the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Jul 2024
Ecosystem leaf area, gross primary production, and evapotranspiration responses to wildfire in the Columbia River Basin
Abstract. Wildfires impact vegetation mortality and productivity and are increasing in intensity, frequency, and spatial area in the western United States. The rates of vegetation recovery after fires play a major role in the reestablishment of biomass and ecosystem functioning (e.g., structure, resilience, and productivity), but such recovery rates are poorly understood. Here we use remotely sensed data products from the Moderate Resolution Imaging Spectroradiometer (MODIS) to quantify the resistance and resilience of leaf area index (LAI), gross primary production (GPP), and evapotranspiration (ET) to 138 wildfires with various burn severity across the Columbia River Basin of the Pacific Northwest in 2015. Increasing burn severity caused lower resistance and resilience for all three variables. Resistance and resilience are highest in grasslands, intermediate in savanna, and lowest in needleleaf evergreen forests, consistent with the adaptation of these vegetation types to fire. LAI has consistently lower resistance and resilience than GPP and ET, which is consistent with physical and physiological mechanisms that compensate for reduced LAI. Resilience is influenced by precipitation, vapor pressure deficit (VPD), and burn severity across all three vegetation types, however, burn severity plays a more minor role in grasslands. Increasing wildfire severity will reduce the resistance and resilience and lengthen the recovery time of vegetation structure and fluxes with climate change, with significant consequences on the provision of ecosystem functioning and implications for model predictions.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(0 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-2217', Tempest McCabe, 04 Sep 2024
General comments:
The paper “Ecosystem leaf area, gross primary production, and Evapotranspiration Responses to Wildfire in the Columbia River Basin” examines how different vegetation types recovered after a severe fire season during 2015, looking at MODIS LAI, GPP and ET products combined with the MBTS burn serenity classification data. The authors targeted concepts of “resistance” and “resilience” by looking at the difference in LAI, GPP, and ET in years following 2015 compared to the interannual variability of years prior to 2015. The authors then use a random forest feature importance to determine if vapor pressure deficit, precipitation, or burn severity were most important in determining how resilient a vegetation type was. They found that grasses were more resistant and resilient than savannahs and needleleaf evergreen forests, and that LAI was more responsive to burn severity than GPP and ET.
I found this paper to address important scientific concepts. The authors’ application of concepts of resilience and resistance was an important and interesting lens to explore how disturbances affect landscapes. I also found the paper to have some major flaws, specifically pertaining to how the authors document uncertainty within their analyses, and the independence of the MODIS LAI, GPP, and ET products. I identified minor flaws in some elements of how the analysis was scoped and discussed.
Notably, the materials I received for review had no supplemental materials, but the paper makes reference in the text to several supplemental elements. The text will need to be edited substantially if the supplement was intentionally excluded.Major comments:
This paper lacks discussion of uncertainty in interpreting the results presented. For example, in Figure 2 the only delineation of uncertainty is that of interannual variability. It’s unclear if that interannual variability is mean to be interpreted as the threshold for “significant difference”, and if that is the case it is in contradiction with some of the writing (see comments for line 355). A single number to describe a year gave me pause as a reviewer. Similarly, Figures 3 and figures 4 present single-point comparative data with nothing to contextualize the uncertainty around that single point. This is particularly relevant for figures 3 and 4, because the range in inter-annual resistance and resilience presented in figure 4 is sometimes larger than the differences in resistance and resilience between different vegetation-types. Interpreting if the differences presented are meaningful is left to the reader, undercutting interesting patterns. In the text, discussion of uncertainty only acknowledges limitations of remotely-sensed LAI, and does not discuss the specific uncertainties and limitations unique to the paper itself. While more appropriate to the methods section, the text also does not discuss spatial uncertainty.
