MONTHLY STREAMFLOW PREDICTION WITH LIMITED HYDRO-CLIMATIC VARIABLES IN THE UPPER MKOMAZI RIVER, SOUTH AFRICA USING GENETIC PROGRAMMING

Streamfiow prediction remains crucial to decision-making especially when it concerns planning and management of water resources systems. The prediction of streamfiow however, comes with various complexities arising from non-linear and dynamic nature of the climatological and hydrological factors. Several modelling studies relating to streamfiow prediction have been carried out using different approaches. However, considering the non-linear and dynamic behaviour of hydro-climatological processes, a significant amount of historical data is required in all these approaches in order to achieve accurate and reliable results. Genetic Programming (GP), a computational intelligence approach based on evolutionary algorithm was employed in this study to predict the response of streamfiow to hydro-climatic variables in the upper Mkomazi River in South Africa using limited amount of datasets. Historical records for a period of nineteen years (1994-2012) were used for the construction and selection of input variables into the GP vector space. Individual monthly models were employed for streamfiow prediction for each month of the year. The performances of the models were evaluated using three statistical measures of accuracy. Results obtained indicate a close agreement and highly positive correlation between observed and predicted values of streamfiow during the training and validation phases for all the twelve models developed. These results further confirm the efficacy of the GP approach in monthly streamfiow prediction despite the use of limited amount of datasets.