Interannual variability of atmospheric CO2 is primarily driven by fluctuations in carbon uptake by terrestrial ecosystems. Some modeling results have shown that global terrestrial carbon uptake is dominated by the sensitivity of productivity to precipitation in semi-arid ecosystems, while atmospheric CO2 measurements indicate that global terrestrial carbon uptake is dominated by the sensitivity of respiration to temperature in tropical ecosystems. There is a need to better understand factors that control the carbon balance of land ecosystems across spatial and temporal scales. Here we used eddy covariance data, remote sensing observations, atmospheric inverse model, and process-based models to investigate how productivity (GPP) and ecosystem respiration (ER) determine net ecosystem exchange (NEE) over the contiguous United States (CONUS). We quantified the spatial sensitivity (βspatial), indicated by the change of GPP or ER in response to change of water availability over space, and temporal sensitivity (βtemporal), indicated by the sensitivity of GPP or ER in response to interannual water variability.
There is a spatial mismatch between the observed GPP by remote sensing and the NEE by atmospheric inverse model in CONUS. Strong correlations between GPP and NEE are evident along the precipitation along the West-East precipitation gradient, with strong positive correlations in semi-arid western ecosystems and weak or negative correlations in more humid eastern ecosystems . Both βspatial and βtemporal of GPP and ER to precipitation exhibit an emergent threshold where GPP is more sensitive to mean annual precipitation in semi-arid western ecosystems(MAP < 750 mm) and ER is more sensitive to precipitation in more humid eastern ecosystems (MAP > 750 mm). This emergent ecosystem threshold was evident in eddy covariance data, remote sensing observations, and atmospheric CO2 inversions. However, analyses from 10 TRENDY models indicate current Dynamic Global Vegetation Models (DGVMs) fail to capture this emergent ecosystem threshold behavior and tend to overestimate the sensitivity of GPP and ER to precipitation across CONUS ecosystems. Our results suggest that NEE was determined by productivity in the low precipitation (< 750 mm) less productive ecosystems, and by respiration in moist (> 750mm) high productive ecosystems. Furthermore our analysis indicates that productivity and respiration are too tightly coupled in DGVM models and that productivity in these models may be overly sensitive to precipitation. These results highlight emergent thresholds at the ecosystem and continental scale that may help reconcile model simulations and observations of terrestrial carbon processes.