novel system

7/29/2019 Novel System

1/19

Journal of Environmental Management 82 (2007) 221239

A novel system for environmental monitoring through a cooperative/

synergistic scheme between bioindicators and biosensors

Frank Batzias, Christina G. Siontorou

Department of Industrial Management and Technology, University of Piraeus, Karaoli & Dimitriou 80, 18534 Piraeus, Greece

Received 26 January 2005; received in revised form 22 December 2005; accepted 23 December 2005

Available online 29 March 2006

Abstract

This paper addresses environmental monitoring through a robust dynamic integration between biomonitor and biosensor systems, a

strategy that has not been attempted before. The two systems are conceptually interrelated and methodologically correlated to a

cooperative/synergistic scheme (CSS) with a view to minimise uncertainty and monitoring costs and increase reliability of pollution

control and abatement. The structures and operations of the biosensor component (in terms of sensitivity, device and method versatility,

nature-mimicking physicochemical mechanisms, prospects and technological input) are such that they reinforce or promote the

structures and operations of the natural component (in terms of biosurveillance, impact assessment, environmental quality indexing,

stress responses, metabolic pathways, etc.) and vice versa.

The bioindicator ontology presented herein, including concepts, relations and controlled vocabulary aiming at establishing an

integrated methodology for mapping/assessing negative environmental externalities, provides a useful tool for the design/development/

implementation of an environmental network for the monitoring of a variety of pollutants over time and space and the assessment of

environmental quality; the collection of the available information and its classification into taxonomic and partonomic relations allows

the construction of a database that links pollutants with organisms response, at a phenomenological and in-depth level, considering

ecological parameters, relations and geomorphologic characteristics. As a result, a computer program has been designed/developed as a

decision support system and has been successfully tested on a representative population of species indigenous to southern Greece.

Significantly, a novel system in the form of a rational framework at the conceptual design level has been developed, that actually

contributes towards achieving a cost-effective long-term monitoring program, with the flexibility to counter on-course any (anticipated

or not) variations/modifications of the surveillance environment. This novel and pioneering approach will further offer a dynamic system

utilised in (a) environmental impact studies and risk assessment (positive/analytic approach), (b) decision-making in the short-run

(normative/tactic approach), and (c) policymaking in the long-run (normative/strategic approach). The proposed CSS, based on the

integration of multiple data sources, can establish a local area network, incorporated into/expanding to a wide area network, thus

offering the potential of better predictive ability and greater lead-time warning at alarm conditions than that provided by separate, stand-

alone surveillance modalities.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Environmental monitoring; Decision support system; Bioindicators; Biosensors; Ontology; Biosurveillance

1. Introduction

The escalating public concern about atmospheric and

water pollution has prompted efforts to establish control

programmes in many countries. In practice, controlling

(anthropogenic) pollutants is a very complex problem:

sources and emissions have to be identified, analytical

methods have to be evaluated, risks have to be assessed,

critical emissions have to be controlled, and economical

aspects have to be integrated (Caughlan and Oakley, 2001;

Srensen et al., 2003). Successful monitoring programs

must be ecologically relevant, statistically credible, and

cost-effective (Hinds, 1984); programs that neglect any one

of these critical areas will face problems and likely fail.

The decision support systems developed to guide local

authorities in quality management can be based on

ARTICLE IN PRESS

www.elsevier.com/locate/jenvman

0301-4797/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.

doi:10.1016/j.jenvman.2005.12.023

Corresponding author. Tel.: +301 210 414 2369;

fax: +301 210414 2366.

E-mail addresses: [email protected] (F. Batzias), [email protected]

(C.G. Siontorou).
http://www.elsevier.com/locate/jenvmanhttp://www.elsevier.com/locate/jenvman


2/19

dispersion modelling (source-orientation, a priori known

emission sources) and field measurements of the emission

(receptor/effect- orientation). As technical field measure-

ments, involving equipment and manpower, are generally

associated with high costs and prolonged analysis time,

dispersion modelling is mostly preferred (Overton et al.,

1995; Argent, 2004). Field measurements, however, shouldbe regarded as necessary and indispensable: they may be

used to validate dispersion models, and the data obtained

may indicate the presence of sources not known or

registered (Wolterbeek, 2002; Uhlenbrook and Sieber,

2005). Although the field of environmental instrumental

analysis has reached a stage of development that is

challenging and promising, owing to the advances in

extraction procedures, microfluidics, chip design, detection,

engineering, and software, the continuous, long-term field

monitoring of a broad range of chemicals simultaneously

has not yet been realised, due to the lack of sufficiently

sensitive and inexpensive devices (Guiochon and Beaver,

2004). In this context, biosensors appear as suitable

alternative or complementary analytical tools.

Biosensors, utilising the selectivity and sensitivity of

biological components (such as whole cells, organelles or

biomolecules) coupled with signal transducers, have the

potential to overcome most of the disadvantages of the

conventional methods by virtue of their high specificity,

fast response times, low cost, ease of use, and continuous

real-time signal (Kress-Rogers, 1997; Malhorta et al.,

2005). These devices can detect individual substances or

groups of substances of environmental interest (e.g.,

industrial emissions, pesticides, insecticides), as well as

biological effects (such as genotoxicity, immunotoxicity,and endocrine responses) (Rogers, 1995; Siontorou et al.,

1997a; OSullivan and Alcock, 1999; Siontorou et al., 2000;

Chiti et al., 2001; Castillo, et al., 2004; Tschmelak et al.,

2005; Rodriguez-Mozaz et al., 2005). The universal

applicability of biosensors places them clearly at the core

of any programme to address future technology for

environmental monitoring in real matrices.

Combining physical components of a sensor platform

with highly specific biological recognition elements may

allow biosensors to respond to both regulatory and

practical requirements (Rogers, 1995; OSullivan and

Alcock, 1999; Castillo et al., 2004); however, commercial

development of biosensors for environmental applications

is not a trivial matter. Although biosensor technology has

indeed the potential for a major impact in the environ-

mental testing market, the commercialisation and utilisa-

tion of these devices for in situ monitoring will have to wait

until microfabrication and nanotechnology engineering

become more accessible and less expensive (Rogers, 1995;

Weetall, 1996; Dannemand Andersen et al., 2004). It is here

that biomonitoring comes in.

The environmental impact of atmospheric deposition has

been extensively studied in many heavily polluted areas, mainly

on lichens and mosses (Bargagli, 1989, 1998; Nimis et al., 1990;

Henderson, 1996; Kirschbaum and Hanewald, 1998;

Conti and Cecchetti, 2001; Bargagli et al., 2002; Carreras

and Pignata, 2002; Figueira et al., 2002). Although many

changes in vegetation are now generally attributed to

atmospheric deposition, doseeffect relationships are

usually poorly known (Heij et al., 1991; Wolterbeek, 2002).

This paper addresses environmental monitoring through

a robust dynamic integration between biomonitor andbiosensor systems, a strategy that has not been attempted

before; the two systems are conceptually interrelated and

methodologically correlated to a cooperative/synergistic

scheme (CSS) with a view to minimise uncertainty and

monitoring costs and increase reliability of pollution

control and abatement. Attention is given to establishing

a fully functioning and reliable network approach for

monitoring all pollutant species (primary and secondary)

and achieving doseresponse relations and calibration of

biomarker vegetation by means of the nature-mimicking

biosensor devices, aiming eventually at the shifting,

partially or totally, from instrumental to natural monitor-

ing. The structures and operations of the biosensor

component (in terms of sensitivity, device and method

versatility, physicochemical mechanisms, prospects and

technological input) are such that they reinforce or

promote the structures and operations of the natural

component (in terms of biosurveillance, impact assessment,

environmental quality indexing, stress responses, metabolic

pathways, etc.) and vice versa. The network initially

consists of both biosensors and bioindicators, as part of

a validation phase, which, by removing most of the former

devices, gives rise to a hybrid-monitoring scheme, engaging

mostly bioindicators and a few biosensors (Fig. 1).

By means of bioindicator recalibration with periodicrevisiting of the biosensors, the scheme progressively

reaches a steady state, thus ensuring reliability and

robustness. This novel and pioneering approach will

further offer a cost-effective dynamic system utilised in

(a) environmental impact studies and risk assessment

(positive/analytic approach), (b) decision-making in the

short-run (normative/tactic approach), and (c) policymak-

ing in the long-run (normative/strategic approach).

