Evaporation modelling of lake Turkana using artificial neural networks (ANNs) based on climate projections

Climate change remains to be one of the major threats to the water sector other than other human factors such as polution and uncontrolled utilization of water resources in sectors such as urbanization and agriculture. The Lake Turkana basin is one of the areas that is under climate change threat adding on to the present severe climatic conditions. Determining the future of the major water source in this region with climate change being a reality is of great importance. This study aimed at estimating the amount of evaporation with Artificial Neural Networks (ANNs) model, to improve the performance of the model and to make domestic and livestock projections. In addition, it is aimed to determine the effects of climate change on lake evaporation and to develop possible adaptation measures by using future forecasts including meteorological data compiled from CORDEX regional climate models in the model. The ANNs model with 6 hidden neurons and a training, validation and testing rates of 80%, 10% and 10% respectively had the best performance in estimating evaporation. The ANNs model with four parameters was able to perform with the highest correlation coefficient of 0.99, least mean square error of 8.13967E-5 and MAPE of 6%. The future average annual lake evaporation was projected to be 23.76, 25.68, 18.89 and 18.57 billion m3 for HADGEM and MPI 4.5 and 8.5 respectively. ANNs was found to be an efficient tool in future forecasts of climate based projections and in water management related decision making.