An algorithm for real-time prediction of river stage dynamics using a Takagi-Sugeno fuzzy system is presented in this paper. The system is trained incrementally each time step and is used to predict one-step and multi-step ahead of river stages. The number of input variables that were considered in the analysis was determined using two statistical methods, i.e. autocorrelation and partial autocorrelation between the variables. Effectiveness of the identification technique was demonstrated by a simulation study on the river stage of the Cilalawi River in Indonesia. The numerical results of the Takagi-Sugeno fuzzy modeling method were compared with the results of a conventional linear regression model. Through inspection of the results it was found that the Takagi-Sugeno fuzzy approach was more accurate in predicting one-step and multi-step ahead of river stage dynamics than the conventional multiple linear regression approach. The Takagi-Sugeno fuzzy system was able to make a proper fuzzy rule from the training data set, which might be considered as one of the main drawbacks of the Takagi-Sugeno fuzzy system. Yet, more substantial improvement certainly should be pursued through further research to improve the forecast results at greater lead times.