Spatio-temporal consistency analysis of AMSR-E soil moisture data using wavelet-based feature extraction and one-class SVM

Soil moisture is one of the most important climatic parameters playing an important role in the global climate system. Soil moisture can be derived from in-situ measurements as well as remotely sensed observations. However, these measurements typically lack the spatial and/or temporal resolutions necessary for modeling and applications. Land surface models (LSM) can be used to simulate the land surface state at hydrologically-relevant spatio-temporal scales. Further, many a LSM also use sophisticated data assimilation schemes to assimilate the soil moisture observations. Before assimilation of soil moisture data into a LSM, the characteristics of the soil moisture data have to be verified. The objective of this paper is to compare the spatio-temporal characteristics of the remotely sensed soil moisture estimates from the Advanced Microwave Scanning Radiometer - EOS (AMSR-E) against in-situ soil moisture measurements from the USDA Soil Climate Analysis Network (SCAN). We have developed a consistency assessment method based on wavelet-based feature extraction and one-class support vector machines (SVM). This method performs a consistency assessment of entire time series in relation to others and provides a spatial distribution of consistency levels whereas conventional approaches typically provide information on every data point individually in relation to its neighbors only. We have applied this new methodology to assess the spatio-temporal characteristics of the soil moisture products from AMSR-E. Spatial distribution of consistency levels are presented as consistency maps showing a region in southeastern United States. These results are verified by correlating with the spatial distributions of average soil moisture, and the cumulative counts of dense vegetation.