Second, it is unclear if the MODIS LAI, GPP, and ET are independent enough to draw separate conclusions. The authors wisely acknowledge in line 384 that the MODIS GPP and ET products use MODIS LAI as an input, along with meteorological data. However, this does call into question if the finding that LAI responds more to burn severity than GPP or ET, because GPP and ET definitionally contain the same information as the LAI product, plus other variables that don’t respond to burn severity. If the GPP and ET product contain information that could respond to burn severity beyond LAI, then that needs to be explicitly stated.
Minor comments:
Methods: Grasses could grow back faster than the “resilience”/ “resistance” increment of 1 year. Is it appropriate to consider grasses’ “resistance” in the first year post fire?
Line 85- Citation how LAI related to prediction uncertainties of earth system models?
Line 160-164 – This seems to contradict methods described in lines 193-195
Line 186 - Difficulty understanding the sentence beginning with “Specifically…” What information is applied? Is this referring to MODIS pixels that span multiple fires, or each fire-MODIS pair?
Line 195 - Table 1 seems more appropriate to a supplement. Perhaps a column indicating the years used in this study?
Line 245 - 248: Citation/ discussion for why 100 trained models were sufficient.
Lines 250-258: This paragraph reads as more relevant to the methods section, and the writing shifts from present tense to past tense.
Figure 1 - Also seems more appropriate for a supplement. More discussion/ visual delineation of uncertainty would make this a more powerful figure. For example, why not include error bars for the years themselves? Finally, while the text claims climate variations in years (is that distinct from interanual variability?) are not confounding, 2019 seems to be different than other years compared to the interannual variability of 2001-2014 in precipitation. This is not discussed, and specifically how years are not significantly different is not defined.
Line 355 - Seems to contradict Figure 2 – 2019 is higher than 2020 in GPP. Also, if “Recover” is defined as reaching the same value within the variability of 2011-2014, then many metrics did not “recover” at all.
Line 321-322 Cite statement about overfitting.
443 - The connection between this research and afforestation and sustainable agriculture is unclear in this current writing.
Conclusions: Potential uses for this research for the calibration of ESM is discussed for the first time in the conclusions.
Technical Corrections:Line 209: - Additional period.
Line 278: “S” denoting severity and “S” denoting supplement both is confusing.
Line 316 important scores -> importance scores
364 - fully - full
395- interacting -> interaction
425 “findings that obtained with” -> findings from, “potential of” -> “potential for”
428 - unclear what is meant by “reasonable data quality controls”
Citation: https://doi.org/10.5194/egusphere-2024-2217-RC1 -
AC1: 'Reply on RC1', Mingjie Shi, 12 Nov 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-2217/egusphere-2024-2217-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Mingjie Shi, 12 Nov 2024
-
RC2: 'Comment on egusphere-2024-2217', Anonymous Referee #2, 19 Sep 2024
General comments:
In "Ecosystem leaf area, gross primary productions, and evapotranspiration responses to wildfire in the Columbia River Basin" the paper compares metrics of resistance and resilience of the MODIS-derived products mentioned in title across 3 primary vegetation types that underwent different burn severities from 138 fires in 2015. Resistance is calculated as the ratio of the 2016 annual maximum value for LAI, GPP, and ET to the average annual max value from the 4 years preceding the fire (2011-2014). Resilience is ratio of the average annual max values of LAI, ET, and GPP during 4 post-fire years (2017-2020) to the average annual max values during the 4 pre-fire years. A random forest approach is used to compare the influence of precipitation, vapor pressure deficit and burn severity on resilience of the three vegetation types (needleleaf evergreen forest, woody savanna, and grasslands). The paper concludes that 1. increased burn severity decreases resistance and resilience of LAI, GPP, and ET across all vegetation types, 2. LAI resistance and resilience is the most impacted by burn severity, 3. grassland LAI, GPP and ET are more resilient and resistant to wildfire than that of evergreen forests and savannas, and 4. burn severity is the primary driver of resilience in evergreen forests and savannas, while VPD and precipitation determine grassland resilience.
