Integration of multi-objective spatial optimization and data-driven interpretation to direct the city-wide sustainable promotion of building-based decentralized water technologies

Abstract

Decentralized water technologies such as rainwater harvesting (RWH) and greywater recycling (GWR) can supplement centralized urban water systems, helping reduce water withdrawal and improve water reliability. These benefits only emerge when decentralized water technologies are widely implemented. Several decision-supporting frameworks have been developed to identify suitable locations for deploying decentralized water technologies in a city. Yet, the support remains inadequate regarding: (1) the evaluation of the trade-off between environmental benefits and economic costs in selecting locations, and (2) the interpretation of the transition of optimal selections from low to high investment to assist in the promotion. This study presents an integrated analytic framework that combines multi-objective optimization and data-driven interpretation to direct the city-wide sustainable promotion of building-based decentralized water technologies. We select single-family houses in the city of Boston and apply the framework to study the promotion of building-based RWH and GWR. The framework starts with multi-objective spatial optimization to identify the non-dominant optimal selections (i.e., Pareto-front) of houses and technologies at the trade-off between maximizing energy savings and minimizing financial investment. Then, we evaluate the impact of the initial selection setting and the community-based maximum water saving constraint on the Pareto-optimal front. The spatial optimization shows that RWH is much more applicable than GWR for single-family house communities in Boston. When interpreting the Pareto-front, two clusters of census blocks stand out based on the change in the percentages of houses selected to invest RWH and GWR in each census block along with different investment levels. One cluster demonstrates its priority of being first selected to deploy RWH. Using Random Forest, critical features explain why one cluster should be selected first for promotion, including the larger demand for non-potable water use, longer distance from the centralized facilities, and larger rooftop for collecting rainwater. Finally, we discuss possible future improvements of the proposed spatial optimization and interpretation framework. Overall, our study can be useful to promote decentralized water technologies in cities.

Keywords: Data-driven interpretation; Greywater recycling; Multi-objective spatial optimization; Rainwater harvesting; Sustainable promotion.