the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Snowmelt Characterization from Optical and Synthetic Aperture Radar Observations in the Lajoie Basin, British Columbia
Sara E. Darychuk
Joseph M. Shea
Brian Menounos
Anna Chesnokova
Georg Jost
Frank Weber
Abstract. Snowmelt runoff serves both human needs and ecosystem services and is an important parameter in operational forecasting systems. Sentinel-1 Synthetic Aperture Radar (SAR) observations can estimate the timing of melt within a snowpack; however, these estimates have not been applied on large spatial scales. We present here a workflow to fuse Sentinel-1 SAR and optical data from Landsat-8 and Sentinel-2 to estimate the onset and duration of snowmelt in the Lajoie Basin, a large watershed in the Southern Coast Mountains of British Columbia. A backscatter threshold is used to infer the point at which snowpack saturation occurs, and the snowpack begins to produce runoff. Multispectral imagery is used to estimate snow free dates across the basin to define the end of the snowmelt period. SAR estimates of snowmelt onset form consistent trends by elevation and aspect on the watershed scale and reflect snowmelt records from continuous SWE observations. SAR estimates of snowpack saturation are most effective on moderate to low slopes (< 30°) in open areas. The accuracy of snowmelt durations is reduced due to persistent cloud cover in optical imagery. Despite these challenges, snowmelt durations agree with trends in snow depths observed in the Lajoie. This approach has high potential for adaptability to other alpine regions and can provide estimates of snowmelt timing in ungauged basins.
- Preprint
(1784 KB) -
Supplement
(2261 KB) - BibTeX
- EndNote
Sara E. Darychuk et al.
Status: closed
-
RC1: 'Comment on tc-2022-89', Carlo Marin, 24 Jun 2022
The paper presents a method to estimate the snowmelt duration exploiting Sentinel-1, Sentinel-2 and Landsat-8 images. The method was applied in the Lajoie basin in British Columbia considering four melting seasons. The method exploits the multitemporal information provided by Sentinel-1 to identify the run-off onset, whereas sentinel-2 was used to identify the end of the melting season. In particular, this was possible by exploiting the relationship between the melting phases and multitemporal SAR backscattering originally described in Marin et al. 2020 and the capability of the multispectral optical data acquired in the visible and short wave infrared bands to identify the snow cover. Differently from the paper of Marin et al. 2020, the Sentinel-1 data has been studies outside the European Alps, with different land cover types, included forested areas, and considering, for some of the analyzed year lower temporal frequency of acquisition i.e., 12 days or more. The evaluation of the results suggests the possibility to performed detailed analysis of the melting season. The paper is well written but there are some points that can be further improved increasing in this way the general interest for the work.
Specific comments
Pag 7 line 173. It would be interesting to know how much the percentage of infilled areas is. Moreover, it is not clear, at least to me in the present from, if the signature always developed also in the cases where the dates were outside of two and a half standard deviation.
Pag 8 line 198. The identification of the end of the season using sparsely acquired high resolution data is a very challenging task. Some methods have been presented in the literature that try to address this problem using HR multi-source (optical-optical and optical-SAR) data and it is worth to mentioning them.
Table 2. It is not clear, at least to me in the present form, what the image frequency is. For example 1-7 days refer to the day in which there is an acquisition?
Table 3. It would be interesting to know the percentage of image for which the cloud cover is less than a given low threshold e.g., 30%.
Figure 2. It is not clear why the shaded blue representing the melt period is not stopping at SWE = 0 for some years. I think this is the rule to be applied once the onset is identified.
Interestingly the onset for the runoff is derived in simplified snow model by considering the average temperature (and the radiation) i.e., degree day model. If air temperatures are available for the Lajoie basin, it would be interesting to discuss the difference between Sentinel-1 in identifying the runoff onset (temperature can be spatialized at high resolution and thresholded accordingly).
The sampling time provided by Sentinel-1 seems to be not adequate, in Shannon sense, to properly sampling the melting which has probably a temporal resolution less than one day. That means that the error could be potentially of several days. How is this uncertainty propagating in the case of snow melt duration analysis when different years are compared? What is the ideal revisit time needed for this kind of analysis?
I’m looking forward to seeing the snow melt duration for all the southern cost mountains of British Columbia at least for one year. Do you think this is possible? In case I would comment the main challenges of this operation in the manuscript.
Citation: https://doi.org/10.5194/tc-2022-89-RC1 - AC1: 'Reply on RC1', Sara Darychuk, 15 Sep 2022
-
RC2: 'Comment on tc-2022-89', Giacomo Bertoldi, 22 Aug 2022
The paper applies the recent approach proposed by Marin et. al. 2020 to estimate snowmelt timing using Sentinel-1 radar backscattering to a mountainous, partially forested catchment. It shows promising results, especially for not too steep areas. The quality of the analysis is good and convincing.
I have only minor comments.
