Chironomid-Based Modern Summer Temperature Data Set and Inference Model for the Northwest European Part of Russia

Round 1

Reviewer 1 Report

Review of “Chironomid-based modern summer temperature data set and inference model for the Northwest European part of Russia” by Nazarova et al.

In this study, the authors aimed to produce a new regional chironomid-based inference model for quantitative palaeoclimatic reconstructions for North-western Russia. They sampled 98 lakes and identified chironomid remains to the finest taxonomic level. Overall, results showed that mean July air temperature (T July), distance from the tree line, water depth, pH, and Altitude explain the most significant variance in chironomid distribution, whereas T July appeared to be the most important environmental variable. The authors, therefore, suggested that their training set can be recommended for application in palaeoclimatic studies in the East of Northern Eurasia.

 

This is a well-conducted and timely study providing interesting and reliable results in the field of palaeoclimatic studies and chironomid ecology. The paper is generally well-written and therefore deserves to be published.

 

I have only a few major concerns that require more work or clarification:

 

- the location map must be revised by adding the following information: elevation, geographical coordinates, published vs. new study sites, and vegetation types. The present map is not informative.

 

- a more thorough description of the study sites should also be added in the result section (PCA?) to better show how the selected sites will help to explore the relationship between climate and chironomid (that is key in this type of study).

 

- it is always a bit surprising to see that, despite the very large latitudinal (and therefore temperature) gradient covered by this study, the explained variance is still very low, only 8.9% for the best model, suggesting that the key driver of variability in chironomid community composition is missing (could also be related to the sampling, not only environmental variables, I don’t know). The authors should at least say a few words about it.

 

- I also do believe that the authors ruled out too quickly any potential biogeographical (or spatial) effects on taxonomic composition, especially since their study covered such large gradients. There are numerous methods that could be used to try to quantify the spatial vs. environmental controls (see also Benito et al. 2019 Lake regionalization and diatom metacommunity structuring in tropical South America).

 

- Finally, the two most interesting (and novel) results of this study are not discussed enough: only from l. 265 to l. 269.

The influence of rare vs. common species on temperature inference and the differences in temperature optima deserves more attention in the draft (maybe how we should adapt the protocol to this point, HC count set, etc.).

Furthermore, the fact that the same morphotype can have different ecological requirements is also very intriguing and must be further explored. Among others, can we also assume that these differences could also have occurred in the past; a morphotype having different ecological requirements because the environmental conditions differed from nowadays (meaning that modern samples would be poor analogs?)

Author Response

Reviewer 1

Comments and Suggestions for Authors

Review of “Chironomid-based modern summer temperature data set and inference model for the Northwest European part of Russia” by Nazarova et al.

In this study, the authors aimed to produce a new regional chironomid-based inference model for quantitative palaeoclimatic reconstructions for North-western Russia. They sampled 98 lakes and identified chironomid remains to the finest taxonomic level. Overall, results showed that mean July air temperature (T July), distance from the tree line, water depth, pH, and Altitude explain the most significant variance in chironomid distribution, whereas T July appeared to be the most important environmental variable. The authors, therefore, suggested that their training set can be recommended for application in palaeoclimatic studies in the East of Northern Eurasia.

This is a well-conducted and timely study providing interesting and reliable results in the field of palaeoclimatic studies and chironomid ecology. The paper is generally well-written and therefore deserves to be published.

I have only a few major concerns that require more work or clarification:

- the location map must be revised by adding the following information: elevation, geographical coordinates, published vs. new study sites, and vegetation types. The present map is not informative.

A: we corrected the map. In addition to the new map we provide now an further information on which sampling regions are old and which are new in the Introduction, as well as in the M&M sections:

“..between our earlier published data from two regions in west European part of Russian arctic (Pechora and Komi, Fig. 1; [16, 19, 23]) with the addition of new sampling regions to the data set: Anzher Solovki (7 lakes), Central European region (3 lakes), Karelian Isthmus and Ladoga (23 lakes), Kola Peninsula (9 lakes), Karelia Zaonezhje (12 lakes), Novaya Zemlya (1 lake), Onega Peninsula (6 lakes) (Fig. 1).”

“.the data collected earlier in the north-eastern part of European Russia (Komi and Pechora region)”.

This information we added as well to the Figure caption. We made an additional Table S1 with more information provided for all sampling regions.

Detailed information on coordinates, vegetation type, and other ecological features of each single investigated lake will be uploaded into Pangea upon (if) the manuscript acceptance for publication.

 

- a more thorough description of the study sites should also be added in the result section (PCA?) to better show how the selected sites will help to explore the relationship between climate and chironomid (that is key in this type of study).

A: We added an additional table to the SEM (Table S1) with the main ecological characteristics of the regions. Locations of the sampling regions are presented at the Map (Fig. 1) that has been adjusted accordingly. Detailed information on lake coordinates, vegetation type, and other ecological features of the investigated lakes will be uploaded into Pangea after (if) the manuscript is accepted for publication.

PCA can`t be used here as the DCA analysis has showed that the data have a unimodal distribution. DCA plot is shown in the manuscript (Fig. 3).

 

- it is always a bit surprising to see that, despite the very large latitudinal (and therefore temperature) gradient covered by this study, the explained variance is still very low, only 8.9% for the best model, suggesting that the key driver of variability in chironomid community composition is missing (could also be related to the sampling, not only environmental variables, I don’t know). The authors should at least say a few words about it.

A: this is not fully correct. The Table 3 provides an information on the features of CCA axes. To avoid the misleading information, we replace the Table 3 by the Table demonstrating the results of forward selection CCAs, where the proportion of the variance explained by each significant variable and the variance explained by the significant environmental variables are provided. The earlier Table 3 is moved into SEM as Table S2. The presented in the Table S2 results are well comparable with the earlier published results of CCA from other regions: e.g. Barley et al., 2006; Nazarova et al., 2011, 2015.

