Conclusions

• All three estimates of Ae based on measurements of free tropos-pheric or canopy air were close to each other (Table 3). Since the spatial representation of the flask-sampling networks is still limited (Tans et al, 1996), the collection of canopy air to deduce Ae can be recommended.

• The strong negative relationship between Ae and A/Gc supported the hypothesis proposed by Buchmann et al (1998). Despite the variations of Ae within a biome, the use of Ae seems promising to detect differences in the ratio of carbon to water fluxes among biomes or changes in this ratio due to climatic or environmental conditions.

• Modeling Ae without consideration of land-use changes and the transient dynamics of clearance, agriculture, abandonment, and succession could contribute to the observed differences between modeled and measured Ae values. A sensitivity analysis of the importance of land-use change at a global scale is therefore necessary. Currently dynamic models are computer-intensive and coarser in PFT-specific-ness, but only future simulations with these models will be able to assess the long-term importance of "memory effect" and isotopic disequilibrium.

• Differences between flask-derived and modeled estimates of ecosystem discrimination (up to 3%o) were due to the lack of measured Ae estimates in certain ecosystems and/or regions and to model parameterization. Field measurements filling the spatial gaps as well as modeling actual vegetation cover and its physiology should be of high priority in global ecology research: According to Fung et al (1997), an underestimate of Ae in global models by 3%o (such as the difference we observed here) translates into a carbon flux of 0.7 Gt C year-1 or into an overestimation of the terrestrial biospheric carbon sink of 20%.

0 0

Post a comment