One billion people worldwide experience intermittent water supply (IWS), in which piped water is delivered for limited durations. Households with IWS must invest in water storage infrastructure and often rely on multiple sources of water; therefore, these household-level purchasing and infrastructure decisions is a critical component of water access. Informed by interviews with IWS households, we use radial basis function networks, a type of artificial neural network, to determine optimal household water management decisions that maximize reliability of water supply while minimizing costs for a representative household in Mexico City that uses municipal piped water, trucked water, and rainwater. We find that securing reliable water supply for IWS households is greatly assisted by installation of household storage tanks of at least 2500 L. In the case of IWS households with limited storage options, the overall cost for water supply is reduced by scheduling water deliveries on nonconsecutive days. Rainwater harvesting systems were shown to be economically viable for households with limited water supply. This study demonstrates the importance of considering the management of multiple sources and household storage infrastructure when evaluating water investments in cities with IWS.
Keywords: Mexico City; coping costs; intermittent water supply; multi-objective optimization; multiple water sources; radial basis function networks; rainwater harvesting; water storage.