Multi-year incubation experiments boost confidence in model projections of long-term soil carbon dynamics

Global soil organic carbon (SOC) stocks may decline with a warmer climate. However, model projections of changes in SOC due to climate warming depend on microbially-driven processes that are usually parameterized based on laboratory incubations. To assess how lab-scale incubation datasets inform model projections over decades, we optimized five microbially-relevant parameters in the Microbial-ENzyme Decomposition (MEND) model using 16 short-term glucose (6-day), 16 short-term cellulose (30-day) and 16 long-term cellulose (729-day) incubation datasets with soils from forests and grasslands across contrasting soil types. Our analysis identified consistently higher parameter estimates given the short-term versus long-term datasets. Implementing the short-term and long-term parameters, respectively, resulted in SOC loss (-8.25.1% or -3.9 +/- 2.8%), and minor SOC gain (1.8 +/- 1.0%) in response to 5 degrees C warming, while only the latter is consistent with a meta-analysis of 149 field warming observations (1.6 +/- 4.0%). Comparing multiple subsets of cellulose incubations (i.e., 6, 30, 90, 180, 360, 480 and 729-day) revealed comparable projections to the observed long-term SOC changes under warming only on 480- and 729-day. Integrating multi-year datasets of soil incubations (e.g., > 1.5 years) with microbial models can thus achieve more reasonable parameterization of key microbial processes and subsequently boost the accuracy and confidence of long-term SOC projections. As the climate warms, soil carbon stores will likely be degraded by microbes and released as CO2, but these predictions are based on laboratory incubations that might not reflect real rates. Here the authors optimize model projections using dozens of short- and long-term incubations in forest and grasslands.