research article assessing the spectral separability of...

Research ArticleAssessing the Spectral Separability of Flue Cured TobaccoVarieties Established on Different Planting Dates and underVarying Fertilizer Management Levels

Ezekia Svotwa1 Anxious J Masuka2 Barbara Maasdorp3 and Amon Murwira4

1 Kutsaga Research Station PO Box 1909 Harare Zimbabwe2 Tobacco Research Board Kutsaga Research Station PO Box 1909 Harare Zimbabwe3Department of Crop Science University of Zimbabwe Harare Zimbabwe4Department of Geography and Environmental Science University of Zimbabwe Harare Zimbabwe

Correspondence should be addressed to Ezekia Svotwa esvotwa2gmailcom

Received 30 September 2013 Revised 24 January 2014 Accepted 11 February 2014 Published 26 March 2014

Academic Editor David Clay

Copyright copy 2014 Ezekia Svotwa et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

TheNDVI was used to discriminate tobacco variety assess fertilizer levels and determine the impact of planting date on separatingcrops A split plot design with four planting dates September October November and December as main plots variety as subplotand fertilizer treatments as sub-subplots was used Radiometric measurements were taken from 5m times 5m sampling plots usinga multispectral radiometer The September October and November crops had significant variety x fertilizer treatment differences(119865 lt 005) from the age of 10 weeks T 66 and KRK26 varieties had similar (119865 gt 005) NDVI values and these were greater (119865 lt 05)than those for K E1 The 100 and the 150 fertilizer treatments were similar (119865 gt 005) and both were greater (119865 lt 005) thanthe 50 fertilizer treatments All of the fertilizer and variety treatments at the December planting dates had similar reflectancecharacteristics (119875 gt 005) which were lower (119875 lt 005) than the September and October planting dates The results showed thatplanting dates varieties and fertilizer levels could be distinguished using spectral data Weeks 10-11 and 15 after the start of theexperiment were optimal for separating the planting date effect

1 Background

Flue cured tobacco variety planting time and fertilizer man-agement can be direct sources of yield and quality differencesThese variances affect plant canopy biophysical propertiesand chemical composition Su et al [1] reported the consider-able effect that different nutrients and fertilizer rates have onplant leaf chemical composition According toMackown et al[2] an understanding of the level of N fertilisation may bea suitable diagnostic test of crop N sufficiency that could beused for yield and quality predictions

It is widely accepted that nitrogen (N) plays a key rolein the production of tobacco Nitrogen is a constituent ofchlorophyll the green coloring matter in plants that playsa role in absorbing light for photosynthesis The molecularstructure of the chlorophyll incorporates a large proportionof total leaf nitrogen in the form of protein Several studies

have found that foliar chlorophyll concentration provides anaccurate indirect estimate of plant nutrient status [3]

Fertilizer management levels affect the crop canopymineral composition and the way the canopy interacts withsolar radiation Interaction between light and interceptingmolecules in the canopy results in scattering and absorptionby the atomic bonds electrons or atoms in the molecule [4]Thewavelength atwhich the processes of absorption or reflec-tion take place is highly specific for the atomic bonds andthe molecular architecture of the intercepting target makingthe identification of individual chemical components withinthe intercepting canopy possible [5]

Canopy reflectance of irradiance can be used as a rapidmeans of assessing the N status of the crop and can be apredictor of final yield [6]The spectral absorbance propertiesof chemical components in plant leaves are manifested in thereflectance spectra of leaves [7]This offers the opportunity of

Hindawi Publishing CorporationInternational Journal of AgronomyVolume 2014 Article ID 219159 9 pageshttpdxdoiorg1011552014219159

2 International Journal of Agronomy

using measurements of reflected radiation as a nondestruc-tive method for quantifying pigments as affected by level ofcropmanagement [3] Similar cases of use of radiometric datafor crop assessment and N estimation were reported over awide range of crops [8 9]

The spectral characteristics of vegetation vary with wave-length Analysing vegetation using remotely sensed datarequires knowledge of the structure and function of vege-tation and its reflectance properties [10] Leaf chlorophyllstrongly absorbs radiation in the red and blue wavelengthsbut reflects in the green wavelength [11] The internal struc-ture of healthy leaves acts as a diffuse reflector of near-infraredwavelengthsMeasuring andmonitoring the infraredreflectance can be useful in determining crop health [12]

Well managed crop canopies appear greenest and becomered or yellow with decrease in chlorophyll content due topoor fertilisation [13] Blackmer and Schepers [8] were ableto separate N treatments using the near 550 nm reflectancemeasured on maize leaves detached from a field N fertilizerexperiment and also found that nutrient stressed canopieshad a lower spectral reflectance in the NIR and higherred wavebands when compared with unstressed canopiesOsborne et al [14] and Graeff et al [15 16] could identifyand quantify nutrient deficiencies by means of reflectancemeasurements based on selected stress specific wavelengthranges Aase et al [17] reported a relationship between wheatin-season dry matter and NIRred ratios suggesting thatreflectancemeasurements could be used to estimate plant leafdry matter

The crop health and yield of tobacco depend on goodagronomic practices which include among other thingsapplying the recommended fertilizer rates and selection ofappropriate planting dates [18] Tobacco planting starts fromthe first day of September and to the end of the second weekofDecember in Zimbabwe depending onwhether the farmerintends to irrigate the crop or rely on seasonal rainfall [19]In Zimbabwe the resource poor farmers resort to applyinghalf the recommended fertilizer rates to save on their costs ofproduction [19]These practices have an impact on the healthphysiological status and final yield of the crop

