the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Sep 2022
Towards long-term records of rain-on-snow events across the Arctic from satellite data
Helena Bergstedt
Georg Pointner
Xaver Muri
Kimmo Rautiainen
Leena Leppänen
Kyle Joly
Aleksandr Sokolov
Pavel Orekhov
Dorothee Ehrich
Eeva Mariatta Soininen
Abstract. Rain-on-Snow (ROS) events occur across many regions of the terrestrial Arctic in mid-winter. Snow pack properties are changing and in extreme cases ice layers form which affect wildlife, vegetation and soils beyond the duration of the event. Specifically, satellite microwave observations have been shown to provide insight into known events. Only Ku-band radar (scatterometer) has been applied so far across the entire Arctic. Data availability at this frequency is limited, however. The utility of other frequencies from passive and active systems need to be explored to develop a concept for long-term monitoring. Active (radar) records have been shown to capture the associated snow structure change based on time series analyses. This approach is also applicable when data gaps exist and bears capabilities to evaluate the impact severity of events. Active as well as passive microwave sensors can also detect wet snow at the timing of a ROS, if an acquisition is available. Wet snow retrieval methodology is, however, rather mature, compared to identification of snow structure change which needs consideration of ambiguous scattering behaviour. C-band radar is of special interest due to good data availability including a range of nominal spatial resolution (10 m–12.5 km) A combined approach is therefore considered and tested for C-band (active, snow structure change) and L-band (passive, wet snow). Results were compared to in situ observations (snow pit records, caribou migration data) and Ku-band products. Ice crusts were found in the snow pack after detected events. The more crusts (events) the higher the winter season backscatter increase at C-band. ROS events captured on the Yamal and Seward peninsulas have had severe impacts on reindeer and caribou, respectively, due to crust formation. Temperature dependence of C-band backscatter observable down to -40 °C is identified as a major issue for ROS retrieval, but can be addressed by combination with passive microwave wet snow indicators (demonstrated for Metop ASCAT and SMOS). Synthetic Aperture Radar (SAR) from specifically Sentinel-1 (C-band) is promising regarding ice layer identification at better spatial details for all available polarizations. The fusion of multiple types of microwave satellite observations is suggested for the creation of a climate data record, but the consideration of performance differences due to spatial and temporal cover as well as microwave frequency is crucial. Retrieval is most robust north of 65° N, in the tundra biome, where records can be used to identify extremes and to apply the results for impact studies at regional scale.
-
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
(15160 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(15160 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Annett Bartsch et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-899', Anonymous Referee #1, 18 Oct 2022
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-899/egusphere-2022-899-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Annett Bartsch, 06 Nov 2022
Many thanks for your valuable comments!
Please find responses to your questions below:
RC1: Area of study: it is apparent from the figures that the authors have applied the approach to not only the entire Arctic but also to land areas extending much further south (eg., Figure 5). However, it is stated in the conclusion that the approach is only recommended for regions north of 66 deg due to coverage issues with SMOS. Perhaps the authors should revise the boundaries of the areas for which ROS detections are presented, or comment on how representative the ROS data are for the lower latitude areas shown in Fig.5?
Reply: The analyses extent was chosen based on the boundaries of the existing database from QuikScat. We describe the differences between sensors in forested/southern regions on lines 377ff and discuss it on lines 452-469. In the latter case, a wrong figure is referred to in the submitted manuscript (#2 instead of #5). However, we agree, an extended discussion on how representative the results are south of the treeline/in the SMOS gap zone would be beneficial.
RC1: The authors have described and presented a wide range of different types of observations, but I think that some of the datasets used do not really add much to the overall goals and conclusions of the study. While I see the need for observations to validate/support the remote sensing data, I think the use of too many different observations, each with their own considerations for ROS detection, makes it at times difficult to follow the main objectives of the study. I would for example recommend reconsidering whether the use of the caribou data are really necessary.
Reply: Our reasons for including the Caribou/Serward peinsula study were to (1) demonstrate that there is stronger variability from year to year regionally than what can be observed for the entire Arctic (Figure 5) and (2) to point to the potential use of the backscatter change magnitude in addition to just event detection. The Caribou herds used some areas where an event was detected, but not areas which exceeded a certain back scatter change value (figure 12c). This could be moved to the discussion part of the paper.
