Reduced stomatal conductance under elevated CO2 results in increased soil moisture, provided all other factors remain constant. Climate models coupled to land-surface modules therefore predict increased global runoff under elevated CO2. In contrast to previous models which use standard physiological response curves to infer transpiration under elevated CO2, we combine transpiration (sap flow) and soil moisture data from the Swiss Canopy Crane FACE experiment with 104 years of daily weather data from an adjacent meteorological station to drive a three-layer bucket model. This allows to test the CO2-sensitivity of runoff to model parameters and precipitation pattern and to identify conditions under which flooding risk may increase.

Results/Conclusions

Over all 104 years, a doubled preindustrial CO2-concentration results in a 2.8 % (5.5 mm/year) increase in ecosystem runoff. Out of all 37986 simulated days (1.1.1901 to 31.12.2004), 576 (1.5 % of all days) produce higher runoff in the elevated CO2-scenario. The maximum increase in runoff due to a physiological CO2-response on a single day was 17 mm on October 5th, 1939. Only 1 out of 17 years produces a CO2-signal greater than 20 mm/year, which mostly results from 7 to 35 single days when runoff under elevated CO2 exceeds runoff under ambient conditions. Several weeks of intermediate, well-spaced precipitation events produce the highest CO2-effect on runoff. Very dry as well as very wet periods cause the difference in soil moisture between the two scenarios to disappear. Increased runoff under elevated CO2 is ca. 9 times more sensitive to changes in rainfall distribution (maximum effect 46 mm/year) than to changes in the model parameters (e.g. soil water capacity, interception; maximum effect 5 mm/year). Therefore, the key driver of increased runoff under future CO2-concentration is the day by day rainfall pattern. However, most modelling studies to date use monthly means or random rain generators to simulate precipitation. We argue that increased runoff due to a first-order plant physiological CO2-effect will be small (ca. 3 %) on average in a landscape dominated by temperate deciduous forest, and will increase flooding risk in catchments in years with exceptional rainfall distribution only.