Brief communication: Soil moisture observations reconcile the discrepancy in detecting tornado early-stage track during the 24&ndash;25 March 2023 Mississippi outbreak

Wang, Jingyu; Wang, Xianfeng; Park, Edward; Lin, Yun

doi:https://doi.org/10.5194/nhess-2023-100

Preprints

https://doi.org/10.5194/nhess-2023-100

Preprints

31 Jul 2023

| 31 Jul 2023

Status: this discussion paper is a preprint. It has been under review for the journal Natural Hazards and Earth System Sciences (NHESS). The manuscript was not accepted for further review after discussion.

Brief communication: Soil moisture observations reconcile the discrepancy in detecting tornado early-stage track during the 24–25 March 2023 Mississippi outbreak

Jingyu Wang, Xianfeng Wang, Edward Park, and Yun Lin

Abstract. A series of tornadoes hit Western Mississippi on 24–25 March 2023, resulting in at least 26 deaths and massive destruction. The most devasting long-track EF4 tornado has been confirmed to touch down at southeast of Mayersville, Mississippi, however considerable inconsistency was found regarding its touchdown location and precise damage track between the onsite spotter reports and aftermath damage survey assessments. Such variation is a combined result of nocturnal occurrence (low nighttime visibility for spotter) and lack of assessment reference on non-vegetated farmland (high uncertainty of damage location for ground survey). This study suggests such discrepancy can be reconciled through the tornado damage scar captured by soil moisture observations.

Received: 20 Jun 2023 – Discussion started: 31 Jul 2023

Download & links

Jingyu Wang, Xianfeng Wang, Edward Park, and Yun Lin

Status: closed

RC1:
'Comment on nhess-2023-100', Anonymous Referee #1, 04 Aug 2023
This brief communication offers soil moisture analysis from remote sensing in an attempt to identify the starting point of the violent tornado that occurred in the spring of 2023. I applaud the authors for submitting a novel idea for peer review, but this article is not suitable for publication. This brief communication leaves open too many concerns about potential variables that impact the results. The methodology that was used for this brief communication has been only tested on one previous tornado, and this referee has serious concerns about the approach.
This brief communication is lacking a fundamental understanding of the tornado reporting and ground survey process that happens in the United States and that is conducted by the NOAA/National Weather Service.
Ln 27: I am not sure what is meant by “four types of tornado track detection methods”. There is really only one type of tornado track detection method and that is ground surveys conducted after suspected tornadoes. Meteorological data analysis, witness reports, and remote sensing data are fed into the process to help those that are conducting ground surveys. The statement that follows and is cited by Bluestein is also not completely true, at this time, tornadoes can only be confirmed by a ground survey.

Figure 2f and associated text and analysis are not correct. SPC storm reports should not have a linear track between them. Tornado reports from the SPC are simply reports and approximate locations. They will be used by the NWS and ground survey teams as a starting point. Actual ground survey information is available from the Damage Assessment Toolkit (DAT). The ground survey team will eventually then finalize the survey and submit it to StormData. SPC Storm reports and information from the DAT should not be used to show discrepancies, as it is expected that there will never be a 1:1 relationship between them.

Additionally ground surveys that are conducted on higher end tornadoes (>EF3) usually undergo multiple days of surveys, that include boots on the ground, remote sensing data from both aerial and space borne assets. It would be highly, highly unlikely that the ground survey teams did not have the beginning of the tornado correct. The authors offer nothing more than coarse and moderate resolution satellite imagery to suggest that their methodology is identifying a discrepancy at the beginning of the tornado track. I would encourage the authors to offer high resolution satellite imagery, photographs showing this discrepancy.

Ln 105: I am confused by this admission, are the authors admitting that their methodology from WANG23 is no longer usable in this case study or has serious flaws that were not previously addressed in WANG23? If this is the case, should additional case studies be conducted to validate the WANG23 methodology?
Ln 146: The statements introduced by the author on the potential issues that arise within 1 month between the Sentinel-2 imagery, suggest to this reviewer that there are several factors that would have prevented the use of imagery for this analysis. The authors even admit that the precip that fell aided the land surface in recovery, so how can the authors continue to use the imagery for their analysis? How does precip impact the soil moisture analysis and how can the authors determine a track is still there? Additionally, just because events weren’t recorded in the DAT, does that mean that the wind didn’t blow top soil around in that time frame or that the farmers were not in the fields preparing the land for planting? Visually figure 2a and 2b show significant changes to the bare fields that lead me to believe that it would be very difficult to identify any remaining tornado tracks.
Overall the spatial resolution of both MODIS and Sentinel-2 make it very difficult to see any changes or potential tornado paths or starting points in any of the figures.
Citation: https://doi.org/10.5194/nhess-2023-100-RC1
- AC3:
  'Reply on RC1', Jingyu Wang, 11 Sep 2023
  We thank the reviewer for recognizing our efforts and providing helpful suggestions and comments. Our point-by-point responses are provided as follows.
  Ln 27: I am not sure what is meant by “four types of tornado track detection methods”. There is really only one type of tornado track detection method and that is ground surveys conducted after suspected tornadoes. Meteorological data analysis, witness reports, and remote sensing data are fed into the process to help those that are conducting ground surveys. The statement that follows and is cited by Bluestein is also not completely true, at this time, tornadoes can only be confirmed by a ground survey.
  
  We appreciate your insights regarding the description of "four types of tornado track detection methods." Your clarification about the primary approach, which involves ground surveys conducted after suspected tornadoes, is indeed accurate. We apologize for any confusion caused by the wording. As suggested, we will make further clarification stating that meteorological data analysis, witness reports, and remote sensing data complement the ground survey process.
  Figure 2f and associated text and analysis are not correct. SPC storm reports should not have a linear track between them. Tornado reports from the SPC are simply reports and approximate locations. They will be used by the NWS and ground survey teams as a starting point. Actual ground survey information is available from the Damage Assessment Toolkit (DAT). The ground survey team will eventually then finalize the survey and submit it to StormData. SPC Storm reports and information from the DAT should not be used to show discrepancies, as it is expected that there will never be a 1:1 relationship between them.
  
