Assessment of Stream Hydrological Response Using Artificial Neural Network: A Case Study of River Kaduna, Nigeria

Abstract

Hydrological alterations may result either from changes in average condition or from changes in the distribution and timing of extreme events. In view of this, the study attempted an evaluation of the hydrological response of River Kaduna at Shiroro Dam site, Nigeria to hypothetical climate change scenarios using the Artificial Neural Network (ANN) paradigm. For the deployment of the ANN, monthly historic hydrometeorological data (i.e., evaporation, rainfall, streamflow and temperature) spanning 33 years were obtained. To this end, four climate change scenarios: +10% rainfall, 2×coefficient of variation in rainfall, -10% rainfall and +30C average temperature were considered. The historical data were used as input to the ANN and selected monthly synthetic streamflow hydrographs in the seasons (i.e., dry and wet) were generated with an average high value of the goodness-of-fit (R2=0.96). The response pattern indicated a variability index for the River to be in the range of 0.85-1.25 while for the recession pattern it is 0.75-0.81. It is imperative to note that the ANN enhanced the generalization of the flow dynamics of the extreme events (peak and low flow regime) with relative predictability capacity values of 103% ( R_max) and 96.35% (R_min), respectively. However, considering the fact that the upgraded temperature and coefficient of variation in rainfall might impact negatively on the average runoff, flow variability, flood frequency and predictability, there is the need for the use of an extensive hydrometeorological data base coupled with the application of associated risk value for effective flood forecasting in real-time.