Scenario Approach for the Optimization of Regularization Parameters in the Direct Variational Data Assimilation Algorithm

The problem of data assimilation for the advection diffusion model is considered. Data assimilation is carried out by choosing an uncertainty function that has the sense of the emission sources. Previously, a direct algorithm for data assimilation with a stabilizer in the cost functional governing the norm of the uncertainty function and its spatial derivative was introduced. In the paper, the assimilation parameters are found for a scenario with a known solution (training sample). The optimization is carried out by a genetic algorithm. The values found are used in scenarios with unknown emission sources (control experiment). The results of numerical experiments on solving a test problem are given.