Crop polarisation behavior derived from Multi-Temporal adeos-polder data

Photonirvachak Journal of the Indian Society of Remote Sensing, Vol. 33, No. 2, 2005

CROP POLARISATION BEHAVIOR DERIVED FROM MULTI-TEMPORAL ADEOS-POLDER DATA V.N. SRIDHAR@, R.P. SINGH, ANJUM MAHTAB A N D R A N E N D U GHOSH Remote Sensing Applications Area Space Applications Centre (ISRO) Ahmedabad-380 015 @Corresponding author : vnsridhar.sac.isro.gov.in

ABSTRACT This paper reports a study on multi-temporal polarized response of wheat crop from spaceborne ADEOS-POLDER sensor over a homogeneous wheat region of Punjab, India. Both the polarized as well as total reflectance of wheat were observed at different scattering angles for two spectral bands i.e. 670 nm and 865 nm during crop growth from November to April in rabi 1996-97 season. Results show that sun-target-viewing geometry plays an important role in polarization property. The top of atmosphere (TOA) polarized reflectance is found to decrease exponentially with increasing scattering angle. Polarized reflectance of crop was found to be an order of magnitude smaller in comparison to the total reflectance. An attempt was also made to model the observed polarized behavior over an agricultural area using a theoretical simplified crop reflectance model and accounting for atmospheric molecular (Rayleigh) contribution in the single scattering approximation. It was found that there was a decrease in the polarized reflectance at the grain filling (heading) stage of wheat crop. This is in accordance with ground- based observations and can be due to the reduction in the specular component of the reflected light during post-heading stage of the crop.

Introduction

D u e to the t r a n s v e r s e n a t u r e o f light (electromagnetic radiation), a complete description o f light and its interactions with matter must involve both the intensity and polarization characteristics. Polarization refers to a directional property o f light, with specific reference to the vibrations o f the

electric vector and its orientation with respect to a reference plane. In particular, in completely linearly polarized light, the plane o f vibration o f the electric v e c t o r is either p e r p e n d i c u l a r or parallel to a reference plane while in partially polarized light, both components are present and the difference between the two c o m p o n e n t s i n d i c a t e s the a m o u n t o f partially polarized light. A good review o f basics and

Special Section on "Modeling for Remote Sensing Applications in Agriculture" Received in final form 18 December, 2002; Accepted on 20 March, 2004

340

V.N. Sridharet al.

role of polarization in remote sensing and polarization properties of natural surfaces is given by Coulson (1988). Much of remote sensing hitherto has focused almost exclusively on remote (aircraft or satellite based) measurements of the intensity of reflected light along with its spectral, spatial, radiometric and temporal variations to infer surface properties. On the other hand, there have been comparatively few studies on polarisation properties of natural surfaces. It has been shown that many surfaces such as soils, minerals, vegetation, etc., exhibit significant polarisation effects when polarisation measurements are carried out in the light reflected from different surfaces (Talmage and Curran, 1986). In case of vegetation, particularly crops, a series of ground based experiments (Vanderbilt et al., 1985, Rondeaux and Herman, 1991, Ghosh et al., 1993., Breon et al,, 1995) have demonstrated that crops do exhibit significant polarisation effects and that it is possible to discriminate between different crop stages on the basis of polarisation signatures. These studies have further demonstrated that the partial polarisation of reflected light from crop canopies is mainly due to specular reflection from surface as modeled by Fresnel equations and modified by canopy structure. It must be emphasised here that the information contained in polarised reflectance is complementary and independent to that of conventional intensity measurements since crop polarisation response is mostly a property of the surface. With the launch of POLDER (Polarisation and Directionality of Earth Reflectances) onboard ADEOS (Advanced Earth Observation Satellite) in 1996, there has been a renewed interest in the study of polarisation characteristics of surfaces from space. Due to the capability of POLDER to measure both intensity and polarisation of reflected light at a number of angles for the same pixel (resolution 6*7 km at nadir), it is possible to obtain realistic estimates of surface BRDF (Bi-directional