This paper’s methods and results are thought provoking and likely of interest to many. In addition to sharing concerns regarding the uncertainty issues addressed by the first reviewer, I have identified other limitations that should be addressed. Namely, the resistance and resilience metrics in this study are produced using different data than in the key cited paper (DeSoto et al., 2020) and thus there must be clear differences in the interpretation of the results.
The short post-fire study period undermines discussion related to resilience. This paper claims that low resilience values relate to low chances of full recovery (438-439). However, a 4-year post-fire study period is too short to assess the likelihood of recovery when forest recovery can take decades. If the authors decided to retain this resilience metric, the interpretation needs to appropriately acknowledge and match the limitations of the method.
Additionally, the authors do not use DeSoto’s ecosystem recovery metric. At the very least, there should be discussion of why that is not used since they otherwise rely on DeSoto’s framework. DeSoto et al. 2020 define recovery as the ratio between the post disturbance value and the value during the disturbance, and thus resilience (post disturbance/pre disturbance) is a product of both resistance (disturbance/pre disturbance) and recovery (post disturbance/ disturbance). Based on figure 2, recovery of LAI, GPP, and ET appear quite similar across the three vegetation types following severe burn. Thus, observed differences in resilience are driven primarily by differences in resistance as opposed to recovery. I believe that it is key that this paper properly discuss the nuanced differences in interpreting recovery (seemingly similar across these ecosystems) and resilience and resistance (seemingly contrasting).
Additional Major Comments:
Burn severities are presented as a key driving factor but I could not find information (in either the article or SI) about the distribution of burn severities across the different primary vegetation types. This would provide useful context for figures 3 and 5. Also, there is no discussion concerning the issues with burn severity classifications in grasslands compared to forests. Are there differences in classifications between vegetation types that readers should consider? Do MTBS algorithms and workflows perform similarly well in grasslands, savannas and forests?
Vegetation classifications are based on a 2015 MODIS dataset. For those areas that burned in 2015, do more recent land cover type datasets suggest a possible change in vegetation cover type? The authors do not discuss possible land cover conversion following different severity burns. Indeed, in figure 3 savannas seem to respond like grasslands at high severities and forests at low severities. While the post-fire study period is not long enough to expect full forest or savanna recovery from severe burn, more explicit discussion on conversion is warranted.
Technical corrections:
61-62: change to: “greater reductions in ET to precipitation ratios (ET/P)”
63: Is “respectively” needed?
63-66: Please improve the wording of this sentence.
70: Start sentence with “There”.
83-85: Rearrange this sentence, possibly split into two sentences.
93: Replace “, experienced” with “across”
95 -105: It would be easier to read if the hypotheses were simply listed. The discussion of concerning results from previous studies may be better earlier in the introduction or later in the discussion section.
108-113: The final part of the sentence about increasing wildfire severity in the CRB should be another sentence.
361: Including “(e.g., grasses)” is misleading. DeSoto et al. 2020 compares gymnosperm and angiosperm tree species, there is no discussion about grassland recovery.
116: Unnecessary first sentence.
130: include “annual 500 m” before “MODIS” and delete the second sentence.
132-135: Combine these two sentences.
153: Replace “burnt” with “burn”
153-156: Start a new sentence after “respectively”.
158: Delete “in 2015”.
168: Replace “water” with “vapor”.
171: Replace “data sets” with “datasets”.
192-194: Strange wording and see reviewer 1’s comment.
Table 1: Agree with reviewer 1: move to SI. Also it’s VPD, not WPD.
209: Delete the extra period.
230: Capitalize Python.
252-254: This should be in the methods section.
Figure 1: Agree with reviewer 1: better in the SI.
310-312: Strange wording.
316-317: First half of the sentence is poorly worded.
356: Replace “till 2020” with “within the post-fire study period”.
364: Delete the “y” on “fully”.
