Estimating potential evapotranspiration of a data scarce region: a case of lake victoria basin of Kenya

This paper examines potential evaporation over the Lake Victoria basin, a region with very few operational meteorological stations hence scarce data. Data stations within the Lake basin provided rainfall and temperature data covering periods of at least twenty years. Five methods of computing Potential Evapotranspiration (PET) (Penman, Hargreaves, Thornwaite, Blaney-Criddle and Pan Evaporation) were studied for their applicability under the same catchment. The homogeneity of rainfall was done using mass curve analysis and the filling of missing data using Markov Model. The averaging of historical data was used to interpolate the missing data. The stated empirical equations were applied to all stations over the 20 year period. C++ Program was developed and used to generate respective PET values using the stated empirical models. Based on the range of difference and average values of both the total absolute differences and the standard deviation, and using the Penman method as the basis of comparison, the Blaney-Criddle method predicted the monthly PET values better and Thornwaite was rated the poorest predictor. This showed that the Blaney-Cridle method could be applied in place of Penman as it predicted closer to Penman where the latter could not work.