High-resolution boreal winter precipitation projections over tropical America from CMIP5 models

Abstract

Climate-change projections for boreal winter precipitation in Tropical America has been addressed by statistical downscaling (SD) using the principal component regression with sea-level pressure (SLP) as the predictor variable. The SD model developed from the reanalysis of SLP and gridded precipitation GPCC data, has been applied to SLP outputs from 20 CGMS of CMIP5, both from the present climate (1971–2000) and for the future (2071–2100) under the RCP2.6, RCP4.5, and RCP8.5 scenarios. The SD model shows a suitable performance over large regions, presenting a strong bias only in small areas characterized by very dry climate conditions or poor data coverage. The difference in percentage between the projected SD precipitation and the simulated SD precipitation for present climate, ranges from moderate to intense changes in rainfall (positive or negative, depending on the region and the SD GCM model considered), as the radiative forcing increases from the RCP2.6 to RCP8.5. The disparity in the GCMs outputs seems to be the major source of uncertainty in the projected changes, while the scenario considered appears less decisive. Mexico and eastern Brazil are the areas showing the most coherent decreases between SD GCMs, while northwestern and southeastern South America show consistently significant increases. This coherence is corroborated by the results of the ensemble mean which projects positive changes from 10°N towards the south, with exceptions such as eastern Brazil, northern Chile and some smaller areas, such as the center of Colombia, while projected negative changes are the majority found in the northernmost part.