Learning Surface Soil Moisture Behavior from Global SMAP and GPM Satellite Data

Since the launch of the Soil Moisture Active Passive (SMAP) satellite mission in 2015, multiple years of coarse-resolution, near-global surface soil moisture data became available. This provides unprecedentedly rich information about this important component of the water cycle over large spatial extents and at a resolution that is commensurate with the scale of application of continental-scale hydrologic models. New studies on the SMAP dataset have been emerging, ranging from the characterization of surface soil moisture dynamics, to rainfall and runoff estimation, to data assimilation applications. The main objective of this study is to use data-driven approaches to discover governing dynamics and characteristics of surface soil moisture at the 36-km SMAP resolution over the entire globe. We refrain from presuming either the structure of soil moisture dynamics (which is typically upscaled from physical laws understood at much smaller scale) or the parameterizations used to describe these dynamics, but attempt to derive this information from the data itself. We combine SMAP data with global precipitation observations from the Global Precipitation Measurement (GPM) mission (launched in 2014) to further inform SMAP data analysis. We discuss the patterns of the learned surface soil moisture behavior from both the spatial (over the globe) and temporal (based on the three years of observations) dimensions. The governing behavior derived from the SMAP data can be compared with that simulated by continental-scale distributed models, thus shedding light on coarse-scale distributed model representation of surface soil moisture.