Potential Discrimination of Toxic Industrial Chemical Effects on Poplar, Canola and Wheat,...

$Page 1: Potential Discrimination of Toxic Industrial Chemical Effects on Poplar, Canola and Wheat, Detectable in Optical Wavelengths 400\u0026#x2013;2450 nm$
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 5, NO. 2, APRIL 2012 563

Potential Discrimination of Toxic Industrial ChemicalEffects on Poplar, Canola and Wheat, Detectable in

Optical Wavelengths 400–2450 nmDerek Rogge, Benoit Rivard, Michael K. Deyholos, Josée Lévesque, Jean-Pierre Ardouin, and Anthony A. Faust

Abstract—This research examined the spectral response ofpoplar (Populus deltoides, Populus trichocarpa), wheat (Triticumaestivum), and canola (Brassica napus) leaves subjected to fumi-gation with gaseous phase toxic industrial chemical gases (TICs).The gases include ammonia (NH ), sulphur dioxide (SO ), hy-drogen sulphide (H S), chlorine (Cl ), and hydrogen cyanide(HCN). This study aimed to determine if: 1) vegetation subjectedto TICs could be distinguished from background vegetationduring varying growth stages and environmental stresses; and,2) different TICs could be distinguished based on the spectral re-sponse of vegetation. The results showed that both environmentaland TICs induced similar spectral features inherent to plants,which are related primarily to chlorophyll and water loss. Thesefeatures include pigments in the visible and cellulose, lignin, lipidsstarches, and sugars in the SWIR. Although no specific spectralfeatures could be tied to individual TICs an analysis of the datausing vegetation indices showed that the TICs and environmentalstresses result in diagnostic trends from healthy mature to highlystressed leaves. In addition combinations of specific indices couldbe used to distinguish the effects of NH , SO , Cl and their effectfrom that of other treatments of the study. The continued goal forthis research program is to develop a remote detection capabilityfor hazardous events such as a toxic gas leak. Our findings at theleaf level suggest that damage can be detected within 48 hrs andshould last for an extended period. Thus, the next experimentalstep is to test if the results shown here at the leaf level can also bedetected with airborne and satellites systems.

Index Terms—Leaf level spectroscopy, spectral indices, toxic in-dustrial chemical gases.

Manuscript received July 18, 2011; revised October 14, 2011; acceptedNovember 20, 2011. Date of publication January 24, 2012; date of currentversion May 23, 2012. This research was conducted as part of the HYMEXproject W7702-05R073/A with funds provided by J.-P. Ardouin of the defenseestablishment at Val-Cartier, Québec, Canada.D. Rogge was with the Department of Earth and Atmospheric Sciences, Uni-

versity of Alberta, Edmonton, Canada, and is now with the German RemoteSensing Data Centre, DLR, Wessling D-82234, Germany (corresponding au-thor, e-mail: [email protected]).B. Rivard is with the Department of Earth and Atmospheric Sciences, Univer-

sity of Alberta, Edmonton, AB, T6G 2E3, Canada (e-mail: [email protected]).M. K. Deyholos is with the Department of Biological Sciences, University of

Alberta, Edmonton, AB, T6G 2E3, Canada (e-mail: [email protected]).J. Levesque is with the Spectral and Geospatial Exploitation Section, De-

fense Research and Development Canada, Valcartier, Québec, G3J 1X5, Canada(e-mail: [email protected]).J.-P. Ardouin is with the Spectral and Geospatial Exploitation Section, De-

fense Research and Development Canada, Valcartier, Québec, G3J 1X5, Canada(e-mail: [email protected]).A. A. Faust is with the Explosives Detection Group, Defence Research and

Development Canada Suffield, Medicine Hat, AB, T6G 2E3, Canada (e-mail:[email protected]).Color versions of one or more of the figures in this paper are available online

at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/JSTARS.2011.2179918

I. INTRODUCTION

R EMOTE sensing has been a useful tool in the assessmentof natural and anthropogenic disasters, which can have

significant effects on the health of vegetation. Early studies ofthe environment have shown that changes in photosynthetic pro-cesses can be indicative of pollutants. Understanding the re-sponses of plants to anthropogenic disasters, such as the releaseof gases into the atmosphere, can allow for better control of thespatial extent and damage of these spills in urban, agriculturalor natural settings (e.g., train derailments). This paper considersspectral responses of plants to anthropogenic stress factors inthe form of airborne toxic industrial chemical gases (TIC). Plantstress detection is assessed at the leaf-level in a controlled lab-oratory environment.A number of scientific publications can be found in various

journals pertaining to plant stress and plant stress detection,where stress factors can be divided into natural and anthro-pogenic factors. Much of the literature has focused on phys-iological effects, processes and responses of plants, but therehave also been numerous studies that demonstrate spectroscopy,via remote sensing, as an effective tool to detect plant stress.Publications by Chapin [1] and Lichtenthaler [2] give a goodoverview of the former. This paper will focus on the latter.Spectroscopy is a commonly used non-invasive approach

in the study of plant physiology, processes, and responses tostress factors. Examples include, Gausman [3] who consideredleaf optical properties pertaining to pigment concentrations,internal leaf structure, water content, maturation, nodal posi-tion, sun versus shaded leaves, pubescence, and senescence.Curran [4] focused on estimating foliar chemical content usingremote sensing. In a series of publications by Carter [5]–[8],leaf spectral responses to plant stress and water content wereconsidered in relation to chlorophyll concentration. Blackburn[9] evaluated the application of remote sensing for quantifyingchlorophyll and carotenoids at leaf and canopy scales. Deter-mining leaf pigment content and activity with a reflectometerwas assessed by Gamon and Surfus [10]. Zarco-Tejada et al.[11] used multiple airborne remote sensing platforms to detectstress in crops from chlorophyll concentrations. Baltzer andThomas [12] studied leaf optical responses to light and soil nu-trient availability in temperate deciduous trees. More recently,Blackburn [13] reviewed the potential of hyperspectral remotesensing for the analysis of concentrations of various pigmentgroups including chlorophyll, carotenoids, and anthocyanins.

