Leaf conductance, isohydric strategy, and 50 shape drought responses of European tree species in a dynamic vegetation model

Increasingly frequent and intense drought events can jeopardize the current and future productivity and health of forests. Consequently, the ability of dynamic vegetation models (DVMs) to simulate drought impacts is paramount to improving their representation of the carbon cycle. To capture the physiological damage inflicted by drought, many state-of-the-art DVMs have implemented representations of plant hydraulic architecture in recent years. Although the understanding of the underlying processes governing hydrodynamic behavior in plants has steadily increased, the parameterization of hydraulic traits for different plant functional types (PFTs) remains a source of uncertainty in model output in part due to limited data availability. Here, we use LPJ-GUESS-HYD, an extension of LPJ-GUESS with new parameters and processes to simulate plant hydraulic architecture, isohydrodynamic water-potential regulation, and hydraulic failure mortality. Using latin hypercube sampling we create 6000 sets of hydraulic parameter combinations based on values found in the literature. Based on these parameter sets, we conduct a comprehensive variance-based sensitivity analysis for a set of 12 common European tree species across 37 sites from the FLUXNET 2020 warm winter dataset, encompassing a wide range of European ecosystems. Subsequently, we determine which parameters and parameter interactions contribute the most to variations in model outputs. Our results indicate that of the seven parameters used in the hydraulic architecture model of LPJ-GUESS-HYD, only a few have a significant effect on the model outcomes. More specifically, 50, the water potential at which 50 percent of conductance is lost, and maximum specific leaf conductance had the largest impact on simulated processes. Parameters related with the isohydric strategy of plants, had a lesser but still substantial role in shaping the model output. These results suggest that certain hydraulic parameters and combinations thereof play a disproportionate role in modulating simulated forest fluxes and states in LPJ-GUESS-HYD. Specific parameterization choices can drastically alter model performance, including whether PFTs can survive in a given climate or not. Aside from encouraging careful consideration of the available trait data when parameterizing new PFTs, our results may guide future experiments in choosing which hydraulic traits to focus on.