Soil Temperature, Organic-Carbon Storage, and Water-Holding Ability Should Be Accounted for the Empirical Soil Respiration Model Selection in Two Forest Ecosystems

Round 1

Reviewer 1 Report

This paper uses an impressive 25-year record of soil respiration and temperature records measured at two forest sites with contrasting soil types to compare 5 different empirical methods for estimating soil respiration. The authors conclude that the best choice of empirical model varies by soil type and also by climate.

Overall, I think this is an interesting study and worth publishing, but unfortunately its presentation falls far below what I consider to be publishable standards. An extensive amount of editorial work is needed to improve the readability prior to publication. Given the extensive nature of the edits – everything from swapping tenses, missing words, spelling errors and sentences that do not make sense – I do not have the time to edit this paper and will leave this to the authors and the publisher. I will only comment on the scientific aspects of the paper.

1.       More clarity is needed throughout the paper. Quite simply, the authors use a 25-year record measured at two sites to evaluate 5 different empirical models for predicting soil respiration that varyingly use temperature, precipitation, or soil organic carbon content. The authors consider both soil and air temperature independently and current and previous year precipitation. The authors use a range of metrics to evaluate what they consider to be the best model and consider wet and dry years as well as freezing periods. This is not clear in the abstract (and should be) and in much of the text.

2.       I have a couple of questions about the data. In Figure 2 soil temperature in 2010 seems to be quite different between the sites yet they are just 9km apart? [ensuring consistent scales on the y-axis will help). I also don’t think that the colors in Figure 2 are helpful. Additionally, table 1 shows soil data for 0-10cm when the models require 0-20cm for SOC?

3.       Methods. It is not explicitly clear to me how Ro is estimated? In Figure 3, monthly SR values are shown when T is approximately 0. These values vary tremendously (argued as a legacy effect and should be ignored? [line 252-255] while I presume the lines are estimated from the intercept of a plot that compares SR with temperature? If I am correct, then this should be clarified – I would also completely omit the data that is not used (due to tree falling) from the figure as it just adds to the confusion. Finally, a major criticism I have of this paper is that the large month-to-month scatter in the data is never explained. This applies both to when Ro is shown and when the scatter plots showing modeled versus predicted SR are shown. The authors should attempt to explain why there is a tremendous amount of scatter in the data especially if the objective of this paper is to improve the accuracy of the empirical modelling.

4.       Data interpretation. The authors use several metrics to evaluate the models, but ultimately, they are subjective when deciding which model is the best. One thing that struck me is that the models generally performed equally well within a dry, wet, or normal year (biggest difference is between the 3 lines – not the models).  Comparing model fits among wet, dry and normal years in more detail would be beneficial. Additionally – in most (all?) models the slope is less than 1 – so why is this the case? Additionally, given that there are different numbers of data points for the 3-year types, the “quality” of an individual metric (e.g. R2) is influenced by n. Given that the level of performance is similar among models (and varies by parameter) and there is so much scatter in the data (e.g. figures 5 and 6) I also question how much of a difference exists in annual soil respiration when they are estimated from the 5 models compared with measured data.  Rather than just comparing two models in figures 8 and 9, I would rather see a table that shows the mean (with standard error) annual SR over the 25-year record along with the estimated model means (table or bar chart) to see if these differences in choice of model make a notable difference in annual SR, especially given the scatter in the data.

5.       My final comment is that the authors could talk about the model fits more critically. For example, it seems apparent to me that in general models underestimate SR at higher rates and overestimate SR at lower rates (Figures 5 and 6). At low respiration rates measured values often do not change much while model predictions show they increase. The same happens at higher rates. The authors should discuss these in more detail. It may be that the main conclusion of te paper is overstated and that while inclusion of SOC etc. may improve some of the model fit metrics but have little impact when assessing annual SR.

The paper needs extensive revision throughout.

Author Response

Dear reviewer, thank you so much for your comments and suggestions! To address your concerns we conducted revision of the paper focusing on the highlighted issues. Please see our answers below. We looked through the text to make it more readable, and we plan to request help from the Forests Journal to improve the manuscript readability prior to publication if needed.

