Challenges and opportunities in building a global model of plant hydraulics

Observations of drought-driven damage to vegetation are widespread but, until recently, large-scale terrestrial models used to study climate-vegetation interactions did not capture the contrasting sensitivities of plants to drought. This is changing with the advent of a generation of models that consider plant hydraulics. Plant hydraulics link plant water status to pedoclimatic conditions; as such, explicit consideration of plant hydraulics should make models more mechanistic and predictive. Models, however, diverge in how they represent the plant water transport pathway and relate it to other plant functions (e.g., photosynthesis), so they further diverge in their parameterisation approach for hydraulic processes. Only at the most basic level do plant hydraulic implementations converge on a common set of measurable traits or parameters: two that describe a hydraulic vulnerability curve (e.g., P₁₂ and P₅₀, the water potentials at which 12% and 50% of a plant’s hydraulic conductivity are lost, respectively), and one that quantifies the efficiency of water movement within the plant (e.g., maximum hydraulic conductance). Regrettably, we do not yet know how to obtain regional- or global-scale hydraulic parameters from local-scale measurements, nor how to connect them to other plant traits. In this study, we propose strategies to leverage cross-species hydraulic diversity when scaling traits from the species level into model parameters. We also emphasise the importance of accounting for (i) within-species trait variability across space (e.g., interactions between hydraulic traits and their environment) and (ii) cross-functional trait covariation (i.e., interactions – or lack thereof – among traits that characterise different functional axes). Beyond advancing regional and global plant hydraulic modelling, efforts to address the suggested strategies would ready models for simulations that capture the resilience of vegetation communities worldwide.