Bayesian fitering-based general data assimilation method

Abstract

The invention discloses a bayesian fitering-based general data assimilation method. The method comprises the steps of: inputting an initial value set into an analysis model in a prediction step so as to obtain a prediction set value; calculating prediction error covariance matrix by using set kalman filtering in an updating step, and updating each prediction set according to the observation value and kalman gain matrix; or, calculating importance weight of each set sample by adopting particle filtering through set prediction value, calculating the number of effective particles by utilizing normalization importance, resampling the set according to the weight to obtain updated analysis value and analysis set; or, calculating prediction error covariance matrix by adopting unscented kalman filtering, and updating each prediction set according to the observation value and kalman gain matrix; conducting next prediction and assimilation by taking the updated analysis set as the initial values of the analysis model, and repeating the prediction step and the updating step. The method can enable Earth remote-sensing observation information and land surface process model information to be effectively integrated, thus forming a land surface process information prediction system with small errors.

Description

technical field [0001] The present invention relates to the field of earth system scientific information processing, in particular to a general data assimilation algorithm based on Bayesian filtering, which enables the effective fusion of earth remote sensing observation information and land surface process model information, thereby forming an Forecasting system for land surface process information (such as soil moisture, soil temperature, etc.). Background technique [0002] The core idea of ​​land surface data assimilation is to integrate the direct and indirect observations from different sources and different resolutions through the data assimilation algorithm within the dynamic framework of the land surface process model, and combine the land surface process model with various observation operators (such as radiation Transmission model) is integrated into a forecast system that continuously relies on observations to automatically adjust the model trajectory and reduce ...

AI Technical Summary

Problems solved by technology

[0005] 1) Both the ensemble Kalman filter and the classic extended Kalman filter assume that the prior probability distribution of the error is a multidimensional Gaussian distribution. Therefore, the accuracy of these two algorithms is not good enough when solving non-Gaussian problems; in addition, it is necessary to focus on It is pointed out that the variables that obey the Gaussian distribution will also show the characteristics of non-Gaussian distribution after being transformed by the nonlinear system

[0006] 2) Variational assimilation methods (such as 3DVar and 4DVar) generally require the adjoint operator of the model, but in the field of land surface data assimilation, since most land surface process models and radiative transfer models are nonlinear, and the land surface process model The discontinuity of the threshold value is widely used (Qiu Chongjian, 1997), so the adjoint operator of the land surface process model is usually difficult to obtain (Reichle et al., 2002)

However, more accurate algorithms are still needed to make the data assimilation system more accurate

Images