1. More clarity is needed throughout the paper. Quite simply, the authors use a 25-year record measured at two sites to evaluate 5 different empirical models for predicting soil respiration that varyingly use temperature, precipitation, or soil organic carbon content. The authors consider both soil and air temperature independently and current and previous year precipitation. The authors use a range of metrics to evaluate what they consider to be the best model and consider wet and dry years as well as freezing periods. This is not clear in the abstract (and should be) and in much of the text.

Answer

Thank you for your comments. Even though the last two paragraphs of the Introduction have already mentioned the hypotheses of the cold vs. warm periods and  the years with different climatic conditions, we agree that more clarification may be needed.  Following your suggestions, we updated the Abstract and the final part of Introduction accordingly. Please see the text below.

Abstract, Lines: 16-21. The weighted non-linear regression was used for model parameter estimations for the normal, wet, and dry years. The slope and intercept of the linear-model comparison between the measured and modeled data were controlled by change in R₀ – SR at zero soil temperature – for improving the model resolutions by magnitude, and the mean bias error (MBE), root-mean-square error (RMSE), and determination coefficient (R²) were used for the estimation of goodness of model performances.

Introduction, Lines: 95-100. To do this the weighted non-linear regression was used to estimate the model coefficients for the normal, wet, and dry years separately to ensure an adequate coverage of different climatic periods. By controlling the slope and intercept of the linear-model comparison between the measured and modeled values, the selected models are being re-adjusted to adequately represent the measurement range during the year.

2. I have a couple of questions about the data. In Figure 2 soil temperature in 2010 seems to be quite different between the sites yet they are just 9km apart? [ensuring consistent scales on the y-axis will help). I also don’t think that the colors in Figure 2 are helpful. Additionally, table 1 shows soil data for 0-10cm when the models require 0-20cm for SOC?

Answer

Thank you for your suggestions. We updated the Figure 2 with the consistent scales for both sites. We would like to note that on the Figure 2 the soil temperatures are brown and the local air temperatures are red. The big difference does occur in the local air temperatures (red) and the soil temperatures (brown) as expected don't change so much. We think the graph coloration is intuitive in this case.

From the air temperature perspective, the 9-km distance can cause the observed change in monthly temperature if for example the cold front passed through the region and was detected by only one of the site measurements (different days of measurements). The measurements are usually conducted once per week and are adequate from the soil-respiration perspective. However,  they can easily cause this discrepancy in the local air temperature measurements. Moreover, in the text we have already noted (colored lines below) that the Haplic Luvisol is located on the hills and elevated about 100-150-m above the river – an approximate altitude for the plain Entic Podzol – which makes the Haplic Luvisol more susceptible to the air-temperature changes. Please see the highlighted lines in the text, clarifying this issue:

Lines:   107-108. The landscapes are plain sandy terraces formed as the result of modern and ancient erosion processes located above the Oka River flood plain

Lines:   111-112. The landscape is hilly with about 100-150-m elevation above the River.

It can be seen that in the original Table 1, there are both 0-10cm and 0-20cm SOC values presented, and in our investigation it was the 0-20cm  SOC value that was used. The choice of the top 20 cm soil layer has already been mentioned in the Methods,

Lines:  177-178. For the initialization of the empirical models depended on SOC, the estimations of its storage in 20-cm layer of Entic Podzol and Haplic Luvisol were used (Table. 1).

To clarify the possible highlighted ambiguity, we removed the 0-10cm values from the Table 1.

3. Methods. It is not explicitly clear to me how Ro is estimated? In Figure 3, monthly SR values are shown when T is approximately 0. These values vary tremendously (argued as a legacy effect and should be ignored? [line 252-255] while I presume the lines are estimated from the intercept of a plot that compares SR with temperature? If I am correct, then this should be clarified – I would also completely omit the data that is not used (due to tree falling) from the figure as it just adds to the confusion. Finally, a major criticism I have of this paper is that the large month-to-month scatter in the data is never explained. This applies both to when Ro is shown and when the scatter plots showing modeled versus predicted SR are shown. The authors should attempt to explain why there is a tremendous amount of scatter in the data especially if the objective of this paper is to improve the accuracy of the empirical modeling.

Answer

Yes, we agree that it is better to remove the data after the tree falldown at the Haplic Luvisol, which is not in use in this investigation and significantly increases the scatter. We updated the Figure 2 accordingly.