1939-1404/$31.00 © 2012 IEEE

564 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 5, NO. 2, APRIL 2012

Although there has been a large number of publications in-vestigating the effects of gases on plants, few papers have exam-ined the spectral responses of these plants. For example, Smith[14], [15], and Noomen [16] measured the spectral responsesof plants related to subsurface gas leaks. More recently withthe advent of CO sequestration, spectroscopy has been ap-plied to monitor CO influx into soils and the resulting plantstress [17]. Papers that have focused specifically on spectralresponses of plants to atmospheric gases are also limited. El-evated O and CO were considered in papers by Carter etal. [7], Kraft et al. [18], and Carter and Knapp [5]. More re-cently Gab [19] used NIR-spectroscopy to investigate foliageresponse to ozone-stressed Fagus sylvatica trees and Meroni[20] looked at early leaf level injuries caused by ozone usingpassive fluorescence and the photochemical reflectance index[21]. Williams and Ashenden [22] looked at spectral character-istics of white clover to gaseous pollutants, such as SO andNO . Schutt [23], Gemmell and Colls [24], and Rudorff et al.[25] observed spectral effects of O and SO on leaf charac-teristics over portions of the 400–1100 nm range. Results fromthese limited studies have shown that spectroscopy can be usedto detect plant stress associated with atmospheric gases basedon changes in leaf structure, chlorophyll and pigment concen-tration and water content. However, directly correlating specificgases to spectral responses has been more problematic as plantsrespond to different stress types in basically the same way [1],[5].In the context of evaluating the ability of reflective hyper-

spectral imaging (400–2450 nm) to identify vegetation that wassubjected to fumigation with gaseous phase TICs (ammonia(NH ), sulphur dioxide (SO ), hydrogen sulphide (H S),chlorine (Cl ), and hydrogen cyanide (HCN)), this researchexamined the spectral responses from individual leaves of threecommon Canadian plants: poplar (Populus deltoides, Populustrichocarpa), wheat (Triticum aestivum), and canola (Brassicanapus). Poplar is distributed throughout both the boreal forestand in the parklands that comprise much of the landscapein Canada. Wheat and canola are the most widely cultivatedplants in Canada, and both have many closely-related wild rel-atives (i.e., grasses and crucifers) that grow widely throughoutCanada and the rest of the world. The research also examinedbackground spectral responses that included spectral variationrelated to developmental (senescence) and environmental ef-fects (drought and NaCl). The larger goal is to develop a remotedetection and monitoring capability for hazardous events suchas airborne toxic gas leaks.The accumulated experimental data was analyzed to address

two objectives: 1) determine if vegetation subjected to TICscould be distinguished from vegetation that is senescing or sub-jected to environmental stresses, and 2) determine if effectsof different TICs on vegetation can be distinguished. Two ap-proaches were used to analyze the spectral responses of eachspecies. The first made use of endmember analysis to capture themost severe response relative to healthy leaves and determine ifspecific TICs produced diagnostic spectral features relative toother TICs and/or environmental stresses. The second approachconsidered whether leaf spectral responses for each TIC differedas a function of the dose response (i.e., from normal growth

to severe damage). In order to observe these temporal spectralchanges we made use of a number of vegetation indices, whichfocused on key spectral bands associated with chlorophyll, pig-ments and water content.

II. MATERIALS AND TREATMENT PROCEDURES

Treatments were designed to allow quantification of the vari-ation in spectra that might be expected due to environmental,developmental, and stochastic effects on the physiological stateof individual plants within each species.

A. Plant Growth

All plants were grown in controlled environment cham-bers at the University of Alberta, using standard conditions.Specifically, plants were grown in Metromix 360 soil mixture(Scotts, Marysville, OH) in controlled environment chambersat 24 C with 50% humidity, and at a light intensity of 200 E(microEinstein) supplied by high output fluorescent bulbs(color rendering index of 85, color temperature of 3500 K)on a 16 h light/8 h dark cycle. Canola (B. napus var. Westar)and wheat (T. aestivum var. HR 5600) were grown from seed,and poplar (P. trichocarpa x P. deltoides clone H11-11) wasgrown from cuttings. Wheat and canola plants were 3–4 weekspost-germination at the time of treatment, and poplar saplingswere generally 8 weeks post-cutting.

B. Environmental Stresses

Each species was subjected to treatments representing en-vironmental factors that could influence spectral responses inthe field, and which may influence the detection of TICs by re-flectance spectroscopy. NaCl treatment was used as a model en-vironmental stress, since it induces ionic stress, as well as os-motic stress responses that are common to drought, freezing,and salinization (e.g., [26], [27]). Acute NaCl stress was ap-plied by drenching roots with a volume of NaCl solution (150 to400 mM, Table I) equivalent to at least three times the volumeof the pot in which the plant was contained. Each treatmentwas compared to a mock-treated parallel control. We also im-posed drought stress, by with-holding of regular watering fromtreated plants, while control plants grown in parallel receivedthe normal amount of water. Drought-treated plants (listed asH O in tables and figures) were analyzed spectrally when thesoil had fully dried and plants began to show visible signs ofstress (e.g., wilting).

C. Application of Toxic Industrial Chemicals (TICs)

Plants were exposed to the following five TICs: ammonia(NH ), sulphur dioxide (SO ), hydrogen sulphide (H S), chlo-rine (Cl ), and hydrogen cyanide (HCN), at atmospheric con-centrations ranging between 0.5 to ppm (Table I). How-ever, owing to practical limitations (e.g., availability of TIC,plant size), not all species were exposed to all TICs. For expo-sure we used gaseous phase exposure (i.e., fumigation) of wholeplants as the exclusive method of application of TICs, since thisis directly relevant to the field situation. Alternative methodsof exposure (e.g., painting or spraying) are difficult to replicateand quantify, require solubilization of the TIC in a liquid, andare more likely to produce artifacts.

