Runoff and sediment yield modeling using ANN and support vector machines: a case study from Nepal watershed

Physics-based models for simulation of runoff and sediment yield from watersheds are relatively composite model based on learning algorithm. Physics-based is complex model due to involvement of tremendous spatial variability of watershed characteristics and precipitation patterns. Recently, pattern-learning algorithms such as the artificial neural networks (ANNs) have gained recognition in simulating the rainfall-runoff-sediment yield processes producing a comparable accuracy. We have simulated daily runoff and sediment yield from a Nepal watershed, Kankaimai (area = 1180 km(2)), with data from 1995 to 1999 for runoff prediction and 2001-2003 of the wet season for sediment yield prediction, using support vector machines (SVMs), a statistical learning theory based pattern-learning algorithm. The performance of the model was evaluated using the root mean square error, correlation coefficient and coefficient of efficiency. The results of SVM were compared to those of ANN and simple regression. ANN being a computationally intensive method, SVM could be used as an efficient alternative for runoff and sediment yield predictions under comparable accuracy in predictions.