Predictability of weather on extended NWP time scales over Kenya using the NCEP GFS model

Abstract

Numerical weather prediction (NWP) portends viable applicability as a tool III forecasting the evolution of atmospheric processes and the associated weather on extended time scales. Proper deciphering and interpretation of these products is critical if the NWP model outputs are to have usefulness in day-to-day socio - economic spheres, thus achieving the intended purpose of the modeling community. This research work sought to investigate the predictability of the National Centres for environmental prediction Global Forecasting System (NCEP GFS) model weather on extended NWP time scales of up to 10-days over Kenya in an attempt to handle the concern at hand. The accuracy and skill of the GFS model was assessed using Root Mean square Error (RMSE), correlation analyses, Equitable Threat Score (ETS), True Skill Score (TSS), Heike Skill Score (HSS), Probability of Detection (POD) and Frequency Bias Index (FBI). Results from the analyses of errors and probabilistic skill score showed that the GFS model was able to replicate spatial and temporal distribution of day to day temperature and rainfall over Kenya using one and two day lead times. The skill and accuracy of the GFS model performed poorly for three-day lead time although it showed some improvement thereafter but to a lesser extent than the first two lead times. Results from the study indicated that the model was able to replicate the temporal and spatial variability for both 10-day total rainfall and 10-day average temperature. This was evident from the low RMSE errors, higher correlation coefficients and significant scores. In conclusion, the NCEP GFS model, from the results was found to have significant skill in predicting rainfall and temperature on extended NWP timescales of up to 10-days over Kenya. The model could therefore be used for predicting the temporal and spatial distribution of both 10-day total rainfall and 10-day average temperature over Kenya and can provide effective early warning tools for application in many climate sensitive sectors.