Discussion

- Changes in climate and land use have a larger direct impact than rising CO2 on global river runoff trends

A decline in runoff in response to rising atmospheric CO₂ (simulation E1) (Fig. 3B) implies a concurrent increase in plant transpiration (PT). Without considering any effects on leaf energy balance, the ecosystem transpiration change between today's CO₂ levels and preindustrial CO₂ levels can be expressed as a function of changes in LAI and in stomatal conductance: PT(CO₂) = PT₀ × [1 + ΔLAI(CO₂)] × [1 − ΔS(CO₂)], where PT(CO₂) is the value of PT under current CO₂ levels, PT₀ is the preindustrial value of the transpiration, and ΔLAI(CO₂) and ΔS(CO₂) are the relative changes in LAI and stomatal conductance, respectively, caused by rising CO₂. This simplified equation assumes a linear relationship between transpiration per leaf area and stomatal conductance. If the term [1 + ΔLAI(CO₂)] × [1 − ΔS(CO₂)] is >1, an increase in transpiration is predicted. On average, an increase in vegetation productivity of 0.13–0.2% per ppm increase in CO₂ (14, 18) and a decreasing stomatal conductance of 0.07% per ppm increase in CO₂ (6) suggest potential increases in transpiration of ≈0.06–0.1% per ppm with rising CO₂, which is comparable with the value of 0.04% per ppm derived in simulation E1. Notwithstanding, projections of future behavior must be carried out with caution because the response of vegetation growth to elevated CO₂ is a nonlinear process (19). Controlled experiments with plants growing in CO₂-rich environments have indeed indicated that LAI increased sharply from low (200 ppm) to ambient (360 ppm) levels of CO₂ but tended to saturate at higher CO₂ levels (7). As a consequence, it is possible that the continuous increase in atmospheric CO₂ will probably lead to a decline in transpiration in the future, as suggested by field experiments (20).

If there is a decrease in runoff and an increase in PT in response to rising CO₂, such as that suggested by our model simulation E1, then these results have important implications for both soil-water storage and plant growth. During the last century, global soil moisture estimated by simulation E1 declined by ≈1% because of elevated CO₂. Such a decrease in soil moisture can exert negative feedbacks on vegetation growth, especially in water-stressed ecosystems. To illustrate the interactions between CO₂-related hydrological changes and vegetation growth, we analyzed the seasonal trends in LAI [monthly values of ΔLAI(CO₂)] and soil-moisture contents for temperate grassland ecosystems (25–50°N) during the last century from simulation E1. The increase in LAI was less pronounced at the end of the growing season than during the early growing season, which is in agreement with the results from FACE experiments (21). Pinter et al. (21) found that elevated CO₂ only led to an increase in LAI in the early and middle parts of the growing season. Such a decrease in the response of plant growth to CO₂ was probably associated with fall soil moisture deficits because of prior enhanced vegetation growth. The largest decrease in modeled soil moisture occurred at the end of growing season.

Uncertainties still exist regarding the global runoff trend over the last century, despite the good agreement between the results of simulation E3 (considering the joint effects of atmospheric CO₂, climate, and land-use change) and the observation-based reconstruction of global runoff change (2). For example, agricultural irrigation plays an important role in the global water cycle, particularly in North America and Eurasia (22, 23), but we have too little knowledge about its potential contribution to the runoff trend on a global scale. Because of the lack of information about historical changes in the agriculture irrigation area and water use for irrigation, we did not consider the effects of changes in irrigation on the global runoff trend. A previous study suggested that irrigation diverts water from runoff to evapotranspiration by the order of 4–5% of annual global runoff (24). Therefore, our simulated historical global runoff trend may be overestimated. However, it also is possible that the negative effect of irrigation on runoff is partly compensated for by an increase in precipitation resulting from irrigation-driven enhancements of evapotranspiration (5, 25). Because the comparison with Labat et al. (2) is performed at the continental scale, the impact of irrigation on the runoff trend should be relatively small. Further studies based on spatially and temporally explicit historic irrigation data sets are needed to quantify the role that irrigation change plays in the regional and global hydrological cycles.