Linkages between stem vulnerability curves and tree demography and their implications for plant physiological modeling.

Vulnerability curves (VCs) have been measured extensively to describe the differences in plant vulnerability to cavitation. Although the roles of hydraulic conductivity (Ks,max) and hydraulic safety (P50, embolism resistance), both of which are parameters of VCs ('sigmoidal' type), in tree demography have been evaluated across different forests, the direct linkages between VCs and tree demography are rarely explored. In this study, we combined measured VCs and plot data of 16 tree species in Panamanian seasonal tropical forests to investigate the connections between VCs and tree mortality, recruitment and growth. We found that the mortality and recruitment rates of evergreen species were most significantly positively correlated with P50. However, the mortality and recruitment rates of deciduous species only exhibited significant positive correlations with parameter a, which describes the steepness of VCs and indicates the sensitivity of conductivity loss with water potential decline, but is often neglected. These differences among evergreen and deciduous species may contribute to the poor performance of existing quantitative relationships (such as the fitting relationships for all 16 species) in capturing tree mortality and recruitment dynamics. Additionally, evergreen species presented a significant positive relationship between relative growth rate (RGR) and Ks,max, while deciduous species did not display such relationship. The RGR of both evergreen and deciduous species also displayed no significant correlations with P50 and a. Further analysis demonstrated that species with steeper VCs tended to have high mortality and recruitment rates, while species with flatter VCs were usually those with low mortality and recruitment rates. Our results highlight the important role of parameter a in tree demography, especially for deciduous species. Given that VC is a key component of plant hydraulic models, integrating measured VC rather than optimizing its parameters will help improve the ability to simulate and predict forest response to water availability.