Based on this concept, this paper presents in Section 2 a

brief review of the current practices in biological and

biosensor-based environmental monitoring, highlighting

ARTICLE IN PRESS

biosensing

biomonitoring

biosensing

START

START

START

END

END

END

biomonitoring

biomonitoring

Start-Up Cycle Steady-State Cycle

biomonitoring

biosensing

biosensing

validation validation

environmental monitoring environmental monitoring

START

END

Fig. 1. The cyclic operation of the cooperative/synergistic system

presented herein; the size of the arrows indicates the relative importance

between biosensing and biomonitoring.

F. Batzias, C.G. Siontorou / Journal of Environmental Management 82 (2007) 221239222


3/19

the unresolved applicability problems and the unattained

critical parameters that hinder further exploitation of these

systems. Section 3 attends to the bioindicator issues,

introducing two novel concepts: (a) the construction of a

bioindicator ontology aimed at establishing an integrated

methodology for mapping/assessing negative environmen-

tal externalities, which by means of taxonomic/partonomicrelations links environmental pollutants to the stress

response of the vegetation species, at both the phenomen-

ological and the cellular/ biochemical level, provides a

pattern (connoting a guide not a prototype or model)

under the form of a group of species, appropriate to

monitor in the long-term specific pollutant(s) in the desired

concentration range, and (b) a methodological framework

for calibrating/validating biomonitoring organisms utilis-

ing biosensor technology. The latter also provides suitable

correlations and proper modelling between ecological

impact and biosensor signal, since interpretation relies on

similar physicochemical bases. The novel concepts pre-

sented and resolutions proposed by the authors are

integrated into a monitoring scheme intended for multi-

elemental monitoring, through a methodological frame-

work introduced in Section 4 and discussed in Section 5.

Throughout the paper, emphasis is put on lichens and

mosses as bioindicators, since they are abundant at areas

around potential pollution sources, whereas the variety of

responses of different species towards pollution enables

detailed patterns to be obtained even at low levels of

pollution: the response of a range of species can be

monitored through separate temporal and/or spatial

Geographical Information System (GIS) layers and subse-

quently correlated by means of a relational database, sothat environmental impact can be recorded either system-

atically or as function of time and/or space (Conti and

Cecchetti, 2001; Bargagli et al., 2002; Carreras and Pignata,

2002; Wolterbeek, 2002; Canto n et al., 2004).

The proposed biosurveillance CSS, based on the

integration of multiple data sources, can establish a Local

Area Network (LAN), incorporated/expanding into a

Wide Area Network (WAN), thus offering the potential

of better predictive ability and greater lead-time warning at

alarm conditions than that provided by separate, stand-

alone surveillance modalities.

2. The current practice

2.1. Biological monitoring

The use of living organisms as indicators for environ-

mental stability has long been widely recognised. Plants,

animals, fungi, and bacteria have been employed as

bioindicators and biomonitors in air, soil and water

pollution surveys over the past few decades (Bargagli,

1989, 1998; Nimis et al., 1990; Heij et al., 1991; Henderson,

1996; Kirschbaum and Hanewald, 1998; Conti and

Cecchetti, 2001; Beeby, 2001; Bargagli et al., 2002; Carreras

and Pignata, 2002; Figueira et al., 2002; Wolterbeek, 2002).

In general, biomonitoring is based on sensitive or

accumulative organisms (Beeby, 2001; Wolterbeek, 2002).

The former may be of the optical type, based on

morphological changes in abundance behaviour related to

the environment and/or upon chemical and physical

aspects, as photosynthetic or respiratory activity modifica-

tions. The accumulative species have the ability to storecontaminants in their tissues; bioaccumulation is the result

of the equilibrium process of biota compound intake/

discharge from and into the surrounding environment.

Lichens are one of the most valuable long-term

biomonitors of atmospheric pollution (Henderson, 1996;

Bargagli, 1989; Beeby, 2001; Conti and Cecchetti, 2001;

Bargagli et al., 2002): they can be used as sensitive

indicators to estimate the biological effects of pollutants,

by measuring changes at the community or population

levels, and as accumulative monitors of persistent pollu-

tants, by assaying trace element content. During the last

15 years, lichen biomonitoring studies showed the re-

appearance of lichens in areas previously devoid of these

organisms (lichen desert) and the improvement of lichen

biodiversity in many urban and industrial areas (Hawks-

worth and McManus, 1989; Seaward and Letrouit-

Galinou, 1991; Loppi et al., 1998; Loppi et al., 2004).

The recolonisation process is of great lichenological

interest, but its evaluation is possible only when old data

or periodic observations are available, allowing compar-

ison of the results at different times. Lichen colonisation

has been attributed to declining SO2 concentrations, and

the diminished lichen biodiversity to the constantly high

levels of NOx (Loppi et al., 2004). In addition to floristic

changes, variations in lichen trace element contents in timecan provide useful evidence for trends in ambient pollution

burdens (Loppi et al., 1998). Because the concentrations of

trace elements in lichen thalli are directly correlated with

environmental levels of these elements (Wolterbeek, 2002),

lichens are very useful for monitoring not only spatial

patterns but also temporal trends of trace element

deposition (Zschau et al., 2003).

Accounts of such applications have been extensively

published, but all too frequently the information is widely

scattered, often in obscure or inaccessible literature

sources, and lacks synthesis (Seaward, 1995). In theory,

the techniques developed could be employed for low

technology environmental monitoring where comparable

on-site instrumentation would be expensive to install and

maintain. Unfortunately, most of the techniques based on

bioaccumulation require sophisticated analytical equip-

ment, together with a fairly detailed understanding of the

taxonomy of one or more groups of organisms. Moreover,

cumulative effects cannot be used for short-term decision-

making and compensation.

Moss transplants have often been used as biomonitors

because of the absence of native species in a study area

(Szczepaniak and Biziuk, 2003): waste incinerators (Carpi

et al., 1994), chlor-alkali plants (Ferna ndez et al., 2000),

urban areas (Vasconcelos and Tavares, 1998), roadsides

ARTICLE IN PRESS

F. Batzias, C.G. Siontorou / Journal of Environmental Management 82 (2007) 221239 223


4/19

(Viskari et al., 1997), and rural sites (Evans and Hutch-

inson, 1996). In the recent years, considerably more studies

have utilised indigenous flora than transplants, and this is

one of the reasons for the difficulties associated with the

interpretation of the results obtained in the latter kind of

study. These difficulties are fundamentally associated with

the controls; the results are expressed as either netenrichment or enrichment factors and therefore interpreta-

tion of the results is made on the basis of the corresponding

control concentrations. The controls comprise part of the

moss contained in the moss bags, but are not exposed to

contamination and are considered as blanks (Vasconcelos

and Tavares, 1998); however, these blanks are also subject

to different sources of variability.

Biological monitoring in the context of an integrated

environmental monitoring program still encompasses the

qualitative or semi-quantitative nature of measurement,

lacking validation, calibration, standardisation, and har-

monised guidelines. Transplants may be used to assess the

current environmental status provided that the environ-

ment baseline is known or can be estimated (from previous

studies) and the bioavailability curves and correlation

factors are validated (Szczepaniak and Biziuk, 2003;

Ferna ndez et al., 2000); since such studies are relatively

recent, assessment is based on ad hoc produced environ-

mental indices (e.g. atmospheric purity index) that cannot

provide the degree of sensitivity and reliability required for

decision-making. On the other hand, indigenous plants can

be used for long-term monitoring, bearing however a major

disadvantage: all processes and all sources act at the same

time and there is no possibility of separating them or

looking for a particular one (C eburnis and Valiulis, 1999).Although lichen biogeography, physiology, biological

cycle, ontogeny, chemotaxonomy and biochemistry have

been extensively studied, the fate of the adsorbed pollutant

in the symbiont and its effect on the metabolism of the

bacterium and the fungus have not been clarified, while

studies on the effect of two or more pollutants acting

simultaneously on the same organism are scarce. Further-

more, calibration relies on the estimation of sensitivity

indices, performed in the laboratory on a single-pollutant

basis, without considering in situ species tolerance and

developed protective mechanisms, combined contamina-

tion, bioavailability variations or synergistic/antagonistic

effects of the surroundings (adhering bark, mosses, other

lichen species, soil particles, etc.). Also, considering the

wide genetic and geographic variation of the symbionts, a

complete, in-depth study of one species and its habitat will

most probably not be applicable to another species in a

distant habitat.

