Abstract. Irrigation is a major issue for water resources management agencies as it is the main component of human fresh water consumption. However, irrigation can be monitored at plot scale but not at larger scales, i.e. from river basin to global scale. Hence, simulating the irrigation process in models is of great interest, not only to forecast the water availability, but also to provide realistic lower boundary conditions for atmosphere and climate models. This process is relatively well represented in agronomical or agro-hydrological models, designed for crop and water management at the plot scale. But this kind of model is not adapted for water management at the basin scale or even larger scale, due to their complexity. Land Surface Models (LSMs) are used for this purpose. However, irrigation is not well represented in LSMs. These models use basic decision rules to estimate irrigation volumes. Most of the time, it only consists in triggering an irrigation event when the soil moisture in the root zone drops below a fixed threshold. This threshold is unique at global scale, being independent of the crop type or the common irrigation practices in the simulated area. Then an irrigation amount is applied based on the volume needed to replenish the soil reservoir to a fixed level. There is no consideration about actual agricultural practices. These simple irrigation schemes do not have the flexibility needed to adapt to the wide variety of crops and irrigation practices encountered at large scales. The present study aims at developing and evaluating an irrigation scheme very similar to the one used in agronomical or agro-hydrological models for the SURFEX-ISBA LSM developed by Meteo-France. Particularly, it allows adapting the triggering threshold spatially and temporally and relating it to the actual phenology of the crop and to the irrigation practices. But increasing the flexibility of a model also means that it needs more input information to constrain it. High-resolution remote sensing products, like those derived from Sentinel-2, can provide part of this information spatially. This study thus presents a method to determine irrigation parameters, and particularly the triggering soil moisture threshold, from high-resolution remotely sensed leaf area index. This method is compared to three other experiments: a reference simulation with the current irrigation scheme of SURFEX-ISBA, a second experiment designed to show the contribution of remotely sensed irrigation period determination in the current scheme and a third which uses a single threshold over the season. The comparison is done on several maize plots in southwestern France. The results show that the method using remote sensing to modulate the triggering soil moisture threshold shows the best performances in estimating annual irrigation volumes. Indeed, it shows a bias around 10 mm per year and a RMSE around 30 mm whereas the standard scheme shows a bias around 50 mm per year and a RMSE around 60 mm. The sensitivity to the estimation of the soil maximal available water content is then performed. It shows that all the experiments are very sensitive when the maximal available water content in the soil is low. Finally, the impact on evapotranspiration is evaluated. It shows small differences between experiments and with the measured evapotranspiration. This study thus shows the potential of using high resolution remote sensing products to improve the irrigation simulation in LSMs. Indeed, it allows increasing the realism of the irrigation scheme while keeping it generic enough to simulate at regional to global scale.
Soil Moisture Monitoring Using Remote Sensing Data and a Stepwise-Cluster Prediction Model: The Case of Upper Blue Nile Basin, Ethiopia
