Stability of Ecosystem CO2 Flux in Response to Changes in Precipitation in a Semiarid Grassland
Round 1
Reviewer 1 Report
Dear Editor & Authors
Overall this is an interesting paper, studying long-term CO2 respiration responses on altered climatological conditions (temperature and precipitation) in China. In general, the paper is well written and logically structured. However, I would like to make following suggestions:
It would be good to add an overview map (of a part of China/ / East Asia) indicating the location of the study sites as well as a climatogram.
I can see you have used an ANOVA to test some differences. However, I’m wondering if you tested your data on normality by using for example a kolmogrov-smirnov test? If this appears not to be the case you will have to use a non-parametric alternative.
In order to highlight long-term overall impacts across the treatments it would be better to express the CO2 efflux respiration in a cumulative way (e.g. fig. 3). Please check the international literature, because this is a commonly used.
I’m not sure if you did also measure the SOC content/stock? But it would be a great idea to do so, because that way you can express your soil CO2 respiration rate per soil organic carbon mass (so, µg respired C / g soil C). This is actually a commonly used method (see international literature) to express and investigate the degree of stability of the soil organic carbon stored in your system.
The English seems OK, but certainly can be further improved. Unfortunately, I cannot help the authors with this task as I’m not a native speaker, but please consult an English-speaking colleague or a language editorial office to help you further improving the English language level of the paper.
Author Response
Overall this is an interesting paper, studying long-term CO2 respiration responses on altered climatological conditions (temperature and precipitation) in China. In general, the paper is well written and logically structured. However, I would like to make following suggestions:
Response: Thanks for your positive comments.
It would be good to add an overview map (of a part of China/ / East Asia) indicating the location of the study sites as well as a climatogram.
Response: We added a map to indicate the location of the study site (Fig. 1)
I can see you have used an ANOVA to test some differences. However, I’m wondering if you tested your data on normality by using for example a kolmogrov-smirnov test? If this appears not to be the case you will have to use a non-parametric alternative.
Response: We tested the normality using Shapiro-Wilk test. The data follow normal distribution. We added this information in Materials and methods section as: We tested normal distribution (Shapiro-Wilk test) and homogeneity of variances (Levene test) for each metric prior to statistical analysis. (Lines 185-186)
In order to highlight long-term overall impacts across the treatments it would be better to express the CO2 efflux respiration in a cumulative way (e.g. fig. 3). Please check the international literature, because this is a commonly used.
Response: Thanks for your suggestion. We have provided a seasonal averaged values of each CO2 flux in each treatment in Fig. 5, which followed the similar pattern with cumulative CO2 flux. So we did not add cumulative flux.
I’m not sure if you did also measure the SOC content/stock? But it would be a great idea to do so, because that way you can express your soil CO2 respiration rate per soil organic carbon mass (so, µg respired C / g soil C). This is actually a commonly used method (see international literature) to express and investigate the degree of stability of the soil organic carbon stored in your system.
Response: We agree that present soil respiration per soil organic carbon would provide more useful information about soil OC dynamics, which was usually used for the soil incubation experiment. However, for the measurement about soil respiration in field condition, soil CO2 flux was largely contributed by root respiration, and present soil CO2 flux based on soil organic carbon content would result in wrong flux. So we did not present soil CO2 flux based on soil organic carbon.
The English seems OK, but certainly can be further improved. Unfortunately, I cannot help the authors with this task as I’m not a native speaker, but please consult an English-speaking colleague or a language editorial office to help you further improving the English language level of the paper.
Response: Thanks for your suggestion. We have asked language help from Nature Research Editing Service to improve the English of this MS.
Reviewer 2 Report
Review of Bao et al.
The paper by Bao et al describes the results of a 3 year field experiment that assessed the effects of altered rainfall on the CO2 flux of a Chinese semi-arid grassland. The experimental design appears to be sound, and the quality of thee data appear to be high. However, the manuscript needs major revision before it is potentially acceptable for publication.