 

- I also do believe that the authors ruled out too quickly any potential biogeographical (or spatial) effects on taxonomic composition, especially since their study covered such large gradients. There are numerous methods that could be used to try to quantify the spatial vs. environmental controls (see also Benito et al. 2019 Lake regionalization and diatom metacommunity structuring in tropical South America).

A: Thank you very much for providing us with the valuable recommendation and interesting references. However, in comparison to other similar studies, we see that the investigated in this study gradient is not that long (9.2 – 18.1 °C) and even needs an extension by including more sites from the cold part of the gradient. We admit this in the manuscript (end of Conclusion, and Discussion, 1^st paragraph). The lakes in the NWE are separated from each other though by big (max. 1900 km) distance, but as well, not as big as e.g. in our earlier published datasets (over 3000 km, Self et al., 2011; over 3700 km Nazarova et l., 2015). Taxonomic distances analysis demonstrates that the regions are taxonomically comparable: median TD=1.02, which mean that at least 50% of the fauna is common for all regions.

 

- Finally, the two most interesting (and novel) results of this study are not discussed enough: only from l. 265 to l. 269.

A: we prefer to stay cautious with making any further conclusions, as our NWE data set is warmer than the earlier investigated North Eurasian dataset (Nazarova et al., 2008, 2011, 2015; Self et al., 2011) where the gradient length is 1.8 to 18.6 °C. It shows that we must further look for colder places across the West Eurasian arctic in order to prove or correct this finding. But still we think that the information we could obtain from the NWE data set should be presented in order to illustrate the current state of art in our investigations.

 

The influence of rare vs. common species on temperature inference and the differences in temperature optima deserves more attention in the draft (maybe how we should adapt the protocol to this point, HC count set, etc.).

A: yes, this is an interesting question, but to try to find an answer on this question it is necessary to perform  “test reconstructions” where we use the same model and core data and apply the model differently: taking into account and not taking into account rare taxa (we do not use “species” in palaeochironomids). Therefore, this subject lays outside of the tasks we have in the current publication. Rare taxa downweighting is a standard feature of the CANOCO, when making a preparatory statistic for the model development.

Our experience showed that exclusion of any of rare taxa from reconstructions does not influence the resulting reconstructions. I tested it for me when making plenty of reconstructions with the same outcome. But this was done just privately, to see if I could achieve any better results when excluding some unusual taxa from the reconstruction. This might be a subject for a future methodological work for sure, but only after a very detailed investigation and analysis of the resulrs.

Count settings for chironomids head capsules are standard. Here in the manuscript, we cite 2 of 3 known experimental works proving that at least 50 HC should be investigated in each sample. It was as well proven that in cold regions (arctic) even 30 HC can be representative for a reliable estimation of faunistic composition:

Heiri, O.; Lotter, A. F. Effect of Low Count Sums on Quantitative Environmental Reconstructions: An Example Using Subfossil Chironomids. J Paleolimnol 2001, 26, 343–350. https://doi.org/10.1023/A:1017568913302.

Quinlan, R.; Smol, J. P. Setting Minimum Head Capsule Abundance and Taxa Deletion Criteria in Chironomid-Based Inference Models. J Paleolimnol 2001, 26, 327–342.

The third experimental work is:

Larocque I. 2001. How many chironomid head capsules are enough? A statistical approach to determine sample size for palaeoclimatic reconstructions. Palaeogeography, Palaeoclimatology, Palaeoecology. 172, 133-142.

 

Furthermore, the fact that the same morphotype can have different ecological requirements is also very intriguing and must be further explored. Among others, can we also assume that these differences could also have occurred in the past; a morphotype having different ecological requirements because the environmental conditions differed from nowadays (meaning that modern samples would be poor analogs?)

A: differences in morphotypes ecological requirements evolve from taxonomical uncertainties that we have when dealing with chironomid remains (Brooks et al., 2007). To date, there are more than 5000 chironomid species known (some estimated say more than 15000) and many of them are indistinguishable by the larvae stage. Most of chironomid larvae can`t be identified beyond genus or species-group level. Additionally, if we deal only with the head capsules, even more details necessary for identification are lacking. For this reason, we deal with the “taxa”, not with species. Detailed description of the “taxa” are provided in Brooks et al., 2007, where the work of all palaeochironomid workers was put together as a standard in order to avoid misleading taxonomy.

This limits the applicability of the calibration sets to the region in which the certain morphotypes are collected as the same morphotype can be in fact different species that inhabit different regions. Another limit is late Quaternary, as a boundary for the insect migration (Brooks et al., 2007). The oldest identifiable specimens I studied and from which I was able to make reliable reconstructions are Emian sediments: ca 130 ka BP (Kienast et al., 2011), 130.9 ka BP (Plikk et al., 2018).

For this reason, it is necessary to develop separate models based on calibration dataset for different regions. E.g. our data set for Siberia (all the area east of Ural mountains) had to be divided into 3 parts based on taxonomic similarities and geographical locations of the investigated ca 300 lakes: west Siberia, East Siberia and Far East (Nazarova et al., 2011, 2015; Self et al., 2011).

In our manuscript here we make comparison of T optima for the same taxa in different calibration sets across the Northern Eurasia (Fig. 6).

There is one more point to the application of the transfer functions. For any reconstructions we make additional tests that show how reliable the reconstruction is (Modern Analogue Technique (MAT), the goodness-of-fit (G-O-F) statistics (Birks et al. 1990); Statistical significance of the reconstruction using the method of (Telford and Birks, 2011). For questionable regions we make reconstructions using several transfer function in order to see which one can fit better (Plikk et al., 2018; Pł´ociennik et al., 2022) and supplement with the significance tests as well. This allows to avoid using un-applicable transfer functions and helps to improve the quality of reconstructions.

