UNCERTAINTIES IN GLOBAL AND REGIONAL CLIMATE CHANGE PROJECTION OF SUMMER MONSOON TEMPERATURE AND PRECIPITATION OVER WEST AFRICA

Abstract

This study aims at investigating climate change scenarios over West Africa with the associated uncertainties to improve the value of climate information to end-users for informed decision making. For the present day (1982-2005), the mean climatology, intermodel variability and spatio-temporal patterns of temperature and precipitation over West Africa from CMIP5, CMIP5_SUBSET (ensemble of GCMs driving CORDEX) and CORDEX multimodel ensembles (MMEs) were first evaluated and intercompared for the monsoon season (June-September). While CORDEX failed to outperform the simulated mean climatology of temperature by the CMIP5 ensembles, it substantially improved precipitation and provided more realistic fine-scale features tied to local topography and landuse. This improved performance over the region depend more on the internal models physics than the driving boundary conditions and results from a more consistent and realistic simulation of monsoon precipitation across the various Regional Climate Models (RCMs). Rotated Empirical Orthogonal Function (REOF) analysis indicated that the CORDEX ensemble captures better the spatio-temporal variability of both temperature and precipitation (first REOF mode), in particular depicting the warming and Sahel precipitation recovery in recent decades over West Africa. On the other hand, the spatial patterns and associated time series of the last two REOF modes in CORDEX mostly follow the CMIP5_SUBSET pointing towards a strong role of the boundary forcing in the RCM simulation of precipitation variability. For the future climate 2070-2099 relative to 1976-2005, a Bayesian model was applied to the three sets of models (CMIP5, CMIP5_Subset and CORDEX) and PDFs of Temperature and precipitation change for two sub region (Sahel and Guinea Coast) were derived. For temperature change over the Guinean Coast, CMIP5_S models under RCP8.5 has a lot of uncertainties showing more bias and less agreement among models but the vi CORDEX seems to reduce those uncertainties. Over the Sahel, only CORDEX under RCP4.5 scenario shows more agreement and less bias. CMIP5 and CMIP_S show multi modal PDF pointing out some uncertainties and less agreement among models. For precipitation change over the Guinean Coast under RCP8.5 and RCP4.5 uncertainties still remain in CORDEX model with an increasing precipitation trend for the late century. There is no significant difference on precipitation change between RCP4.5 and RCP8.5. CORDEX has a wide PDF curve under RCP4.5 and RCP8.5 scenario showing the persistence of uncertainties. Two sources of uncertainty in climate projection from CMIP5, CMIP5_Subset and CORDEX were also examined for temperature and precipitation. An ordinary least square was used to fit each decadal anomalies prediction of CMIP5, CMIP5_Subset and CORDEX with a fourth-order polynomial over the years of 2006-2099 for the two scenarios RCP45 and RCP85. The anomalies were computed with the reference period of 1976-2005. The new generation of models had an added value compare to the driving GCMs (CMIP_S) and CMIP5 MMEs by reducing the Internal and Inter Model Variability over the West African region. Inter Model Variability was the dominant source of uncertainties and is explaining up to 90 % of total uncertainty. The study conclude that for temperature under the two scenarios, the change is robust (Signal to Noise ratio greater than one) over most of West African countries with more spatial details and improved signal to noise ratio with CORDEX MMEs compare to CMIP5 and CMIP5_S MMEs. Over West Africa, CORDEX under RCP4.5 has a signal to noise ratio greater than one with an increasing trend of precipitation while the noise dominates the signal under RCP8.5, in CMIP5, CMIP5_S and CORDEX. An assessment of climate change information over West African region needs to rely on the careful evaluation and compounded information deriving from multiple sources.