Developing an Agent-Based Model of Dual-Flush Toilet Adoption

Abstract

The spread of individual water conservation behaviors within a population can have large impacts on overall water demand. Agent-based models (ABMs) represent individual actors that update their behaviors over time in response to their environment and other agents, and ABMs have been applied to model the adoption of water conservation behaviors and technology. Existing ABM approaches are calibrated based on cumulative water demand data and use assumptions about household-level adoption behaviors. This research develops an ABM of water appliance (dual-flush toilets) adoption and introduces a new approach to calibrate the ABM while allowing for stochasticity and heterogeneity in agent parameters and adoption decisions. The calibration approach uses a noisy genetic algorithm (NGA), and the ABM is calibrated to match household survey data that was collected in Jaipur, India, in 2015. The NGA is applied multiple times to explore variability in the search, and five solutions were found with similar error values. The best-performing solution is applied to project adoption over a 100-year period for varying climate scenarios, and results show quicker adoption rates for dry climates. Sensitivity analysis was conducted for a parameter that represents a delay in adopting dual-flush toilets and a parameter that represents the importance of drought in making adoption decisions. The model presented in this research can be used to aid water resource planning and to anticipate potential impacts of water conservation policies, such as rebate programs or media campaigns.