory analysis 361
To further investigate the ability of multispectral imaging analysis to describe color and textural 362
quality of packaged wild rocket, a multispectral approach was used. The average spectra from 363
three packaged samples of wild rocket are seen in Fig. 3. These samples were first subjected to 364
sensory analysis and then to multispectral imaging. In the red yellow VIS region (550-700 nm), 365
which includes chlorophyll absorbance peaks (Lunadei et al., 2012), the HighOTR20-6 sample had 366
higher mean reflection intensities than the Control and the LowOTR20-6 samples (Fig. 3). This 367
result shows that more light was reflected to the instrument and thus less absorbed by the leaves. 368
This result indicates that the chlorophyll content was lower in these leaves as compared to the 369
19
content of the Control and the LowOTR20-6 samples. The control sample differed from the High 370
and the LowOTR20-6 samples in the NIR regions (Fig. 3), while no differences were seen in the 371
400-500 nm regions (Fig. 3), which is the region of carotenoid detection. A lower reflectance in the 372
NIR region may be due to breakdown of cell walls during senescence (Nicolai et al., 2007). 373
Furthermore, the reflectance in the NIR regions is highly influenced by the water content (Buning-374
Pfaue, 2003; Cen and He, 2007). As the cells collapse during senescence and fermentation, cell 375
water is released into the intercellular space (Toivonen and Brummell, 2008), and this will lead to 376
additional free cellular water and thus change the reflectance of water from the sample. The lower 377
reflection in the NIR bands of these samples may therefore be due to more free cellular water also 378
related to the breakdown of cell walls. The two phenomena in the spectra cannot be separated 379
from each other (Nicolai et al., 2007). 380
381
382
Fig. 3. Average spectra from the VideometerLab of control, LowOTR20-6 and HighOTR20-6 samples 383
(average of 12). The bar indicates the region of the visual (VIS) and the near infrared (NIR) region of 384
the spectrum. The pictures show trays of packaged wild rocket used for the experiment. 385
386
400 500 600 700 800 900 10000
20
40
60
80
100
120
nm
Mean in
tensity
Control
HighOTR20-6
Low OTR20-6Control
HighOTR20-6
LowOTR20-6
VIS NIR
20
The main variation of the sensory evaluation was also visualized in a PCA, and along PC1 387
(principal component 1) two groups of samples moving from fresh to less fresh were found (Figs. 388
4A and 4B). Along PC2, a separation of senescent and fermented packages was found. This 389
difference along PC2 was due to the two different packaging materials that induce very different 390
degradation processes: High OTR packages had signs of senescence resulting in change from 391
green leaves to yellow leaves, brown cut edges, rotten leaves, loss of texture and a heterogeneous 392
appearance in both color and texture. Low OTR packages showed signs of anaerobic respiration 393
resulting in fermentation and a rather uniform texture with olive brown leaves, a smoked odor and 394
loss of tissue integrity. Shortly, PC1 mainly explained the textural degradation of the product and 395
PC2 explained the difference in color and other quality faults experienced due to the two packaging 396
materials. 397
Interestingly, the scores plot from the average spectra reveals the same trends as found in the 398
scores plot from the sensory data. The scores plot of the average spectra from the multi-spectral 399
images (Fig. 4C) showed a trend from right to left along PC1 reflecting increased degradation of 400
leaves due to senescence or fermentation. Samples packaged in high OTR film were placed at the 401
top of the plot whereas samples packaged in low OTR film were placed at the bottom. This position 402
of the samples in a PCA was similar to previous results with a PCA of sensory data (Fig. 4A). The 403
detail that the scores plots of the PCAs of the sensory and the multispectral data were similar is a 404
strong indication for a correlation between these two approaches of measuring the quality of green 405
leafy vegetables. The loadings in Fig. 4D revealed mainly the contribution of NIR bands to PC1. 406
The NIR bands were therefore associated with increased senescence of the samples. Samples 407
placed to the right in the scores plot (Fig. 4C) had high signals in the NIR region (Fig. 4D) while 408
samples placed to the left (Fig 4C) had low signals in the region. PC2 described the difference 409
between the two packaging materials reflecting the differences in color of these samples; this is 410
manifested mainly by the VIS bands (Fig. 4C). This indicates that the wavelengths in the NIR 411
region, in contrary to the CIELAB values, gave additional information on product quality. This 412
means that the combination of information from both VIS and NIR wavelengths appears to hold the 413
information of both color and texture of packaged wild rocket. Control samples were characterized 414
by having high signals in the NIR region an415
(HighOTR20-6) were characterized by low NIR and high VIS signals while fermented samples 416
(LowOTR20-6) were characterized by low NIR and VIS signals. 417
418
419
Fig. 4. Scores plot (A) and loadings plot (B) from 420
in packages stored at different temperatures (reported in421
loadings plot (D) from a PCA of average422
taken by VideometerLab. See Table 1 for an explanation of symbols in Figs. 4A and 4C.423
means that the combination of information from both VIS and NIR wavelengths appears to hold the
information of both color and texture of packaged wild rocket. Control samples were characterized
by having high signals in the NIR region and low signals in the VIS region. Senescent samples
6) were characterized by low NIR and high VIS signals while fermented samples
6) were characterized by low NIR and VIS signals.