The Normalised Difference Vegetation Index (NDVI) isthe most commonly used spectral index in assessing cropvigor vegetation cover and biomass production from mul-tispectral satellite data [20 21] The index is calculated usingthe formula NDVI = (NIRminusRed)(NIR+Red) As describedby Jackson et al [20] the index is considered a good indicatorof the amount of vegetation and hence useful in distin-guishing vegetation from soil Benedetti and Rossini [22]used the NDVI in assessing wheat growth vigor and appliedthe information in estimating yield Research at Kutsaga hasestablished that the index is positively correlated with mostbiophysical characteristics of flue cured tobacco [23]

Monitoring of crop phenology by remote sensing isvery important for many practical agronomic applicationsand notably for yield forecasting purposes [24] Accurateanalysis of temporal and spatial variations throughout theseason can be used to interpret crop spectral response toagronomic management as emphasized by Tucker et al [25]It is essential to quantify the relationship between spectral

properties of tobacco and agronomic parameters in orderto utilize the full potential of remote sensing for assessingcrop condition and for forecasting yield The use of effectiveremote sensing techniques in crop growth studies wouldeliminate the need for extensive field sampling by giving goodnutrient adequacy detection capability [14]

This paper presents the results of an experiment whichwas conducted to correlate the spectral indices with theagronomic variables of three flue cured tobacco varietiesfour planting dates and three fertilizer management levelsBy measuring the energy that is reflected by a crop canopysurface over a variety of different wavelengths it is hypothe-sized that spectral signatures for the varying fertilizer levelsvarieties and planting dates can be distinguished from theadjacent bare ground It is also hypothesized that from spec-tral characteristics of the canopies of the four planting datesone can identify the most suitable temporal windows formeaningful measurements This information could be veryuseful in estimating the area covered by each planting datetreatment and in large area crop yield and quality predictionsusing remote sensing

2 Method

21 Study Area This experiment was conducted for threeseasons at Kutsaga Research Station near Harare Zimbabwe(17∘551015840S 31∘081015840E Altitude 1480m above sea level [26])during summer and between 2010 and 2012 The experimentwas carried out under irrigation on a sandy loam soil (728sand 88 silt and 184 clay)The long-term average annualrainfall for the station is 882mmBefore 2011 the research areahad been planted to tobacco followed by four years of Chlorisgayana cv Katambora Rhodes grass [27] The range averagemonthly temperature for the station is 8∘C

A split plot design with four planting dates SeptemberOctober November and December as main plots variety assubplot and fertilizer treatments as sub-subplots was used(Table 1) Three tobacco varieties KRK26 T 66 and K E1developed by Kutsaga Research Station were used whilethree fertilizer management levels (50 100 and 150recommended) were applied by hand (Table 2) The recom-mended compound fertilizer rate from soil test results was700 kgha while that for ammonium nitrate (345 N) was96 kgha at 4 weeks after planting and 75 kgha after topping

An initial overhead preirrigation of 50ndash60mm wasapplied at about 60 days before transplanting the tobacco Attransplanting 4-5 liters of water was applied in each plantinghole A 15mm settling in irrigation run was applied just afterplanting followed by a 28-day stress period to promote rootdevelopment after which 25minus30mm is applied to facilitate anitrogenous fertilizer application

The experimental plots were located on well-drainedgranitic sands and were low in available nitrogen mediumin available phosphorus and high in available potassiumcontent throughout the profile During February of 2010 and2011 the plots were disced after a three-year Katambora grassfallow period to incorporate grass Agricultural lime wasapplied using recommendations given from soil test results

International Journal of Agronomy 3

Table 1 Soil analysis and fertilizer recommendations for experimental plots

Lab Your Crop N P K fertilizers Required nutrients Suggested fertilizer Lime (kgha)N P K Rate (kgha) Application top dressing

113315 Land block 1 Tobacco lowast lowast lowast 140ndash160 70ndash100 lowast 20 17 800 (ie 2 cup no 22plant) 760113316 Land block 2 Tobacco lowast lowast lowast 100ndash120 70ndash100 lowast 20 17 600 (ie cup no (30 + 5)plant) NIL113317 Land block 3 Tobacco lowast lowast lowast 100ndash120 70ndash100 lowast 20 17 600 (ie cup no (30 + 5)plant) 600113301 Land block 4 Tobacco lowast lowast lowast 80ndash100 70ndash100 lowast 20 17 500 (ie cup no 30plant) 600

Table 2 Variety and fertilizer treatments

Variety FertilizerKRK26 50 fertilizerKRK26 100 fertilizerKRK26 150 fertilizerT 66 50 fertilizerT 66 100 fertilizerT 66 150 fertilizerK E1 50 fertilizerK E1 100 fertilizerK E1 150 fertilizer

to raise the soil pH from 53 to 63 levels optimum fortobacco production (Table 1) For the three years precedingthe 2010 experiments the sites were under Katambora grassto control nematodes Recommended management practiceswere followed [26] except for N P K levels and plantingtimes which were treatments in the experiment Fertilizerand lime application was done using the results of the soilanalysis provided by the soil chemistry laboratory at Kutsaga(Table 1)

22 Fertilizer Treatments TheN P K treatmentswere hand-applied in bands about 10 cm deep and 30 cm to each side ofa row at planting while N treatments were applied at about4 weeks after transplanting after topping and at 6 weeks afterplanting

23 Procedure Radiometric measurements were taken on5m times 5m square sampling plots using a hand-held multi-spectral radiometer (Cropscan MSR-5 450ndash1750 nm) withthe FOV centering over rows Each sampling plot consistedof 5 rows each with 32 plants spaced at 56 cm The interrowdistance was 12m Normalised Difference Vegetation Index(NDVI) was calculated from the spectral bands obtained inChannels 3 and 4 which correspond to the red (630ndash690 nm)and near-infrared (nir) (60ndash900 nm) regions of the MSR 5respectively using the following formula

NDVI = nir minus rednir + red

(1)

Fasheun and Balogun [28] suggested that NDVI was sen-sitive to the biochemical properties of leaf and phenologicalstage of crop respectively

