Evaluation of 22 precipitation and 23 soil moisture products over a semiarid area in Southeastern Arizona

Precipitation and soil moisture are rigorously measured or estimated from a variety of sources. Here, 22 precipitation and 23 soil moisture products are evaluated against long-term daily observed precipitation (Pobs) and July-September daily observationally constrained soil moisture (SM) datasets over a densely monitored 150 km² watershed in southeastern Arizona, United States. Gauge-radar precipitation products perform best, followed by reanalysis and satellite products, and the median correlations of annual precipitation from these three categories with Pobs are 0.83, 0.68, and 0.46, respectively. Precipitation results from phase 5 of the Coupled Model Intercomparison Project (CMIP5) are the worst, including an overestimate of cold season precipitation and a lack of significant correlation of annual precipitation with Pobs from all (except one) models. Satellite soil moisture products perform best, followed by land data assimilation systems and reanalyses, and the CMIP5 results are the worst. For instance, the median unbiased root-meansquare difference (RMSD) values of July-September soil moisture compared with SM are 0.0070, 0.011, 0.014, and 0.029m³ m^-3 for these four product categories, respectively. All 17 (except 3) precipitation [17 (except 2) soil moisture] products with at least 20 years of data agree with Pobs (SM) without significant trends. The uncertainties associated with the scale mismatch between Pobs and coarser-resolution products are addressed using two 4-km gauge-radar precipitation products, and their impact on the results presented in this study is overall small. These results identify strengths and weaknesses of each product for future improvement; they also emphasize the importance of using multiple gauge-radar and satellite products along with their uncertainties in evaluating reanalyses and models.