ADJOINT SENSITIVITY-BASED DATA ASSIMILATION METHOD

摘要

The present invention relates to a data assimilation method based on the accompanying model sensitivity, and more specifically, (1) integrating the numerical model from the start time to the end time of the assimilation window using the first guess as the initial condition Generating an example guarantee of the end time; (2) generating a reference state using the observation data at the end time and the guarantee; (3) defining a response function using a forecast error that is a difference between the generated reference field and the forecast field; (4) integrating the accompanying model from the end time to the start time to generate the accompanying model sensitivity to the forecast error of the moving picture window start time; (5) determining the magnitude of the generated accompanying model sensitivity; And (6) generating an improved initial condition using the initial estimate and the magnitude of the associated model sensitivity. According to the data assimilation method based on the accompanying model sensitivities proposed in the present invention, by generating the initial sensitivity of the model for the forecast error calculated by integrating the numerical model and generating the improved initial condition using the numerical model and the accompanying model The initial condition can be generated while remarkably reducing the computational cost compared to the 4-dimensional reanalysis data assimilation method, and the initial guess error and the observation Observations errors are not associated and more precise initial conditions can be generated.