Land-atmosphere interactions and regional Earth system dynamics due to natural and anthropogenic vegetation changes

Abstract: Observation and modelling studies have indicated that the global land surfaces have been undergoing significant changes in the past few decades, driven by both natural and anthropogenic factors, such as changes in ecosystem productivity, fire and land use. Land surface changes can potentially influence local and regional climate through land-atmosphere interactions. Continued greenhouse gas emissions and current socioeconomic trends are expected to drive further land cover changes in the future, thus further understanding of land-atmosphere interactions including different feedback mechanisms is necessary to understand how future climate change will continue unfolding. Land-atmosphere interactions vary under different conditions. The strength of local land-atmosphere interactions depends on the capabilities of different land covers to control surface energy and mass exchanges, including latent and sensible heat, water and carbon. Local feedbacks can also influence regional to global climate, such as circulation changes that affect regional energy and moisture transport, or cloud cover that affects incoming radiation. Regional Earth system models (RESMs) with high resolution dynamical downscaling approaches and incorporating individual-based vegetation dynamics add value to the traditional global climate modelling studies for regions with highly complex topography or/and pronounced seasonal water deficits, potentially allowing for more refined land-atmosphere interactions studies thanks to more realistic vegetation dynamics and biophysical feedbacks, more accurate regional climate dynamics and overall richer spatial detail.In this thesis, I investigated regional land surface changes due to natural and anthropogenic vegetation changes and their impacts on land-atmosphere interactions, by applying a dynamical downscaling approach with RCA-GUESS, a RESM that couples the Rossby Centre regional climate model RCA4 to LPJ-GUESS, an ecosystem model that combines an individual-based representation of vegetation structure and dynamics with process-based physiology and biogeochemistry. Europe, Africa and South America were chosen as research domains. In the land surface study based on LPJ-GUESS simulations, I showed that future changes in the fire regime over Europe, driven by climate and socioeconomic change, were important for projecting future land surface changes. Fire-vegetation interactions and socioeconomic effects emerged as important uncertainties for future burned area. My study on land-atmosphere interactions based on RCA-GUESS simulations indicated that the hydrological cycle in the tropics was sensitive to land cover changes over semi-arid regions in Africa, and that biophysical feedbacks were important through their modulation of regional circulation patterns. A study based on the analysis of empirical datasets and CMIP5 ESMs outputs revealed that simulated climate biases are the main cause of model-data discrepancies. Models and data shared a marked hydrological relationship that suggested that decreased precipitation and land use change constituted the largest threats to the future Amazon forest. A study based on RCA-GUESS simulations with a realistic land use scenario identified both positive and negative impacts of land use on natural ecosystem productivity in the Amazon through its effects on the local and the regional climate.