| Soil water is not only an important content of hydrology, atmosphere and land surface, been an important link of the surface water and groundwater, but also an important parameter to describe the energy exchange of land, atmosphere and vegetation growth. Accurate estimate of soil moisture is of great significance of food security and water and soil conservation in arid areas, but accurately estimate the soil water in real case is a very difficult and complicated work, however the remote sensing technology makes the long time and large area estimate possible. Using remote sensing technology and model of heat and water transfer to inversion information of soil moisture also has been widely used. Data assimilation technology has increasingly become the forefront of the research of ecological and hydrological processes and of remote sensing inversion, been a bridge to the coordination and integration of remote sensing and ecological modeling. It can integrate multi-source remote sensing data and the model simulation results efficiently, thus improve the accuracy of soil moisture prediction.By using MODIS and Landsat TM as data sources, this paper take advantage of TVDI(Temperature- Vegetation Dryness Index) as the observation operator, kalman filtering method been applied to one-dimensional hydrological model(HYDRUS-1D) to simulate surface soil moisture, the main conclusions are as follows:1 Compared with the commonly used NDVI(Normalized Difference Vegetation Index) Index, RVI(Ratio Vegetation Index) Index is more adapt to the high Vegetation coverage area, has sensitive response and instructions to the region of vegetation coverage is more than 50%, but insensitive to lower coverage area.2 Surface temperature, soil moisture and vegetation index have significantly relationship, but it takes a long time series of vegetation data. Used in the inversion of soil moisture TVDI model has higher correlation with surface soil, indicate that the TVDI is sensitive to surface soil moisture.3 According to the environment of study area as well as the predecessors’ experience, combined with empirical parameters and SCE UA method,the initial boundary conditions of model been estimated and determined after debugging, then output the simulation results of soil moisture. HYDRUS is more convenient to simulate the soil moisture, to reflect the status of soil moisture in the study area, and played a certain reference function to estimate soil moisture in large area.4 Ensemble kalman filter can well deal with nonlinear problem, when the remote sensing data is available, set the model empirical parameters and and the forecast of soil moisture as input, using soil moisture inversed by TVDI model as the initial input values to update model operator, using ensemble kalman filter to initialize the model parameters by updated soil moisture for next assimilation. Do next observation while remote sensing inversion data is available.5 Certain changes of soil moisture content, residual volume moisture content and saturated hydraulic conductivity have been made in the process of assimilation, The adjustment amount of each parameter have different trend over time. In the 0 to 10 cm, the adjustment of the parameter value of assimilation has certain effect.Assimilation effect lie in the continuously updated variables of remote sensing data of soil moisture inversion after collection, the continuously update of HYDRUS model parameters. It also means the adjustment of model parameters and state variables and reduce the prediction error of the model. Compared with using HYDRUS-1D model alone, assimilation accuracy of surface soil moisture improved obviously, the root mean square error reduced 0.0153, average error reduced 0.0426, multi-source remote sensing data shows great potential in the study of surface soil moistures simulation. |
PDF Full Text Request