Estimate Surface Heat And Water Fluxes By Assimilating MODIS LST Products

In this paper, we develop a land data assimilation scheme based on ensemble Kalman filter and Common Land Model version 1.0. This scheme is used to improve the estimation of soil temperature profile. The leaf area index (LAI) is also updated dynamically with MODIS LAI product; the MODIS land surface temperature (LST) product are assimilated into CLM.Sensitivity analysis: (1) Adding 1% stochastic noise to surface temperature, compare the output of Common Land Model for surface temperature, sensible heat flux and latent flux. Results indicate that, sensible and latent heat fluxes changed while surface temperature changed. The largest change is corn surface. When the RMSE of surface temperature changed about 3.05K, sensible and latent heat fluxes changed 36.03Wm-2,55.86 Wm-2 resepectively; The smallest change is for forest surface, sensible and latent heat fluxes changed 14.98Wm-2, 16.74 Wm-2 respectively. (2) Setting 2.0 as the reference of leaf area index (LAI), we test the results on the following conditions, (a) LAI=0.5; (b) LAI=3.5; (3) LAI=5.0. Results indicate that, Common Land Model is more sensitive to LAI when LAI is getting smaller gradually. We introduced LAI products of MODIS (MOD15A2) to CLM.Assimilation system: we choose Common Land Model (CLM) as the model operator and ensemble kalman filter as the optimization method to construct the assimilation system. We relate the MODIS LST and the CLM surface temperature; we take it as the observation operator of the assimilation scheme. We compare the ground temperature in situ observations with the MODIS LST, and take the RMSE as the observation error.Test data: The scheme was tested and validated using observations from three automatic weather stations (Blackhill, Bondville and Brookings) of the Ameriflux network. We assimilated MODIS LST and introduced MODIS LAI into CLM, choose the data of QC=0 as test data.Assimilation result: Results indicate that data assimilation method improves the estimation of surface temperature (0cm) and heat flux. The RMSE reduced from 81.5 Wm-2 to 58.4 Wm-2 at Blackhill site, from 47.0 Wm-2 to 31.8 Wm-2 at Bondville, from 46.5 Wm-2 to 45.1 Wm-2 at Brookings. The RMSE latent fluxes reduced from 88.6 Wm-2 to 57.7 Wm-2 at Bondville, from 53.4 Wm-2 to 47.2 Wm-2 at Blackhill site. In addition, assimilation of MODIS land surface temperature into land surface model is a practical way to improve the estimation of surface heat and water fluxes.