The statistical analysis requires more information before the author’s interpretations can be accepted. First, it is unclear why the authors used a 2-way ANOVA when the experiment was a randomized block design with repeated measures. At a minimum the authors should have used a repeated-measures ANOVA to appropriately deal with time variable and the repeated measures. Second, the authors graciously provided the numerator degrees of freedom for their ANOVAs, but unfortunately omitted the denominator (error) degrees of freedom, making it impossible to determine whether the error mean square, and thus, F was calculated correctly. Failing to report both the numerator and error df renders the ANOVA results worthless, and the reader is left to trust that the authors calculated the MS and F correctly. Third, the rationale for selecting the various regression equations was lacking. Why were quadratic and/or linear regressions conducted? Was the point to fit curves or to learn something more about how temperature and moisture effect ecosystem physiology? If the former, given the scatter of points (Fig. 6), just about any model would probably explain the variability as well as any other, which really does not help with understanding how moisture and temperature affect CO2 flux. I suggest that the authors spend a little more time looking for models that have a well-known physiological basis (e.g., hyperbolas) so we can understand how the treatments affected ecosystem physiology.
The authors insist that there was no direct effect of their precipitation treatments on CO2 flux, but there appears to be a statistically significant interaction between treatment and season. So, just because there was not significant direct effect does not mean that there was no effect. In fact, if there is a statistically significant interaction, the direct effect, whether significant or not, does not matter. Also, there appears to be differences in the model coefficients that indicate that the response of a given CO2 flux component to temperature (moisture) is not independent of the effects of moisture (temperature). There are ways to test if two or more like equations have coefficients that are statistically different, and I suggest that the authors use these statistical tests to determine whether they response function differ in response to their treatments.
The figure quality was poor, with missing axes, units, and information on the sample size used to calculate the standard errors. There were also lapses in English grammar, which made the paper hard to understand. Please improve the figures and English grammar before resubmission.
Detailed comments
L25: which components?
L35: Vulnerable in terms of what?
L43: “the” focus or “a” focus?
L59-60: This statement makes no sense.
L70: which CO2 flux components?
L135-163: please comments on the temperature increase during the 2.5 min measurement period and the attenuation of PAR by the chamber. Also, was there any fogging of the chamber during measurement?
Fig. 1: The x-axis quality is poor and incomplete.
L1156: why a quadratic function rather than some other function that has a better theoretical understanding (hyperbola)? Was this chosen for the best fit? Quadratic functions are difficult to interpret and have very little physiological meaning.
L173: Why a linear equation? Is there are theoretical argument why a linear equation should be used here?
L178-180: Due to the repeated measurement of the plots, a repeated-measures ANOVA is more appropriate. Also, the experimental design is a randomized block. It would be good to see if there was any effect of blocking.
Fig. 2: Why dies the new year start in March?
Table 1 (and all ANOVA tables); Please display the error degrees of freedom for all direct effects and interactions.
Fig. 3 (an d all figs with standard errors): Since the SE = SD/ square-root of n, the sample size (n) must also be displayed.
L217-219: Assuming that the p-values are correct, there appeared to be an interaction between rainfall treatment and season for NEE (Table 1). So, the treatment effect depended on season, which is an interesting result. Why?
L236-240: so did CO2 flux decrease due to a lower temperature or higher soil moisture? There is undoubtedly an interaction between soil moisture and temperature, but it is unknown because no test was dine to see in the model coefficients were significantly different between the treatments.
Tables 2 and 3: I am curious why these models were selected and whether there were any significant differences between coefficients. This would illustrate a temperature vs moisture interaction, which is the whole point of this experiment. Also, it is hard to believe that with such low R2 values that the p-values for these models were < 0.01 (see Table 3). What was the sample size (n) for these regressions? Were all data for each treatment lumped together?
Fig. 6: again, unclear why certain models and/or transformations were used because it seems like several models could equally explain the scatter. Were all data per treatment pooled?
L268-270: yes, I agree. But were coefficients significantly different?
L293: more negative NEE does not equal low NEE.
L294-295. This statement makes no sense.
L302-305: no you didn't demonstrate this because there appears to be an interaction between rainfall treatment and season.
L312: This statement makes no sense.
L314-316: so what? If soil moisture was low in the DP treatment it still could have caused a decline in GEP, growth, etc. even though soil moisture was within the "normal range" of values.
