How do carbon, nitrogen, and phosphorus cycles regulate climate system feedbacks, and how sensitive are these feedbacks to model structural uncertainty?

This experiment has dual objectives: first, examining how more complete treatments of nutrient cycles affect carbon–climate system feedbacks, with a focus on tropical systems; and second, investigating the influence of alternative model structures for below-ground reaction networks on global-scale biogeochemistry–climate feedbacks. The focus on nutrients will center on the addition of phosphorus (P) to existing models of soil biogeochemistry, testing the hypothesis that P availability limits tropical ecosystem production and plays an important role in regulating global-scale feedbacks connecting CO2 concentration, temperature, and the hydrologic cycle. The P cycle is introduced on top of existing vertically resolved carbon-nitrogen (C-N) biogeochemistry within the CLM 4.5 framework.

Our near-term experiments will be run using the P cycle in conjunction with two independently developed C-N reaction networks: the converging trophic cascade (CTC) model (Thornton et al., 2007) and the CENTURY model (Parton et al., 1987). These two reaction networks are accessible within CLM 4.5 via a single compile-time switch. By evaluating the introduction of P dynamics in both networks, we will assess the influence of model structural uncertainty on the sign and magnitude of biogeochemistry–climate feedbacks in the coupled global system. Since the success of the experiment does not critically depend on ultrahigh resolution, these simulations can commence using current well-tested, and considerably less expensive, model grids.

The structural uncertainty dimension of this experiment (testing alternative below-ground reaction networks) is expected to have important consequences across the globe. Of particular interest, as we look toward the much more mechanistically complete representation of biogeochemistry processes that characterizes our 10-year vision, is how these structural differences manifest in high-latitude systems. We will use the experiment described here to explore the influence of choice of C-N reaction network on thermal hydrology and carbon fluxes in Arctic tundra and other high-latitude systems.

A possible experimental protocol consists of four pairs of simulations, exercising the fully coupled climate and biogeochemistry system, using the low-resolution model configuration (section 2.1) with modifications as described here. The two members of each pair are conducted with alternate representations of the belowground reaction networks (CTC versus CENTURY), enabling a first-order resolution of the influence of model structural uncertainty. The four simulation pairs are as follows:

  1. A pair of fixed-forcing control simulations, using preindustrial (circa 1850 AD) boundary conditions. Those boundary conditions include land cover, aerosols, nitrogen deposition over land and ocean, and atmospheric concentrations of non-CO2 greenhouse gases. This pair of simulations uses prognostic land and ocean biogeochemistry as implemented in the CMIP6 protocols, including a dynamic nitrogen cycle, but not including phosphorus dynamics in the land model. Atmospheric CO2 concentration is prognostic in this simulation, but is expected to remain quite stable, based on the results of a biogeochemistry spin-up procedure (described below). Duration of this experiment is a minimum of 250 model-years. If small drifts in climate and/or CO2 concentration remain after spinup, this simulation will be used as a reference case for the removal of long-term residual trends from the following transient-forcing simulations.
  2. A pair of transient-forcing control simulations, using historical forcings for the period 1850–2004, and a high radiative forcing scenario (RCP8.5, from the CMIP6 protocol) for the period 2005–2100. Transient forcings include fossil fuel and industrial emissions of CO2, land use and land cover change, atmospheric nitrogen deposition over land and oceans, aerosols, and non-CO2 greenhouse gas concentrations. This pair of simulations uses the same configurations of below-ground reaction networks as the fixed-forcing controls, with a dynamic nitrogen cycle on land but no representation of phosphorus dynamics.
  3. A pair of fixed-forcing C-N-P simulations, configured exactly as the fixed-forcing controls, but replacing the standard carbon-nitrogen reaction networks with new carbon-nitrogen-phosphorus (CN-P) networks, implemented in each of the reaction networks (CTC and CENTURY). This simulation requires a spun-up state, which is independent of the spin-up used for the fixed-forcing control. The spin-up procedures are discussed below.
  4. A pair of transient-forcing C-N-P simulations, configured exactly as the transient-forcing control experiments, but replacing the C-N networks with their C-N-P equivalents.

The ACME team has at its disposal spun-up states for fully coupled climate–biogeochemistry simulations using C-N dynamics in the CTC reaction network. Since some model components are slated for rapid modification in the initial six months of the project, it is likely that additional model spin-up time will be required to meet typical steady-state metrics. Spun-up offline simulations (land driven by data atmosphere) are available for the C-N dynamics in the CENTURY reaction network, and we will commence a standard of partially coupled simulations to equilibrate the biogeochemistry with this configuration. We expect these spin-up simulations to be ready within six months of project launch, allowing execution of the fixed-forcing control simulations in Q3/Q4 of the first project year.

Spun-up offline C-N-P simulations in the CTC reaction network are available today, and these will be used to initiate the fully coupled spin-up sequence immediately upon project launch. C-N-P integration within the CENTURY reaction network is not yet underway; this is the highest-priority model development task, slated for completion in Q2 of the first project year. Fixed-forcing and transient-forcing C-N-P simulations can then initiate in Year 1 Q3.

Our experimental results will be made more robust if we can carry out ensembles for the transient-forcing experiments. We have set a target minimum of four ensemble members for each of these simulations. The fixed-forcing controls do not require ensemble members, but could be extended in length if long-period variation emerges.