1. Introduction. Please define in a more precise way the paper's aims and formulate clear research questions.
2. Section 4.2 snow disappearance times. Since Landsat has a low overpass time and images are often cloud covered, please better quantify the errors and uncertainties in days for the snow disappearance time. c
3. Section 5.2 Sensitivity - L233 - "the least accurate approximation in 2019" - please quantify in numbers, variance ...
4. L255 - besides slope, does the accuracy of results change also with the aspect?
5. Paragraph at line 315 - Interesting! Maybe an additional Figure can support this!
6. L324 - What do you mean by Data Fusion? Please explain better!
Citation: https://doi.org/10.5194/tc-2022-89-RC2 - AC2: 'Reply on RC2', Sara Darychuk, 23 Sep 2022
Status: closed
-
RC1: 'Comment on tc-2022-89', Carlo Marin, 24 Jun 2022
The paper presents a method to estimate the snowmelt duration exploiting Sentinel-1, Sentinel-2 and Landsat-8 images. The method was applied in the Lajoie basin in British Columbia considering four melting seasons. The method exploits the multitemporal information provided by Sentinel-1 to identify the run-off onset, whereas sentinel-2 was used to identify the end of the melting season. In particular, this was possible by exploiting the relationship between the melting phases and multitemporal SAR backscattering originally described in Marin et al. 2020 and the capability of the multispectral optical data acquired in the visible and short wave infrared bands to identify the snow cover. Differently from the paper of Marin et al. 2020, the Sentinel-1 data has been studies outside the European Alps, with different land cover types, included forested areas, and considering, for some of the analyzed year lower temporal frequency of acquisition i.e., 12 days or more. The evaluation of the results suggests the possibility to performed detailed analysis of the melting season. The paper is well written but there are some points that can be further improved increasing in this way the general interest for the work.
Specific comments
Pag 7 line 173. It would be interesting to know how much the percentage of infilled areas is. Moreover, it is not clear, at least to me in the present from, if the signature always developed also in the cases where the dates were outside of two and a half standard deviation.
Pag 8 line 198. The identification of the end of the season using sparsely acquired high resolution data is a very challenging task. Some methods have been presented in the literature that try to address this problem using HR multi-source (optical-optical and optical-SAR) data and it is worth to mentioning them.
Table 2. It is not clear, at least to me in the present form, what the image frequency is. For example 1-7 days refer to the day in which there is an acquisition?
Table 3. It would be interesting to know the percentage of image for which the cloud cover is less than a given low threshold e.g., 30%.
Figure 2. It is not clear why the shaded blue representing the melt period is not stopping at SWE = 0 for some years. I think this is the rule to be applied once the onset is identified.
Interestingly the onset for the runoff is derived in simplified snow model by considering the average temperature (and the radiation) i.e., degree day model. If air temperatures are available for the Lajoie basin, it would be interesting to discuss the difference between Sentinel-1 in identifying the runoff onset (temperature can be spatialized at high resolution and thresholded accordingly).
The sampling time provided by Sentinel-1 seems to be not adequate, in Shannon sense, to properly sampling the melting which has probably a temporal resolution less than one day. That means that the error could be potentially of several days. How is this uncertainty propagating in the case of snow melt duration analysis when different years are compared? What is the ideal revisit time needed for this kind of analysis?
I’m looking forward to seeing the snow melt duration for all the southern cost mountains of British Columbia at least for one year. Do you think this is possible? In case I would comment the main challenges of this operation in the manuscript.
Citation: https://doi.org/10.5194/tc-2022-89-RC1 - AC1: 'Reply on RC1', Sara Darychuk, 15 Sep 2022
-
RC2: 'Comment on tc-2022-89', Giacomo Bertoldi, 22 Aug 2022
The paper applies the recent approach proposed by Marin et. al. 2020 to estimate snowmelt timing using Sentinel-1 radar backscattering to a mountainous, partially forested catchment. It shows promising results, especially for not too steep areas. The quality of the analysis is good and convincing.
I have only minor comments.
1. Introduction. Please define in a more precise way the paper's aims and formulate clear research questions.
2. Section 4.2 snow disappearance times. Since Landsat has a low overpass time and images are often cloud covered, please better quantify the errors and uncertainties in days for the snow disappearance time. c
3. Section 5.2 Sensitivity - L233 - "the least accurate approximation in 2019" - please quantify in numbers, variance ...
4. L255 - besides slope, does the accuracy of results change also with the aspect?
5. Paragraph at line 315 - Interesting! Maybe an additional Figure can support this!
6. L324 - What do you mean by Data Fusion? Please explain better!
Citation: https://doi.org/10.5194/tc-2022-89-RC2 - AC2: 'Reply on RC2', Sara Darychuk, 23 Sep 2022
Sara E. Darychuk et al.
Sara E. Darychuk et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 423 | 157 | 13 | 593 | 44 | 5 | 4 |
- HTML: 423
- PDF: 157
- XML: 13
- Total: 593
- Supplement: 44
- BibTeX: 5
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1