Plumbum contamination detecting model for ...

Plumbum contamination detecting model for agricultural soil using hyperspectral data Xiangnan Liu*a Fang Huangb Ping Wangb School of Information Engineering, China University of Geosciences, Beijing,100083, China; b School of Urban and Environmental Sciences, Northeast Normal University, Renmin Street 5268, Changchun 130024, China a

ABSTRACT The issue of environmental pollution due to toxic heavy metals in agricultural land has caused worldwide growing concern in recent years. Being one of toxic heavy metals, the accumulation of Plumbum (Pb) may have negative effects on natural and agricultural vegetation growth, yield and quality. It can also constitute short-term and long-term health risks by entering the food chain. In this study, we analyze the relationships between physical and chemical characteristics, biological parameters of soil-vegetation system and hyperspectral spectrum responses systematically. The relation between hyperspectral data and the biological parameters of Pb polluted wheat canopy such as leaf pigments, leaf moisture, cell structure and leaf area index (LAI) are discussed. We detect the changes in the wheat biological parameters and spectral response associated with Pb concentration in soil. To reveal the impact mechanisms of Pb concentration on agricultural soil, six models including chlorophyll-leaf moisture model, chlorophyll-cell structure model, chlorophyll-LAI model, leaf moisture-cell structure model, leaf moisture-LAI model, cell structure- LAI model are explored. We find that changes in Pb concentration present various features in different models. Pb contamination in agricultural soil can be identified and assessed effectively while integrating the characteristics of those developed models. Keywords: Plumbum contamination model, crop, hyperspectral remote sensing, multiple discriminant analysis 1.

INTRODUCTION

Soils form the interface between atmosphere and surface waters to groundwater systems. Soil is also as a storage and source media of pollutants including heavy metals and organic compounds. Nowadays, the non-point source contamination in agricultural field is one of vital ecological environment issues. It has threatened global environment quality, human being subsistence and food security as well[1]. Traditionally, screening or monitoring the non-point source contamination in agricultural land relies on chemical analyses by well-established and accurate laboratory techniques (for instance the AAS, ICP-MS, chromatography etc.), mostly based on point sampling[2]. However, these methods are time consuming and costly. Owing to the wide spatial coverage and internal consistency of data sets, remote sensing techniques will help to investigate the spatial distribution of contaminants and evaluate the contaminated soil in large area rapidly and effectively[3-5]. Note that the non-point source contamination in agricultural soil is affected by not only the physical variations (i.e. geographical, climatic, hydrological and geological conditions) but also the diversity of pollutants sources (i.e. fossil fuel, coal combustion, industrial effluents, solid waste disposal, fertilizers and mining and metal processing)[6]. These factors may interact with each other and involve stochastic process and effects, which cause the remotely sensed information mechanism of contaminated soil more complicated. Meanwhile, soil polluted may not be distinguished by externally visible characteristics since the soil contamination information is sort of accumulated tiny variable in a large area. Pollution of agricultural field can be characterized by complexity and persistent concealment. The uncertainty and stochastic characteristic of contaminated soil information may lead to obvious difficulties in pollution identification[7]. Heavy metals such as Fe, Cu, Zn, Ni, Pb, Cd and other trace elements are important for proper functioning of biological systems. In past a few decades, due to the increasing industrial waste emission, polluted water irrigation and overuse of pesticide and fertilizer, agricultural soil contaminated by those heavy metals including Pb has been exacerbating and *[email protected]

Geoinformatics 2008 and Joint Conference on GIS and Built Environment: Monitoring and Assessment of Natural Resources and Environments, edited by Lin Liu, Xia Li, Kai Liu, Xinchang Zhang, Yong Lao Proc. of SPIE Vol. 7145, 71450P · © 2008 SPIE · CCC code: 0277-786X/08/$18 · doi: 10.1117/12.813002 Proc. of SPIE Vol. 7145 71450P-1 2008 SPIE Digital Library -- Subscriber Archive Copy

spreading. The toxic metal entering the ecosystem may lead to geoaccumulation, bioaccumulation and biomagnification. The potential accumulation of Pb in biosystems through contaminated soil, water and air may not only have negative effects on natural and agricultural vegetation growth, yield and quality but also constitute short-term and long-term health risks by entering the food chain[8]. Threats to soil beyond the limits standards will cause an immense damage in various aspects, affecting life quality of society and the social and economic development. The issue of environmental pollution due to toxic heavy metals in agricultural land due has begun to cause worldwide growing concern. Researches on heavy metals absorption and transfer mechanisms of plant in polluted soils, rehabilitation of soil contaminated by heavy metals and soil improvement have been carried out in recent years. Yet how to discover and evaluate the widearea agricultural soil polluted by heavy metals quickly has not been reported effectively. Vegetation is regarded as a sentinel to indicate contamination presence or grade since plants hardly maintain normal growing status in polluted soil[9]. Hyperspectral data can describe the sensitive indicator of soil-vegetation system. To explore the quantitative remote sensing models based on hyperspectral data of the soil-vegetation system is a promising approach of wide-area soil contamination detection. Previous study indicates that plant growth will be influenced while the concentrations of heavy metals in soils reach certain values. The vegetation in acid-alkali or heavy metal toxic condition will lead to the physiological and ecological changes, thereafter vegetation spectrum response varies. The aims of this study include (1) to analyze the remote sensing information mechanism of heavy metal contaminated agricultural soil systematically; and (2) to propose the models of heavy metal detection for agricultural land based on hyperspectral data. 2.

