In this paper by applying system identification tools a neural network model of an irrigation main canal pool is obtained. The complete system identification procedure, from experimental design to model validation, taking into account prior physical information, is developed. It is established that a nonlinear model with NARX structure can adequately describe the dynamic behavior of an irrigation main canal pool. The model validation results show that the model obtained reproduces with high accuracy the observed data and therefore it can be applied in the design of nonlinear control systems and/or for prediction purposes.