RC1: Line 175: why were different terms/hardness scales used at Yamal compared to the Scandinavian sites? Why not just use a standard scale for all sites?
Reply: : The surveys come from different (partially long-term) monitoring programs, carried out by different institutions which follow different schemes. It would be indeed very beneficial if future surveys would follow the same scheme.
RC1: Line 296: “location specific threshold” - does this mean that a threshold is determined for individual pixels, or for regions?
Reply: Yes, the threshold is defined individually for each grid point.
RC1: Table 2: Events represent November 2021 to February 2022; why are values from only 1 year/winter of observations used?
Reply: Thanks for spotting! It should read November – February, years 2011-2022 (note that figure 1 is also positioned in the main text, similar to table 1)
RC1: Figure 4: Could the authors comment on the event confirmed by SMOS occurring in the start of December 2016? Here the AWS data show very low temperature (approx. -15 deg.C), no precipitation and increasing snow depth in the following days. What could be the reason for detection of wet snow?
Reply: The SMOS detection actually refers to a smaller event (LRI < 1mm) two days before the ASCAT detection. The ASCAT detection does however represent a period of temperature drop (following the rain event). The impact of liquid precipitation several days before is not captured. This case shows the disadvantage of using a 3 day window for the SMOS masking, but it is necessary to account for data gaps and nature of the radar retrieval scheme, as discussed on lines 304 in the methods part. We agree that the discussion on the choice of the detection window could be more extensive and included in the discussion.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC1 -
AC3: 'Reply on RC1 - correction (first response was for RC2)', Annett Bartsch, 06 Nov 2022
Please apologize the mixing up with comments to RC2.
Also thanks to you for your valuable comments!
Please find our response to your main comments below:
RC1: ... this work lacks a quantitative validation of the ROS retrieval from ice layer I snowpits, maybe % of commission/omission from the algorithm could be calculated.
Reply: It is unfortunately not possible to infer from the snowpit records what the reason for formation of hard layers was. It is very common that wind compaction leads to hardness 4 and 5 at the sites with the snow pits. An error of commission/omission can therefore not be derived. We agree that this should be explained in more detail in the manuscript for clarification.
RC1: I would also propose to compare to another passive ROS detection algorithm to evaluate the benefit of the method. ... L310. Consider comparing your method to a passive-based ROS (Dolant et la., 2016, Pan et al., 2018) retrieval to show the improvement your method could deliver.
Reply: Other algorithms for ROS detection from passive microwave L-band have not been published yet to our knowledge. ROS retrieval results from passive microwave records in general have (to our knowledge) so far only documented with figures in publications and datasets of events not published. One event was however documented by one of our co-authors (Sokolov et al. 2016) with passive microwave (AMSR-E) previously. We discuss this on lines 395 and 510. We fully agree that comparisons to other passive microwave records/frequencies would be useful and specifically exploitation of SSMI/I records in order to be able to go more back in time (see lines 474ff), but this is beyond the scope of this study. The advantage of using radar is however (1) the potential to analyse the severity of the event (e.g. figure 12c) and (2) the option to go to higher spatial resolution with SAR. Current SAR missions do not allow for operational circumpolar retrieval, but it might become feasible in the future. Note that in a preceding study results from fusion of Ku-band backscatter change with AMSR-E detection are documented in Semmens et al. (2013). It was found that many events found over Alaska were due to fog instead of ROS: "... fog occurrence is viewed as a proxy for warm air mass intrusion which creates condensation on the snow surface resulting in melt that is detected by the passive microwave" (Semmens et al. 2013, page 9). We briefly discuss this on lines 475-477.
