Avaliação da extensão da janela temporal de um sistema de assimilação de dados 4D-Var na costa sudeste brasileira

Abstract

Obtaining the best estimate of sea state from available observed data and physicalmathematical models is a crucial part of any operational ocean modeling and forecasting system. As many parameters of a 4DVAR assimilation system, the choice of the assimilation window length is widely based on a heuristic approach. In this work, an established 4DVAR data assimilation system for the southeastern oceanic region of Brazil is used as a test platform for the objective assessment of the assimilation window length. Based on a twin experiments approach, three windows (4, 7 and 10 days) are tested using real observations - SST, SSH and in situ hydrographic profiles - over a period of 6 months. Objective error metrics and cost function minimization analyzes are related to oscillations of nonlinearity in ocean dynamics to support discussions on the validity of the tangent linear assumption. The performance of the experiments with 4 and 7-days window lenght were superior to the 10-days and presented smaller errors especially in periods with high turbulent kinetic energy and consequent difficulty of the systems to keep the validity of the linear assumption.