Statistical modeling of daily urban water consumption in Hong Kong: Trend, changing patterns, and forecast

Abstract

This study attempted to address statistical properties and forecast of daily urban water consumption in Hong Kong from 1990 to 2007. A statistical model was formulated to differentiate the effects of five factors on water use, i.e., trend, seasonality, climatic regression, calendar effect, and autoregression. The postulate of the statistical model is that total water use is made up of base, seasonal, and calendrical water use. Daily urban water consumption in Hong Kong from 1990 to 2001 was modeled and the developed statistical model explains 83% of the variance with six factors: trend (8%), seasonality (27%), climatic regression (2%), day-of-the-week effect (17%), holiday effect (17%), and autoregression (12%). The model was further validated using an independent data set from 2002 to 2007, yielding a R2 of 76%. The results indicated good performance of the developed statistical model in depicting the temporal variations of the urban water use in Hong Kong, offering an improved insight into urban utilization of water resources and acting as the theoretical support for effective urban water resource management in Hong Kong under the changing environment. Copyright 2010 by the American Geophysical Union.