SPECTRAL PARAMETER SELECTION AND MULTIPLE DISCRIMINANT ANALYSIS METHOD

In general, there seems no apparent difference in agricultural soil polluted by heavy metal (e.g. Pb). However, crop will be in poisoning due to the excessive absorption of these metals, thereafter the toxic accumulation will result in changes in the biophysical characteristics including chlorophyll content, leaf moisture, leaf area index (LAI) and leaf cell structure etc. These variations might be better reflected using hyperspectral data, which is regarded as the theoretical basis of pollution detection by remote sensing in agricultural land. To explore the mechanism of contaminated soils by hyperspectral remote sensing, we employ the reflectance in specific bands and some spectral variables of plant in this study. The selected bands range from 10 (447 nm) to 57 (925nm), 79 (932 nm) to 115 (1296 nm), 135 (1498 nm) to 163(1780 nm), and 185 (2002 nm) to 224 (2395 nm) with the exception of those atmospheric water vapor absorbance bands between 1400nm and 1900nm. Table 1 shows those spectral variables, which are potentially sensitive to chlorophyll content, leaf moisture and leaf cell structure. Theoretically, these mentioned narrow bands and variables might reveal the biophysical changes of crop to some extent. But the relevant questions are required to be answered, including which spectral bands and variables are sensible and effective to detect Pb pollution in wheat field? How to select a set of defensible, operational indicators from these voluminous parameters? In this paper, we adopt a multiple discriminant analysis (MDA) method to filter these bands and variables which are sensitive to wheat chlorophyll, wheat leaf moisture, wheat cell structure and wheat LAI. MDA approach follows the similar principle and way of principle component analysis method. In a multiple dimension variable space, the row data in the matrix are transformed into spatial point data. Then a linear combination is performed based on the predicting variable. The discriminant function is defined as the variable with best performance of separation to obtain this matrix. To select the most suitable discriminant function which can identify the variations in wheat chlorophyll, wheat leaf moisture, wheat cell structure and wheat LAI polluted by Pb, the MDA approach are applied to the analyze those various reflectance and spectral index. The maximum distance between most alike groups are defined. The F probability is regarded as the discriminant criteria, after that those variables with low relationship with Pb contamination in wheat field are eliminated. 3.

THE MECHANISM OF PLUMBUM CONTANMATION INDENTIFYING BY REMOTE SENSING FOR AGRICULRAL SOIL

3.1. Reflectance spectrum of crop Generally, there are many crests and troughs of the crop reflectance curve like natural vegetation affected by different factors such as leaf pigments and water absorption and so on (Fig.1). The visible regions of the spectrum, namely 350490nm and 650-700nm are the high absorption sections of leaf carotene and chlorophyll. Because the absorption of chlorophyll-a occur at 680nm and 700nm, the troughs of reflectance (i.e. red trough) can be found. In the infrared

Proc. of SPIE Vol. 7145 71450P-2

regions of the spectrum, the reflectance increases abruptly and dramatically in the spectra between 700nm and 750nm ( i.e. red edge). Owing to the weak absorption of pigments and liquid water, together with multiple reflections and scattering due to leaf cell structure, the reflectance between 750 and 1300nm maintain high values with an undulating trend. Obvious wave troughs of the reflectance occur at two regions i.e. 1300-1600nm (1400nm) and 1830-2008nm (1900nm), while the wave crests could be found at 1600-1830 nm (1650 nm) and 2008-2350nm (2200nm). Table.1 The selected spectral parameters associated with crop biophysical characteristics for pollution detection Representation

Variables Depth of the 671 nm absorption

Chlorophyll

band (Depth671)

763 nm)

[10]

[(ρ701nm-ρ671nm)-0.2(ρ701nm-

Daughtry et al.

in Reflectance Index(MCARI)

ρ549nm)](ρ701nm/ρ671nm)

[11]

(ρ864nm-ρ671nm)/(ρ864nm+ρ671nm)

Rouse et al. [12]

Normalized Difference Vegetation Index (NDVI) (RVSI) Normalized pigments chlorophyll ratio index (NPCI) Depth of the 983nm absorption

