A new multiple regression model for predictions of urban water use

Abstract

Shortages of freshwater have become a serious issue in many regions around the world, partly due to rapid urbanisation and climate change. Sustainable city development should consider minimising water use by people living in cities and urban areas. The purpose of this paper is to improve our understanding of water-use behaviour and to reliably predict water use. We collected appropriate data of daily water use, meteorological parameters, and socioeconomic factors for the City of Brossard in Quebec, Canada, and analysed these data using multiple regression techniques. The techniques represent a new approach to predictions of daily water use; its base use component is predicted using a function of socioeconomic factors, as opposed to a function of time as in existing approaches. The quality of the new approach is quantitatively demonstrated. Time series of predicted daily water-use captures observed characteristics very well, and improves the results of the weighted coefficient of determination, the relative index of agreement and the root mean square error from the existing approaches. Water use in the city exhibits a downward trend possibly due to an increasing annual charge for water use. Water use increases due to weekend effect. It decreases in the occurrence of rainfall; the decrease is more sensitive to previous-day than current-day rainfall. The analysis procedures reported in this paper can be applied to analyse water use in any other cities. The new approach would be a useful tool for decision makers to manage water use by adjusting water consumption policies and price.