L320-322: but is water is limiting, higher soil temperature will exacerbate this limitation. Also, GEP may be more sensitive to air temperature than soil temperature
L323-324: This statement makes no sense.
L326-327: GEP only explained < 50% of the variability in Re and <18% of the variability in Rs. There appeared to be an interaction between T and soil moisture (thus, rainfall)
L344-345: Interaction between treatment and season?
L349-351: so rainfall did or did not affect CO2 flux?
L353: Temperature limitation? Too high or too low?
L354-356: This statement makes no sense.
L366-367: yes!
Author Response
The statistical analysis requires more information before the author’s interpretations can be accepted. First, it is unclear why the authors used a 2-way ANOVA when the experiment was a randomized block design with repeated measures. At a minimum the authors should have used a repeated-measures ANOVA to appropriately deal with time variable and the repeated measures. Second, the authors graciously provided the numerator degrees of freedom for their ANOVAs, but unfortunately omitted the denominator (error) degrees of freedom, making it impossible to determine whether the error mean square, and thus, F was calculated correctly. Failing to report both the numerator and error df renders the ANOVA results worthless, and the reader is left to trust that the authors calculated the MS and F correctly. Third, the rationale for selecting the various regression equations was lacking. Why were quadratic and/or linear regressions conducted? Was the point to fit curves or to learn something more about how temperature and moisture effect ecosystem physiology? If the former, given the scatter of points (Fig. 6), just about any model would probably explain the variability as well as any other, which really does not help with understanding how moisture and temperature affect CO2 flux. I suggest that the authors spend a little more time looking for models that have a well-known physiological basis (e.g., hyperbolas) so we can understand how the treatments affected ecosystem physiology. The authors insist that there was no direct effect of their precipitation treatments on CO2 flux, but there appears to be a statistically significant interaction between treatment and season. So, just because there was not significant direct effect does not mean that there was no effect. In fact, if there is a statistically significant interaction, the direct effect, whether significant or not, does not matter.
Response: (1) Thanks for your suggestion about using repeated measurement ANOVA. When we included date into the model, the effects of precipitation and season and their interaction were similar to 2-way ANOVA. We replaced the original Table 1 with this new Table. We used block as random effect when conducting the ANOVA.
(2) We added the degrees of freedom of the model and error for Tables 1 and S1.
(3) We selected various regression equations for different metrics based on previous findings by other researchers in the same research topic. We added citations to support our selection (Lines 161, 169 and 178). We agree that using physiological basis models could provide understandings on the ecosystem physiology, however, to make our results comparable with others, we kept using these regression equations.
(4) We agree your comments. We found that most of the CO2 flux was not affected by precipitation treatment, except NEE that was affected by the interaction between precipitation and season. We described this interaction in the text, and revised throughout the text that most component of CO2 flux was not affected by precipitation. (Lines 22, 217, 235, 318 and 320)
Also, there appears to be differences in the model coefficients that indicate that the response of a given CO2 flux component to temperature (moisture) is not independent of the effects of moisture (temperature). There are ways to test if two or more like equations have coefficients that are statistically different, and I suggest that the authors use these statistical tests to determine whether they response function differ in response to their treatments.
Response: We selected these equations according previous reports for each specific component of CO2 flux. In our study, there are 2 of 32 fittings were not significant (P=0.09 and 0.06) (Tables 2 and 3), however, the parameters of these fittings are comparable with other researchers. In Tables 2 and 3, we provided the SE of each estimate so that we can test if these estimate were statistically differ among treatments. We added lower case letters for each estimate to indicate whether they are statistically differ among treatment. We added how we examine these differences in the notations of Tables 2 and 3.
The figure quality was poor, with missing axes, units, and information on the sample size used to calculate the standard errors. There were also lapses in English grammar, which made the paper hard to understand. Please improve the figures and English grammar before resubmission.
Response: We revised the figures to improve the quality. We asked language help from Nature Research Editing Service to improve the English.
Detailed comments
L25: which components?
Response: We clarified this as: GEP, RE and Rs, however, decreased with soil moisture. (Lines 25-26)
L35: Vulnerable in terms of what?