Regarding the suggested considerations of Dolant et al. (2016) and Pan et al. (2018): Note that I contributed a review of ROS retrieval methods in Serreze et al. (2021), Table 1, which also provides some basic details of the two papers. Pan et al. is commented there also on page 12, left column. Dolant et al. focused for validation (community observations of ROS from three settlements close to each other) on one winter, 2010/11, which is excluded from our ASCAT/SMOS analyses due to issues in the SMOS records at the beginning of the mission. Pan et al. focus on Alaska and combine passive microwave observations with MODIS to identify if snow is on ground or not. Validation was based on precipitation proxies. These differences in validation strategy across existing studies could be added to the introdcution and discussion. Pan et al. included a figure for one specific example for a ROS in 2013, which we could extract from our records for comparison.
References
Dolant, C., Langlois, A., Montpetit, B., Brucker, L., Roy, A., & Roy, A. (2016) Development of a rain-on-snow detection algorithm using passive microwave radiometry. Hydrological Processes, 30, 3184-3196. DOI: 10.1002/hyp.1082
Pan, C. G, Kirchner, P. B, Kimball, J. S, Kim, Y. & Du, J. (2018). Environ. Res. Lett. 13 075004, DOI : 10.1088/1748-9326/aac9d3
Serreze, M. C., Gustafson, J., Barrett, A. P., Druckenmiller, M. L., 660 Fox, S., Voveris, J., Stroeve, J., Sheffield, B., Forbes, B. C., Rasmus, S., Laptander, R., Brook, M., Brubaker, M., Temte, J., McCrystall, M. R., and Bartsch, A.: Arctic rain on snow events: bridging observations to understand environmental and livelihood impacts, Environmental Research Letters, 16, 105 009, https://doi.org/10.1088/1748-9326/ac269b, 2021
Semmens, K. A., Ramage, J., Bartsch, A., and Liston, G. E.: Early snowmelt events: detection, distribution, and significance in a major sub-arctic watershed, Environmental Research Letters, 8, 014 020, https://doi.org/10.1088/1748-9326/8/1/014020, 2013.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC3 -
AC4: 'Reply on RC1 - addressing 'specific comments'', Annett Bartsch, 18 Nov 2022
Please find our further responses to the 'specific comments' below.
L35. I would not use the term “aging of snow” _since it is not accurate. Please refer to the vapour flux from temperature gradient.
- Reply: We agree to modify this accordingly in the revised version.
L38. “The mapping of snow changes afterwards instead of wet snow circumvents”, I’m not sure what is meant in the sentence. Please consider modification for improved clarity.
- Reply: we suggest the following rephrasing
- Old: The mapping of snow changes afterwards instead of wet snow circumvents this issue but requires the use of wavelengths which are sensitive to changes in snow properties, this means comparably short wavelengths with respect to the typical grain size of snow
- New: The mapping of snow structure changes as a result of events instead of wet snow during an event circumvents this issue but requires the use of wavelengths which are sensitive to changes in snow properties, this means comparably short wavelengths with respect to the typical grain size of snow
L40. Include citation on wavelength and snow grain size.
- Reply: We agree to modify this accordingly in the revised version.
l65. Do you mean “With ROS, associated” _
- Reply: yes
L77-80. Perhaps the objective should be modify or addressed more clearly in the conclusion. Were you able to correctly answer (1) with this method? How was (3) evaluated?
- Reply: “(1) gain insight into recent occurrence of rain on snow events across the Arctic” – this refers to the ROS cases with known impact which are detailed in the paper. We suggest to add ‘specific’ before ‘rain’. Regarding (3), the impact of ROS on snow properties was investigated using hardness from snow pit records. This is briefly referred to on line 522 in the conclusions, but we agree that it could be extented.
L150. Consider adding a statement on how these data can be subjective and what was done to avoid this.
- Reply: e.g. “Hardness measurements can be subjective. Specific schemes have been developed to judge hardness (see table 1). For long-term measurement sites such as Varanger, Saariselkä and Sodankylä people doing the measurements undergo training in using these schemes. “
L207. Please clarify this sentence “ROS using wet snow from C-band”. Do you mean wet snow detection?
- Reply: yes. Rephrasing suggestion: “ROS identification based on wet snow detection from C-band …”
L.246. Please reword the beginning of the sentence.
- Reply: please find our suggestion below
- Old: Passive microwave observations as available from SMOS provide two polarizations ..
- New: Passive microwave observations commonly provide two polarizations ...
