A Mechanistic-Stochastic Approach to Classify Water Consumers and Simulate Urban Water Demand

Stochasticity in urban water demand arises due to the unpredictability and randomness of consumer behavior, which is influenced by population growth, climatic conditions, and conservation programs. Most urban water demand estimation methodologies are based on end-use models or stochastic models. End-use models describe the uses of water by households at the appliance level and require extensive and detailed data about water activities and water appliances. Stochastic models, however, predict water use using empirical relationships based on predictors, such as population size and water pricing. Integration of mechanistic end-use modeling with stochastic modeling can aid in better understanding of consumers' water use behavior and, therefore, can aid in better estimation of water availability in planning and management of urban water resources. A novel mechanistic-stochastic water demand model is developed here through the integration of an end-use model and a stochastic model. The model is developed using residential customer billing records from two water utilities. The historical water billing records are fitted to a gamma distribution based on the Akaike Information Criterion (AIC) values compared to exponential, extreme value, and log-normal distribution. Consumers are categorized into different groups from the distribution of water billing records and aggregated demand is estimated for the water system. To validate the modeled customer categories, housing survey data is collected and analyzed. Integration of mechanistic and stochastic modeling along with linkage of multiple data sources through this methodology can provide a powerful tool for efficient and sustainable water resources management.