Relative importance of climatic variables, soil properties and plant traits to spatial variability in net CO₂ exchange across global forests and grasslands

Compared to the well-known drivers of spatial variability in gross primary productivity (GPP), the relative importance of climatic variables, soil properties and plant traits to the spatial variability in net ecosystem exchange of CO₂ between terrestrial ecosystem and atmosphere (NEE) is poorly understood. We used principal component regression to analyze data from 147 eddy flux sites to disentangle effects of climatic variables, soil properties and plant traits on the spatial variation in annual NEE and its components (GPP and ecosystem respiration (RE)) across global forests and grasslands. Our results showed that the largest unique contribution (proportion of variance only explained by one class of variables) to NEE variance came from climatic variables for forests (24%-30%) and soil properties for grasslands (41%-54%). Specifically, mean annual precipitation and potential evapotranspiration were the most important climatic variables driving forest NEE, whereas available soil water capacity, clay content and cation exchange capacity mainly influenced grassland NEE. Plant traits showed a small unique contribution to NEE in both forests and grasslands. However, leaf phosphorus content strongly interacted with soil total nitrogen density and clay content, and these combined factors represented a major contribution for grassland NEE. For GPP and RE, the majority of spatial variance was attributed to the common contribution of climate, soil and plant traits (50% - 62%, proportion of variance explained by more than one class of variables), rather than their unique contributions. Interestingly, those factors with only minor influences on GPP and RE variability (e.g., soil properties) have significant contributions to the spatial variability in NEE. Such emerging factors and the interactions between climatic variables, soil properties and plant traits are not well represented in current terrestrial biosphere models, which should be considered in future model improvement to accurately predict the spatial pattern of carbon cycling across forests and grasslands globally.