Response: We added in the last half sentence that they are vulnerable to global climate change. (Lines 35-36)
L43: “the” focus or “a” focus?
Response: We revised this to “a” focus. (Line 42)
L59-60: This statement makes no sense.
Response: We removed “, one of the most influential factors for CO2 flux due to precipitation treatment,”.
L70: which CO2 flux components?
Response: We clarified this as: We also hypothesized that increased precipitation would enhance the dependence of CO2 flux components (photosynthesis and respiration) on temperature. (Lines 69-70)
L135-163:please comments on the temperature increase during the 2.5 min measurement period and the attenuation of PAR by the chamber. Also, was there any fogging of the chamber during measurement?
Response: We added following information in the Materials and Methods section: Our measurement showed that change in air temperature inside the chamber was less than 0.2 ℃ during the 2.5 min measurement, and thus the change in air temperature could be neglected. The attenuation of PAR by the chamber was less than 5%. There was no fog on the chamber during the measurement. (Lines 138-141)
Fig. 1: The x-axis quality is poor and incomplete.
Response: We revised the x-axis to improve the quality of this figure.
L1156: why a quadratic function rather than some other function that has a better theoretical understanding (hyperbola)? Was this chosen for the best fit? Quadratic functions are difficult to interpret and have very little physiological meaning.
L173: Why a linear equation? Is there are theoretical argument why a linear equation should be used here?
Response to above two comments: We agree that using physiological basis models could provide understandings on the ecosystem physiology, however, to make our results comparable with others, we kept using these quadratic equations for GEP and NEE, exponential equations for Re and Rs to fit the relationships to temperature, and linear equations for relationships of CO2 flux to soil moisture, respectively.
L178-180: Due to the repeated measurement of the plots, a repeated-measures ANOVA is more appropriate. Also, the experimental design is a randomized block. It would be good to see if there was any effect of blocking.
Response: Thanks for your suggestion about using repeated measurement ANOVA. When we included date into the model, the effects of precipitation and season and their interaction were similar to 2-way ANOVA. We replaced the original Table 1 with this new Table. We used block as random effect when conducting the ANOVA.
Fig. 2: Why dies the new year start in March?
Response: We reported the results during growing season. The growing season starts in March.
Table 1 (and all ANOVA tables); Please display the error degrees of freedom for all direct effects and interactions.
Response: We added the error degrees of freedom in the notation of Table 1 and Table S1. (Table 1: Degree of freedoms of the model, date, season, treatment and interaction were 8, 1, 2, 2 and 4, respectively. Degree of error was 679. Table S1: Degree of freedoms of the model and error were 3 and 646, respectively. Lines 233-234, 391)
Fig. 3 (an d all figs with standard errors): Since the SE = SD/ square-root of n, the sample size (n) must also be displayed.
Response: We added sample size for Fig. 4. (The sampling size for each treatment is 4. Lines 214-215).
L217-219: Assuming that the p-values are correct,there appeared to be an interaction between rainfall treatment and season for NEE (Table 1). So, the treatment effect depended on season, which is an interesting result.
Response: We described the interactive effect of treatment and season on NEE here (NEE decreased (less negative) in the early stage of the season (+15 and +19%) but increased (more negative) in the late stage of the season (-15 and -19%) in both precipitation treatments. Lines 221-223).
L236-240: so did CO2 flux decrease due to a lower temperature or higher soil moisture? There is undoubtedly an interaction between soil moisture and temperature, but it is unknown because no test was dine to see in the model coefficients were significantly different between the treatments.
Response: Yes, they do have interaction in influence CO2 flux (Table S1). We discussed the interaction between soil moisture and temperature in Discussion section. (Lines 381-387)
Tables 2 and 3:I am curious why these models were selected and whether there were any significant differences between coefficients. This would illustrate a temperature vs moisture interaction, which is the whole point of this experiment. Also, it is hard to believe that with such low R2 values that the p-values for these models were < 0.01 (see Table 3). What was the sample size (n) for these regressions? Were all data for each treatment lumped together?
