An evaluation of actual evapotranspiration in tropical East Africa

Monthly actual evapotranspiration (AET) for four humid catchments in Kenya, East Africa was evaluated using the Morton and Grindley models. The area of each catchment was less then 100 km(2) and all catchments lie around the equator. Three of the catchments are chiefly vegetated with pasture, annual and perennial crops, whereas one is largely under forest. The AET estimates from the aforementioned models were compared with those based on a water balance analysis. A total of 34 data years for daily rainfall and run off for all. the catchments were used for analysis. The results indicated that both models tended to overestimate AET in relation to the water balance-based values. The Grindley model (AET(G)) overestimated such that the estimates were either equal or close to the Penman potential evapotranspiration (PET) values in all the catchments. The Morton model (AET(M)) performed better, and AET estimates by this method, although marginally higher, were closer to the water balance-based estimates. The overall overprediction by the Grindley model on a monthly basis was of the order of 32% whereas by the Morton model it was only 8%. Although the mean values from the Morton model are only 8% higher than the water balance values on a monthly basis, values of the RMSE (root mean square error) range between 25 and 47 mm. The additional merit of the Morton model lies in its ability to provide estimates of AET based solely on meteorological data, which are readily available in Kenya, East Africa. (C) 1997 by John Wiley & Sons, Ltd.