[13] Blackburn G

(R680 - R460)/ (R680 + R460) Continuum removal method (edges at 933 nm and

A.[14] Clark & Roush

1094 nm; 1094 nm and 1286 nm)

[10]

Continuum removal method（1094 nm and 1286

Clark & Roush

bands (Depth1205)

nm)

(LWVI) Normalized Difference Water Index (NDWI)

Apan et al.[16]

(ρ864nm-ρ1245nm)/(ρ864nm+ρ1245nm)

Gao [17]

First Derivative and Red Edge

Savitzky-Golay smoothing method (691 nm to

Position (FDREP)

763 nm)

Red-edge Vegetation Stress Index (RVSI)

Galva˜o[15]

(ρ803nm+ρ549nm)/(ρ1659nm+ρ681nm)

R870/R950

Index (NDVI)

[10] Leˆnio Soares

(ρ1094nm-ρ893nm)/(ρ1094nm+ρ983nm)

Water index (WI)

Normalized Difference Vegetation

(LAI)

Merton Huntington

Depth of the 1205nm absorption

Disease Water Stress Index (DSWI)

Leaf area index

((ρ712nm+ρ752nm)/2)-ρ732nm

bands (leaf liquid water)(Depth983)

Leaf Water Vegetation Index

Vegetation structure

Reference Clark & Roush

Modified Chlorophyll Absorption

Red-Edge Vegetation Stress Index

Water content

Equation Continuum removal method (edges at 569 nm and

(ρ864nm-ρ671nm)/(ρ864nm+ρ671nm) ((ρ712nm+ρ752nm)/2)-ρ732nm

Ogunjemiyo, S[18] Tsai & Philpot [19] Rouse et al. [12] Merton & Huntington [13]

Ratio vegetation index (RVI)

(ρ864nm/ ρ671nm)

Wiegand, C.[20]

Ratio vegetation index (RVI)

(ρ864nm/ ρ671nm)

Wiegand, C.[20]

Note：ρ is the reflectance. In the equations, the wavelengths indicated are from Hyperion bands.

3.2. Spectral response to the leaf pigments of Pb contaminated wheat In the process of photosynthesis, chlorophyll-a, chlorophyll-b and carotenoids (carotene and lutein) of green vegetation absorb light and perform photosynthesis, of which chlorophyll-a has direct relation with the efficiency of slight energy use. The responses of leaf and canopy reflectance and transmission on photosynthetic pigments can be regarded as the important indicators to not only monitor photosynthesis and nitrogen but also evaluate moisture and the stress of disease and pollution. Therefore, the chlorophyll content may be calculated by measuring the reflectance, transmission and absorption of leaf. Generally, there are differences in three spectrum regions of vegetation namely blue band, red band with intensive absorption due to chlorophyll and shortwave infrared band associated with water absorption. High absorbance of chlorophyll results in the troughs in visible regions of hyperspectral spectrum of healthy green vegetation. The visible effect of green color might weaken in case the concentration of leaf pigments decrease. “Red edge” is the most significant characteristics of vegetation spectrum. It occurs between 680nm and 750nm, locating in the biggest


slope of vegetation reflectance curve. The reflectance ranges from the absorbance trough in red band to the crests in near infrared regions.

Reflectance/%

near infrared reflectance shoulder

reflectance crest

red edge

_,/__\\'¼/:

reflectance

green peak

water absorption trough

red trough

350

550

750

' *w&&Gcrest

950

1150

1350

water absorption trough

750

1550

wavelength (nm)

1950

2150

2350

Fig.1 Reflectance spectrum of wheat

0.10

0.50

0.09

0.40

NDVI

Red2（660-681nm）

Pb is apparently not a necessary element for crop growth. When Pb enters and is absorbed by wheat, the crop will have some poisoning symptoms. Low Pb concentration will disturb the metabolic process of wheat, and the growth will be restrained. Severe Pb poisoning in wheat will lead to death. Pb does a lot of harm to the wheat, especially the intensity of plant photosynthesis and transpiration. With Pb contamination aggravating, the activity of plant and transpiration will weaken gradually. Pollutants damage the chloroplast particularly chlorophyll-a, thus plant photosynthesis will lower. As a consequence of increasing Pb concentration, chlorophyll content of wheat will descend. The destroying mechanism of chlorophyll in wheat by heavy metals including Pb may summarize the following types: 1) when entering the plant leaf, the heavy metals will accumulate locally, and they destroy the structure and activities of chloroplast directly by combining the active-groups i.e. -SH of proteins or substituting useful elements such as Fe, Zn, Cu and Mg. and so on ;2) the heavy metals disturb the absorption and transport of essential Fe, Zn, Cu and Mg by antagonism indirectly, which block the foliage transportation of those nutritious component thereafter chlorophyll synthesis is obstructed; 3) the heavy metals strengthen enzyme activity of chlorophyll and the decomposition of chlorophyll will accelerate, in consequence the chlorophyll content decreases.

