Statistical Downscaling Techniques and Projection of Future Climate Extremes in the Sudano Sahelian Environment

The Sudano-Sahelian Region of Cameroon (SSRC) faces significant challenges due to high rainfall variability and rapid population growth. Despite long-standing adaptation strategies, the region's vulnerability to climate variability and change remains a critical concern, prompting extensive research and calls for greater adaptation funding. In Sahelian West Africa, the decline in rainfall stations has significantly reduced data availability, hindering the calibration and evaluation of climate models and limiting their ability to accurately represent the region's climate. However, there are notable discrepancies between global and regional models regarding projected changes in precipitation patterns across specific regions and seasons, particularly in areas like the Eastern Sahel region, which includes the SSRC. Bias correction (BC) and downscaling (DS) are crucial, as these bias can be propagated into impact models. This study aims to fill the gap of localized and reliable information for climate services in the Sudano Sahelian region.

Using high-resolution rainfall data from NoCORA—daily interpolated rainfall maps for Northern Cameroon based on 418 stations (1948–2022) at 0.01° resolution (Jérémy et al., 2023)—the 25-km resolution regional climate models derived from a CMIP5 model are employed to better capture the climatology of extreme precipitation events, with kilometer-scale bias correction applied to outputs over the study area. Additionally, a subset of CMIP6 simulations is statistically downscaled to evaluate local-scale model uncertainties and compare the effectiveness of statistical and dynamical downscaling methods.

This study evaluates the performance of four state-of-the-art statistical downscaling techniques namely Linear Scaling, CDF-t, Quantile Mapping and Quantile DeltaMapping using different metrics and compares extreme precipitation changes under climate change scenarios to identify a suitable method for correcting bias in climate models projections for the period 2005-2100. The findings of this study will help impact modelers by enhancing the application of bias adjustment methods, thereby supporting the development of robust local climate change impact assessments in agriculture and hydrology domains.

Keywords : extreme precipitation, biais correction, Statistical downscaling, climate models