EVALUATION OF MULTIPLE INCIDENCE ANGLE X-SAR IMAGES COVERING PARTS OF GUJARAT FOR ROUGHNESS, SOIL MOISTURE AND VEGETATION STUDIES

S.Mohan, P.Patel, D.R. Rajak, H.S. Srivastava, R.L.Mehta and N.S.Mehta

Remote Sensing Applications Group

Space Applications Centre ( ISRO )

AHEMEDABAD -380 053 , INDIA

1.0 INTRODUCTION

Estimation of land surface parameters from imaging radar had been of interest in recent past. Estimation of soil moisture, surface roughness and vegetation parameters from satellite measurements is of primary importance for agricultural, hydrology and meteorological applications. For the retrieval of these information, a need exists to examine the microwave response to various surface properties. A key element in understanding the ability of microwave data for retrieving surface parameter is the ability of a model in explaining the variability in the signal due to various surface parameters. A few of the surface scattering models like Kirchhoeff"s model, small perturbation model does predict the trend of radar backscatter due to roughness/ soil moisture..In certain cases, their validity is limited owing to the assumption involved in theoretical model (Oh et al.,1992). Parallel efforts are being made by various investigators on the development of semi-empirical and empirical models. The theoretical model developed by Fung and Chen (1992) accounted for larger range of surface roughness in the computation of backscattering coefficient. .Some of the recent empirical and semi-empirical models (Oh et al., 1992., Dubois and Engman, 1995) are able to explain the backscattering coefficient variability due to surface parameters like roughness/soil moisture. However, these models require multipolarisation data. In the present scenario, SAR data is available from ERS-1/2 or Radarsat on an operational basis. Among these, ERS-1/2 SAR is a fixed parameter system whereas Radarsat - SAR provides images of the earth's surface in multiple incidence angle imaging mode. The fixed parameter SAR has limitation in resolving contribution due to roughness in backscattering values that influence relationship between backscatter and soil moisture.. Thus, need exists to examine the methods/ models that are based on angular SAR response to surface characteristics. SIR-C/X-SAR mission has provided oppurtinity to experiment with multiple incidence angle SAR data at different frequency and polarisation. The present paper brings out the results of such an evaluation for multiple incidence angle X-SAR ( VV ) data covering parts of Gujarat.

2.0 STUDY AREA AND DATA SET

The study area is dominated by deltaic alluvium soils covering parts of Bhavanagar, Gujarat, India. X-SAR data was acquired on 11th, 12th and 14th April 1994 at three angles of incidence. During the period of SAR data acquisition, agricultural fields were mostly fallow. Theirs measured roughness (rms height) was varying up to 2.2 cm. Soil moisture measurements were done using gravimetric method during each sequence of the pass. During the period of SAR data acquisition of four days, soil roughness was same and roughness measurements were done using gridded plate method (Ulaby et al., 1986). Among the vegetation, Acacia plantation was mostly characterised in to three vegetation density classes (2.75, 4 and 6.25 plants per hundred meter square). Other land cover classes like creeks, barren land and habitation were also present in the study area. For the purpose of ground truth and data analysis, support data like topographic map and Indian Remote Sensing Satellite data were also used.

3.0 DATA PROCESSING AND ANALYSIS

The X-SAR sub-images were extracted at three angles of incidence and coregistered. Images were subjected to speckle noise suppression using Frost filter (Frost et al.,1982). The data was calibrated using the procedure described by Zink et al. (1995). Next task of analysis involves identification of groundtruth locations and extraction of backscattering coefficient values. The extracted values of backscattering coefficient at three angles of incidence were analysed for determining X-SAR response to various surface characteristics. The details of modelling efforts are described in the following sections

3.1 X-SAR response to bare soil

In order to determine suitability of X-SAR in multiple incidence angle imaging mode for the estimation of surface parameters, backscattering coefficient values were computed using Kirchoeff's small perturbation model for the groundtruth locations. Fig.1 shows comparison between X-SAR measured and model computed values of backscattering coefficient. Considering the validity of model, backscattering values were simulated for various roughness/ moisture conditions (Fig.2). It is noted that the variability in backscattering coefficient due to roughness is of the same order as that arising due to soil moisture. Therefore, single parameter X-SAR would not be able to discriminate the variability arising due to roughness or soil moisture. Because of this reason correlation between X-SAR measured backscattering coefficient and soil moisture is found to be poor at all the angle of incidence (Fig.3).

