Estimating Surface Heat And Water Fluxes Using Ensemble Kalman Filter

Surface water and heat flux are important parts of the energy exchange in the Earth system. They also play extremely important roles in environmental monitoring, agricultural production and climate change research when being estimated by a variety of methods in various spatial and temporal scales. Current methods to estimate these surface fluxes including land surface model simulations, ordinary site observations and satellite data inversion. Three methods have their own advantages and disadvantages, but only model simulation method can be used to obtain temporally and spatially continuous surface water and heat fluxes. In recent years, the use of the land data assimilation method in land surface model can improve the simulation results of the land surface model by introducing observations, which is becoming the best way to get an accurate estimate of continuous surface fluxes.In this thesis, sensible and latent heat flux were obtained by direct assimilation of site observations into NCAR Community Land Model3.0(CLM3.0) in order to improve the global land surface model flux estimates for a preliminary study. First, we selected data from three Ameriflux sites and Heihe remote sensing experiment station, Arou to measure and assess the analogous performance of the global land surface model and the effects of model simulations for estimating regional surface water and heat fluxes. Second, data assimilation framework was built based on CLM3.0as a dynamic framework, which selected ensemble Kalman filter algorithm. Three sites in the Heihe River Basin (Arou, Guantan and Yingke) were obtained and used to appraise the newly built data assimilation framework.This study drew the following conclusions:(1) Specific tests to verify the performance of CLM3.0under different underlying surfaces in the mid-latitudes showed model’s ability to accurately simulate the daily averaged sensible heat flux and latent heat flux and the accurate performance on describing the flux intra-day trend. However, there exist systematic errors of CLM3.0when simulating heat fluxes from the underlying surface of the forest, frozen soil and crop fields. (2) The spatial distribution of simulated global seasonal and annual averaged sensible and latent heat flux to the surface flux had direct correlations with the noon solar altitude and atmospheric humidity. Intensive solar radiation in the region associated with high flux while flux exchange in the polar region away from direct insolation point is close to zero. Areas of high water vapor content have greater latent heat exchange than the sensible heat exchange, arid region flux exchange of sensible heat flux.(3) This paper is an initial attempt to establish a direct assimilation system of surface heat flux for CLM3.0. Validation and assessment for assimilation system to improve the accuracy of heat flux estimation in three different underlying surfaces were processed. As can be seen from the results, regardless of the performance of original simulation result, assimilation algorithm is capable of introducing the observed new information into model to improve the simulation, but the improvement is subject to the behavior and characteristics of land model.(4) Assimilated simulation results are sensitive to the initial perturbation amplitudes of observations and atmospheric forcing data while little responses can be found in the result when perturbing the initial condition. With the observation error decreases and the atmospheric forcing data error increases, the assimilated simulation approached to observations. Accurate estimation of a priori observing error can lead to more accurate assimilation results obtained which are closest to the real situation, but if continue to reduce the a prior observing error the assimilated simulation changed very little. In the case of the atmospheric forcing data, we haven’t found such situation.