As simple as possible but not simpler?: the case of irrigation modeling at catchment scale in southwestern France

The estimation of irrigation water requirements (IWR) amount and timing is crucial for designing water management strategies at the regional scale. Irrigation requirements can be estimated with different types of models: very complex and detailed crop models, agent-based models, or simplified modeling approaches. Because simplified approaches are often preferred, in this study, we evaluate the consequences of using simplified approaches for IWR assessment at a catchment scale and the consequences of various modeling choices, providing information on the uncertainties. To this end, different simple modeling approaches based on the CropWat model are compared with an agent-based approach (MAELIA), which serves as a benchmark. To assess simulations in detail, partial variance is calculated for several indicators characterizing daily simulated irrigation. Our sensitivity analysis, applied over a sub-catchment of the Aveyron River (southwestern France), shows a high variability in simulations produced by CropWat between the modeling assumptions tested, principally explained by the rules for irrigation triggering and the quantification of daily irrigation. The analysis also shows that several simplified approaches are able to reproduce the irrigation simulated by the high-accuracy MAELIA model, but not necessarily corresponding to an optimal irrigation scheme. Hence, this study confirms the possibility of assessing daily irrigation with simplified approaches, but warns about high modeling uncertainties, reflecting uncertainty in effective irrigation practices. This uncertainty can be taken into account by water managers and modelers through the combination of a set of irrigation models.