In order to resolve the roughness/soil moisture ambiguity , there is a need to address the problem using multiple incidence angle SAR system. Considering the angular scattering properties from bare soil surfaces, it can be deduced that the signature variability also depends on the roughness of the surface. Therefore, ratio of backscattering coefficient values at two incidence angles could be related to roughness of the surface (Autret et al., 1989). Fig. 4 shows results of such simulation at X- band. It may be seen that the roughness could be described well by angular scattering coefficient. In view of this, soil moisture retrieval model may include retrieved roughness along with backscattering coefficient.Considering this,soil moisture relation with different combination of backscattered values were attempted. It is observed that there is a drastic improvement in value of correlation coefficient (Fig.5,6). Such techniques could be useful for retrieving soil moisture from operational satellites like Radarsat etc. which provide multiple incidence angle data.

In one of the situation, L-band SAR has shown very high contrast in areas around creek, as compared to X-band (Fig.6). In this case, creek was having very dry soil cover upto 30 cm depth. Layers below 30 cm were mostly wet. This reconfirms the penetration ability of L-SAR under dry soil cover (Elachi, 1982). Thus, L-SAR could be exploited in future for identifying such zones.

3.2 X-SAR response to vegetation

For the analysis of vegetation classes, the angular backscattering coefficient variation in relation to the density of Acacia plantation is being studied. In the study area, three vegetation densities (2.75 plants, 4 plants and 6.25 plants per hundred meter square) were present. These classes were characterised by their angular backscattering coefficient values. Fig.7 shows trends of backscattering coefficient values for three density classes. It may be seen that the higher incidence angle data is able to discriminate the classes in three density levels, whereas lower angle of incidence provides two levels of density classes. For the evaluation of vegetation density relation with backscattering coefficient value, regression analysis was done for various combination of backscattering coefficient values. Highest value of correlation coefficient was observed when a combination of three angle data is used (Fig.8).

From the analysis, iconclusion can be drawn that the multiple incidence angle data is able to provide better information than that of single incidence angle data.

4.0 CONCLUSION

The present attempt is towards demonstrating the use of multiple incidence angle data for roughness / soil moisture and vegetation studies. It is observed that angular SAR signatures are able to explain variability in the backscattering coefficient due to roughness/ soil moisture and vegetation density. For the bare soil case, experimental observations have been explained by theoretical models. For a typical case, L-SAR is able to sense moist soil conditions at a depth of 30 cm under a top dry soil cover condition. Further efforts are being made for extending the evaluation technique at L and C-band for the same site.

REFERENCES

Autret, M.,Bernard, R. and Vadal Madjar, D., 1989, Theoretical study of the sensitivity of the microwave backscattering coefficient to the soil surface parameters, Int. J. Remote Sensing, 10, 171-179

Dubois, P.C., Zyl, J.V. and Engman, T., 1995, Measuring soil moisture with imaging radar, IEEE Trans. Geosci. And Remote Sensing, 33, 915-926

Elachi, C., 1982, Shuttle imaging radar experiment, Science, vol.218, 996-1002

Frost,V.S., Stiles, J.A., Schuanamugam, K.S. and Holtzman, J.C., 1982, Model of radar images and its application to digital filtering of multiplicative noise, IEEE Trans. Pattern. & Anal. Machine Intelligence, PAMI - 4, 157-166

Fung,A.K. and Chen, K.S., 1992, Dependence of the surface backscattering coefficients on roughness, frequency and polarisation studies, Int J. Remote Sensing, 13, 1663-1680

Oh, Y., Sarabandi, K. and Ulaby, F.T., 1992, An empirical model and an inversion technique for radar scattering from bare soil surfaces, IEEE Trans. Geo.Sci. and Remote Sensing, GE-30, 370-

Ulaby, F.T., Moore, R.K. and Fung., A.K., 1986, Microwave Remote Sensing, Artech., Norwood, mass

Zinc, M. and Balmer, R., 1995, X-SAR radiometric calibration and data quality, IEEE Trans. Geo.Sci. and Remote Sensing, GE-33, 840-84