A methodology for improved hydrologic forecasting of streamflows in headwater basins has been developed. The methodology is based on the mathematical modeling of the hydrologic catchment processes via conceptual hydrologic models. Real-time updating procedures have been designed for the on-line incorporation of past observed discharges in the forecasts. Updating is accomplished via a state estimator that exploits a priori knowledge on the statistical characteristics of model input and parameter errors. Hydrologic expertise is the only essential requirement for routine application. The value of the methodology in real-time hydrologic forecasting is illustrated for two hydrologically different headwater basins: the Bird Creek basin in Oklahoma, U.S.A., and the Orgeval basin in France. Verification of the methodology with data from the aforementioned basins was performed as part of a World Meteorological Organization (WMO) workshop held in Vancouver, British Columbia, Canada, during the period: 30 July-8 August, 1987. Real-time conditions were simulated with verification data made known after forecasts were issued. The verification results appear encouraging especially in light of the fact that all the forecasts were obtained in an automatic fashion (without manual intervention). The methodology developed has been implemented on the National Weather Service River Forecast System forecast component at the Hydrologic Research Laboratory, National Weather Service (NWS), and is currently undergoing testing in a true real-time environment. © 1990.