The Influence of Burn Severity on Post-Fire Spectral Recovery of Three Fires in the Southern Rocky Mountains
Round 1
Reviewer 1 Report
General comments:
This was a well written paper, and the topic is of broad interest. The analysis was thorough in that the authors considered a range of variables that might influence vegetation resilience. The manuscript was well organised, and I could not see any major issues with the approach followed.
I’ve included some specific comments below to help the authors with their revision.
Specific comments:
Abstract:
Page 1, line 8. Suggest rewording this to: ‘…Landsat time series and random forest models…’
Page 1, lines 10-11. Severity still had a statistically significant impact of recovery though, so it might be worth including another sentence here.
Introduction:
Page 1, line 23. delete ‘then’.
Page 1, line 34. Suggest deleting the full stop in the middle of this line and replacing with a comma. ‘…compared to lower elevation forests, making them good …’
Page 2, line 43 (and line 47). I noticed the citations were not in order here.
Page 2, lines 52-53. Suggest rewording ‘..more completely eliminate the seedbank…’ to either ‘completely eliminate the seedbank…’ or simply ‘reduce the seedbank…’
Page 2, line 60. Replace ‘characterize’ with ‘characterizing’
Page 2, line 64. Replace ‘…percentage of trees that are killed…’ with ‘…percentage of vegetation that is killed…’
Page 2, lines 78 and 79. ‘…severity of fires over ≥405 hectares…’ and ‘…fires over ≥202 hectares…’
Page 2, lines 83-85. The definitions here for moderate and high severity are worded the same. Surely there is some difference in the degree of canopy mortality?
Page 3, Table 1. In the ‘Relationship to ecological damage’ column, the definitions for moderate and high severity are the same?
Table 1 caption. I believe acronyms should be written out in full in the table caption. What is USGS?
Materials and Methods:
Page 4, line 130. Reference to Table 1 here, should be Table 2.
Page 4, line 133-137. These sentences would really be better placed in the introduction section, or deleted to avoid repetition.
Page 4, Table 2 caption. ‘Stand Characteristics’ suggests you will present some characteristics regarding the forest stands?
Page 4, Table 2. Surely aspect varies somewhat over the 10,600 ha area? Maybe dominant aspect might be a better heading? Similarly, perhaps elevation should be presented as a range, rather than a single value?
I feel the statement at the bottom of Table 2 is a bit of a cop-out. As a reader I would be keen to see more detailed information about the fires. Perhaps this could be summarized here?
Page 4, line 139. Seems a bit strange to have to separate ‘Materials and Methods’ headings.
Page 4, line 143. The word ‘time-series’ was not hyphenated earlier. Please be consistent.
Page 4, line 158. I think ‘LST’ should read ‘LTS’, based on line 143. I noticed ‘LST’ was also used on lines 163, 164, 165.
Page 4, line 159. Perhaps this should be a new sub-heading (also applies to the other phases).
Page 4, line 160. This was already defined (line 95).
Page 5, line 170. ‘…allowed up to calculate…’ should read ‘…allowed us to calculate…’
Also, ‘timeseries’ was written as two separate words earlier.
Page 5, line 177. It is not clear what is meant by ‘distance 2’ and ‘distance 1’. I think some additional wording is needed here.
Page 5, line 180. What is PRISM data? This needs to be made clear for the reader, or at least some reference provided.
Page 6, Table 3. Is this table needed? I don’t think the variable names (abbreviations) are used elsewhere. This table could be replaced with some text (two sentences) simply describing the climate variables. In fact, Table 3 is not even cited in the text, so should be removed.
Page 6, line 206. Is the SRTM acronym needed here? I could not see these used elsewhere. Same applies to MFRI on line 210.
Page 6, line 221. Is the CART acronym needed here? I could not see these used elsewhere. Please avoid unnecessary use of acronyms, there are enough in this manuscript already!
Pages 5 and 6. It would have been nice if phases 2 and 3 could be combined somehow. That way it would give a better understanding of the relative contribution of climate variables as opposed to topography, burn severity, fire regimes and geological data. However, I realise this might be complex to do. Could selected climatic variables (e.g. mean and minimum temperature) be included in the Random Forest analysis?
