Intégration de l’identité et de la diversité fonctionnelle dans un modèle de surface terrestre : une étude de la productivité et de la stabilité d’un écosystème dans un contexte de changement climatique

Land surface models (LSMs), such as the ORCHIDEE model, represent vegetation using fixed parameters associated with plant functional types (PFTs), in a simplified approach that neglects functional identity and diversity within ecosystems. This limitation is particularly marked for permanent grasslands, which are biodiversity hotspots. In this thesis, I integrate functional metrics at the scale of plant communities, in particular the community-weighted means (CWM) and the community-weighted variances (CWV) of traits, into the ORCHIDEE model in order to evaluate the primary productivity of French permanent grasslands (FPGs) and its stability at the national scale. In the first part of the thesis (Chapter II), I studied the effect of the spatial variability of functional traits at the community scale (that is, the CWM maps) on the simulated net primary productivity (NPP) and gross primary productivity (GPP) over the period 2001-2019, using three key traits : specific leaf area (SLA), leaf nitrogen content (LNC), and leaf lifespan. The results show that accounting for the spatial variability of traits reduces the overall productivity dynamics across the territory by decreasing productivity in the most productive lowland areas (plains) and increasing it in the less productive areas (mountainous zones). These results are in coherence with satellite data (MODIS) and in-situ observations in several French regions. The second part of the thesis (Chapter III) explores the influence of functional identity via the CWMs on productivity and stability in the historical period (1960-2019) and future (2010-2099), according to two climate scenarios (RCP 4.5 and RCP 8.5). Stability is analyzed through its components : constancy and resistance (according to Mahaut et al., 2023), as well as a new component, resistance M10, introduced in this work. The objective is to compare the simulated relationships between productivity and stability with those derived from observations. The results reveal the complexity of these relationships, which vary greatly depending on the types of grasslands (calcareous, mesic, ruderal, and mountainous). The introduction of CWM traits makes it possible, in several French regions, to obtain more spatially coherent patterns of productivity and stability compared to simulations with fixed traits. Moreover, the evolution of average annual NPP, constancy, and resistance between 1960-2010 shows trends different from those of the 2010-2099 period, depending on the EURO-CORDEX climate model used.Finally, the last part of the thesis (Chapter IV) explores the temporal evolution of traits by combining CWMs and CWVs to simulate pseudo-species in ORCHIDEE. The objective is to generate a set of simulations based on the sampling of the SLA and Vcmax traits under three hypotheses of ecosystem adaptation (full adaptation, partial adaptation, and absence of adaptation). The temporal evolution of traits, NPP, constancy, and resistance is analyzed. The results indicate that trait-based adaptation strategies significantly influence the simulated productivity and its stability. In conclusion, this thesis shows that it is possible to integrate functional identity and diversity into a biosphere model such as ORCHIDEE. It demonstrates how trait-based modeling can improve the representation of the productivity and stability of permanent grasslands subjected to climate change and extreme events. These advances are crucial for improving the predictive capacities of Earth system models and land surface models, and for promoting sustainable land management while preserving biodiversity, environmental resources, and ecosystem services.