Modeling water quality and hydrological variables using ARIMA: a case study of Johor River, Malaysia

Abstract

Long-term trends in water quality and hydrological variables of natural systems reveal information about physical, chemical and biological changes and variations due to manmade and seasonal interventions. The objective of this study was to develop suitable stochastic models for predicting river water quality and hydrological variables through the establishment of dynamic relationship among the variables using transfer function modeling approaches. Autoregressive integrated moving average (ARIMA) model containing autoregressive (AR), integrated (I) and moving average (MA) was used for this purpose. The water quality variables, namely pH, color (TCU), turbidity (ppm), Al3+ (ppm), Fe2+ (ppm), NH4 + (ppm) and Mn2+ (ppm), and hydrological variables, namely rainfall and river discharge for Johor River, Malaysia, recorded for the period 2004–2007 were used in the study. Results showed that except Al3+, Fe2+, NH4 + and rainfall, all other variables are stationary. The non-stationary time series can be fitted with ARIMA (p, 1, q), while the stationary time series can be fitted with AR model with 1–5 time lags. The autocorrelations of all the samples were found within the 95% confidence bounds and the model residuals were found to follow normal probability distribution, which indicate the suitability of the models in forecasting water quality and hydrological variables. It is expected that the modeling approach developed in this paper can be replicated in other river basins for reliable prediction of river water quality due to changes in rainfall–runoff processes.