How important is the selection of computational analysis method to the accuracy of rainwater tank behaviour modelling?

As the concept of sustainable urban water management is incorporated into the practice of urban water resource managers, actions, such as the utilization of roof runoff via rainwater tanks, which have multiple benefits, are increasingly being built into urban water systems. Modelling tools are frequently used to predict the yield from rainwater tanks and to estimate the storage capacity required to achieve a given potable supply reduction level, with these estimates used in both urban water resources policy development and engineering design. Therefore, it is important that the accuracy of commonly used models is understood. This paper investigates the impact of computational time step, computational operating rule, initial storage level, and the length of simulation period on the accuracy of the storage–yield–reliability relationship calculated using a simple rainwater tank behaviour model. Four time steps (ranging from 6 min to 24 h), two operational rules (supply before spillage and supply after spillage), two initial storage level states (empty and full), and three simulation periods (50 years, 10 years and 1 year) were applied to a wide range of rainwater tank system configurations and three different locations in Australia. It was found that the supply‐after‐spillage computational operating rule is preferable, while the ratio of the average demand volume in a single computational time step divided by the storage capacity (ΔD/S) can be used to assess whether a given combination of demand, storage, inflow, and computational time step will provide long‐term yield estimates that are within ± 5% of the values produced by a simulation that used a 50‐year time series of climate, 6‐min time step, and a supply‐after‐spillage operational rule (50‐6‐YAS). Copyright © 2007 John Wiley & Sons, Ltd.