376: Delete “the result reveals”.
378: Replace “the findings” with “previous work showing”.
395: Delete this first sentence.
396: “Though, burn severity is less important to grassland resilience,”
399: Use “interaction”.
407: Delete “across”.
438-439: How does this study imply that these ecosystems may have extremely low chances of full recovery? 4 years is not a very long post-year period.
Citation: https://doi.org/10.5194/egusphere-2024-2217-RC2 -
AC2: 'Reply on RC2', Mingjie Shi, 12 Nov 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-2217/egusphere-2024-2217-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Mingjie Shi, 12 Nov 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-2217', Tempest McCabe, 04 Sep 2024
General comments:
The paper “Ecosystem leaf area, gross primary production, and Evapotranspiration Responses to Wildfire in the Columbia River Basin” examines how different vegetation types recovered after a severe fire season during 2015, looking at MODIS LAI, GPP and ET products combined with the MBTS burn serenity classification data. The authors targeted concepts of “resistance” and “resilience” by looking at the difference in LAI, GPP, and ET in years following 2015 compared to the interannual variability of years prior to 2015. The authors then use a random forest feature importance to determine if vapor pressure deficit, precipitation, or burn severity were most important in determining how resilient a vegetation type was. They found that grasses were more resistant and resilient than savannahs and needleleaf evergreen forests, and that LAI was more responsive to burn severity than GPP and ET.
I found this paper to address important scientific concepts. The authors’ application of concepts of resilience and resistance was an important and interesting lens to explore how disturbances affect landscapes. I also found the paper to have some major flaws, specifically pertaining to how the authors document uncertainty within their analyses, and the independence of the MODIS LAI, GPP, and ET products. I identified minor flaws in some elements of how the analysis was scoped and discussed.
Notably, the materials I received for review had no supplemental materials, but the paper makes reference in the text to several supplemental elements. The text will need to be edited substantially if the supplement was intentionally excluded.Major comments:
This paper lacks discussion of uncertainty in interpreting the results presented. For example, in Figure 2 the only delineation of uncertainty is that of interannual variability. It’s unclear if that interannual variability is mean to be interpreted as the threshold for “significant difference”, and if that is the case it is in contradiction with some of the writing (see comments for line 355). A single number to describe a year gave me pause as a reviewer. Similarly, Figures 3 and figures 4 present single-point comparative data with nothing to contextualize the uncertainty around that single point. This is particularly relevant for figures 3 and 4, because the range in inter-annual resistance and resilience presented in figure 4 is sometimes larger than the differences in resistance and resilience between different vegetation-types. Interpreting if the differences presented are meaningful is left to the reader, undercutting interesting patterns. In the text, discussion of uncertainty only acknowledges limitations of remotely-sensed LAI, and does not discuss the specific uncertainties and limitations unique to the paper itself. While more appropriate to the methods section, the text also does not discuss spatial uncertainty.
Second, it is unclear if the MODIS LAI, GPP, and ET are independent enough to draw separate conclusions. The authors wisely acknowledge in line 384 that the MODIS GPP and ET products use MODIS LAI as an input, along with meteorological data. However, this does call into question if the finding that LAI responds more to burn severity than GPP or ET, because GPP and ET definitionally contain the same information as the LAI product, plus other variables that don’t respond to burn severity. If the GPP and ET product contain information that could respond to burn severity beyond LAI, then that needs to be explicitly stated.
Minor comments:
Methods: Grasses could grow back faster than the “resilience”/ “resistance” increment of 1 year. Is it appropriate to consider grasses’ “resistance” in the first year post fire?
Line 85- Citation how LAI related to prediction uncertainties of earth system models?
Line 160-164 – This seems to contradict methods described in lines 193-195
Line 186 - Difficulty understanding the sentence beginning with “Specifically…” What information is applied? Is this referring to MODIS pixels that span multiple fires, or each fire-MODIS pair?
