Proportional integral (PI) control is the most frequently used method for controlling water levels in irrigation canals. For design purposes, simplified linear models are typically used to describe the series of pools that, separated by control gates, compose the whole canal. Common models are made by series connection of integrator plus time delay or integrator with time delay plus a zero, these models being obtained from the Saint-Venant equations. However, it is known that hydraulic processes in irrigation canals are strongly nonlinear and present dynamics that are time varying depending on the operating point and the hydraulic conditions of the pool, so that the performance of PI controls may be reduced as the operating point moves far from the nominal design state. This paper presents the application of an adaptive predictive expert control (ADEX) for the distant downstream level control in irrigation canals trying to exploit their adaptive capability to efficiently achieve changes in the setpoint levels and disturbance rejection. Both PI and ADEX controllers are compared in simulation and in real field tests in the control of the end pool of the Canal Imperial de Aragón in Spain.KEY WORDS: predictive control; adaptive predictive control; irrigation canal control; water resources During the last 20 years, a significant number of researchers from different groups, mainly in academic or management centers and research environments related to water, have developed, for this problem, various methods of control in the theoretical context. A state of the art prior to 2000 can be seen in [1,2]. Some significant further studies are summarized below. A robust tuning method for proportional-integral (PI) classical control is presented in [3]. In [4], fractional PI control is applied.