I’ve read the preprint for "Historical variation in normalized difference vegetation index compared with soil moisture at a taiga forest ecosystem in northeastern Siberia” with great interest and first of all I would like to congratulate the authors with their valuable paper.

The authors present extensive records of various remote sensing, biotic and abiotic field monitoring datasets from a long-term ecological monitoring station in the Siberian Arctic. Continuous datasets such as the one presented are crucial to understand the role of the vast and carbon rich Siberian permafrost ecosystems in the context of global change and I am very happy to see strong international collaboration on this topic continue.

I found the study well written and informative. The combined use of ndvi, rwi, abiotic monitoring and foliar properties provides detailed insight into what mechanisms are more or less likely to change in response to extreme wet events. The trend breaks in association between ndvi/rwi and swe after the 2007 wet events are very striking and well-illustrated in figures 3 – 5.

I have several points for improvement that I think will need to be addressed before publication. Below I will list several main comments, and a list of smaller line comments. If the authors revise these issues I would recommend the study for publication in Biogeosciences.

• In the introduction, discussion and particularly the conclusion, the authors mostly discuss earlier findings from Spasskaya Pad, and hardly touch upon potential similarities or dissimilarities with other regions. This makes it very hard for the reader to assess to what extent the findings presented here may hold lessons for the other boreal forests on permafrost. In my view, your results hold important lessons for the potential impacts of increased precipitation variability in northern forests, also beyond Siberian larch forests! Precipitation variability is increasing rapidly in this region (see also https://doi.org/10.1016/j.jhydrol.2021.126865) so it is important to discuss what your findings imply for the future functioning of Siberian larch forests and potentially boreal forests in general. You also demonstrate a clear “legacy effect” that could be related to recent insights regarding duration of the impacts of extremes (see for instance https://onlinelibrary.wiley.com/doi/pdfdirect/10.1111/gcb.16078). You still find divergence in NDVI over ten years after an extreme event. This is a major legacy effect, that has important implications for knowledge on Arctic greening/browning and should be stressed more strongly in the conclusion and abstract!
  
  
  
  
• The described aim of the research is to assess how the local forest has changed over time, but throughout the methods you have decided a priori to split up the data into a pre-2007 and post-2007 segment based on an extreme event. Hence, it seems more appropriate to either first statistically evaluate and demonstrate whether there is a significant trend break. I do not doubt this would be the case if you would try it, but it would provide a back up for your methodological choice. Alternatively (maybe this is easier) you could reframe the research aim to explicitly investigate the effect of this wet event. This would make sense, since the subdivision of forest types within the transect seems to already be based on forest damage and regeneration stadia, and the introduction already extensively discussed observed effects of the 2007 wet period.
  
  
  
  
• The ecophysiological meaning of the d15N, d13C and C/N ratio data, as well as the methods through which they were derived, are completely lacking. The reader will need more background to understand the presented patterns and the methods are not reproducible here. 
  
  
  
  
• I have some concerns about confounding effects of seasonal availability of landsat ndvi data in shaping the temporal dynamics of ndvi and affecting relationships with other site data. In the line comments, I have added some examples and suggestions on how to deal with this. I think with an additional figure or potentially addition of covariates/interactions such issues could be resolved quite well.
  
  
  
