Evaluation of four bias correction methods and random forest model for climate change projection in the Mara River Basin, East Africa

This study evaluates the performance of four bias correction methods based on CORDEX (coordinated regional climate downscaling experiment) domain six regional climate models (RCMs) at the Mara River Basin. A suitable bias correction method was considered to develop the future climate scenario. The performance of bias correction methods was evaluated by various statistical metrics based on the historical period and revealed that the distribution mapping (DM) techniques have strong performance under the different climatic conditions. The effectiveness of the DM method is found better in capturing the coefficient of variation and standard deviation of observed rainfall and temperature. Therefore, this study considers the future climate (2026-2095) from bias-corrected RCMs output using DM techniques. The results from bias-adjusted RCMs unfolds an increase of rainfall (+118.3%) and temperature (+2.91) in the future climate under Representative Concentration Pathways (RCPs) 2.6, 4.5 and 8.5. In addition, this study tested the random forest (RF) method to determine the capacity of each bias-corrected RCMs for reproducing the future rainfall and temperature under the RCP 4.5 and RCP 8.5 scenario. The results demonstrate that the RF can reproduce the climate variable with its average correlation (R2) of 0.93 for rainfall and 0.95 for temperature.

Automated summary

This study evaluates the performance of four bias correction methods based on CORDEX domain six regional climate models in the Mara River Basin and finds that distribution mapping techniques have strong performance. The results from bias-adjusted RCMs show an increase of rainfall and temperature in the future climate. Additionally, the random forest method demonstrates good capacity for reproducing future climate variables.