0.08

0.30

Y = −0.0173X + 0.5817

Y = 0.0954X + 5.275 r = +0.8482

0.07

0.20

N = 90

N = 90

Sigmif.level = 0.01

Sigmif.level = 0.01 0.06 0.10

0.20

r = −0.7102

0.30 0.40 0.50 (a) Red edge((691-763nm)

0.10 0.06

0.08 0.10 0.12 0.14 0.16 0.18 0.20 Depth671 (b)

Fig.2 M DA discriminant models (a) the relationship of reflectivity of Red edge(691-763nm) and reflectivity of Red2(660-681nm) (b) the relationship of depth of the 671 nm absorption band and Normalized Difference Vegetation Index (NDVI)


On the basis of above analysis, we select the following spectral bands and parameters to investigate the response on the changes in pigments of wheat contaminated by Pb. The reflectance in seven spectrum regions include Blue (447-498nm (B10-B15)), Green1(508-528nm(B16-B19)), Green2(538-569nm(B20-B22)),Green3(B579-B599nm(23-25)),Red1(610650nm (B26-B30)), Red2 (660-681nm (B31-B33)) and Red edge(691-763nm (B34-B41)). Five indexes are also employed such as depth of the 671 nm absorption band (Depth671), Modified Chlorophyll Absorption in Reflectance Index (MCARI), Normalized Difference Vegetation Index (NDVI), Red-Edge Vegetation Stress Index (RVSI) and Normalized Pigments Chlorophyll Ratio Index (NPCI). The relationship between seven reflectance and Pb contamination, five indexes and Pb contamination are established respectively before the multiple discriminant analysis (MDA) is performed. For the calculation convenience, the actual Pb concentration ranging from 0 to 5000mg/kg are converted to contamination data from 0 to 5. The range of 0-1 represents the Pb concentration of 0-1000mg/kg, and the values between 1 and 2 are assigned to the 1000-2000mg/kg intervals. The rest may be deduced by analogy. Several discriminant functions are obtained while using MDA method. Fig. 2 shows the discriminant result plots for Red edge, Red2, Depth671 and NDVI. Deriving from the various functions, the response coefficients of the above mentioned seven spectral bands and indices corresponding with Pb concentration are illustrated respectively in Fig. 3 (a) and (b). From Fig.3 (a), the response coefficients lines from the highest to the lowest, namely Red edge > Red2> Green2> Blue > Red1> Green3> Green1 resulted from the various reflectance and absorption of chlorophyll in different spectra regions. The significant response is found in the following bands: Red edge, Red2, Green and Blue. In red edge region，the slope of its spectral curve is related to the chlorophyll-a, b content per leaf area. When wheat is contaminated by Pb, the spectrum in red edge region is most sensitive to the concentration thereby the response coefficients reaches the highest peak. Red2 region is the chlorophyll absorption band, and the highest absorbance of chlorophyll-a may be found around 680nm in particular. The response in Green2 region is associated with the strong absorption of phycoerythrin in the spectra between 530 and 590nm, but the region around 550nm is related to strong reflectance of chlorophyll. Blue band shows considerable absorbance of carotenoids. Note that the content of a certain element in different soil types may extremely differ from in the same region. Therefore, contamination is closely related to soil type, soil physical and chemical characteristics and vegetation types as well. There might be changes in the response relationship between Pb contamination and wheat chlorophyll. According to the analysis of Fig.3 (b), the response coefficients of remote sensing indicators are listed from high to low as follows: MCARI> NPCI> RVSI> NDVI> Depth671. MCARI synthetically reflects the variation of chlorophyll. Since wheat chlorophyll polluted by Pb is more sensitive in Red edge region, MCARI may highly respond with Pb contamination. NPCI represents the response of wheat chlorophyll on Pb concentration in red and blue absorbance troughs. Because the changes in other indices do not simply relate to wheat chlorophyll variations, there are no their obvious responses of such changes to Pb polluted wheat chlorophyll. Thus, we employ MCARI to detect the leaf chlorophyll variations of Pb contaminated wheat as shown as Fig.4 (a), wherein the values of MCARI have been standardized. Since the roots of plant could fix or absorb partial Pb passively, the transportation of Pb from roots to stem and leaf is restricted. From Fig.4 (a), it is found that there are no big changes in MCARI values when Pb concentration of wheat ranges from 0 to 2000mg/kg.With the Pb concentration increasing, the values of MCARI decreases with small amplitude. This indicates that plant grows normally and the leaf pigments are not destroyed severely. It is also shown that wheat has strong resistibility to Pb contamination. Within the Pb concentration between 2000 and 3000mg/kg, the MCARI curve has a stable shoulder, indicating that wheat is in a state of Pb-resistance. From 3000 to 5000mg/kg, MCARI decreases dramatically and the pigments are destroyed severely. 3.3. Spectral response to the leaf moisture of Pb contaminated wheat When the wheat is stressed by Pb contamination, the cell membrane and the endomembrane system of various orthogenetic will swell or break. The osmosis of membrane is destroyed, and penetration pressure between the interior and exterior of the membrane is unbalanced. In this case, the transportation of moisture molecule and nutritious elements will be obstructed, thereafter not only stimulates the decreasing absorption of moisture but also results in the metabolic disturbance.


