Spatio-temporal variability of global soil moisture products

Being an important variable for various applications, for example hydrological and weather prediction models or data assimilation, a large range of global soil moisture products from different sources, such as modeling or active and passive microwave remote sensing, are available. The diverse measurement and estimation methods can lead to differences in the characteristics of the products. This study investigates the spatial and temporal behavior of three different products: (i) the Soil Moisture and Ocean Salinity (SMOS) Level 2 product, retrieved with a physically based approach from passive microwave remote sensing brightness temperatures, (ii) the MetOp-A Advanced Scatterometer (ASCAT) product retrieved with a change detection method from radar remote sensing backscattering coefficients, and (iii) the ERA Interim product from a weather forecast model reanalysis. Results show overall similar patterns of spatial soil moisture, but high deviations in absolute values. A ranking of mean relative differences demonstrates that ASCAT and ERA Interim products show most similar spatial soil moisture patterns, while ERA and SMOS products show least similarities. For selected regions in different climate classes, time series of the ASCAT product generally show higher variability of soil moisture than SMOS, and especially than ERA products. The relationship of spatial mean and variance is, especially during wet periods, influenced by sensor and retrieval characteristics in the SMOS product, while it is determined to a larger degree by the precipitation patterns of the respective regions in the ASCAT and ERA products. The decomposition of spatial variance into temporal variant and invariant components exhibits high dependence on the retrieval methods of the respective products. The change detection retrieval method causes higher influence of temporal variant factors (e.g. precipitation, evaporation) on the ASCAT product, while SMOS and ERA products are stronger determined by temporal invariant factors (e.g. topography, soil characteristics). The investigation of the effect of changing scales on spatial variance in three different areas indicates that the variance does not vary with increasing support scale. Increasing extent scales from 250 to 3000 km raise spatial variance of all products and all study areas according to a power law, which is varying seasonally. ERA shows a consistent scaling behavior with a constant power scale factor and similar intercepts across all study regions. In general, the investigated products show overall different spatial and temporal statistics which are induced by their different estimation methods and which are important to be aware of for the selection of a product for application and for their up- or downscaling