The scour around abutments is a major damage of bridge which appears during the flood hazard. Accurate prediction of scour depth at abutment is very essential to estimate foundation level for a cost-effective design. The accuracy of conventional method is low for prediction of temporal scour depth. However, in this study, two robust techniques, adaptive neuro-fuzzy inference system (ANFIS) and artificial neural networks (ANNs), were employed to estimate temporal scour depth at abutment. All experiments were conducted under clear-water conditions. Extensive data sets were collected from present and previous studies. To determine the best method, two models of ANNs, feed forward back propagation (FFBP) and radial basis function (RBF), and two kinds of ANFIS, subtractive clustering and grid partition, were investigated. The results showed that the accuracy of the FFBP with two hidden layers (RMSE = 0.011) is higher than that of RBF (RMSE = 0.055), multiple linear regression method (RMSE = 0.049) and previous empirical equations. A comparable prediction was provided by the ANFIS-grid partition method with RMSE = 0.041. This research highlights that the ANN-FFBP and ANFIS-grid partition can be successfully employed for prediction of scour hazard and reduction in bridge failure.