  We agree with the reviewer that SPC reports only represent the approximate locations but do not include track. The purpose of linking adjacent SPC points is to demonstrate the possible tornado track if SPC reports were fully trusted. Therefore, we are illustrating the SPC linear track to showcase its deviation from the DAT track, in order to highlight the uncertainty associated with this method.
  Additionally ground surveys that are conducted on higher end tornadoes (>EF3) usually undergo multiple days of surveys, that include boots on the ground, remote sensing data from both aerial and space borne assets. It would be highly, highly unlikely that the ground survey teams did not have the beginning of the tornado correct. The authors offer nothing more than coarse and moderate resolution satellite imagery to suggest that their methodology is identifying a discrepancy at the beginning of the tornado track. I would encourage the authors to offer high resolution satellite imagery, photographs showing this discrepancy.
  
  We understand the accuracy of ground survey is very high. However, as mentioned in the manuscript, this method is still evidence-based, or reference-based. According to the DAT data, the two points that determine the DAT track in Figure 2f both correspond to the damage description of Trees: Hardwood (TH) Trunks snapped. The intention of this study is not to question the authority of the DAT survey. Instead, it proposes the inclusion of soil moisture as an additional piece of evidence for consideration. We apologize that the Sentinel-2 is the highest resolution image we can find. But further onsite surveys with photographs (both visible and infrared) is proposed for winter tornado study over this area.
  Ln 105: I am confused by this admission, are the authors admitting that their methodology from WANG23 is no longer usable in this case study or has serious flaws that were not previously addressed in WANG23? If this is the case, should additional case studies be conducted to validate the WANG23 methodology?
  
  We have commenced a partnership with CIWRO to investigate all similar incidents across the Mississippi Alluvial Plain. However, we hold the opinion that the 2023 MS tornado outbreak discussed in this paper serves as a well-timed confirmation of the WANG23 approach and warrants being published.
  Ln 146: The statements introduced by the author on the potential issues that arise within 1 month between the Sentinel-2 imagery, suggest to this reviewer that there are several factors that would have prevented the use of imagery for this analysis. The authors even admit that the precip that fell aided the land surface in recovery, so how can the authors continue to use the imagery for their analysis? How does precip impact the soil moisture analysis and how can the authors determine a track is still there? Additionally, just because events weren’t recorded in the DAT, does that mean that the wind didn’t blow top soil around in that time frame or that the farmers were not in the fields preparing the land for planting? Visually figure 2a and 2b show significant changes to the bare fields that lead me to believe that it would be very difficult to identify any remaining tornado tracks.
  
  First, as we mentioned in the manuscript, the 1-month recovery image on 30 April is not used for the retention of tornado track, but serves as the reference of undisturbed land surface. During the 1-month period, all the meteorological and non-meteorological factors can certainly wipe out the remaining tornado tracks as suggested by the reviewer. As a result, by contrasting the 1-day post-tornado image (26 March) with the 1-month recovery image, the tornado track can be unraveled on the post-tornado image, not the recovery image.
  Overall the spatial resolution of both MODIS and Sentinel-2 make it very difficult to see any changes or potential tornado paths or starting points in any of the figures.
  
  Once more, we express our gratitude to the reviewer for the valuable recommendations and regret any unsettling remarks. The primary objective of this research is to suggest the incorporation of soil moisture as an extra supporting factor for forthcoming DAT surveys. We will exercise caution in our choice of words as we undertake the revision process.
  
  Citation: https://doi.org/10.5194/nhess-2023-100-AC3
AC1:
'Comment on nhess-2023-100', Jingyu Wang, 12 Aug 2023
We thank the reviewer for recognizing our efforts and providing helpful suggestions and comments. Our point-by-point responses are provided as follows.
Ln 27: I am not sure what is meant by “four types of tornado track detection methods”. There is really only one type of tornado track detection method and that is ground surveys conducted after suspected tornadoes. Meteorological data analysis, witness reports, and remote sensing data are fed into the process to help those that are conducting ground surveys. The statement that follows and is cited by Bluestein is also not completely true, at this time, tornadoes can only be confirmed by a ground survey.

We appreciate your insights regarding the description of "four types of tornado track detection methods." Your clarification about the primary approach, which involves ground surveys conducted after suspected tornadoes, is indeed accurate. We apologize for any confusion caused by the wording. As suggested, we will make further clarification stating that meteorological data analysis, witness reports, and remote sensing data complement the ground survey process.
Figure 2f and associated text and analysis are not correct. SPC storm reports should not have a linear track between them. Tornado reports from the SPC are simply reports and approximate locations. They will be used by the NWS and ground survey teams as a starting point. Actual ground survey information is available from the Damage Assessment Toolkit (DAT). The ground survey team will eventually then finalize the survey and submit it to StormData. SPC Storm reports and information from the DAT should not be used to show discrepancies, as it is expected that there will never be a 1:1 relationship between them.

We agree with the reviewer that SPC reports only represent the approximate locations but do not include track. The purpose of linking adjacent SPC points is to demonstrate the possible tornado track if SPC reports were fully trusted. Therefore, we are illustrating the SPC linear track to showcase its deviation from the DAT track, in order to highlight the uncertainty associated with this method.
Additionally ground surveys that are conducted on higher end tornadoes (>EF3) usually undergo multiple days of surveys, that include boots on the ground, remote sensing data from both aerial and space borne assets. It would be highly, highly unlikely that the ground survey teams did not have the beginning of the tornado correct. The authors offer nothing more than coarse and moderate resolution satellite imagery to suggest that their methodology is identifying a discrepancy at the beginning of the tornado track. I would encourage the authors to offer high resolution satellite imagery, photographs showing this discrepancy.