L248. Algorithms for ROS detection using 37 and 19 GHz are also sensitive to dry snow surface change into ice crust and ice layer. Consider using does to improve the algorithm since L band is useless when no liquid water is present.
- Reply: In this section we list published wet snow detection schemes. But we agree, that this should be mentioned in the outlook when referring to potential use of other passive microwave data .
L521-523 Can you provide a quantitative validation of the method to detect ROS events?
- Reply: The LRI record for Sodankylä can be summarized.
L527. the phrasing with the comma is confusing, do you mean ... “play a role on what should be considered”
- Reply: We mean that the role of frequency and polarization should be studied in more detail.
L526. Maybe consider using a passive observation with 19 and 37 GHz to improve sensitivity to ice crust and dry snow surface change. Once the liquid water is frozen and the temporal timing of the ROS event could not be detected with SMOS, those frequencies could help to detect surface change while C-band can provide info at high resolution.
- Reply: see response to ‘L248’
L529. “The magnitude of specific extreme events can be documented by the use of ASCAT alone, without fusion with SMOS.” I thought you showed you need wet snow detection and ASCAT alone cannot detect ROS.
- Reply: what we mean is that if it is known from other observations that it was a ROS situation, then the cross-check that it was not a ‘temperature-drop’ misclassification is not needed. It refers to the Alaska example, where the ROS occurred rather south, where there was a gap in SMOS.
Figure 6. This figure is hard to understand. what are H4 1, H4 2 and H4 3? Why not add all layers so we have a better understanding of the whole snowpack?
- Reply: In case of use of 1, 2, and 3, there have been 3 separate layers with type H4 in the snowpack. Other layers could be added, but would be less relevant.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC4
-
AC1: 'Reply on RC1', Annett Bartsch, 06 Nov 2022
-
RC2: 'Comment on egusphere-2022-899', Anonymous Referee #2, 27 Oct 2022
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-899/egusphere-2022-899-RC2-supplement.pdf
-
AC2: 'Reply on RC2', Annett Bartsch, 06 Nov 2022
Many thanks for your valuable comments!
Please find responses to your questions below:
RC2: Area of study: it is apparent from the figures that the authors have applied the approach to not only the entire Arctic but also to land areas extending much further south (eg., Figure 5). However, it is stated in the conclusion that the approach is only recommended for regions north of 66 deg due to coverage issues with SMOS. Perhaps the authors should revise the boundaries of the areas for which ROS detections are presented, or comment on how representative the ROS data are for the lower latitude areas shown in Fig.5?
Reply: The analyses extent was chosen based on the boundaries of the existing database from QuikScat. We describe the differences between sensors in forested/southern regions on lines 377ff and discuss it on lines 452-469. In the latter case, a wrong figure is referred to in the submitted manuscript (#2 instead of #5). However, we agree, an extended discussion on how representative the results are south of the treeline/in the SMOS gap zone would be beneficial.
RC2: The authors have described and presented a wide range of different types of observations, but I think that some of the datasets used do not really add much to the overall goals and conclusions of the study. While I see the need for observations to validate/support the remote sensing data, I think the use of too many different observations, each with their own considerations for ROS detection, makes it at times difficult to follow the main objectives of the study. I would for example recommend reconsidering whether the use of the caribou data are really necessary.
Reply: Our reasons for including the Caribou/Serward peinsula study were to (1) demonstrate that there is stronger variability from year to year regionally than what can be observed for the entire Arctic (Figure 5) and (2) to point to the potential use of the backscatter change magnitude in addition to just event detection. The Caribou herds used some areas where an event was detected, but not areas which exceeded a certain back scatter change value (figure 12c). This could be moved to the discussion part of the paper.
RC2: Line 175: why were different terms/hardness scales used at Yamal compared to the Scandinavian sites? Why not just use a standard scale for all sites?
Reply: : The surveys come from different (partially long-term) monitoring programs, carried out by different institutions which follow different schemes. It would be indeed very beneficial if future surveys would follow the same scheme.
RC2: Line 296: “location specific threshold” - does this mean that a threshold is determined for individual pixels, or for regions?
Reply: Yes, the threshold is defined individually for each grid point.
