Coupled carbon-water exchange of the Amazon rain forest, I. Model description, parameterization and sensitivity analysis
Résumé
A bi-modal distribution of leaf area density with a total leaf area index of 6 is inferred from several observations in Amazonia. Predicted light attenuation within the canopy agrees reasonably well with observations made at different field sites. A comparison of predicted and observed canopy albedo shows a high model sensitivity to the leaf optical parameters for near-infrared short-wave radiation (NIR). The predictions agree much better with observations when the leaf reflectance and transmission coefficients for NIR are reduced by 25?40%. Available vertical distributions of photosynthetic capacity and leaf nitrogen concentration suggest a low but significant light acclimation of the rain forest canopy that scales nearly linearly with accumulated leaf area.
Evaluation of the biochemical leaf model, using the enclosure measurements, showed that recommended parameter values describing the photosynthetic light response, have to be optimized. Otherwise, predicted net assimilation is overestimated by 30?50%. Two stomatal models have been tested, which apply a well established semi-empirical relationship between stomatal conductance and net assimilation. Both models differ in the way they describe the influence of humidity on stomatal response. However, they show a very similar performance within the range of observed environmental conditions. The agreement between predicted and observed stomatal conductance rates is reasonable. In general, the leaf level data suggests seasonal physiological changes, which can be reproduced reasonably well by assuming increased stomatal conductance rates during the wet season, and decreased assimilation rates during the dry season.
The sensitivity of the predicted canopy fluxes of energy and CO2 to the parameterization of canopy structure, the leaf optical parameters, and the scaling of photosynthetic parameters is relatively low (1?12%), with respect to parameter uncertainty. In contrast, modifying leaf model parameters within their uncertainty range results in much larger changes of the predicted canopy net fluxes (5?35%).