Atmospheric observations inform CO2 flux responses to enviroclimatic drivers

Understanding the response of the terrestrial biospheric carbon cycle to variability in enviroclimatic drivers is critical for predicting climate-carbon interactions. Here we apply an atmospheric-inversion-based framework to assess the relationships between the spatiotemporal patterns of net ecosystem CO2 exchange (NEE) and those of enviroclimatic drivers. We show that those relationships can be directly observed at 1° × 1° 3-hourly resolution from atmospheric CO2 measurements for four of seven large biomes in North America, namely, (i) boreal forests and taiga; (ii) temperate coniferous forests; (iii) temperate grasslands, savannas, and shrublands; and (iv) temperate broadleaf and mixed forests. We find that shortwave radiation plays a dominant role during the growing season over all four biomes. Specific humidity and precipitation also play key roles and are associated with decreased CO2 uptake (or increased release). The explanatory power of specific humidity is especially strong during transition seasons, while that of precipitation appears during both the growing and dormant seasons. We further find that the ability of four prototypical terrestrial biospheric models (TBMs) to represent the spatiotemporal variability of NEE improves as the influence of radiation becomes more dominant, implying that TBMs have a better skill in representing the impact of radiation relative to other drivers. Even so, we show that TBMs underestimate the strength of the relationship to radiation and do not fully capture its seasonality. Furthermore, the TBMs appear to misrepresent the relationship to precipitation and specific humidity at the examined scales, with relationships that are not consistent in terms of sign, seasonality, or significance relative to observations. More broadly, we demonstrate the feasibility of directly probing relationships between NEE and enviroclimatic drivers at scales with no direct measurements of NEE, opening the door to the study of emergent processes across scales and to the evaluation of their scaling within TBMs.