RC2: Table 2: Events represent November 2021 to February 2022; why are values from only 1 year/winter of observations used?
Reply: Thanks for spotting! It should read November – February, years 2011-2022 (note that figure 1 is also positioned in the main text, similar to table 1)
RC2: Figure 4: Could the authors comment on the event confirmed by SMOS occurring in the start of December 2016? Here the AWS data show very low temperature (approx. -15 deg.C), no precipitation and increasing snow depth in the following days. What could be the reason for detection of wet snow?
Reply: The SMOS detection actually refers to a smaller event (LRI < 1mm) two days before the ASCAT detection. The ASCAT detection does however represent a period of temperature drop (following the rain event). The impact of liquid precipitation several days before is not captured. This case shows the disadvantage of using a 3 day window for the SMOS masking, but it is necessary to account for data gaps and nature of the radar retrieval scheme, as discussed on lines 304 in the methods part. We agree that the discussion on the choice of the detection window could be more extensive and included in the discussion.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC2
-
AC2: 'Reply on RC2', Annett Bartsch, 06 Nov 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-899', Anonymous Referee #1, 18 Oct 2022
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-899/egusphere-2022-899-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Annett Bartsch, 06 Nov 2022
Many thanks for your valuable comments!
Please find responses to your questions below:
RC1: Area of study: it is apparent from the figures that the authors have applied the approach to not only the entire Arctic but also to land areas extending much further south (eg., Figure 5). However, it is stated in the conclusion that the approach is only recommended for regions north of 66 deg due to coverage issues with SMOS. Perhaps the authors should revise the boundaries of the areas for which ROS detections are presented, or comment on how representative the ROS data are for the lower latitude areas shown in Fig.5?
Reply: The analyses extent was chosen based on the boundaries of the existing database from QuikScat. We describe the differences between sensors in forested/southern regions on lines 377ff and discuss it on lines 452-469. In the latter case, a wrong figure is referred to in the submitted manuscript (#2 instead of #5). However, we agree, an extended discussion on how representative the results are south of the treeline/in the SMOS gap zone would be beneficial.
RC1: The authors have described and presented a wide range of different types of observations, but I think that some of the datasets used do not really add much to the overall goals and conclusions of the study. While I see the need for observations to validate/support the remote sensing data, I think the use of too many different observations, each with their own considerations for ROS detection, makes it at times difficult to follow the main objectives of the study. I would for example recommend reconsidering whether the use of the caribou data are really necessary.
Reply: Our reasons for including the Caribou/Serward peinsula study were to (1) demonstrate that there is stronger variability from year to year regionally than what can be observed for the entire Arctic (Figure 5) and (2) to point to the potential use of the backscatter change magnitude in addition to just event detection. The Caribou herds used some areas where an event was detected, but not areas which exceeded a certain back scatter change value (figure 12c). This could be moved to the discussion part of the paper.
RC1: Line 175: why were different terms/hardness scales used at Yamal compared to the Scandinavian sites? Why not just use a standard scale for all sites?
Reply: : The surveys come from different (partially long-term) monitoring programs, carried out by different institutions which follow different schemes. It would be indeed very beneficial if future surveys would follow the same scheme.
RC1: Line 296: “location specific threshold” - does this mean that a threshold is determined for individual pixels, or for regions?
Reply: Yes, the threshold is defined individually for each grid point.
RC1: Table 2: Events represent November 2021 to February 2022; why are values from only 1 year/winter of observations used?
Reply: Thanks for spotting! It should read November – February, years 2011-2022 (note that figure 1 is also positioned in the main text, similar to table 1)
RC1: Figure 4: Could the authors comment on the event confirmed by SMOS occurring in the start of December 2016? Here the AWS data show very low temperature (approx. -15 deg.C), no precipitation and increasing snow depth in the following days. What could be the reason for detection of wet snow?
