This study is to perform real-time probabilistic flood stage forecasting. The proposed method consists of a deterministic stage forecast derived from the support vector machines, and a probability distribution of forecast error based on the fuzzy inference model. The probabilistic flood stage forecasts can then be obtained by combining the deterministic stage forecasts with the error probability distributions. The proposed approach is applied to the Lang-Yang River in Taiwan pertaining to validation events of six flash floods. The probability distributions of stage forecasts one to six hours ahead are made, and the predictive uncertainty information is presented and discussed in various aspects. Forecasting results examined by forecast hydrographs with a 95% confidence interval, and the percentages of data included in the confidence region, indicate the effectiveness of proposed methodology.