Abstract:
Numerical Weather Predictions (NWPs) are subject to systematic errors and biases. Hence, the continuous verification of NWP model outputs in order to contribute to model improvement became very important over recent years. Verification results provide numerical measures of how well NWP models perform, in an objective way. It also allows for monitoring of how NWPs improve over time. In the day-to-day operation of weather forecasting one might find biases in either forecasts generated by the NWP model, or biases that result from the weather forecaster’s interpretation of NWP output, or both. The use of verification statistics might help to identify the source of these biases, which might lead to research targeted to improve the scientific understanding of the underlying processes required to improve NWP forecasts. This study investigates the potential of the 20km x 20km resolution Limited Area Model over Africa (Africa LAM) developed by the United Kingdom Meteorological Office (UK Met Office) to be used as supplementary tool to improve weather forecast output to end-users over the Western Sahel (WS) and Nigeria. In the study, Africa LAM T+24h forecasts dataset was verified against daily observed rainfall, maximum and minimum temperature data, of 36 selected meteorological point stations over the WS from January 2005 to December 2006. 12 meteorological point stations were selected across each of the three identified climate zones of the WS, namely (1) Wet Equatorial (WE) climate zone (from the southern coastline up to a latitude of 8.00ºN), (2) Wet and Dry Tropics (WDT) climate zone (between latitude 8.00ºN and 12.00ºN) and (3) Semi-Arid (SA) climate zone (between latitude 12.00ºN and 15.00ºN). The dataset was also stratified into four seasons, namely (1) January-February-March (JFM), (2) April-May-June (AMJ), (3) July-August-September (JAS) and (4) October-November-December (OND). The verification algorithms and measures used in this study are in accordance with the WMO NWP verification standards. The verification results indicate that the Africa LAM model temperature forecasts show skill, more so during the raining seasons (AMJ and JAS) than during the dry seasons (JFM and OND) over the WS. The model rainfall forecasts, however, show more skill during the dry seasons (JFM and OND) than during the raining seasons (AMJ and JAS). The results further indicate that, on a regional basis, the model temperature forecasts show more spatial skill over the WE climate zone than over the WDT and SA climate zones of the WS, while rainfall forecasts show more skill over the SA climate zone than over the WDT and WE climate zones of the WS. Additional results from simple bias corrections and Model Output Statistics (MOS) which are some of the suggested post-processing techniques in this study are presented. These results give a better understanding of the model forecast errors, and also provide the feedback necessary for a possible improvement of Africa LAM forecasts by scientists at the UK Met Office. Copyright