Results:
Page 9, line 252. It is stated that there was no significant difference in NBR spectral values. Please provide the statistics to back this up.
Figure 2. B and C in this figure both have the heading of ‘2002 Fire’. I think sub-figure C should be for the 2006 fire? Additional information in the figure caption would help here.
Table 5. Please report consistent number of decimal places with the values in this table. E.g. 56.169 should be 56.17 for consistency. One decimal place would be fine for the figures in this table.
Page 10, line 272, 275, 276. The word ‘fire’ does not need to be capitalised. ‘Fire’ should be ‘fire’.
Page 10, line 272. The word ‘in’ is repeated in this sentence.
Table 6. There is an unnecessary row in this table – row four, (1999 fire with high severity) with a single value of 0.33. I think this can be deleted.
Table 6. Suggest the heading ‘Fired’ should be ‘Fire’
Page 12, line 289-290. Add a comma in this sentence (i.e. In the random forest analysis, burn severity….), and suggest joining the two sentences (i.e. ‘…(Table 7), indicating that burn severity…’)
Page 12, Table 7, caption. More detail is needed in this caption. What is ‘Forest Mag and Rate’? Also explain heading ‘mtry’ in the table or the caption. I think that you should also state that magnitude and rate refer to post-fire NBR recovery.
Page 12, line 289-291. These random forest analysis results conflict with earlier results. Here you state that burn severity plays a large role in resilience. However, on lines 251-254 it is stated that ‘When comparing NBR spectral values 15 years
post-fire, there was no significant difference in NBR spectral values among fire severities in individual fire events, indicating that burn severity may not be the most important variable in determining resilience.’ Similarly, the abstract states that burn severity doesn’t strongly impact recovery. These mixed messages are confusing to the reader.
Discussion:
Page 13, line 337. ‘…use of a percentages predisposes…’ should read ‘…use of percentages predisposes…’
Page 14, line 359-361. This sentence does not make sense to me and should be more clearly worded.
Page 14, line 360. Suggest re-wording to ‘..pine forest following three large, high-severity fires.’
There could be other factors (in addition to climate) that add to the variation between the three fires? Perhaps there are also biological factors that have not been considered that can also influence ecosystem recovery. For example, predation of regeneration, competitive interactions etc. While I agree that climate is likely the strongest driver here, I think it would be worthwhile mentioning that there could be other drivers that cannot be easily measured with remote sensing datasets.
Author Response
Reviewer 1:
General Comments:
Comment: This was a well written paper, and the topic is of broad interest. The analysis was thorough in that the authors considered a range of variables that might influence vegetation resilience. The manuscript was well organized, and I could not see any major issues with the approach followed.
Response: Thank you for your feedback. We have made the revisions you suggested below. We have included the changes in track changes.
Abstract:
Comment: Page 1, line 8. Suggest rewording this to: ‘…Landsat time series and random forest models…’
Response: Done.
Comment: Page 1, lines 10-11. Severity still had a statistically significant impact of recovery though, so it might be worth including another sentence here.
Response We have added this.
Introduction:
Comment: Page 1, line 23. delete ‘then’.
Response: Done.
Comment: Page 1, line 34. Suggest deleting the full stop in the middle of this line and replacing with a comma. ‘…compared to lower elevation forests, making them good …’
Response: Done
Comment: Page 2, line 43 (and line 47). I noticed the citations were not in order here
Response: We have corrected the citations.
Comment: Page 2, lines 52-53. Suggest rewording ‘..more completely eliminate the seedbank…’ to either ‘completely eliminate the seedbank…’ or simply ‘reduce the seedbank…’
Response: We went with reduce the seedbank.
Comment: Page 2, line 60. Replace ‘characterize’ with ‘characterizing’
Response: Done
Comment: Page 2, line 64. Replace ‘…percentage of trees that are killed…’ with ‘…percentage of vegetation that is killed…’
Response: Done
Comment: Page 2, lines 78 and 79. ‘…severity of fires over ≥405 hectares…’ and ‘…fires over ≥202 hectares…’
Response: Done.
Comment: Page 2, lines 83-85. The definitions here for moderate and high severity are worded the same. Surely there is some difference in the degree of canopy mortality?