Author Response File: Author Response.pdf

Reviewer 2 Report

Review of the article by Nazarova et al.

Chironomid-based modern summer temperature data set and inference model for the Northwest European part of Russia

The article is devoted to establishing the relationship between modern chironomid assemblages and environmental conditions (predominantly mean July air temperature) in North-Eastern Eurasia (North-Western Russia), and this study fits the profile of “Water” journal.

The study under consideration is relevant: 1) it fills a gap in the study of chironomids of Northern Eurasia (the lakes of North-Eastern Europe have not been investigated, with some exceptions); 2) obtained results can later form the basis of paleoclimatic reconstructions within the study area.

The goal of the MS is clearly stated: studying the influence of various environmental factors on the distribution and abundance of chironomids in a recently collected “chironomids-environment” dataset of lakes in North-Western Russia and developing a regional inference chironomid -based model for the reconstruction of regional climatic and ecological changes of the past. Research tasks are solved by applying standard methods, and only the territory where the inference model can be applied is expanded. There are no comments for this part.

However, I have some remarks on study design.

In the introduction, authors note that chironomid-based inference models can hardly be used outside of the regions where they were developed. That is why “it was imperative to develop a regional chironomid-based inference model for application in the East European North” (lines 50-51). At the same time, in the presented MS, 98 studied lakes are combined into one regional dataset, despite the location on a vast territory – more than 1600 km in both latitude and longitude directions. This territory, as noted by the authors, includes such different natural zones as arctic desert, tundra, forest tundra, northern taiga, and taiga. It is obvious that these natural zones differ significantly in their landscape and climatic conditions.

It would be nice if the authors argue or substantiate their choice on lakes locations for building chironomid-based inference model. The administrative location of all studied lakes in Russian Federation seems not a convincing argument: the natural conditions of the Kola Peninsula (NW Russia) are much closer to those of the high latitude Sokli site (N Finland) than to those of the Central European region (Russia).

Unfortunately, the authors do not show the boundaries of the natural zones where 98 studied lakes are located. Instead, an uninformative map of administrative boundaries within the Russian Federation is provided in Fig. 1. The distance from tree line (DTL) is one of the important parameters to achieve the study goals. Visualization of the boundaries of natural zones including tree line (in Fig. 1) would make clearer, for example, the negative values of the DTL in Table 1. Information on how many lakes are located in each of the zones is not provided neither in the figure 1, nor in the table 1.

Only 2 out of 98 studied lakes have been excluded as outliers, but the authors do not analyze what natural zones these lakes are from and what are characteristics of these lakes. This information is important and should be included.

By giving this additional information, the authors will help the reader to better navigate the presented material.

In my opinion, the article requires a minor revision and, after correcting a few remarks, it could be published, since it is devoted to a little-known part of Northern Eurasia and is of scientific interest.

 

Specific remarks

In the figure 1, with the location of the lakes, it is desirable to show the boundaries of natural zones and the graticule or scale bar (the graticule on the globe in the inset is poorly visible).

 

Lines 60-62.    

The authors state: “The data set includes 98 lakes, which are situated across a wide latitudinal and longitudinal ranges and several environmental zones (arctic desert, tundra, forest tundra, northern taiga, taiga) (Table 1).” However, Table 1 (line 77) does not provide information on the belonging of lakes to certain natural zones or on the number of lakes in each of the natural zones. The caption under the table (lines 78-79) just indicates the following: *Vegetation types are not presented in the table and include: arctic desert; tundra; forest tundra; northern taiga; taiga. Thus, reference on the Table 1 at the end of sentence mentioned above is misleading.

 

Line 147 - Please check the spelling of the names of lake locations: Zaonezhje, Anzhes Solovli, Onegy Peninsula. They differ from the names in the caption to Fig. 1: Anzher Solovki (AS); Karelia Zaonezhje (KZ); Onega Peninsula (OP).

 

The first author’s research and publication activity deserve sincere respect; however, according to the ethical standards of citation, the number of self-citations should not exceed 30%. 30 self- cited publications out of 75 (40%) involuntarily attracts attention. Perhaps it would be better to introduce self-citation into accepted norms.

Author Response

Reviewer 2

The article is devoted to establishing the relationship between modern chironomid assemblages and environmental conditions (predominantly mean July air temperature) in North-Eastern Eurasia (North-Western Russia), and this study fits the profile of “Water” journal.

The study under consideration is relevant: 1) it fills a gap in the study of chironomids of Northern Eurasia (the lakes of North-Eastern Europe have not been investigated, with some exceptions); 2) obtained results can later form the basis of paleoclimatic reconstructions within the study area.

The goal of the MS is clearly stated: studying the influence of various environmental factors on the distribution and abundance of chironomids in a recently collected “chironomids-environment” dataset of lakes in North-Western Russia and developing a regional inference chironomid -based model for the reconstruction of regional climatic and ecological changes of the past. Research tasks are solved by applying standard methods, and only the territory where the inference model can be applied is expanded. There are no comments for this part.

However, I have some remarks on study design.

In the introduction, authors note that chironomid-based inference models can hardly be used outside of the regions where they were developed. That is why “it was imperative to develop a regional chironomid-based inference model for application in the East European North” (lines 50-51). At the same time, in the presented MS, 98 studied lakes are combined into one regional dataset, despite the location on a vast territory – more than 1600 km in both latitude and longitude directions. This territory, as noted by the authors, includes such different natural zones as arctic desert, tundra, forest tundra, northern taiga, and taiga. It is obvious that these natural zones differ significantly in their landscape and climatic conditions.

A: We made some adjustments in the introduction in order to better explain the background of the research and the steps that we have done to reach the main goal of the study – development of a regional transfer function.