We understand the accuracy of ground survey is very high. However, as mentioned in the manuscript, this method is still evidence-based, or reference-based. According to the DAT data, the two points that determine the DAT track in Figure 2f both correspond to the damage description of Trees: Hardwood (TH) Trunks snapped. The intention of this study is not to question the authority of the DAT survey. Instead, it proposes the inclusion of soil moisture as an additional piece of evidence for consideration. We apologize that the Sentinel-2 is the highest resolution image we can find. But further onsite surveys with photographs (both visible and infrared) is proposed for winter tornado study over this area.
Ln 105: I am confused by this admission, are the authors admitting that their methodology from WANG23 is no longer usable in this case study or has serious flaws that were not previously addressed in WANG23? If this is the case, should additional case studies be conducted to validate the WANG23 methodology?

We have commenced a partnership with CIWRO to investigate all similar incidents across the Mississippi Alluvial Plain. However, we hold the opinion that the 2023 MS tornado outbreak discussed in this paper serves as a well-timed confirmation of the WANG23 approach and warrants being published.
Ln 146: The statements introduced by the author on the potential issues that arise within 1 month between the Sentinel-2 imagery, suggest to this reviewer that there are several factors that would have prevented the use of imagery for this analysis. The authors even admit that the precip that fell aided the land surface in recovery, so how can the authors continue to use the imagery for their analysis? How does precip impact the soil moisture analysis and how can the authors determine a track is still there? Additionally, just because events weren’t recorded in the DAT, does that mean that the wind didn’t blow top soil around in that time frame or that the farmers were not in the fields preparing the land for planting? Visually figure 2a and 2b show significant changes to the bare fields that lead me to believe that it would be very difficult to identify any remaining tornado tracks.

First, as we mentioned in the manuscript, the 1-month recovery image on 30 April is not used for the retention of tornado track, but serves as the reference of undisturbed land surface. During the 1-month period, all the meteorological and non-meteorological factors can certainly wipe out the remaining tornado tracks as suggested by the reviewer. As a result, by contrasting the 1-day post-tornado image (26 March) with the 1-month recovery image, the tornado track can be unraveled on the post-tornado image, not the recovery image.
Overall the spatial resolution of both MODIS and Sentinel-2 make it very difficult to see any changes or potential tornado paths or starting points in any of the figures.

Once more, we express our gratitude to the reviewer for the valuable recommendations and regret any unsettling remarks. The primary objective of this research is to suggest the incorporation of soil moisture as an extra supporting factor for forthcoming DAT surveys. We will exercise caution in our choice of words as we undertake the revision process.
Citation: https://doi.org/10.5194/nhess-2023-100-AC1
RC2:
'Comment on nhess-2023-100', Anonymous Referee #2, 08 Sep 2023
The submitted manuscript describes an attempt to use remotely sensed soil moisture measurements to locate the initial portion of the track of a violent tornado in the southeastern United States in 2023.
Such a method for the detection of tornado occurrence in areas with little vegetation and no EF scale damage indicators would be a useful addition to the arsenal of techniques that are used to confirm and document tornadoes. The method is described in a publication that the lead author previously led (Wang et al., 2023) that used satellite-based measurements of soil moisture before and after a violent tornado outbreak in the United States to detect one location with elevated soil moisture, after the event, but for which no tornado track had been confirmed. There was one local storm report of a house destroyed in that area. Thus, Wang et al., 2023 believe that a new tornado had been detected using their method. However, in the areas where tornadoes had been documented, the technique produced little evidence of tornado occurrence – something that is not explained by the authors, aside from the fact that the technique is expected to work well only in soils with a clay-rich near-surface layer. The depth of soil removal required for this near-surface layer to be exposed is also not provided, but one assumes only a strong to violent tornado would be able to scour bare soil to that level. In any case, though peer-reviewed, I would conclude that the technique is not proven in any robust sense. And it’s use (in quite a different way) in this study renders the technique even more suspect.
Unfortunately, the present study is poorly conceived, the paper is not well written, and the conclusions are based on flawed analysis. My recommendation is rejection.
What follows are general comments on the fundamental flaws with the study, then some detailed comments about the manuscript.

GENERAL COMMENTS
The study has a number of fundamental problems, listed below:
The method outlined by Wang et al., 2023 could not be used because the satellite data available was compromised by severe rainfall in the area, presumably with the tornado outbreak. So the authors use satellite data from shortly after the event and then in the recovery period, well after the event. They also use a different type of satellite data, though it is arguably better. It’s not clear how appropriate the choice of the recovery-period data is because such a technique is not fully explained. How dry should the newly exposed clay soil appear a month after the event? Not discussed. Also, with the severe rainfall the authors describe, how do the authors know that any ground scouring was not due to intense runoff from this rainfall?

The focus of this article is on the start location of a documented tornado. Near the start of its lifetime, a tornado would typically be at a low intensity (e.g. EF0-level winds) – not at the intensity necessary for ground scouring. This puts in doubt that any enhanced soil moisture in the location of interest is actually due to a tornado.

The authors attempt to convince the reader that a few areas with the most significant change in soil moisture (Fig. 2f) are related to tornado ground scouring, even while in Fig. 2c they show ‘visual detection’ of a track in a somewhat different location and with a somewhat different orientation. They then attempt to discount the reliability of a damage observation made at ground level during an NWS survey since it does not fit well with their data. Again, does not inspire confidence in the analysis.

The authors try to convince the reader of the validity of this revised method using only one highly questionable case.

Overall, there is not enough evidence presented by the authors to support their conclusions.

