Model driven data analysis
Soil temperature response is a major driver of shifts in soil carbon stocks in Earth system models. This was the first project to see a direct change in soil organic carbon stocks in a set of field warmed experiments. While Earth system models fell within this temperature response, they did not reflect the variability in the soil temperature response seen in the field. Implying that models underestimate the uncertainty of the shift in soil carbon stocks over the 21st century.
Post-hoc corrections to land carbon in Earth system model
Earth system models require a huge amount of personal and CPU time. We used a reduced complexity framework to apply post-hoc corrections to the land carbon calculation of the CMIP5 Earth system models. These corrections showed that nutrient (N and P) limitation on net primary productivity would likely reduce future land carbon inputs to such an extent that land carbon stocks may become a source of atmospheric carbon dioxide over the next century. In addition, the soil carbon response is inline with experimental data but likely not reflective of real world uncertainty.
Drivers of soil dynamics in Earth system models
Soil carbon dynamics in Earth system models are primarily driven by differences in inputs (net primary production), soil temperature, and parameterization of intrinsic decay rates and the soil temperature sensitivity. Model structural differences like the number of soil carbon pools and their connections, as well as other environmental sensitivities like moisture, did not drive model differences. This effective structural homogeneity is surprising, but due to the fact that soil carbon stocks in most grid cells are close to steady state, and the basic mathematical structure of these models.
Benchmarking soils in Earth system models
In general, soil carbon models match global soil carbon stock estimates and then fall off in their performance at finer scales (biome and grid-by-grid comparison). However, they do match experimental temperature response curves.