Thank you for your correction! We clarified the highlighted confusion in the text that in our research the R0 was not directly estimated from the measurements at 0 < T_soil < 1 °С but it was compared to them. The R0 was estimated from the models when the intercept of lm is positive and close to zero (these two parameters are directly interconnected to each other) – the technique we use to ensure that the measurements and modeling have similar magnitudes and there is no underestimation in lower SR values (could be observed in cold period) when the intercept of lm is negative. We updated the highlighted text.

Lines:   233-238. In our research,  was estimated from modeling by the T– TP – TPP – TPC – TPPC models in the following way:  is directly interconnected with the intercept of lm and by controlling and lowing  the intercept of lm is being readjusted to become positive and closer to zero and at the same time the slope of lm increases and becomes closer to unity. With the negative intercept of lm there will be an underestimation of the soil respiration in the winter period.

Lines:  249-255. The obtained  values (colored lines, Figure 3) are two-three times smaller than the earlier obtained values by Raich and Potter (1995) and Kurganova et al. (2019) for the Entic Podzol. The obtained  values They can be directly compared with SR measurements at 0 ≤ T_soil < 1 °С in the autumn–winter period (colored dots with labels, Figure 3) when there is no any freezing of the top-soil level, and by the selection of the temperature interval they should be located closer to the lower SR observed at the lower temperatures (T_soil ≈ 0).

Trying to investigate the large scatter in the near zero temperatures, we found some evidences that the SR values depend both on the monthly precipitation and soil properties together. For the Entic Podzol (sandy soil with poor water holding ability and larger pores) the lower precipitation periods have the extremely low SR – good drainage easily dries out the soil. However, for Haplic Luvisol (loamy soil with good water holding ability and smaller pores) the lower precipitation periods are actually associated with the higher SR – the soil is over saturated with water in cold period. This to some extent can explain the observed scattering in SR and points out the importance to account for the soil properties in SR modeling. To illustrate this we updated the Figure 3 with the color scheme from red (smallest precipitation) to blue (largest precipitation) for individual measurements - points. The following text was updated to address this issue:

Lines:  296-307. The individual values of the measured SR at near zero temperatures when there is no any freezing of the top-soil level occurred (colored dots, Figure 3) are generally associated with the end-of-the-year cold periods with not very large monthly precipitation (Figure 2). Investigating the large scattering of these SR values, we found some evidences that they depend both on the monthly precipitation and soil properties together. For the Entic Podzol (sandy soil with poor water-holding ability and larger pores) the lower precipitation periods (red and brown dots, Figure 3 left) have the extremely low SR – good drainage easily dries out this soil. However, for Haplic Luvisol (loamy soil with good water-holding ability and smaller pores) the lower precipitation periods (red and brown dots, Figure 3 right) are actually associated with the higher SR – this soil is over saturated with water [36] in cold periods. This points out the importance accounting for the soil properties in SR modeling.

4. Data interpretation. The authors use several metrics to evaluate the models, but ultimately, they are subjective when deciding which model is the best. One thing that struck me is that the models generally performed equally well within a dry, wet, or normal year (biggest difference is between the 3 lines – not the models). Comparing model fits among wet, dry and normal years in more detail would be beneficial. Additionally – in most (all?) models the slope is less than 1 – so why is this the case? Additionally, given that there are different numbers of data points for the 3-year types, the “quality” of an individual metric (e.g. R2) is influenced by n. Given that the level of performance is similar among models (and varies by parameter) and there is so much scatter in the data (e.g. figures 5 and 6) I also question how much of a difference exists in annual soil respiration when they are estimated from the 5 models compared with measured data.  Rather than just comparing two models in figures 8 and 9, I would rather see a table that shows the mean (with standard error) annual SR over the 25-year record along with the estimated model means (table or bar chart) to see if these differences in choice of model make a notable difference in annual SR, especially given the scatter in the data.

Answer

Yes, the models were fitted separately for the normal, wet, and dry years, and this is the reason why the combination of them performs well over the entire period. The modeling performance was significantly worse if the single fit could have been used. In the Results section we describe the choice of the better models by the metrics based on the normal, wet, and dry periods separately (please see Figure 7), which should address your concern of subjectivity – the inter-comparison among the sets of models and the selection of the better models are done on the same data sets ensuring the same quality of the individual metrics and the selection process over all.

