Forecast of urban water consumption under the impact of climate change

Abstract

The proper planning of future urban water supply is essential to sustainable development. This paper answers two questions: What quantity of water will be needed in the long term? To what extent will water consumption be affected by climate change? We forecast water consumption using Bayesian statistics methods. A clustering analysis of observed daily water consumption and climate variables splits observations into base water use and seasonal water use, on the basis of the correlation between water consumption and air temperature. We show that the base water use is independent of climate change, but is subject to weekend effects. The seasonal water use depends on daily air temperature and total precipitation. Our forecast allows for uncertainties in climate variables and model parameters. The results from Bayesian linear regression give a probability distribution of daily water use. We obtained climate projections from multiple general circulation models and downscaled them for Greater Montreal. Bias corrections were made to the downscaled daily minimum temperature, maximum temperature and total precipitation. Using these corrected data as input to the Bayesian linear regression model, we forecast water consumption for the next three decades. The forecast results show a trend of increasing seasonal water use over time.