This paper propose a neuro-fuzzy Smith predictor controller (NFSP) for effective control of water distribution in an irrigation main canal, which allows to improve the efficiency and to reduce the water losses. The dynamic model of this process is obtained from an identification procedure based in real-time data. A neuro-fuzzy controller combined with a Smith predictor is therefore designed which behaves robustly against the effect of external disturbance and plant dynamical parameters variations. Simulated results compare the performance of the proposed controller with a standard PI controller, also combined with a conventional Smith predictor for several operation conditions. These results show that the proposed controller outperforms the standard controller in terms of performance and robustness.