Reflectance Distribution Function) and BPDF (Bidirectional Polarisation Distribution Function). The POLDER sensor images the earth in 9 wavelengths and can provide reflectance and polarised reflectance data at a number of sun and view angles (up to a maximum of 14 angles) (Deschamps et al., 1994). Due to its unexpected failure in June 1997, ADEOS-POLDER data has been available for the months November 1996-May 1997. A few studies exploring the bi-directional properties of surfaces and corrections for angular effects (Leroy and Hautecoeur., 1999; Sridhar et al., 2001) and polarisation signatures (Nadal and Breon., 1999) have been reported using multi-date POLDER data. In this paper, we have attempted to study behavior of crop polarisation using temporal POLDER data over Punjab state. We have attempted a simplistic and empirical analysis of the polarised reflectance data both as a function of sun-targetsensor geometry,and date of acquisition. In addition, a model of vegetation polarisation response incorporating Fresnel reflection and canopy structure (Rondeaux and Herman, 1991) as well as polarised atmospheric molecular effects in the single scattering approximation (Leroy et aL, 1997) have been used to compute TOA polarised response and compare with POLDER observed polarised reflectance.

Methodology The five date ADEOS-POLDER data used in the analysis span the months November 1996-April 1997 of the following dates: November 7, 1996, December 18, 1996, January 28, 1997, March 10, 1997 andApril 20, 1997. Data corresponding to February 1997 was found to be cloudy over India and hence was not used in the analysis. The relevant POLDER data pertaining to India (Latitude 8-37 ~, Longitude 68-97 ~) was extracted from six date ADEOS-POLDER global data. The images were displayed on IBM RS 6000 workstation

Crop Polarisation Behavior Derived from... and a representative crop pixel was selected from a simultaneous perusal of six images of Punjab state. It is to be emphasized here that the crop pixel was selected from a relatively large homogeneous region so that it represents the average behavior of crop in the region.

341

surface contribution, ii) polarized atmospheric m o l e c u l a r contributions and iii) polarized atmospheric aerosol contributions. We have taken into account the first two contributions to the POLDER observed signal in this analysis.

Surface contribution Information regarding solar zenith, view zenith relative azimuth and associated scattering angles along with the corresponding polarized reflectance at 670 and 865 nm were extracted for the crop pixel from pixel record information. A representative plot of polarized reflectance as a function of scattering angle for 670 and 865 nm is given in figure 1 for December 1996 data. The scattering angle ct is related to sun zenith (0~), view zenith (0v) and relative azimuth angles as follows: Cos ct= -cos0s* cos0V- sin 0s* sin 0v'cos (~-~v) (I) where ~s - d~v is the relative azimuth between sun and view directions. There is a monotonic decrease o f polarized reflectance with scattering angle. As a simple and compact descriptor of crop polarized reflectance behavior, we have fitted a two parameter exponential curve to the data of the form PR~ = ax*exp (-bxet)

(2)

where PR~ is polarized reflectance at wavelength ~, cx is the scattering angle and a x and bx are fitted coefficients.

As mentioned above, the primary mechanism responsible for polarization of reflected light from a surface is Fresnel specular reflection from optically smooth surfaces. In many crops such as wheat, corn etc., the wax deposits o f leaves form a smooth amorphous film on the cuticle (Vanderbilt et al., 1985) which forms an optically smooth surface, thereby leading to partial polarisation of reflected light. Using Fresnel equations and a simple model for the canopy, Rondeaux and Herman (1991) have derived the following expression for crop polarised

reflectance (PRveg): PRve~ = F(i)/(4(cos0s + cOS0v))

(3)

where F(i) is Fresnel reflection coefficient at incidence angle i and is given by F(i) = 0.5*[{(A-cosi)/(A+cosi)}- {(N2cosi-A)/ (N2cosi + A)}] where A = (N 2- sin2i) t/2, N = 1.5 is the refractive index of vegetation. The incidence angle i is related to scattering angle tx by i = (r~ - c0/2, and 0 s and 0 v are solar and view zenith angles respectively.

Atmospheric molecular (Rayleigh) contribution

Except in case of December data for 865 nm, the coefficient of determination (R 2) is found to be more than 90 per cent in all cases. The values of fitted coefficients for the two wavelengths are given in Table 1 for all the five dates.