Response: Thank you for catching this, we have changed it.
Comment: Page 3, Table 1. In the ‘Relationship to ecological damage’ column, the definitions for moderate and high severity are the same?
Response: Yes we just put it in a table to clarify.
Comment: Table 1 caption. I believe acronyms should be written out in full in the table caption. What is USGS?
Response: United States Geological Survey
Materials and Methods:
Comment: Page 4, line 130. Reference to Table 1 here, should be Table 2.
Response: We have made this change.
Comment: Page 4, line 133-137. These sentences would really be better placed in the introduction section, or deleted to avoid repetition
Response: We wanted to leave this a distinct section since it is a short section on the ecosystem and cites another source for more detailed information. This section isn’t related to Remote Sensing and we chose to leave it separate.
Comment: Page 4, Table 2 caption. ‘Stand Characteristics’ suggests you will present some characteristics regarding the forest stands?
Response: We retitled this to clarify that it is of burned forest stands in this paper.
Comment: Page 4, Table 2. Surely aspect varies somewhat over the 10,600 ha area? Maybe dominant aspect might be a better heading? Similarly, perhaps elevation should be presented as a range, rather than a single value?
Response: We have changed it to ‘dominant aspect’ and mean elevation.
Comment: I feel the statement at the bottom of Table 2 is a bit of a cop-out. As a reader I would be keen to see more detailed information about the fires. Perhaps this could be summarized here?
Response: This was clarified to reference more about the field validation post-fire regeneration. Lodgepole pine in this region are talked about extensively in the introduction.
Comment: Page 4, line 139. Seems a bit strange to have to separate ‘Materials and Methods’ headings.
Response: We chose to leave it this way. We believe it makes it easier to understand where the methods begin vs. where there is background on the study location.
Comment: Page 4, line 143. The word ‘time-series’ was not hyphenated earlier. Please be consistent.
Response: We made it not hyphenated.
Comment: Page 4, line 158. I think ‘LST’ should read ‘LTS’, based on line 143. I noticed ‘LST’ was also used on lines 163, 164, 165
Response: We corrected it to LTS.
Comment: Page 4, line 159. Perhaps this should be a new sub-heading (also applies to the other phases).
Response: We added a sub-heading.
Comment: Page 4, line 160. This was already defined (line 95).
Response: We removed the duplicate definition.
Comment: Page 5, line 170. ‘…allowed up to calculate…’ should read ‘… allowed us to calculate…’
Response: We corrected this.
Comment: Also, ‘timeseries’ was written as two separate words earlier.
Response: We made it two separate words.
Comment: Page 5, line 177. It is not clear what is meant by ‘distance 2’ and ‘distance 1’. I think some additional wording is needed here.
Response: We corrected this figure and made distance 1 and distance 2 more clear.
Comment: Page 5, line 180. What is PRISM data? This needs to be made clear for the reader, or at least some reference provided.
Response: We added the definition of PRISM (Parameter-elevation Regressions on Independent Slopes Model)
Comment: Page 6, Table 3. Is this table needed? I don’t think the variable names (abbreviations) are used elsewhere. This table could be replaced with some text (two sentences) simply describing the climate variables. In fact, Table 3 is not even cited in the text, so should be removed.
Response: We agree. We already had the variables listed in lines 183-185. We removed the table.
Comment: Page 6, line 206. Is the SRTM acronym needed here? I could not see these used elsewhere. Same applies to MFRI on line 210.
Response: We removed the acronyms.
Comment: Page 6, line 221. Is the CART acronym needed here? I could not see these used elsewhere. Please avoid unnecessary use of acronyms, there are enough in this manuscript already.
Response: We removed CART.
Comment: Pages 5 and 6. It would have been nice if phases 2 and 3 could be combined somehow. That way it would give a better understanding of the relative contribution of climate variables as opposed to topography, burn severity, fire regimes and geological data. However, I realize this might be complex to do. Could selected climatic variables (e.g. mean and minimum temperature) be included in the Random Forest analysis?
Response: We used climate in Random Forest in another paper Guz et al. 2021 that is cited in this paper that is based on field data. However, in this paper the climate variables in this paper are continuous and the topography, burn severity, fire regimes, and geological data are not continuous. We did think they could be combined in a meaningful way and chose to leave them in separate analysis.