DETAILED COMMENTS

For such a short article, there are many minor problems – due to inaccurate and/or unsupported statements, and spelling, grammar and formatting problems. I will only highlights some of the worst offenses below.
Line 32                  “interpretation of misaligned landscapes” – not at all sure what this means
Line 110                “The analysis primarily focused on the reported initial touchdown location, which is believed to strongly influence its progression to the EF4 level in later stage” – the authors appear to be saying here that the start location of the tornado influenced its intensity, which is I’m fairly certain not the case
Line 170                “The discrepancy in the DAT track may be attributed to the magnetic effect of the damage point” – I think the authors are trying to say that this one point was leading the NWS surveyors astray, but it is only incorrectly worded speculation
Line 174                “Snapped trunks can also be caused by derechos or gusty winds in close vicinity to the tornado” – it’s clear from this that the authors do not know what a derecho is (and they should)

REFERENCES
Wang, J., Lin, Y., McFarquhar, G. M., Park, E., Gu, Y., Su, Q., et al.: Soil moisture observations from shortwave infrared

channels reveal tornado tracks: A case in 10–11 December 2021 tornado outbreak. Geophysical Research Letters, 50,

e2023GL102984. https://doi.org/10.1029/2023GL102984, 2023.
Citation: https://doi.org/10.5194/nhess-2023-100-RC2
- AC2:
  'Reply on RC2', Jingyu Wang, 11 Sep 2023
  The submitted manuscript describes an attempt to use remotely sensed soil moisture measurements to locate the initial portion of the track of a violent tornado in the southeastern United States in 2023. Such a method for the detection of tornado occurrence in areas with little vegetation and no EF scale damage indicators would be a useful addition to the arsenal of techniques that are used to confirm and document tornadoes. The method is described in a publication that the lead author previously led (Wang et al., 2023) that used satellite-based measurements of soil moisture before and after a violent tornado outbreak in the United States to detect one location with elevated soil moisture, after the event, but for which no tornado track had been confirmed. There was one local storm report of a house destroyed in that area. Thus, Wang et al., 2023 believe that a new tornado had been detected using their method. However, in the areas where tornadoes had been documented, the technique produced little evidence of tornado occurrence – something that is not explained by the authors, aside from the fact that the technique is expected to work well only in soils with a clay-rich near-surface layer. The depth of soil removal required for this near-surface layer to be exposed is also not provided, but one assumes only a strong to violent tornado would be able to scour bare soil to that level. In any case, though peer-reviewed, I would conclude that the technique is not proven in any robust sense. And it’s use (in quite a different way) in this study renders the technique even more suspect.
  
  We thank the reviewer for the summary and the questions. Our point-by-point responses are provided as follows.
  General Comments
  The method outlined by Wang et al., 2023 could not be used because the satellite data available was compromised by severe rainfall in the area, presumably with the tornado outbreak. So the authors use satellite data from shortly after the event and then in the recovery period, well after the event. They also use a different type of satellite data, though it is arguably better. It’s not clear how appropriate the choice of the recovery-period data is because such a technique is not fully explained. How dry should the newly exposed clay soil appear a month after the event? Not discussed. Also, with the severe rainfall the authors describe, how do the authors know that any ground scouring was not due to intense runoff from this rainfall?
  
  We do not have access to the in-situ measurements of soil moisture data, therefore we cannot answer how dry the newly exposed clay soil were after one month recovery. It is possible the soil moisture scars were associated with the surface runoff gutters caused by severe precipitation.
  The focus of this article is on the start location of a documented tornado. Near the start of its lifetime, a tornado would typically be at a low intensity (e.g. EF0-level winds) – not at the intensity necessary for ground scouring. This puts in doubt that any enhanced soil moisture in the location of interest is actually due to a tornado.
  
  According to DAT data, the first touchdown tornado was EF1, which sufficed to cause ground scouring.
  The authors attempt to convince the reader that a few areas with the most significant change in soil moisture (Fig. 2f) are related to tornado ground scouring, even while in Fig. 2c they show ‘visual detection’ of a track in a somewhat different location and with a somewhat different orientation. They then attempt to discount the reliability of a damage observation made at ground level during an NWS survey since it does not fit well with their data. Again, does not inspire confidence in the analysis.
  
  The authors did not 'discount' the damage observation during the ground survey, instead, they wanted to highlight the limitation of the 'damage-based' assessment method. According to the DAT data, the two points that determine the DAT track in Figure 2f both correspond to the damage description of Trees: Hardwood (TH) Trunks snapped. The intention of this study is not to question the authority of the DAT survey. Rather, it proposes the inclusion of soil moisture as an additional piece of evidence for consideration.
  The authors try to convince the reader of the validity of this revised method using only one highly questionable case.
  
  We agree with the reviewer that one or two cases are definitely insufficient to validate the method. However, the winter tornadoes over Mississippi Alluvial Plain are extremely rare. We will try to conduct further onsite surveys with photographs (both visible and infrared) for the future events.
  Line 32                  “interpretation of misaligned landscapes” – not at all sure what this means
  
  Changed to 'interpretation of disturbed landscapes'.
  Line 110                “The analysis primarily focused on the reported initial touchdown location, which is believed to strongly influence its progression to the EF4 level in later stage” – the authors appear to be saying here that the start location of the tornado influenced its intensity, which is I’m fairly certain not the case
  
  This sentence has been removed.
  Line 170                “The discrepancy in the DAT track may be attributed to the magnetic effect of the damage point” – I think the authors are trying to say that this one point was leading the NWS surveyors astray, but it is only incorrectly worded speculation
  
  Changed to 'The discrepancy in the DAT track may be attributed to the fact that one point was leading the NWS surveyors astray'.
  Line 174                “Snapped trunks can also be caused by derechos or gusty winds in close vicinity to the tornado” – it’s clear from this that the authors do not know what a derecho is (and they should)
  
  We thank the reviewer for catching the error. The suspicion of derechos has been removed.
  
