Statistical downscaling estimation of recent rainfall trends in the eastern slope of the Andes mountain range in Argentina

Abstract

Statistical models for rainfall downscaling based on multiple linear regression techniques have been developed and tested in the Andean Region of west Argentina, an extended mountainous region where three different rain regimes predominate and rainfall has great spatial and temporal variability. The verification procedure was focused on the model’s ability to reproduce observed rainfall trends in recent decades. In the northwest of Argentina, domain of the tropical summer rain regime, the monthly rainfall variance accounted for by downscaling models was 77% on average and models reproduced satisfactorily the negative linear trend observed in the last two decades of the past century. In the arid central-west Argentina, a region of rapid transition between two different rain regimes, model performance was rather poor (an average of 50% of explained variance), even so models were able to capture outstanding differences in the linear trend between the northern and southern sectors of the region. In the southwest of Argentina, domain of the mid-latitude winter rain regime, the monthly variance accounted for by downscaling models was 71% on average and models were capable to reproduce a singular change in the onset of the rainy season that occurred during the 1990s. The results achieved demonstrate that it is feasible to establish significant and useful statistical relationships between atmospheric variables and rainfall at monthly and river basin scales, even for a topographically complex region like western Argentina.