The atmospheric carbon dioxide (CO₂) growth rate varies interannually in response to climate variations. Most of this variation arises from terrestrial ecosystems, where climate affects vegetation productivity (net primary productivity, NPP) and the rate of respiration of organic matter to CO₂ (heterotrophic respiration, HR). While remote sensing observations are available to constrain variations in NPP, there are no large-scale observations that can be used to infer heterotrophic respiration rates. We therefore investigate the signature that heterotrophic respiration leaves on atmospheric CO₂, which reflects both NPP and HR, at several timescales. We use a novel testbed approach to simulate HR and, in turn, an atmospheric tracer transport model to simulate the imprint of this process on atmospheric CO₂. The testbed includes two models that are microbially explicit (MIMICS and CORPSE) and one that is microbially-implicit (CASA).

We quantified the interannual variability, seasonality, and climate sensitivity of HR fluxes across four ecoregions and of the fraction of CO₂ variability owing to HR over six latitude bands. The microbially explicit models do a better job at capturing the magnitude of interannual variability when evaluated against CO₂ observations from NOAA marine boundary layer sites. HR is an important component of this modeled variability: at interannual timescales, the magnitude of simulated CO₂ variability from HR is comparable to that of CO₂ simulated only from NPP across the Northern Hemisphere. At seasonal timescales, the amplitude of CO₂ from HR is about 50-75% of that from NPP. CASA, the microbially implicit model, showed the smallest variations both seasonally and interannually. We found that NPP fluxes in warmer regions are more variable, while HR fluxes are more variable in colder regions. Given dominant patterns of atmospheric transport, this means that the apparent influence of NPP or HR on patterns of atmospheric CO₂ is different. In general, the three testbed formulations suggest that CO₂ from boreal HR is highly correlated with the observed CO₂ variations, suggesting a possible path to use atmospheric CO₂ observations together with productivity data to constrain hemispheric HR.