  Citation: https://doi.org/10.5194/nhess-2023-100-AC2

Status: closed

RC1:
'Comment on nhess-2023-100', Anonymous Referee #1, 04 Aug 2023
This brief communication offers soil moisture analysis from remote sensing in an attempt to identify the starting point of the violent tornado that occurred in the spring of 2023. I applaud the authors for submitting a novel idea for peer review, but this article is not suitable for publication. This brief communication leaves open too many concerns about potential variables that impact the results. The methodology that was used for this brief communication has been only tested on one previous tornado, and this referee has serious concerns about the approach.
This brief communication is lacking a fundamental understanding of the tornado reporting and ground survey process that happens in the United States and that is conducted by the NOAA/National Weather Service.
Ln 27: I am not sure what is meant by “four types of tornado track detection methods”. There is really only one type of tornado track detection method and that is ground surveys conducted after suspected tornadoes. Meteorological data analysis, witness reports, and remote sensing data are fed into the process to help those that are conducting ground surveys. The statement that follows and is cited by Bluestein is also not completely true, at this time, tornadoes can only be confirmed by a ground survey.

Figure 2f and associated text and analysis are not correct. SPC storm reports should not have a linear track between them. Tornado reports from the SPC are simply reports and approximate locations. They will be used by the NWS and ground survey teams as a starting point. Actual ground survey information is available from the Damage Assessment Toolkit (DAT). The ground survey team will eventually then finalize the survey and submit it to StormData. SPC Storm reports and information from the DAT should not be used to show discrepancies, as it is expected that there will never be a 1:1 relationship between them.

Additionally ground surveys that are conducted on higher end tornadoes (>EF3) usually undergo multiple days of surveys, that include boots on the ground, remote sensing data from both aerial and space borne assets. It would be highly, highly unlikely that the ground survey teams did not have the beginning of the tornado correct. The authors offer nothing more than coarse and moderate resolution satellite imagery to suggest that their methodology is identifying a discrepancy at the beginning of the tornado track. I would encourage the authors to offer high resolution satellite imagery, photographs showing this discrepancy.

Ln 105: I am confused by this admission, are the authors admitting that their methodology from WANG23 is no longer usable in this case study or has serious flaws that were not previously addressed in WANG23? If this is the case, should additional case studies be conducted to validate the WANG23 methodology?
Ln 146: The statements introduced by the author on the potential issues that arise within 1 month between the Sentinel-2 imagery, suggest to this reviewer that there are several factors that would have prevented the use of imagery for this analysis. The authors even admit that the precip that fell aided the land surface in recovery, so how can the authors continue to use the imagery for their analysis? How does precip impact the soil moisture analysis and how can the authors determine a track is still there? Additionally, just because events weren’t recorded in the DAT, does that mean that the wind didn’t blow top soil around in that time frame or that the farmers were not in the fields preparing the land for planting? Visually figure 2a and 2b show significant changes to the bare fields that lead me to believe that it would be very difficult to identify any remaining tornado tracks.
Overall the spatial resolution of both MODIS and Sentinel-2 make it very difficult to see any changes or potential tornado paths or starting points in any of the figures.
Citation: https://doi.org/10.5194/nhess-2023-100-RC1
- AC3:
  'Reply on RC1', Jingyu Wang, 11 Sep 2023
  We thank the reviewer for recognizing our efforts and providing helpful suggestions and comments. Our point-by-point responses are provided as follows.
  Ln 27: I am not sure what is meant by “four types of tornado track detection methods”. There is really only one type of tornado track detection method and that is ground surveys conducted after suspected tornadoes. Meteorological data analysis, witness reports, and remote sensing data are fed into the process to help those that are conducting ground surveys. The statement that follows and is cited by Bluestein is also not completely true, at this time, tornadoes can only be confirmed by a ground survey.
  
  We appreciate your insights regarding the description of "four types of tornado track detection methods." Your clarification about the primary approach, which involves ground surveys conducted after suspected tornadoes, is indeed accurate. We apologize for any confusion caused by the wording. As suggested, we will make further clarification stating that meteorological data analysis, witness reports, and remote sensing data complement the ground survey process.
  Figure 2f and associated text and analysis are not correct. SPC storm reports should not have a linear track between them. Tornado reports from the SPC are simply reports and approximate locations. They will be used by the NWS and ground survey teams as a starting point. Actual ground survey information is available from the Damage Assessment Toolkit (DAT). The ground survey team will eventually then finalize the survey and submit it to StormData. SPC Storm reports and information from the DAT should not be used to show discrepancies, as it is expected that there will never be a 1:1 relationship between them.
  
  We agree with the reviewer that SPC reports only represent the approximate locations but do not include track. The purpose of linking adjacent SPC points is to demonstrate the possible tornado track if SPC reports were fully trusted. Therefore, we are illustrating the SPC linear track to showcase its deviation from the DAT track, in order to highlight the uncertainty associated with this method.
  Additionally ground surveys that are conducted on higher end tornadoes (>EF3) usually undergo multiple days of surveys, that include boots on the ground, remote sensing data from both aerial and space borne assets. It would be highly, highly unlikely that the ground survey teams did not have the beginning of the tornado correct. The authors offer nothing more than coarse and moderate resolution satellite imagery to suggest that their methodology is identifying a discrepancy at the beginning of the tornado track. I would encourage the authors to offer high resolution satellite imagery, photographs showing this discrepancy.
  
  We understand the accuracy of ground survey is very high. However, as mentioned in the manuscript, this method is still evidence-based, or reference-based. According to the DAT data, the two points that determine the DAT track in Figure 2f both correspond to the damage description of Trees: Hardwood (TH) Trunks snapped. The intention of this study is not to question the authority of the DAT survey. Instead, it proposes the inclusion of soil moisture as an additional piece of evidence for consideration. We apologize that the Sentinel-2 is the highest resolution image we can find. But further onsite surveys with photographs (both visible and infrared) is proposed for winter tornado study over this area.
  Ln 105: I am confused by this admission, are the authors admitting that their methodology from WANG23 is no longer usable in this case study or has serious flaws that were not previously addressed in WANG23? If this is the case, should additional case studies be conducted to validate the WANG23 methodology?
  
