Franklin J. Opijah*, Joseph N. Mutemi* and Laban A. Ogallo**
*University of Nairobi, Department of Meteorology
**IGAD Climate Prediction and Application Centre
Franklin J. Opijah
University of Nairobi, Department of Meteorology
(Received 14 May 2016, received in revised form 30 May 2017, Accepted 30 May 2017)
Seasonal climate prediction over Kenya poses a considerable challenge to the modeling community due to the intricate interactions among the atmospheric, oceanic and land surface processes. This paper assesses the performance of the Regional Spectral Model (RSM) in downscaling the European Centre-Hamburg (ECHam) global model outputs from 1970 to 1999 over Kenya with respect to rainfall and temperature prediction using standard verification techniques. The results show that the accuracy of simulating the annual cycle and spatial distribution of convection and precipitation over the country is still poor. The seasonal rainfall predictability over Kenya by the RSM is better during the October-December season (correlation coefficient [r] of 26%; proportion correct [PC] of 60%; Frequency Bias Index [FBI] of 111%) than in the March-May season (r of 8%; PC of 54%; FBI of 83%), but the prediction for temperature is better in the March-May season (r of 25%; PC of 53%; FBI of 124%) than the OND season (r of -11%; PC of 46%; FBI of 100%). The predictability for rainfall during the cool-dry June-August period is still low (r of -4%; PC of 49%; FBI of 52%) but that for temperature has better skill as compared to the March-May and October-December seasons (r of 49%; PC of 70%; FBI of 90%). There is need to improve the development of convective processes that govern tropical precipitating systems in the region through sensitivity analysis of cloud simulation modules in the RSM applied as well as address rare systems that episodically influence the weather over the country and the region.
Key words: skill score, root mean square error, correlation, RSM, ECHam, seasonal prediction, rainfall, temperature, Kenya