Reply: The SMOS detection actually refers to a smaller event (LRI < 1mm) two days before the ASCAT detection. The ASCAT detection does however represent a period of temperature drop (following the rain event). The impact of liquid precipitation several days before is not captured. This case shows the disadvantage of using a 3 day window for the SMOS masking, but it is necessary to account for data gaps and nature of the radar retrieval scheme, as discussed on lines 304 in the methods part. We agree that the discussion on the choice of the detection window could be more extensive and included in the discussion.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC1 -
AC3: 'Reply on RC1 - correction (first response was for RC2)', Annett Bartsch, 06 Nov 2022
Please apologize the mixing up with comments to RC2.
Also thanks to you for your valuable comments!
Please find our response to your main comments below:
RC1: ... this work lacks a quantitative validation of the ROS retrieval from ice layer I snowpits, maybe % of commission/omission from the algorithm could be calculated.
Reply: It is unfortunately not possible to infer from the snowpit records what the reason for formation of hard layers was. It is very common that wind compaction leads to hardness 4 and 5 at the sites with the snow pits. An error of commission/omission can therefore not be derived. We agree that this should be explained in more detail in the manuscript for clarification.
RC1: I would also propose to compare to another passive ROS detection algorithm to evaluate the benefit of the method. ... L310. Consider comparing your method to a passive-based ROS (Dolant et la., 2016, Pan et al., 2018) retrieval to show the improvement your method could deliver.
Reply: Other algorithms for ROS detection from passive microwave L-band have not been published yet to our knowledge. ROS retrieval results from passive microwave records in general have (to our knowledge) so far only documented with figures in publications and datasets of events not published. One event was however documented by one of our co-authors (Sokolov et al. 2016) with passive microwave (AMSR-E) previously. We discuss this on lines 395 and 510. We fully agree that comparisons to other passive microwave records/frequencies would be useful and specifically exploitation of SSMI/I records in order to be able to go more back in time (see lines 474ff), but this is beyond the scope of this study. The advantage of using radar is however (1) the potential to analyse the severity of the event (e.g. figure 12c) and (2) the option to go to higher spatial resolution with SAR. Current SAR missions do not allow for operational circumpolar retrieval, but it might become feasible in the future. Note that in a preceding study results from fusion of Ku-band backscatter change with AMSR-E detection are documented in Semmens et al. (2013). It was found that many events found over Alaska were due to fog instead of ROS: "... fog occurrence is viewed as a proxy for warm air mass intrusion which creates condensation on the snow surface resulting in melt that is detected by the passive microwave" (Semmens et al. 2013, page 9). We briefly discuss this on lines 475-477.
Regarding the suggested considerations of Dolant et al. (2016) and Pan et al. (2018): Note that I contributed a review of ROS retrieval methods in Serreze et al. (2021), Table 1, which also provides some basic details of the two papers. Pan et al. is commented there also on page 12, left column. Dolant et al. focused for validation (community observations of ROS from three settlements close to each other) on one winter, 2010/11, which is excluded from our ASCAT/SMOS analyses due to issues in the SMOS records at the beginning of the mission. Pan et al. focus on Alaska and combine passive microwave observations with MODIS to identify if snow is on ground or not. Validation was based on precipitation proxies. These differences in validation strategy across existing studies could be added to the introdcution and discussion. Pan et al. included a figure for one specific example for a ROS in 2013, which we could extract from our records for comparison.
References
Dolant, C., Langlois, A., Montpetit, B., Brucker, L., Roy, A., & Roy, A. (2016) Development of a rain-on-snow detection algorithm using passive microwave radiometry. Hydrological Processes, 30, 3184-3196. DOI: 10.1002/hyp.1082
Pan, C. G, Kirchner, P. B, Kimball, J. S, Kim, Y. & Du, J. (2018). Environ. Res. Lett. 13 075004, DOI : 10.1088/1748-9326/aac9d3
Serreze, M. C., Gustafson, J., Barrett, A. P., Druckenmiller, M. L., 660 Fox, S., Voveris, J., Stroeve, J., Sheffield, B., Forbes, B. C., Rasmus, S., Laptander, R., Brook, M., Brubaker, M., Temte, J., McCrystall, M. R., and Bartsch, A.: Arctic rain on snow events: bridging observations to understand environmental and livelihood impacts, Environmental Research Letters, 16, 105 009, https://doi.org/10.1088/1748-9326/ac269b, 2021
Semmens, K. A., Ramage, J., Bartsch, A., and Liston, G. E.: Early snowmelt events: detection, distribution, and significance in a major sub-arctic watershed, Environmental Research Letters, 8, 014 020, https://doi.org/10.1088/1748-9326/8/1/014020, 2013.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC3 -
AC4: 'Reply on RC1 - addressing 'specific comments'', Annett Bartsch, 18 Nov 2022
Please find our further responses to the 'specific comments' below.