2.2. Environmental biosensors

Biosensors are analytical devices incorporating a biolo-

gical material, a biologically derived material or a

biomimic in intimate contact with a physicochemical

transducer or a transducing microsystem (The venot et al.,

1999). The principle of detection involves the specific

binding of the target analyte to the complementary

biorecognition material; the specific interaction results in

a change in one or more physicochemical properties (pH

change, electron transfer, mass change, heat transfer,

uptake or release of gases/ions), detected and measured

by the transducer and converted to an electronic signal,which is a function of the concentration of the analyte,

allowing for both quantitative and qualitative measure-

ments in real time. In theory, and verified to a certain

extent in the literature, any biological sensing element may

be paired with a physical transducer, and any target

analyte or group of analytes can be detected with the

desired sensitivity and selectivity.

Emerging from Clarks enzyme electrode in 1962,

progress in biosensors was marked by (a) advances in

biological surface science regarding the attachment of the

biomaterial on the transducer, starting from entrapment

using a dialysis membrane, moving to covalent fixation and

reaching direct immobilisation and integration, (b) trans-

ducer development, from electrodes to ultraminiaturised

chips and from fluorescence to surface plasmon resonance

and (c) advances in material science, expanding the range

of bioreceptors from enzymes and whole cells to immuno-

systems and nucleic acids and from natural elements to

semi-synthetic and recently fully synthetic mimics (Rogers,

1995; Kress-Rogers, 1997; Malhorta et al., 2005; Castillo

et al., 2004).

Commercialisation and success came from clinical

applications (eg. glucose meters), and research is still

focusing on medical devices, although environmental issues

have also been considered. Because biosensor technologylends itself to fast, economical and continuous monitoring

capabilities, development of these systems to complement

conventional analytical measurements is expected to result

in a substantial cost benefit, especially when sample

turnaround time and cost per analysis are important issues

(Rogers, 1995; OSullivan and Alcock, 1999; Malhorta

et al., 2005; Rodriguez-Mozaz et al., 2005).

Notwithstanding the high technological input, the

majority of environmental biosensors is only just starting

to move from the proof-of-concept stage to field-testing,

mass production and commercialisation (World Biosensor

Markets, 1997; OSullivan and Alcock, 1999). The regu-

lated and emerging contaminants that can currently be

detected by biosensors include heavy metals (Ivask et al.,

2002), PCBs and dioxins (Shimomura et al., 2001), phenols

(Parellada et al., 1998), surfactants (Nomura et al., 1998),

PAHs (Koening et al., 1997), nitrates (Glazier et al., 1998)

and gases (Hart et al., 2002; Gil et al., 2002), endocrine

disrupters (Rodriguez-Mozaz et al., 2004; Gobi et al.,

2004), antibiotics (Patel, 2002) biological threat agents

(Iqbal et al., 2000), and pesticides (Tschmelak et al., 2005).

In addition to specific chemical analysis, fast determination

of quality parameters, such as BOD, and assessment of

biological contamination by pathogenic organisms, are

currently possible by biosensor-based methods (Liu and

ARTICLE IN PRESS



5/19

Mattiasson, 2002; Koubova et al., 2001). Other commercial

biosensors for environmental applications include the

versatile BIACOREs (Bioacore AB, Uppsala, Sweden),

and Spreetas (Texas Instruments Inc., Dallas, USA).

For environmental pollution risk assessment, the inte-

gration of both chemical and effect-related analyses

(toxicity, endocrine disruption activity, etc.) is essential.Many efforts have been made during the last years to

develop biosensors for toxicity evaluation of water

samples. As a result, various biosensors based on whole

organisms have been commercialised, such as Cellsenses

(Euroclon Ltd., Yorkshire, UK), an amperometric biosen-

sor which incorporates Escherichia coli bacterial cells for

rapid ecotoxicity analysis (Farre and Barcelo , 2003).

The wide gap between research and marketplace is

mainly due to stability issues and quality assurance

(Weetall, 1996; World Biosensor Markets, 1997); biosen-

sors are custom-designed to accommodate the performance

specifications and constraints of the biological macromo-

lecules, which are not particularly rugged, especially after

their immobilisation on the transducer (Rogers, 1995;

Weetall, 1996; Yang et al., 2003; Castillo et al., 2004). Used

in real applications, i.e., in a continuously changing

environment and under conditions of uncertainty, perfor-

mance faults are anticipated, most commonly due to

fouling or device ageing. The latter is often one of the

deciding factors as to the commercial viability of these

sensor devices (Rogers, 1995; OSullivan and Alcock, 1999;

Dannemand Andersen et al., 2004; Malhorta et al., 2005;

Rodriguez-Mozaz et al., 2005): the immobilised proteinac-

eous molecules are subject to activity loss due to

denaturation and/or deactivation, thus diminishing the lifeof the sensor. The operational stability of the device is

further influenced by the stability of the support between

the biomaterial and the transducer, as well as the stability

of the transducer; the reported shelf life of an algal

biosensor for volatile organic compounds is 1 month

(Podola and Melkonian, 2003) whereas the stability of a

hydrogenase biosensor is only two weeks (Qian et al.,

2002).

Performing within an integrated monitoring system

under these conditions, biosensors should be accompanied

by an intelligent system for fault detection and fault

compensation, as well as a suitable replacement/mainte-

nance program. This is one area that has not received

sufficient attention. An account of the causes of biosensors

signal-to-noise ratio increase (expressed as a noisy or a

dead sensor, commonly encountered during operation) has

been recently attempted by the authors (Batzias and

Siontorou, 2005). A methodological framework for devel-

oping a suitable replacement and maintenance program,

incorporating the biosensor lifetime and the stress level

expected to endure during operation, has been also

presented (Batzias and Siontorou, 2004), aiming at the

techno-economical optimisation of the system for long-

term field monitoring. The only well studied, designed and

tested, integrated field environmental biosensor reported is

the Automated Water Analyser Computer Supported

System (AWACSS), a multi-analyte immunoassay-based

system, using intelligent remote surveillance and control

that allows unattended continuous monitoring of several

organic pollutants simultaneously (Tschmelak et al., 2004).

Needless to say that besides the lack of technical support

and long-term management considerations of biosensornetworks, a serious barrier that has to be removed is the

cost of the device. Although the development at a bench-

scale is indeed cheap, industrial-scale production is still

very expensive, as it will take time to achieve economies of

scale and to develop the most efficient fabrication methods.

The devices developed are usually generic and extremely

versatile, since one transducer system can be linked to a

variety of bioelements for the detection of the correspond-

ing target analytes; for example, a lipid-based biosensor

can be used (a) as is, for the detection of triazine herbicides

(Siontorou et al., 1997a), (b) modified with haemoglobin or

methaemoglobin for the detection of CO2 (Siontorou et al.,

1997b) or cyanide ions (Siontorou and Nikolelis, 1996),

respectively, and (c) incorporated with DNA for the

detection of hydrazines (Siontorou et al., 1998). However,

most biosensors utilise precious metal electrodes, whereas

their construction usually involves several steps of con-

siderable complexity and precise methods of immobilisa-

tion of the biological element, which at mass production

will increase the cost of the product. In addition,

harmonised protocols should be established for field-

testing, evaluation and validation, if the new methodology

is to be approved for regulated applications, such as

control of environmental pollutants (Rogers, 1995;

Weetall, 1996).

3. The principles of CSS at the design level

Biomonitoring, in a general sense, may be defined as the

use of bio-organisms/materials to obtain (quantitative)

information on certain characteristics of the biosphere. In

that context both indicator/monitor organisms and bio-

sensors fit the definition. However, in developing biomo-

nitoring into a fully accepted quantitative tool, the

quantification problem of the monitor organisms should

be resolved, whereas the great potential and advanced

technology of biosensors should be put into practical use.

Unquestionably, the impact of pollution (short- or long-

term) on living organisms is the most comprehensive

indicator of the degree of ecosystem damage and its

correlation (direct or indirect) to human health. The

relative ease of sampling, the absence of any need for

complicated and expensive technical equipment, and the

accumulative and time-integrative behaviour of the organ-

isms make them the less costly means of environmental

surveillance. In effect, all the monitors needed have been

long installed and are still fully functional in measuring,

assessing and storing information, and all that is needed is

the ability to read/comprehend the information.

ARTICLE IN PRESS



6/19

3.1. Bioindicator ontology on environmental exposure

Seen from an ontological point of view, the relation

between exposure and response (phenomenological at surface

level and physicochemical at deep level) should be elucidated.