  

~~~ Line comments ~~~



Abstract:

L. 29: Could you reflect briefly on the implications of your results to place them in a wider context? Parts of the Siberian Arctic show record browning in recent decades, as you undoubtedly know better than anyone. Perhaps you could reflect on the potential role of moisture dynamics, drought and waterlogging in this browning trend? (Just a suggestion). 



Introduction:

L. 31-32 "occupy a large forest area, approximately 27 % (Fao, 2020)" --> I assume you mean 27% of the world's forest cover? Could you rewrite this to make it clearer what the statistic refers to? Also consider writing "FAO" instead of "Fao" as you also write it in the reference list. 

L. 39 "and change the ecosystem" --> Could you provide a few concise examples and references?

L. 31 - 66: Please consider adding some thematic structuring to the introduction; the introduction seems to give an overview of earlier work that is mostly focused on C-exchange, while the knowledge gap decsribed on L. 65-66 focuses on NDVI and foliar parameters.

L. 67 - 70: The research aim is described as "assessing how the forest has changed", which seems unnecessarily vague. Could you provide more specific aims or research questions and (optionally) hypotheses? Setting more specific aims may also help provide structure and direction to the introduction paragraph above. 

Materials & Methods:

L. 78: "consists of deciduous species" --> any information which ones? do they occupy a significant share of the canopy compared to dominant larch vegetation?

L. 80 " and other grasses" --> please remove "other" (as the shrubs mentioned before are not grasses)

L. 95: "Regenerating forests RF-2 had moderate forest conditions between RF-1 and DF" --> what do you mean by moderate forest conditions?

L. 108 - 110: " The transect plots, which consist of pixels not attributed to quality pixels (clear terrain, low-confidence cloud, and low-confidence cirrus) in the quality assessment bit index band according to Landsat Surface Reflectance product guides, were excluded from the analysis.  --> due to the structure of this sentence it reads to me as though all transect plots ndvi values were excluded from analysis, but as the text continuous you describe how it was used in further analysis, so I assume you only removed pixels ( or transect plots?) that were flagged in the QA product? Perhaps you could rephrase this more clearly (e.g. that "pixels flagged in the quality assessment bands were omitted from analysis"? or that "transect plots that contained pixels flagged in the quality assessment bands were omitted from analysis"?). 

L. 120: can you provide an assessment of fit among the different sensors, e.g. on days for which multiple products are available? how accurate is the estimate for the one sensor based on another sensor compared to the actual value? Roy et al 2016 recommend to use a locally parameterized regression, although it would be understandable if insufficient overlap in acquisitions among different sensors prevents establishment of specific regression parameters for your site. 

L. 133-136: this paragraph lacks context of the ecological or physiological meaning of isotope ratios and C/N ratio. More explanation and literature is needed for the non-expert reader to assess what the d15N, d13C, C/N ratios and ring widths actually mean and what questions you are answering by including these data (alternatively, you could also already explain how the different types of datasets relate to the research aims in the final introduction paragraph)

L. 133-136: There seems to be no explanation of how the d15N, d13C and C/N ratios were derived, Add methodology (which tissues were sampled, how many grams, how were they analyzed, on which instrument, against which isotope standards at what precision?). If the data come from an existing dataset or study, please cite it so the reader can understand how the values were derived. 

L. 150: can you explain why you chose a pearson correlation, rather than a spearman correlation or crosscorrelation function (which in my experience are more appropriate choices for relatively short timeseries)? Not that I doubt the outcome of your analysis (you present very clear visual and temporal patterns), but the backing of your choices could be stronger. 

L. 152: "differences between the two groups" --> which groups are you referring to? there are more than two types of forests mentioned in earlier in the methods. It is also unclear to me why an unpaired test was selected if data from the same years or acquisitions is available for different forests. I am probably misunderstanding what you are describing here, so perhaps that is an indication that better explanaiton is needed.

Results:

L. 163-165: "The seasonal maximum of each year was observed from 25 June to 13 August, except for 1999 (shown in Table S2). The maximum transect NDVI in 1999 was observed on 27 August (0.75 ± 0.02, n = 34) because the Landsat data in 1999 were limited to the latter half of August. " --> landsat scene availability throughout the summer can be highly limited. to what extent is the seasonal maximum an artfeact of data availability (e.g. it would obviously fall in June if no data from July and August are available, even if the true maximum would fall in july or august). Please add an indication or statistical backing (maybe in SI) of how the timing of the seasonal maximum relates to scene availability, because otherwise it cannot be called "year to year variation" and it would be unclear whether the time series you describe in fig. 2a is robust, or merely an artefact of seasonal timing. 

