Research On Ensemble Data Assimilation Algorithm Based On Fuzzy Logic

In the face of such an extremely complex earth surface system,how to express the dynamic evolution of various processes accurately and reduce the estimation error as much as possible has become a hot issue in the field of geoscience.As an optimal method,data assimilation needs to be carefully addressed in the process of effectively integrating observational data and geosciences models through assimilation algorithms.In the actual data assimilation,the background error covariance is underestimated because of the spurious correlation which overestimates the correlation between the observation points and the state update points that are far away even if the true correlation is small,and then the observation weight is so smaller and smaller that the analysis values are difficult to express the true state of the system,from which the filter divergence is generated.To solve the above problems,localized methods have been proposed to improve assimilation performance.In this paper,based on the ensemble Kalman filtering framework,a new localization scheme based on fuzzy logic is proposed.(1)The observation error theory,ensemble Kalman filter theory,localization algorithm and fuzzy logic theory in data assimilation are reviewed.(2)Based on the ensemble Kalman filter algorithm,the fuzzy logic controller is designed to realize the coupling with fuzzy logic idea and localized algorithm,and the validity of the new method under model error,background error covariance and Kalman gain is tested based on nonlinear lorenz-96 model.The results show that the new method can effectively restrain spurious correlation and optimize assimilation analysis.(3)Based on the lorenz-96 model,the sensitivity and robustness of the new method under the change of ensemble size,observation number,covariance inflation factor,and forcing factor are demonstrated.The results show that,the new method is less sensitive to covariance inflation factor and performs well at a smaller overall scale compared to the traditional method,and further compared and analyzed the Power spectrum density values of the new method under different observation errors.Overall,the new method has strong robustness.(4)Based on the QG model,the applicability of the new method in the actual assimilation system is verified,and the true value field,the analysis field and the error field state distribution characteristics of the new and old methods in the high dimensional state space are analyzed in detail.The results show that compared with the traditional distance-based localization,fuzzy logic-based localization not only achieves better performance in theory,but also is more convenient to implement and use in more practical atmospheric models.To sum up,the experimental results of this thesis strongly demonstrate the superiority of the new method in the ensemble data assimilation system.although it is desirable to use fuzzy logic algorithms in theoretical and experimental simulations,the importance of using fuzzy logic in practical assimilation remains unclear and needs to be further explored in future studies.