Probing structural differences in contemporary soil carbon models

Link To Share This Poster: https://cdmcd.co/93Y8DY

Matthew E. Craig and Anthony P. Walker, Environmental Sciences Division and Climate Change Science Institute, Oak Ridge National Laboratory, Oak Ridge, TN, Benjamin N. Sulman, Climate Change Science Institute and Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN, Melanie A. Mayes, Oak Ridge National Laboratory, Oak Ridge, TN

Background/Question/Methods
Hundreds of models have been developed to describe the formation and decomposition of soil organic carbon (SOC), representing different assumptions about the dominant processes governing SOC dynamics. Moreover, in an effort to reflect emerging consensus from empirical studies, recent models employ more complicated representations of microbial and mineral interactions. This has led to an explosion of potential SOC model configurations. Unsurprisingly, different process representations among existing models have led to divergent predictions about SOC responses to global change. As SOC models continue to evolve, there is a critical need to probe the drivers of this process-level uncertainty. Here, we integrated contemporary SOC models (e.g. MIMICS, MEND, CORPSE) into the multi-assumption architecture and testbed (MAAT)—a modular modeling code that can easily vary model process representations. We then embedded and evaluated alternative process representations for each of these models to uncover sources of process-level uncertainty.
Results/Conclusions
Preliminary results indicate that models produce substantially different predictions about SOC responses to global change drivers (e.g. warming and increased inputs). Much of this uncertainty can be attributed to specific processes. For example, embedding 12 alternative temperature response functions resulted in warming-induced (+5°C) SOC losses ranging from 13.7 to 28.9 % in a 10-yr simulation. We find that unifying the representation of key processes brings model predictions into alignment. For example, representing microbial density-dependence in MEND leads to predictions that more closely mirror other models (e.g. MIMCS) in the context of increased carbon inputs. We will discuss how these analyses are informing the development of a multi-assumption SOC model. Such efforts will be critical in efficiently directing future empirical studies and model development.