Line 195 - Table 1 seems more appropriate to a supplement. Perhaps a column indicating the years used in this study?
Line 245 - 248: Citation/ discussion for why 100 trained models were sufficient.
Lines 250-258: This paragraph reads as more relevant to the methods section, and the writing shifts from present tense to past tense.
Figure 1 - Also seems more appropriate for a supplement. More discussion/ visual delineation of uncertainty would make this a more powerful figure. For example, why not include error bars for the years themselves? Finally, while the text claims climate variations in years (is that distinct from interanual variability?) are not confounding, 2019 seems to be different than other years compared to the interannual variability of 2001-2014 in precipitation. This is not discussed, and specifically how years are not significantly different is not defined.
Line 355 - Seems to contradict Figure 2 – 2019 is higher than 2020 in GPP. Also, if “Recover” is defined as reaching the same value within the variability of 2011-2014, then many metrics did not “recover” at all.
Line 321-322 Cite statement about overfitting.
443 - The connection between this research and afforestation and sustainable agriculture is unclear in this current writing.
Conclusions: Potential uses for this research for the calibration of ESM is discussed for the first time in the conclusions.
Technical Corrections:Line 209: - Additional period.
Line 278: “S” denoting severity and “S” denoting supplement both is confusing.
Line 316 important scores -> importance scores
364 - fully - full
395- interacting -> interaction
425 “findings that obtained with” -> findings from, “potential of” -> “potential for”
428 - unclear what is meant by “reasonable data quality controls”
Citation: https://doi.org/10.5194/egusphere-2024-2217-RC1 -
AC1: 'Reply on RC1', Mingjie Shi, 12 Nov 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-2217/egusphere-2024-2217-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Mingjie Shi, 12 Nov 2024
-
RC2: 'Comment on egusphere-2024-2217', Anonymous Referee #2, 19 Sep 2024
General comments:
In "Ecosystem leaf area, gross primary productions, and evapotranspiration responses to wildfire in the Columbia River Basin" the paper compares metrics of resistance and resilience of the MODIS-derived products mentioned in title across 3 primary vegetation types that underwent different burn severities from 138 fires in 2015. Resistance is calculated as the ratio of the 2016 annual maximum value for LAI, GPP, and ET to the average annual max value from the 4 years preceding the fire (2011-2014). Resilience is ratio of the average annual max values of LAI, ET, and GPP during 4 post-fire years (2017-2020) to the average annual max values during the 4 pre-fire years. A random forest approach is used to compare the influence of precipitation, vapor pressure deficit and burn severity on resilience of the three vegetation types (needleleaf evergreen forest, woody savanna, and grasslands). The paper concludes that 1. increased burn severity decreases resistance and resilience of LAI, GPP, and ET across all vegetation types, 2. LAI resistance and resilience is the most impacted by burn severity, 3. grassland LAI, GPP and ET are more resilient and resistant to wildfire than that of evergreen forests and savannas, and 4. burn severity is the primary driver of resilience in evergreen forests and savannas, while VPD and precipitation determine grassland resilience.
This paper’s methods and results are thought provoking and likely of interest to many. In addition to sharing concerns regarding the uncertainty issues addressed by the first reviewer, I have identified other limitations that should be addressed. Namely, the resistance and resilience metrics in this study are produced using different data than in the key cited paper (DeSoto et al., 2020) and thus there must be clear differences in the interpretation of the results.
The short post-fire study period undermines discussion related to resilience. This paper claims that low resilience values relate to low chances of full recovery (438-439). However, a 4-year post-fire study period is too short to assess the likelihood of recovery when forest recovery can take decades. If the authors decided to retain this resilience metric, the interpretation needs to appropriately acknowledge and match the limitations of the method.