As you suggested to clarify the better model selection issue we added the Figure 8 with the mean over 25-year SR (means and annual standard deviations), which shows an advantage of the models we selected: TPPrh and TPPC. It should be noted that the mean annual SR difference between the measured and modeled data for different fits has already been presented in the Figure 7 and the tables in the Appendix 1 as the mean bias error (MBE) metrics. And the same MBE metrics for the best combined models over the 25-year period is presented in the Table 2. Moreover, Table 2 illustrate the difference between the cold and warm periods and temperature sources. It shows that for the model performance analysis it is important to separate between these periods to exclude an excessive effect of the larger summer-value differences on the metrics. The following text was added to address the scattering issue

   Lines:  417-428. These conclusions are also supported by the comparison of the mean annual SR measurements with modeled values (Figure 8). While the TPPrh model shows a better annual performance for T_soil, for T_air the TPPC model becomes slightly better for Entic Podzol – the respective SR measured and modeled values stay within the standard deviation ranges of each other. However, it should be noted that these conclusions based on the annual means can be biased toward the larger summer SR values underestimating the effects the smaller winter SR. The underestimation of the mean annual SR is also in the agreement with the fact that the model-comparison slopes are smaller than unity for both Entic Podzol (Figure 5) and Haplic Luvisol (Figure 6) causing a possible underestimation of the larger summer-time SR values. In the next section we'll see that the lower winter-time SR values are actually adequately modeled by our procedure and should not influence the respective model behavior showed on the Figure 8.

It should be noted that there is a general problem with conducting the modeling against measured data with the non-linear regression that the magnitude of the modeling is less than the magnitude of the data. We have already highlighted this well observed issue in the following text

Lines:  223-227.  In the ideal model the intercept of lm should lay in the origin of the coordinate system and the slope of lm should be equal to 1. However, as a rule for the real models the slope < 1 and intercept > 0 due to insufficient magnitude of the modeled data in comparison with the measurements – the extreme values are not represented well by the modeling [20, 31, 42].

Our approach of connecting  with the intercept described above is able to mitigate this magnitude issue to most extent reducing the intercept close to zero and by the same time increasing the slope to the values close to 0.9. 

5. My final comment is that the authors could talk about the model fits more critically. For example, it seems apparent to me that in general models underestimate SR at higher rates and overestimate SR at lower rates (Figures 5 and 6). At low respiration rates measured values often do not change much while model predictions show they increase. The same happens at higher rates. The authors should discuss these in more detail. It may be that the main conclusion of te paper is overstated and that while inclusion of SOC etc. may improve some of the model fit metrics but have little impact when assessing annual SR.

Answer

Yes, looking and the annual model performance as you suggested (Figure 8), we can confirm that the models underestimate the mean annual SR. This occurred due to the underestimation at high SR values (summer time) and can be connected to the slopes being less than unity. However, our approach of connecting the R₀ with the intercept is able to adequately resolve the lower SR without overestimations and by this our conclusions on the cold- and warm-period model comparisons  and the better combined model selection are still held intact. We updated the following text to clarify the issue

Lines:  484-487. It should be noted that the (intercept → 0+) approach which we developed guarantees an adequate estimation of the cold-period SR and good magnitude resolution of the model results (confirmed by Figure 8) in comparison to the measurements – an often observed inefficiency of the standard parameterization approaches [20, 31, 42].

Thank you for your comments. As you suggested we investigated this large scattering issue on the near zero temperature SR values (Figure 3) and found that it can be partially explained by the amount of precipitation at the respective months and the soil properties. They also can be associated with the snow period and the soil freezing-thawing cycles introducing noise in the measurements. The models follow the temperature trends and don't reproduce this noise. We have already discussed this in the following updated text

Lines:  492-496. The low R² values for the cold period (Table 2) are directly associated with the high variability of the observations (see Figure 3) typically occurred during the winter-time measurements due to snow presence on the ground and freezing-thawing cycles and also due to changes in precipitation causing CO₂ accumulation in soil and interrupting gas exchange.

Sincerely yours, Sergey Kivalov

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors, below are comments related to the review.