Consequently, the construction of an ontology, linking multi-

directionally the effect of the pollutant (regulated orsuspected) species, alone and in combination, upon the

monitor organisms is essential. In knowledge management,

ontology is an explicit formal specification of how to

represent the object, concepts and other entities that are

assumed to exist in some area of interest and the relation-

ships that hold among them. Ontology resembles faceted

taxonomy but use richer semantic relationships among terms

and attributes, as well as strict rules about how to identify/

specify/analyse/synthesise terms and relationships (Neches,

et al., 1991; Sullivan et al., 2004). Since ontologies do more

than just control a vocabulary, they are thought of as

knowledge representation, suitable to provide information

for (a) decision making in complex interdisciplinary

domains (like surveillance technology and ecosystems) and

(b) problem solving within these domains. Exploiting the

alteration of the physicochemical/metabolic pathways of the

bioindicators due to each pollutant species and their possible

combinations, the gap between external characteristics,

bioaccumulation and bioavailability will be bridged.

Although some of these relations have been studied in

vitro and in vivo even to gene level with respect to single

exposure (one pollutant), the available information is

scattered and highly unorganised (Seaward, 1995; Bargagli,

1998; Conti and Cecchetti, 2001; Wolterbeek, 2002;

Srensen et al., 2003; Sullivan et al., 2004). Among themost studied, in number and extent, is the exposure of

lichens to sulphur dioxide or nitrogen oxides, providing

doseresponse relations, sensitivity indices, bioavailability

correlations, bioaccumulation levels, and metabolic/ bio-

chemical pathways of infestation (Seaward, 1995; Loppi

et al., 1998; Beeby, 2001; Conti and Cecchetti, 2001;

Wolterbeek, 2002; Zschau et al., 2003; Szczepaniak and

Biziuk, 2003); however, the effect of the two pollutants in

combination has only been studied at a phenomenological

level (morphological characteristics). Various other infor-

mation can be retrieved from species classification studies,

as for example the constraints affecting the association of

bacteria with fungi at the cellular level (Boissiere et al.,

1987; Seaward and Letrouit-Galinou, 1991), or behaviour

studies (Loppi et al., 1998; Conti and Cecchetti, 2001;

Wolterbeek, 2002; Bargagli et al., 2002; Loppi et al., 2004).

Also, toxicological studies reveal the genetic variations

caused by long-term exposure (Bargagli, 1989, 1998; Heij

et al., 1991; Szczepaniak and Biziuk, 2003).

The collection of the available information and its

classification into taxonomic and partonomic relations

would provide a database linking pollutants with response

and sensitivity indices, bioavailability correlations, bioac-

cumulation levels, biochemical effects, approximate inhibi-

tion patterns in the presence of other substances, etc., all

considered with respect to seasonal variation, species

variability, and ecosystem parameters (climatic conditions,

nutrient availability, background pollution level, antago-

nistic/synagonistic relations of species, geomorphologic

characteristics). The authors have constructed a lichens

ontology based on exposure using a relational database

management system (Fig. 2). The first domain containsinformation on the biochemical effect of a pollutant (a) on

the phytobiont (primary attack point), (b) on the transpor-

tation of materials to the fungus (secondary attack point),

and (c) on the fungus (tertiary attack point), taking into

account spot characteristics (eg. bark pH and nutrient

availability) and locale parameters (eg. climatic conditions),

linking each level to morphological transformations (eg.

colour change, percentage coverage), behaviour alterations

(eg. species diversity, antagonistic/synagonistic relations),

and tolerance limits (eg. non-toxic, subtoxic and hypertoxic

ambient concentrations of the pollutant). The combinator-

ial biochemical effect of several pollutants is provided from

the second domain; the information used derives from in

vivo or in vitro studies, in the absence of which possible

pathways derive from models with suitable approximations,

either existing or ad hoc produced utilising the data from

the first domain. The first two domains are connected to the

existing lichens gene libraries in order to provide more in-

depth correlations/associations.

The third domain relates all the information to

geographic zones and seasons (GIS-mapping), considering

lichen taxonomy (species, families and classes) and

ecosystem (lichen coverage and diversity). The fourth

domain contains the validation parameters of the bioindi-

cator output (sensitivity indices and pollutant valuesgiven by the sentinel organisms), with respect to ambient

pollutant concentration, predicted and observed lichen

response, as well as observed response and lab analysis for

accumulation (bioavailability correlations); this platform

further provides comparisons of biomonitoring data from

different geographic areas and species.

As the first two domains refer to the biochemical/cellular

basis of pollution impact, the semantic links are mainly of

the is-a or has-a type; as the last two domains refer to the

systematic response within the ecosystem, the semantic

links is-part-of-a and means-of-a (is-expressed/examined-

by) dominate. The ontology can be used to retrieve species

specific for one or several pollutants, ranked by their

sensitivity indices and their availability in the area of

interest, or otherwise indicate suitable species for trans-

plantation. It also provides information on the dose

response relations, type of response (visual, lab determina-

tion) and variation, expected biochemical mechanisms,

validation parameters, and frequency of sampling.

Although highly promising, the bioindicator ontology is

currently condemned to limited use owing to lack of

necessary information. The morphological and/or genetic

differences between different species in the same locale or

the same species in different locales affect significantly the

response towards the same pollutant, however most studies

ARTICLE IN PRESS



7/19

have a local character. Extensive biomonitoring data

comparisons are not available and biochemical mechan-

isms are still not well understood, seasonal variation and

species variability are not considered in depth, not to

mention that harmonised monitoring and validation

protocols do not exist. Further research is required,

focused mainly on the cellular and subcellular response

to stress, especially due to combination of pollutants, and

guidelines should be agreed upon so that comparisons and

temporal/spatial analyses could be possible.

3.2. Bioindicator calibration using biosensors

When biosensors entered the scientific field they were

characterised as nature-mimicking sensors, implying their

relation to the natural sensing mechanisms, eg. the

olfactory system (Krull and Thompson, 1985); a lot of

research has been in fact focused on creating artificial

organs (artificial pancreas, electronic nose, etc.) (Kress-

Rogers, 1997). In view of that and considering that

biosensors can be tailor made to accommodate any

biological element available (or many bioelements con-

nected in series or in parallel) for detecting any target

analyte(s) at the desired level of sensitivity or selectivity,

the authors argue that they can be utilised in mimicking

bioindicators for calibration/validation purposes.

Although the cost for the mass production of biosensors is

presently quite large, the development of these devices at

bench-scale is indeed cheap, due to the inherent versatility of

biosensor technology. Employing one type of transducer

linked to a data analysis system, one can construct as many

biosensors as the number of bioelements that can be

immobilised on the surface the transducer (for example,

Siontorou and Nikolelis, 1996; Siontorou et al., 1997a,b,

1998). A complete study of any biosensor system for research

purposes, involving signal optimisation, selectivity/sensitivity

studies, calibration, mechanism of signal generation, inter-

ference and real sample studies, can be concluded within a

few months, or in the case of similar devices, within a few

weeks. Also, incorporating one bioelement into different set-

ups or transducers, different types of response can be

obtained for the same target analyte (cumulative, concentra-

tion-dependant, threshold-triggered, etc.) or different dis-

plays for the same response (transient signal, step-increase/

decrease, etc.) (for example, Siontorou et al., 1998, 2000).

Moreover, studies for cascade reactions (bioelements in

series) detecting a singular response for one analyte, or

simultaneous reactions (bioelements in parallel) sorting many

responses for different levels of one analyte or for a number

ARTICLE IN PRESS

GIS mappingValidation

Genetic taxonomy

combinatorial biochemical effects of several pollutantsbiochemical effects of a pollutant

on the fungus

on the transportation of

materials to the fungus

on the phytobiont

environmental

parameters

symbiotic conditions

models

in vitro studies

in vivo studies

alteration of

metabolic pathways inhibition

synergy

alteration of

bioavaialability

seasonal variation

topography

lichen

taxonomy

ecosystem

geographic zones

inter-/intra-species

comparisons

bioaccumulation

sensitivity indices

geographic

response variation

predicted/observed

species response

to pollutants

gene-based

nutrient-basedWeb-databases

Lichen gene libraries

Decision on optimal selection of biomonitoring

species with respect to type an d concentration

ranges of primary and secondary (suspected)

pollutants

Fig. 2. Flow diagram of the bioindicator ontology, designed for optimal decision making in environmental surveillance. The upper layers are based on

taxonomic relations, whereas the lower layers on partonomic.