60 Band response coefficient (%)

Spectral bands 50

Blue (447-498nm (B10-B15))

b gGreen1 (508-528nm (B16-B19)) gGreen2 (538-569nm (B20-B22)) gGreen3 (579-599nm (B23-B25)) rRed1 (610-650nm)(B26-B30)) rRed2 (660-681nm)(B31-B33)) rRed edge (691-763nm (B34-B41))

40 30 20 10 0 60

60

50

50 Response coefficient of spectral parameters(%)

Response coefficient of spectral parameters(%)

(a) The band response relationship of Pb polluted wheat in leaf pigment

40 30 20 10

Depth671 MCARI NDVI 1

0

RVSI

NPCI

(b) The parameter response relationship of Pb polluted wheat in leaf pigment

Band response coefficient (%)

70 60 50

40 30 20 10 0

Depth Depth LWVI DSWI NDWI WI 983 1205 (c) The parameter response relationship of Pb polluted wheat in water content Spectral bands Red R edge (691-763nm (B34-B41))

N N NIR-3 (864-892nm (B51-B53)) N NIR-4(894-972nm(B54-B57、B80-B83)) N NIR-5(976-993nm)(B84-B85)) N NIR-6 (996-1194nm)(B86-B105)) N NIR-7(1205-1255nm (B106-B111)) N NIR-8(1265-1295nm (B112-B115)) N NIR-1(782-803nm (B43-B45)) NIR-2(813-862nm (B46-B50))

40 30 20 10 0

(d) The band response relationship of Pb polluted wheat in vegetation structure

Response coefficient of spectral parameters(%)

60 0 50 0 40 0

Spectral parameters FDREP

30 0

NDVI

20 0

RVSI

0 10

0

0

RVI

(e) The parameter response relationship of Pb polluted wheat in vegetation structure

Fig.3 The response relationships of spectral and parameters of Pb polluted wheat


Two pronounced troughs can be observed in the regions larger than 1300nm of the vegetation spectra, namely 13001600nm (1400nm) and 1830-2008nm (1950nm) due to considerable moisture absorption. They can be sensitive bands to monitor crop water content. Additionally, we choose to six indices to investigate moisture variations, including Depth of the 983 nm absorption band (Depth983), Depth of the 1205 nm absorption band (Depth1205), Leaf Water Vegetation Index (LWVI), Disease Water Stress Index (DSWI), Normalized Difference Water Index (NDWI) and Water index(WI).In theory, these parameters can be used to retrieve crop moisture. Again the analysis should be performed to determine the indices closely related to moisture of Pb contaminated wheat. The result using multiple discriminant analysis is shown in Fig.3(c).The response coefficients of above indices are listed from high to low: NDWI>WI>DSWI> Depth983>LWVI >Depth1205 (Fig.3(c)). NDWI is more sensitive to Pb contaminated wheat moisture, which can be regarded as the response factors of crop moisture. Fig.4 (b) illustrates the results of wheat moisture variation using NDWI. According to Fig.4 (b), the changes in NDWI value is not much when the Pb concentration ranges from 0 to 500mg/kg. NDWI is basically high within this extent, suggesting that the wheat moisture has not been affected by Pb contamination. The variation may be derived from Pb adsorption by wheat root. When Pb concentration is between 500 and 2000mg/kg, the NDWI shows a decreasing tend. But the declining amplitude is relatively small, probably because of the increasing reflectance at 1245nm. Once wheat is polluted by Pb, the intensity of oxidation tends to lower. Therefore, the absorption of moisture decreases while vegetation reflectance rises. There occurs a breakpoint of NDWI when Pb concentration reaches 2000mg/kg. When the Pb concentration is between 2000 and 3000mg/kg，NDWI values keep steady, indicating a resistance stage of wheat. When Pb concentration exceeds 3000mg/kg, NDWI will drop rapidly which suggests that Pb contamination hampers growth of plant, and may lead to the death of crop finally. 3.4. Spectral response to the cytoarchitecture of Pb contaminated wheat If wheat is polluted by Pb, the contamination may induce changes in the structure and function of cell membrane also. Because the membrane is the main objects influenced by Pb contamination, the cell membrane will be in peroxidation causing membrane damnification and osmosis variations. Pb pollutants will affect the photosynthesis system of plant, enzyme activity and chlorophyll synthesis as well. Moreover, since Pb contamination destroys the structure of chloroplast; the antagonism action will lead to the maladjustment of elements in plant. The inducing nutrition stress may affect the growth of plant indirectly. The reflectance of typical leaf is between 40 and 50 percent in the near infrared spectra region (750～1300nm), due to the multiple reflection by leaf interior structure. Nine spectral regions are adopted to determine the cytoarchitecture variation of wheat influenced by Pb contamination in this study, namely Red edge (691-763nm (B34- B41)), NIR1(782803nm(B43-B45)), NIR2(813-862nm(B46-B50, B71-B72)), NIR3(864-892nm(B51-B53, B73-B75)), NIR4(894-972nm (B54-B61,B76-B83)), NIR5 (976-993nm (B62-B63,B84-B85)), NIR6(996-1194nm(B64-B105)), NIR7(1205-1255nm (B106-B 111)), NIR8 (1265-1295nm(B112-B115)). We select four indices including First Derivative and Red Edge Position (FDREP), Normalized Difference Vegetation Index (NDVI), Red-edge Vegetation Stress Index (RVSI) and Ratio Vegetation Index (RVI) to retrieve the cytoarchitecture variation of plant. Likewise, multiple discriminant analysis is performed to determine the indices which are closely related to wheat cytoarchitecture variation influenced by Pb concentration. The results are shown in Fig.3 (d)-(e). According to Fig.3 (d), the sequence of the spectral response coefficients from high to low is described as follows: NIR5>Red edge>NIR1>NIR3>NIR7>NIR6>NIR8>NIR2>NIR4. It indicates that wheat cytoarchitecture variation in the NIR5 bands is more sensitive than that in Red edge region. Because of the considerable decrease of reflectance in the NIR5 bands, the calculated response coefficient is relative higher. On the whole, both NIR5 and Red edge are more sensitive to wheat cytoarchitecture variation detection while Pb contaminating. From Fig.3 (e), the response coefficients of four indices can be listed as the following order: NDVI> RVSI> FDREP> RVI. The response of NDVI to cytoarchitecture variation of Pb-polluted wheat is extremely high. RVSI can also reflect the wheat structure affected by Pb contamination. Within the Red edge spectra region in particular, RVSI is highly responding with Pb pollutants to some degree because Pb-polluted wheat cytoarchitecture are sensitive in this band.