L. 191 - 192: "To consider the historical variation in the NDVI of typical forests in our study area, the TF NDVI and observed parameters were compared (Fig. 2 and 3)." --> I would strongly urge you to account for landsat scene availability throughout the season, for instance by adding the date within the season as a covariate or interaction. This would give additional information of the association with other parameters may vary across the season and would account for the possibility that the temporal dynamics of ndvi are influenced more by scene availabiltiy than annual dynamics in site conditions. 

L. 197: "TF NDVI did not show any correlation with summer temperature" --> you present correlations of NDVI values at different seasonal timings (june / july / august) to overall JJA temperatures. wouldn't it make more sense to compare the ndvi to mean temperatures of degree days up until the moment of ndvi acquisition?

L. 218: the header of the next section accidentally ended up in the figure caption here.

Discussion

L. 275: "In most years before 2007, the NDVI values in RF and DF were higher than those in TF" --> could this be related to topgraphy; i.e. DF and RF are damaged by floods since they occur in depressions and hence suffer less from drought but more from flooding? the role of terrain is hardly touched upon but potentially very important. It might also be helpful to present some indication of terrain variability; what is the magnitude of elevation differences between typical DF and Tf sites, for example? 

paragraph 4.1: Please discuss whether waterlogging may have influence ndvi directly, independent from tree properties, due to its influence on near infrared reflectance. 

L. 312 - 317: I know it is very likely the case, but here you seem to derive causation from the presented correlations. Tone down these causal statements (e.g. "which likely contributed" instead of "which contributed"), or provide more backing for why carbon storage in previous years should be the cause of NDVI dynamics in this period. 

L. 327 - 328: "The mechanism by which plant δ13 C responds to changes in light and water availability has been well explained in previous studies (e.g., Farquhar et al., 1989). " --> I don't doubt it, but it is very difficult to place your findings on isotope ratios in the appropriate context without some minimum amount of explanation of their meaning and key processes driving isotope fractionation in trees. Please add this (or see comments regarding lines 133-136) at some point so the reader can understand the meaning of the presented work on isotope and c/n ratios to some degree without having to refer to cited work. 

L. 329: "Under drought stress during 2001–2002, there was a decrease in needle stomatal conductance" --> this is another example of a conclusive statement that does not seem to be backed up by data or a reference. Please check the entire discussion for statements like these and either back them up or tone them down ("has likely decreased stomatal conductance, as suggested by d13C values")

L. 354 - 346: "Therefore, the decrease in the TF NDVI in wet years may be due to factors other than the carbon assimilation process" --> here you should probably discuss the direct influence of water on near infrared reflectance and ndvi. 

L. 400 - 401: "However, the TF NDVI and RWI were not significantly correlated after 2007, whereas there was a significant positive correlation before 2007. " --> please consider alternative explanations. For instance, the use of detrending methods in tree ring width series can remove long-term decreases or increases from the time series, and your RWI likley only reflects year-to-year variation in ring width. In this sense, do you think the RWI series reflect any long-term decreases due to for instance waterlogging events and comprimised growth over longer timescales?

L. 432-434: "To better understand changes in the forest, long-term observation of variations in soil N availability depending on soil moisture and other factors is necessary" --> Perhaps we would also need better understanding and forecasting of precipitation extremes or weather extremes in general?

Conclusion

L. 435-452: In general , I think the conclusion presents some statements that rely on interpretation quite a lot, and presents a lot of statements that are merely repetition of the results. I do not disagree with your interpretations (I think they are well found), but it should be clear for the reader which statements are interpretations and which are not (e.g. by adding "which we attribute to .."). Also see my main comment; the conclusion does not go beyond the distinct physiological response observed in this ecosystem and does not discuss implications. To be of value to a wide readership, please try to "zoom out" a bit beyond Spasskaya Pad. Maybe mention and discuss the importance of findings such as the long-term alteration of relationships between moisture availability and tree performance, or provide recommendations for future studies. 

Tables and figures

