Abstract: By accounting for the time-delay effects of multiple offtake disturbances along irrigation canals, previous studies modified the Integrator Delay (ID) model and proposed the Generalized Integrator Delay (GID) model. However, the computational method for its key parameter (i.e., the optimal offtake location) has not been established. To thoroughly investigate the impact of offtake location on the prediction accuracy of the ID model across different types of open canals, four typical canal sections were selected. These canal sections were derived from the following sources: the mild-slope and steep-slope test canal cases proposed by American Society of Civil Engineers, the Middle Route of the South-to-North Water Diversion Project, and an irrigation canal from a large-scale irrigation district in China. Assuming each canal section contains only one offtake, the prediction accuracy of the ID model was evaluated by comparing its results with those of unsteady flow numerical simulations. Numerical simulation experiments were conducted under various offtake locations and operating flow conditions to investigate the relationship between the prediction accuracy of the ID model and offtake location. The results revealed the applicable conditions of the ID model: (1) large open canal systems with extremely mild bed slopes, and (2) low-flow operation scenarios in other common types of open canals. In other scenarios, the prediction accuracy of the ID model was significantly affected by the offtake location. It is recommended to adopt the GID model to account for the time-delay effects of offtake disturbances along the canal. Based on extensive experimental results presented above, an empirical formula for determining the optimal offtake location was developed. This empirical formula demonstrates high accuracy and broad applicability while requiring only known parameters for rapid computation. Compared with the ID model, it eliminates the need for tedious parameter calibration, thereby significantly reducing the computational burden for implementing the GID model. Model predictive controllers were then designed based on the ID model and the GID modes, respectively, with their control performance validated using the benchmark case proposed by the American Society of Civil Engineers. The advanced controller based on the GID model showed significant performance improvements. For the classical single-pool test case, the controller achieved 52.7% enhancement in water level control stability, 31.7% improvement in water supply reliability, 31.2% and 28.1% reduction in water level fluctuations and overshoot, respectively. For the classical multi-pool test case, the controller exhibited 42.0% increase in water level control stability, 3.8% improvement in water supply reliability, 40.4% and 33.2% mitigation of water level fluctuations and overshoot across canal pools. These improvements indicated more precise and stable water delivery to users along the irrigation canals. The GID model offers a relatively straightforward control modeling method, significantly enhancing water level control performance without increasing parameter identification complexity, demonstrating strong potential for precision irrigation applications. The research achievement can provide guidance for applying the GID model for the control modeling of irrigation open canals.