Fig.4 (c) illustrates the wheat cytoarchitecture variation associated with Pb pollutants using NDVI. When Pb concentration is between 0 and 2000mg/kg, there is no big change in NDVI. Because of the Pb contamination, the reflectance in near infrared decline while increase in red spectra regions. However, wheat cytoarchitecture do not change obviously corresponding with this concentration levels. When Pb concentration reaches 2000-3000 mg/kg, NDVI values decrease fast suggesting there is severely damage of wheat cytoarchitecture. Within the concentration of 3000-4000 mg/kg, the NDVI curve occurs a steady shoulder which indicates that cytoarchitecture of wheat is damaged to a certain degree. Wheat is in the state of resistance with slow reaction to Pb concentration. Once Pb concentration exceeds 4000 mg/kg, NDVI drops rapidly. The crop cytoarchitecture will be damaged badly till the death of wheat. 3.5. Spectral response to the LAI of Pb contaminated wheat Pb contamination may lead to some certain changes in wheat modality and physiological features. The modality damnification refers to the exterior shape variations such as defoliation, ramification decrease or death and foliage emerging necrotic speckles. We employ leaf area index (LAI) to represent the wheat modality changes. In this study, RVI is used to retrieve LAI of wheat. The response relationship between RVI and Pb concentration is established and shown in Fig.4 (d), wherein the values of MCARI have been standardized. 0.8

1.0

0.7

0.9 0.8 Computation result

0.5 MCARI

0.3 0.2

0.5 0.4 0.3 0.2

0.1

0.1 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Pb pollution concentration coefficient

0

(b) Response curve of Pb pollution and NDWI

0.9

0.9

0.8

0.8

0.7

0.7

0.6

0.6

0.4 NDVI

0.2 0.1

RVI

0.4 0.3 0.2 0.1

(c) Response curve of Pb pollution and NDVI

8

4

4

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Pb pollution concentration coefficient 6

0

8

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Pb pollution concentration coefficient

4

0

0.5

0

0.3

Computation result

Computation result

(a) Response curve of Pb pollution and MCARI

0.5

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Pb pollution concentration coefficient

6

0

NDWI

0.6

8

0.4

0.7

4

Computation result

0.6

(d) Response curve of Pb pollution and RVI

Fig.4 The response curves of Pb pollution and vegetation indices

Based on Fig.4 (d), when Pb concentration ranges from 0 to 2000mg/kg, RVI value decreases slightly associated with the rising concentration. It is also suggested that the decreasing amplitude of LAI is small and the variation in wheat modality is not obvious. When Pb concentration is between 2000 and 3000mg/kg, the RVI curve is in a stable state while the RVI value drops a lot. Both LAI and the density of wheat leaf decrease because vegetation structure is destroyed by Pb contamination inducing defoliation. The relative steady in RVI indicates that wheat can resist certain


Pb concentration. When Pb concentration is larger than 3000mg/kg, RVI declines rapidly and LAI decrease. Most wheat leaves will fall off. 4.

