This study presents an analysis of CO2 mole fraction data collected at the ZOTTO measurement site in Siberia. These data are an incredible resource for the community. The authors use a variety of statistical tools and an inverse method to interpret these mole fraction data and come to some conclusions – some that agree and some that appear to be at odds with previous studies. I have noted some weaknesses in their analysis that need to be addressed before I can support this study being accepted for publication. There are a number of grammatical errors that should be addressed before publication. 

Line 52: this list is not comprehensive. These authors appear to be leaning on CUP/CRP ideas presented elsewhere but not cited properly. This reviewer notes that some of the primary references are used by Kariyathan et al, 2023. This is important because the primary references outline some of the issues raised by the present study. There is no theoretical way to remove a linear trend from any time series that describes a range of stationary and non-stationary processes – there will always be leakage between these variations irrespective of the method used to disentangle the individual signals. As a consequence of this, it is also difficult to directly split apart changes in maxima and minima associated with seasonal cycles. Although the authors have used error correlations to study some seasonal relationships. The CCGCRV method is a complicated but effective tool that involves fitting a series of harmonics but it cannot easily address non-stationary processes, which is most noticeable during rapidly varying environmental conditions that affect the amplitude and phase of the time series. This should at least be acknowledged somewhere, particularly because the authors study the influence of a heat wave and anomalous fire year.

This reviewer is more than a little concerned with their method used by the authors to calculate their linear slopes, given the large year to year variations. Values will be disproportionately influenced by outliers. More robust methods include the Seigel or Theil-Sen estimators. These methods will provide a more rigorous assessment of any observed trends, particularly given the length and noisiness of the ZOTTO time series. As a consequence of using a simple linear regression, this reviewer is wondering whether the results presented will remain statistically inconsistent with previous studies. Certainly, eyeballing some of their figures it is hard to imagine assigning any non-zero trend. Depending on what they find, using a refined method to determine trends may influence results described later about correlations between seasonal temperature anomalies. 

Re inversions: these authors will be well aware that translating changes in atmospheric mole fractions to regional CO2 fluxes is complex, which in this case involves seasonal variations in atmospheric transport from lower latitudes. The imbalance between the size and distribution of observation networks at higher and lower latitudes may also render the posterior solution problematic. Is there a noticeable improvement in the model performance in describing the CUP and CRP metrics from the prior to the posterior fluxes?

On a related note, this reviewer is not surprised by the influence of including this one site into a global inversion. Despite claims to the contrary, there are large gaps in our knowledge about the carbon cycle at high northern latitudes and across the tropics. It would be useful for this reader to understand what had to change elsewhere (via mass balance) due to this albeit small change in the regional carbon balance due to using ZOTTO data. 

Last point, re conclusions: improved flux estimates will also come from satellite data, with data collected primarily during summer months when the observing geometry is favourable but also during other months via improved regional estimates of upwind budgets. Using a higher resolution transport model may improve regional estimates using sparse ground-based data, but whether the net impact is positive is debatable.

Minor points

• Line 162: which 78 sites? Suggest they are listed somewhere.
• Line 263: 1958 to 1961? 
• Grammatical errors throughout. Worth a closer check by the authors when they revise their manuscript. 

This manuscript investigates the interannual variability of the carbon uptake and carbon release period in Central Siberia. The presented analysis is based on CO₂ observations carried out at the Zotino Tall Tower Observatory (ZOTTO) in central Siberia, during the period 2010 – 2021. This reviewer acknowledges the relevance of the ZOTTO dataset for the scientific community as it is collected in a region that is presently experiencing significant impacts of climate change.

Results based on CO₂ measurements revealed that amplitude and length of carbon release and carbon uptake period increased during the analysed period (2010 – 2021). However, data show that the growth of the amplitude of carbon release period is larger than the growth in carbon uptake period amplitude, suggesting that the enhanced carbon uptake during the growing season was offset by the autumn/winter carbon release.   

The manuscript is generally well written, and presented results are an important contribution to the knowledge on the effect of climate change on terrestrial ecosystems. However, some aspects of data analysis and related discussion should be clarified:

• Time series of the target tank is used to evaluate the quality of CO₂ mole fraction measured at ZOTTO. This reviewer noted a small jump in the time series of target tank between 2018 and 2019. The average value before the jump seems to be lower than the value for the following period. In the opinion of the reviewer, there could be a potential bias introduced in the ambient measurements. Could the authors comment on this?
• Lines 122-125. Authors are applying a despiking methodology to remove unreliable CO₂ observations but it is not clear to this reviewer the meaning of “unreliable”: is this the definition assigned to concentrations mainly affected by local sources? Moreover, this reviewer is wondering if the removal of unreliable data is impairing the capability of CO₂ mole fraction dataset to detect the effect of extreme events. Finally, this reviewer advises authors to show the percentage of removed observations.
• Lines 135-136. Has the consistency of the assumption that there are not significant changes in the curve shape of the season over the years been tested?
• Section 2.5. The choice of the threshold value of the spatial root mean square (RMS), used to determine the region of influence, should be explained. Moreover, the impact of different RMS thresholds on the extension of the region of influence should be addressed. Finally, this reviewer suggests including a description of the land cover in the region of influence. This could help readers to get an idea on the type of ecosystems embedded in this area and affecting observations collected at ZOTTO.
• Section 3.1. Authors found that there is not a significant trend in the timings of CUP, while there is a significant increase of the CUP length. How is it possible to have an increase in the CUP length when the timings (onset and termination) are not changing? Moreover, authors are stating that the heat wave in 2020 induced an early onset of CUP, but the error bar associated to the estimated CUP onset in 2020 (Figure 5) is very large, casting doubt on the author’s statement. Finally, authors are claiming for a significant jump in the CUP length in 2020 but looking at Figure 7 the jump is visible in the CUP amplitude (and CUP rate), not in the CUP length.

Minor points to be addressed are listed below:

• This reviewer advises authors to cite the data repository where they retrieved the CO₂ mole fractions measured in atmospheric stations used for both inversions s10v2022 and s10v2022+Allstations.
• Lines 120-122. Ranges of short-term and long-term cut-off values tested are different from those reported in Table B1.
• Line 205: change “later” to “layer”.
• Caption of Figure 4: change “Thoning et al. (1996)” in “Thoning et al. (1989)”.
• Line 374. Add “are” after “s10v2021+ZOT”.