Predicción estacional del clima en Centroamérica mediante la reducción de escala dinámica. Parte II: aplicación del modelo MM5V3

Abstract

In the first part of this work it was determined that general circulation model (GCM) ECHAM4.5 shows more ability than CCM3.6 to simulate key climate features of Central America. For such reason, output from ECHAM4.5 was used to perform a dynamical downscaling experiment using the regional model MM5v3, in which a set of high-resolution simulations (of up to 30-km horizontal resolution) was generated for January 2000. The results of the dynamical downscaling allow to conclude that MM5v3 is able to suitably reproduce aspects of the Central American climate that GCMs cannot capture because of their coarse horizontal resolution, their limitations in representing both the regional topography and the mesoscale dynamical interactions. Comparison with data derived from observations indicates that the MM5v3 simulates the region of maximum low-level wind that is related to the Intra-Americas Seas Low Level Jet, although the regional model underestimates its intensity. Regarding the precipitation patterns, they agree with those derived from the observations (drier areas in the Pacific, wetter areas in the Caribbean). Nevertheless, there is a generalized overestimation in the amount of simulated rain. The analysis of the standard deviation for a twelve-member sample shows areas in which MM5v3 has greater dispersion or uncertainty (mainly to the south of Panama). Keywords: numerical models, seasonal climate prediction, dynamical downscaling, climate, climate variability.