8/19

of analytes, are feasible (Rogers, 1995; Kress-Rogers, 1997;

Castillo et al., 2004; Rodriguez-Mozaz et al., 2004, 2005;

Tschmelak et al., 2005; Malhorta et al., 2005).

The bioindicator stress response can be directly linked to

cellular and subcellular changes due to biochemical

alterations (Bargagli, 1998; C eburnis and Valiulis, 1999;

Conti and Cecchetti, 2001; Wolterbeek, 2002; Srensen

et al., 2003; Szczepaniak and Biziuk, 2003; Loppi et al.,

2004). For example the response of lichens to sulphur

dioxide stress is presented in Fig. 3a; the sulphate ion

participates in the redox reactions of the cytochrome P

system of the phytobiont. A similar mechanism is utilised

for the detection of SO2 from a screen-printed gas phase

amperometric biosensor (Fig. 3b).

The algorithmic procedure especially designed and

developed by the authors for lichen calibration/valida-

tion, includes the stages shown below, considering one

pollutant substance. Fig. 4 illustrates the connection

of stages, represented by the corresponding number or

letter, used to designate activity or decision nodes,

respectively.

1. Decomposition of the bioindicator to biochemical

systems, relevant to pollution impact.

2. Determination of the environmental parameters/vari-

ables and their limits (pH, ionic strength, temperature,

etc.).

3. Selection of transducer, set-up, and data analysis

system, suitable for the biochemical systems of the

bioindicator and concentration-dependant response.

4. Selection of the best method for immobilisation of

each bioelement on the surface of the transducer

selected at (3).

5. Immobilisation of the bioelements on the transducers

selected at (3).

P. Is immobilisation successful?

6. Set-up of the concentration-dependant devices.

7. Preliminary experiments to check functionality within

the required conditions.

Q. Is the device functional under these conditions?

8. Determination of device response to various concen-

trations of the pollutant.

ARTICLE IN PRESS

Fig. 3. (a) Possible mechanism for non-visible injury and SO2 detoxification (often in response to low pollution levels): The algal symbiot (phytobiont)

hydrolyzes SO2 to H2SO3 and HSO

3 , which are converted to SO24 in the fungal symbiot (mycobiont). The process generates superoxide radicals which

trigger the lichen antioxidant system: superoxide dimutase (SOD) catalyses the conversion of the radicals to hyperoxide, which is converted to water and

oxygen by peroxidase (POD) and/or reduced glutathione (GSH). The latter is produced from glutathione disulfite (GSSG); the NADPH required for the

production of GSH is covered by the oxidative pentose phosphate cycle, during the conversion of glucose-6-phosphate (G6P) to 6-P-gluconate, as well as

during the degradation of mannitol, through the action of mannitol dehydrogenase (MDH) and mannitol-1-phosphate dehydrogenase (M1PDH). The

utilisation of the mannitol cycle consumes the sugar alcohols of the phytobiont. (b) Schematic of a gas phase amperometric biosensor incorporating sulfite

oxidase as the biorecognition element and cytochrome c as mediator.



9/19

9. Correlation of the device response to organism

response determined at (28).

R. Is the correlation satisfactory?

S. Is the response concentration-dependant?

10. Calibration of the biosensor.

11. Selection of model for biosensor/bioindicator correla-

tion, using the bioavailability parameters determined

at (25) and the time exposure determined at (30).

12. Correlation of biosensor calibration to bioindicator

calibration.

ARTICLE IN PRESS

Fig. 4. The algorithmic procedure designed/developed for the calibration/validation of bioindicators by means of biosensors.



10/19

U. Is there another model?

13. Production of a model ad hoc.

T. Is correlation satisfactory?

14. Investigation of mechanism of signal generation.

V. Is sensitivity adequate?

15. Correlation of bioindicator morphological character-

istics to biosensor response on increasing concentra-

tions of the pollutant determined at (29).

W. Is correlation suitable for monitoring purposes?

X. Is there another bioelement available?

16. Implementation of interference studies.17. Selection of transducer, set-up, and data analysis

system, suitable for the biochemical systems of the

bioindicator and cumulative response.

18. Selection of the best method for immobilisation of

each bioelement on the transducer surface selected at

(17).

19. Immobilisation of the bioelements on the transducers

selected at (17).

P. Is immobilisation successful?

20. Set-up of the cumulative-response device.21. Preliminary experiments to check functionality within

the required conditions.

Q. Is the device functional under these conditions?

22. Determination of device response to increasing con-

centrations of the pollutant determined at (29).

23. Selection of a model for biosensor/bioindicator

response based on the bioaccumulation levels deter-

mined at (31) and tolerance limits determined at (30).

24. Correlation of the device cumulative response to

organism bioaccumulation determined at (31).

R. Is the correlation satisfactory?

U. Is there another model?

25. Estimation of bioavailability parameters.

26. Production of a bioavailability correlation model ad

hoc.

27. Investigation of the physiological response of the

organism to pollutant under simulated environment.

Y. Is the response expressed as morphological altera-

tions?

28. Investigation of the response of the organism in

various concentrations of the pollutant.

29. Investigation of the response of the organism in

increasing concentrations of the pollutant.

30. Determination of time exposure and maximum toler-

ance limits.

31. Determination of bioaccumulation levels.32. Creation/enrichment of a Knowledge Base, receiving

information internally (Fig. 4) and externally via an

intelligent agent (Batzias and Markoulaki, 2002).

33. Production of ad hoc model for bioindicator response

in the field.

The methodology presented can be easily extended

for mosses, whereas it can be adapted for microorganisms

and lower invertebrata, as well as marine biomonitors.

The information required for the development of the

biosensor systems (stages 3, 4, 17, and 18) is retrieved from

a biosensors Knowledge Base. The lichen ontology

described in Section 3.2 is used to retrieve information

for the bioindicator species and the biochemical pathways

(stage 1), whereas it can be enriched by the output of the

calibration/validation procedure (stage 32).

The organisms response to a given pollutant can be thus

studied in depth by decomposing the organism and

examining, qualitatively and quantitatively the effect of

any pollutant upon each of the isolated biochemical

systems or sub-systems. It is feasible to use the enzymes

or the systems participating in stress-response as bioele-

ments for developing biosensors. In that way, the pollutant

effect can be qualified and quantified for each step of itsway through the organism on the basis of the biosensors

response.

Developing biosensors for concentration-dependent

response, biochemical systems that are not affected (no

response from the device), mostly affected (devices that

provide the greatest response) or threshold-triggered

(device that responds only above a certain concentration

of the pollutant) can be identified, whereas tolerance limits

and doseresponse relations can be estimated. The devel-

opment of biosensors for cumulative response provides

useful correlations to the actual levels of pollutants

contained within the organism (bioaccumulation). By

investigating the mechanism of signal generation of the

biosensors, the impact of the pollutant upon the metabo-

lism of the organism will be clarified. By means of

interference and matrix effect studies, the effect of other

substances present upon the adsorption of the target

pollutant by the organism (effect upon the signal of the

device, competition for the bioelement binding sites) can be

qualified. Apart from impact assessment, biosensors will

provide the means for calibration and validation of the

organism. The doseresponse relations can be linked to

morphological alterations, whereas sensitivity indices can

be derived from the comparison of the performance of the

biosensors developed from different species. Establishing

ARTICLE IN PRESS



11/19

and maintaining a study scheme as such, the investigation

of seasonal variations, species variability, pollution back-

ground, combination of pollutants, spot variables, etc.,

upon bioindicator performance becomes realistic.

Correlation between bioindicator stress response and

biosensor signal should take into consideration cellular

membrane functions since most biochemical reactions inliving organisms are performed within or through lipid

membranes. This problem could be possibly overcome by

utilising bilayer lipid membrane biosensors (Krull and

Thompson, 1985; Nikolelis et al., 1998). Lipid biosensors

could be further used for bioavailability estimation, which

is currently quite problematic in biomonitoring, as it is

based on correlations between ambient concentration

and bioaccumulation (Vasconcelos and Tavares, 1998;

Wolterbeek, 2002) thus requiring frequent verification

and adjustment. Bioavailability mostly depends on

solubility and membrane permeability, the latter being

difficult to estimate (Burger et al., 2003); utilising artificial

membrane biosensors permeation kinetics can be investi-

gated.