The multispectral radiometer (MRS 5) was positionedfacing vertically downward at 1m above crop canopies and

measurements were taken around solar noon to minimizethe effect of diurnal changes in solar zenith angle In total 10measurements were taken per sampling unit and reflectancemeasurements were averaged for each sampling plot to esti-mate a single reflectance value Three-dimensional positionslatitude longitude and altitude for the whole experimentalarea and for each treatment plot were taken using a GarminPersonal Navigator (GPS V) to enable repeated sampling atthe same locationThe averageNDVI data for the two seasonswere used for the analysis

24 Data Analysis NDVI data was analysed by analysis ofvariance and statistically significant treatment effects wereseparated using least significant differences (LSD) The datawas analysed using the Genstat 92 statistical package at 5level of significance Studentrsquos t-test calculations were doneto compare the planting date effect and graphs were plottedusing Excel 2007

3 Results

The September planted croprsquos NDVI reached a peak of07ndash085 at 10ndash13 weeks after planting (Table 3) Duringthe juvenile stages all the variety fertilizer managementtreatments were statistically similar (119875 gt 005) There weresignificant (119875 lt 005) fertilizer level differences in the NDVIstarting from the age of 10 weeks after planting At peakthe NDVIs for 100 and the 150 fertilizer levels werestatistically similar and were both significantly greater thanthe 50 fertilizer treatmentsTheNDVI for the three varietieswere also significantly different (119875 lt 005) from 7 weeks afterplanting when canopy reflectance measurements startedThevarieties T 66 and KRK26 NDVIs were statistically similar(119875 gt 005) and were significantly greater than that for K E1treatments There was no variety x fertilizer level interactioneffect (119875 gt 005)This trend was maintained beyond the peakstage and as the NDVI fell to the lowest levels at 16-17 weeksof age

The NDVI for all the treatments in the October plantedcrop rose sharply from week 6 after planting to peak atapproximately 9 weeks after planting (Table 4) Beyond thepeak the NDVI also fell gently to reach the minimum at 13weeks of age Like the September planting all the treatmentshad similar (119865 gt 005) NDVI from 6 weeks after plantinguntil the age of 10 weeks when a peak NDVI was attainedThe 100 and the 150 fertilizer levels of T 66 and KRK26were also statistically similar (119875 gt 005) and were sig-nificantly greater than both their 50 fertilizer and all


Table 3 The NDVI temporal profiles for the September 15 planted crop

Variety Fertilizer Weeks after planting7 8 10 12 13 15 17 19 21 22

KRK26 50 fertilizer 065 079 079 078 070 054 041 032 037 050KRK26 100 fertilizer 069 082 083 082 070 055 041 029 034 043KRK26 150 fertilizer 069 082 085 082 076 061 048 040 035 043T 66 50 fertilizer 064 080 079 078 068 054 037 029 040 050T 66 100 fertilizer 065 082 085 083 075 067 052 041 046 040T 66 150 fertilizer 065 082 084 085 079 070 060 034 039 047K E1 50 fertilizer 061 075 076 075 065 052 023 027 034 058K E1 100 fertilizer 061 073 073 075 063 053 030 030 042 059K E1 150 fertilizer 063 076 078 080 070 065 053 032 039 053

Variety lowast fertilizer119865-Probability 084 062 001 026 010 033 011 009 021 032

SED 003 002 002 002 002 005 006 005 004 005LSD 007 005 003 004 005 010 013 010 008 010

Fertilizer

50 fertilizer 063 078 078 077 067 053 033 029 037 053100 fertilizer 065 079 080 080 070 058 041 033 041 048150 fertilizer 065 080 082 082 075 065 054 035 038 048

119865-Probability 051 032 0001 0004 lt0001 0002 lt0001 010 015 015SED 002 001 0009 001 001 003 004 003 002 003LSD 004 003 0019 002 003 006 008 006 004 006

Variety

K RK26 068 081 082 081 072 057 043 034 035 045T 66 064 081 083 082 074 064 050 035 042 046K E1 062 075 076 077 066 056 035 029 038 057

119865-Probability 001 lt0001 lt0001 0002 lt0001 003 0004 013 003 0004SED 002 001 0009 001 001 003 004 003 002 003LSD 004 003 0019 002 003 006 008 006 004 006

the K E1 treatments The varietal NDVI differences were lesspronounced when compared with those for the Septembercrop Like the September crop there was no interaction effectobserved (119875 gt 005)

The November fertilizer treatment effect also followed acomparable trend to the September and October plantings(Table 5)Therewas a gentler decline of theNDVI values afterthe peak stage (9-10 weeks after planting)Theminimum 05-06 was attained at 15 weeks after planting From the crop ageof 12 weeks after planting the varieties showed significant(119875 lt 005) NDVI differences The variety x fertilizer levelinteraction was also absent

The December crop peak was attained at 8 weeks afterplanting and was maintained to the age of 10 weeks afterplanting (Table 6)There were no significant (119865 lt 005) treat-ment differences from the first sampling to the end of reapingin week 13 after planting

The NDVI values for the September and October cropswere similar (119905 gt 005) and both were significantly greaterthan those for the November and the December crops(Figure 1) There was a general fall of the NDVI with laterplantings At all fertilizer levels K E1 had the highest NDVIin the October planted crop

The September crop had the longest field life span of 22weeks as compared to the 17 15 and 12 weeks for theOctober

0

01

02

03

04

05

06

07

08

09

6 11 16 21 26

ND

VI

Weeks after September 15 planting

SeptemberOctoberNovember

DecemberGround

Figure 1 NDVI profiles for the September October November andDecember crops

November and December crops respectively (Figure 1)The September planted crop could be separated from laterplantings in the first 9 weeks of establishment Between 6and 16 weeks of the tobacco growing season the September