Results:
Comment: Page 9, line 252. It is stated that there was no significant difference in NBR spectral values. Please provide the statistics to back this up.
Response: We removed significant. We included all spectral points and didn’t need to include the word significant.
Comment: Figure 2. B and C in this figure both have the heading of ‘2002 Fire’. I think sub-figure C should be for the 2006 fire? Additional information in the figure caption would help here.
Response: We have fixed this figure.
Comment: Table 5. Please report consistent number of decimal places with the values in this table. E.g. 56.169 should be 56.17 for consistency. One decimal place would be fine for the figures in this table.
Response: We have made this table consistent.
Comment: Page 10, line 272, 275, 276. The word ‘fire’ does not need to be capitalised. ‘Fire’ should be ‘fire’.
Response: We made it ‘fire’.
Comment: Page 10, line 272. The word ‘in’ is repeated in this sentence.
Response: We removed ‘in’.
Comment: Table 6. There is an unnecessary row in this table – row four, (1999 fire with high severity) with a single value of 0.33. I think this can be deleted.
Response: We are unclear why we would remove the R-squared for only the 1999 high severity fire. We left the 0.33.
Comment: Table 6. Suggest the heading ‘Fired’ should be ‘Fire’.
Response: Done
Comment: Page 12, line 289-290. Add a comma in this sentence (i.e. In the random forest analysis, burn severity….), and suggest joining the two sentences (i.e. ‘…(Table 7), indicating that burn severity…’)
Response: Done
Comment: Page 12, Table 7, caption. More detail is needed in this caption. What is ‘Forest Mag and Rate’? Also explain heading ‘mtry’ in the table or the caption. I think that you should also state that magnitude and rate refer to post-fire NBR recovery.
Response: We have corrected this to forest magnitude and rate which is discussed in section 2.2.1. We include ‘We used mtry, the number of variables randomly sampled as candidates at each split to parameterize’ to the table.
Comment: Page 12, line 289-291. These random forest analysis results conflict with earlier results. Here you state that burn severity plays a large role in resilience. However, on lines 251-254 it is stated that ‘When comparing NBR spectral values 15 years post-fire, there was no significant difference in NBR spectral values among fire severities in individual fire events, indicating that burn severity may not be the most important variable in determining resilience.’ Similarly, the abstract states that burn severity doesn’t strongly impact recovery. These mixed messages are confusing to the reader.
Response: Yes, we understand why this may not be clear. We have added a revision to clarify this. 295-301 where we clarify – “This makes since as ecologically lodgepole pine have prolific initial post-fire regeneration [6]. However, when comparing post-fire spectral values 15 fires post fire there isn’t a significant impact. Indicating that fire severity may impact regeneration following high severity fires. However, when put in context of the other results, fire se-verity may not impact long-term spectral reflectance.”
Discussion:
Comment: Page 13, line 337. ‘…use of a percentages predisposes…’ should read ‘…use of percentages predisposes…’
Response: Done
Comment: Page 14, line 359-361. This sentence does not make sense to me and should be more clearly worded.
Response: Done
Comment: Page 14, line 360. Suggest re-wording to ‘..pine forest following three large, high-severity fires.
Response: Done.
Comment: There could be other factors (in addition to climate) that add to the variation between the three fires? Perhaps there are also biological factors that have not been considered that can also influence ecosystem recovery. For example, predation of regeneration, competitive interactions etc. While I agree that climate is likely the strongest driver here, I think it would be worthwhile mentioning that there could be other drivers that cannot be easily measured with remote sensing datasets.
Response: Yes, there could be other variables that influenced regeneration besides climate, such as soil moisture, finer levels of geology, and other biological factors. We have added a statement before the conclusion.
Author Response File: 
Author Response.docx
Reviewer 2 Report
General comment
This manuscript focused on Post-fire spectral recovery of three fires in the Southern Rocky Mountains. I agreed the significance of this topic in remote sensing. The framework and presentation of this manuscript are easy to know and I have some special comments as follows.
Special comment
- Title and Abstract are suitable that could reflect whole text.
