Analysis of Stream Inflow and Peak Flow of Kainji Lake Using Stochastic Models

Abstract

The research worked on flood forecasting of Kainji Lake using stochastic models by making use of average monthly inflow data for 30 years from the period of 1990 to 2021 and average annual peak flows data for 21 years from 2000 to 2021 collected from Kainji Dam meteorological station. MINITAB and SPSS software were used for the analysis. The potential models selected for the analysis were ARIMA Models of order (2,1,2) and (2,1,0) for inflows and (1,1,1) and (1,1,0) for peak flows. The selection of these models was done by identifying their features using Auto and Partial Autocorrelation functions of having the lowest values of Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Sum of Squares (SS) and Mean Squares (MS) when compared to the other models. Furthermore, the analysis of the residuals for Auto and Partial Autocorrelation functions, normal probability and Histogram plots were obtained and used for the validation of the models, the results show ARIMA of order (2,12) and (1,1,1) for in-flow and peak flow were the best. Twelve and a half (12.5) and five (5) years of forecast data for the two cases were obtained. The forecast result showed that the months of August to October 2023 have high inflow values with September having the highest inflow with a value of 3471.33 (m3/sec). This highlighted the importance and usefulness of these models in warning the communities around the study area of likely impending flood events from the months of September to October and also the land around the study area can be used for agricultural purposes during the months of March to July due to low flows.