plot (A) and loadings plot (B) from a PCA of average sensory evaluation
in packages stored at different temperatures (reported in Løkke et al. (2012)
average spectra reading of eight treatments of packaged wild rocket
See Table 1 for an explanation of symbols in Figs. 4A and 4C.
21
means that the combination of information from both VIS and NIR wavelengths appears to hold the
information of both color and texture of packaged wild rocket. Control samples were characterized
d low signals in the VIS region. Senescent samples
6) were characterized by low NIR and high VIS signals while fermented samples
sensory evaluation of wild rocket
)). Scores plot (C) and
reading of eight treatments of packaged wild rocket
See Table 1 for an explanation of symbols in Figs. 4A and 4C.
22
424
3.4. Prediction of color and textural quality by multispectral imaging 425
In order to further investigate the relation between the data from the multispectral images and the 426
sensory analysis, a PLS model was calculated from the data matrices of multispectral (X) and 427
sensory data (Y) from averages of the 12 samples (4 sessions × 3 panelists) (Fig. 5A). The 428
correlation loadings plot (Fig. 5A) can be used to interpret the degree of explained variance by the 429
first two PLS components. The position of the variables with respect to the dotted lines indicates 430
how much variation of the individual variables that is described by the first two PLS components. 431
Hence, the more connected a spectral variable is to the main variations described by the model, 432
the further away from the center it is. Likewise, the further away the sensory variables are located 433
from the center, the more variance is explained by the model. The sensory perception of texture 434
was correlated with the NIR bands (Fig. 5A), as they were located in the same part of the plot, 435
whereas the sensory perception of ‘yellow leaves’ and ‘brown cut edges’, were associated with the 436
VIS bands. From the plot, it seemed possible to predict 'yellow leaves' from variable 590-660 nm 437
and 'sprinklyness' from variable 780 nm. ‘Green leaves’ and freshness before and after opening 438
were also placed close to the outer circle, but no X-variables were placed close to these attributes. 439
‘Color and texture homogeneity’ were placed oppositely to ‘yellow leaves’, and therefore also in the 440
590-660 nm VIS bands. This result is probably more a sign of negative correlation to 'yellow 441
leaves' than actually correlation to these X-variables. 442
23
443
444
24
Fig. 5. Correlation loadings from the PLS model (A) with eight average spectra as X (black dot) and 445
average sensory data as Y (red ∆). Explained Y-variation component 1: 58% component 2: 29%. 446
Correlation loadings from PLS model (B) with spectra from single packages (96) as X (black dots) 447
and average sensory data as Y (red ∆). Explained Y-variation component 1: 40%, component 2: 23%. 448
The numbers in the plots refer to wavelengths. 449
450
3.5 Prediction of color and textural quality by multispectral imaging using single packages 451
It is of interest to determine if the sensory quality can be predicted by an instrumental method such 452
as the VideometerLab. There was a correlation between the results of the sensory analysis and the 453
multispectral images, when average values were used in the models. However, in everyday 454
measurements, it is unrealistic to use 12 samples for each prediction of color and textural quality; 455
therefore, such predictions should be possible to perform on the basis of single packages. 456
Furthermore, the use of a low number of samples in correlations increases the risk of wrong 457
conclusions (Kjeldahl and Bro, 2010). A preliminary study showed that the variation in spectral data 458
from repeated sampling on the same package was smaller than the variation between packages 459
(data not shown). When studying the spectral data from a single sample compared to average 460
values from additional samples, results showed that these data covered a much larger variation 461
