A radial basis function approach to retrieve soil moisture and crop variables from x-brand scatterometer observations

Abstract

An outdoor crop-bed was prepared to observe scatterometer response in the angular range of 20° to 70° at VV- and HH-polarization. The soil moisture and crop variables like plant height, leaf area index and biomass of crop ladyfinger were measured at different growth stages of the crop ladyfinger. Temporal variation in scattering coefficient was found highly dependent on crop variables and observed to increase with the increase of leaf area index and biomass for both polarizations. In this paper, a novel approach is proposed for the retrieval of soil moisture and crop variables using ground truth microwave scatterometer data and artificial neural network (ANN). Two different variants of radial basis function neural network (RBFNN) algorithms were used to approximate the function described by the input output relationship between the scattering coefficient and corresponding measured values of the soil moisture and crop variables. The new model proposed in this paper gives near perfect approximation for all three target parameters namely soil moisture, biomass and leaf area index. The retrieval with minimal error obtained with the test data confirms the efficacy of the proposed model. The generalized regression network was observed to give minimal system error at a much lower spread constant.