Response: We used commonly used regression equations for specific CO2 flux by other researchers in the same research topic. We pooled all the measurements together to fit these equations. We added sampling size in Tables 2 and 3. (The sampling size for each treatment is 232. Lines 290 and 297-298).
Fig. 6:again, unclear why certain models and/or transformations were used because it seems like several models could equally explain the scatter. Were all data per treatment pooled?
Response: We used commonly used regression equations for specific CO2 flux by other researchers in the same research topic. We used all the measurement to generate this Fig.
L268-270: yes, I agree. But were coefficients significantly different?
Response: We provided the P values for each comparison in the text to indicate whether they are significant or not. We also provided SE of each estimate in Table 2 and 3, which make us to evaluate whether the difference among treatment are significant or not. Please see our response to your overall comments.
L293: more negative NEE does not equal low NEE.
Response: We revised this as: produce more negative NEEs. (Line 305)
L294-295. This statement makes no sense.
Response: We removed the last half sentence “providing evidence that respiration was driven by photosynthesis in this grassland ecosystem”.
L302-305:no you didn't demonstrate this because there appears to be an interaction between rainfall treatment and season.
Response: We revised this as: These results indicated that most components of CO2 flux were relatively stable in this semiarid grassland across the various precipitation scenarios, supporting our first hypothesis. (Lines 319-321)
L312: This statement makes no sense.
Response: This statement is the basis for our explanation, so we kept this.
L314-316:so what? If soil moisture was low in the DP treatment it still could have caused a decline in GEP, growth, etc. even though soil moisture was within the "normal range" of values.
Response: Soil moisture will not cause a decline in GEP, growth, etc. if soil moisture was within the "normal range" of values.
L323-324:This statement makes no sense.
Response: This is an ending sentence of this paragraph, so we kept this sentence.
L326-327:GEP only explained < 50% of the variability in Re and <18% of the variability in Rs. There appeared to be an interaction between T and soil moisture (thus, rainfall)
Response: Yes. We provided the ANOVA for the interactive effects of temperature and soil moisture in Table S1. We discussed such interaction in the late part of the Discussion section: In other words, CO2 flux components depend more on temperature (higher slope of the response function) when soil moisture was higher (in IP treatment) but less on temperature (lower slope of the response function) when soil moisture was lower (in DP treatment), compared with AP treatment (Table 2). This interaction could be due to the co-limitation of temperature and moisture in the ecosystem. (Lines 383-387)
L344-345: Interaction between treatment and season?
Response: Yes, our results showed such an interaction (Table 1). However, here we want to say that NEE and GEP had similar response pattern to precipitation changes rather than the interaction. We revised this as: Furthermore, NEE and GEP had similar response pattern to precipitation changes in early and late season (Fig. 5). (Lines 362-363)
L349-351: so rainfall did or did not affect CO2 flux?
Response: We indicated in the last half sentence that rainfall did influence the dependence of CO2 flux on temperature. (Lines 369-370)
L353: Temperature limitation? Too high or too low?
Response: We revised this as: in a low-temperature limited ecosystem. (Line 372)
L354-356: This statement makes no sense.
Response: We revised this as: This result was consistent with other findings that increases in temperature weakened the temperature limitation in temperate ecosystems, or those at high latitude or altitude. (Lines 372-374)
L366-367: yes!
Response: Thanks.
Reviewer 3 Report
This study tested the response of ecosystem CO2 flux in a semiarid grassland to increasing and decreasing precipitation. Ecosystem CO2 flux measurements were undertaken during the growing season for a period of 3 years. The results showed that precipitation changes only had a small impact on ecosystem CO2flux. However, the study revealed a shift in the response to soil temperature and soil moisture as a result of precipitation changes.
The manuscript provides novel insights into the direction and magnitude of the ecosystem CO2flux response to increasing and decreasing precipitation changes in semiarid/arid grasslands, which is important to predict future responses of ecosystem CO2fluxes to climate change.
The research is mostly well described and presented. My main concerns are related to the response of ecosystem CO2 fluxes to soil moisture, which is an important part of this manuscript. The authors discuss in great depth how CO2 fluxes respond to soil moisture changes, however looking at Fig. 6 as well as 3 there is a lot of scatter and I think there needs to be at least some discussion about the uncertainty related to that. Further, the authors used different fits for different response functions and it is not clear how these were chosen (best fit and if so were different functions tested? based on previous findings?).