ROGGE et al.: POTENTIAL DISCRIMINATION OF TOXIC INDUSTRIAL CHEMICAL EFFECTS ON POPLAR, CANOLA AND WHEAT 565

TABLE IENVIRONMENTAL STRESSES AND TICS

The duration and concentration of exposures were major con-siderations in the experimental design. Short-term exposures(i.e., 30 min) were used, because this seemed to best reflect thefield conditions that wewere trying to mimic, and short-term ex-posures also minimized confounding experimental effects thatmight be induced by growing plants for a long period within aPlexiglas box. The range of exposure concentrations was deter-mined by several practical and theoretical considerations. Forpractical reasons, the concentrations needed to be within therange that could be detected by the industrial hygiene monitorsdescribed below (0.5–200 ppm, depending on the TIC sensor).The concentrations also needed to be higher than might be en-countered incidentally under ambient conditions, but lower thanwhat might cause an acute safety risk to laboratory personnel.For theoretical reasons, we also considered what concentrationsmight be encountered within the landscape, in case of TIC re-lease in the field, based on limited reports modeling accidental

gas releases such as methyl isocyanate in the Bhopal disaster[28]. Finally, the target treatment concentrations were adjustedbased on the results of concurrent spectral analysis, to obtain arepresentative range of responses induced by the treatment con-centrations listed in Table I.Plants were exposed to each TIC within a custom-built

36 cm 36 cm 75 cm Plexiglas box, which, during expo-sures, was covered with a loose-fitting Plexiglas lid. Poplarplants were exposed individually, while up to eight wheat andcanola plants could be treated simultaneously. Two 11 cm,12 V computer fans were used to circulate the atmospherewithin the box. TICs (except for HCN, as described below)were introduced into the box through a (internal diam-eter) tube, which used a stream of house-compressed, driedand filtered air as a carrier. During mock treatments, only thecarrier gas was applied to the chamber (without any TICs).NH , SO , H S, and Cl were supplied from compressed gascylinders (lecture bottles), through a stainless steel regulator(Sigma Aldrich P/N Z148594). Because a commercial massflow controller that was suitable for corrosive gases ultimatelycould not be found, the outlet pressure on the regulator wasadjusted manually throughout the course of the treatments tomaintain the concentration of TIC within the treatment boxwithin a target range. We monitored the concentration of TICwithin each chamber throughout the duration of the exposuresby using a Drager PacIII personal safety monitor that had beenfitted with the appropriate sensor for each TIC. In some cases,the concentration of the TIC briefly exceeded the saturationlimit of the detector. Although the regulator was thoroughlypurged with dry air after each treatment, the regulator needed tobe replaced every few months, due to corrosion from the TICs.Because we were unable to purchase a convenient volume

of HCN in a compressed cylinder, plants were exposed togaseous HCN vapors as follows. Potassium cyanide crystals(KCN) (7.5 g for 50 ppm) were placed in a 200 ml glass beakerwith a magnetic stir bar, and the beaker was placed on top ofa solid-state magnetic stirrer within the treatment chamber,which also contained the plants for treatment. The lid wasplaced on the container, while 50% glacial acetic acid (v/v inwater) was dripped, with stirring, into the beaker from outsidethe chamber, through a polypropylene tube, until the targetconcentration in the chamber atmosphere was obtained.

III. SPECTRAL DATA

A. Data Collection

Spectral measurements were acquired using an AnalyticalSpectral Devices (ASD®) Fieldspec FR spectrometer. This in-strument operates from 350 to 2500 nm, where full width halfmaximum is 3 nm at 700 nm and 10 nm at 1400 and 2100 nm.The sampling interval is 1.4 nm between 350 nm and 1050 nmand 2 nm between 1000 nm and 2500 nm.Spectral measurements of leaves made use of an ASD®

leaf clip viewing a circular leaf area of 10 mm radius illu-minated with a halogen bulb emitting a color temperature of2911 10 K. Leaf reflectance was calculated by dividing leafradiance by that of a white reference (99% Spectralon re-flectance panel, Labsphere, Inc.) viewed and illuminated under


the same conditions. The white reference is on a reversible plat-form with the opposite side containing a dark panel, which hasan average reflectance of 0.025% and used as the backgroundduring leaf measurement.Poplar and canola leaves were sufficiently large to fill the

field of view of the leaf clip. However, owing to the shape ofwheat leaves, these leaves only covered a portion of the field ofview resulting in a reflectance spectrum that is a combination ofthe leaf and dark panel. This does not change the shape of thespectrum and thus, did not affect the analysis of the spectra thatis described below.The number of plants in each experiment varied between four

and nine, with the majority of experiments including six plantsper treatment and six control plants (e.g., untreated). Generallythree leaves per plant were measured. For this project we focuson healthymature leaves, which dominate each plant. For poplarand canola, a number of leaves at various stages of senescence(low, medium, high) were also measured. Measurements of newgrowth comprised only the top few leaves.The spectrum collected at each leaf location consisted of an

average of 10 scans spanning 1–2 seconds, reducing randommeasurement noise. For each canola and poplar leaf, spectrawere collected from three different locations approximatelyhalfway between the main leaf vein and the leaf edge. Forwheat, three adjacent measurements were collected along theleaf at approximately its maximum width. The measurementsfrom each leaf were then averaged accounting for spectralvariability across the leaf.Measurements were taken 24 and 48 hours after treatment.

For some experiments, such as drought, measurements weretaken at periods greater than 48 hours. If, for a given exper-iment, a high degree of stress was visually observed after 24hours, measurements at 48 hours were not taken.

B. Pre-Processing of Spectral Data

Prior to data analysis erroneous measurements were removedeither due to experimental error (e.g., leaf improperly placed inleaf clip), or instrumental errors (e.g., abnormally low signal tonoise). Spectral measurements obtained from each leaf were av-eraged to produce a representative leaf spectrum used for anal-ysis. Lastly, noisy bands below 400 nm and above 2450 nmwereremoved.

IV. DATA ANALYSIS METHODOLOGY

A. Endmember Analysis and Vegetation Indices

The responses observed following treatment were variableand in a given plant included relatively healthy leaves and leaveswith a severe response. Thus, the spectral data for each treat-ment include a representation of the variability of leaf responses.Endmember analysis was used to capture the spectral signatureof leaves with the most severe response. The premise is thatif a given TIC results in a unique spectral signature, it shouldbe visible in the leaves with the most severe response. Spec-tral endmembers are spectral extremes that lie at the vertices ofthe data volume, or simplex, in n-dimensional space. Fig. 1 il-lustrates this concept using data for poplar displaying spectraltrends from healthy mature leaves to new growth and to fully

Fig. 1. Volume of poplar data shown in two dimensions (x-axis 1394 nm, y-axis701 nm). Dotted line represents an approximation of the simplex that definesthe data volume encompassed by healthy mature, new growth, and senescenceleaves as endmembers. Each point represents the spectral values of a measure-ment. New growth A, B, C and represent new leaves sampled at the end tip ofthe plant and progressively down from that point. Secondary leaves where mea-sured from the first branch observed below the tip of the plant.

