2024 Abstracts

A Framework for Evaluating Process Uncertainty Among Soil Carbon Models


Matthew E. Craig1* (craigme@ornl.gov), Benjamin N. Sulman1, Melanie A. Mayes1, Rose Z. Abramoff2,  Lianhong Gu1, Jeff D. Wood2, Jana R Phillips1, Anthony P. Walker1, Paul J. Hanson1, Daniel Ricciuto1


1Environmental Sciences Division and Climate Change Science Institute, Oak Ridge National Laboratory, Oak Ridge, TN; 2Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA



Microbial and mineral interactions mediate soil carbon (C) responses to environmental change. These interactions are numerous and often result in complicated or counterintuitive soil responses. The newest generation of soil C models attempt to predict these responses by explicitly representing some microbial and mineral processes. However, these models make vastly different predictions. This model uncertainty is driven largely by process-knowledge uncertainty. That is, different models represent different combinations of hypotheses about the processes governing soil C dynamics. Reducing this process uncertainty and homing in on optimal process representation in soil C models requires: (1) identification of key processes that drive divergence among models; and (2) empirical efforts targeted toward a better understanding of these hypotheses. Here, researchers discuss the progress on developing and applying a multi-assumption soil C model within the multi-assumption architecture and testbed (MAAT) to evaluate the factors that underlie uncertainty among several microbially explicit soil C models. MAAT is a modular modeling code that can easily vary model process representations, with built in tools for model calibration and process- and parameter-level sensitivity analyses. Researchers have implemented several soil C models in MAAT and found that these models simulate markedly different responses of soil carbon to alterations in inputs, temperature, and moisture. Alternative hypotheses about mineral-associated C turnover and microbial biomass dynamics account for a large portion of this uncertainty in response to altered soil C inputs. Researchers will discuss progress in calibrating models against a common dataset to control for differences in model parameterization that can confound the comparison of process representation among models.

Future work will focus on implementing nutrients into the multi-assumption soil C model and combining model development and optimization with forthcoming experimental incubations and field datasets.