than represented by average values (Fig. 6). Furthermore, in the correlation loadings plot of a PLS 462
model with single spectra (Fig. 5B), the sensory attributes were placed much closer to the middle 463
than in the model with the average spectra (Fig.5A). Hence, the sensory attributes were less well 464
described by the first two components in the model from single spectra, than by the first two 465
components in the model from average spectra. This means that the variation between the 466
packages complicated the model. 467
25
468
Fig. 6. Scoreplot from a PCA of multispectra from single packages (triangles and squares). Averages 469
of equal packages are predicted (bold ’O’ connected with a line, average of 12 samples). All symbols 470
are colored according to temperature sum at elevated temperature (see Table 1). 471
The possibility of predicting sensory attributes by individual PLS models from spectra of single 472
packages (Table 3) was investigated. ‘Initial freshness’ could be predicted with an RMSECV of 1.5 473
using 4 PC’s and explaining 92% of the variation in the y-data, and variables from both the VIS- 474
and NIR range were important for the model. ‘Green leaves’ and ‘yellow leaves’ were well 475
predicted from the multispectral images (Table 3). Leaves deteriorated from fermentation, seen as 476
olive brown leaves, were difficult to predict using multi-spectral imaging. This is partly due to the 477
category having only two y-values different from the lowest value, which makes it a poor candidate 478
for PLS modeling. 479
Color is an important appearance attribute for initial perception of freshness and at the point of 480
purchase, however, only appearance can be judged; however, appearance does not always agree 481
with perceived overall freshness after opening (Løkke et al. 2012). The overall impression of 482
overall freshness after opening depends on appearance, odor and texture. Texture attributes such 483
as ‘sprinklyness’ could also be predicted using data in the NIR region (Table 3). Overall, the 484
26
multispectral images using both VIS and NIR wavelengths showed promising results in predicting 485
color and textural quality of packaged wild rocket. 486
Table 3. Results from PLS models calculated from spectra of single packages. 487
Attribute Range RMSECV PC Expl Y-var
Initial freshness 0.2-11.8 1.5 4 92%
Green leaves 0.6-12.9 1.6 4 91%
Yellow leaves 0.3-11.2 1.5 3 89%
Olive brown leaves 1.1-13.5 - - -
Sprinklyness 0.4-12.4 1.7 4 88%
488
4. Conclusion 489
Sensory determined color and texture changes of packaged wild rocket could be predicted using 490
multispectral images. Multispectral images gave a more comprehensive description than CIELAB 491
values of the variation in color and textural quality of senescent and fermented, packaged wild 492
rocket. The combination of information from both VIS and NIR wavelengths holds the information 493
of color and texture, respectively, and thus the ability to quantify changes in product quality during 494
storage. The possibility of subtracting background and a larger measurement area made the 495
multispectral images more comprehensive than measurements with a spectrophotometer. 496
Acknowledgements 497
The authors would like to thank Jens Michael Madsen, Aarhus University, Department of Food 498
Science, for assistance with the experimental work. Yding Grønt is greatly acknowledged for 499
providing the packages of wild rocket and Scanstore for supplying the packaging materials. The 500
Danish Ministry of Science, Technology and Innovation and the Danish Agency for Science, 501
Technology and Innovation are acknowledged for financial support provided to the Innovation 502
consortium (Innovationskonsortium – Produkttilpasset pakning af frisk frugt og grønt, J.nr. 08-503
034100). 504
27
References 505
Able, A.J., Wong, L.S., Prasad, A., and O'Hare, T.J., 2005. The physiology of senescence in 506
detached pak choy leaves (Brassica rapa var. chinensis) during storage at different temperatures. 507
Postharvest Biol.Technol. 35, 271-278. 508
