Abstract :
[en] The Earth's climate depends on the interactions among multiple interconnected components of the Earth system, including the atmosphere, hydrosphere, cryosphere, biosphere, and anthroposphere. Each of these components exerts varying degrees of influence on the others, which can either amplify or dampen one system responses. These complex interactions and feedbacks make it challenging to monitor, study, and model the climate and ecosystems, as their effects can occur locally or globally and operate over short to long timescales. For instance, drought can locally accelerate leaf shedding in forests, whereas rising global temperatures can gradually increase tree mortality and induce species migration.
The land-atmosphere interface represents one of the key hotspots of these interactions. Yet, regional studies often rely on models that simulate only one side of the land-atmosphere boundary while prescribing the other, thereby limiting the representation of coupled feedback.
This thesis investigates how remote sensing data and process-based modelling can improve the representation of land-atmosphere interactions in regional models. Specifically, we integrate satellite observations into the regional climate model MAR and couple it with the dynamic vegetation model CARAIB to provide more dynamic and physically consistent surface conditions.
First, we assessed the sensitivity of MAR to vegetation cover density by replacing its standard leaf area index input derived from a fixed MERRA2 monthly climatology at 50 km resolution with satellite leaf area index data from MODIS. This replacement enhanced both spatial resolution (500 m) and temporal frequency (8-day). Despite not explicitly modelling vegetation dynamics, MAR’s internal modules responded to changes in vegetation density, showing increased maximum daily temperatures during droughts and reduced extreme precipitation, although effects over longer periods upon the present climate remained less clear.
Building on these findings, we developed a two-way coupling between MAR and CARAIB using file exchange, enabling each model to use the outputs of the other as inputs. This coupling enables a more comprehensive representation of feedbacks between the land surface and the atmosphere. The coupled system exhibited stable behaviour over Western Europe, showing no climate drift and exhibiting good agreement with observations. Compared to the uncoupled configuration, it produced denser vegetation, enhanced day-to-day vegetation variability, and a stronger response to drought events.
Finally, we integrated satellite microwave observations into MAR to estimate the occurrence of wet snow over Antarctic ice shelves. As remote sensing remains one of the only viable means to monitor surface-melt extent across Antarctica, we applied a temperature nudging approach to align MAR’s simulated wet snowpack with satellite-derived observations over the Antarctic Peninsula. The results revealed that MAR systematically underestimates wet-snow extent, particularly during the early and late melt seasons. Substantial increase in the snowpack temperature was required to match satellite observations, suggesting possible underestimation of surface melt quantity or excessive refreezing and percolation processes. Despite the strong increase in surface melt induced by the integration of the satellite data, the simulated surface mass balance remained largely unchanged due to MAR's snowpack absorption capacity.
In summary, both the coupling and integration experiments improved the representation of land-atmosphere interactions in MAR by providing it with dynamical data and more representative observed surface conditions. These developments are essential steps toward a better understanding of surface processes and reducing uncertainties in long-term regional climate projections.