  We have commenced a partnership with CIWRO to investigate all similar incidents across the Mississippi Alluvial Plain. However, we hold the opinion that the 2023 MS tornado outbreak discussed in this paper serves as a well-timed confirmation of the WANG23 approach and warrants being published.
  Ln 146: The statements introduced by the author on the potential issues that arise within 1 month between the Sentinel-2 imagery, suggest to this reviewer that there are several factors that would have prevented the use of imagery for this analysis. The authors even admit that the precip that fell aided the land surface in recovery, so how can the authors continue to use the imagery for their analysis? How does precip impact the soil moisture analysis and how can the authors determine a track is still there? Additionally, just because events weren’t recorded in the DAT, does that mean that the wind didn’t blow top soil around in that time frame or that the farmers were not in the fields preparing the land for planting? Visually figure 2a and 2b show significant changes to the bare fields that lead me to believe that it would be very difficult to identify any remaining tornado tracks.
  
  First, as we mentioned in the manuscript, the 1-month recovery image on 30 April is not used for the retention of tornado track, but serves as the reference of undisturbed land surface. During the 1-month period, all the meteorological and non-meteorological factors can certainly wipe out the remaining tornado tracks as suggested by the reviewer. As a result, by contrasting the 1-day post-tornado image (26 March) with the 1-month recovery image, the tornado track can be unraveled on the post-tornado image, not the recovery image.
  Overall the spatial resolution of both MODIS and Sentinel-2 make it very difficult to see any changes or potential tornado paths or starting points in any of the figures.
  
  Once more, we express our gratitude to the reviewer for the valuable recommendations and regret any unsettling remarks. The primary objective of this research is to suggest the incorporation of soil moisture as an extra supporting factor for forthcoming DAT surveys. We will exercise caution in our choice of words as we undertake the revision process.
  
  Citation: https://doi.org/10.5194/nhess-2023-100-AC3
AC1:
'Comment on nhess-2023-100', Jingyu Wang, 12 Aug 2023
We thank the reviewer for recognizing our efforts and providing helpful suggestions and comments. Our point-by-point responses are provided as follows.
Ln 27: I am not sure what is meant by “four types of tornado track detection methods”. There is really only one type of tornado track detection method and that is ground surveys conducted after suspected tornadoes. Meteorological data analysis, witness reports, and remote sensing data are fed into the process to help those that are conducting ground surveys. The statement that follows and is cited by Bluestein is also not completely true, at this time, tornadoes can only be confirmed by a ground survey.

We appreciate your insights regarding the description of "four types of tornado track detection methods." Your clarification about the primary approach, which involves ground surveys conducted after suspected tornadoes, is indeed accurate. We apologize for any confusion caused by the wording. As suggested, we will make further clarification stating that meteorological data analysis, witness reports, and remote sensing data complement the ground survey process.
Figure 2f and associated text and analysis are not correct. SPC storm reports should not have a linear track between them. Tornado reports from the SPC are simply reports and approximate locations. They will be used by the NWS and ground survey teams as a starting point. Actual ground survey information is available from the Damage Assessment Toolkit (DAT). The ground survey team will eventually then finalize the survey and submit it to StormData. SPC Storm reports and information from the DAT should not be used to show discrepancies, as it is expected that there will never be a 1:1 relationship between them.

We agree with the reviewer that SPC reports only represent the approximate locations but do not include track. The purpose of linking adjacent SPC points is to demonstrate the possible tornado track if SPC reports were fully trusted. Therefore, we are illustrating the SPC linear track to showcase its deviation from the DAT track, in order to highlight the uncertainty associated with this method.
Additionally ground surveys that are conducted on higher end tornadoes (>EF3) usually undergo multiple days of surveys, that include boots on the ground, remote sensing data from both aerial and space borne assets. It would be highly, highly unlikely that the ground survey teams did not have the beginning of the tornado correct. The authors offer nothing more than coarse and moderate resolution satellite imagery to suggest that their methodology is identifying a discrepancy at the beginning of the tornado track. I would encourage the authors to offer high resolution satellite imagery, photographs showing this discrepancy.

We understand the accuracy of ground survey is very high. However, as mentioned in the manuscript, this method is still evidence-based, or reference-based. According to the DAT data, the two points that determine the DAT track in Figure 2f both correspond to the damage description of Trees: Hardwood (TH) Trunks snapped. The intention of this study is not to question the authority of the DAT survey. Instead, it proposes the inclusion of soil moisture as an additional piece of evidence for consideration. We apologize that the Sentinel-2 is the highest resolution image we can find. But further onsite surveys with photographs (both visible and infrared) is proposed for winter tornado study over this area.
Ln 105: I am confused by this admission, are the authors admitting that their methodology from WANG23 is no longer usable in this case study or has serious flaws that were not previously addressed in WANG23? If this is the case, should additional case studies be conducted to validate the WANG23 methodology?

We have commenced a partnership with CIWRO to investigate all similar incidents across the Mississippi Alluvial Plain. However, we hold the opinion that the 2023 MS tornado outbreak discussed in this paper serves as a well-timed confirmation of the WANG23 approach and warrants being published.
Ln 146: The statements introduced by the author on the potential issues that arise within 1 month between the Sentinel-2 imagery, suggest to this reviewer that there are several factors that would have prevented the use of imagery for this analysis. The authors even admit that the precip that fell aided the land surface in recovery, so how can the authors continue to use the imagery for their analysis? How does precip impact the soil moisture analysis and how can the authors determine a track is still there? Additionally, just because events weren’t recorded in the DAT, does that mean that the wind didn’t blow top soil around in that time frame or that the farmers were not in the fields preparing the land for planting? Visually figure 2a and 2b show significant changes to the bare fields that lead me to believe that it would be very difficult to identify any remaining tornado tracks.