4. Realisation of the CSS

The design, development and establishment of a

cooperative/synergistic monitoring system engaging

living organisms and biosensors, considering the diversity

and the complexity of the former and the performance

problems of the latter, especially in long-term field

measurements, could be quite problematic. Depending on

ones role, there may be subtle or distinctly different

priorities for outcomes and even on criteria for measuringwhether the outcomes have been achieved. These differ-

ences may be necessary for the proper functioning of

the system as a whole but they make development

and management difficult. Any system as such should

(a) establish practical, valid and equitable evaluation

criteria by which qualitative and quantitative changes of

pollutants can be monitored and assessed, and (b) involve

methodological pluralism (including both qualitative and

quantitative methods) to ensure rigour and comprehen-

siveness in assessment.

The methodological framework designed and developed

by the authors to address the implementation problems of

the proposed CSS is presented below, interconnected as in

Fig. 5.

1. Collection of information on the pollutants released

from similar sources, and on their reactivity after

release (in combination with source-specific releases

or/and releases from other near-by sources or/and the

community).

2. Determination of specific source activities that con-

tribute directly or indirectly to pollution (air, liquid).

3. Determination of primary pollution species.

4. Selection of relevant phenomenological models

regarding the formation of secondary pollution

species, resulting from the combination of (inert or

active) chemicals from the source under study or from

near-by sources. Creation of a pollution Knowledge

Base.

5. Ranking of models in order of decreasing likelihood

(to be valid in the certain situation), according to

technical literature and experts opinion.6. Determination of the species to be measured and the

corresponding significance level (for each species)

required for model validation.

7. Determination of (usual) concentration ranges, diffu-

sion/transport patterns and frequency of release.

8. In field measurements to verify concentration ranges

and diffusion/transport patterns.

K. Are ranges and diffusion/transport patterns verified?

9. Setting alarm levels, maximum tolerance levels, threat

levels for each pollutant species, as well as sensitivityfor each level.

10. Choice (from Lichen Ontology) of the grouped

patterns of indigenous vegetation species to be utilised

as bioindicators. Criteria: (a) coverage of the desired

pollutant concentration range, (b) well-established

sensitivity to pollution, (d) densely spread through

the area under consideration.

11. Testing by operating in simulated environment under

laboratory-controlled conditions as per individual

pollutants and combinations; assignment of sensitivity

indices (stages 2731 of the algorithmic procedure

illustrated in Fig. 4).

L. Is the operation of the bioindicators in simulated

environment satisfactory?

12. Calibration of bioindicator response to pollutant(s)

levels: (a) external alterations (related to low, medium,

high levels of pollutions, related to stage 9) and

(b) bioaccumulation levels; estimation of species

bioavailability and calculation of available/bioavail-

able correlation factor (through the algorithmic

procedure illustrated in Fig. 4).

M. Are the specifications set at stage 9 met?

13. Retrieve relevant physicochemical modes from the

Lichen Ontology (phytobiont, transport, fungus).

14. Choice (from the Biosensors Knowledge Base), pre-

paration and laboratory testing of the dedicated

biosensor, utilising bioindicator-mimicking operation.

15. Test biosensors in simulated environment under

laboratory-controlled conditions as per individual

pollutants and combinations; determination of bio-

sensor response (transient and cumulative) and

investigation of mechanism of signal generation.

ARTICLE IN PRESS



12/19

N. Is the operation of the biosensors in simulated

environment satisfactory?

16. Correlation of biosensor transient response to bioin-

dicator external response and physiochemical altera-

tions; correlation of biosensor cumulative response to

bioaccumulation levels and bioavailability (Data

retrieved from the Biosensor Calibration/Validation

Knowledge Base).

O. Is correlation adequate to cover partially or totally

the specifications set at stage 9?

17. Calibration of bioindicators (Data retrieved from the

Biosensor Calibration/Validation Knowledge Base).

18. Biosensor development for operating in real environ-

ment (determination of stability, shelf life, calibration,

fault detection/compensation).

19. Testing biosensors by operating in real environment

under most likely natural conditions.

P. Is the operation in real environment satisfactory?

Q. Is there another species left unexamined?

20. Design of a network for gathering and analysing

biosensor data.

21. Parallel measurement of all species at the same time by

using the biosensors that passed the tests so far and the

network developed.

R. Is the model used valid at a predetermined

significance level?

S. Is there another model left unused?

22. Selection of the most likely tested model by relaxing

the statistical significance criterion.

ARTICLE IN PRESS

Fig. 5. The methodological framework designed/developed for integrated environmental biomonitoring through the multi-level synergy between

biondicators and biosensors.



13/19

23. Packaging the set of approved biosensors within a kit

for protection and convenience (i.e. easiness of

handling during inspection/maintenance/ replace-

ment).

24. Cost estimation per kit, including placement/checking/

maintenance/ replacement.

25. Construction of grid within each GIS layer, under theconstraint of the budget available, corresponding to

each pollutant species considered as input variable of

the model.

26. Optimal placement of biosensors according to experi-

mental design techniques. Selection of bioindicator

grids.

27. Small-scale preliminary operation for testing and

design of a maintenance/replacement program.

T. Are the results of preliminary testing satisfactory?

28. Full-scale operation, long enough (at least one year) topermit adequate output data collection of both

biosensors and bioindicators, in the same course.

Stages 2938, as well as the intermediate decision nodes

U, V, W, X, run in series or in parallel, for each pollutant

layer Gi (i 1,y,k, where k is the number of total

pollutants involved):

29. Extension of the full operation period to obtain

additional biosensor measurements to be compared

with corresponding estimated values based on bioin-

dicator external response and accumulation levels and

the learning set obtained in the previous stage.

30. Comparison between biosensor measurements and

corresponding bioindicator estimated values for the

Gi pollutant GIS layer (based on the pollution

prediction model outcome and the interpolation

method currently employed).

U. Is the comparison satisfactory, for each Gi pollutant

GIS layer, at a predetermined significance level?

31. Bioindicator sampling and lab testing. Estimation ofbioindicator replacement program.

V. Are species resistant?

32. Application of a program for saving economic

resources by removing most biosensors from this

system (to be used within a similar system, taking

advantage of the knowledge/experience acquired) but

keeping a number of them required to maintain

predictability within an acceptable region.

33. Estimation of time periods Ti (duration di and

frequency fi) that biosensors are established to execute

a full program of measurements together with the

bioindicators which act on a permanent basis.

34. Realisation ofTi, di, and fi.

35. Comparison between biosensor measurements and

corresponding estimated values within each Gi pollu-

tant GIS layer, at time period Ti.

W. Is the comparison satisfactory, or should we extend

the biosensor measurement period?

36. Creation/enrichment of a Knowledge Base (KB)

receiving information internally (Fig. 5) and externally

via an intelligent agent (Batzias and Markoulaki,

2002).

37. Intra-and inter-net searching via an intelligent agent

based on an ontological adaptable interface.

38. Periodic comparison of measured pollution levels with

the corresponding values obtained by the model

estimations, as depicted on a separate GIS layer(based on the pollution prediction model outcome

and the interpolation method currently employed).

X. Is the comparison satisfactory, for this Gi pollutant

GIS layer, at a predetermined significance level?

Y. Is the packaged kit performance satisfactory?

The framework can be utilised/adjusted to any environ-

mental system for multi-elemental monitoring in both

phases, the gas and the liquid, encompassing a regional

framework around a pollution source in the form of a

LAN, with on-line data collection and mining, as regards

instrumental monitoring (stages 20 and 21). Biologicalmonitoring requires periodic field observations, the fre-

quency of which can be modified according to the signal

retrieved by the network; for example, an alarming

situation revealed by the biosensors (increasing trend of

pollution load) would prompt an in situ investigation of the

lichen response and the priming, if necessary, of the

appropriate countermeasures. The correlation of biosensor

signal to bioindicator response can be also used to confirm

the validity of the correlation model (stage 30), especially in

cases of abrupt pollution load increases (peaks) where the

physiological response of the organisms could differ

significantly from that observed in the simulated environ-

ment.