Table 4 The NDVI temporal profiles for the October 15 planted crop



V lowast F119865-Probability 022 099 088 099 083 011 045 087 028 031

SED 003 002 003 002 002 003 008 005 004 003LSD 006 004 006 004 005 006 019 011 009 006

Fertilizer


119865-Probability 027 0007 015 0007 0005 lt0001 015 084 002 054SED 002 001 002 001 001 002 005 003 003 002LSD 003 002 003 002 003 003 011 006 005 003

Variety

KRK26 035 081 077 081 079 073 058 048 068 055T 66 038 083 078 083 081 078 060 054 066 056K E1 040 084 080 084 081 079 062 051 064 056

119865-Probability 0008 004 023 004 019 0005 071 023 047 061SED 002 001 002 001 001 002 005 003 003 002LSD 003 002 003 002 003 003 011 006 005 003

October and November crops could be separated because oftheir significant (119905 lt 005) average NDVI differences Thiswas well before the December crop was established TheSeptember crop declined at 15 weeks after planting whenthe October crop was at peak Only the December plantedcrop was actively growing in the field beyond 24 weeks afterthe start of the experiment A summary of the approximatetimes for collecting reflectance data for separating the cropsis presented in Table 7

4 Discussions

The general NDVI profile followed the same pattern as theleaf area index [25] Similar to the findings of Peng et al[29] the spectral characteristics of tobacco in the luxuriantgrowth stages between 0 and 10weeks after plantingwere verysimilar making it difficult to separate variety and fertilizereffect using a single-phase remote sensing image

The September and October crops had higher averagereflectance values than the November and December cropsThis could have resulted from the dry conditions underwhichthese crops are established The crops are generally subjectedto 4ndash6 weeks of dry conditions before irrigation is applied orrains received [26] The moderate stress is considered ben-eficial to the tobacco plants as deeper root development is

encouraged in preparation for the rapid growth phase Theadditional root development results in increases in yield andquality of the cured leaf [18] In addition the stimulation ofroot development subsequently promotes the attainment ofdesirable chemical composition

Generally periods of severe water stress however cancause a negative spectral response as argued by Negeswara-Rao et al [30] and in the NDVI profiles Such responseswere evidenced by the negative oscillations during weeks9 and 10 of the September planted crop The general declineof theNDVIwith late plantingwas similar to the documentedresponse of tobacco yield to planting time [26] thus showingthe potential for NDVI to be used in developing yieldforecasting models [31]

The separation of flue cured varieties and fertilizer levelsusing canopy reflectance could be done at the age of atleast 10 weeks after planting when NDVI differences amongthese became significant The 10-week stage coincided withthe period of intense reaping for most flue cured tobaccovarieties At this stage the potential of the different crops inthe field as determined by variety and fertilizermanagementcould easily be detected and considered in overall yieldestimation

The higher NDVI for both 100 and 150 fertilizer leveltreatments for T 66 and KRK26 could be an indication of


Table 5 The NDVI temporal profiles for the November 15 planted crop



VAR lowast fertilizer119865-Probability 065 071 072 025 052 034 058 058 074 049

SED 003 006 005 004 004 003 003 003 004 003LSD 006 013 011 009 009 007 007 007 008 007

Fertilizer


119865-Probability 046 028 029 lt0001 lt0001 lt0001 lt0001 lt0001 062 014SED 002 003 003 002 002 002 002 002 002 002LSD 004 007 006 005 005 004 004 004 005 004

058 059 056

Variety

KRK26 025 052 063 069 063 062 058 065 065 058T 66 029 053 069 071 066 068 066 061 059 057K E1 027 051 068 068 064 067 061

001 003 035119865-Probability 010 083 012 061 031 003 001 002 002 002

SED 002 003 003 002 002 002 002 004 005 004LSD 004 007 006 005 005 004 004

higher yield potential for these treatments According to Jianget al [32] the NDVI reflects the growing status of greenvegetation thusmaking the task of cropmonitoring and cropyield estimation by remote sensing realizable The differentrates of fall of NDVI after the peak could be due to fertilizer xvariety treatment related ripening rates [32]

A long life span in the field is important in increasingthe crop leaf area duration (LAD) which is very essential forthe crop to complete the development of sinks [33 34] Inthe experiment crop life span generally decreased with lateplanting in the experiment

The similarity of average NDVI for the three varietiesin September could be an indication of equal yield poten-tial under such growing conditions making it possible toestimate combined area and yield forecast for the threeThe difference between the bare ground and crop canopyreflectance could be directly related to biomass and shouldbe considered in yield forecasting models

The NDVI profiles for the four planting dates couldbe used to select window periods at which the single-phase NDVIs were significantly different This would enableseparate monitoring of crops for specific planting dates

As Jiang et al [35] pointed out crop area estimation forthe different planting dates could be realised using remotesensing techniques on the basis of time serial NDVI datatogether with agriculture calendars

5 Conclusions

Tobacco varieties and fertilizer management effect could bedistinguished using spectral data after the attainment of peakcanopy reflectance Crops planted on different planting dateshad different temporal profiles Periods where these weresignificantly different could be used in calculating crop areaforecasts

Varieties T 66 and KRK26 could not be distinguished byboth single-day remote sensing imagery and temporal NDVIprofile in the field while KE 1 could be separated from atleast 10 weeks after planting Under-fertilized tobacco couldbe distinguished in the field by its lower NDVI than theoptimally and over-fertilized crop from at least 10 weeks afterplanting

The results of this study indicated that the second tolast week of November and the period between late Januaryand late February to early March were the optimal times for