Following Leroy et al. (1997), the polarized reflectance due to Rayleigh scattering in the single scattering approximation (PRmol) can be written as

Model fit to observations

where "~R(~.) is the Rayleigh optical depth at wavelength ~, and QR(~,,~) is the Rayleigh phase function and is equal to 3/4(1+cos2o0 for total (unpolarised) scattering and 3/4(cos2a-1) for

The polarized signal observed at the TOA can be modeled as a sum of three terms viz. i) polarized

PRmoI=(TR(X)/(4COS(0v))*QR(X,~)

(4)

342

V.N. Sridhar et al.

linearly polarized scattering at wavelength ~. Since we do not have in-situ measurements of optical depth at the two wavelengths over the area o f interest, representative values o f optical depth at sea level for a tropical atmosphere have been taken from Bucholz (1995). The values are XR(670) = 0.044 and "tR(865 ) = 0.013.

other dates. The range of scattering angles across the five dates spans the range - 90-158 degrees. n = 43.769e-0 m ] 0~77xy LE] 865

o~'4 F, 3

An illustrative plot of modeled polarized TOA reflectance using eqns (3) and (4) and the POLDER observed polarized TOA reflectance is given in figure 2 for 865 nm for December data.

r

oJ

~N 2 __m O

m

1

Temporal behavior of polarized reflectance For a fixed scattering angle of approximately 121 degrees we have plotted both Rayleigh (molecular) corrected polarized and unpolarised reflectance (in the single scattering approximation) at 670 and 865 nm as a function of date of acquisition starting from November 1, 1996. The scattering angle is a function of sun zenith, view zenith and relative azimuth angles and therefore varies with day of year. For scattering angle around 121 degrees, observations were available on all five dates for 670 and 865 nm and these were used to analyze both unpolarised and polarized behavior of crop. Results and Discussion

It is seen from figure 1 that the TOA (top of atmosphere) polarized reflectance shows a maximum at low scattering angles and smaller values in the backscattering direction. A similar behavior of polarized reflectance is reported in Nadal et aL, (1999). Further, the magnitude o f polarized reflectance is more in 670 nm compared to 865 nm. Since these are top o f a t m o s p h e r e (TOA) reflectance, atmospheric contributions, in particular the Rayleigh polarized scattering contribution is expected to be more in 670 nm than in 865 nm. Though we have not shown here the polarized reflectance for all dates, the trend is similar for the

50

70

90

110

130

150

Scattering Angle (degrees)

Fig. 1. Spectral dependence ofTOA polarised reflectance as a function of scattering angle It is seen from Table 1 that the 'a' coefficients are larger for 670 nm compared to 865 nm and that the ' b ' coefficients are o f the same order o f magnitude across the five dates. The two-parameter empirical fits to the polarized reflectance data can be used to estimate polarized reflectance at any scattering geometry. For instance, for a sun angle of 45 degrees and nadir view angle, the scattering angle is 135 ~. The expected polarized reflectance at TOA for this geometry is 0.92 % and 0.60 % for 670 and 865 nm for the month o f November and 0.90 % and 0.7 % for the month o f January. The modeled polarized reflectance in fig. 2 (using eqns (3) and (5)) shows qualitatively similar behavior to that o f observed polarized TOA reflectance. The deviations between observed and modeled reflectance vary with scattering angle ranging from - 8 % at low scattering angles to - 3040 % at higher scattering angles. A better model of surface polarized reflectance incorporating multiple scattering contributions within crop canopy as well as inclusion o f atmospheric polarized aerosol

343

Crop Polarisation Behavior Derived from... T a b l e 1: Two parameter (a, b) empirical, exponential fit to observations X(nm)

Nov 96

Dee 96

Jan 97

Apr 97

M a r 97

a

b

a

b

a

b

a

b

a

b

670

46.36

.029

43.77

.028

30.32

.026

72.55

.034

521.12

.043

865

41.20

.0313

25.34

.025

10.44

.02

46.43

.033

205.5

.045

contribution is expected to improve the model fit to POLDER observed reflectance.

60

~t670nm o

50

Dt865nm

- ~ - ~ 670nmx40

3.5,

•

3.

~ * * *

"

~

o~ 40

- - -865nmobs

- ~ - ~ 865nmx40

865 nm mod

~ 2.5 2O ~ 1.5

10 0

~. 0.5

0 75

100

125

Fig. 2. Comparison of observed and modeled polarised reflectances as a function of scattering angle for 865 nm

Temporal behavior of polarized reflectance In figure 3, it is seen that that the total reflectance at 670 and 865 nm follows expected behavior during the temporal evolution of the crop. The peak at 865 nm is observed at about 125 days after November 1, 1996 i.e. around mid-February. The crop polarized reflectance at 865 nm (multiplied by 40 for visual display) shows a pronounced decrease at the same time. Previous work on crop polarized reflectance in ground based experiments (Vanderbilt, 1985; Ghosh et al., 1993) has shown that there is a significant decrease in crop polarized reflectance after heading stage. This is because heads, not being smooth, are poor specular reflectors and polarizes of incident light. It is interesting to note that the same behavior of crop polarized reflectance is seen with POLDER measured crop polarized reflectance.