TABLE IIVEGETATION INDICES USED IN THIS STUDY

senescent leaves defining the endmembers of this system. End-members were captured with the spatial spectral endmember ex-traction tool (SSEE) [29] without the spatial component.It is also possible that key distinctions in environmental stress

and/or TIC reside in the spectral trends observed from healthy tostressed leaves. To examine this possibility the data were anal-ysed using a number of common vegetation indices that focuson bands related to leaf pigments and water content (Table II).As stated above past studies have shown that spectroscopy canbe used to detect plant stress based on changes in chlorophylland pigment concentration and water content. The indices usedincluded normalizing factors, such that the resulting trends werebased on the shape and not the amplitude of spectra.For each type of treatment all experiments that showed signif-

icant spectral changes (see Table I) were combined into a singlespectral library for subsequent analysis using vegetation indices.This was done to capture a well-defined trend from healthy tohighly stressed leaves. The associated control plants for eachtreatment were also combined into a single library. Senescedleaves for poplar and canola were included in this analysis. Veg-etation indices were calculated for each compiled library. Only


environmental stresses and treatments that resulted in notabletrends are presented.

B. Testing Spectral Significance

Spectral measurements acquired for each experiment werecompared to control samples to determine if significant spec-tral changes occurred for the stressed leaves. To determine thespectral significance of each experiment we calculated the spec-tral angle [30] between the mean reflectance spectra for thegiven experiment and the mean reflectance spectra for the con-trol plants. The spectral angle was calculated for five spectralintervals starting with the full spectral range (400–2450 nm)followed by specific intervals. These included the visible re-gion focusing on pigments (400–716 nm), a region focusing onthe red edge (717–975 nm), and finally three regions primarilyassociated with water content (976–1265 nm, 1266–1770 nm,1771–2450 nm). While defining the spectral regions of the anal-ysis we also considered the ASD spectrometer detection capa-bilities. If the angle exceeded a set threshold for any spectralrange the experiment was considered to contain significant spec-tral changes and warranted further analysis. Threshold valuesfor each spectral range were determined by calculating the meanpair-wise spectral angle for all leaves in control sets for eachspecies. Table I highlights individual experiments (measure-ment time in hours shown in bold) with significant spectral dif-ferences compared to control samples.

V. RESULTS

A. Endmember Analysis

Endmembers derived from the various experiments reflectedthe range of spectral response to the given stress. Each exper-iment presented a range of leaf responses from healthy ma-ture to the most severely stressed leaves and the range of re-sponses varied across experiments. Within experiments, spec-tral responses varied for individual plants and spatially acrosssingle leaves. Thus endmembers were derived from each exper-iment and compiled for all experiments for a given plant treat-ment (e.g., NaCl for canola). In doing so we aimed to captureextreme responses and determine if specific spectral signaturesor absorption features could be related to a given TIC. In thesection below we provide, for each species, a summary of end-member spectra for a set of treatments.Poplar: Results from the endmember analysis for the poplar

experiments showed that the drought stress and the NaCl andNH treatments had a significant effect on the reflectancespectra of leaves when compared to the mean spectra of healthyleaves and the mean spectra of senescent leaves (Fig. 2). Spec-tral changes were observed across the full spectral range usedin this analysis. In the short wave infrared (SWIR, 1000–2500nm) the predominant difference relative to a mature healthy leafwas an increase in reflectance. Absorption features attributedto cellulose and lignin, as seen near 2110 nm, become apparentdue to leaf dehydration [8]. Dehydration also enhanced narrowabsorption features near 1680, 1730, 2310 and 2350 nm, relatedto cellulose, oils, starches, and sugars [4]. In the near infrared(NIR, 700–1000 nm) the slope of the short wavelength end ofthe infrared plateau ( 750–1100 nm) decreased. The change in

Fig. 2. Endmember spectra for poplar experiments H O, NaCl and NH . Theserepresent extreme responses to the given treatment with arrows highlightingnotable absorption features. Also shown are mean spectra of control (black line,mature healthy) and senescence (purple line, high senescence) for reference.Black dotted box denotes smoothing of red edge.

spectral slope was variable but significant in all experiments. Inthe visible (400–700 nm) the key change with respect to healthyleaves was a deepening of the chlorophyll trough near 680 nmowing to increased reflectance between 550 and 650 nm, whichcan be attributed to chlorophyll (A, B) loss (chlorosis). Assuch, a number of narrow absorption features became visibleand/or enhanced (e.g., 540 and 615 nm), related to otherpigments such as ß-carotene, pycoerythrin and phycocyanin.These changes in the visible compare with those observed inthe spectra of senescent leaves.Canola: For canola experiments, not all of the applied treat-

ments (NaCl, H S, HCN,NH , Cl , and SO ) and drought stresshad a significant effect on leaf reflectance spectra. Treatmentsthat showed significant changes across the full spectral rangeincluded NH , Cl , and SO (Fig. 3), with significant changesonly occurring over a small spectral range for NaCl, H S, HCN(not shown), and drought. For NH , Cl , and SO treatments,spectral features in the visible (chlorophyll and pigments) andin the SWIR (lignin, cellulose, oils, starches, and sugars) sim-ilar to those described for poplar were observed. As observed forpoplar, the depth of a given absorption feature, the increase in re-flectance, or the slope change of the red edge were variable. Forthe NaCl, H S, HCN, and dehydration experiments spectra dis-played an increased slope between the green peak (540 nm) andchlorophyll trough and an overall minor increase in reflectancein the SWIR.Wheat: The response of wheat to the various treatments and

the drought stress was minimal compared to poplar and canola.Treatments that showed significant changes across the full spec-tral range were restricted to HCN, SO and H O (Fig. 4) withsignificant changes only occurring over a small portion of the


Fig. 3. Endmember spectra for canola experiments H O, NH , Cl , and SO .These represent extreme responses to the given treatment with arrows high-lighting notable absorption features. Also shown are mean spectra of control(black line, mature healthy) and senescence (purple line, high senescence) forreference. Black dotted box denotes smoothing of red edge.

spectral range for NaCl, NH and Cl . For HCN and SO , anddrought, similar spectral features as those described for poplarand canola were observed in the SWIR (lignin, cellulose, oils,starches, and sugars). Notable spectral changes in the visible(chlorophyll and pigments) and NIR (red edge slope) were re-stricted to the SO treatment. For the NaCl, Cl and NH treat-ments minor spectral changes included an increased slope be-tween the green peak and chlorophyll trough, and, a minor in-crease in reflectance in the SWIR.Implications of Endmember Analysis: The endmember

spectra shown in Figs. 2, 3, and 4 indicate that the TIC treat-ments and environmental stresses (NaCl and drought) showa range of known spectral absorption features common instressed vegetation. Many absorption features become visibleand/or enhanced primarily because of chlorosis (visible) anddehydration (SWIR). This is not unexpected as plant phys-iological responses to stress share many common featuresregardless of the cause of stress. The occurrence of diagnosticspectral features that can be correlated directly with a givenTIC or environmental stress was not evident from this analysis.However, depending on the level of stress experienced by agiven plant, the depth of a given absorption feature, the increasein reflectance, or the slope change of the red edge varied. Thiscan be attributed to variability in the response of the leavesto the given stress (e.g., chlorosis versus dehydration). Thus,the differences observed between the experiments reflect howa plant, or individual leaf, responds over time to the exposureof a treatment. As seen below the use of vegetation indiceshighlights spectral differences across treatments that can becaptured for populations of leaves.

