Study On Observation Error Of Observation Data Assimilation System

In data assimilation(DA)scheme,observation error is one of the main sources of system uncertainty.In order to get more accurate forecasting value which requires the more accurate values of current state,but the current state value updating depends on weather the observed value can provide data effectively,so the key of observation error directly affects the assimilation finally effective.How to use the ground observation data effectively and extract from the valuable information of model,then put them and other data fusion initial data needed for numerical weather prediction,improve the prediction skill of numerical model,there is a lot of relevant workers are trying to solve the problem.In the case of assimilation,there would lead to spurious correlations,if the observation position has long-distance from the state and the ensemble numbers is too small.In order to reduce the observation error and eliminate spurious correlations,the main research work of this paper is divided into the following aspects:(1)From the data assimilation theories,analyzes and discusses the source and mechanism of data assimilation observation error and spurious correlations,then put forward the Fuzzy Control algorithm,which can measure the distance between observation position and status updates,and output the corresponding weights.(2)Through coupled the Fuzzy Control algorithm with data assimilation algorithm,that can get more accurately analysis,in order to solve this problem and in the high dimensional chaotic Lorenz-96 model,coupling the Fuzzy Control algorithm with Ensemble Transform Kalman Filter compared with Local Ensemble Transform Kalman Filter in ensemble numbers,observation numbers,model steps,inflation numbers,forcing parameters.Then comparison the local ensemble transform kalman filter coefficients,and the covariance localization gain matrix assimilation effect.The results show that the coupling Fuzzy Control algorithm can effectively improve the filtering accuracy in the assimilation system and get more accurate prediction values.(3)By comparison of the coupled Fuzzy Control algorithm that obtain better filtering effect,and the analysis value is more close to the real value,but in the process of experiment,because each state update requires Fuzzy Control algorithm for distance judgment on the observation position,so in the process of calculation,need extra time to get the final analysisvalue.We apply the basis of widely used Ensemble Transform Kalman Filter in data assimilation scheme,and coupled with fuzzy mathematics and fuzzy theory as the basic idea,proposed fuzzy control data assimilation scheme that handle observation error in data assimilation process.The results of the study provides a reference value for using in actual application of data assimilation scheme in the future.