How to Model an Intermittent Water Supply: Comparing Modeling Choices and Their Impact on Inequality

Intermittent water supply (IWS) networks have distinct and complicated hydraulics. During periods without water supply, IWS networks drain, and consumers rely on stored water; when supply resumes, pipes and consumer storage are refilled. Draining, storage, and filling are not easily represented in standard modeling software. We reviewed 30 ways modelers have represented the hydraulics of IWS in open-source modeling tools and synthesized them into eight distinct methods for quantitative comparison. When selecting methods, modelers face two critical choices: (1) whether to ignore the filling phase, and (2) how to represent consumers as attempting to withdraw their demand: as fast as possible (unrestricted), as fast as possible until a desired volume is received (volume-restricted), or just fast enough to receive a desired volume by the end of supply (flow-restricted). We quantify these choices' impact on consumer demand satisfaction (volume received/volume desired) and inequality using three test networks under two supply durations, implemented in two different hydraulic solvers (EPANET and EPA-SWMM). Predicted inequality and demand satisfaction were substantially affected by the choice to represent consumer withdrawals as unrestricted, volume-restricted, or flow-restricted, but not by the specific implementation (e.g., three different flow-restricted methods agreed within 0.01%). Volume-restricted methods predict wider inequalities than flow-restricted methods and unrestricted methods predict excessive withdrawal. Modeling filling delayed water provision unequally, reducing the volume received by some consumers (by similar to 20%), especially where water supply is brief. All else being equal, we recommend using volume-restricted methods, especially when modeling system improvements, and including the filling process when studying inequalities. Understanding how water distribution networks perform is complicated, particularly when the network is operated intermittently. Pipes in intermittent networks periodically fill up, supply water, and then drain, forcing consumers to adapt, e.g., by storing water. Intermittent networks represent about 20% of the world's water pipes, but there are not yet standardized, accessible methods for practitioners and utilities to model the key features of intermittent networks by using or adapting off-the-shelf hydraulic software. Previous methods of adapting hydraulic software for IWS networks disagree on two key choices: how consumers behave and whether to ignore the initial filling of pipes. We quantitatively compare how these varied choices affect the resultant hydraulic predictions. First, we find that when consumers are assumed to fill their storage as fast as they can (volume-restricted), water delivery is less equal than if consumers are assumed to withdraw slower (flow-restricted). Second, we found that including pipe filling delays water supply and magnifies predicted inequalities. Accordingly, we recommend that modelers carefully select methods that reflect how consumers behave in their context. We provide accessibly packaged Python code to enable other hydraulic modelers to efficiently compare, extend, and/or adopt the compared intermittent modeling methods.