“In this paper we present the results of our work that has been done to reanalyse and standardise the taxonomy between our earlier published data from two regions in west European part of Russian arctic (Pechora and Komi, Fig. 1; [16, 19, 23]) with the addition of new sampling regions to the data set: Anzher Solovki (7 lakes), Central European region (3 lakes), Karelian Isthmus and Ladoga (23 lakes), Kola Peninsula (9 lakes), Karelia Zaonezhje (12 lakes), Novaya Zemlya (1 lake), Onega Peninsula (6 lakes) (Fig. 1). Following taxonomic standardisation and analysis of taxonomic distances between the sampling regions, we merged the data sets. Merging the datasets has essential advantages: extending the environmental and geographical gradients, increasing the representation of taxa in the calibration set, improving the performance and widening the applicability of the chironomid-temperature inference model by providing more reliable estimates of the environmental optima of chironomid taxa [5,23]. Inclusion of the greater number of the geographically and ecologically suitable lakes into the model increases the probability of better analogues between present and past assemblages [5]. “

 

The lakes in the NWE are separated from each other though by big distance, but as well, not as big as in our earlier published datasets (over 3000 km, Self et al., 2011; over 3700 km Nazarova et l., 2015). Taxonomic distances demonstrated that the regions are taxonomically comparable: median TD=1.02, which mean that at least 50% of the fauna is common for all regions.

Our study has shown that the investigated in this study gradient is not long enough (9.2 – 18.1 °C) and needs an extension by including more sites from the cold part of the gradient. We admit this in the manuscript (end of Conclusion, and Discussion, 1st paragraph) as our plan for future development.

 

It would be nice if the authors argue or substantiate their choice on lakes locations for building chironomid-based inference model. The administrative location of all studied lakes in Russian Federation seems not a convincing argument: the natural conditions of the Kola Peninsula (NW Russia) are much closer to those of the high latitude Sokli site (N Finland) than to those of the Central European region (Russia).

A: administrative units are of no importance in the current study. This work was not a result of any special project or effort of a special group, but a result of volunteering efforts of the authors. In this study we used the lakes that we were able to reach within the expeditions that took place in recent years, and in which, collecting of the surface sediment samples was mostly a secondary task, or just a good will of the expedition members. We are extremely grateful to all participants of expeditions, who provided us with the samples, and we admitted them in the Acknowledgements with a big pleasure and gratitude. This is how the material was collected, and we simply were happy to have the samples we had in the end. If you allow, we would prefer to skip this information from the manuscript and leave this point as it is, just widening the introductory part by additional explanation of the steps we had to do.

We thought as well that merging the data we have with the lakes from Finish (Prof. T. Luoto) or Norwegian data set (S. Brooks) could be the next step. But this is the subject we have to discuss more with the owners of the Finish or Norwegian data. There is as well a published Swedish data set (Larocque, 2001) that could be geographically very suitable, but the data were collected and proceeded before the taxonomic standardization took place (Brooks et al., 2007), which means the whole data set has to be re-analysed. However, the owner of the data set, is now out of scientific work and the task seems to look unrealistic. Another possibility could be to see if the Polish data set could be suitable for merging with our data, but as well, this is a subject for the future discussions with the data set owner (Prof. M. Płóciennik).

However, prior to merging the parts of the dataset we made taxonomical standartisation, and statistical analysis, including taxonomic distances, that have shown that the investigated regions can be merged together into a single dataset. This approach we have used earlier when developing transfer functions for Siberia and Far East.

 

Unfortunately, the authors do not show the boundaries of the natural zones where 98 studied lakes are located. Instead, an uninformative map of administrative boundaries within the Russian Federation is provided in Fig. 1. The distance from tree line (DTL) is one of the important parameters to achieve the study goals. Visualization of the boundaries of natural zones including tree line (in Fig. 1) would make clearer, for example, the negative values of the DTL in Table 1. Information on how many lakes are located in each of the zones is not provided neither in the figure 1, nor in the table 1.

A: We adjusted the map according to suggestions of all reviewers. We added the elevation, the tree-line and deleted the administrative boundaries. Boundaries of the natural zones would make the map much more difficult to read. Instead we prepared an additional Table S1, in which we provide more information for each sampling region. Detailed information on coordinates, vegetation type, and other ecological features of each single investigated lake will be uploaded into Pangea upon (if) the manuscript acceptance for publication.

 

Only 2 out of 98 studied lakes have been excluded as outliers, but the authors do not analyze what natural zones these lakes are from and what are characteristics of these lakes. This information is important and should be included.

A: In our study the lakes were defined as outliers based on their absolute residual of the samples that exceeded the standard deviation of T July in all models [Birks et al., 1990]. As, extremely unfortunately, we do not have a full spectrum of ecological parameters for the investigated lakes, we couldn`t use an ecological approach to elimination of the lakes from the data set, as we did it in our previous studies, and had to rely solely on statistical results.

 

By giving this additional information, the authors will help the reader to better navigate the presented material.In my opinion, the article requires a minor revision and, after correcting a few remarks, it could be published, since it is devoted to a little-known part of Northern Eurasia and is of scientific interest.

 

Specific remarks

In the figure 1, with the location of the lakes, it is desirable to show the boundaries of natural zones and the graticule or scale bar (the graticule on the globe in the inset is poorly visible).

A: we corrected the map and added all possible information.

We now provide an additional information on which sampling regions are old and which are new in the Introduction, as well as in the M&M sections:

“..between our earlier published data from two regions in west European part of Russian arctic (Pechora and Komi, Fig. 1; [16, 19, 23]) with the addition of new sampling regions to the data set: Anzher Solovki (7 lakes), Central European region (3 lakes), Karelian Isthmus and Ladoga (23 lakes), Kola Peninsula (9 lakes), Karelia Zaonezhje (12 lakes), Novaya Zemlya (1 lake), Onega Peninsula (6 lakes) (Fig. 1).”

