Wood You Be-Leaf It? The First Trait-Based Map of Global Vegetation Water Storage

Climate change is drastically affecting the health and composition of terrestrial vegetation through increasing average temperatures and altered water availability. Terrestrial vegetation is a key mediator of global water fluxes through processes such as evapotranspiration and photosynthesis, meaning that the peril posed by the degradation of vegetation will be felt at the global scale. Because of this, there is mounting interest in modelling global vegetation water storage (S_veg) as a means to understand the role that vegetation plays in maintaining a stable climate and how this will be affected by the climate crisis.

Currently, estimates of global S_veg can be derived from satellite imagery using microwave remote sensing. The microwave signal is attenuated by the amount of water contained in vegetation and is then related to S_veg through a look-up table of land cover-specific values, known as the b parameter. Along with water storage, the b parameter is influenced by the biomass and structure of the vegetation, however the interaction between these variables and their influence on the b parameter is poorly understood. Therefore, to further elucidate the physiological component that the b parameter represents in satellite derived estimates of S_veg, it is necessary to generate independent physiologically derived S_veg estimates. Here, we present the first globally explicit trait-based map of S_veg, with vegetation separated into two physiological components: wood and leaf tissue. Phylogenetically imputed species-level values for wood density (WD) and specific leaf area (SLA) were used for 46,309 plant species, derived from field and laboratory-measured data. Wood water storage was estimated through a linear relationship with WD and biomass. Leaf water storage was estimated through a non-linear relationship with SLA and scaled to the canopy with leaf area index. Comparing our S_veg estimates with independently derived plot-level trait-based estimates demonstrated a strong correlation (R² 0.94), suggesting our phylogenetic imputation approach to be robust and scalable.