Table 6 The NDVI temporal profiles for the December 15 planted crop

Variety Fertilizer Weeks after planting5 6 7 8 9 11 12 13

KRK26 50 fertilizer 034 043 054 063 051 048 040 038KRK26 100 fertilizer 044 061 067 061 064 051 031 036KRK26 150 fertilizer 037 047 057 053 055 045 040 039T 66 50 fertilizer 039 051 061 058 059 051 035 036T 66 100 fertilizer 039 052 061 051 063 049 037 041T 66 150 fertilizer 035 054 061 066 068 053 038 040K E1 50 fertilizer 028 039 050 049 056 048 034 039K E1 100 fertilizer 039 051 063 060 061 053 036 040K E1 150 fertilizer 031 050 058 058 066 056 027 041

V lowast F119865-Probability 053 047 045 004 051 030 017 074

SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer

50 fertilizer 034 044 055 057 055 049 036 038100 fertilizer 041 055 064 057 063 051 035 039150 fertilizer 034 050 059 059 063 052 035 040

119865-Probability 005 008 005 079 007 055 079 047SED 003 004 003 004 004 003 003 002LSD 006 009 007 007 007 005 006 004

Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


Table 7 Optimum times for discriminating crops for different planting times

Weeks from September 1 Spectrally dominant crop Crop area (A)0ndash9 September (s) 119860 = 119860 s only10ndash12 September (s) + October (o) 119860 = 119860 s + 119860o (119860 s gt 119860o)

13ndash15 September (s) October (o) and November (n) 119860 = 119860 s + 119860o + 119860nNDVIo gt NDVIs gt NDVIn

16ndash18 October (o) and November (n) 119860 = 119860o + 119860nNDVIo ge NDVIn

18ndash20 October (o) November (n) and December (d) 119860 = 119860o + 119860n + 119860d(NDVIo = NDVIn) ge NDVId

20ndash22 November (n) and December (d) 119860 = 119860o + 119860n + 119860dNDVIo = NDVIn = NDVId

23ndash26 December (d) 119860 = 119860dNDVId only

One has the following(1) 119860 crop area(2) 119860s 119860o 119860n and 119860d areas for the September October November and December crops(3) NDVIs NDVIo NDVIn and NDVId NDVI for the September October November and December crops

discriminating the September-October planted tobacco fromthe nonirrigated November-December tobacco Althoughthe December crop remained in the field beyond 24 weeksafter the September 1 planting date spectral confusion due

to weeds other crops and even sucker regrowth from earlierplanted crop could make its separation difficult

Using time serial NDVI data with agricultural calendarstobacco planted on different planting dates can be separated


and each crop area can be separately possibly determinedusing remote sensing and agronomic techniques

The temporal windows for planting date separation estab-lished in this research have to be tested in future tobacco croparea forecast research More work is also needed to establishthe relationship between the Cropscan MSR 5 reflectanceand reflectances for selected satellite platforms like ModisLandsat 5 and Landsat TM which have been used for thesame purpose over large areas in other crops [36]

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors are grateful to the Tobacco Research BoardKutsaga Research Station for funding this series of exper-iments on ldquodeveloping flue cured tobacco crop area andyield forecastingmodels using remote sensing and agronomictechniquesrdquo The Tobacco Research Board is a researchInstitution and the first author of the paper is registeredDPhil student with the University of Zimbabwe with theother three as supervisors This paper is the third of fivepapers developed from the five objectives of the DPhilstudiesThe TRB strives to publish all research works that arecarried out not for financial gain

References

[1] F Su L Fu H Chen and L Hong ldquoBalancing nutrient use forflue-cured tobaccordquo Better Crops vol 90 no 4 pp 23ndash25 2006

[2] C T MacKown S J Crafts-Brandner and T G Sutton ldquoEarly-season plant nitrate test for leaf yield and nitrate concentrationof air-cured burley tobaccordquoCrop Science vol 40 no 1 pp 165ndash170 2000

[3] G A Blackburn ldquoRemote sensing of forest pigments usingairborne imaging spectrometer and LIDAR imageryrdquo RemoteSensing of Environment vol 82 no 2-3 pp 311ndash321 2002

[4] O Mutanga Hyperspectral remote sensing of tropical grassquality and quantity [PhD thesis]WageningenUniversity ITCWageningen The Netherlands 2004

[5] J G FerwerdaMeasuring andmapping the variation of chemicalcomponents in foliage using hyperspectral remote sensing [PhDthesis] Wageningen University ITC Wageningen The Nether-land 2005

[6] E Boegh H Soegaard N Broge et al ldquoAirborne multispectraldata for quantifying leaf area index nitrogen concentrationand photosynthetic efficiency in agriculturerdquo Remote Sensing ofEnvironment vol 81 no 2-3 pp 179ndash193 2002

[7] A A Gitelson andM N Merzlyak ldquoRemote sensing of chloro-phyll concentration in higher plant leavesrdquo Advances in SpaceResearch vol 22 no 5 pp 689ndash692 1998

[8] T M Blackmer and J S Schepers ldquoUse of a chlorophyll meterto monitor nitrogen status and schedule fertigation for cornrdquoJournal of Production Agriculture vol 8 no 1 pp 56ndash60 1995

[9] P M Hansen and J K Schjoerring ldquoReflectance measurementof canopy biomass and nitrogen status inwheat crops using nor-malized difference vegetation indices and partial least squares

regressionrdquo Remote Sensing of Environment vol 86 no 4 pp542ndash553 2003

[10] G P Asner ldquoBiophysical and biochemical sources of variabilityin canopy reflectancerdquo Remote Sensing of Environment vol 64no 3 pp 234ndash253 1998

[11] C S T Daughtry C L Walthall M S Kim E B De Colstounand J E McMurtrey III ldquoEstimating corn leaf chlorophyll con-centration from leaf and canopy reflectancerdquo Remote Sensing ofEnvironment vol 74 no 2 pp 229ndash239 2000

[12] O W Liew P C J Chong B Li and A K Asundi ldquoSignatureoptical cues emerging technologies for monitoring planthealthrdquo Sensors vol 8 no 5 pp 3205ndash3239 2008

