A Spectral Analysis Based Methodology to Detect Climatological Influences on Daily Urban Water Demand

Abstract

Urban water demand (UWD) is highly dependent on interacting natural and socio-economic factors, and thus a wide range of data analysis and forecasting methods are required to fully understand the issue. This study applies, for the first time, the continuous wavelet transform to determine changes in the temporal pattern of UWD and its potential meteorological drivers for three major Canadian cities: Calgary, Montreal, and Ottawa. This analysis is complemented by Fourier and cross-spectral analysis to determine inter-relationships and the significance of the patterns detected. The results show that the annual (365 days) cycle provides the most consistent and significant relationship between UWD and meteorological drivers. Wavelet analysis shows that UWD is only sensitive to air temperature in the summer months when mean daily temperatures are greater than 10 to 12 °C. For the three cities studied, the UWD increases by between 10 ML (Montreal) and 50 ML (Calgary) per day with every 1 °C increase in air temperature. In an area with low precipitation (Calgary), there is an inverse relationship between UWD and precipitation during summer months. Wavelet transform and Fourier analysis also detected a 7-day cycle in UWD, particularly in the more industrialized city of Montreal, which is related to the working week. In general, applying the season dependent linear relationships between UWD and temperature is suggested as perhaps being more appropriate and potentially successful for forecasting, rather than continuous complex nonlinear algorithms that are designed to explain variability in the entire UWD record.