The direction and magnitude of change of soil organic carbon (SOC) stocks due to global warming depend strongly on the temperature sensitivity (e.g., Q₁₀) of carbon mineralization. To date, most multi-pool SOC models assume a general Q₁₀ of 2 despite experimental evidence suggesting different Q₁₀ for different carbon fractions. The aim of this study was to test if the use of experimentally derived pool specific Q₁₀ values improves the performance of SOC models. Five contrasting data sets from three field experiments and two laboratory incubations were used to study the link between carbon pool recalcitrance and Q₁₀ using two different approaches: a) Bayesian calibration of the Daisy SOC model parameters to infer Q₁₀ of SOC and crop-litter pools, and b) using measured Q₁₀ values of carbon degrading enzymes as proxies for Q₁₀ of different Daisy pools. Namely β-glucosidase (median Q₁₀ of 1.82) was assigned to metabolic litter and phenol/peroxidase (1.35) to structural litter and both SOC pools. To partition litter-carbon and SOC into model pools, the lignin-to-nitrogen ratio and the ratio of aliphatic/aromatic-carboxylate carbon were used, respectively. Measurements included soil microbial biomass, soil carbon dioxide (CO₂) evolution and remaining carbon in soils and crop-litter. In the Bayesian calibration, strong differences in inferred Q₁₀ values of the same pools between experiments suggested that intrinsic substrate recalcitrance was not the main driver of temperature sensitivity. For field experiment simulations, both the Q₁₀ values derived by Bayesian calibration and measured enzyme Q₁₀ were centered around values below 2, contrasting with high Q₁₀ values for mineralization under laboratory incubations (close to 3). Furthermore, assigning measured phenol/peroxidase Q₁₀ values to the slow crop-litter as well as both SOC pools and β-glucosidase to the fast crop-litter pool (approach b), could significantly improve model performance compared to using the default Q₁₀ value of 2 for all pools. Root-mean-squared-deviation reductions were between 3 and 10% for field experiments, with no change in the laboratory experiments. Thus, site specific Q₁₀ values of soil enzymes show potential as proxies for pool specific Q₁₀. We present a new conceptual framework to explain the observed differences in temperature sensitivities between experiments as a result of two fundamental driving factors classified in a) state variables, that fluctuate in time, and b) soil properties, that are constant over decades. Measured enzyme Q₁₀ values were interpreted as a proxy incorporating both factors. More than intrinsic substrate recalcitrance, the state variables such as physical protection, substrate abundance and unfavorable conditions for microorganisms control temperature sensitivity of mineralization. To reduce the uncertainty in global SOC simulations under a changing climate, their relative contributions should be disentangled and then implemented into SOC models.

由于全球变暖，土壤有机碳（SOC）储量变化的方向和幅度在很大程度上取决于碳矿化的温度敏感性（例如Q ₁₀）。迄今为止，尽管有实验证据表明不同碳含量的Q ₁₀不同，但大多数多池SOC模型仍假定Q ₁₀为2 。这项研究的目的是测试使用实验得出的池特定Q ₁₀值是否可以改善SOC模型的性能。来自三个野外实验和两个实验室培养的五个对比数据集用于研究碳库顽固性与Q ₁₀之间的联系使用两种不同的方法：a）将菊花SOC模型参数的贝叶斯校准来推断Q ₁₀的SOC的和作物垃圾池，和b）使用测得的Q ₁₀碳降解酶的值作为对代理Q ₁₀不同雏菊池。将β-葡萄糖苷酶（Q ₁₀中位数为1.82）分配给代谢垃圾，将酚/过氧化物酶（1.35）分配给结构垃圾和两个SOC库。要将枯枝落叶碳和SOC划分为模型库，分别使用了木质素与氮的比率以及脂肪族/芳香族羧酸盐碳的比率。测量包括土壤微生物生物量，土壤二​​氧化碳（CO ₂）的演变以及土壤和农作物凋落物中的残留碳。在贝叶斯校准中，实验之间相同库的Q ₁₀推断值存在很大差异，这表明固有的底物顽固性不是温度敏感性的主要驱动因素。对于现场实验模拟，通过贝叶斯校准得出的Q ₁₀值和测得的酶Q ₁₀都以低于2的值为中心，而在实验室温育下矿化的Q ₁₀值较高（接近3）。此外，分配测得的苯酚/过氧化物酶Q ₁₀与将所有作物的默认Q ₁₀值设为2相比，慢速作物凋落物以及SOC池的值和快速作物凋落物池的β-葡萄糖苷酶（方法b）的值都可以显着改善模型性能。现场实验的均方根偏差减少了3％至10％，实验室实验没有变化。因此，土壤酶的位点特异性Q ₁₀值显示出潜在的池特异性Q ₁₀代理。我们提供了一个新的概念框架，以解释由于两个基本驱动因素而在实验之间观察到的温度敏感性差异。这些驱动因素归类为a）状态变量（随时间波动）和b）土壤特性，这些因素在数十年内保持不变。测得的酶Q ₁₀值被解释为包含两个因素的代理。除了固有的底物顽固性外，状态变量（例如物理保护，底物丰度和微生物的不利条件）还控制着矿化的温度敏感性。为了减少气候变化情况下全球SOC模拟的不确定性，应将它们的相对贡献分解开，然后将其应用于SOC模型。