The selection of the biomonitoring species (stage 10) is

realised through a program constructed ad hoc from the

Lichen Ontology (see Section 3.1), designed to provide a

grouped pattern suitable to cover the desired concentration

range: a range of organisms with various levels of tolerance

are selected so that the concentration of the pollutant in the

area under consideration is estimated by the existing

pattern, i.e., the absence of certain species and the presence

(and abundance) of other species. The data input and the

intermediate/final results output (correlations), referring to

species selection in a particular area to cover the desired

pollution levels (ranging from low to high) and their

ARTICLE IN PRESS



14/19

ranking as per sensitivity, are stored into a Relational

Database Management System (RDBMS), an adaptation

of the ABEPE Database (Batzias et al., 2004). The

RDBMS model provides a user friendly and efficient tool

to handle large quantities of data, and compatibility for

hosting GIS ready data. In addition to the interoperability,

the RDBMS wisely utilises the powerful programminglanguage (Structured Query Language: SQL) as well as the

unrivalled performance and scalability of the GIS. More-

over, since SQL is an ANSI standard, widely accepted by

the computer industry, a large amount of other commercial

software can cooperate with SQL applications and

increase/expand functionality. Since there are natural

language interpreters translating voice or keyboard entered

commands into SQL statements (e.g. Microsoft English

Query), the potential bioindicator ontology user can

submit queries in a natural language. The broad avail-

ability of MS Access and its friendly interface with the GIS

used renders this RDB the most suitable development

environment. Not to mention that such an implementation

of the database could be easily upsized to a multiprocessor

system running the SQL Server RDBMS because of the

transparent migration of an Access application to the SQL

Server.

An SQL Select query was created as an ADD-ON

software device to combine the species availability with the

tolerance data and calculate the sensitivity potential for all

lichen species per pollutant, region, season, and altitude.

In fact, the use of a RDBMS has been proven effective,

since the formation of complex queries comes easily

through the exploitation of SQL.

The program searches the database for lichens indigen-ous to a specific geographic area, selected either from the

pre-determined options in the region button (that covers

districts, prefectures, towns, etc.) or by specifying the

point-source coordinates (Fig. 6a). The user defines the

pollutant or the combination of pollutants (by activating

the pollutant 2 button), the required concentration range,

as well as the desired sensitivity (low, medium, high,

depending on the species response towards the given

pollutants). The concentration ranges are entered as

normal, low, low to intermediate, intermediate, and high,

following the current international norms. For example,

regarding sulphur dioxide, normal(N) iso3 ppb, low (L) is

340 ppb, low to intermediate (LI) is 40100 ppb, inter-

mediate (I) is 100200 ppb, and high (H) is 4200 ppb.

There are also options for season, altitude and pH that can

further specify the query. The program retrieves grouped

patterns, in which the combination of species response

offers the best estimation for the pollutant(s) concentra-

tion, ranked by selectivity/sensitivity and robustness.

Moving down from one alternative to the other, the

uncertainty is increased; this is compensated by increasing

the number of species participating in the monitoring

scheme. The second, third, etc., alternatives are used as a

pool, retrieved by activating the corresponding button,

providing the suitable replacements for the species in the

first grouped pattern, that for any reason, cannot function

as required.

The program has been implemented for a limited

number of cases, one of which is presented in Fig. 6. The

monitoring of SO2 in Crete, during summer and at

moderately acid conditions (pH: 4.75.8), can be realised

using the first grouped pattern shown in Fig. 6a, retrievedby the program as most suitable to cover the whole

concentration range. The numbers in the right columns

represent the degree of coverage of the species response

towards the concentration level shown in the first row (as a

fraction of 1); for example, Collema nigrescens covers

100% (ie., 1.0) of the normal to low-intermediate level and

10% (ie., 0.1) of the intermediate level, which can also

indicate that this species would be present at ca. 100 ppb

SO2 but most probably absent at 140 ppb SO2. The

estimation of the sulphur dioxide level is given by

combining the species response: the presence of Aspicilia

cheresina var. microspora indicates a clean environment

(o3 ppb SO2), in the absence of which the level of sulphur

dioxide would be up to ca. 30 ppb if Lecanora pruinosa is

present; in case the latter are absent but the two other

Lecanora species are found in considerable numbers,

sulphur dioxide levels are ca. 50 ppb. If the latter are

absent, the presence of Toninia candida sets pollution at

7080 ppb, in the absence of which Cladonia and Collema

species raise this level to 100110 ppb. In this way,

uncertainty is decreased significantly, especially towards

the boundaries of the pollution levels: when moderate

pollution is expected or predicted, a new optimal grouped

pattern is provided through fine tuning (Fig. 6b), by

replacing some species in the previous grouped patternwith others in order to confer better sensitivity for the

range 3200 ppb. At that point, by activating the abun-

dance button, the same grouped pattern appears, thus

concluding the selection procedure.

In case the available information on the indigenous

vegetation species is not adequate, the extension button

will activate a new query for locating similar species,

inhabiting areas with similar climatic, geomorphologic,

and geophysicochemical parameters, with established

response towards the pollutant of interest, in order to

enrich knowledge. The same can apply if indigenous species

with adequate sensitivity towards the pollutant of interest

are scarce; in that case, transplants should be considered,

whereas the use of moss bags could be useful as a pre-

screening activity, in order to evaluate level of contamina-

tion with respect to availability parameters.

Upon correlation of instrumental and biological

responses (stage 30), deviations could be attributed to

modifications of the organisms growing conditions (e.g.

change in soil mineral content or pH) due to reasons other

than the monitored pollution. The lab analysis of the

species (stage 31) should clarify that point, indicating the

shift towards another bioindicator grouped pattern in case

the selected species show systematic or behavioural

deviations that render them unsuitable for monitoring

ARTICLE IN PRESS



15/19

purposes (enabling the utilisation of the replacements

obtained in stage 10) or, in the case of unanticipated

environmental alterations, a change in the correlation

model (stage 16); the latter involves a detailed investigation

of the origins, extent and fluctuation of the alteration, and

depending on its severity, would call for re-calibration/

validation (stage 17).

The core of the framework is the removal of most

biosensors from the field (stage 32) and the shift to

biomonitoring for saving economic resources, although a

number of them should be kept in order to maintain

predictability within an acceptable region. The devices

removed will be placed again in the field as part of the

recalibration/revalidation phase of biomonitoring (stage 38).

Taking advantage of the knowledge/experience acquired,

they can also be used, including the supporting equipment,

for initiating/maintaining a similar surveillance system in

another area.

5. Discussion

The methodological framework presented considers both

evaluation and management tasks supporting decision

making. It permits quantitative and qualitative analyses

of the environment in a cost-effective manner, since the

main objective is the gradual withdrawal of the analytical

instrumentation, and the shifting of the monitoring system

to nature.

ARTICLE IN PRESS

Fig. 6. Sample screenshots of the developed computer program for selecting grouped patterns to monitor SO2 in the island of Crete (Greece): (a) best

grouped pattern for covering all concentration ranges, and (b) best grouped pattern to account for low to intermediate concentrations. In the latter case,

some species are added (as Evernia prunastriand Opegrapha varia to account for pollution levels towards 2535 and 4050 ppb, respectivey) and/or replace

others (as Caloplaca Saxicola replaces Aspicillia cheresina, sinceo3 ppb accuracy is not a prime concern) in order to increase sensitivity and reliability of

measurements.



16/19

Most monitoring programs or methodological frame-

works reported are ad hoc developed, based on the

substances to be monitored, the conditions of the area

under surveillance, the equipment available and the

recourses (Caughlan and Oakley, 2001; Conti and

Cecchetti, 2001; Szczepaniak and Biziuk, 2003; Argent, 2004;

Guiochon and Beaver, 2004). Limited budget compelsmost of the time the selection of simplistic approaches as

more affordable and acceptable by the authorities that

finance the program. The proposed scheme of a modular

structure that will enable decisions to be made regarding

costs as instrumental monitoring can be increased/reduced

over time according to the performance of the program and

the needs arisen and can encourage the local production of

ready-to-market/use devices, utilising indigenous materials

and taking advantage of scale economies due to large

production. Furthermore, the experience obtained and the

developed set of biosensors and supporting equipment used

for engaging and maintaining the monitoring scheme in

one area can be used, mostly as is, in starting-up/

maintaining a biomonitoring scheme in another area.

It is worthwhile noting that in the time-course, the

expected validation cost will decrease as a result of

(a) human experience accumulation and (b) incorporating

know-how within the system itself. This is a common

characteristic in systems approaches to multi- and inter-

disciplinary issues, as it has been stressed by several

authors engaged with solving problems in different

disciplines; e.g. Senge (1992) and Bellamy et al. (2001),

argue about a kind of learning organisation that by means

of a systemic view leads to knowledge enrichment, implying

progressive improvement of the system itself. This is onlycountered by a possible self-organising of the ecosystem

that may develop resistant species; in that case, new

grouped patterns (stage 30 of the algorithmic procedure

shown in Fig. 5) are required, as well as more extensive

revalidation in comparison with the periodic one.

