Improving spatial patterns prior to land surface data assimilation via model calibration using SMAP surface soil moisture data

Authors: ZHOU, J. - Hohai University; WU, Z. - Hohai University; Crow, Wade; DONG, J. - US Department Of Agriculture (USDA); HAI, H. - Hohai University

Submitted to: Water Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/18/2020
Publication Date: 10/6/2020
Citation: Zhou, J., Wu, Z., Crow, W.T., Dong, J., Hai, H. 2020. Improving spatial patterns prior to land surface data assimilation via model calibration using SMAP surface soil moisture data. Water Resources Research. 56:10. https://doi.org/10.1029/2020WR027770.
DOI: https://doi.org/10.1029/2020WR027770

Interpretive Summary: The assimilation of satellite-based surface soil moisture estimates into a land surface model is commonly used to improve the quality of model-based soil moisture estimates for agricultural drought monitoring. However, to date, these efforts have been based on assimilation approaches that utilize only temporal information contained in the satellite estimates. This is inefficient because these retrievals also contain neglected spatial information. This paper describes a new calibration-based data assimilation approach that allows a water balance model to benefit from both spatial and temporal information contained in satellite-based soil moisture products. The paper describes the approach and, for verification purposes, applies it to a region of China with a dense ground-based soil moisture network. Results illustrate that the approach is capable of better leveraging spatial and temporal information contained in satellite-based soil moisture products than existing approaches and provides for the more efficient correction of error in model-based estimates.

Technical Abstract: Spatial information described in patterns of remotely sensed soil moisture (SM) retrievals is generally discarded after rescaling to a model climatology in model calibration and data assimilation applications. This study explores an alternative approach for bias correction of modeled SM that provides the benefits of traditional rescaling while also retaining spatial information present in remotely sensed SM. To conduct the model calibration procedure, a daily Variable Infiltration Capacity (VIC) model simulation with a spatial resolution of 25 km^2 is constructed in the Huai River Basin, China. Two VIC soil parameters (bulk density and hydraulic conductivity related constant) are then calibrated against the Soil Moisture Active Passive (SMAP) Level 3 surface SM product. Through validation against in-situ observations acquired from a dense SM monitoring network, the benefits of such model calibration are investigated. Results show that the model calibration method successfully corrects the bias between VIC SM and the SMAP retrievals, and thecalibrated soil parameters lead to more accurate VIC-simulated SM spatial patterns and slightly improved VIC-simulated SM time series. Furthermore, these SM improvements eventually translate into enhanced VIC streamflow estimates and have only a very limited impact on VIC evapotranspiration (ET) estimates. While relative spatial patterns present in SMAP retrievals (after the removal of basin-scale biases) are of benefit for the model calibration, basin-scale mutual biases between SMAP and VIC SM products must still be corrected to avoid degraded streamflow estimates and discontinuous ET estimates.