The observed interannual variability of atmospheric CO 2 reflects short-term variability in sources and sinks of CO 2 . Analyses using 13CO 2 and O 2 suggest that much of the observed interannual variability is due to changes in terrestrial CO 2 exchange. First principles, empirical correlations and process models suggest a link between climate variation and net ecosystem exchange, but the scaling of ecological process studies to the globe is notoriously difficult. We sought to identify a component of global CO 2 exchange that varied coherently with land temperature anomalies using an inverse modeling approach. We developed a family of simplified spatially aggregated ecosystem models (designated K-model versions) consisting of five compartments: atmospheric CO 2 , live vegetation, litter, and two soil pools that differ in turnover times. The pools represent cumulative differences from mean C storage due to temperature variability and can thus have positive or negative values. Uptake and respiration of CO 2 are assumed to be linearly dependent on temperature. One model version includes a simple representation of the nitrogen cycle in which changes in the litter and soil carbon pools result in stoichiometric release of plant-available nitrogen, the other omits the nitrogen feedback. The model parameters were estimated by inversion of the model against global temperature and CO 2 anomaly data using the variational method. We found that the temperature sensitivity of carbon uptake (NPP) was less than that of respiration in all model versions. Analyses of model and data also showed that temperature anomalies trigger ecosystem changes on multiple, lagged time-scales. Other recent studies have suggested a more active land biosphere at Northern latitudes in response to warming and longer growing seasons. Our results indicate that warming should increase NPP, consistent with this theory, but that respiration should increase more than NPP, leading to decreased or negative NEP. A warming trend could, therefore increase NEP if the indirect feedbacks through nutrients were larger than the direct effects of temperature on NPP and respiration, a conjecture which can be tested experimentally. The fraction of the growth rate not predicted by the K-model represents model and data errors, and variability in anthropogenic release, ocean uptake, and other processes not explicitly represented in the model. These large positive and negative residuals from the K-model may be associated with the Southern Oscillation Index.scales. Some of the variation is clearly explained

No abstract

No abstract