| Soil moisture is a crucial variable in agriculture, hydrology, meteorology and plays an important role in water cycle. It affects surface evaporation, runoff, sensible and latent heat distribution. How to get the high precision continuous soil moisture information has been always the study focus. Since 1990 s, data assimilation method has been gradually introduced into the earth surface science and hydrology field by scientists carrying out the research work of data assimilation of soil moisture. Although soil moisture data assimilation has experienced nearly 20 years of development, there are still some problems need to be solved, such as root zone soil moisture estimation; spatial heterogeneity of soil properties parameters; model prediction error and observation error covariance matrix estimation in assimilation system. In particular, the soil properties parameters uncertainties of the present model limits the improvement of soil moisture assimilation accuracy. In view of the above questions, this dissertation sets Ensemble Kalman Smoother(EnKS) and the Ensemble Kalman Filter(EnKF) assimilation algorithms as the foundation, and uses the upper reaches of the Heihe River Arou freezing and thawing observation station data to carry out the state-parameter simultaneous estimation of soil moisture assimilation research.The main research contents and conclusions are as follows:1) The effects of soil properties parameters uncertainties and soil moisture observation frequencies on the performance of soil moisture assimilation system. Based on the Simple Biosphere Model 2(SiB2), the use of surface soil moisture observation data, the study compared the EnKS and EnKF for soil moisture estimation ability by the degree of soil properties parameters uncertainty and soil moisture observation frequencies effect. The results show that the soil properties parameters uncertainties affect the results of the simulation of the surface soil moisture most, impact on the deep is relatively small; using the EnKF and EnKS assimilation of surface soil moisture observation data, can significantly increase the surface and root zone soil moisture estimation accuracy; En KS is affected by the parameters uncertainties and the observation frequencies less than EnKF, and the former is better than the latter for the estimation of the surface soil moisture. In the case of sparse observation data, the performance of EnKS algorithm is more prominent.2) The simultaneous estimation of soil moisture and soil properties parameters. Based on the SiB2 model, EnKS and EnKF algorithms, two dual-pass methods are developed for the state-parameter simultaneous estimation base on dual EnKS and EnKS-EnKF. Three numerical experiments were carried out based on the site observation data, experiment on soil properties parameter hierarchical optimization, and to study the effect of observation frequencies of dual EnKS and EnKS-EnKF and the two methods of root zone soil moisture estimation ability. The experimental results show that the dual-pass method can optimize the parameters of the surface and subsurface layer by using the observed data of surface soil moisture; EnKS-EnKF and dual EnKS in the parameter estimation influenced by the observation frequencies and the number of error is less than dual EnKF, i.e. EnKS in the parameter optimization ability is more robust than EnKF. Combined with the results of three numerical experiments, the estimation of the dual EnKS on the surface and root zone soil moisture is optimal, and EnKS-EnKF is second. When the observed data is sufficient, taking into account the storage and time consuming problems, EnKS-EnKF can be used to replace the dual EnKS, when the data are scarce, the use of dual EnKS is more appropriate.3) The modified method to reduce the influence of the forcing data error, the background error and the model structure uncertainty on the parameter estimation. Based on En KS-EnKF, a simple and effective method is introduced to reduce the influence of other errors in the process of parameter optimization. The sensitivities of the improved EnKS-EnKF method in the aspects of set size, observation frequencies, forcing data and assimilation time window length have been analyzed. The results show that the improved EnKS-EnKF can reduce the influence of error on parameter estimation, is sensitive to the observed frequencies, forcing data and the length of the assimilation time window, less than the standard EnKS-EnKF with higher parameter estimation accuracy and stronger robustness; After verification, when the collection number is more than 50, the improved EnKS-EnKF estimation in parameters and soil moisture can stabilize; Further analysis shows that the improved EnKS-EnKF method is more applicable.4) The estimation of model prediction error and observation error covariance matrix in EnKF algorithm. With the evolution of time, the EnKF algorithm often underestimate the forecast error covariance matrix, and model generation of ensemble prediction often do not contain the true value, so on the basis of the Study 3, the inflation algorithm with the feedback information is used to improve the EnKF to optimize forecast error and observation error covariance matrix estimation. The results show that the EnKF with the inflation algorithm plays a role in the improvement of soil properties parameters and soil moisture estimation.5) Application of improved EnKS-EnKF method in the estimation of soil moisture in farmland. In soil moisture assimilation system based on Study 3, we use COsmic-ray soil moisture observing system(COSMOS) and COsmic-ray soil moisture interaction code(COSMIC) inversion of soil moisture observation data to simultaneously estimate soil moisture and soil properties parameters. The results show that the improved EnKS-EnKF method can effectively reduce the influence of the precipitation error on the parameter estimation during the irrigation period; The improved EnKS-EnKF method is superior to the standard EnKS-EnKF method for the estimation of soil properties parameters and soil moisture. |
PDF Full Text Request