- I have three comments for introduction. First, here put formula that is not suitable (Line 76). Second, usually, Table 1 should omit in this chapter. Third, to add the study purpose in the end of this chapter is necessary.
- I suggest adding map in Study area chapter. It is too simple in this chapter. Data type should have more description, such as, Table 5 and 6, only 3 years (periods)? 1999, 2002 and 2006? However, Figure 1 is from 1995 to 2010? The data were not equal in above analysis. It should have present in detail.
- I still could not clearly understand why this study only choice1999, 2002 and 2006? Because now is in the end of 2021, the distance from now to 2006 is near 15 years.
Author Response
Reviewer 2:
General comments:
Comment: This manuscript focused on Post-fire spectral recovery of three fires in the Southern Rocky Mountains. I agreed the significance of this topic in remote sensing. The framework and presentation of this manuscript are easy to know and I have some special comments as follows.
Response: Thank you for your feedback. We responded to your comments below.
Special comment:
Comment: Title and Abstract are suitable that could reflect whole text
Response: We changed the title to include burn severity and added to the abstract.
Comment: I have three comments for introduction. First, here put formula that is not suitable (Line 76). Second, usually, Table 1 should P 0 WORDS omit in this chapter. Third, to add the study purpose in the end of this chapter is necessary.
- We are not sure what P 0 WORDS are in this chapter. So we didn’t omit anything from the title in Table 1.
- We agree the formula wasn’t necessary and was removed.
- We decided to leave the purpose of this paper. We find it a useful reading device for readers scanning the paper and link it when we are reading papers.
Comment: I suggest adding map in Study area chapter. It is too simple in this chapter. Data type should have more description, such as, Table 5 and 6, only 3 years (periods)? 1999, 2002 and 2006? However, Figure 1 is from 1995 to 2010? The data were not equal in above analysis. It should have present in detail.
Response:
- We have added a study area map
- Figure 1 is a theoretical framework to show how the % NBR was calculated and to show change in spectral reflectance before and after a fire. Table 1 is not resulting. It is purely an example. We have edited the figure to make this clearer.
- Tables 5 and 6 are results. Table 5 shows % recovered 5-, 10-, and 15-years post fire. Table 6 is again results showing how climate influences regeneration. We recorded all of the variables.
Comment: I still could not clearly understand why this study only choice1999, 2002 and 2006? Because now is in the end of 2021, the distance from now to 2006 is near 15 years.
Response: We added another sentence about this in the end and it was stated in the introduction. We only studied fires that had field validation in Guz et al. 2021. We wanted burned sites that weren’t logged or had evidence of a secondary disturbance. Additionally, we need at least 15 years to indicate how the fires are recovery. We originally had more fires, but when comparing the spectral results to field data we would identify evidence of secondary disturbances and remove those fires.
Author Response File: Author Response.docx
Reviewer 3 Report
This is a very interesting study about forest resilience of different burn severity. The experiment was very well presented and I only have a few minor considerations.
1) regarding burn severity: three levels of burn severity were defined, and low level severity was defined as burn with damaged under story herbaceous vegetation. If only understory vegetation is affected, it will be difficult to detect low level burns with optical sensors, and therefore the changes in NBR before and after low level burns may not be evident.
2) lines 83, 84, you have the same definition for moderate and high severity, please verify.
3) lines 118-119, and line 160, the definition of forest resilience was repeated several times. since you have already given the definition previously, it feels redundant.
4) Figure 1. “…Percent NBR change in 2010 was defined as distance 2 divided by distance 1 multiplied by 100”, please indicate in the figure what distance 1 and distance 2 refer to.
5) Figure 2, a typo in c; it is 2006 instead of 2002 fire.
6) Table 5, in both medium and low severity fire, the post-fire %NBR recovery increases with years, however, high severity fire recovery decreases with years, especially for 1999 and 2002 fire events. Please explain.
7) Table 6, regarding the significant explanatory variables for %NBR change, please indicate significance level. For example, the variable Intercept ppt in summer Season for high severity fire in 1999 had p-value of 0.09, it is not significant at 95% confidence level.
8) Table 7, The mtry value is missing for the model with response variable of Rate, and High severity. The Mae values are very different between Magnitude and Rate models, please provide explanation. Also please explain why some values in percent variance explained are negative.
