Validation of an irrigation scheme inside ORCHIDEE land surface model at global scale

Irrigation activities play a key role in food production and consume 90% of freshwater withdrawal worldwide. These activities have a strong impact on water and energy budgets and associated biogeochemical cycles, and can have effects on local and regional climate. Furthermore, irrigation activities are projected to increase due to population growth and climate change. This context has encouraged the inclusion of irrigation in land surface models (LSMs), which simulate the continental branch in earth system models.

Here we present an irrigation scheme within the ORCHIDEE LSM that replicates flood and drip techniques. Water demand is calculated as the soil moisture deficit with respect to a target value. This deficit is estimated in the root zone of the crop and grass fraction (which contains both irrigated and rainfed crops), but the demand is limited by the fraction equipped for irrigation and by the water supply, i.e. water available in rivers and aquifers reduced to preserve a minimum volume in each water store for ecosystems. In addition, the scheme prioritizes water abstraction by source (surface or groundwater) according to the Siebert et al. map (2010). Hence, in a gridcell with little groundwater pumping infrastructure, most of the water will be extracted from the river, even if the water demand is not fully supplied. The water finally withdrawn for irrigation is allocated on the surface of the soil column for infiltration, and a maximum irrigation rate is set to prevent runoff production. The user-defined parameters that drive the scheme's response are the root zone depth and soil moisture target, the minimum volume left for ecosystems, and the maximum irrigation rate.

For validation, we use this scheme inside ORCHIDEE to run global offline simulations without and with irrigation. We use a set of parameter values that tries to fit the irrigation rates reported by AQUASTAT, while reducing the bias of evapotranspiration in irrigated areas with respect to the satellite-based products. We explore the possible reduction of bias in other variables like leaf area index, water storage anomalies and observed discharge. Finally, we correlate the bias reduction with landscape features to gain insights on the shortcomings of the irrigation scheme and ORCHIDEE.