Nonparametric Model for the Retrieval of Soil Moisture by Microwave Remote Sensing

Abstract

Currently, the retrieval of soil moisture is necessary through microwave remote sensing and it is the simplest and most reliable model for applications such as drought, flood, irrigation management, and scheduling. For this purpose, a soil test site has been prepared with different moisture content and constant soil surface roughness. The bistatic scatterometer measurements have been carried out in the angular range of 20-70 degrees steps of 10 degrees at HH- and VV-polarization for the X-band. The observed data sets (bistatic scattering coefficients and soil moisture content) have been divided into calibration (75%) and validation (25%) of data sets. The radial basis function artificial neural network (RBFANN) has been calibrated for the retrieval of soil moisture using bistatic scatterometer data. The performance of the RBFANN model has been evaluated using statistical performance indices %Bias, root mean square error, and Nash-Sutcliffe efficiency. The estimated values of soil moisture by the RBFANN model and the experimentally observed values of soil moisture are found very close during calibration and validation of the model. © 2016 Elsevier Inc. All rights reserved.