The continuous decline in lake water levels is not only a major concern but also a daunting challenge to policymakers, demanding a backup technological and policy interventions in context of broader political and socio-economic realities. This study used Lake Volta hydrological system to shed light on the extensive and flexible modelling and simulation capabilities of stochastic models to understand the bigger picture of water level (WL) dynamics. The study used Autocorrelation Regressive Integrated Moving Average (ARIMA) and Kalman Filtering (KF) techniques as the proposed optimal stochastic models for the study area. The first order ARIMA (0, 1, 1) was found suitable for predicting the future monthly Lake Volta WL in the presented study based on expert advice and recommendations from existing studies. The statistical performance indicators used were minimum residual error (rmin), maximum residual error (rmax), arithmetic mean error (AME), arithmetic mean squared error (AMSE), arithmetic mean absolute percentage deviation (AMAPD), and arithmetic standard deviation (ASD). Based on the results achieved in this study, ARIMA (0, 1, 1) achieved AME, AMSE, AMAPD and ASD of -0.1268 m, 0.0037 m, 0.5749 m, and 0.0033 m respectively. Ensemble of ARIMA and KF was further used to forecast the upcoming monthly WL trends up to December 2048. ARIMA (0, 1, 1) model is found suitable for forecasting Lake Volta WL which shows positive trend up to December 2048. The study further predicted that Lake Volta WL will increase from the current average level of 0.2272 m to an average of 9.1366 m for the next 28 years. The ensuing conclusions stressed the need for checks against over-release of WL for hydropower production and measures for sustainable land and water management in the entire basin. This study can potentially enhance our understanding of hydrodynamic processes in Lake Volta and support water resource management.
Lake Volta, Kalman Filter, Stochastic Models, Time Series Analysis, Water Levels