Study on water level prediction using ARMA and ARIMA models: a case study of the river Benue

Flooding and its impacts are a persistent global problem, especially in developing countries. In Nigeria, for instance, studies on how to control the devastating impact of this extreme event have concentrated on remote sensing and geospatial techniques. Recently studies have taken a different direction towards tackling the menace caused by the hazards posed by floods globally, thereby proposing water level forecasting for effective water resources management and so prevention or mitigation of flood disasters. In this work, as well as descriptive analysis, we implement a novel application of time series analysis, to predict monthly water level discharges in Nigeria between 2011-2016, using data collected from Nigeria Hydrological Services Agency for three different water stations along the river Benue, including Ibi, Makurdi and Umaisha water stations, using Autoregressive Moving Average (ARMA) and Autoregressive Integrated Moving Average (ARIMA) models. The performance of these time series models was tested using three different performance measures, including mean absolute error, root mean square error and Nash-Sutcliffe efficiency to find an appropriate model which will be used to forecast the water level discharges from the three water stations. Such predictions will provide information to the populace of what to expect in future to mitigate the impact of flood disasters when they occur.