Table 1: The added value of this table relative to the clear patterns in fig 2b, are unclear to me. I also find it unclear why only TF and Rf1 are presented. Due to nestedness (transect plots within years within groups), the p-values should be corrected for pseudoreplication. A visual overview might be stronger here and you could consider replacing or omitting this table. 

Figures 4 & 5: "p-values and R2 describe the significance and the degree of variability of the regression  models, respectively" --> degree of variability is probably not the appropriate term here, I assume this is a coefficient of determination?

SI tables S4-S5: How reliable are the p values derived for differences among degraded forest and other forest types, if there were only two transect plots with data for degraded forests? I also find it hard to understand why the others use pairwise tests rather than anova/kruskal-wallis tests with post-hoc tests? Throughout the supporting tables S4-S10, you perform very large amounts of t-test and if you want to use these values to support your findings, you should discuss the role of Type I errors. 

In "Historical variation in normalized difference vegetation index compared with soil moisture at a taiga forest ecosystem in northeastern Siberia” the authors investigated the variation in NDVI among forest conditions ( typical mature, TF; regenerating-1, RF-1;  regenerating-2, RF-2; and damaged forests, DF) and field-observed parameters (from 1998 to 2019) such as RWI, soil moisture, changes of larch needles (δ13C, δ15N, C/N), air temperature, and precipitation. The authors determined that prior to the 2007 extreme wet event, wet areas like DF and RF had higher NDVI values than dry TF sites due to greater water availability. However, following 2007, the TF had a greater NDVI than the DF and RF, although being visibly unaffected by the wet event.

Studying historical variations in NDVI compared with soil moisture at a taiga forest ecosystem in north-eastern Siberia is important for several reasons. Firstly, NDVI data can provide valuable information about temporal and spatial changes in vegetation distribution, productivity, and dynamics, which allows for the monitoring of habitat degradation and fragmentation. Secondly, the comparison of historical variations in NDVI with soil moisture can provide insights into the impact of extreme weather events on vegetation, such as the extreme wet event in 2007, which resulted in high tree mortality and a decrease in NDVI at affected sites. Understanding the ecological effects of climatic disasters such as drought or fire can be assessed using NDVI data, making it a valuable tool for monitoring changes in vegetation due to climate change. Overall, studying historical variations in NDVI and soil moisture in a taiga forest ecosystem can provide valuable insights into the impact of extreme weather events on vegetation and the effects of climate change on vegetation dynamics. Therefore, this paper has the potential to make an important contribution to the body of knowledge concerning the impacts of global change on sensitive and complex permafrost ecosystems.

It is my opinion that the authors used sound methods to address the study aims and presented the research findings clearly and concisely and they used appropriate figures to illustrate the NDVI values of the forest types and the trends in the transect and 10-km plot, which could be useful for researchers and policymakers. However, I agree with referee 1 about their main points raised as well as the minor comments provided. To avoid repetition and in the interest of brevity, I will not be going over them again in this review, but I strongly advise the authors to make the corrections already suggested. Instead, I will just add a few points concerning the discussion section that I would like to see addressed before publication. When the authors revise these issues, I recommend the study for publication in Biogeosciences.

In the discussion, the authors considered the probable reasons for the differences in NDVI values among the forest types, such as the change in vegetation and the presence of surface water and saturated soil. However, the section could benefit from a more critical evaluation of the results and their implications. For example, the article does not address the limitations of using NDVI as a proxy for vegetation health and productivity, which could impact the accuracy of the results. NDVI measures the amount of chlorophyll in the uppermost layers of vegetation. This means that it may not accurately represent the health and productivity of plants with lower canopies or those that are hidden from view. The limitations of using NDVI as a proxy for vegetation health and productivity may be particularly relevant in taiga/permafrost ecosystems due to their complex vegetation structure and sensitivity to environmental changes.

Additionally, the article does not explore the broader ecological implications of these findings, such as how changes in vegetation health and productivity may impact ecosystem services or the ability of forests to sequester carbon. Finally, while the article notes the potential for using the observational data for analyses of ecosystem changes at the plot and regional scales, it does not explicitly state what these analyses might entail or why they would be valuable. A more explicit discussion of the practical applications of the research could make the findings more accessible to a wider audience.