Assessment of Non-parametric Methods for Soil Moisture Retrieval from Active Microwave Data

Active microwave remote sensing observations hold the potential for efficient and reliable mapping of spatial soil moisture distributions. However, soil moisture retrievals from microwave remote sensing techniques are typically complex because of the inherent difficulty in characterizing the interactions among land surface parameters that contribute to the retrieval process. Therefore adequate physical mathematical descriptions of the interaction of microwave radiation with parameters such as land cover, vegetation density, and soil characteristics are not readily available. On the other hand it may possible to use non-parametric classifiers like neural networks, fuzzy logic and multiple regression models to retrieve soil moisture distributions. In this study we make use such classifiers after using soil moisture data derived using ESTAR for training the non-parametric models due to limited availability of in-situ soil moisture measurements. The fuzzy logic and neural network models performed better when compared to multiple regression models. It was also seen that the inclusion of the vegetation and soil characteristics, derived from infrared and visible measurements, in these models have significant positive impact on soil moisture retrievals with RMSE being reduced by around 30% in the retrievals. Finally the soil moisture derived from these models was compared with ESTAR soil moisture (RMSE ~4.0%) and field soil moisture measurements (RMSE ~6.5%). Additionally, the study showed that soil moisture retrievals from highly vegetated areas are less accurate than that from bare soil areas. Key Terms:Soil Moisture, Active Microwave, Neural Network, Fuzzy Logic, Vegetation.