Spatiotemporal variation in carbon use efficiency derived from eddy-covariance measurement of global terrestrial biomes

Vegetation carbon use efficiency (CUE), the ratio between net primary productivity (NPP) and gross primary productivity (GPP), provides insight into the ability of ecosystems to transfer large amounts of carbon (C) from the atmosphere to potential C-sinks. Although the patterns and feedback of CUE on climate change have been previously studied, large uncertainties remain due to methodological constraints. To address this problem, we proposed a new method that enables the separation of autotrophic respiration (R_a) from ecosystem respiration (R_e) by assuming that R_a is related to the lower bound of the relationship between R_e and GPP. By applying this method, we analyzed flux data acquired from 195 sites globally in an investigation of spatiotemporal dynamics in CUE. The results revealed a global average CUE of 0.50 ± 0.13, with the greatest values corresponding with croplands and the lowest with mixed forests. Spatially, CUE was greatest for Mediterranean and subtropical regions, and least for tropical regions. Temporally, CUE exhibited seasonal fluctuations across most biomes, with CUE increasing during the early growing season and then decreasing as the season progressed. We also investigated CUE's response to variations in several environmental drivers (e.g., air temperature, soil moisture, and incident solar radiation), with the help of machine learning, specifically extreme gradient boosting (xgboost) and a SHapley Additive exPlanation (SHAP)-value based interpretation of the results. A negative relationship was shown to exist between ambient CO₂ concentrations and CUE, confirming hypotheses that relate translocation and accumulation of nonstructural carbohydrates in plant tissues. These findings highlight the feasibility and value of leveraging flux data through advanced methods in deepening our understanding of CUE dynamics and their regulation at a global scale.