Method for improving ocean forecasting precision based on analysis of four-dimensional set variation

Abstract

The inveniton provides a method for improving ocean forecasting precision based on analysis of four-dimensional set variation. The method comprises the following steps: respectively adding disturbance values to an initial field guess value and an external forced field to form a set sample, rewriting an evolution equation for analyzing the four-dimensional set variation according to the set sample, and expanding a disturbance quantity until recursion to an initial moment; when the initial field guess value is irrelevant to the external forced field statistics, obtaining two special background field error covariance matrixes respectively containing tangent linear evolution matrixes and corresponding adjoint matrix information; performing disturbance expansion near the initial field guess value to obtain a relationship between a state variable at any moment in the future and a tangent linear evolution matrix and a disturbance quantity, and constructing a target function after disturbance; solving the gradient of the disturbed target function to the disturbance quantity; and the obtained initial field guess value is the optimal initial field guess value until the gradient of the disturbed target function to the disturbance quantity infinitely approaches zero. According to the invention, the applicability of the four-dimensional set variation method is improved, and the prediction error is effectively reduced.

Description

technical field [0001] The invention relates to a method for improving the accuracy of ocean forecasting. In particular, it relates to a method for improving the accuracy of ocean forecasting based on analytical four-dimensional ensemble variation. Background technique [0002] As an important part of the data assimilation system, data assimilation algorithms have made great research progress in the past few decades. However, in high-dimensional space, multi-scale, nonlinear, non-Gaussian, complex uncertainty and state quantity space Under the comprehensive effect of various factors such as the nature of data, the shortcomings of the existing data assimilation algorithms are becoming more and more prominent. [0003] To obtain an accurate prediction result, two conditions must be met: first, the numerical prediction model can simulate the evolution process of the ocean system accurately enough, that is, the physical laws controlling the movement of seawater are well describ...

AI Technical Summary

Problems solved by technology

However, when the existing four-dimensional ensemble variational method is applied to a system with strong nonlinearity, the effect will not be as good as the traditional four-dimensional variational method.

Images