A Multivariate Regression Model Between the October Rainfall Anomalies in Central America and the Tropical Pacific and Atlantic Ocean

Abstract

Principal Component Analysis was used to identity common anomaly patterns amongst 72 rainfall gauge stations in Central America during October, in order to identify stations to from October Rainfall Indices. October was chosen because it represents the center of the rainfall seasons, and previous studies have shown how ocean-atmosphere relations with Central America rainfall vary according to the rainfall month. Five rainfall regions where identifies through this process, and the standardized rainfall anomaly time series was calculated for each region. A Multivariate Regression model was fitted to quantify the ocean-atmosphere interaction between the Tropical Pacific and Atlantic Ocean indices and each of the Rainfall indices. These models show that the Niño 4 region has the largest influence over the region when compared with the influence of other candidate ocean indices, Niño 4 having negative correlation with all the Rainfall Indices. In addition, the Tropical North Atlantic index was found to have positive correlations with some of the Central America October rainfall regions. This work shows that the variability of the Tropical Eastern Pacific sea surface anomaly (SSTA) presents stronger associations with the October Central America rainfall, than the Tropical North Atlantic SSTA. This is in contrast to stationary studies, which was previously shown to be more strongly associated with the Tropical North Atlantic SSTA. It is though that the October result is mainly related to the Pacific SSTAs modifying the depth and degree of development of mesoscale convection.