Improving Seasonal Rainfall Prediction over West Africa using Dynamic Climate Models

Abstract

Reliable prediction of seasonal rainfall is crucial for decision-making in various socio-economic sectors in West Africa, but obtaining reliable forecasts poses a big challenge to weather forecasters across the region, because their seasonal forecasts are mostly based on empirical models. While several recent studies are suggesting that the use of dynamic climates models may be a solution to the challenge, there is a dearth of information on how well these models simulate parameters like rainfall onset date (ROD), rainfall cessation date (RCD) and length of rainy season (LRS) over West Africa. The present study evaluates the performance of both global and regional climate models (GCMs and RCMs) in simulating these parameters over the study domain in the past and present climate. These datasets are from the China Meteorological Administration (CMA) Sub-seasonal to Seasonal (S2S) prediction, the UK Met Office Unified Model (MetUM), and 8 of RCMs that participated in the Coordinated Regional Climate Downscaling Experiment (CORDEX). The study also examines how a further modification of the Betts-Miller Janjic (BMJ) convective scheme in the Weather Research and Forecasting (WRF) model can improve the prediction of seasonal rainfall over West Africa. This study further assesses the potential impacts of 1.5°C and 2°C global warming levels (GWL15 and GWL20) on ROD, RCD and LRS in West Africa. Using common definitions within the sub-region, the simulated RODs, RCDs and LRS are compared with observation from satellite datasets, and the models’ capability to reproduce the inter-annual variability of these parameters over the climatological zones in the sub-continent is statistically quantified. The impacts of GWL15 and GWL20 on each parameter were also quantified and compared. The outcomes of the study show that all the models have some biases in their simulations although they do produce convincing results. The CMA model realistically simulates the observed spatial pattern and the interannual variability of RODs in the study area, as well as the observed seasonal movement of the West African Monsoon (WAM) and its associated rainfall patterns. The MetUM also reproduces the latitudinal progression of the observed RODs, RCDs and LRS over West Africa suitably, but performs poorly in simulating their inter-annual variability, even though there is improvement in the simulations of the new versions. It was also found that the CORDEX RCM ensemble correctly replicated and captured the essential features in the observed RODs, RCDs and LRS in the historical climate, and the RCM spread also enclosed the observed values. Most of the selected convection schemes reliably simulate the observed spatial distribution of RODs, RCDs, and LRS in the study area but overestimated the average monthly rainfall over the entire West African region. A new version of the BMJ scheme outperforms the default scheme in the sub-continent. The study project the western and eastern Sahel as hot-spots for a delayed ROD and reduced LRS in the 1.5°C and 2°C warmer climate under the Representative Concentration Pathway (RCP4.5 and RCP8.5) scenarios. The results of this study will be beneficial for agricultural and water resources planning decision-making and in reducing the impacts of global warming over West Africa.