“.the data collected earlier in the north-eastern part of European Russia (Komi and Pechora region)”.

This information we added to the Figure caption. We created an additional Table S1 with more information provided for all sampling regions. Detailed information on coordinates, vegetation type, and other ecological features of each single investigated lake will be uploaded into Pangea upon (if) the manuscript acceptance for publication.

Lines 60-62.    

The authors state: “The data set includes 98 lakes, which are situated across a wide latitudinal and longitudinal ranges and several environmental zones (arctic desert, tundra, forest tundra, northern taiga, taiga) (Table 1).” However, Table 1 (line 77) does not provide information on the belonging of lakes to certain natural zones or on the number of lakes in each of the natural zones. The caption under the table (lines 78-79) just indicates the following: *Vegetation types are not presented in the table and include: arctic desert; tundra; forest tundra; northern taiga; taiga. Thus, reference on the Table 1 at the end of sentence mentioned above is misleading.

A: Information on distribution of the lakes along the certain environmental gradients is presented in the Figure 2, from which it can be seen how many lakes are found in each part of each gradient. Even though not it is not presented in exact numbers, but even more importantly, the figure shows how evenly lakes are distributed along the gradients. Moreover, we controlled all lakes for skewness of their distribution along the environmental gradients and significantly skewed data, were log transformed (Altitude and WD). From the Figure 2 it is obvious that the lakes are not evenly distributed along the altitude and especially water depth. These data were log transformed to reach normality of distribution.

Additionally, we added information on ecological features of the lakes in each region where sampling took place (Table S1). List of the lakes with the full information will be uploaded into Pangea after (if) the manuscript is accepted for publication. This statement will be added to the manuscript upon the manuscript acceptance.

 

Line 147 - Please check the spelling of the names of lake locations: Zaonezhje, Anzhes Solovli, Onegy Peninsula. They differ from the names in the caption to Fig. 1: Anzher Solovki (AS); Karelia Zaonezhje (KZ); Onega Peninsula (OP).

A: we corrected, thank you.

 

The first author’s research and publication activity deserve sincere respect; however, according to the ethical standards of citation, the number of self-citations should not exceed 30%. 30 self- cited publications out of 75 (40%) involuntarily attracts attention. Perhaps it would be better to introduce self-citation into accepted norms.

A: Sorry for not taking this into account. We reduced self- citation. The thing is, that hardly anybody else has been working on palaeochironomids in Russia apart from the LN; for this reason, she is a co-author of practically all palaeostudies that includes chironomids in Russian arctic. We refer in many cases to these studies and make comparison with studies in northern Eurasia where chironomids were part of the research.

Reviewer 3 Report

Sorry for the delay, we have a lot of work to do.

 

The work is interesting and useful. The perception of this article is prevented by numerous misprints.

Climate research is an extremely relevant question, especially in our times. The using of classical methods and the involvement of new approaches can improve the accuracy of temperature determination. It is known that each region needs to develop its own assessment model. Chironomids appear to be a very convenient object. They are found in almost all reservoirs, have a high diversity of species, and their head capsules are available for analyzing for a long time. The authors explored a lot of lakes and obtained interesting data.

 

Figure 4. Where are the designations for the axes? - Relative abundance?

Table 1. “sqew” misprint – “skew and skewness”, like in line 90?

line 83. “HC” is the Head Capsules?

line 105. “To access the taxonomic distances” – assess?

line 158. This abbreviation is not deciphered - North West European Russia (NWE), only in line 207.

We have not found an explanation of what unites lakes with similar, extreme characteristics, for example, with a pH below 6?

 

Author Response

Reviewer 3

Sorry for the delay, we have a lot of work to do.

 

The work is interesting and useful. The perception of this article is prevented by numerous misprints.

Climate research is an extremely relevant question, especially in our times. The using of classical methods and the involvement of new approaches can improve the accuracy of temperature determination. It is known that each region needs to develop its own assessment model. Chironomids appear to be a very convenient object. They are found in almost all reservoirs, have a high diversity of species, and their head capsules are available for analyzing for a long time. The authors explored a lot of lakes and obtained interesting data.

 

Figure 4. Where are the designations for the axes? - Relative abundance?

A: axis X is a relative abundance and the sign “%”  is provided under the axis. Axis Y is mean July temperature and the title “T July, ° C” is provided

 

Table 1. “sqew” misprint – “skew and skewness”, like in line 90?

A: thank you, we corrected.

 

line 83. “HC” is the Head Capsules?

A: yes, we added “head capsules”

 

line 105. “To access the taxonomic distances” – assess?

A: yes. We re-phrase the sentence.

 

line 158. This abbreviation is not deciphered - North West European Russia (NWE), only in line 207.

A: we put the clarification to the line 158.

 

We have not found an explanation of what unites lakes with similar, extreme characteristics, for example, with a pH below 6?

A: This can be an interesting point to explore. In our study the lakes were defined as outliers based on their absolute residual of the samples that exceeded the standard deviation of T July in all models [Birks et al., 1990]. As, extremely unfortunately, we do not have a full spectrum of ecological parameters for the investigated lakes, we couldn`t use an ecological approach to elimination of the lakes from the data set, as we did it in our previous studies, and had to rely solely on statistical results.

Though it is important to find out how other than T parameters influence the faunistic characteristic of the lakes, in our dataset we focus on T July air temperature that appeared to be the most important parameter. In our dataset we have median pH =6.8, with st.dev 0.86 and skew -0.05 (which is very low). As there were only 4 lakes with low pH (≤6), we can`t be sure that we can draw some pattern out of such a small number of lakes. But it might be interesting to clear this point out when we will work on the bigger data set that will include (presumably) all our lakes in all parts of Northern Eurasia. The current dataset is supposed to be a part of this bigger ecological study.