PLUMBUM CONTANMATION MECHANISM MODEL OF AGRICULRAL SOIL BY REMOTE SENSING

Though there might be not visible characteristics of agricultural soil polluted by heavy metals such as Pb, the crop will change in some biophysical feature associated with the contamination. The hyperspectral remote sensing data and the indices derived from the hyperspectral data can respond with such variation effectively. After analyzing the response relationship between Pb polluted soil and the variations in wheat pigments, moisture, cytoarchitecture and LAI, six models are established to describe the impact mechanisms of Pb concentration on agricultural soil using the above four hyperspectral indices. They can be used to identify non-point source pollution of agricultural soil. These models include MCARI-NDWI model, MCARI-NDVI model, MCARI-RVI model, NDWI-NDVI model, NDWI-RVI model and NDVI-RVI model shown in Fig.5. 1.0

0.8 0.8 0-2.0

0.6 NDVI

NDWI

0.6 2.0-4.0

0.4

0-2.0 0.4

2.0-4.0 4.0-5.0

0.2

0.2 4.0-5.0 0

0.2

0.4 MCARI

0.6

0

0.8

0.8

NDVI

0.4 4.0-5.0

0-2.0

2.0-4.0 0.4

0.2

0.4 MCARI

0.6

0

0.8

0-2.0

0.8

3.0-5.0

0.2

0.4 0.6 NDWI

0.6

1.0

0-2.0

RVI

2.0-3.0 RVI

0.8

0.8

0.6

0.4

0.8

2.0-4.0

0.2

0.2

0.6

0.6

0-2.0 RVI

0.4 MCARI

0.8

0.6

0

0.2

3.0-5.0

0.4 2.0-4.0 0.2

0.2

0

0.2

0.4 0.6 NDWI

0.8

1.0

0

3.0-5.0

0.2

0.4 NDVI

0.6

0.8

Fig.5 The remote sensing mechanism models of Pb polluted wheat Figures in circles represent the Pb concentration of the points locating in this area Proc. of SPIE Vol. 7145 71450P-9

1.0

From Fig.5, we find that these remote sensing parameters are correlated to each other. The low Pb concentration (02000mg/kg) is related to the high levels of these indices, while medium Pb concentration (2000-3000 mg/kg) is associated with the regions of medium values of those parameters. Moreover, when Pb concentration is between 2000 and 3000 mg/kg, crop often resist to the contamination. It can be observed in the response relationship of each parameter. Therefore, all the remote sensing indices may have certain value, which could be regarded as a “Wheat Resistibility Point”. High Pb concentration (3000-5000mg/kg) is associated with the lower remote sensing indices. Pb contamination for large area agricultural soil can be identified and assessed on the basis of these models. 5.

CONCLUSION

The issue of environmental pollution due to toxic heavy metals in agricultural land due has begun to cause worldwide growing concern. In this paper, the remote sensing information mechanism of heavy metal contaminated agricultural soil is discussed by integrating Hyperion imagery and ground sample measurements. We analyze the relationships between physical and chemical characteristics, biological parameters of soil-vegetation system and hyperspectral spectrum responses. The multiple discriminant analysis (MDA) method is presented to filter these bands and variables which are sensitive to wheat chlorophyll, wheat leaf moisture, wheat cell structure and wheat LAI. The changes in the wheat biological parameters and spectral response associated with Pb concentration in soil. We establish six models namely MCARI-NDWI model, MCARI-NDVI model, MCARI-RVI model, NDWI-NDVI model, NDWI-RVI model and NDVI-RVI model to explore the impact mechanisms of Pb concentration on agricultural soil. When Pb concentration is between 2000 and 3000 mg/kg, there will occur “Wheat Resistibility Point”. High Pb concentration (3000-5000mg/kg) is associated with the lower remote sensing index value. The use of vegetation as sentinels to indicate presence/absence of contaminants provides an ideal mechanism for a wide-area detection scenario. Plants have the ability to interrogate their environment spatially and temporally through roots and leaf stomata. MCARI, NDWI, NDVI and RVI can be adopted to detect the wheat land contaminated by Pb. On the basis of the six mechanisms models for Pb contamination detecting proposed by this study, we may adjust the indices values in terms of actual soil and wheat varieties practically so that Pb contamination for large area agricultural soil can be identified and assessed.

ACKNOWLEDGEMENTS This work was supported by the National Natural Science Foundation of China (40771155) and National High-tech R&D Program of China (863 Program) (2007AA12Z174).

