[A-AS02] From weather predictability to controllability

[AAS02-P05] Evaluation of CMIP6 GCMs for Water Resources Modeling in The Tana River Basin, Kenya.

*Daniel Mwendwa Wambua¹, - Hiroaki Somura¹ (1.Okayama University)

Keywords:CMIP6, GCMs, KGE, pr, tasmax, tasmin

Eastern Africa is one of the most terrain complex areas of the world ranging from its costal strips, through plains and plateaus across the rift valley and up the highest peaks in Africa. Due to this terrain complexity and the coarse resolution of General Circulation models (GCMs) products, it becomes quite difficulty for most GCM products to capture the spatial variations of weather parameters especially on short temporal intervals in such terrain complex areas. The newly released set of the World Climate Research Programme (WCRP) Coupled Model Intercomparison Project 6 (CMIP6) GCMs have been shown to have improved representation of weather parameters across the globe compared to the Coupled Model Intercomparison Project 5 (CMIP5) products. The aim of this study is to evaluate a set of 19 CMIP6 GCM for precipitation (pr), maximum temperature (tasmax), and minimum temperature (tasmin) in the Tana river Basin of Kenya, east Africa. The Kling Gupta Efficiency (KGE) statistic was computed for each parameter and GCM and for each station located at different points in the basin for the GCM hindcast. To determine the best performing GCMs, the multiparameter Multi-station KGE was calculated to select the best performing GCM at the Basin scale. Most of the GCMs showed excellent performance in the prediction of tasmin and tasmax. NorESM2-MM with a KGE of 0.819 in Kitui and 0.734 as an average for all stations performed best for tasmax. For tasmin, GFDL-ESM4 performed best with a KGE of 0.648 when averaged across all stations. However, the performance in the prediction of pr was very low for all the GCMs except MPI-ESM1-2-LR which gave a KGE statistic of .0799 for Kitui met station and 0.347 when averaged across all three stations. The final selection narrowed down to a set of five GCMs namely, CMCC-ESM2, MPI-ESM1-2-HR, ACCESS-CM2, NorESM2-MM, GFDL-ESM4 with a the multiparameter Multi-station KGE statistics of 0.455, 0.464, 0.470, 0.482 and 0.511 respectively. A set of best performing CMIP6 GCMs was identified in this study. The selected set of the CMIP6 showed an average performance for the selected weather parameters. The relatively low KGE was attributed to the relatively low ability of the GCMs to predict pr for the study area. The future predictions for each of the selected GCMs for ssp126, ssp245 and ssp585 scenarios will be accessed from the WCRP site, manipulated using climate data operator (CDO) in Cygwin and downscaled using SD GCM software for water resources and crop water demand evaluation in the Tana river basin.