SOIL MOISTURE RETRIEVAL ALGORITHMS. USING SAR DATA

Abstract

The scattering of microwave radiation from soil surface are greatly affected by the local terrain as well as from different soil parameters. The dielectric property of soil being related to the soil moisture provides a good mean to find out the soil moisture. For this various researchers have developed many algorithms to exploit this feature of soil. However the SAR data available is not purely reflected from soil surface but also vegetation and other land cover classes. This challenge of discovering new techniques to normalize the effect of vegetation and find out the backscattering value only dependent on the soil surface requires normalizing vegetation effect. SAR data is a polarized data having different wavelength HH, VV, HV and VH. These wavelengths interact differently with different properties of soil from which we can use the best possible parameters that related to soil moisture. But first the land cover is classified into various classes so as to identify different regions. From classification it can be seen that land belonging to same land cover class exhibits similar surface properties. So this classification is carried out using optical data. The optical data consists of spectral information of the class in visible region, infra-red and red region. And the spectral reflectance of surface is different for different land cover classes. Using this property of soil various indices are applied like the Normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), Modified soil-adjusted vegetation index (MSAVI), Global environment monitoring index (GEMI), Purified adjusted vegetation index (PA VI) etc. are used. With the help of these indices various land cover are studied. After classification the next main task done was estimating the soil moisture. Using the classified MODIS image the PALSAR image is superimposed in order to classify the PALSAR image. After classification the inversion approach is applied to the radar signal for estimating the soil moisture. In this thesis three inversion approaches are implemented, Bayesian approach, the neural network approach and an empirical approach. The result of the inversion is then checked out to find the limits of error on surface region. From this error estimation a minimum error limit is defined on received values from which surface regions having almost similar soil moisture can be classified into regions having similar moisture values.