[13] D Haboudane J R Miller E Pattey P J Zarco-Tejada andI B Strachan ldquoHyperspectral vegetation indices and novelalgorithms for predicting green LAI of crop canopies modelingand validation in the context of precision agriculturerdquo RemoteSensing of Environment vol 90 no 3 pp 337ndash352 2004

[14] S L Osborne J S Schepers D D Francis and M R Schlem-mer ldquoDetection of phosphorus and nitrogen deficiencies incorn using spectral radiancemeasurementsrdquoAgronomy Journalvol 94 no 6 pp 1215ndash1221 2002

[15] S Graeff D Steffens and S Schubert ldquoUse of reflectancemeasurements for the early detection of N P Mg and Fedeficiencies in Zea mays Lrdquo Journal of Plant Nutrition and SoilScience vol 164 pp 445ndash450 2001

[16] S Graeff andW Claupein ldquoQuantifying nitrogen status of corn(Zea mays L) in the field by reflectance measurementsrdquoEuropean Journal of Agronomy vol 19 no 4 pp 611ndash618 2003

[17] J K Aase A B Frank and R J Lorenz ldquoRadiometricreflectance measurements of northern great plains rangelandand crestedwheatgrass pasturesrdquo Journal of RangeManagementvol 40 no 4 pp 299ndash302 1987

[18] F StocksTobacco Production in Zimbabwe Nehanda Publishers(Pvt) Harare Zimbabwe 1994

[19] D Magadlela ldquoA smoky affair challenges facing some small-holder burley tobacco producers in Zimbabwerdquo Zambezia vol24 no 1 pp 13ndash30 1987

[20] R D Jackson P N Slater and P J Pinter Jr ldquoDiscriminationof growth and water stress in wheat by various vegetationindices through clear and turbid atmospheresrdquo Remote Sensingof Environment vol 13 no 3 pp 187ndash208 1983

[21] J U H Eitel R F Keefe D S Long A S Davis and L AVierling ldquoActive ground optical remote sensing for improvedmonitoring of seedling stress in nurseriesrdquo Sensors vol 10 no 4pp 2843ndash2850 2010

[22] R Benedetti and P Rossini ldquoOn the use of NDVI profiles asa tool for agricultural statistics the case study of wheat yieldestimate and forecast in Emilia Romagnardquo Remote Sensing ofEnvironment vol 45 no 3 pp 311ndash326 1993

[23] E Svotwa BMaasdorp AMurwira andAMasuka ldquoSelectionof optimum vegetative indices for the assessment of tobaccofloat seedlings response to fertilizer managementrdquo ISRN Agron-omy vol 2012 Article ID 450473 10 pages 2012

[24] Y Curnel and R Oger ldquoAgrophenology indicators from remotesensing state of the artrdquo in ISPRS Archives XXXVI-8W48Workshop Proceedings Remote Sensing Support to Crop YieldForecast and Area Estimates pp 31ndash38 Stresa Italy November-December 2006

[25] C J Tucker ldquoRed and photographic infrared linear combina-tions for monitoring vegetationrdquo Remote Sensing of Environ-ment vol 8 no 2 pp 127ndash150 1979


[26] Tobacco Research Board (TRB) Flue Cured Tobacco Recom-mendations TRB Harare Zimbabwe 2010

[27] U Mazarura and C Chisango ldquoEffects of long term croppingsystems on soil chemical propertiesrdquo Asian Journal of Agricul-ture and Rural Development vol 2 no 4 pp 632ndash640 2012

[28] T A Fasheun and E E Balogun ldquoSome spectral signatures ofgrain amaranthus (Amaranthus cruentus L)rdquo InternationalJournal of Remote Sensing vol 13 no 11 pp 2009ndash2015 1992

[29] G Peng L Deng W Cui T Ming and W Shen ldquoRemotesensing monitoring of tobacco field based on phenologicalcharacteristics and time series image-A case study of chengjiangcounty Yunnan Province ChinardquoChinese Geographical Sciencevol 19 no 2 pp 186ndash193 2009

[30] R CNageswara-Rao J HWilliamsMV K Sivakumar andKD RWadia ldquoEffect of water deficit at different growth phases ofgroundnut II Response to drought during pre-flowering phasrdquoAgronomy Journal vol 80 pp 431ndash438 1988

[31] J M Moore ldquoIrrigation a proper technique guide to irrigatingtobaccordquo 2010 httpwwwtobaccofarmquarterlycom

[32] I Gondola ldquoWeather conditions and nitrogen supply on ripen-ing rate yield and quality of flue-cured tobaccordquo in CORESTACongress Harare Zimbabwe 1994

[33] R Devndra Y S Veeraj M U Kumar and K S K Sastry ldquoLeafarea duration and its relationship to productivity in early ricecultivarsrdquo Proceedings of the National Academy of Sciences IndiaSection B vol B49 no 6 pp 693ndash696 1983

[34] M SaleemMMaqsood andMUl-Hassan ldquoLeaf area durationand total dry matter responses of cotton to integrated plantnutrition and irrigation schedulingrdquo Crop Management vol 8no 1 2009

[35] D Jiang N-B Wang X-H Yang and J-H Wang ldquoStudy onthe interaction between NDVI profile and the growing statusof cropsrdquo Chinese Geographical Science vol 13 no 1 pp 62ndash652003

[36] C Yang J H Everitt and JM Bradford ldquoUsing high resolutionQuickBird satellite imagery for cotton yield estimationrdquo inASAE Annual International Meeting 2004 pp 893ndash904 papernumber 041119 August 2004

Submit your manuscripts athttpwwwhindawicom

Nutrition and Metabolism

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Food ScienceInternational Journal of

Agronomy


International Journal of



Microbiology

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

AgricultureAdvances in


PsycheHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

BiodiversityInternational Journal of


ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GenomicsInternational Journal of


