Global climate models are converging on the prediction that many tropical forests will experience significant drying over this century. Land surface models are one of the most promising tools available for assessing the implications of such drying on the carbon cycle of the tropics. However, their ability to capture drought induced tree mortality remains largely untested and poorly constrained. The 1982-1983 El Nino drought at Barro Colorado Island (BCI) in Panama is an ideal test case for evaluating mortality predictions by land surface models because (a) it reflects one potential drying scenario for the tropics, which was a lengthening and extreme intensification of the dry season, and (b) mortality rates of all trees > 1 cm dbh have been monitored in a 50-ha plot on the island since 1981. An increase in mortality rates was observed in the census following the 1982-83 El Nino for trees belonging to the canopy-dominant and gap-pioneer demographic groups. In contrast, mortality rates of the canopy-subdominant and gap-generalist demographic groups were less affected by the drought. In this analysis, the Ecosystem Demography (ED2) model, a land surface model with demographic and hydrodynamic processes explicitly represented, was evaluated for its ability to capture the observed mortality patterns of the BCI forest caused by the 1982-83 El Nino drought.
Results/Conclusions
Preliminary results from the model simulations of the BCI forest response to the 1982-83 El Nino drought are promising. The model correctly captured the observed increase in mortality rates of trees parameterized as pioneer compared to trees parameterized as late successional. The model also correctly predicted increased mortality rates among larger tree size-classes during the drought. In addition, the hydrodynamic formulation predicts that trees parameterized as drought intolerant have higher mortality rates compared to tolerant trees. The drought tolerance of trees in the model is based on differences in hydraulic traits. These results now provide testable hypotheses about how plant traits should differ between species at BCI that had observed differences in mortality rates caused by the 1982-83 El Nino. These results also show how model sensitivity is related to uncertainty regarding soil boundary conditions and plant traits associated with growth, mortality and hydraulics.