Allende, A., Luo, Y.G., Mcevoy, J.L., Artes, F., and Wang, C.Y., 2004. Microbial and quality 509
changes in minimally processed baby spinach leaves stored under super atmospheric oxygen and 510
modified atmosphere conditions. Postharvest Biol.Technol. 33, 51-59. 511
Barrett, D.M., Beaulieu, J.C., and Shewfelt, R., 2010. Color, flavor, texture, and nutritional quality of 512
fresh-cut fruits and vegetables: desirable levels, instrumental and sensory measurement, and the 513
effects of processing. Crit.Rev.Food Sci. 50, 369-389. 514
Brookfield, P.L., Nicoll, S., Gunson, F., Harker, F., and Wohlers, M., 2011. Sensory evaluation by 515
small postharvest teams and the relationship with instrumental measurements of apple texture. 516
Postharvest Biol.Technol. 59, 179-186. 517
Buning-Pfaue, H., 2003. Analysis of water in food by near infrared spectroscopy. Food Chem. 82, 518
107-115. 519
Cen, H. and He, Y., 2007. Theory and application of near infrared reflectance spectroscopy in 520
determination of food quality. Trends Food Sci.Tech. 18, 72-83. 521
Cruzcastillo, J.G., Ganeshanandam, S., Mackay, B.R., Lawes, G.S., Lawoko, C.R.O., and Woolley, 522
D.J., 1994. Applications of Canonical Discriminant-Analysis in Horticultural Research. Hortscience 523
29, 1115-1119. 524
De Ketelaere, B., Stulens, J., Lammertyn, J., Cuong, N.V., and De Baerdemaeker, J., 2006. A 525
methodological approach for the identification and quantification of sources of biological variance in 526
postharvest research. Postharvest Biol.Technol. 39, 1-9. 527
Dissing, B.S., Nielsen, M.E., Ersboll, B.K., and Frosch, S., 2011. Multispectral Imaging for 528
Determination of Astaxanthin Concentration in Salmonids. Plos One 6, e19032. 529
Ferrante, A. and Maggiore, T., 2007. Chlorophyll a fluorescence measurements to evaluate 530
storage time and temperature of Valeriana leafy vegetables. Postharvest Biol.Technol. 45, 73-80. 531
Francois, I.M., Wins, H., Buysens, S., Godts, C., Van Pee, E., Nicolai, B., and De Proft, M., 2008. 532
Predicting sensory attributes of different chicory hybrids using physico-chemical measurements 533
and visible/near infrared spectroscopy. Postharvest Biol.Technol. 49, 366-373. 534
Gowen, A.A., O'Donnell, C.P., Cullen, P.J., Downey, G., and Frias, J.M., 2007. Hyperspectral 535
imaging - an emerging process analytical tool for food quality and safety control. Trends Food 536
Sci.Tech. 18, 590-598. 537
28
Hertog, M.L.A.T., Lammertyn, J., De Ketelaere, B., Scheerlinck, N., and Nicolai, B.A., 2007. 538
Managing quality variance in the postharvest food chain. Trends Food Sci.Tech. 18, 320-332. 539
Kim, J.G., Luo, Y.G., and Gross, K.C., 2004. Effect of package film on the quality of fresh-cut salad 540
savoy. Postharvest Biol.Technol. 32, 99-107. 541
Kjeldahl, K. and Bro, R., 2010. Some common misunderstandings in chemometrics. J.Chemometr. 542
24, 558-564. 543
Koukounaras, A., Siomos, A.S., and Sfakiotakis, E., 2006. 1-Methylcyclopropene prevents 544
yellowing of rocket ethylene induced leaves. Postharvest Biol.Technol. 41, 109-111. 545
Koukounaras, A., Siomos, A.S., and Sfakiotakis, E., 2007. Postharvest CO2 and ethylene 546
production and quality of rocket (Eruca sativa Mill.) leaves as affected by leaf age and storage 547
temperature. Postharvest Biol.Technol. 46, 167-173. 548
Lawless, H.T. and Heymann, H., 2010. Sensory evaluation of food: principles and practices, 549
second ed. Springer, New York. 550
Lleo, L., Barreiro, P., Ruiz-Altisent, M., and Herrero, A., 2009. Multispectral images of peach 551
related to firmness and maturity at harvest. J.Food Eng. 93, 229-235. 552
Løkke, M.M., Seefeldt, H.F., and Edelenbos, M., 2012. Freshness and sensory quality of packaged 553
wild rocket salad. Postharvest Biol.Technol. Submitted. 554
Lu, R., 2004. Multispectral imaging for predicting firmness and soluble solids content of apple fruit. 555
Postharvest Biol.Technol. 31, 147-157. 556
Lu, S.M., 2007. Effect of packaging on shelf-life of minimally processed Bok Choy (Brassica 557
chinensis L.). LWT.Food Sci.Technol. 40, 460-464. 558
Lunadei, L., Diezma, B., Lleo, L., Ruiz-Garcia, L., Cantalapiedra, S., and Ruiz-Altisent, M., 2012. 559