Author Response File: Author Response.pdf

Reviewer 4 Report

Chironomid-based modern summer temperature data set and inference model for the Northwest European part of Russia. Nazarova et al.

I find it an interesting and well-developed study in general terms. The manuscript is easy to read and understand and the figures and tables are adequate.

Temperature inference models is of relevance in climate change studies and the authors present a model for a region highly affected by this climate change.

My first concerns refer to the statistical results. The % of explained variance derived from CCA is very low. I would like this is to be discussed or thinking in alternative analysis.

I have some other comments on the manuscript:

Water is a general journal about water science and technology. It is not a journal specializing in paleoecology. Therefore, some methods and results should be explained at greater length so that they are understandable to all potential readers.

In general, I found too many acronyms in the text, some of them are not necessary

 

INTRODUCTION

Introduction can be implemented and extended, for example explaining why it is important to make this model for reconstructing temperature

It could also be implement explaining why chironomid-based inference models can hardly be used outside of the regions in which they were developed.

Likewise, explain why chironomid remains from the sediment is appropriate to do this research and better than alive chironomids, for example.

It is important to say in the introduction tan the modern chironomid assemblages, environmental conditions, and inference models for reconstructing T July is based in chironomid remains preserved in lake sediments. Explain that these remains are head capsules.  Similarly, the word sediment should be included in the abstract and keywords and maybe in the tittle too.

Abstract- “We investigated the chironomid fauna of a 98-lake data”. Please specify that the chironomid data set is obtained from chironomid remains from surficial sediment.

 

M&M

Do lakes stratify in summer?

 Which lakes are stratified and which are not? Are the stratified lakes in the same region and the non-stratified lakes in a separate region?

 

Line 83- 50 HC. Please, write chironomid head capsule . Anyway 50 HC seems to me a low number

 

Line 91- Environmental parameters should be standardized before applying statistical analysis. An usual transformation is z-scores or relative-scale variable if CCA is applied.

(see for example p205-207 from  Legendre P, Birks HJB (2012) From classical to canonical ordination. In: Tracking environmental change using lake sediments. In: Birks HJB, Lotter AF, Juggins S, Smol JP (eds) Data handling and numerical techniques, vol 5. Springer, Dordrecht, pp 201–248)

Line 105- Which are the new data set and the old data set?

Line 110- Please rephrase this sentence.

 

RESULTS

Line 144- “DCA (Fig. 3) and TD (Table 2) analysis demonstrated that NZ is the most taxonomically distinct region” .  There is only one lake en the NZ group , so it is difficult to say NZ is the most taxonomically distinct region

Line 146- Rephrase the sentence

Line 159. NEW? Specify

 Line 176-7-   The CCA with five variables had  EIGENVALUE OF   CCA axis 1 of 0.154 and CCA axis 2 of 0.086 explain 4.7% and 2.6% of the variance in the data (Table 3). The word eigenvalue is missing?

The % of explained variance is very low. It seems there are other variables affecting the distribution of species or perhaps it is necessary to reduce the number of lakes and make a more homogeneous set of lakes.

Maybe it would be better to have more than one transfer function, it is a very large region and heterogeneous group of lakes. I would recommend to do it and check if results are improved.

The lakes have a very wide depth range (range 0.7 to 140 m) and I suppose that some  lakes are stratified in summer and others are not. If this is so, perhaps there is no relationship between the air temperature and the cold hypolimnion temperature in stratified lakes. Could the depth of the lakes have affected the results of the transfer function? For example, Eggermont and Heiri (2012) discuss this topic.

Line 186- T July explains the most significant part of the data variance but this part is a low percentage. Please discuss these results in this sense.

Line 195- r2 boot shows a low value

FIGURES

Fig. 2- Please, specify what is WD

DISCUSSION

I would like to read something in the discussion about the depth of the lakes and its possible influence on the results of the transfer function.

I will also like to read about the statistical results as I have commented in the results section

Author Response

Reviewer 4

Chironomid-based modern summer temperature data set and inference model for the Northwest European part of Russia. Nazarova et al.

I find it an interesting and well-developed study in general terms. The manuscript is easy to read and understand and the figures and tables are adequate.

Temperature inference models is of relevance in climate change studies and the authors present a model for a region highly affected by this climate change.

My first concerns refer to the statistical results. The % of explained variance derived from CCA is very low. I would like this is to be discussed or thinking in alternative analysis.

A: this is not fully correct. The Table 3 provides an information on the features of CCA axes. To avoid the misleading information, we replace the Table 3 by the Table demonstrating the results of forward selection CCAs, where the proportion of the variance explained by each significant variable and the variance explained by the significant environmental variables are provided. The earlier Table 3 is moved into SEM as Table S2. The presented in the Table S2 results are well comparable with the earlier published results of CCA from other regions: e.g. Barley et al., 2006; Nazarova et al., 2011, 2015.

 

I have some other comments on the manuscript:

Water is a general journal about water science and technology. It is not a journal specializing in paleoecology. Therefore, some methods and results should be explained at greater length so that they are understandable to all potential readers.

A: we re-write the introduction according to suggestions, changed Table 3 to a more self-explaining one, changed the map and made some additions additional changes in the text as advised.

 

In general, I found too many acronyms in the text, some of them are not necessary.

A: We use the following acronyms: water depth (WD), Northwest European data set (NWE), mean July air temperature (T July), distance from the tree line (DTL) and reduced the names of the sampling site regions as we thought they are difficult to read and remember. But probably it really might be useful to retain the full names if the sampling regions in the text, as they appear not that often as the acronyms WD, T July, DTL (that are as well pretty standard) and the name of the data set. We retain the acronyms in the figures and table captions in order to make the figures and tables better readable.

