The study compared the performance of four di erent cumulus parametrization schemesrnin the ICTP's Regional Climate Model RegCM4 for the period 1989-2008 over thernHorn of Africa. The Horn of Africa has been delineated in to 12 homogeneous regionrnbased on clustering method that minimizes a sum of dissimilarities. Grell-FC schemernbest captured the mean climatology. Within the 12 clustered regions positive/negativernprecipitation biases are obtained in Emanuel/Grell-AS and Kuo convective schemesrnwhile intermediate values closer to zero are obtained in Grell-FC. Grell-FC schemernis then selected for customization over the Horn of Africa. To improve the simulationrnof rainfall using the selected Grell-FC scheme, several sensitivity experimentsrnare performed by adjusting the minimum and maximum precipitation e ciency, thernminimum and maximum shear e ect on precipitation e ciency, the minimum andrnmaximum convective heating and the available buoyant energy removal time scale.rnSetting the minimum and maximum shear e ect on precipitation e ciency to 0.10rnand 0.30 respectively improve the correlation, which implies a better simulation ofrnprecipitation. Setting the minimum and maximum convective heating to -150 andrn400 respectively minimize the positive bias over most of the regions. 17 years longrnperiod run revealed that using the new values of shear e ect on precipitation e ciencyrnimprove the correlation and representation of the inter-annual variability