Additionally, the authors do not use DeSoto’s ecosystem recovery metric. At the very least, there should be discussion of why that is not used since they otherwise rely on DeSoto’s framework. DeSoto et al. 2020 define recovery as the ratio between the post disturbance value and the value during the disturbance, and thus resilience (post disturbance/pre disturbance) is a product of both resistance (disturbance/pre disturbance) and recovery (post disturbance/ disturbance). Based on figure 2, recovery of LAI, GPP, and ET appear quite similar across the three vegetation types following severe burn. Thus, observed differences in resilience are driven primarily by differences in resistance as opposed to recovery. I believe that it is key that this paper properly discuss the nuanced differences in interpreting recovery (seemingly similar across these ecosystems) and resilience and resistance (seemingly contrasting).
Additional Major Comments:
Burn severities are presented as a key driving factor but I could not find information (in either the article or SI) about the distribution of burn severities across the different primary vegetation types. This would provide useful context for figures 3 and 5. Also, there is no discussion concerning the issues with burn severity classifications in grasslands compared to forests. Are there differences in classifications between vegetation types that readers should consider? Do MTBS algorithms and workflows perform similarly well in grasslands, savannas and forests?
Vegetation classifications are based on a 2015 MODIS dataset. For those areas that burned in 2015, do more recent land cover type datasets suggest a possible change in vegetation cover type? The authors do not discuss possible land cover conversion following different severity burns. Indeed, in figure 3 savannas seem to respond like grasslands at high severities and forests at low severities. While the post-fire study period is not long enough to expect full forest or savanna recovery from severe burn, more explicit discussion on conversion is warranted.
Technical corrections:
61-62: change to: “greater reductions in ET to precipitation ratios (ET/P)”
63: Is “respectively” needed?
63-66: Please improve the wording of this sentence.
70: Start sentence with “There”.
83-85: Rearrange this sentence, possibly split into two sentences.
93: Replace “, experienced” with “across”
95 -105: It would be easier to read if the hypotheses were simply listed. The discussion of concerning results from previous studies may be better earlier in the introduction or later in the discussion section.
108-113: The final part of the sentence about increasing wildfire severity in the CRB should be another sentence.
361: Including “(e.g., grasses)” is misleading. DeSoto et al. 2020 compares gymnosperm and angiosperm tree species, there is no discussion about grassland recovery.
116: Unnecessary first sentence.
130: include “annual 500 m” before “MODIS” and delete the second sentence.
132-135: Combine these two sentences.
153: Replace “burnt” with “burn”
153-156: Start a new sentence after “respectively”.
158: Delete “in 2015”.
168: Replace “water” with “vapor”.
171: Replace “data sets” with “datasets”.
192-194: Strange wording and see reviewer 1’s comment.
Table 1: Agree with reviewer 1: move to SI. Also it’s VPD, not WPD.
209: Delete the extra period.
230: Capitalize Python.
252-254: This should be in the methods section.
Figure 1: Agree with reviewer 1: better in the SI.
310-312: Strange wording.
316-317: First half of the sentence is poorly worded.
356: Replace “till 2020” with “within the post-fire study period”.
364: Delete the “y” on “fully”.
376: Delete “the result reveals”.
378: Replace “the findings” with “previous work showing”.
395: Delete this first sentence.
396: “Though, burn severity is less important to grassland resilience,”
399: Use “interaction”.
407: Delete “across”.
438-439: How does this study imply that these ecosystems may have extremely low chances of full recovery? 4 years is not a very long post-year period.
Citation: https://doi.org/10.5194/egusphere-2024-2217-RC2 -
AC2: 'Reply on RC2', Mingjie Shi, 12 Nov 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-2217/egusphere-2024-2217-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Mingjie Shi, 12 Nov 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 213 | 0 | 0 | 213 | 0 | 0 |
- HTML: 213
- PDF: 0
- XML: 0
- Total: 213
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Mingjie Shi
Nate McDowell
Huilin Huang
Faria Zahura
Lingcheng Li
Xingyuan Chen
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.