Controls on SOC across space and time: Models with different acclimation schemes make similar spatial predictions but divergent warming predictions

Researchers use spatial gradients to estimate long-term ecosystem responses to perturbations. This approach is commonly applied to soil organic carbon (SOC) stocks which change slowly but store the majority of terrestrial carbon. Climate warming may reduce SOC stocks if higher temperatures increase decomposition rates. Yet, it is uncertain how vulnerable SOC is to warming, and whether the same factors - such as organo-mineral associations, climate, or plant inputs - determine SOC stocks across space and time. In order to test the "space for time" concept, we developed two versions of the Substrate-Mineral-Microbe Soil (SuMMS) model - one with microbial temperature acclimation and one without - to analyze observed SOC stocks at 24 sites spanning a wide range of soil types and climate. Both model predictions of SOC were strongly correlated with observations (R2 > 0.9), because mineral sorption capacity was the dominant control over steady-state SOC stock as determined by a Random Forest model. However, the two model versions made fundamentally different predictions of the change in SOC following 5°C soil warming from 2016 to 2100 because the initial mean annual temperature (MAT) was the dominant control over the SOC response. The model with microbial acclimation predicted that SOC would decline 10% at all sites along the transect, while the model with no acclimation predicted large surface SOC losses at high latitude sites and SOC gains at low latitude sites where microbial exoenzymes were already at or near their temperature optimum. These simulations suggest that gradient studies cannot be used to infer site-level responses to warming, because the dominant controls on SOC at steady state (i.e., mineral sorption capacity) are different than the dominant controls on the SOC response to a warming perturbation (i.e., initial MAT, capacity for acclimation).