Representing Soil Moisture Spatial Distribution by Assimilating Satellite-based Soil Moisture Data into a Land Surface Model

Description

The spatial distribution of soil moisture (SM) is one of the key drivers in land-atmosphere interaction (LAI) processes. The heterogeneity of SM can modulate the magnitude and scale of the local circulations in the planetary boundary layer through surface heat fluxes, and subsequently influence the formations of cloud and precipitation. In the state-of-the-art atmospheric models, SM conditions are usually updated through land surface models (LSMs), which can also be run as a stand-alone model with prescribed atmospheric forcing. Nevertheless, the quality of atmospheric forcings and input land surface properties can directly impact how well the SM is represented in the LSM simulations. Satellite-based SM data is advantageous in its spatial coverage as opposed to much sparser in-situ observations, but remotely sensed SM datasets cannot fully describe the physical state of the land/soil system, in particular temperature and SM at deeper soil levels. Therefore, in this study, we aim to generate high-resolution SM analysis by assimilating NASA’s Soil Moisture Active and Passive (SMAP) SM data into the Noah multiparameterization (Noah-MP) LSM.

The Noah-MP LSM is configured with a domain encompassing the eastern part of the continental U.S. at 1-km grid spacing. Over the period from March 2015 to December 2016, the model-predicted soil moisture is constrained hourly by the SMAP soil moisture retrievals. We assess the impact of SMAP data assimilation (DA) by using the in-situ observations collected by the Oklahoma Mesonet and ARM’s soil temperature and moisture profiles (STAMP) system. The results indicate our approach does significantly improve the top-layer SM representation by reducing the overall dry bias in the area. Sensitivity experiments emphasize the benefits of using longer assimilation window, proper interpolation for meteorological forcing, and higher-resolution SMAP product (9-km versus 36-km). It has shown potential in better initializing Large Eddy Simulation (LES) and mesoscale models for studies of land-atmosphere-cloud (LAC) coupling.

(Supported by ICLASS SFA, Jerome Fast PI)