Author Response
Reviewer 3:
General Comments:
Comment: This is a very interesting study about forest resilience of different burn severity. The experiment was very well presented, and I only have a few minor considerations.
Response: Thank you for your feedback. We have responded to your suggestions below.
Comment: regarding burn severity: three levels of burn severity were defined, and low level severity was defined as burn with damaged under story herbaceous vegetation. If only understory vegetation is affected, it will be difficult to detect low level burns with optical sensors, and therefore the changes in NBR before and after low level burns may not be evident.
Response: Yes, we agree with this. The degree in spectral reflectance was low for low severity fires.
Comment: lines 83, 84, you have the same definition for moderate and high severity, please verify.
Response: This has been corrected.
Comment: lines 118-119, and line 160, the definition of forest resilience was repeated several times. since you have already given the definition previously, it feels redundant.
Response: We removed the repeated definition.
Comment: Figure 1. “…Percent NBR change in 2010 was defined as distance 2 divided by distance 1 multiplied by 100”, please indicate in the figure what distance 1 and distance 2 refer to.
Response: We have corrected this.
Comment: Figure 2, a typo in c; it is 2006 instead of 2002 fire.
Response: We have fixed it.
Comment: Table 5, in both medium and low severity fire, the post-fire %NBR recovery increases with years, however, high severity fire recovery decreases with years, especially for 1999 and 2002 fire events. Please explain.
Response: We have added an explanation on lines 253 – 257. “ In Table 5, both medium and low severity fire, the post-fire %NBR recovery increases with years. This may be because the fire thinned out the stand and allowed other trees to grow more or because the fire was strong enough to release some seeds from serotinous cones. Lodgepole pine have thin bark and are note resistant to fire [6].”
Comment: Table 6, regarding the significant explanatory variables for %NBR change, please indicate significance level. For example, the variable Intercept ppt in summer Season for high severity fire in 1999 had p-value of 0.09, it is not significant at 95% confidence level.
Response: We added “significance level was based on 90% confidence level” at the end of Table 6.
Comment: Table 7, The mtry value is missing for the model with response variable of Rate, and High severity. The Mae values are very different between Magnitude and Rate models, please provide explanation. Also please explain why some values in percent variance explained are negative.
Response:
- Part 1: Thank you for catching that. We added the mtry value.
- Part 2: MAE is the mean absolute error. This measures the average magnitude of the errors in a set of predictions, without considering their direction. Magnitude and rate are measuring different attributes of the same event. Magnitude can be calculated for each segment based on the vertex time and spectral dimensions. This is similar to Figure 1. Also see image below. While rate is the trajectory of a pixel time series prior to disturbance segments that we’ve attributed to fire. The magnitude and the rate could be drastically different input numbers to our random forest model.
- Part 3: Typically, a negative % variance is evidence of a poor model and there likely aren’t any strong predictive relationships. Alternatively, there may be nonlinear relationships between a model and the response. A simple interpretation of a negative R² (rsq), is that you are better off predicting any given sample as equal to overall estimated mean, indicating very poor model performance. The models consistently have negative percent variance explained for medium and low severity fires, indicating that there may other variables influencing low and medium burn severity stands.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Comment
In this revised version, some of my suggestions have improved. I only have two comments for authors.
- Abstract is too simple. It should add more information here, such as background information, importance and results.
- To put formula and Table in introduction is not suitable because usually introduction contents background information, importance and study purpose.
- Figure 1. Map of study area should add scale and source (citation)
- In Figure 3, the title and number of X and Y axis are too small, it should be improved.
Author Response
Comment 1: Abstract is too simple. It should add more information here, such as background information, importance, and results.
Response 1: We have elaborated in the abstract to include information about the study area, background on spectral recovery, and highlight important results.
Comment 2: To put formula and Table in introduction is not suitable because usually introduction contents background information, importance and study purpose.
Response 2: We have removed the formula.
Comment 3: Figure 1. Map of study area should add scale and source (citation)
Response 3: We have added a scale We made this map and so do not need a unique citation.
Comment 4: In Figure 3, the title and number of X and Y axis are too small, it should be improved.
Response 4: We have made the title, number of X and Y axis larger.
Author Response File: Author Response.docx