Fig. 4. Endmember spectra for wheat experiments H O, HCN, Cl and SO .These represent extreme responses to the given treatment with arrows high-lighting notable absorption features. Also shown are mean spectra of control(black line, mature healthy) for reference.

B. Vegetation Indices

Eight vegetation indices (see Table II) were used to analyzespectral trends displayed by healthy mature to highly stressedgroup of plants. The objective was to determine if specific TICtreatments resulted in diagnostic trends, with respect to othertreatments and environmental stress.Poplar: Fig. 5 shows scatter plots for a series of vegetation

indices for poplar treatments NaCl and NH , drought, senes-cence, and control plants. The results clearly demonstrated thattrend lines related to NaCl, NH and senescence were distinctgiven the proper combination of indices. H O followed a trendsimilar to that seen for NaCl in most plots.Canola: Fig. 6 relates to canola treatments Cl and SO ,

drought, senescence, and control plants. The various vegeta-tion indices scatter plots showed that Cl and SO treatmentscould be distinguished from drought and senesced plants. In theplots NDVI versus Red Edge, MCARI1 versus Red Edge, andPSRI versus NPQI, trends for Cl and SO were slightly dif-ferent and distinguishable. In the plot PSRI versus SWIR1 it ap-peared that SO displayed two different trends. Thus the resultsfor individual SO experiments were examined in as shown inFig. 7. Shown are results for three experiments with 50 ppm(exp 1), 100 ppm (exp 2), and 60–120 ppm (exp 8), of SO(Table I). Experiments 1 and 2 showed a similar trend, but dif-fered in dosage. Experiment 8 showed a different trend from ex-periments 1 and 2, but had variable dosage spanning the dosageof experiments 1 and 2. Thus, dosage did not appear to havebeen the controlling factor on the trends observed between thedifferent SO experiments. Closer examination of the NDVIversus NDWI plot showed that the key difference between theSO experiments was water content (as expressed by NDWI).


Fig. 5. Vegetation indices for poplar treatments NaCl and NH , dehydration(H O), senescence, and controls plants. Refer to Table II for the description ofeach index.

Thus, we could speculate that initial water content of the leavesmay have been the controlling factor in the trends observed inthese experiments rather than the treatment type.Wheat: The results from the vegetation indices scatter plots

for wheat (Fig. 8) did not show trends that were as well-definedas for poplar and canola plants. This was not unexpected asthe wheat plants were consistently less affected by applicationof TICs. There was considerable data scatter for the wheat,however, general trends that discriminate SO , HCN, anddrought were visible in the plots NDVI versus NDWI, PRIversus NDWI, PSRI versus NDWI, NDWI versus SWIR1, andPRI versus SWIR1. In each of these cases leaf dehydrationappeared to be an important discriminating feature as expressedby NDWI and SWIR1. A slightly different trend was alsovisible between drought and SO and HCN in the PRI andSWIR1 plot. Leaves treated with Cl did not show a responseas extreme as for HCN, SO , and drought and did not revealany particular trend.Indices Across Species: Trends displayed by scatter plots of

vegetation indices, with respect to each species, showed thatit was possible in many cases to distinguish between environ-mental stresses and TICs, and in some cases between TICs.Given that a potential practical application of this work may in-volve detection in natural landscapes of mixed plant species, thecombined results for poplar, canola, and wheat were examined.Fig. 9 shows a selection of vegetation indices across all

species for treatments NaCl, NH , SO , Cl , HCN, and,drought, senescence and controls plants. In Fig. 9 plot A

Fig. 6. Vegetation indices for canola treatments Cl and SO , dehydration(H O), senescence, and controls plants. Refer to Table II for the descriptionof each index.

Fig. 7. Vegetation indices PSRI versus SWIR1 and NDVI versus NDWI forcanola treatments Cl and SO , and controls plants. Results for individual SOexperiments are displayed. Refer to Table II for the description of each index.

(NDVI versus NDWI) it was difficult to determine trendsthat were consistent across species. The key reason is that thecontrol samples for each species defined distinct clusters. Thus,trends from healthy to highly stressed leaves for each specieshave different starting points, which causes overlap with otherdata trends complicating the discrimination of treatments andstresses. For the remaining plots the control samples for thethree species overlap to varying degree, and the trends ob-served have a similar starting point. In plots where the controlplants cluster more it is evident that the environmental and TICtreatments show distinctive trends that are not a function ofthe plant species. This is most evident in plots E-F, where acombination of indices distinguishes NH , SO , and Cl (fromeverything else) but not HCN. In the case of NaCl and H Othe two trends tend to overlap in each example. This is notunexpected as NaCl induces ionic stress and an osmotic stressresponse that is common to drought. It is important to note


Fig. 8. Vegetation indices for wheat treatments Cl , HCN, SO , dehydration(H O), and controls plants. Refer to Table II for on the description of each index.

that not all species included all TIC treatments, and as such,it is possible that differences could occur across species for agiven TIC. However, based on the results accumulated in thisstudy and shown in Fig. 9 it appeared that an appropriate set ofvegetation indices (where control samples overlap) allows forthe discrimination between environmental stresses and TICs,and in some cases between TICs.