Monitoring of fresh-cut spinach leaves through a multispectral vision system. Postharvest 560
Biol.Technol. 63, 74-84. 561
Lunadei, L.L., Galleguillos, P., Diezma, B., Lleo, L., and Ruiz-Garcia, L., 2011. A multispectral 562
vision system to evaluate enzymatic browning in fresh-cut apple slices. Postharvest Biol.Technol. 563
60, 225-234. 564
Luo, Y.G., Mcevoy, J.L., Wachtel, M.R., Kim, J.G., and Huang, Y., 2004. Package atmosphere 565
affects postharvest biology and quality of fresh-cut Cilantro leaves. Hortscience 39, 567-570. 566
Martinez-Sanchez, A., Allende, A., Cortes-Galera, Y., and Gil, M.I., 2008. Respiration rate 567
response of four baby leaf Brassica species to cutting at harvest and fresh-cut washing. 568
Postharvest Biol.Technol. 47, 382-388. 569
29
Martinez-Sanchez, A., Allende, A., Bennett, R.N., Ferreres, F., and Gil, M.I., 2006a. Microbial, 570
nutritional and sensory quality of rocket leaves as affected by different sanitizers. Postharvest 571
Biol.Technol. 42, 86-97. 572
Martinez-Sanchez, A., Luna, M.C., Selma, M.V., Tudela, J.A., Abad, J., and Gil, M.I., 2012. Baby-573
leaf and multi-leaf of green and red lettuces are suitable raw materials for the fresh-cut industry. 574
Postharvest Biol.Technol. 63, 1-10. 575
Martinez-Sanchez, A., Marin, A., Llorach, R., Ferreres, F., and Gil, M.I., 2006b. Controlled 576
atmosphere preserves quality and phytonutrients in wild rocket (Diplotaxis tenuifolia). Postharvest 577
Biol.Technol. 40, 26-33. 578
Mcguire, R.G., 1992. Reporting of Objective Color Measurements. Hortscience 27, 1254-1255. 579
Nicolai, B.M., Beullens, K., Bobelyn, E., Peirs, A., Saeys, W., Theron, K.I., and Lammertyn, J., 580
2007. Nondestructive measurement of fruit and vegetable quality by means of NIR spectroscopy: A 581
review. Postharvest Biol.Technol. 46, 99-118. 582
Olesen, M.H., Carstensen, J.M., and Boelt, B., 2011. Multispectral imaging as a potential tool for 583
seed health testing of spinach (Spinacia oleracea L.). Seed Sci.Technol. 39, 140-150. 584
Penchaiya, P., Bobelyn, E., Verlinden, B.E., Nicolai, B.M., and Saeys, W., 2009. Non-destructive 585
measurement of firmness and soluble solids content in bell pepper using NIR spectroscopy. J.Food 586
Eng. 94, 267-273. 587
Peng, Y. and Lu, R., 2008. Analysis of spatially resolved hyperspectral scattering images for 588
assessing apple fruit firmness and soluble solids content. Postharvest Biol.Technol. 48, 52-62. 589
Piagentini, A.M., Guemes, D.R., and Pirovani, M.E., 2002. Sensory characteristics of fresh-cut 590
spinach preserved by combined factors methodology. J.Food Sci. 67, 1544-1549. 591
Schulz, H., Drews, H.-H., Quilitzsch, R., and Krüger, H., 1998. Application of near infrared 592
spectroscopy for the quantification of quality parameters in selected vegetables and essential oil 593
plants. J.Near Infrared Spec. 6, A125-A130. 594
Seefeldt, H.F., Løkke, M.M., and Edelenbos, M., 2012. Effect of variety and harvest time on 595
respiration rate of broccoli florets and wild rocket salad using a novel O2 sensor. Postharvest 596
Biol.Technol. 69, 7-14. 597
Taghizadeh, M., Gowen, A., Ward, P., and O'Donnell, C.P., 2010. Use of hyperspectral imaging for 598
evaluation of the shelf-life of fresh white button mushrooms (Agaricus bisporus) stored in different 599
packaging films. Innov.Food Sci.Emerg. 11, 423-431. 600
Tijskens, L.M.M., Konopacki, P., and Simcic, M., 2003. Biological variance, burden or benefit? 601
Postharvest Biol.Technol. 27, 15-25. 602
Toivonen, P.M.A. and Brummell, D.A., 2008. Biochemical bases of appearance and texture 603
changes in fresh-cut fruit and vegetables. Postharvest Biol.Technol. 48, 1-14. 604
30
Wold, S., Esbensen, K., and Geladi, P., 1987. Principal Component Analysis. 605
Chemometr.Intell.Lab. 2, 37-52. 606
Wold, S., Sjostrom, M., and Eriksson, L., 2001. PLS-regression: a basic tool of chemometrics. 607
Chemometr.Intell.Lab. 58, 109-130. 608
Xue, L.H. and Yang, L.Z., 2009. Deriving leaf chlorophyll content of green-leafy vegetables from 609
hyperspectral reflectance. ISPRS J.Photogramm. 64, 97-106. 610