Longitudinal dispersion coefficient is a key parameter in determining the distribution of pollution concentration; especially in temporally time varying source cases after that full cross sectional mixing has occurred. Several studies have been carried out to present simple formulas for its prediction. However, they may not always result in accurate prediction due to the complexity of the phenomena. In this study, M5´ model tree was used to develop a new model for prediction of the longitudinal dispersion coefficient. The main advantages of the model trees are that they (a) provide transparent formulas and offer more insight into the obtained formulas and (b) are more convenient to develop and employ compared to other soft computing methods. To develop the model tree, extensive field data sets consisting of hydraulic and geometrical characteristics of different rivers were used. The performance of the model was also compared with those of other existing equations using error measures. Overall, results showed that the developed model outperforms the existing formulas and can serve as a valuable tool for prediction of the longitudinal dispersion coefficient.