The distributed model intercomparison project (DMIP): motivation and experiment design

The distributed model intercomparison project (DMIP) was formulated as a broad comparison of many distributed models amongst themselves and to a lumped model used for operational river forecasting in the US. DMIP was intended to provide guidance on research and implementation directions for the US National Weather Service as well as to address unresolved questions on the variability of rainfall and its effect on basin response. Twelve groups participated, including groups from Canada, China, Denmark, New Zealand, and the US. Numerous data sets including seven years of concurrent radar-rainfall and streamflow data were provided to participants through web access. Detailed modeling instructions specified calibration and verification periods and modeling points. Participating models were run in ‘simulation’ mode without a forecast component. DMIP proved to be a successful endeavour, providing the hydrologic research and forecasting communities with a wealth of results. This paper presents the background and motivations for DMIP and describes the major project elements.