A behavioural approach for household outdoor water use modelling

Abstract

Reliable predictions of household outdoor water use are important inputs for effective design and management of urban water systems. This paper evaluates and enhances the behavioural approach (BA) for modelling outdoor water use. The underlying premise in the BA is that outdoor water use is governed by people's probabilistic behavioural response to recent weather conditions (rainfall and temperature).The BA models used in this paper were evaluated using a 12 year dataset of monthly outdoor water use for 135 homes in the Newcastle region of New South Wales, Australia. The BA model of Coombes et al (2000) was found to found to outperform traditional linear regression techniques, after calibration using a new simulated likelihood calibration approach. However, it was found to be over-parameterised and underestimated observed variability by 22%. An enhanced BA model was more parsimonious and better simulated the observed variability (only 9% underestimation). Conditioning behavioural response on daily rainfall and maximum temperature did not provide good model performance. Rather, the major drivers of household outdoor water use variability were found to be long dry periods (for 80% of homes), while a smaller number (20%) additionally responded to the long hot periods (characterised by the degree day concept).