Specific comments
L80 20%, 35% were these numbers chosen randomly?
L131 bare soil: was vegetation removed? If so how and when?
L235 Did you test different fits?
L245 this should be Fig. 6
L397 This needs to be interpreted with caution given the large scatter
Fig. 1 The labels on the x-axis are not all visible
Author Response
This study tested the response of ecosystem CO2 flux in a semiarid grassland to increasing and decreasing precipitation. Ecosystem CO2 flux measurements were undertaken during the growing season for a period of 3 years. The results showed that precipitation changes only had a small impact on ecosystem CO2 flux. However, the study revealed a shift in the response to soil temperature and soil moisture as a result of precipitation changes.
The manuscript provides novel insights into the direction and magnitude of the ecosystem CO2flux response to increasing and decreasing precipitation changes in semiarid/arid grasslands, which is important to predict future responses of ecosystem CO2fluxes to climate change.
Response: Thanks for your positive comments.
The research is mostly well described and presented.My main concerns are related to the response of ecosystem CO2 fluxes to soil moisture, which is an important part of this manuscript. The authors discuss in great depth how CO2 fluxes respond to soil moisture changes, however looking at Fig. 6 as well as 3 there is a lot of scatter and I think there needs to be at least some discussion about the uncertainty related to that. Further, the authors used different fits for different response functions and it is not clear how these were chosen (best fit and if so were different functions tested? based on previous findings?).
Response: (1) We discussed the uncertainty about the scatters in the response of CO2 flux to soil moisture in Materials and Methods section as: In this study, although there are lots of scatters in soil moisture, nearly all the fittings of CO2 flux to soil moisture were significant at P<0.05. Therefore, our fittings reflected the general response pattern of CO2 flux to soil moisture. However, we recommended that such measurement should be intensified to avoid scatters. (Lines 187-190)
(2) We selected different response functions for different CO2 flux components based on previous findings by other researchers. We added citations to support our selection (Lines 161, 169 and 178).
Specific comments
L80 20%, 35% were these numbers chosen randomly?
Response: We designed to increase and decrease precipitation by 30% of annual precipitation. However, our measurement showed that the precipitation in DP treatment was 35% lower than AP, while that in IP were 20% increased than AP. Here we reported the measured percent changes of precipitation.
L131 bare soil: was vegetation removed? If so how and when?
Response: It is not bare soil. We revised this as: … and inserted 7 cm into the soil in each plot. The plant in the PVC base was removed by hand a week prior the start of the experiment to exclude the effects of plant respiration and photosynthesis on the Rs measurement. (Lines 130-132)
L235 Did you test different fits?
Response: We selected different response functions for different CO2 flux components based on previous findings by other researchers. So we did not test different fits. We added citations to support our selection (Lines 161, 169 and 178). Please see our response to your overall comments.
L245 this should be Fig. 6
Response: Thanks. We shifted the order of Fig. 6 and Fig. 7 (Figs 7 and 8 in the revised manuscription).
L397 This needs to be interpreted with caution given the large scatter
Response: We discussed this in Material and Methods section (Please see our response to your overall comments, Lines 187-190). Here we added the detailed changes in soil moisture to support our statement as: The DP resulted in 16% decrease in soil moisture, strengthening the limitation of soil moisture. In contrast, IP resulted in 21% increase in soil moisture, which weakened the limitation of soil moisture and, thus, the dependence of GEP on soil moisture (Table 3). (Lines 409-411)
Fig. 1 The labels on the x-axis are not all visible
Response: We revised clarified the labels of all the Figs in the revised manuscript.
Round 2
Reviewer 2 Report
The authors have thoroughly revised the paper according to my suggestions. While I do not agree completely with the author's rationale for using polynomial functions to model their data, I do understand their desire to be consistent with previous studies. Furthermore, the statistical analyses on the model coefficients helps elucidate how precipitation changes alter CO2 flux response functions. There are some awkward phrases in the revised version, but I suspect that these will be addressed by good copy editing.
