Statistical analysis and combination of active and passive microwave remote sensing methods for soil moisture retrieval

Knowledge about soil moisture and its spatio-temporal dynamics is essential for the improvement of climate and hydrological modeling, including drought and flood monitoring and forecasting, as well as weather forecasting models. In recent years, several soil moisture products from active and passive microwave remote sensing have become available with high temporal resolution and global coverage. However, for the improvement of a soil moisture product and for its proper use in models or other applications, validation and evaluation of its spatial and temporal patterns are of great importance. In chapter 2 the Level 2 Soil Moisture and Ocean Salinity (SMOS) soil moisture product and the Advanced Scatterometer (ASCAT) surface soil moisture product are validated in the Rur and Erft catchments in western Germany for the years 2010 to 2012 against a soil moisture reference created by a hydrological model, which was calibrated by in situ observations. Correlation with the modeled soil moisture reference results in an overall correlation coefficient of 0.28 for the SMOS product and 0.50 for ASCAT. While the correlation of both products with the reference is highly dependent on topography and vegetation, SMOS is also strongly influenced by radiofrequency interferences in the study area. Both products exhibit dry biases as compared to the reference. The bias of the SMOS product is constant in time, while the ASCAT bias is more variable. For the investigation of spatiotemporal soil moisture patterns in the study area, a new validation method based on the temporal stability analysis is developed. Through investigation of mean relative differences of soil moisture for every pixel the temporal persistence of spatial patterns is analyzed. Results indicate a lower temporal persistence for both SMOS and ASCAT soil moisture products as compared to modeled soil moisture. ASCAT soil moisture, converted to absolute values, shows highest consistence of ranks and therefore most similar spatio-temporal patterns with the soil moisture reference, while the correlation of ranks of mean relative differences is low for SMOS and relative ASCAT soil moisture products