A COMPARISON OF DIFFERENT CUMULUS PARAMETERIZATION SCHEMES IN WEATHER RESEARCH AND FORECASTING (WRF) MODEL OVER WEST AFRICA

Abstract

This research examined the performance of four different cumulus parameterization schemes (CPS) in the Weather Research and Forecasting (WRF) model for simulating three heavy rainfall events. The first over Ouagadougou in Burkina Faso on 1 September 2009 (Case study 1; C1), the second over Dakar in Senegal on 26 August 2012 (Case study 2; C2) and third over central region of Benin on 17 June 2005 (Case study 3; C3). The modeled rainfall was compared with 3-hourly satellite observation and objectively evaluated using the Root Mean Square Error (RMSE) and Coefficient of Efficiency (COE) verification techniques. The techniques are based on how close the model is able to quantitatively reproduce the observed precipitation feature of the selected events. All simulation of the three case studies was run for 72 hours starting at 0000 GMT 31 August 2009, 25 August 2012 and 16 June 2005 over the Burkina Faso, Senegal and Benin Republic domains, respectively. The four different CPSs used are the Kain-Fritsch (KF), Betts- Miller-Janjic (BMJ), Grell-3Dimension ensemble (GR) and Grell Devenyi Ensemble (GDE) scheme. Generally, all four schemes produced negative biases in terms of coverage in the first two case studies (C1 and C2), except in the third case study (C3) where the KF scheme produced a small positive bias. While the GR scheme shows some success in C1 and C2, it produced high location bias in C3 leading to its less COE. Also, KF scheme has better COE in C3 due to its ability to reproduce similar feature as observed but had some location bias in C1 and C2 resulting to its less efficiency in these case studies. The performance of the GDE scheme appears to be fair with COE of 0.36 and 0.31in C1 and C3 also it has the least COE of 0.22 in C2 where the correlation coefficient (0.07) was very low whereas the BMJ scheme performed poorly in all the three case studies with the least COE of 0.31 and 0.26 in C1 and C3, however GR schemes simulated the observed drastic increase of rainfall at 0900 GMT 1 September 2009 over Ouagadougou in C1 while the KF scheme was able to reproduce similar feature as observed in C3 at 2100 GMT 17 June and 0000 GMT 18 June 2005 over the central parts of Benin Republic. Although, all the four schemes were unable to reproduce the observed drastic rain rate at 0600 GMT 26 August 2012 over Dakar in C2, but the better performance of KF scheme in C3 and that of the GR scheme in C1 and C2 suggests that suitability of CPSs may be case dependent also the poor performance of the profile adjustment BMJ scheme suggests that mass flux KF, GR and GDE schemes are better in representing heavy rainfall events over the selected regions. However, based on the subjective verification and objective evaluation, it can be concluded that the performance of the four schemes is highly case-dependent and the performance of the better GR and KF schemes cannot be generalized for all the three case studies.