L35. I would not use the term “aging of snow” _since it is not accurate. Please refer to the vapour flux from temperature gradient.
- Reply: We agree to modify this accordingly in the revised version.
L38. “The mapping of snow changes afterwards instead of wet snow circumvents”, I’m not sure what is meant in the sentence. Please consider modification for improved clarity.
- Reply: we suggest the following rephrasing
- Old: The mapping of snow changes afterwards instead of wet snow circumvents this issue but requires the use of wavelengths which are sensitive to changes in snow properties, this means comparably short wavelengths with respect to the typical grain size of snow
- New: The mapping of snow structure changes as a result of events instead of wet snow during an event circumvents this issue but requires the use of wavelengths which are sensitive to changes in snow properties, this means comparably short wavelengths with respect to the typical grain size of snow
L40. Include citation on wavelength and snow grain size.
- Reply: We agree to modify this accordingly in the revised version.
l65. Do you mean “With ROS, associated” _
- Reply: yes
L77-80. Perhaps the objective should be modify or addressed more clearly in the conclusion. Were you able to correctly answer (1) with this method? How was (3) evaluated?
- Reply: “(1) gain insight into recent occurrence of rain on snow events across the Arctic” – this refers to the ROS cases with known impact which are detailed in the paper. We suggest to add ‘specific’ before ‘rain’. Regarding (3), the impact of ROS on snow properties was investigated using hardness from snow pit records. This is briefly referred to on line 522 in the conclusions, but we agree that it could be extented.
L150. Consider adding a statement on how these data can be subjective and what was done to avoid this.
- Reply: e.g. “Hardness measurements can be subjective. Specific schemes have been developed to judge hardness (see table 1). For long-term measurement sites such as Varanger, Saariselkä and Sodankylä people doing the measurements undergo training in using these schemes. “
L207. Please clarify this sentence “ROS using wet snow from C-band”. Do you mean wet snow detection?
- Reply: yes. Rephrasing suggestion: “ROS identification based on wet snow detection from C-band …”
L.246. Please reword the beginning of the sentence.
- Reply: please find our suggestion below
- Old: Passive microwave observations as available from SMOS provide two polarizations ..
- New: Passive microwave observations commonly provide two polarizations ...
L248. Algorithms for ROS detection using 37 and 19 GHz are also sensitive to dry snow surface change into ice crust and ice layer. Consider using does to improve the algorithm since L band is useless when no liquid water is present.
- Reply: In this section we list published wet snow detection schemes. But we agree, that this should be mentioned in the outlook when referring to potential use of other passive microwave data .
L521-523 Can you provide a quantitative validation of the method to detect ROS events?
- Reply: The LRI record for Sodankylä can be summarized.
L527. the phrasing with the comma is confusing, do you mean ... “play a role on what should be considered”
- Reply: We mean that the role of frequency and polarization should be studied in more detail.
L526. Maybe consider using a passive observation with 19 and 37 GHz to improve sensitivity to ice crust and dry snow surface change. Once the liquid water is frozen and the temporal timing of the ROS event could not be detected with SMOS, those frequencies could help to detect surface change while C-band can provide info at high resolution.
- Reply: see response to ‘L248’
L529. “The magnitude of specific extreme events can be documented by the use of ASCAT alone, without fusion with SMOS.” I thought you showed you need wet snow detection and ASCAT alone cannot detect ROS.
- Reply: what we mean is that if it is known from other observations that it was a ROS situation, then the cross-check that it was not a ‘temperature-drop’ misclassification is not needed. It refers to the Alaska example, where the ROS occurred rather south, where there was a gap in SMOS.
Figure 6. This figure is hard to understand. what are H4 1, H4 2 and H4 3? Why not add all layers so we have a better understanding of the whole snowpack?
