Improved ET assimilation through incorporating SMAP soil moisture observations using a coupled process model: A study of U.S. arid and semiarid regions

摘要

An accurate estimation of evapotranspiration (ET) is essential for characterizing the water budget in arid and semiarid ecosystems. Although various soil moisture data have been used to improve the hydrological process modeling, only a few studies improved ET estimation at a global scale by utilizing satellite soil moisture active passive SMAP data, particularly targeting arid and semiarid areas. To address this issue, this paper proposes a process-based assimilation scheme (LPJ-SM_A) to simulate daily ET at 0.25° spatial resolution for the water limited areas. First, an integrated model (LPJ-PM) is constructed by the updated Priestley-Taylor Jet Propulsion Laboratory model (PT-JPL_(SM)) and the Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM). As the PT-JPL_(SM) model establishes a connection between soil moisture (SM) and ET, the SMAP data could be integrated into LPJ-PM. Second, with the estimated ET_(PM) (3-day interval) in LPJ-PM as "observations", the original ET_(LPJ) (daily) estimated by LPJ-DGVM could be well constrained in a water-limited area through data assimilation. The results showed that: (1) the ET_(PM) with SMAP information performed better and had a higher accuracy than ET_(LPJ). (2) After assimilating ET_(PM) into LPJ-DGVM, the assimilated-ET (ETDA) showed a superior performance (R = 0.75,RMSD = 0.72 mm/d) to ET_(LPJ) (R = 0.55, RMSD = 1.02 mm/d) and ET_(PM) (R = 0.70, RMSD = 0.93 mm/d) when evaluated against in situ observations at a 95 significance level. Our proposed assimilation system (LPJ-SM_A) was applied to arid and semiarid regions in the United States, and the results illustrate that the spatial distribution and annual value of the LPJ-SM_A ET was very similar to NLDAS-2 ET products, which have a higher precision over North America than other global 'reference' products. The proposed LPJ-SM_A system can be used to optimize model simulation performance and effectively improve ET prediction accuracy. This method can be used