Modelling and Forecasting of Dissolved Oxygen Using the ARIMA Model: The case of the Ledrat Dam (Medea Province-Algeria)

Abstract

Dissolved oxygen (DO) is a key indicator of water quality. This study focuses on the modelling of the temporal evolution of dissolved oxygen (DO) mg/L in the waters of the Ledrat dam, located in the wilaya of Medea, using the ARIMA model.

The descriptive analysis shows significant variability in the dissolved oxygen values during the period of 2012-2022 with seasonal fluctuations and outliers greater than 12 mg/L, as well as the decomposed time series shows a nonlinear trend.

The ARIMA model shows a good performance with an RMSE of 1.4191, an MAE of 1.0727 and a MAPE of 13.28%, which indicates adequate accuracy for the forecasts. The model validation using the L-Jung-Box test (Q*=22.448, p-value=0.2625) confirms the independence of the residues, attesting to the quality of the fitted model. The results obtained prove that the ARIMA model is effective in capturing dissolved oxygen (DO) levels and providing relevant forecasts that can be used as a decision support tool to support the optimized management of the dam's water resources and preserve water quality.