Considerable amounts of water can be saved by automating irrigation canals. The design of most of the practical automatic controllers rely on a simplified model of the irrigation canal. This model can be obtained from measured data (identification) or can be formulated (white box models) assuming simplifications in the physical concepts and using the canal geometry. Several models of this kind are presently available. Moreover, short canals reveal a resonance problem, due to the back and forth of waves. This paper is focused on how to choose a suitable model for short canal pools with the purpose of control design. Four simple models are applied to two different types (resonant and non-resonant) of short canals: First order transfer function based on the Hayami model, Muskingum model, Integrator Delay (ID), and Integrator Delay plus Zero (IDZ). Model predictive controllers are developed based on these models and they are tested numerically and experimentally in order to evaluate their contribution to the control effectiveness. The controllers based on the ID and IDZ model showed the best performance.