Abstract :
[en] The functioning of temperate terrestrial ecosystems is increasingly shaped by the combined impacts of climate change and land-use transitions, with critical implications for carbon cycling, ecosystem resilience, and sustainable land management. This thesis investigates how land-use dynamics, species-level plant traits, and adaptive forest interventions influence carbon balance, water-use efficiency (WUE), and drought resilience in Wallonia, Belgium. To address these questions, a high-resolution simulation framework was developed using the CARAIB Dynamic Vegetation Model (DVM), integrating machine learning–based land cover classifications from Landsat imagery, physiological trait data from the TRY database, and regionally downscaled climate projections. Simulations were performed at 1-km resolution for the period 1980–2070, covering both retrospective validation and forward-looking scenario analyses. Historical simulations (1980–2020) showed strong agreement with satellite and field observations, with gross primary productivity (GPP) yielding $R^{2} > 0.85$ against MODIS and GOSIF datasets, and aboveground biomass correlating with GEDI (R\textsuperscript{2} = 0.77) and ESA-CCI (R\textsuperscript{2} = 0.91).
Two complementary scenario modeling strategies were implemented to explore future ecosystem trajectories. The first focused on land-use change and included five spatially explicit scenarios based on regional planning assumptions: a business-as-usual baseline, a conservation-oriented strategy, and three afforestation pathways with varying levels of ambition. These scenarios provided a range of land conversion pressures and restoration potentials against which ecosystem functioning was assessed. The second strategy incorporated socio-economic projections from the ReCOVeR SmartPop model a cellular automata framework simulating demographic and economic drivers of land use. These projections were harmonized into CARAIB compatible land cover categories at 1-km resolution. Both strategies were coupled with species-specific silvicultural interventions including thinning, clear-cutting, and regeneration to evaluate adaptive forest management under RCP2.6 and RCP8.5 climate scenarios.
Simulation results revealed several key trends. Forested regions showed progressive gains in ecosystem carbon stocks across both historical and future simulations. Afforestation and conservation scenarios enhanced GPP, net ecosystem productivity (NEP), and biomass accumulation relative to baseline trends. Although CO\textsubscript{2} fertilization contributed to increased productivity, these gains were increasingly constrained under RCP8.5 due to intensifying heat and drought stress. Grasslands exhibited the highest seasonal GPP averaging over 2480 gCm\textsuperscript{--2}yr\textsuperscript{--1} while forests maintained stable annual productivity (1800–2200 gCm\textsuperscript{--2}~yr\textsuperscript{--1}) and played a dominant role in long-term carbon sequestration. Soil carbon declines were most pronounced in areas of sustained forest loss, underscoring the role of woody vegetation in supporting belowground carbon pools.
To better represent ecological mechanisms, simulations incorporated adaptive forest management rules for twelve dominant tree species. Functional outcomes varied substantially among species, influenced by hydraulic strategy, carbon allocation patterns, and climate sensitivity. Species exhibited a continuum of stomatal regulation strategies, from more isohydric to more anisohydric behavior. For instance, \textit{Fagus sylvatica}, \textit{Quercus robur}, and \textit{Alnus glutinosa} often classified as intermediate generally sustained or improved carbon use efficiency (CUE) and WUE under elevated CO\textsubscript{2} and active management. In contrast, strongly anisohydric species such as \textit{Populus nigra} and \textit{Pinus sylvestris} experienced marked declines under unmanaged, high-emission conditions. Thinning and regeneration improved WUE in most species by enhancing water availability and reducing canopy competition. These interventions also supported biomass recovery and long-term carbon storage, although their effectiveness declined under extreme climatic stress in RCP~8.5. Notably, \textit{Alnus glutinosa} exhibited consistently high drought resilience, evidenced by elevated RI\textsubscript{CUE} values, underscoring its functional stability under environmental stress. Leaf mass fraction (LMF) the proportion of biomass allocated to leaves emerged as a robust predictor of interspecific variation in CUE, linking plant morphology with carbon efficiency.
To evaluate vulnerability under climate extremes, species-level drought response indices (RI\textsubscript{CUE}, RI\textsubscript{WUE}) were developed to quantify functional changes during drought years relative to normal baselines. These indicators revealed substantial spatial and interspecific variability, with some species exhibiting strong average performance but low drought stability. Mapping these metrics enabled the identification of resilience hotspots and provided a spatially targeted basis for ecological risk assessment.
The modeling framework also incorporated uncertainty analysis using type-1 fuzzy logic to represent variability of the ecosystem to trait parameterization. By integrating species traits, dynamic land-use scenarios, adaptive forest management, and resilience diagnostics, this thesis presents a spatially explicit and ecologically mechanistic assessment of terrestrial ecosystem functioning under changing climate and land-use regimes. The results offer actionable insights for policymakers, forest managers, and conservation practitioners aiming to promote carbon resilience, ecological stability, and sustainable land stewardship in temperate regions and beyond.