First, as we mentioned in the manuscript, the 1-month recovery image on 30 April is not used for the retention of tornado track, but serves as the reference of undisturbed land surface. During the 1-month period, all the meteorological and non-meteorological factors can certainly wipe out the remaining tornado tracks as suggested by the reviewer. As a result, by contrasting the 1-day post-tornado image (26 March) with the 1-month recovery image, the tornado track can be unraveled on the post-tornado image, not the recovery image.
Overall the spatial resolution of both MODIS and Sentinel-2 make it very difficult to see any changes or potential tornado paths or starting points in any of the figures.

Once more, we express our gratitude to the reviewer for the valuable recommendations and regret any unsettling remarks. The primary objective of this research is to suggest the incorporation of soil moisture as an extra supporting factor for forthcoming DAT surveys. We will exercise caution in our choice of words as we undertake the revision process.
Citation: https://doi.org/10.5194/nhess-2023-100-AC1
RC2:
'Comment on nhess-2023-100', Anonymous Referee #2, 08 Sep 2023
The submitted manuscript describes an attempt to use remotely sensed soil moisture measurements to locate the initial portion of the track of a violent tornado in the southeastern United States in 2023.
Such a method for the detection of tornado occurrence in areas with little vegetation and no EF scale damage indicators would be a useful addition to the arsenal of techniques that are used to confirm and document tornadoes. The method is described in a publication that the lead author previously led (Wang et al., 2023) that used satellite-based measurements of soil moisture before and after a violent tornado outbreak in the United States to detect one location with elevated soil moisture, after the event, but for which no tornado track had been confirmed. There was one local storm report of a house destroyed in that area. Thus, Wang et al., 2023 believe that a new tornado had been detected using their method. However, in the areas where tornadoes had been documented, the technique produced little evidence of tornado occurrence – something that is not explained by the authors, aside from the fact that the technique is expected to work well only in soils with a clay-rich near-surface layer. The depth of soil removal required for this near-surface layer to be exposed is also not provided, but one assumes only a strong to violent tornado would be able to scour bare soil to that level. In any case, though peer-reviewed, I would conclude that the technique is not proven in any robust sense. And it’s use (in quite a different way) in this study renders the technique even more suspect.
Unfortunately, the present study is poorly conceived, the paper is not well written, and the conclusions are based on flawed analysis. My recommendation is rejection.
What follows are general comments on the fundamental flaws with the study, then some detailed comments about the manuscript.

GENERAL COMMENTS
The study has a number of fundamental problems, listed below:
The method outlined by Wang et al., 2023 could not be used because the satellite data available was compromised by severe rainfall in the area, presumably with the tornado outbreak. So the authors use satellite data from shortly after the event and then in the recovery period, well after the event. They also use a different type of satellite data, though it is arguably better. It’s not clear how appropriate the choice of the recovery-period data is because such a technique is not fully explained. How dry should the newly exposed clay soil appear a month after the event? Not discussed. Also, with the severe rainfall the authors describe, how do the authors know that any ground scouring was not due to intense runoff from this rainfall?

The focus of this article is on the start location of a documented tornado. Near the start of its lifetime, a tornado would typically be at a low intensity (e.g. EF0-level winds) – not at the intensity necessary for ground scouring. This puts in doubt that any enhanced soil moisture in the location of interest is actually due to a tornado.

The authors attempt to convince the reader that a few areas with the most significant change in soil moisture (Fig. 2f) are related to tornado ground scouring, even while in Fig. 2c they show ‘visual detection’ of a track in a somewhat different location and with a somewhat different orientation. They then attempt to discount the reliability of a damage observation made at ground level during an NWS survey since it does not fit well with their data. Again, does not inspire confidence in the analysis.

The authors try to convince the reader of the validity of this revised method using only one highly questionable case.

Overall, there is not enough evidence presented by the authors to support their conclusions.

DETAILED COMMENTS

For such a short article, there are many minor problems – due to inaccurate and/or unsupported statements, and spelling, grammar and formatting problems. I will only highlights some of the worst offenses below.
Line 32                  “interpretation of misaligned landscapes” – not at all sure what this means
Line 110                “The analysis primarily focused on the reported initial touchdown location, which is believed to strongly influence its progression to the EF4 level in later stage” – the authors appear to be saying here that the start location of the tornado influenced its intensity, which is I’m fairly certain not the case
Line 170                “The discrepancy in the DAT track may be attributed to the magnetic effect of the damage point” – I think the authors are trying to say that this one point was leading the NWS surveyors astray, but it is only incorrectly worded speculation
Line 174                “Snapped trunks can also be caused by derechos or gusty winds in close vicinity to the tornado” – it’s clear from this that the authors do not know what a derecho is (and they should)

REFERENCES
Wang, J., Lin, Y., McFarquhar, G. M., Park, E., Gu, Y., Su, Q., et al.: Soil moisture observations from shortwave infrared

channels reveal tornado tracks: A case in 10–11 December 2021 tornado outbreak. Geophysical Research Letters, 50,

e2023GL102984. https://doi.org/10.1029/2023GL102984, 2023.
Citation: https://doi.org/10.5194/nhess-2023-100-RC2
- AC2:
  'Reply on RC2', Jingyu Wang, 11 Sep 2023
  The submitted manuscript describes an attempt to use remotely sensed soil moisture measurements to locate the initial portion of the track of a violent tornado in the southeastern United States in 2023. Such a method for the detection of tornado occurrence in areas with little vegetation and no EF scale damage indicators would be a useful addition to the arsenal of techniques that are used to confirm and document tornadoes. The method is described in a publication that the lead author previously led (Wang et al., 2023) that used satellite-based measurements of soil moisture before and after a violent tornado outbreak in the United States to detect one location with elevated soil moisture, after the event, but for which no tornado track had been confirmed. There was one local storm report of a house destroyed in that area. Thus, Wang et al., 2023 believe that a new tornado had been detected using their method. However, in the areas where tornadoes had been documented, the technique produced little evidence of tornado occurrence – something that is not explained by the authors, aside from the fact that the technique is expected to work well only in soils with a clay-rich near-surface layer. The depth of soil removal required for this near-surface layer to be exposed is also not provided, but one assumes only a strong to violent tornado would be able to scour bare soil to that level. In any case, though peer-reviewed, I would conclude that the technique is not proven in any robust sense. And it’s use (in quite a different way) in this study renders the technique even more suspect.
  
