The Improvement Of POD/NLS-4DVar Method And Its Application In The Land Data Assimilation System

Data assimilation is an effective way to estimate the soil moisture with the model simulation and observation information. Data assimilation algorithm is the methodology mixing the simulated and observed data. The POD/NLS-4DVar algorithm owns the advantages of the two current mainstream data assimilation algorithms,that is, EnKF and 4DVar.Though many advantages, this paper aims at finding out the strategies to solve the problems within the POD/NLS-4DVar algorithm.The background error covariance( B) plays an important role in data assimilation system. To improve the background error covariance estimation, using ensemble forecast statistics to produce B is being a more attractive approach. The ensemble-based data assimilation methods have the ability to evolve flow-dependent estimates of forecast error covariance(i.e., the background error covariance B) by forecasting the statistical characteristics and its final algorithm after the localization implementation will be certainly changed with the choice of its localization scheme. However, the use of finite ensembles to approximate the error covariance inevitably introduces sampling errors that are seen as spurious correlations over long spatial distances. Such spurious correlations could be ameliorated through the localization process. This paper puts forward two localization strategies,i.e., the adaptive localization and the expanding sample localization, applying to the POD/NLS-4DVar algorithm testing with the Lorenz-96 model,and the results are as follows:The purpose of this study is to describe an economical approach to an existing adaptive localization technique and its implementation in the POD(proper orthogonal decomposition)-based ensemble four-dimensional variational assimilation method(POD-4DVar). Owing to the applications of the sparse processing and the EOF decomposition techniques, the computational costs of this proposed Sparse Flow-Adaptive Moderation localization(SFAM) scheme are significantly reduced. The effectiveness of the POD-4DVar with the SFAM localization is demonstrated by using the Lorenz-96 model in comparison with the SENCORP(i.e., Smoothed ENsemble Correlations Raised to a power) and static localization schemes, respectively. The performance of POD-4DVar with SFAM localization shows a moderate improvement over the schemes with SENCORP and static localization with low computational costs under the imperfect model(F=9).As a step in the improvement of the NLS-4DVar method, we implement an expanding sample localization scheme into NLS-4DVar with an aim at removing the additional linear assumption adopted in the original NLS-4DVar algorithm. This paper derives an improved iterative scheme for NLS-4DVar by the expanding sample localization scheme avoiding the extra linearization assumption with the stable algorithm convergence. Numerical experiments with a nonlinear model of the Lorenz-96 equation show that the proposed new iterative scheme, avoiding the direct inverse of a matrix, produces slight better solutions against the original one without increasing the computational costs.This thesis based on LDAS-IAP/CAS introduces the adaptive localization technique to filter out spurious correlations between the microwave brightness temperature observations and the soil moisture content in the 10 soil layers effectively. The LDAS-IAP/CAS is implemented and evaluated through a regional assimilation experiment over the southern North China(110°-120°E, 34°-40°N) from July 2005 to June 2010. The AMSR-E microwave brightness temperature data are used to run this system. And the results from assimilation experiments indicate that the assimilation of soil moisture from LDAS-IAP/CAS which is implemented the adaptive localization scheme performs better for the in-situ measurements of soil moisture and can capture the temporal evolution of the observed soil moisture reasonably well.