EVALUATION OF THREE EMPIRICAL REFERENCE EVAPOTRANSPIRATION MODELS AT A TROPICAL STATION UNDER THREE SKY CONDITIONS USING TWO SOLAR RADIATION ESTIMATION METHODS

An existing solar radiation model developed at Ilorin and found to be more reliable than Angstrom-type and Hargreaves solar radiation equations was used in the FAO Penman-Monteith reference evapotranspiration model (FAOPM) to obtain daily reference crop evapotranspiration (ETo) for a 32-year (1970 to 2001) period. The number of days having all the required input meteorological data was 9335. The sky conditions of the days were classified as clear, partially cloudy or cloudy depending on the cloudiness index i.e. the ratio of diffuse solar radiation to total solar radiation. The ETo values obtained with FAOPM were compared with predictions of three simpler empirical ETo models namely the Hargreaves (HGRV), Jensen and Haise (JHSE) and Blaney-Morin-Nigeria (BMN) models. When the more reliable solar radiation model was used in HGRV and JHSE, their performances were better than when the solar radiation equation of Hargreaves was used. Generally the three simpler models overpredicted ETo. The bias, root mean square difference (RSMD) and absolute error of prediction deteriorated with sky cloudiness when the solar radiation equation of Hargreaves was used. Linear regression equations with zero intercepts were developed for the estimation of FAOPM predictions from those of the simpler ETo models. The regression equations relating the predictions of FAOPM to those of HGRV generally yielded the highest coefficients of determination and the lowest standard errors of regression. The predictions of HGRV were also the closest to the corresponding FAOPM predictions under the various sky conditions. Based on the outcome of the regression analysis and the ease of application of HGRV, the FAOPM-versus-HGRV regression equations were recommended for the estimation of FAOPM predictions of daily ETo when the use of FAOPM is necessary but not feasible because of incomplete input data.