  We thank the reviewer for the summary and the questions. Our point-by-point responses are provided as follows.
  General Comments
  The method outlined by Wang et al., 2023 could not be used because the satellite data available was compromised by severe rainfall in the area, presumably with the tornado outbreak. So the authors use satellite data from shortly after the event and then in the recovery period, well after the event. They also use a different type of satellite data, though it is arguably better. It’s not clear how appropriate the choice of the recovery-period data is because such a technique is not fully explained. How dry should the newly exposed clay soil appear a month after the event? Not discussed. Also, with the severe rainfall the authors describe, how do the authors know that any ground scouring was not due to intense runoff from this rainfall?
  
  We do not have access to the in-situ measurements of soil moisture data, therefore we cannot answer how dry the newly exposed clay soil were after one month recovery. It is possible the soil moisture scars were associated with the surface runoff gutters caused by severe precipitation.
  The focus of this article is on the start location of a documented tornado. Near the start of its lifetime, a tornado would typically be at a low intensity (e.g. EF0-level winds) – not at the intensity necessary for ground scouring. This puts in doubt that any enhanced soil moisture in the location of interest is actually due to a tornado.
  
  According to DAT data, the first touchdown tornado was EF1, which sufficed to cause ground scouring.
  The authors attempt to convince the reader that a few areas with the most significant change in soil moisture (Fig. 2f) are related to tornado ground scouring, even while in Fig. 2c they show ‘visual detection’ of a track in a somewhat different location and with a somewhat different orientation. They then attempt to discount the reliability of a damage observation made at ground level during an NWS survey since it does not fit well with their data. Again, does not inspire confidence in the analysis.
  
  The authors did not 'discount' the damage observation during the ground survey, instead, they wanted to highlight the limitation of the 'damage-based' assessment method. According to the DAT data, the two points that determine the DAT track in Figure 2f both correspond to the damage description of Trees: Hardwood (TH) Trunks snapped. The intention of this study is not to question the authority of the DAT survey. Rather, it proposes the inclusion of soil moisture as an additional piece of evidence for consideration.
  The authors try to convince the reader of the validity of this revised method using only one highly questionable case.
  
  We agree with the reviewer that one or two cases are definitely insufficient to validate the method. However, the winter tornadoes over Mississippi Alluvial Plain are extremely rare. We will try to conduct further onsite surveys with photographs (both visible and infrared) for the future events.
  Line 32                  “interpretation of misaligned landscapes” – not at all sure what this means
  
  Changed to 'interpretation of disturbed landscapes'.
  Line 110                “The analysis primarily focused on the reported initial touchdown location, which is believed to strongly influence its progression to the EF4 level in later stage” – the authors appear to be saying here that the start location of the tornado influenced its intensity, which is I’m fairly certain not the case
  
  This sentence has been removed.
  Line 170                “The discrepancy in the DAT track may be attributed to the magnetic effect of the damage point” – I think the authors are trying to say that this one point was leading the NWS surveyors astray, but it is only incorrectly worded speculation
  
  Changed to 'The discrepancy in the DAT track may be attributed to the fact that one point was leading the NWS surveyors astray'.
  Line 174                “Snapped trunks can also be caused by derechos or gusty winds in close vicinity to the tornado” – it’s clear from this that the authors do not know what a derecho is (and they should)
  
  We thank the reviewer for catching the error. The suspicion of derechos has been removed.
  
  Citation: https://doi.org/10.5194/nhess-2023-100-AC2

Jingyu Wang, Xianfeng Wang, Edward Park, and Yun Lin

Viewed

Total article views: 467 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
331	108	28	467	15	14

HTML: 331
PDF: 108
XML: 28
Total: 467
BibTeX: 15
EndNote: 14

Views and downloads (calculated since 31 Jul 2023)

Month	HTML	PDF	XML	Total
Jul 2023	29	5	1	35
Aug 2023	137	26	11	174
Sep 2023	65	24	7	96
Oct 2023	25	12	2	39
Nov 2023	10	4	0	14
Dec 2023	14	9	1	24
Jan 2024	28	3	2	33
Feb 2024	22	23	4	49
Mar 2024	1	2	0	3

Cumulative views and downloads (calculated since 31 Jul 2023)

Month	HTML	PDF	XML	Total
Jul 2023	29	5	1	35
Aug 2023	137	26	11	174
Sep 2023	65	24	7	96
Oct 2023	25	12	2	39
Nov 2023	10	4	0	14
Dec 2023	14	9	1	24
Jan 2024	28	3	2	33
Feb 2024	22	23	4	49
Mar 2024	1	2	0	3

Viewed (geographical distribution)

Total article views: 439 (including HTML, PDF, and XML) Thereof 439 with geography defined and 0 with unknown origin.

Country	#	Views	%

Discussed

Latest update: 05 Mar 2024

Short summary

Building upon the findings in a preceding study by the authors (Wang et al., 2023), this brief communication successfully applied the soil moisture-based tornado damage track detection method to the 24–25 March 2023 Mississippi outbreak. This study also found that the notable discrepancies between spotter reports and ground survey assessments at the tornado early stage can be reconciled using the new method.


Total:	0
HTML:	0
PDF:	0
XML:	0