Abstract: The accuracy of flood forecasts obtained from spatially lumped hydrological models is highly affected by the uncertainty in the estimates of area-mean rainfall that depend on the error type and magnitude of rain gauges and their density and pattern of space distribution. The effect of these errors is more prominent in a large basin with a small number of rainfall gauges. To improve the accuracy of real-time flood forecasting, we herein describe an updating method for area-mean rainfall estimation using rainfall system response curves. This new method takes a hydrological model as a response system between input and output, and updates the area-mean rainfall using the rainfall system response curves and least square method to minimize the error between the observed and simulated flow series. Then, the stream flow is recalculated using the updated area-mean rainfall series. Application of the method is demonstrated in combination with the Xinanjiang model tests, via a synthetic case and 14 historical flood events that occurred in the Fuchunjiang basin in Zhejiang during 2010–2014. The results show that the method can significantly improve the accuracy of real-time flood forecasts without increasing model structure complexity or the number of model parameters.