Skill and Global Trend Analysis of Soil Moisture from Reanalyses and Microwave Remote Sensing

Abstract In situ soil moisture measurements from 2007 to 2010 for 196 stations from five networks across the world (United States, France, Spain, China, and Australia) are used to determine the reliability of three soil moisture products: (i) a revised version of the ECMWF Interim Re-Analysis (ERA-Interim; ERA-Land); (ii) a revised version of the Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis from NASA (MERRA-Land); and (iii) a new, microwave-based multisatellite surface soil moisture dataset (SM-MW). Evaluation of the time series and anomalies from a moving monthly mean shows a good performance of the three products in capturing the annual cycle of surface soil moisture and its short-term variability. On average, correlations (95% confidence interval) are 0.66 (±0.038), 0.69 (±0.038), and 0.60 (±0.061) for ERA-Land, MERRA-Land, and SM-MW. The two reanalysis products also capture the root-zone soil moisture well; on average, correlations are 0.68 (±0.035) and 0.73 (±0.032) for ERA-Land and MERRA-Land, respectively. Global trends analysis for 1988–2010 suggests a decrease of surface soil moisture contents (72% of significant trends are negative, i.e., drying) for ERA-Land and an increase in surface soil moisture (59% of significant trends are positive, i.e., wetting) for MERRA-Land. As the spatial extent and fractions of significant trends in both products differ, the trend reflected in the majority of grid points within different climate classes was investigated and compared to that of SM-MW. The latter is dominated by negative significant trends (73.2%) and is more in line with ERA-Land. For both reanalysis products, trends for the upper layer of soil are confirmed in the root-zone soil moisture (first meter of soil).