Ajuste de un modelo VAR como predictor de los campos de anomalías de precipitación en centroamérica

Abstract

Cluster analysis was used to identify common patterns of 15 precipitation points of the region, using their anomaly time series as grouping variables. Five clusters where identified through this process. A Vector Auto Regressive model was fitted tothe data to quantify the ocean-atmosphere interaction between the oceanic indices of the Tropical North and South Atlantic, the Tropical Eastern Pacific and the first empirical orthogonal functions of the regional rainfall clusters. This model shows that the Tropical North Atlantic has the largest influence over the region when compared with the influence of the other indices, having positive correlation with all the rainfall. The Tropical South Atlantic and the Niño 3 indices, instead, were found to have no correlation with the rainfall of the region when an stationary model is fitted. This work shows that the variability of the Tropical North Atlantic sea surface temperature anomaly presents stronger associations with the Central America rainfall than the Tropical Eastern Pacific sea surface temperature anomaly. The association is mainly related to the degree of development of the Tropical Upper Tropospheric Trough.