Linking temperature sensitivities of soil enzymes to temperature responses of different organic matter pools in the DAISY model

Soil organic carbon (SOC) losses under a changing climate are driven by the temperature sensitivity of SOC mineralization (usually expressed as Q₁₀, the multiplier of activity with 10 °C temperature increase). The activation energy theory (AET) suggests that, due to higher activation energies, the more complex the carbon, the higher is mineralization Q₁₀. However, studies on Q₁₀ have been inconsistent with regard to AET. Measurements of potential soil enzymes activity Q₁₀ even contradicted AET: Phenoloxidase (representing complex carbon) had consistently lower Q₁₀ than the more labile xylanase and glucosidase. This study used two approaches of examining Q₁₀ in SOC modeling: 1) Bayesian calibration (BC) and 2) using different measured enzyme Q₁₀ as proxies for mineralization Q₁₀ of different SOC pools. The SOC model was DAISY (S. Hansen et al., 2012). BC informed Q₁₀ by field measured data, while the second approach tested if directly using enzyme Q₁₀ (of phenoloxidase, glucosidase and xylanase) for DAISY pools improved simulation results. Both approaches used the temperature sensitive measurements of CO₂ evolution and soil microbial biomass. The measured enzyme Q₁₀ were from field manipulation experiments with bare fallow and vegetated plots in the two regions of Kraichgau and Swabian Jura in Southwest Germany. The enzyme-derived Q₁₀ were used for modelling those fields and furthermore for in‑situ litterbag decomposition experiments at 20 sites in the same region. Two further laboratory experiments with temperature manipulation were included: an incubation of the field residues into soil and an incubation of bare soil from the start and year 50 of a long duration bare fallow (from Ultuna). The BC made use of CO₂ and microbial data to inform about the range of Q₁₀ of different carbon pools for the individual experiments and combined data.

The BC of the residue incubation experiment constrained Q₁₀ for metabolic (~3) and structural litter (~2). Estimated 95% credibility intervals did not overlap. The BC for Ultuna could constrain the slow and fast SOC pool with Q₁₀ ~2.8 and ~3, respectively, but credibility intervals of both pools overlapped. The Q₁₀ of field experiments, which had most abundant data, could not be constrained by BC, probably because their annual temparature variability was too low. However, the model errors of the field experiment could be reduced by the second approach, when the Q₁₀ of phenoloxidase was used for to the structural litter pool as well as for the fast and slow SOC pools. Thus regional enzyme Q₁₀ improved the model fit but only for regional simulations. Therefore, they could be useful proxies when natural temperature range is too small to inform temperature sensitivity by BC. Any trends found in this study contradicted AET, both from measured enzymes and BC of the incubation experiments. This calls for alternative Q₁₀ hypotheses and the need for individual Q₁₀ values for different SOC pool rather than a general one. BC approaches would benefit from a wider temperature range of field experiments and understanding what causes variable enzyme Q₁₀ could help to improve future SOC models.