Dual state/rainfall correction via soil moisture assimilation for improved streamflow simulation: Evaluation of a large-scale implementation with SMAP satellite data

Abstract

Abstract. Soil moisture (SM) measurements contain information about both pre-storm hydrologic states and within-storm rainfall estimates, both are essential for accurate streamflow simulation. In this study, an existing dual state/rainfall correction system is extended and implemented in a large basin with a semi-distributed land surface model. The latest Soil Moisture Active Passive (SMAP) satellite surface SM retrievals are assimilated to simultaneously correct antecedent SM states in the model and rainfall estimates from the latest Global Precipitation Measurement (GPM) mission. While the GPM rainfall is corrected slightly to moderately, especially for larger events, the correction is smaller than that reported in past studies because of the improved baseline quality of the new GPM satellite product. The streamflow is corrected slightly to moderately via dual correction across 8 Arkansas-Red sub-basins. The correction is larger at sub-basins with poorer GPM rainfall and poorer open-loop streamflow simulations. Overall, although the dual data assimilation scheme is able to nudge streamflow simulations in the correct direction, it corrects only a relatively small portion of the total streamflow error. Systematic modeling error accounts for a larger portion of the overall streamflow error, which is uncorrectable by standard data assimilation techniques. These findings suggest that we may be reaching a point of diminishing returns for applying data assimilation approaches to correct random errors in streamflow simulations. More substantial streamflow correction would rely on future research efforts aimed at reducing the systematic error and developing higher-quality satellite rainfall products.

A Unified Data‐Driven Method to Derive Hydrologic Dynamics From Global SMAP Surface Soil Moisture and GPM Precipitation Data

“…Deriving the governing hydrologic dynamics from satellite data is potentially useful for informing large-scale hydrologic model development and parameterization. Recent studies [e.g., Mao et al, 2019aMao et al, , 2019bNavari et al, 2019;Crow et al, 2019] have pointed out that data assimilation using SMAP to update modeled soil moisture states may offer limited improvement without a better model representation of hydrologic processes and an accompanying reduction in model systematic error. Since dominant hydrologic processes may be drastically different at various spatial scales, satellite SSM data provides direct hydrologic information at a resolution that is commensurate with the resolution of continental-scale distributed hydrologic models.…”