Controls of soil organic matter on soil thermal dynamics in the northern high latitudes

Permafrost warming and potential soil carbon (SOC) release after thawing may amplify climate change, yet model estimates of present-day and future permafrost extent vary widely, partly due to uncertainties in simulated soil temperature. Here, we derive thermal diffusivity, a key parameter in the soil thermal regime, from depth-specific measurements of monthly soil temperature at about 200 sites in the high latitude regions. We find that, among the tested soil properties including SOC, soil texture, bulk density, and soil moisture, SOC is the dominant factor controlling the variability of diffusivity among sites. Analysis of the CMIP5 model outputs reveals that the parameterization of thermal diffusivity drives the differences in simulated present-day permafrost extent among these models. The strong SOC-thermics coupling is crucial for projecting future permafrost dynamics, since the response of soil temperature and permafrost area to a rising air temperature would be impacted by potential changes in SOC.

The presence of soil organic carbon (SOC) reduces thermal diffusivity (D), a key parameter in the soil thermal regime, and consequently cools the deeper soils during summer. In this study, the authors derive D from depth-specific measurements of monthly soil temperature at about 200 sites in the high latitude regions. They find that, among the tested soil properties including SOC, soil texture, bulk density, and soil moisture, SOC is the dominant factor controlling the variability of D among sites. Analysis of the CMIP5 model outputs reveals that the parameterization of D dominates the large-scale performance of these models in simulating permafrost extent. I have some questions, but most are minor: 1. In my opinon, the title seems to be too vast and should be given some necessary qualifiers. There is a very important concept in geography, i.e., temporal and spacial scale. The carbon in permafrost deposits mostly began to accumulate in the late Pleistocene except that northern peatlands developed mostly in the last deglaciation, especially during the Holocene. And the manuscript only pay attention to the circumarctic, not including the Tibetan Plateau.
Indeed, permafrost soil carbon has been accumulated at long time scales, during which the climate, vegetation, and hydrologic and sedimentary environment have varied. We agree that the previous title, 'controls…on permafrost carbon dynamics', is not appropriate. And it's also true that we did not include permafrost at Tibetan Plateau, meanwhile, some cold sites outside today's permafrost zones were included. We therefore revise the title as "Controls of soil organic matter on soil thermal dynamics in the northern high latitudes". Fig. 2(a)? Why is the regression coefficient bigger than 1.0? What's the meaning/implication of this? Please explain the reasons.

What's the linear regression equation in
In Fig. 2a, the x-axis reflects bias in the simulated present-day air temperature by CMIP5 models, while y-axis shows bias in modelled permafrost areas, reflecting bias in soil temperature. The significant correlation in Fig. 2a shows that, as expected, any bias in air temperature will propagate into soil temperature in the models. The slope being higher than 1 simply means that, among the CMIP5 models, those who have a cold bias in air temperature tend to simulate a larger A p /A MAAT<0 . This is probably a pattern by chance, because A p /A MAAT<0 is largely determined by soil thermal diffusivity in these models (Fig. 2b), a feature in the land surface component of the CMIP5 Earth system models, whereas bias in air temperature is a feature in the atmospheric component. The slope does not seem to have a meaningful implication, so we did not give the regression equation in Fig. 2a. Anyhow, we have now provided the data underlying Fig. 2 so that readers could easily derive the regression equation.
3. The form of the fitted linear equations needs to be consistent in figures, with significant levels given.
We have now added the p-values in figures that contain a correlation analysis.
"The form of the fitted linear equations needs to be consistent" -we suppose this refers to the regression equation in Fig. 2a, in which case, please see our response above. We checked that in other figures, the fitted equations are given in figure captions in a consistent form.
4. Why is thermal diffusivity given in form of log10 but not of real values in figures? THe real values are more readable for readers.
We revise Fig.1 and Fig.2b to display the original D values at a log-axis, as shown below. Note that the regression equations are still derived after log-transform, as suggested by a previous reviewer in the first round, due to the log-normal distribution of the original D values.