Detailed comments are found in attached pdf.

The manuscript presents an updated environmental modifier for the soil organic matter decay function within Yasso07 model. The authors use the field measurements of soil organic carbon (SOC) and soil heterotrophic respiration in an upland-peatland complex in southern Finland to calibrate the environmental modifier function, which simulates the effects of soil temperature and moisture on the organic matter decay. The study is timely and well-suited for publication in the GMD following authors’ addressing the comments and suggestions outlined below.

It was somewhat surprising to see the parameters in Tables 1 and 2 not aligning very well. Q10’s estimated with the data assimilation using SOC and SOC+CO2 flux were much higher compared with those estimated using NLS approach using the heterotrophic CO2 flux alone and SWCopt were much lower. Authors attributed the difference to different basal respiration rates in Yasso07 and NLS, however I would argue that Q10 and SWCopt control the curvature the respiration’s curve, and basal respiration rate, being a scaling parameter, should not affect the values of Q10 and SWC opt in such a profound way. I think such a difference in these parameter values may be attributed to the weighing of the observations in the data assimilation approach: SOC stocks appear to have a much larger influence on the posterior parameters compared to the CO2 flux. The posterior values of parameters a and b from the equation 6 were not reported (I assume they were estimated, because there are prior values reported in the Table S1), so it was impossible for me to evaluate whether that may have been the case. I would suggest doing one more calibration experiment to explore whether the data weighing is an issue and calibrate the Yasso07 with CO2 observations alone. If the parameters are similar to the NLS, the weighing of the observations is likely the culprit. If it is, I would suggest weighing the observations by their individual errors: the larger is the error associated with the observation’s mean value, the lesser weight it should be attributed within the calibration algorithm. This way the algorithm would not “hack” itself to produce the smallest error, but rather be forced to gain information from the more precise observations.

It was not clear from the methods whether the data used for model validation were the same as the data used for model calibration. If the data were the same, I would suggest re-doing the calibration with less observations and reserving a portion of the observations for model validation. If the two observation sets are already separated, please include this information in the methods section.

The units of respiration are in g CO2, and within the model the units associated with C transfers are in g C, were the units of respiration converted to gC before calibration?

Below is the list of the minor comments and suggestions:

L23-24: “ “…calibrated against SOC and CO2 data using Bayesian MCMC approach showed …..”

L70: “…in underestimation…”

L55: what metal are the collars made of? Can it affect respiration rate?

L66-67: please include the depth increments

L176: “Breast height…”

L214: if inputs and pools are functions of time, I suggest adding (t) next to the vector elements

L216-218: I suggest revision of this statement. it's a product of a column vector by a row vector C(t), where the elements of the column vector are the fractions that were not transferred among the pools.

L240: the focus of this publication is different, I think a more appropriate reference is this one: https://doi.org/10.5194/gmd-5-1259-2012

L278: N instead of n?

L440-443: this statement does not align with the results of the NLS regression of the respiration data performed in this study

L283: “the mean volumetric…”

Figure 4: please include legend for colors