Data assimilation system applied to Short-range Forecast System

This research carries out an evaluation of the 3DVAR, 3DEnVAR and 4DEnVAR methods applied to Short-range Forecast System (SisPI) with the objetive of determining which scheme presents the best results for short-term forecasting purposes. For this, three study cases are selected with initializations at 00:00 and 12:00 UTC, assimilating in a combined way PrepBufr and radiances data. The assimilation was carried out on the domain with the highest spatial resolution (3km), using a domain-dependent covariance matrix (BEC) generated from 15 days prior to each case study. Multiple outers loops are used in the case of the 3DVAR method, where the multiplicative weight of its control variables is also modified. This does not applied to hybrid schemes, a selection based on results obtained in international studies. On the other hand, the ensemble required for the hybrid methods was obtained from the previous outputs of SisPI, establishing a weight of 75/25 in relation to the perturbations of the ensemble and the BEC. The results suggest that 3DVAR modifies the background field poorly, which causes its results to rapidly converge to the solution without assimilation. On the contrary, the contribution of the flow-dependent errors in combination with the static errors contained in the BEC, in the hybrid schemes, show a significantly modify the first guess, leading to the assimilation effect being prolonged by an approximate threshold of 6 to 12 hours. The performance of the 3DEnVAR method is unstable, as it can lead to very realistic forecasts or others comparable with 3DVAR in the same situation. The 4DEnVAR scheme turns out to be the most appropriate because, although it does not always turn out to be the one that exhibits the most realistic solutions, it is the only one whose forecast always presented fewer errors with respect to the run without assimilation.