Applying statistical downscaling to CMIP6 projections of precipitation for South America: Analysis of pre and post-processed simulations

Abstract

Global Climate Models (GCMs) are fundamental for simulating future climate conditions. However, such tools have limitations like their coarse resolution, systematic biases, and considerable uncertainties and spread among the projections generated by different models. Thus, raw outputs from GCMs are insufficient for regional-scale studies, which can be solved using downscaling techniques. These methods are particularly relevant for South America (SA), given the continent's climate regimes and topographic complexity. Moreover, critical socio-economic activities developed in SA, such as rainfed agriculture and hydroelectric power generation, are highly dependent on climate conditions and susceptible to extreme events, which can lead to intense droughts or floods depending on the region. Given the background, this study aims to analyze the performance of the statistical downscaling technique Quantile Delta Mapping (QDM) applied to precipitation projections simulated by an ensemble composed of eight GCMs from the Coupled Model Intercomparison Project Phase 6 (CMIP6) for SA. In this manner, we evaluate both the original precipitation projections from the GCMs, and after applying the QDM statistical downscaling technique. Daily precipitation data from the Climate Prediction Center (CPC), with a horizontal resolution of 0.5°, and from the Multi-Source Weighted-Ensemble Precipitation version 2 (MSWEPV2), with a horizontal resolution of 0.1°, are used as a reference, so the final resolution of the GCMs (and the ensemble) projections after the QDM technique application is the same from the different validation databases. Preliminary results with CPC indicate a satisfactory performance of the technique on precipitation simulations over SA. The authors thank the CAPES, the R&D Program regulated by ANEEL, and the companies Engie Brasil Energia and Energética Estreito for their financial support.