- Reply: In case of use of 1, 2, and 3, there have been 3 separate layers with type H4 in the snowpack. Other layers could be added, but would be less relevant.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC4
-
AC1: 'Reply on RC1', Annett Bartsch, 06 Nov 2022
-
RC2: 'Comment on egusphere-2022-899', Anonymous Referee #2, 27 Oct 2022
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-899/egusphere-2022-899-RC2-supplement.pdf
-
AC2: 'Reply on RC2', Annett Bartsch, 06 Nov 2022
Many thanks for your valuable comments!
Please find responses to your questions below:
RC2: Area of study: it is apparent from the figures that the authors have applied the approach to not only the entire Arctic but also to land areas extending much further south (eg., Figure 5). However, it is stated in the conclusion that the approach is only recommended for regions north of 66 deg due to coverage issues with SMOS. Perhaps the authors should revise the boundaries of the areas for which ROS detections are presented, or comment on how representative the ROS data are for the lower latitude areas shown in Fig.5?
Reply: The analyses extent was chosen based on the boundaries of the existing database from QuikScat. We describe the differences between sensors in forested/southern regions on lines 377ff and discuss it on lines 452-469. In the latter case, a wrong figure is referred to in the submitted manuscript (#2 instead of #5). However, we agree, an extended discussion on how representative the results are south of the treeline/in the SMOS gap zone would be beneficial.
RC2: The authors have described and presented a wide range of different types of observations, but I think that some of the datasets used do not really add much to the overall goals and conclusions of the study. While I see the need for observations to validate/support the remote sensing data, I think the use of too many different observations, each with their own considerations for ROS detection, makes it at times difficult to follow the main objectives of the study. I would for example recommend reconsidering whether the use of the caribou data are really necessary.
Reply: Our reasons for including the Caribou/Serward peinsula study were to (1) demonstrate that there is stronger variability from year to year regionally than what can be observed for the entire Arctic (Figure 5) and (2) to point to the potential use of the backscatter change magnitude in addition to just event detection. The Caribou herds used some areas where an event was detected, but not areas which exceeded a certain back scatter change value (figure 12c). This could be moved to the discussion part of the paper.
RC2: Line 175: why were different terms/hardness scales used at Yamal compared to the Scandinavian sites? Why not just use a standard scale for all sites?
Reply: : The surveys come from different (partially long-term) monitoring programs, carried out by different institutions which follow different schemes. It would be indeed very beneficial if future surveys would follow the same scheme.
RC2: Line 296: “location specific threshold” - does this mean that a threshold is determined for individual pixels, or for regions?
Reply: Yes, the threshold is defined individually for each grid point.
RC2: Table 2: Events represent November 2021 to February 2022; why are values from only 1 year/winter of observations used?
Reply: Thanks for spotting! It should read November – February, years 2011-2022 (note that figure 1 is also positioned in the main text, similar to table 1)
RC2: Figure 4: Could the authors comment on the event confirmed by SMOS occurring in the start of December 2016? Here the AWS data show very low temperature (approx. -15 deg.C), no precipitation and increasing snow depth in the following days. What could be the reason for detection of wet snow?
Reply: The SMOS detection actually refers to a smaller event (LRI < 1mm) two days before the ASCAT detection. The ASCAT detection does however represent a period of temperature drop (following the rain event). The impact of liquid precipitation several days before is not captured. This case shows the disadvantage of using a 3 day window for the SMOS masking, but it is necessary to account for data gaps and nature of the radar retrieval scheme, as discussed on lines 304 in the methods part. We agree that the discussion on the choice of the detection window could be more extensive and included in the discussion.
Citation: https://doi.org/10.5194/egusphere-2022-899-AC2
-
AC2: 'Reply on RC2', Annett Bartsch, 06 Nov 2022
Peer review completion
Journal article(s) based on this preprint
Annett Bartsch et al.
Annett Bartsch et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 441 | 160 | 21 | 622 | 4 | 5 |
- HTML: 441
- PDF: 160
- XML: 21
- Total: 622
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(15160 KB) - Metadata XML