Plant GenomicsInternational Journal of


Biotechnology Research International


Forestry ResearchInternational Journal of


Journal of BotanyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EcologyInternational Journal of


Veterinary Medicine International


Cell BiologyInternational Journal of


Evolutionary BiologyInternational Journal of



using measurements of reflected radiation as a nondestruc-tive method for quantifying pigments as affected by level ofcropmanagement [3] Similar cases of use of radiometric datafor crop assessment and N estimation were reported over awide range of crops [8 9]

The spectral characteristics of vegetation vary with wave-length Analysing vegetation using remotely sensed datarequires knowledge of the structure and function of vege-tation and its reflectance properties [10] Leaf chlorophyllstrongly absorbs radiation in the red and blue wavelengthsbut reflects in the green wavelength [11] The internal struc-ture of healthy leaves acts as a diffuse reflector of near-infraredwavelengthsMeasuring andmonitoring the infraredreflectance can be useful in determining crop health [12]

Well managed crop canopies appear greenest and becomered or yellow with decrease in chlorophyll content due topoor fertilisation [13] Blackmer and Schepers [8] were ableto separate N treatments using the near 550 nm reflectancemeasured on maize leaves detached from a field N fertilizerexperiment and also found that nutrient stressed canopieshad a lower spectral reflectance in the NIR and higherred wavebands when compared with unstressed canopiesOsborne et al [14] and Graeff et al [15 16] could identifyand quantify nutrient deficiencies by means of reflectancemeasurements based on selected stress specific wavelengthranges Aase et al [17] reported a relationship between wheatin-season dry matter and NIRred ratios suggesting thatreflectancemeasurements could be used to estimate plant leafdry matter

The crop health and yield of tobacco depend on goodagronomic practices which include among other thingsapplying the recommended fertilizer rates and selection ofappropriate planting dates [18] Tobacco planting starts fromthe first day of September and to the end of the second weekofDecember in Zimbabwe depending onwhether the farmerintends to irrigate the crop or rely on seasonal rainfall [19]In Zimbabwe the resource poor farmers resort to applyinghalf the recommended fertilizer rates to save on their costs ofproduction [19]These practices have an impact on the healthphysiological status and final yield of the crop

The Normalised Difference Vegetation Index (NDVI) isthe most commonly used spectral index in assessing cropvigor vegetation cover and biomass production from mul-tispectral satellite data [20 21] The index is calculated usingthe formula NDVI = (NIRminusRed)(NIR+Red) As describedby Jackson et al [20] the index is considered a good indicatorof the amount of vegetation and hence useful in distin-guishing vegetation from soil Benedetti and Rossini [22]used the NDVI in assessing wheat growth vigor and appliedthe information in estimating yield Research at Kutsaga hasestablished that the index is positively correlated with mostbiophysical characteristics of flue cured tobacco [23]

Monitoring of crop phenology by remote sensing isvery important for many practical agronomic applicationsand notably for yield forecasting purposes [24] Accurateanalysis of temporal and spatial variations throughout theseason can be used to interpret crop spectral response toagronomic management as emphasized by Tucker et al [25]It is essential to quantify the relationship between spectral

properties of tobacco and agronomic parameters in orderto utilize the full potential of remote sensing for assessingcrop condition and for forecasting yield The use of effectiveremote sensing techniques in crop growth studies wouldeliminate the need for extensive field sampling by giving goodnutrient adequacy detection capability [14]

This paper presents the results of an experiment whichwas conducted to correlate the spectral indices with theagronomic variables of three flue cured tobacco varietiesfour planting dates and three fertilizer management levelsBy measuring the energy that is reflected by a crop canopysurface over a variety of different wavelengths it is hypothe-sized that spectral signatures for the varying fertilizer levelsvarieties and planting dates can be distinguished from theadjacent bare ground It is also hypothesized that from spec-tral characteristics of the canopies of the four planting datesone can identify the most suitable temporal windows formeaningful measurements This information could be veryuseful in estimating the area covered by each planting datetreatment and in large area crop yield and quality predictionsusing remote sensing

2 Method

21 Study Area This experiment was conducted for threeseasons at Kutsaga Research Station near Harare Zimbabwe(17∘551015840S 31∘081015840E Altitude 1480m above sea level [26])during summer and between 2010 and 2012 The experimentwas carried out under irrigation on a sandy loam soil (728sand 88 silt and 184 clay)The long-term average annualrainfall for the station is 882mmBefore 2011 the research areahad been planted to tobacco followed by four years of Chlorisgayana cv Katambora Rhodes grass [27] The range averagemonthly temperature for the station is 8∘C

A split plot design with four planting dates SeptemberOctober November and December as main plots variety assubplot and fertilizer treatments as sub-subplots was used(Table 1) Three tobacco varieties KRK26 T 66 and K E1developed by Kutsaga Research Station were used whilethree fertilizer management levels (50 100 and 150recommended) were applied by hand (Table 2) The recom-mended compound fertilizer rate from soil test results was700 kgha while that for ammonium nitrate (345 N) was96 kgha at 4 weeks after planting and 75 kgha after topping

An initial overhead preirrigation of 50ndash60mm wasapplied at about 60 days before transplanting the tobacco Attransplanting 4-5 liters of water was applied in each plantinghole A 15mm settling in irrigation run was applied just afterplanting followed by a 28-day stress period to promote rootdevelopment after which 25minus30mm is applied to facilitate anitrogenous fertilizer application

The experimental plots were located on well-drainedgranitic sands and were low in available nitrogen mediumin available phosphorus and high in available potassiumcontent throughout the profile During February of 2010 and2011 the plots were disced after a three-year Katambora grassfallow period to incorporate grass Agricultural lime wasapplied using recommendations given from soil test results











(1)