 

INTRODUCTION

Introduction can be implemented and extended, for example explaining why it is important to make this model for reconstructing temperature

It could also be implement explaining why chironomid-based inference models can hardly be used outside of the regions in which they were developed.

Likewise, explain why chironomid remains from the sediment is appropriate to do this research and better than alive chironomids, for example.

It is important to say in the introduction tan the modern chironomid assemblages, environmental conditions, and inference models for reconstructing T July is based in chironomid remains preserved in lake sediments. Explain that these remains are head capsules.  Similarly, the word sediment should be included in the abstract and keywords and maybe in the tittle too.

A: were-write the introduction by adding the information you advise.

 

Abstract- “We investigated the chironomid fauna of a 98-lake data”. Please specify that the chironomid data set is obtained from chironomid remains from surficial sediment.

A: we made correction: “We investigated the subfossil chironomid remains from surface sediments of a…”

M&M

Do lakes stratify in summer? Which lakes are stratified and which are not? Are the stratified lakes in the same region and the non-stratified lakes in a separate region?

A: as the most of the lakes are difficult to reach, during short expeditions there was no time to deeply investigate each single lake. The answer on the question could be only a speculation based on a single visit to the lakes: in most of cases not stratified. Unfortunately, more information can`t be provided. Sorry for this. We have to admit that we didn`t have this information for any of our previous studies resulted in chironomid based transfer function (Nazarova et al., 2011, 2015; Self et al., 2011)

 

Line 83- 50 HC. Please, write chironomid head capsule. Anyway 50 HC seems to me a low number.

A: Thank you, we wrote “chironomid head capsules”.  “50 HC at least” is a standard. Here in the manuscript we cite 2 of 3 known experimental works proving that at least 50 HC should be investigated in each sample. It was as well proven that in cold arctic regions even 30 HC can be representative for a reliable estimation:

Heiri, O.; Lotter, A. F. Effect of Low Count Sums on Quantitative Environmental Reconstructions: An Example Using Subfossil Chironomids. J Paleolimnol 2001, 26, 343–350. https://doi.org/10.1023/A:1017568913302.

Quinlan, R.; Smol, J. P. Setting Minimum Head Capsule Abundance and Taxa Deletion Criteria in Chironomid-Based Inference Models. J Paleolimnol 2001, 26, 327–342.

The third experimental work is:

Larocque I. 2001. How many chironomid head capsules are enough? A statistical approach to determine sample size for palaeoclimatic reconstructions. Palaeogeography, Palaeoclimatology, Palaeoecology. 172, 133-142.

 

Line 91- Environmental parameters should be standardized before applying statistical analysis. An usual transformation is z-scores or relative-scale variable if CCA is applied.

(see for example p205-207 from  Legendre P, Birks HJB (2012) From classical to canonical ordination. In: Tracking environmental change using lake sediments. In: Birks HJB, Lotter AF, Juggins S, Smol JP (eds) Data handling and numerical techniques, vol 5. Springer, Dordrecht, pp 201–248)

A: Thank you very much for a valuable advice. We use in our work the protocol used in practically all studies aimed in producing a chironomid-based transfer-functions: earlier references are presented in Nazarova et al., 2011, ESM 1. To stay in line with existing approach, we controlled all available environmental variables for skewness [39] and the data with skewed distributions were log transformed. Remaining parameters were left untransformed. Application of the standard methods that have been used in all our models provide an additional value. We are able to better compare the results of our studies that include already multiple regions across the arctic Eurasia.

 

Line 105- Which are the new data set and the old data set?

A: provide now a further information on which sampling regions are old and which are new in the Introduction, as well as in the M&M sections:

“..between our earlier published data from two regions in west European part of Russian arctic (Pechora and Komi, Fig. 1; [16, 19, 23]) with the addition of new sampling regions to the data set: Anzher Solovki (7 lakes), Central European region (3 lakes), Karelian Isthmus and Ladoga (23 lakes), Kola Peninsula (9 lakes), Karelia Zaonezhje (12 lakes), Novaya Zemlya (1 lake), Onega Peninsula (6 lakes) (Fig. 1).”

“.the data collected earlier in the north-eastern part of European Russia (Komi and Pechora region)”.

 

Line 110- Please rephrase this sentence.

A: we change the sentence to: “We used the environmental variable explaining most variance in the data set (as indicated by the CCAs) in order to develop quantitative transfer functions based on weighted averaging partial least squares (WA-PLS) methods”.

 

RESULTS

Line 144- “DCA (Fig. 3) and TD (Table 2) analysis demonstrated that NZ is the most taxonomically distinct region” .  There is only one lake en the NZ group , so it is difficult to say NZ is the most taxonomically distinct region

A: yes, you are right, we didn`t have any more lakes from such a hard to access region but still we wanted to see, if the investigated lake will bring some value to the data set in terms of extension of the temperature gradient. It was as well interesting to see if we find any similarities of the NZ with any other regions of investigation. Of course, if the lake would be an absolute outlier based on the results of statistical investigation, it would be excluded from the model. However, the lake, though demonstrate a certain degree of distinctness (Table 2), it showed similarities with some lakes from the region north of the Ural Mountains (Fig. 3) and other northernmost lakes in the data set. It didn`t appear as outlier in the statistical analysis. Here we wanted to stress, that the lake though unique, but still can be used in the model. Additionally, we (our colleagues) already obtained another lake from the Novaya Zemlya, and even from the Franz Josef Land (north of the NZ). The samples will be investigated and are planned to be used in the future analysis.

 

Line 146- Rephrase the sentence

A: we improved the wording: “Taxonomically closest regions (the lowest TD) are those in the middle of the DCA plot (Fig.3). These are the geographically close regions”.

