A generalized stochastic hydrometeorological model for flood and flash‐flood forecasting: 2. Case studies

The short‐ and long‐term performance of a stochastic‐dynamic hydrometeorological model in real‐time flood flow forecasting is examined with 6‐hourly data from a 2344‐km 2 basin in Oklahoma. The model couples a local precipitation model with a soil and a channel model through (1) the conservation of mass law and (2) an extended Kalman filter. The input consists of surface meteorological data interpolated from observations at nearby National Weather Service stations. Hydrograph peak time and magnitude errors are quantified in an effort to judge performance from an operational hydrologist's point of view. Least squares criteria are also utilized. The effect of the use of forecasts of the input variables on model performance, at various forecast lead times, is indicated. Good overall performance of the model is observed with obvious hydrologic performance deterioration when no precipitation component is used.