Workshop 5 : Avignon 2025

On the use of economics and hydraulic trait data in vegetation models:  implicit assumptions, supporting evidence, and consequences for predictions 

A workshop co-organized by Nicolas Martin, Isabelle Maréchaux, and Miquel de Cáceres (CREAF, Spain) was held in Avignon from 9 to 11 April 2025, bringing together 21 participants.

The workshop was motivated by the following observation: process-based models, and plant hydraulic models in particular, are increasingly being used for applications in ecology and environmental sciences (e.g., land surface models, dynamic vegetation models, forest dynamics models, potentially including forest management). A major advantage of these models is that they are trait-based. However, despite the increasing availability of hydraulic trait data for a wide range of plant systems in dedicated databases, measurement difficulties and biological variability across different scales make the goal of universally characterizing these traits unrealistic.

Consequently, model parameterization often relies on a series of steps aimed at compiling, in more or less opportunistic ways, all the values required. Although these steps are often regarded as merely technical prerequisites to the modelling process itself, parameterization actually relies on numerous biological, methodological, and ecological assumptions that often remain implicit and are rarely discussed. These assumptions may nevertheless have important consequences for simulation results. Moreover, their implicit nature limits the transparency and generalizability of modelling exercises, which are increasingly advocated, including for operational applications.

During the meeting, we first compiled a list of traits commonly used in models of plant gas exchange and hydraulics and briefly discussed the sources of their variability. We then worked in subgroups with the following objectives:

1. Identify the implicit assumptions made by modellers when parameterizing their models.
    
2. Explore relationships among traits that could potentially be used for imputation.
    
3. Determine which traits are sensitive to which assumptions.
    
4. Propose modelling exercises to test the impact of these assumptions on different model outputs (e.g., fluxes, productivity, drought-induced mortality risk, long-term community assembly).

A publication, with an associated R package supporting the parameterization of models with traits, is currently in preparation.