During the past decade hydrologists have become increasingly aware of theproblems of fluvial flood prediction during periods of intense convection,particularly in urbanised catchments whose rainfall-runoff responses tendto be rapid. Existing approaches to deterministic, short-range rainfallprediction are often deficient in their treatment of convectiveprecipitation because they cannot resolve individual convective clouds oreffectively model their evolution. In 1994 the UK Met. Office establisheda joint R&D programme with the Environment Agency (responsible forflood prediction in England and Wales) to explore the benefits of anObject-Oriented conceptual Model (OOM) of convection in the nowcastingof fluvial floods. This involved the development of an automated nowcastingsystem (GANDOLF) designed to run the OOM during episodes of air massconvection. This paper describes the structure and function of the GANDOLFsystem and compares the performance of the OOM with that of two otherprecipitation models routinely used by Thames Region of the Agency.