Comment on “Optimized stomatal conductance of vegetated land surfaces and its effects on simulated productivity and climate” by A. Kleidon

Abstract

[1] Using a simple dynamic vegetation model coupled to an intermediate complexity climate model Kleidon [2004] suggests that an optimum stomatal conductance exists at which vegetation productivity is maximized. It is suggested that this maximum is the result of two competing processes: 1) the increased supply of CO2 with increased stomatal conductance and 2) increased cloud cover, associated with higher latent heat fluxes, that reduces sunlight. Here, it is argued that this conclusion is an artifact of the simple vegetation model used. Kleidon's vegetation model does not take into account that diffused radiation increases productivity so in his model an increase in cloud cover always affects productivity adversely due to decrease in available radiation. Also, transpiration, interception, and soil evaporation are not calculated separately with the result that a decrease in stomatal conductance always leads to a decrease in total evapotranspiration, while more complex models suggest that decrease in transpiration can be compensated by an increase in soil evaporation. [2] It is well established that vegetation maximizes carbon gain per unit water loss at the leaf [Cowan and Farquhar, 1977] as well as the ecosystem [Mediavilla and Escudero, 2003] scale by adjusting stomatal conductance and photosynthesis rates. In has also been suggested [Huxman et al., 2004] that as the supply of moisture is reduced the rain use efficiency (RUE, productivity per unit rain) approaches some maximum value that is independent of the vegetation structure. However, the suggestion by Kleidon [2004] that plants maintain an optimum stomatal conductance to control latent heat fluxes and thus cloud cover is questionable. [3] The concept of optimum stomatal conductance as a balance between CO2 uptake and increasing cloud cover to obtain maximum vegetation productivity is questionable because: 1) an increase in cloud cover can actually increase vegetation productivity via diffused radiation; 2) a decrease in transpiration and evaporation of canopy intercepted water is compensated by increase in soil evaporation; and 3) cloud cover is largely controlled by the large scale atmospheric circulation and not by local transpiration. These three points are briefly discussed below. [4] Several researchers have demonstrated that plant productivity is enhanced in the presence of clouds due to an increase in the diffused fraction of the total radiation [Hollinger et al., 1994; Baldocchi et al., 2002] although eventually the benefits of diffused radiation are outweighed by the overall decrease in solar radiation. The role of diffused radiation in enhancing net ecosystem productivity (NEP) has also been invoked to explain the decrease in the rate of growth of atmospheric CO2 after the eruption of Mount Pinatubo [Roderick et al., 2001; Arora, 2003]. The simple vegetation model used by Kleidon [2004] does not, however, make the distinction between the diffuse and direct fraction of solar radiation so that an increase in cloud cover always affects productivity adversely due to decrease in available radiation. [5] A reduction in stomatal conductance will lead to a decrease in transpiration but this may not necessarily lead to a substantial decrease in total evapotranspiration. Kleidon [2004] runs his simulations for 100 years for dynamic vegetation to reach steady state for range of stomatal conductance values. A reduction in stomatal conductance thus also leads to lower vegetation biomass and leaf area index (LAI) via lower CO2 uptake. Simulations with complex soil-vegetation-atmosphere-transfer (SVAT) schemes show that reduction in transpiration and the evaporation of intercepted water from the canopy is compensated by an increase in soil evaporation. For instance when testing the sensitivity to different leaf area indices, Tian et al. [2004] find that increasing LAI over the Amazonian region in the NCAR climate model with the CLM2 SVAT scheme does not increase latent heat fluxes significantly for exactly this reason. The model used by Kleidon [2004] does not treat transpiration and soil evaporation separately and thus a decrease in dimensionless stomatal conductance (gs) always leads to decrease in total evapotranspiration. [6] There is no observational evidence that plants are able to sense changes in clouds that occur in response to changes in local transpiration via stomatal control. More importantly, there is little evidence that local transpiration alone regulates the cloud cover. Rather, over large parts of the world, large-scale dynamics plays a dominant role in regulating the cloud amount [e.g., Wu and Moncrieff, 2001]. Although a reduction in transpirational flux at the global scale implies reduced cloud cover, plants are not expected to adjust their stomatal conductances in synchrony with each other or in a way such that they all benefit from each other's individual stomatal adjustment. [7] That an increase in cloud cover always affects the plant productivity adversely (without considering the benefits of diffused radiation) and that a decrease in stomatal conductance always leads to a decrease in evapotranspiration (without considering any increase in soil evaporation) and thus cloud cover is therefore built into the model structure. If the effect of diffused radiation is modelled explicitly, vegetation productivity will first increase with increasing cloud cover (due to benefits of diffused radiation) and then decrease (due to decrease in available radiation) as compared to a continuously decreasing function of increasing cloud cover as is assumed by Kleidon [2004]. In addition, if transpiration, evaporative interception losses, and soil evaporation are modelled separately then a decrease in stomatal conductance will lead to relatively smaller changes in cloud cover (since the decrease in transpiration is compensated to some extent by increase in soil evaporation). This is expected to shift the optimum stomatal conductance or may not result in identification of a clear optimum at all, indicating that the optimum stomatal conductance identified by Kleidon [2004] is a function of model parameterizations. [8] Useful comments from George Boer, Greg Flato, John Fyfe, and Francis Zwiers on an earlier version of this manuscript are appreciated.