VI. DISCUSSION

The larger goal of this research program is to develop a re-mote detection and monitoring capability for hazardous eventssuch as airborne toxic gas leaks. This study was conducted atthe leaf level to ensure good experimental control. We notedthat variation in the spectral properties of individuals of somespecies could be detected within treatment groups and betweenreplicate treatments (e.g., SO canola). Although the sources ofthis variation is unknown, we speculate that a major componentcan be attributed to starting conditions of the plants, such asdifferences in water status of individual plants resulting fromthe sum of minor variations in ambient humidity, respirationrate, stomatal opening, soil evaporation, time since watering,light intensity (which decreases as fluorescent tubes age), andminor stochastic events. Because all plants in this study weregrown under nominally identical conditions within controlledenvironment chambers, we expect that in the field, differencesbetween individual plants could be magnified. To determinewhether the net effect of this variation would be significant out-doors at the canopy level requires further experimentation. De-

Fig. 9. Vegetation indices for all species and treatments NaCl, NH , SO , Cl ,HCN, dehydration (H O), senescence and controls plants. Refer to Table II foron the description of each index.

tection of chemical effects on vegetation from an airborne plat-form will likely also be impacted by natural variability in watercontent and other factors over a given area. Thus, the resultsfrom this analysis should represent a reasonable approximationof natural conditions.The duration and concentration of toxic industrial chemical

exposures were also designed to reflect natural field concentra-tions based on practical and theoretical considerations. How-ever, owing to the variable responses of each species, numerousexperiments using different dosages were necessary to invokethe desired plant stress level responses. The result was that anumber of experiments showed little or no spectral responseto exposure, whereas in other cases the response was extreme.These upper and lower limits can be considered to reflect vari-ability in gas levels spatially distributed across a region, whichwould likely be encountered using a detection campaign froman airborne platform. Note also that because of the time require-ments of each experiment, duplicates of each experiment at agiven dosage were not always possible. Thus, in some cases wewere not able to verify, through duplication, spectral responsesto some TICs.Detection for a TIC release in the natural environment would

also be confronted with a range of growth stages for a givenspecies. This variability would be inherent to the analysis ofdata acquired by an airborne platform. However, it is reasonableto expect that during an airborne acquisition of spectral data aparticular growth stage (e.g., healthy mature) would dominate


at a given time. The exception will be at times of the year suchas spring and fall where plants species experience significantphysiological changes (e.g., new growth or senescence).We detected marked differences in physiological and spec-

tral responses for combinations of species and TICs. Differentresponses might be attributed in part to inherent differences inthe structure and function of each species (e.g., stomatal den-sity, cuticle thickness, leaf architecture, cell wall composition,growth and respiration rate), or the existence and efficiency ofspecific metabolic pathways for detoxifying TICs. Whatevertheir source, the existence of species-specific responses of veg-etation to TICs presents both a challenge and an opportunity forregional remote sensing. For the plants of this study we wereable to discern different spectral trends for NH , SO , Cl andthus to distinguish these TICs. The trends result from the vari-able leaf response within plants, between plants and betweenspecies and it is expected that much of the variability observedwithin species would be preserved or even enhanced in nature.As such, the results are encouraging for the possible detectionof TIC effects on natural vegetation using airborne imagery.However the variability introduced from species to species thatalso enhances the TIC spectral trends would be a function of thelandscape investigated and encompassed by a given data set (orgeographical area) analyzed.The larger goal is to develop a remote detection and moni-

toring capability for hazardous events such as a toxic gas leak.Our findings at the leaf level suggest that damage can be de-tectedwithin 48 hrs and should last for an extended period, prob-ably up to a week. Thus, the next experimental step is to test ifthe results shown here at the leaf level can also be detected withairborne and satellites systems. Many airborne systems are nowavailable that include both high spectral (e.g., 10 nm) and spa-tial (e.g., 1–4 m) resolution, which are likely to have the bestprobability of replicating the results from this study. Existingsatellites (Landsat, ASTER, EO-1 ALI) may not be good can-didates as they lack the necessary detailed spectral information.The EnMAP hyperspectral satellite currently set to launch in2015 may be appropriate given that it will offer 30 m pixels, 220bands at 10 nm bandwidth spanning the 420–2450 nm spectrum,and a revisit time of 4 to 23 days [37]. Because we developeda detection methodology from leaf observations there is an im-portant knowledge gap that needs to be addressed using fieldtrials to test if the findings of this study can extend to the detec-tion of leaks in the natural environment. The principal unknownis the effect of varying vegetation canopy structural parameters(e.g., canopy gaps, leaf area) and background properties (litterand soil reflectance) on the specific TIC data trends that wereidentified.

VII. CONCLUSION

This research examined the spectral response of indi-vidual leaves of three common Canadian plant species: poplar(Populus deltoides, Populus trichocarpa); wheat (Triticumaestivum); and canola (Brassica napus). These plants weresubjected to fumigation with gaseous phase TICs. The twoobjectives of this study were to determine if: 1) vegetationsubjected to TICs could be distinguished from backgroundvegetation during varying growth stages (new growth to senes-

cence) and environmental stresses; and 2) different TICs couldbe distinguished based on the spectral response of vegetation.Treatments were designed to allow quantification of the vari-ation in spectra that might be expected due to environmental,developmental, and stochastic effects on the physiologicalstate of individual plants within each species. All plants weregrown in controlled environment chambers at the University ofAlberta, using standardized conditions.The results showed that both environmental and TICs in-

duced similar spectral features inherent to plants that can berelated primarily to chlorophyll and water loss. These includepigments in the visible and cellulose, lignin, lipids starches, andsugars in the SWIR. Although no specific spectral features couldbe tied to individual TICs an analysis of the data using vege-tation indices, which focused on key spectral bands associatedwith chlorophyll, pigments and water content, showed that theTICs and environmental stresses result in diagnostic trends fromhealthy mature to highly stressed leaves. In addition, furtheranalysis showed that combinations of specific indices could beused to distinguish the effects of NH , SO , Cl and their effectfrom that of other treatments of the study.

REFERENCES[1] F. S. Chapin, “Integrated responses of plants to stress,” Bioscience, vol.

41, pp. 29–36, Jan. 1991.[2] H. K. Lichtenthaler, “The stress concept in plants: An introduction,” in

Stress of Life—From Molecules to Man, P. Csermely, Ed., 1998, vol.851, pp. 187–198.

[3] H. W. Gausman and V. R. Weidner, Plant Leaf Optical Parameters inVisible and Near-Infrared Light. Lubbock, TX: Texas Tech Univer-sity Press, 1985, vol. 29.

[4] P. J. Curran, “Remote-sensing of foliar chemistry,” Remote Sens. Env-iron., vol. 30, pp. 271–278, Dec. 1989.

