1. Detection of a direct carbon dioxide effect in continental river runoff records N. Gedney, P. M. Cox, R. A. Betts, O. Boucher, C. Huntingford & P. A. Stott

2. Possible reasons for river runoff increase climate change and variability Deforestation (land cover change) solar dimming (aerosol effect) direct atmospheric CO 2 effects on plant transpiration Runoff is the movement of landwater to the oceans, chiefly in the form of rivers, lakes, and streams. Transpiration is the evaporation of water into the atmosphere from the leaves and stems of plants.

3. On annual and longer timescales, runoff (R) = precipitation (P) – evapotranspiration (ET) Evapotranspiration is a function of energy and water availability, near-surface atmospheric conditions and the control of transpiration by plants. Climate change and variability modify precipitation patterns and near-surface meteorology. Increased aerosol concentration reduces surface solar radiation, leading to reductions in open-pan evaporation. Landcover changes modify the depth of the soil from which plants can extract water and the available energy by changing the land- surface albedo. Stomatal apertures on plant leaves are observed to close partially under increased CO 2, leading to suppression of transpiration

4. Met Office Surface Exchange Scheme (MOSES), using a monthly observational data set of the twentieth-century climate. Five simulations of twentieth-century surface hydrology:  All factors vary  Fixed climate  Fixed aerosol  Fixed atmospheric CO 2  Fixed land use

5. precipitation and modelled and observed runoff trends over the century  post-1960 observed runoff trends for all regions tend to be more positive  a clear disparity between the CLIM modelled runoff and the observed runoff trends  ALL appears closer to the observations than the CLIM response except Europe  modelled response from climate forcing, aerosols and direct CO 2 effects appear to be important over some regions  Land use (excluding irrigation) has a small effect on the modelled continental water balance

6. Use a standard optimal fingerprinting technique to see which factors are likely to be driving the long-term changes in continental runoff by comparing the modelled and observed annual anomalies relative to the long-term mean β scale factors obtained from the optimal fingerprinting technique. The 5 to 95 percentile ranges are shown for CLIM, AER, CO2 and LUSE. >0: detected at the 5% significance level >1:modelled component is consistent with the observed response Individual region regression results:  CLIM effect is detected over all regions at the 5% significance level, and in South America and Asia it is consistent with observation.  Direct CO2 effect is detected over Africa at the 5% significance level.  Aerosols and the CO2 effect are both consistent with the observed trends over Asia.  β(CO2)>0, β(CO2)> β(AER)> β(LUSE) All regions regression results: Only climate and the direct CO2 effect are detected at the 5% significance level. The simulation of CO2 effect is consistent with that observed, whereas the model slightly over-estimates the climate effect.

7.  Runoff enhancement due to suppression of transpiration is detected in the observational records.  Other mechanisms like expanding irrigation and human water consumption could also contribute to changes in observed runoff, but cannot explain the difference between observed and climatedriven simulated runoff.

8. Conclusion Raised CO2 levels are already having a direct influence on the water balance at the land surface. As the direct CO2 effect reduces surface energy loss due to evaporation, it is likely to add to surface warming as well as increasing freshwater availability. The existence of a direct CO2 signal in river runoff records also opens up the intriguing possibility of using long-term river records to monitor CO2 effects on the land carbon sink.

9. Huntington(2008): CO2-induced suppression of transpiration cannot explain increasing runoff Research findings from five major areas are consistent with an increase in the rate of annual actual evapotranspiration (AET):  river-basin water-balance  massive weighing lysimeters  pan evaporation  length of the growing season  plant responses to climate change

10. Plant responses to elevated CO2 that stimulate stomatal closure may induce an increase in leaf and canopy temperature. Increases in plant canopy temperatures will result in an increase in the vapour pressure deficit between the plant canopy and the surrounding atmosphere, thereby increasing transpiration. As temperature increases in conjunction with the increase in CO2, the increased heat burden in the canopy can result in the need for increased stomatal opening to provide evaporative cooling. Such adaptation in the canopy could offset some of the effect of CO2 on stomatal closure. The evidence supporting an increasing rate of AET over many temperate regions during the 20 th century indicates that the attribution of increasing runoff to decreasing transpiration fluxes is unlikely.

11. It is more likely that the increases in runoff observed for many basins are a result of an intensification of the hydrologic cycle whereby increases in precipitation, and in some cases net melting of ice and snow, have more than offset increases in AET. Finding the causes of trends towards increasing runoff is important for understanding hydrologic responses to ongoing climate warming. Discussion: Is it reliable to use optimal fingerprinting technique to estimate CO2 effect in the increasing river runoff trend? Using long-term river records to monitor CO2 effects on the land carbon sink, is it possible?