Nonetheless, the main objective of the framework is the

shift to natural monitoring. The relative ease of sampling,

the absence of any need for complicated and expensive

technical equipment, and the accumulative and time-

integrative behaviour of the monitor organisms (Beeby,

2001; Conti and Cecchetti, 2001; Wolterbeek, 2002;

Bargagli et al., 2002) renders biomonitoring mostly suited

for large-scale surveys. The construction of an environ-

mental impact ontology, as the lichen ontology presented,

can provide a useful tool for the design/development/

implementation of an environmental network for the

monitoring of a variety of pollutants over time and space

and the assessment of environmental quality. The collec-

tion of the available information and its classification

into taxonomic and partonomic relations provided a

database linking multi-functional, phenomenological and

in-depth, pollutants with response, with respect to ecolo-

gical parameters, relations and geomorphologic character-

istics. Such an ontology allows indicator responses to

become understandable, links biomonitoring to policy-

making and provides a significant utility to costbenefit

assessments. In that context, biomonitoring can support

management decisions or quantify the success of past

decisions.

The quantitative assessment of elements requires well-

defined doseresponse and bioavailability relationships,

and knowledge of accumulation, retention and releaseprocesses (Beeby, 2001; Wolterbeek, 2002). The authors

have presented a validation and monitoring scheme that

could address these issues. The calibration of the bioindi-

cators by means of biosensors is the central component of

the framework rather than an activity external to the

program process. It forms the basic platform for the system

development in order to address problems such as

biomonitoring validation. Since field biosensors are used

for validating the performance of bioindicators, the

correlation between the former and the latter in only

possible through inter-calibration. In that way problems

such as misinterpretation of data, incorrect or biased

signals, and uncertainty of measurements are diminished.

However, much effort is still required in order to bridge

laboratory findings and field results.

Although monitoring programs are, in general, not

designed to elucidate causal relationships, biomonitoring

utilises associations that may be causal. The correlation of

biotic indicators to instrumental analysis has been proven

difficult to estimate and is suspect when detected because of

spatial and temporal disjunctions in sampling, differing

scales of sampling units, the lack of information on

diagnostic responses, and the lack of data on alternative

causes or causal cofactors. Even when monitoring reveals

strong associations, those correlations cannot be taken torepresent causation. However, using a similar system for

validation, mechanistic relations can be revealed through

the investigation of the physicochemical basis of biosensor

signal generation.

The linking of the GIS-supported framework to knowl-

edge bases, as the lichen ontology developed by the

authors, enhances the semantic interoperability of the

latter, since ontologies seem to be an adequate methodol-

ogy that helps to define a common ground between

different information communities. This paper describes

an example of biosurveillance and an ontology that serves

as the basis for carrying out such schemes. It is focused on

the way ontologies can support environmental monitoring

and the how the provision and (re)use of ontologies can aid

decision making and cost-effectiveness. Access through the

Internet could enable users to construct common ontol-

ogies across GIS boundaries; the increased need for mobile

interoperable tools makes the ontology-based approach

valuable for field GIS.

The concept of integrated biosurveillance to detect

indications and warnings for a possible peak event is also

covered by the proposed methodology. Current biosurveil-

lance systems fail to provide pre-event predictive informa-

tion and are stage-piped across multi-sector domains. The

integration of multiple data sources (a variety of organisms

ARTICLE IN PRESS



17/19

and a variety of biosensors) into a LAN offers the potential

for better predictive ability and greater lead-time warning

than that provided by separate stand-alone monitoring

systems, not only in the area under consideration but

also in neighbouring areas that might be affected from

transportation phenomena. The value of the proposed

integrated biosurveillance will be realised when lead-timefor response planning is expanded by providing pre-event

alerts that enable priming of the response system. The

LAN, as part of a WAN and through appropriate

transportation/dispersion models, can provide the basis

of or trigger rectifying/prevention measures that cover a

wide range in short time. Comparison of multiple data

sources suggestive of a bio-event may also decrease the

number of false positive alerts. Holistic integration of the

proposed methodology to address impacts on plants,

animals and humans across foreign and domestic domains

is critical to further develop this approach.

Furthermore, statistical indices can be introduced to

account for both bioindicator-species richness and the

evenness with which individuals are spatially distributed

among species. By doing so, a reliable/representative

statistic parameter is obtained as a quantified criterion

for abundance, which can play a significant role in the

hierarchical criteria choice of the optimal grouped pattern

in as much as this criterion may reflect also an impact of

pollution on the population of the species comprising the

group: in the time course, the value of the parameter may

decrease, decreasing at the same time its significance as a

criterion, while increasing its importance as part of the

suffering ecosystem. The Simpsons, Shannon-Weiner, and

Brillouin indices may fit the situation, while the lognormaldistribution provides probably the best statistical back-

ground in practice, although it does not satisfy certain

strict mathematical criteria (He and Legendre, 2002).

6. Conclusions

With environmental concerns escalating, environmental

planning and monitoring programmes are increasingly

being promoted globally as means for assessing environ-

mental quality. This paper has sought to overcome some of

the significant challenges to validation and evaluation in a

complex monitoring system, such as biomonitoring, and to

the management of information collected from the sites of

intensive monitoring. The bioindicator ontology on envir-

onmental exposure presented links pollution to morpholo-

gical (surface) response and biochemical (mechanismic)

alterations, also providing a useful tool for species selection

to fit a robust and reliable biomonitoring system. Based on

this procedure, a computer program has been designed/

developed as a decision support system and has been

successfully tested on a representative population of species

indigenous to southern Greece.

Significantly, a novel system (scheme) in the form of a

rational framework at the conceptual design level has been

developed to guide improvements in such initiatives, that

actually contributes towards achieving a cost-effective

long-term monitoring program, with the flexibility to

counter on-course any (anticipated or not) variations/

modifications of the surveillance environment. It is thereby

proven that the combination of permanent biomonitors

with periodic biosensors (which can be produced locally

utilising indigenous materials) may contribute substantiallyto designing a strategy for saving resources; further to

minimising capital costs and maintenance costs, this

combination creates a local network that apart from

contributing to environmental surveillance can also moni-

tor the growth rates of native vegetation.

Acknowledgements

This work was performed within the framework of

Pythagoras II EU-GR Research Programme (Section:

Environment) for the design/development/implementation

of biosensors/bioindicators.

References

Argent, R.M., 2004. An overview of model integration for environmental

applicationscomponents, frameworks and semantics. Environmental

Modelling & Software 19, 219234.

Bargagli, R., 1989. Determination of metal deposition patterns by

epiphytic lichens. Toxicological and Environmental Chemistry 18,

249256.

Bargagli, R., 1998. Trace Elements in Terrestrial Plants: An Ecophysio-

logical Approach to Biomonitoring and Biorecovery. Springer, Berlin,

Germany.

Bargagli, R., Monaci, F., Borghini, F., Bravi, F., Agnorelli, C., 2002.

Mosses and lichens as biomonitors of trace metals. A comparisonstudy on Hypnum cupressiforme and Parmelia caperata in a former

mining district in Italy. Environmental Pollution 116, 279287.

Batzias, F.A., Markoulaki, E.C., 2002. Restructuring the keywords

interface to enhance CAPE knowledge via an intelligent agent.

Computer-Aided Chemical Engineering 10, 829834.

Batzias, F.A., Siontorou, C.G., 2004. Computer aided optimal determina-

tion of biosensors replacement program for keeping a cleaner

environment in underground mine operations. Seventh Conference

on Process Integration, Modelling and Optimisation for Energy Saving

and Pollution Reduction (PRES 2004), 2226 August, Prague, Czech

Republic.

Batzias, F.A., Sidiras, D.K., Spyrou, E.K., 2004. Evaluating livestock

manures for biogas production: a GIS based method. Renewable

Energy available online at www.sciencedirect.com.

Batzias, F.A., Siontorou, C.G., 2005. Investigating the causes of biosensorSNR decrease by means of fault tree analysis. IEEE Transaction on

Instrumentation and Measurement 54, 13951406.

Beeby, A., 2001. What do sentinels stand for? Environmental Pollution

112, 285298.

Bellamy, J.A., Walker, D.H., McDonald, G.T., Syme, G.J., 2001.

A syste

novel system

Documents