[5] G. A. Carter and A. K. Knapp, “Leaf optical properties in higher plants:Linking spectral characteristics to stress and chlorophyll concentra-tion,” Amer. J. Botany, vol. 88, pp. 677–684, Apr. 2001.

[6] G. A. Carter and R. L. Miller, “Early detection of plant stress by dig-ital imaging within narrow stress-sensitive wavebands,” Remote Sens.Environ., vol. 50, pp. 295–302, Dec. 1994.

[7] G. A. Carter, “Responses of leaf spectral reflectance to plant stress,”Amer. J. Botany, vol. 80, pp. 239–243, Mar. 1993.

[8] G. A. Carter, “Primary and secondary effects of water-content on thespectral reflectance of leaves,” Amer. J. Botany, vol. 78, pp. 916–924,Jul. 1991.

[9] G. A. Blackburn, “Quantifying chlorophylls and caroteniods at leaf andcanopy scales: An evaluation of some hyperspectral approaches,” Re-mote Sens. Environ., vol. 66, pp. 273–285, Dec. 1998.

[10] J. A. Gamon and J. S. Surfus, “Assessing leaf pigment content andactivity with a reflectometer,” New Phytologist, vol. 143, pp. 105–117,Jul. 1999.

[11] P. J. Zarco-Tejada, A. Berjon, and J. R. Miller, “Stress detectionin crops with hyperspectral remote sensing and physical simulationmodels,” presented at the Airborne Imaging Spectroscopy Workshop,Bruges, Belgium, Oct. 8, 2004.

[12] J. L. Baltzer and S. C. Thomas, “Leaf optical responses to light and soilnutrient availability in temperate deciduous trees,” Amer. J. Botany,vol. 92, pp. 214–223, Feb. 2005.

[13] G. A. Blackburn, “Hyperspectral remote sensing of plant pigments,” J.Experimental Botany, vol. 58, pp. 855–867, Mar. 2007.

[14] K. L. Smith, M. D. Steven, and J. J. Colls, “Plant spectral responsesto gas leaks and other stresses,” Int. J. Remote Sens., vol. 26, pp.4067–4081, Sep. 2005.

[15] K. L. Smith,M. D. Steven, and J. J. Colls, “Use of hyperspectral deriva-tive ratios in the red-edge region to identify plant stress responses to gasleaks,” Remote Sens. Environ., vol. 92, pp. 207–217, Aug. 2004.

[16] M. F. Noomen, A. K. Skidmore, F. D. van der Meer, and H. H. T. Prins,“Continuum removed band depth analysis for detecting the effects ofnatural gas, methane and ethane on maize reflectance,” Remote Sens.Environ., vol. 105, pp. 262–270, Dec. 2006.


[17] E. J. Male, W. L. Pickles, E. A. Silver, G. D. Hoffmann, J. Lewicki,M. Apple, K. Repasky, and E. A. Burton, “Using hyperspectral plantsignatures for CO leak detection during the 2008 ZERT CO seques-tration field experiment in Bozeman, Montana,” Environmental EarthSciences, vol. 60, pp. 251–261, Mar. 2010.

[18] M. Kraft, H. J. Weigel, G. J. Mejer, and F. Brandes, “Reflectance mea-surements of leaves for detecting visible and non-visible ozone damageto crops,” J. Plant Physiol., vol. 148, pp. 148–154, Apr. 1996.

[19] M. Gab, K. Hoffmann, M. Lobe, R. Metzger, S. van Ooyen, G. Elbers,and B. Kollner, “NIR-spectroscopic investigation of foliage of ozone-stressed Fagus sylvatica trees,” J. Forest Res., vol. 11, pp. 69–75, Apr.2006.

[20] M. Meroni, V. Picchi, M. Rossini, S. Cogliati, C. Panigada, C. Nali,G. Lorenzini, and R. Colombo, “Leaf level early assessment of ozoneinjuries by passive fluorescence and photochemical reflectance index,”Int. J. Remote Sens., vol. 29, pp. 5409–5422, 2008.

[21] J. A. Gamon, J. Penuelas, and C. B. Field, “A narrow-waveband spec-tral index that tracks diurnal changes in photosynthetic efficiency,” Re-mote Sens. Environ., vol. 41, pp. 35–44, Jul. 1992.

[22] J. H.Williams and T.W.Ashenden, “Differences in the spectral charac-teristics of white clover exposed to gaseous-pollutants and acid mist,”New Phytologist, vol. 120, pp. 69–75, Jan. 1992.

[23] J. B. Schutt, R. A. Rowland, and H. E. Heggestad, “Identification of in-jury resulting from atmospheric pollutants using reflectance measure-ments,” J. Environ. Quality, vol. 13, pp. 605–608, 1984.

[24] F. M. Gemmell and J. J. Colls, “The effects of sulfur-dioxide on thespectral characteristics of leaves of VICIA-FABA L,” Int. J. RemoteSens., vol. 13, pp. 2547–2563, Sep. 1992.

[25] B. F. T. Rudorff, C. L.Mulchi, E. H. Lee, R. Rowland, and C. Daughtry,“Effects of O-3 and SO on leaf characteristics in soybeans grownunder ambient- and enriched-carbon dioxide atmosphere,” inAir Toxicsand Water Monitoring, G. M. Russwurm, Ed., 1995, vol. 2503, pp.89–100.

[26] K. Shinozaki and K. YamaguchiShinozaki, “Gene expression andsignal transduction in water-stress response,” Plant Physiol., vol. 115,pp. 327–334, Oct. 1997.

[27] H. Knight and M. R. Knight, “Abiotic stress signalling pathways:Specificity and cross-talk,” Trends in Plant Science, vol. 6, pp.262–267, Jun. 2001.

[28] M. Sharan and S. G. Gopalakrishnan, “Bhopal gas accident: A numer-ical simulation of the gas dispersion event,” Environ. Model. Softw.,vol. 12, pp. 135–141, 1997.

[29] D. Rogge, B. Rivard, J. Zhang, A. Sanchez, J. Harris, and J. Feng,“Integration of spatial-spectral information for the improved extractionof endmembers,” Remote Sens. Environ., pp. 287–303, Oct. 15, 2007.

[30] J. C. Price, “How unique are spectral signatures,” Remote Sens. Env-iron., vol. 49, pp. 181–186, Sep. 1994.

