Disentangling the influence of vegetation structure and physiology on land-atmosphere coupling

Terrestrial vegetation is a key component of the Earth system as it mediates the exchange of carbon, water and energy between the land and the atmosphere. Thereby, the vegetation affects the climate through changes in its structure (such as leaf area index, LAI) and its physiology (such as stomatal conductance); However, their relative contributions and respective processes on the land-atmosphere coupling are not yet understood. For instance, increased LAI, referred to as structural changes, promotes transpiration and vegetation productivity, and increases the surface albedo in most cases. In contrast, decreased surface conductance, referred to as physiological changes, could reduce transpiration and productivity. Therefore, the overall feedback of vegetation to climate change via water, carbon and energy exchange will depend on the relative importance of structural and physiological responses. Here we study to what extent dynamic changes in global vegetation structure and physiology modulate land-atmosphere coupling using satellite remote-sensing, data-driven, and earth system modelled vegetation data, as well ashydro-meteorological reanalysis. The land-atmosphere coupling is quantified through the correlation between soil moisture and lagged vapor pressure deficit determined with a moving time window. We employ random forests to quantify vegetation physiology by accounting for functional variability (e.g. GPP and ET) explained by hydro-meteorological data but not by the vegetation structure. Then using an explainable machine learning approach (SHAP), we determine the contributions of vegetation structure and physiology where we find overall larger contributions of structure on regulating land-atmosphere coupling during the growing season. The relative importance of vegetation structure differs across ecosystems, with stronger contributions in dry ecosystems. Furthermore, we analyze the variations of the relevance of vegetation structure over time and in particular during warm and dry periods. The results are partially backed up by using in-situ measurements of physiological traits to interpret the large-scale observed physiological patterns.