Temporal and spatial variation in modeled carbon-climate feedbacks: implications for future global land carbon source-sink dynamics
We used the Energy Exascale Earth System Model (E3SM), in a coupled physics and biogeochemistry configuration with dynamic carbon and nutrient cycles in the land model component, to explore the historical and future trends in carbon-climate feedbacks. A series of five simulations was performed, spanning the period 1850-2014 using historical forcings for atmospheric CO2 concentrations, land use and land cover change, and atmospheric nitrogen deposition, and then extending from 2015-2100 using the SSP585 scenario for future forcings. The five simulations include a control with all forcings held constant at 1850 levels, and four transient simulations with biogeochemical and radiative effects of increasing CO2 either held constant or allowed to vary. From these five simulations, it is possible to estimate feedback characteristics describing the response of temperature to rising CO2, the response of land carbon flux to rising CO2, and the response of land carbon flux to changing temperature. We used a moving window regression approach to estimate how these feedback characteristics changed in time, at the global scale, and how they varied in space and time. Our global scale analysis suggests that land carbon uptake in response to rising CO2 is declining now, and is expected to continue declining through 2100 under the SSP585 scenario. At the same time, land carbon loss in relation to ongoing warming under SSP585 is expected to increase. Global temperature sensitivity to increasing CO2 concentration is expected to decline as well under SSP585, mitigating somewhat the adverse effects of the transient land responses. Spatial patterns show strong changes in the land feedback characteristics over the tropics, with significant shifts as well in temperate, boreal, and Arctic regions. As seen with previous modeling exercises of this sort, land carbon-climate feedback responses for CO2 and warming are moderated by the inclusion of nutrient dynamics, in the current case encompassing both nitrogen and phosphorus cycle dynamics, when compared to carbon-only models.