Retrieval of Bare Soil Surface Parameters from Simulated Data Using Neural Networks Combined with IEM

KG Zhao,JC Shi,LX Zhang,LM Jiang,ZJ Zhang,J Qin,YJ Yao,JC Hu
DOI: https://doi.org/10.1109/igarss.2003.1295301
2003-01-01
Abstract:Many attempts have been made to retrieve soil surface parameters, such as soil moisture (SM), surface roughness parameters, by regressions or other statistical methods and some other techniques like neural networks (NNs) and genetic algorithms. The NN is proved to be an effective method for retrieval problems; much effort has been devoted to it. In this study, our goal is to estimate the bare surface soil moisture and surface roughness at Advanced Microwave Scanning Radiometer (AMSR/E) frequencies. First, a preliminary analysis was conducted based on a sensitivity analysis of surface parameters by simulating AMSR/E emissivity data of V, H polarizations, which were generated by the Integral Equation Model (IEM) for AMSR/E viewing angle of 55 degrees. We employed NNs to be first trained with part of the sensitive data determined by the above sensitivity analysis, then the trained NNs were used to retrieve the parameters that we need, especially soil moisture, from the simulated data. Analysis of the difference between the retrieved parameters and the simulated ones is presented. In addition, because the retrieval accuracy of NNs is supposed to be extremely sensitive to "noise" - the difference between the model and measurements, we introduced random noise to the simulated data. At the same time, we carried out a sensitive analysis of the input noise. We also selected the most sensitive frequencies, 6.9 and 10.7 GHz, to soil moisture in our retrieval scheme. This study demonstrates the great potential of NNs in estimating soil surface parameters from passive microwave remotely sensed data again.
What problem does this paper attempt to address?