Surrogate model optimization on fast data assimilation

Abstract

Building a simulation to reproduce observed data from the real world is an important issue, and data assimilation is an essential process before its operation. The model bridge method has been proposed for the fast data assimilation of target data by using a surrogate model and the results of data assimilation to past data. In this study, we propose a structure optimization method for the surrogate model based on the model selection and the feature selection to improve data assimilation accuracy.