[31] J. W. Rouse, R. H. Haas, J. A. Schell, and D. W. Deering, “Monitoringvegetation systems in the great plains with ERTS,” presented at the 3rdEarth Resources Technology Satellite-1 Symp., Greenbelt, MD, 1975.

[32] D. Haboudane, J. R. Miller, E. Pattey, P. J. Zarco-Tejada, and I. B.Strachan, “Hyperspectral vegetation indices and novel algorithms forpredicting green LAI of crop canopies: Modeling and validation in thecontext of precision agriculture,” Remote Sens. Environ., vol. 90, pp.337–352, Apr. 2004.

[33] M. N. Merzlyak, A. A. Gitelson, O. B. Chivkunova, and V. Y. Rak-itin, “Non-destructive optical detection of pigment changes during leafsenescence and fruit ripening,” Physiologia Plantarum, vol. 106, pp.135–141, May 1999.

[34] J. D. Barnes, L. Balaguer, E. Manrique, S. Elvira, and A. W. Davison,“A reappraisal of the use of DMSO for the extraction and determinationof chlorophylls-a and chlorophylls-b in lichens and higher-plants,” En-vironmental and Experimental Botany, vol. 32, pp. 85–100, Apr. 1992.

[35] G. Guyot and F. Baret, “Utilisation de la haute resolution spectralepour suivre l’etat des couverts vegetaux (Utilisation of the high spec-tral resolution to follow the status of the covered vegetation),” in 4thInt. Colloq. Spectral Signatures of Objects in Remote Sensing, Aussois,France, 1988, pp. 279–286.

[36] B. C. Gao, “NDWI—A normalized difference water index for remotesensing of vegetation liquid water from space,” Remote Sens. Environ.,vol. 58, pp. 257–266, Dec. 1996.

[37] T. Stuffler, C. Kaufmann, S. Hofer, K. P. Forster, G. Schreier, A.Mueller, A. Eckardt, H. Bach, B. Penne, U. Benz, and R. Haydn,“The EnMAP hyperspectral imager—An advanced optical payloadfor future applications in Earth observation programmes,” Acta Astro-nautica, vol. 61, no. 1–6, pp. 115–120, 2007.

Derek Rogge received the Ph.D. degree in the De-partment of Earth and Atmospheric Sciences, Uni-versity of Alberta, Edmonton, Canada, in 2007.He is currently a research scientist with the

Applied Spectroscopy Group at the German RemoteSensing Data Center, German Aerospace Center(DLR), Wessling, Germany. His present researchinterests include algorithm development related toendmember extraction, spectral unmixing, imple-mentation of existing algorithms for large multiflight-line hyperspectral data sets, and hyperspectral

flight-line leveling. Applied remote sensing interests include geological appli-cations, such as regional mapping, target mapping for mineral exploration, andmine waste and environmental monitoring.

Benoit Rivard received the Ph.D. degree in earth andplanetary sciences from Washington University, St.Louis, MO, in 1990.He is currently a Professor in the Department of

Earth and Atmospheric Sciences of the University ofAlberta, Canada. He is a geologist with particular in-terest in the development of applied geological re-mote sensing. His key research preoccupation is todevelop the analysis of hyperspectral sensing (field,airborne and spaceborne) to improve the effective-ness of oil/mining industry, and mapping agencies

to delineate and manage their targeted resources. In this respect he has beenworking to: 1) automate the hyperspectral analysis of rock cores and wall rocktoward mineral mapping and rock type classification; 2) improve the analysis ofhyperspectral imagery for northern regions that are remote and difficult to ac-cess; and 3) to improve the analysis of hyperspectral data for boreal and tropicalforests. Past research interests have included the use of Radar remote sensingfor lithologic and structural mapping, and the development of methodologiesfor precise measurement of emissivity.

Michael Deyholos received the Ph.D. degree inplant molecular biology from McGill University,Montreal, Canada, in 2000.He is currently an Associate Professor in the

Department of Biolgoical Sciences of the Universityof Alberta, Canada. His research interests includethe cellular responses of plants to abiotic stresses,including water-deficits and ionic shocks. He worksprincipally with Arabidopsis and with crop plantsincluding canola, flax, and various cereals. Thelab applies techniques of genomics (transcriptome

analysis and whole genome sequencing) and metabolomics (FT-IR, GC-MS,LC-MS) to analysis of plant cells.

Josee Levesque received the B.Sc. degree in ge-ography from Université du Québec à Montréal,Canada, the M.Sc. degree in geography and remotesensing from Université de Sherbrooke, Canada, andthe Ph.D. degree in earth sciences from CarletonUniversity, Ottawa, Canada in 1989, 1993, and 2001,respectively.She has published more than 40 papers in the area

of optical remote sensing applied to environmentalgeology and vegetation mapping. She is currently aresearch scientist at Defense Research and Develop-

ment Canada (DRDC) and an Associate Professor at Laval University in QuébecCity, Canada. Her more recent work revolves around the use of hyperspectralremote sensing to improve knowledge of the terrain in a military context and tomonitor environmentally sustainable training on Canadian military bases.


Jean-Pierre Ardouin received the M.A.Sc. degreein electrical engineering (ultrasonic image pro-cessing) from the University of Toronto, Canada, in1985.He joined Defence R&D Canada (DRDC) in 1986

as a Defence Scientist and is currently Group Leaderfor the Sensor Data Processing group. He has beeninvolved in many image exploitation and spectrora-diometric measurements projects from the visible tothe thermal infrared region. Since 1998 he has beenmore specifically conducting R&D work in hyper-

spectral remote sensing, and from 2004 to 2011 he was Project Manager forDRDC HYperspectral iMage Exploitation (HYMEX) technology demonstra-tion project aimed at demonstrating the military utility of solar reflective hyper-spectral imagery to the Canadian Forces.

Anthony A. Faust received the B.Sc. (Honours)from the University of Victoria, Canada, in physicsand mathematics in 1994, and the Ph.D. degree inphysics from the University of Alberta, Canada,in 1999, specializing in experimental high energyparticle physics.He is head of the Explosives Detection Group at

Defence R&D Canada, Suffield, has been working inthe field of explosive hazard detection for the Cana-dian Forces since 1999. His principal research hasfocused on the development of nuclear-based tech-

niques used to confirm the explosive content of buried explosive hazards.

Date post:	24-Nov-2023
Category:	Documents
Upload:	independent
View:	1 times
Download:	0 times

Potential Discrimination of Toxic Industrial Chemical Effects on Poplar, Canola and Wheat,...

Documents