Evaluation of Selected Numerical Weather Prediction Models for a Case of Widespread Rainfall over Central and Southern Nigeria

Abstract

Precipitation forecasts from four Numerical Weather Prediction (NWP) models are evaluated for a case of widespread rainfall event over Central and Southern Nigeria on the 21st of March 2015. The four models evaluated are the European Centre for Medium Range Weather Forecast (ECMWF) with a resolution of 25 km, The UKMET model 20 km, NCEP Global Forecast System (GFS) 50 km and the Weather Research and Forecast Model (WRF) with 10 km resolution. Precipitation forecasts are compared with observed precipitation at station and gridded observation points for different rainfall amount thresholds using the Method of Objective-based Diagnostic Evaluation (MODE), Grid statistics and Point Statistics. The global models ECMWF, UKMET and GFS underestimated the rainfall amount when compared to the WRF regional model. The global models recorded a critical success index (CSI) of less than 10% while the regional WRF model recorded a critical success index of 40% for rainfall amount greater than 25 mm. MODE analysis showed that the WRF model also recorded a 93% relationship between observed and forecast precipitation events of 21st March 2015 over Nigeria when compared with the ECMWF, UKMET and GFS models which showed 88%, 88% and 87% relationship respectively. Our findings suggest that dynamically downscaling a global model using the WRF model added value and gave a better skill of precipitation forecast for the event under study.