Discovering heterogeneous water demand responses under non-stationary hydroclimatic, social, and economic stressors. A 20-year analysis in Costa Mesa (California)

Non-stationary hydroclimatic, social, and economic stressors can potentially have temporary or permanent effects on water consumption behaviors at the individual, community, and global scale. The on-going COVID19 pandemic with prolonged Shelter in Place orders, for instance, has already transformed and is expected to further transform lifestyles and work patterns globally. Understanding how individuals change their water demand in response to evolving external conditions would provide us with better information on water demand flexibility, along with the possibility to evaluate the effects of demand management strategies (e.g., water use restrictions) and inform future operations and management of water infrastructure. Yet, existing behavioral studies on water consumption change are often limited in size (only a few households are considered), spatial scale (water demand is often aggregated at the district/city scale), or temporal scale (length of water demand time series). These limitations so far prevented a consistent comparison of the potentially heterogeneous responses of households with different socio-demographic background to different external stressors, along with a quantification of the duration of such changes.

In this work, we investigate how individual and community-scale water consumption behaviors changed for 8871 customers in the city of Costa Mesa, California (USA) from 2002 to 2020. Three types of stressors impacted the Costa Mesa area in the considered time span: the 2008-10 and 2012-16 California droughts, the 2009 economic recession, and the first COVID-19 lockdown in 2020. Our analysis is based on bi-monthly water billing data collected at the individual account level. We developed a data-driven behavioral analysis for customer segmentation that integrates the following sequential modules: (i) quantitative water consumption change assessment for individual accounts under each of the three stressors (i.e., droughts, economic recession, and COVID-19). We identify similar behaviors by means of state-of-the-art unsupervised clustering techniques (agglomerative hierarchical clustering); (ii) pattern analysis of water consumption changes. We analyze deviations from baseline water consumption patterns using regression models; and (iii) identification of relevant socio-economic determinants as potential determinants of water consumption behavior change. We explore different subsets of explanatory determinants by means of scenario discovery algorithms. This research contributes behavioral insights on urban water consumption under non-stationary hydroclimate and socio-economic scenarios. Such insights on human-water interactions in urban areas can be ultimately exploited by utilities and decision makers alike to design and implement optimized and tailored water demand management strategies targeting short-term resilience of urban water systems under rapidly changing water demand patterns, or longer-term behavioral changes.