 

Line 159. NEW? Specify

A: we clarified: In the 98 investigated lakes from North West European Russia (NWE)

 

 Line 176-7-   The CCA with five variables had  EIGENVALUE OF   CCA axis 1 of 0.154 and CCA axis 2 of 0.086 explain 4.7% and 2.6% of the variance in the data (Table 3). The word eigenvalue is missing?

A: many thanks. We added the word “eigenvalue”

 

The % of explained variance is very low. It seems there are other variables affecting the distribution of species or perhaps it is necessary to reduce the number of lakes and make a more homogeneous set of lakes.

A: this is not fully correct. The Table 3 provides an information on the features of CCA axes. To avoid the misleading information, we replace the Table 3 by the Table demonstrating the results of forward selection CCAs, where the proportion of the variance explained by each significant variable and the variance explained by the significant environmental variables are provided. The earlier Table 3 is moved into SEM as Table S2. The presented in the Table S2 results are well comparable with the earlier published results of CCA from other regions: e.g. Barley et al., 2006; Nazarova et al., 2011, 2015.

 

Maybe it would be better to have more than one transfer function, it is a very large region and heterogeneous group of lakes. I would recommend to do it and check if results are improved.

A: The main shortcoming of the dataset is not the heterogeneity and not the distances between the lake, but though unskewed but still uneven distribution of the lakes along the temperature gradient. The presented data set and the resulting inference model aims at presenting current state of research that results from nearly 20 years of research (taking into account the data collected in Pechora and Komi regions that took place in 2001 and 2006). Our current study revealed the shortcomings and showed the direction where we have to move on in order to improve the results. We admit this in Discussion (Paragraph 1) and in the conclusion: “Some skewness in the distribution of the investigated lakes along the T gradient and paucity of the lakes from the colder part of the gradient challenges our future work on complementation of the NWE with more lakes from colder regions of NW Eurasia.”

 

The lakes have a very wide depth range (range 0.7 to 140 m) and I suppose that some  lakes are stratified in summer and others are not. If this is so, perhaps there is no relationship between the air temperature and the cold hypolimnion temperature in stratified lakes. Could the depth of the lakes have affected the results of the transfer function? For example, Eggermont and Heiri (2012) discuss this topic.

A: yes, the WD has the most skewed distribution. For this reason, this parameter was log transformed prior to the analyses and appeared that the WD still plays an important role in chironomid distribution.

Thank you for advising a very interesting publication (Eggermont and Heiri, 2012). Though it still will be very difficult to trace, if the lakes are or aren`t stratified, in future we should obligatorily try to pay attention on the influence of the water depth on the T July transfer functions.

 

Line 186- T July explains the most significant part of the data variance but this part is a low percentage. Please discuss these results in this sense.

A: Table 3 provides an information on the features of CCA axes. To avoid the misleading information, we replace the Table 3 by the Table demonstrating the results of forward selection CCAs, where the proportion of the variance explained by each significant variable and the variance explained by the significant environmental variables are provided. The earlier Table 3 is moved into SEM as Table S2. The presented in the Table S2 results are well comparable with the earlier published results of CCA from other regions: e.g. Barley et al., 2006; Nazarova et al., 2011, 2015.

 

Line 195- r2 boot shows a low value.

A: yes, - r² _boot 0.6 shows a relatively low value compared to, e.g. our Siberian models, but as we write in the discussion, is comparable with the values obtained in the T July models world-wide (Nazarova et al., 2011, SEM 4) where the r² varies from 0.382 to 0.91 in different models. In the early version of Siberian chironomid based inference model (Nazarova et al., 2011) we had r² _boot =0.62. In the later version of the model we were able to improve the model behaviour and reach the r² = 0.81 (East Siberia) and r² = 0.87 (Kamchatka). Therefore, the current model shows state of art in our investigation, shows  where to go and what efforts has to be implemented in order to improve  the quality of the transfer function.

We rephrase the last, before the conclusions, paragraph:

 

 The obtained NWE chironomid-based mean July air temperature inference model has fairly moderate coefficients of determination and good RMSEP when compared with other chironomid-based mean July air temperature inference models [12-16]. An earlier version of the NWE has been used in palaeoclimatic reconstruction [29] and showed high sensitivity revealing T July fluctuations during the Late Weichselian and Holocene in Eastern Europe. Though improvements are still required in order to reach better transfer function parameters, the NWE model can be applied for palaeoclimatic reconstructions in most northern and north-western Eurasia.

 

FIGURES

Fig. 2- Please, specify what is WD

A: we added „WD for water depth”

DISCUSSION

I would like to read something in the discussion about the depth of the lakes and its possible influence on the results of the transfer function.

A: investigation of the influence that the water depth can have on the T July transfer function lays outside of our current investigation, mainly because the dataset is not too big and the water depth has a very skewed distribution in the data set which would hamper such a study. Here, as well as earlier, we found out that water depth is an important parameter and earlier we developed a chironomid based transfer function based on the data set from East Siberian (Yakutian) lakes (Nazarova et al., 2011). However, there we had a much bigger dataset and a much bigger range of the lakes with the water depth having unskewed distribution.

So, unfortunately we can`t add any information here in terms of the influence of the water depth on the functioning of the T July transfer function. Our experience has shown that the earlier developed by us chironomid based T July transfer functions works well with the very deep (Bolshoe Schuchje lake, 140 m deep, part of the current dataset, reconstruction has been published Lenz et al., 2022 doi: 10.10020qs.3400) and with the very shallow (swampy palaeolake, Nazarova et al., 2021, doi: 10.1134/S1995425521030094) lakes and even with the a coastal exposure (Kienast et al., 2011, doi:10.1016/j.quascirev.2010.11.024).

 

I will also like to read about the statistical results as I have commented in the results section

A: we made some corrections in the text and replaced the Table 3 with the table that demonstrate better the obtained results and the significance of each of the ecological parameters. As well, some corrections have been done in the discussion.

Author Response File: Author Response.pdf