3 Results








SED 003 002 002 002 002 005 006 005 004 005LSD 007 005 003 004 005 010 013 010 008 010

Fertilizer



Variety

K RK26 068 081 082 081 072 057 043 034 035 045T 66 064 081 083 082 074 064 050 035 042 046K E1 062 075 076 077 066 056 035 029 038 057







0

01

02

03

04

05

06

07

08

09

6 11 16 21 26

ND

VI



DecemberGround








SED 003 002 003 002 002 003 008 005 004 003LSD 006 004 006 004 005 006 019 011 009 006

Fertilizer



Variety

KRK26 035 081 077 081 079 073 058 048 068 055T 66 038 083 078 083 081 078 060 054 066 056K E1 040 084 080 084 081 079 062 051 064 056



4 Discussions












SED 003 006 005 004 004 003 003 003 004 003LSD 006 013 011 009 009 007 007 007 008 007

Fertilizer



058 059 056

Variety

KRK26 025 052 063 069 063 062 058 065 065 058T 66 029 053 069 071 066 068 066 061 059 057K E1 027 051 068 068 064 067 061

001 003 035119865-Probability 010 083 012 061 031 003 001 002 002 002

SED 002 003 003 002 002 002 002 004 005 004LSD 004 007 006 005 005 004 004






5 Conclusions









SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer



Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


















Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014


































(1)





3 Results








SED 003 002 002 002 002 005 006 005 004 005LSD 007 005 003 004 005 010 013 010 008 010

Fertilizer



Variety

K RK26 068 081 082 081 072 057 043 034 035 045T 66 064 081 083 082 074 064 050 035 042 046K E1 062 075 076 077 066 056 035 029 038 057







0

01

02

03

04

05

06

07

08

09

6 11 16 21 26

ND

VI



DecemberGround








SED 003 002 003 002 002 003 008 005 004 003LSD 006 004 006 004 005 006 019 011 009 006

Fertilizer



Variety

KRK26 035 081 077 081 079 073 058 048 068 055T 66 038 083 078 083 081 078 060 054 066 056K E1 040 084 080 084 081 079 062 051 064 056



4 Discussions












SED 003 006 005 004 004 003 003 003 004 003LSD 006 013 011 009 009 007 007 007 008 007

Fertilizer



058 059 056

Variety

KRK26 025 052 063 069 063 062 058 065 065 058T 66 029 053 069 071 066 068 066 061 059 057K E1 027 051 068 068 064 067 061

001 003 035119865-Probability 010 083 012 061 031 003 001 002 002 002

SED 002 003 003 002 002 002 002 004 005 004LSD 004 007 006 005 005 004 004






5 Conclusions









SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer



Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


















Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014





























SED 003 002 002 002 002 005 006 005 004 005LSD 007 005 003 004 005 010 013 010 008 010

Fertilizer



Variety

K RK26 068 081 082 081 072 057 043 034 035 045T 66 064 081 083 082 074 064 050 035 042 046K E1 062 075 076 077 066 056 035 029 038 057







0

01

02

03

04

05

06

07

08

09

6 11 16 21 26

ND

VI



DecemberGround








SED 003 002 003 002 002 003 008 005 004 003LSD 006 004 006 004 005 006 019 011 009 006

Fertilizer



Variety

KRK26 035 081 077 081 079 073 058 048 068 055T 66 038 083 078 083 081 078 060 054 066 056K E1 040 084 080 084 081 079 062 051 064 056



4 Discussions












SED 003 006 005 004 004 003 003 003 004 003LSD 006 013 011 009 009 007 007 007 008 007

Fertilizer



058 059 056

Variety

KRK26 025 052 063 069 063 062 058 065 065 058T 66 029 053 069 071 066 068 066 061 059 057K E1 027 051 068 068 064 067 061

001 003 035119865-Probability 010 083 012 061 031 003 001 002 002 002

SED 002 003 003 002 002 002 002 004 005 004LSD 004 007 006 005 005 004 004






5 Conclusions









SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer



Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


















Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014





























SED 003 002 003 002 002 003 008 005 004 003LSD 006 004 006 004 005 006 019 011 009 006

Fertilizer



Variety

KRK26 035 081 077 081 079 073 058 048 068 055T 66 038 083 078 083 081 078 060 054 066 056K E1 040 084 080 084 081 079 062 051 064 056



4 Discussions












SED 003 006 005 004 004 003 003 003 004 003LSD 006 013 011 009 009 007 007 007 008 007

Fertilizer



058 059 056

Variety

KRK26 025 052 063 069 063 062 058 065 065 058T 66 029 053 069 071 066 068 066 061 059 057K E1 027 051 068 068 064 067 061

001 003 035119865-Probability 010 083 012 061 031 003 001 002 002 002

SED 002 003 003 002 002 002 002 004 005 004LSD 004 007 006 005 005 004 004






5 Conclusions









SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer



Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


















Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014





























SED 003 006 005 004 004 003 003 003 004 003LSD 006 013 011 009 009 007 007 007 008 007

Fertilizer



058 059 056

Variety

KRK26 025 052 063 069 063 062 058 065 065 058T 66 029 053 069 071 066 068 066 061 059 057K E1 027 051 068 068 064 067 061

001 003 035119865-Probability 010 083 012 061 031 003 001 002 002 002

SED 002 003 003 002 002 002 002 004 005 004LSD 004 007 006 005 005 004 004






5 Conclusions









SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer



Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


















Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014





























SED 005 007 006 006 006 004 005 003LSD 011 016 012 013 013 009 011 007

Fertilizer



Variety

KRK26 038 051 059 059 057 048 037 038T 66 038 052 061 058 064 051 037 039K E1 033 047 057 056 061 052 032 040


















Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014





























Acknowledgments


References









































Journal of




Agronomy





Microbiology




Volume 2014






































Journal of




Agronomy





Microbiology




Volume 2014
























research article assessing the spectral separability of...

Documents