| 10 Jul 2026
Multi-year convection-permitting irrigation impacts across the European continent
Abstract. Irrigation representation in weather and climate modeling is advancing and becoming more relevant. Global, regional and local studies that included irrigation in Earth system models already demostrated the effects of irrigation in different variables. Some regional and local simulations have studied the irrigation impact in some parts of Europe with simulations limited to the duration of one growing season. Therefore, long-term irrigation simulations that cover the whole European continent are still missing. This study quantifies the long-term impact of irrigation on surface and atmospheric variables over the EURO-CORDEX domain using convection-permitting ICON simulations that cover a period of 12 years. Our findings indicate that the magnitude of the irrigation impact is limited by the irrigated region and the year of study in the EURO-CORDEX domain, showing that the land-atmosphere coupling is key to determine the irrigation effects. For instance, the cold and wet summer of 2017 in south Europe weaked the irrigation cooling in the Alps. In contrast, the heat wave of 2018 in Central Europe enabled the influence of irrigation on surface variables in this region.
This study investigates the impacts of irrigation on the European climate using 3-km convection-permitting ICON simulations for 2010-2022. It examines how irrigation modifies temperature, humidity, surface energy fluxes, and heat-wave characteristics across different European regions. The manuscript is well written and well structured, and I have only a few minor concerns that I would like the authors to clarify.
-The irrigation prescription is idealized. The irrigation timing varies by region and crop. For example, the authors acknowledge that night irrigation is common in the Ebro basin, while their irrigation occurs at 05 UTC. This is not a minor uncertainty: the timing of irrigation controls the partitioning between evaporation, sensible heat, boundary-layer development, and potentially convection. At 3-km resolution, timing becomes particularly important because irrigation can influence the diurnal boundary layer. I would quantify the uncertainty in the magnitude of irrigation and uncertainty in the timing and diurnal distribution of irrigation, which can alter the atmospheric response even if the annual irrigation volume is correct.
-The paper uses maize as the representative crop while their irrigation map contains irrigated croplands generally, not maize-specific irrigated areas. So, there is a conceptual mismatch? Different European crops have varying rooting depths and seasonal water requirements, which can strongly influence the calculated RAW and, in turn, the irrigation volume.
-What is the actual irrigation water applied? Can the authors show the annual irrigation amount by region and compare with independent irrigation water use estimates? Otherwise, it is difficult to know whether the climate response is physically realistic.
-The irrigation water is sourced externally and is not subtracted from other water reserves. This means the experiment is adding water to the terrestrial system without accounting for river withdrawal, groundwater depletion, reservoir storage, etc. For an atmospheric irrigation-impact experiment, this can be acceptable if the purpose is specifically to isolate the atmospheric response to irrigation. Still, the paper sometimes discusses the results as if they represent realistic irrigation. The distinction needs to be much clearer.
-The paper uses Siebert et al. (2013). Does this mean that the experiment doesn’t capture the expansion of irrigation during the recent period? What about changes in crop type and irrigation technology?
-Does an irrigated grid cell become entirely “irrigated cropland”? At 3 km, many grid cells may contain cropland, forest, grassland, water, and non-irrigated agriculture. Replacing the land-cover class could introduce a land-use perturbation that is not irrigation alone.
-The paper validates the simulations against 3-hourly quality-controlled observations from ICON DREAM reanalysis. But ICON-DREAM is a reanalysis, not an observational dataset. More details are required on which observation network, stations, number of stations, spatial distribution, and whether the station observations are independent of the reanalysis.