REFERENCES Croft, B., Magarey, Disease management, Bureau of Sugar Experiment Stations, Brisbane, 2000. Pinter, P. J., Jackson, R. D. and Gausman, H. D., " Sun-angle and canopy-architecture effects on the spectral reflectance of six wheat cultivars," International Journal of Remote Sensing. (6), 1813– 1825(1985). 3 Goel, P. K., Prasher, S. O., Landry, J. A., Patel, R. M., Bonnell, R. B., Viau, A. A. and Miller, J. A., " Potential of airborne hyperspectral remote sensing to detect nitrogen deficiency and weed infestation in corn," Computer and Electronics in Agriculture. (38), 99–124(2003). 4 Haboudane, D., Miller, J. R., Tremblay, N., Zarco-Tejada, P. J. and Dextraze,L., " Integrated narrow-band vegetation indices for prediction of crop chlorophyll content for application to precision agriculture," Remote Sensing of Environment. (81), 416– 426(2002). 5 Rudorff, B. F. T. and Batista, G. T., " Yield estimation of sugarcane based on agrometeorological spectral models," Remote Sensing of Environment. (33), 183– 192(1990). 6 Osborne, S. L., Schepers, J. S., Francis, D. D. and Schlemmer, M. R., " Use of spectral radiance to estimate inseason biomass and grain yield in nitrogen- and water-stressed corn," Crop Science. (42), 165– 171(2002). 7 Roberts, D. A., Dennison, P. E., Gardner, M. E., Hetzel, Y., Ustin, S. L. and Lee, C. T., " Evaluation of the potential of Hyperion for fire danger assessment by comparison to the Airborne Visible/Infrared Imaging Spectrometer," IEEE Transactions on Geoscience and Remote Sensing. (41), 1297–1310(2003). 1 2


Hansen, P. M. and Schjoerring, J. K., " Reflectance measurement of canopy biomass and nitrogen status in wheat crops using normalized difference vegetation indices and partial least squares regression," Remote Sensing of Environment. (86), 542– 553(2003). 9 Moran, M. S., Pinter, P. J., Clothier, B. E. and Allen, S. G., " Effect of water stress on the canopy architecture and spectral indices of irrigated alfalfa," Remote Sensing of Environment. (29), 251–261(1989). 10 Clark, R. N. and Roush, T. L., " Reflectance spectroscopy: Quantitative analysis techniques for remote sensing applications," Journal of Geophysical Research. (89), 6329– 6340(1984). 11 Daughtry, C. S. T., Walthall, C. L., Kim, M. S., Colstoun, E. B. and McMurtrey, J. E., " Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance," Remote Sensing of Environment. (74), 229–239(2000). 12 Rouse, J. W., Haas, R. H., Schell, J. A. and Deering, D. W., " Monitoring vegetation systems in the Great Plains with ERTS. Proceedings of Third ERTS-1 Symposium," NASA SP-351, Washington. (1), 309–317(1973). 13 Merton, R. and Huntington, J., " Early simulation of the ARIES-1 satellite sensor for multi-temporal vegetation research derived from AVIRIS," Summaries of the Eight JPL airborne earth science workshop. 99(17), 299– 307(1999). 14 Blackburn, G. A., " Spectral indices for estimating photosynthetic pigment concentrations: A test using senescent tree leaves," International Journal of Remote Sensing. (19), 657– 675(1998). 15 Leˆnio Soares Galva˜o*, Antoˆnio Roberto Formaggio and Daniela Arnold Tisot, " Discrimination of sugarcane varieties in Southeastern Brazil with EO-1 Hyperion data," Remote Sensing of Environment. (94), 523–534(2005). 16 Apan, A., Held, A., Phinn, S. and Markley, J., " Detecting sugarcane ‘orange rust’ disease using EO-1 Hyperion hyperspectral imagery," International Journal of Remote Sensing. (25), 489–498(2004). 17 Gao, B. C., " NDWI—a normalized difference water index for remote sensing of vegetation liquid water from space," Remote Sensing of Environment. (58), 257–266(1996). 18 Ogunjemiyo, S., Roberts, D. A., Keightley, K., Ustin, S. L., Hincley, T. and Lamb, B., " Evaluating the relationship between AVIRIS water vapor and poplar plantation evapotranspiration," Journal of Geophysical Research. (D107), ACL20.1– ACL20.15(2002). 19 Tsai, F. and Philpot, W., " Derivative analysis of hyperspectral data," Remote Sensing of Environment. (66), 41– 51(1998). 20 Wiegand, C.anderson, G., Lingle, S. and Escobar, D., " Soil salinity effects on crop growth and yield: Illustration of an analysis and mapping methodology for sugarcane," Journal of Plant Physiology. (148), 418– 424(1996). 8