Paper entitled "Soil temperature, organic-carbon storage, and water-holding ability should be accounted for the empirical soil-respiration model selection in two forest ecosystems" by: Sergey Kivalov, Valentin Lopes de Gerenyu, Dmitry Khoroshaev, Tatiana Myakshina, Dmitry Sapronov, Kristina Ivashchenko and Irina Kurganova is clear and with minor additions can contribute to the understanding of soil respiration as the main component of the carbon cycle in terrestrial ecosystems, whereby when evaluating the optimal model of soil respiration SOC content and water retention capacity should be selected for complete analysis. Since the assessments were made on two types of soil, further research should be directed to other types of soil, which would complete the research and contribute to a wider application.

On the basis of which research was the type of soil determined, and who and when conducted the pedological research?

It is necessary to state the applied test methods shown in Figure 1, and link them to the references from which the methods derive.

The vast majority of documented references, out of a total of 60, are from older publications (before 2010) which should be reviewed and checked for more recent relevant sources.

Citation of references should be reviewed and harmonized with proper, uniform citation.

Author Response

Dear reviewer, thank you so much for your suggestions! Yes we will continue our investigation to include more different soils and ecosystems in the research. For this research we focus on these two 25-year long records to show the feasibility of the conducted study and its approach to the empirical modeling.

On the basis of which research was the type of soil determined, and who and when conducted the pedological research?

To determine type of soils the full soil observations were conducted in 2001-2002 (the first period of our study). The observation included: morphological description of soil profiles, granulometric and basic chemical analyses. The C and N stocks were also determined in 2010 and 2016. Being pedologist according to master and PhD diploma, Dr., Prof. Irina Kurganova was responsible for the soil part of our research.

For this study we use published data (see, please footprints for in the Table 1):

1. Kurganova I.N., V.O. Lopes de Gerenyu, T.N. Myakshina, D.V. Sapronov, I.V. Romashkin, V.A. Zhmurin, and V.N. Kudeyarov Native and model assessment of Respiration of forest sod-podzolic soil in Prioksko-Terrasny Biospheric Reserve. Contemporary Problems of Ecology, 2020, Vol. 13, No. 7, pp. 813–824
2. Kurganova I.N.; Lopes De Gerenyu V.O.; Myakshina T.N.; Sapronov D.V.; Savin I.Y.; Shorohova E.V.; Carbon balance in forest ecosystems of southern part of Moscow region under a rising aridity of climate, Contemporary problems of ecology 2017; Volume 10(7), pp. 748–760
3. Kurganova, I.; Lopes de Gerenyu, V.; Khoroshaev, D.; Myakshina, T.; Sapronov, D.; Zhmurin, V. Temperature Sensitivity of Soil Respiration in Two Temperate Forest Ecosystems: The Synthesis of a 24-Year Continuous Observation. Forests 2022; 13, 1374

It is necessary to state the applied test methods shown in Figure 1, and link them to the references from which the methods derive.

We updated the following text describing the SR measurement method steps with the reference source.

Lines: 144-157.  For the current research, we use 25-year-long SR-measurement time series conducted by the chambers of closed type (SR –­ top, Figure 2) once a week. The standard chamber-measurement approach is described in [20]. Firstly, the repetitive with 10-min intervals gas measurements using syringe-sample collection are conducted at sites at 5 nearby locations to account for the SR heterogeneity. Secondly, these samples are analyzed for CO₂ concentrations in laboratory using gas chromatograph. The obtained changes in CO₂ concentration are recalculated into the SR fluxes by applying the chamber volumetric correction [19, 22]. Simultaneously with the SR, the soil temperature at 5-cm depth and the air temperature at 1-m height were measured at sites (T_soil – brown, T_air – red; middle, Figure 2). The monthly averaged data for the air temperature (T_air) and precipitation (Prec) was collected from the Complex Background Monitoring Station, which is situated nearby the coniferous-deciduous forest site (blue; bottom, Figure 2). All the data was quality checked and prepared averaging on the monthly base to be fitted into the models.

The vast majority of documented references, out of a total of 60, are from older publications (before 2010) which should be reviewed and checked for more recent relevant sources.

Thank you we attempt to review the references to include the more recent sources. It should be noted that about 40% of the references are after 2010 year.

Citation of references should be reviewed and harmonized with proper, uniform citation.

Thank you we attempt to harmonize the citations.

Sincerely yours, Sergey Kivalov

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

My comments have been addressed. 

There are still editorial issues - even in the text they have changed.