How much can we simplify irrigation in an integrated modelling purpose? A case study in southern France

The estimation of irrigation amounts and timings is crucial for the design of water management strategies at the regional scale. However, simplified modelling approaches are often preferred even though very complex and high-accuracy crop models or agent-based models exist. In this study, we develop a sensitivity analysis to evaluate the impacts of simplifications and hypotheses in irrigation modelling. For this, different simple modelling approaches based on the CropWat model were compared to a multi-agent based approach (Maelia), which served as a benchmark. To make an in-depth comparison between simulations, several indicators characterizing daily simulated irrigation were calculated and a decomposition of variance was carried out to measure impacts of diverse factors on irrigation. Applied over a downstream portion of the Aveyron River (southern France), the sensitivity analysis shows a high variability between simulations in function of modelling assumptions. It also shows that several simplifying approaches were able to reproduce the high-accuracy model estimation of irrigation. Decisive variation factors we identified are rules of triggering and quantification of daily irrigation, irrigation period definition and evapotranspiration estimation. Recommendations to take into account highlighted variability linked to farmers’ irrigation practices are introduced in this work, consisting in a combining a set of irrigation models. In function of advancement, a complete integrated agro-hydrologic modelling chain might be presented.