i

i

w

100

150

200

Days after November 1, 1996

150

Scattering Angle (degrees)

i

50

Fig. 3. Temporal behavior of wheat total and polarised reflectances for 670 and 865 nm (tot : total reflectance, rp : polarised reflectance) Conclusions

The following are the main conclusions of the analysis: Crop polarised reflectance is an order of magnitude smaller than total reflectance and is typically 10-20 times smaller than total reflectance. A simple, purely empirical two parameter exponential model can describe b e h a v i o r TOA polarised reflectance over an agricultural area quite well with coefficient of determination (R 2) more than 90 per cent. These empirical relations can be used to estimate crop-polarised reflectance at 670 and 865 nm for any scattering angle. A physical model incorporating Fresnel equations and a simple description of canopy structure and inclusion of atmospheric polarised Rayleigh contribution in the single scattering approximation can explain to an

344

V.N. Sridhar et aL

extent the qualitative behavior o f observed crop polarised reflectance with scattering angle. A decrease in crop polarised reflectance is observed in the post-heading stage at both 670 and 865 nm during the month of February 1997.

Acknowledgements We thank CNES, France for ADEOS-POLDER data and associated software. We thank Dr V.K. Dadhwal, Head, Crop Inventory & Modeling Division, Space Applications Centre, Ahmedabad for inviting us to write this article and for support and encouragement. The authors would like to thank Dr Ajai, Group Director, Forestry, Land use Planning and Photogrammetry Group, Space Applications Centre, Ahmedabad for support and encouragement. We would like to thank the anonymous referee for critical comments and useful suggestions.

References Breon, F., Marie, T.D., Pierre, L. and Maurice, H. (1995). Polarised Reflectance of bare soils and vegetation: Measurements and Models. IEEE Trans Geoscience and Remote Sensing, 33: 487-499.

Scientific Objectives. 1EEE Trans Geoscience and Remote Sensing, 32(3): 598-613. Ghosh, R., Sridhar, V.N., Venkatesh, H., Mehta, A.N. and Patel, K.I. (I 993). Linear Polarisation measurements of a wheat canopy. Int. d. Remote Sensing, 14: 2501. Leroy, M., Deuze, J.L., Breon, F.M., Hautecoeur, O., Herman, M., Buriez, J.C., Tanre, D., Bouffies, S., Chazette, P. and Roujean, J.L. (1997). Retrieval of atmospheric properties and surface bidirectional reflectanees over land from POLDER/ADEOS. d. Geophysical Research, 102( D ! 4): 17,023-17,037. Leroy, M. and Hautecoeur., O. (1999). AnisotropyCorrected Vegetation Indexes Derived from POLDER/ADEOS. 1EEE Trans Geoscience and Remote Sensing, 37(3): 1698-1706. Nadal, F. and Breon, F. (1999). Parametrisation of Surface Polarised Reflectance derived from POLDER Spaceborne Measurements. IEEE Trans Geoscience and Remote Sensing, 37:1709-1718. Rondeaux, G. and Herman M. (1991). Polarisation of light reflected by crop canopies. Remote Sensing Env., 38: 63-75.

Bucholz, A. (1995). Rayleigh-scattering calculations for the terrestrial atmosphere. AppL Optics, 34: 2785.

Sridhar, V.N., Ghosh, R. and Anjum, M. (2001). Atmospheric and angular effects on NDVI Temporal Profiles derived from ADEOS-POLDER data over India. IEEE Trans Geoscience and Remote Sensing, 39:1099-1102.

Coulson, K.L. (1988). Intensity and Polarisation of Light in the Atmosphere. A. Deepak Publishing, Hampton, Virginia, USA.

Talmage, D.A. and Curran, P.J. (1986). Remote Sensing using partially polarised light. Int. d. Remote Sensing, 7: 47-64.

Deschamps, P., Francois, M., Leroy, M., Podaire, A., Bricaud, A., Buriez, J.C. and Seze., G. (1994). The POLDER Mission: Instrument Characteristics and

Vanderbilt, V.C., Grant, L. and Daughtry, C.S.'I: (1985). Polarisation of light scattered by vegetation. Proc IEEE, 73: 1012-1024.