Spaceborne thermal infrared observations of Arctic sea ice leads at 30 m resolution
Abstract. Sea ice leads are elongated fractures within sea ice cover, playing an important role in the heat exchange from the ocean to the overlying atmosphere. Narrow leads less than a hundred meters in width contribute considerable heat fluxes, requiring fine-scale observation of Arctic leads. With the launch of Sustainable Development Science Satellite 1 (SDGSAT-1) by China on 5 November 2021, the on-board Thermal Infrared Spectrometer (TIS) provides thermal infrared imagery at an unprecedented resolution of 30 m in a swath of 300 km. We propose a method adapted to the TIS high-resolution infrared images for lead detection in the Arctic. For the first time, the spatial resolution of leads by infrared remote sensing increases from the scale of kilometers to tens of meters. For the Beaufort Sea cases in April 2022, the detection is consistent with the Sentinel-2 visible images, yielding an overall accuracy of 96.30 %. Compared with the Moderate-Resolution Imaging Spectroradiometer (MODIS), the TIS presents more leads with width less hundreds of meters than the results based on the MODIS data. For the three infrared bands of the TIS, the B2 (10.3–11.3 µm) and B3 (11.5–12.5 µm) bands, show similar performances in detecting leads. The B1 band (8.0–10.5 µm) can be complementary to the other two bands, as the temperature measurement sensitivity is different from the other two, benefiting better detection by combining the three bands. This study demonstrates that SDGSAT-1 TIS data at 30 m resolution is well applicable for observing previously unresolvable ice leads, and will provide insight into the contribution of narrow leads to rapid sea ice changes in the Arctic.
Yujia Qiu et al.
Status: final response (author comments only)
RC1: 'Comment on egusphere-2022-1506', Anonymous Referee #1, 01 Feb 2023
CC1: 'Reply on RC1', Yujia Qiu, 09 Mar 2023
- AC1: 'Reply on RC1', Xiaoming Li, 16 Mar 2023
- AC1: 'Reply on RC1', Xiaoming Li, 16 Mar 2023
- CC1: 'Reply on RC1', Yujia Qiu, 09 Mar 2023
RC2: 'Comment on egusphere-2022-1506', Anonymous Referee #2, 13 Feb 2023
- CC2: 'Reply on RC2', Yujia Qiu, 09 Mar 2023
- AC2: 'Reply on RC2', Xiaoming Li, 16 Mar 2023
Yujia Qiu et al.
Yujia Qiu et al.
Viewed (geographical distribution)
The article provides an overview of a new satellite that allows for detection of sea ice leads at a spatial resolution not possible with any other satellite. This is novel and interesting, however, the results as presented appear to over-emphasize the accuracy and precision of the results given the limited extent of the analysis.
The comparisons against other moderate resolution products is incomplete. The focus of the analysis presented is on narrow leads that go undetected at moderate resolution (1km). The equally important but overlooked analysis would be the moderate (1km) resolutions results interpolated into higher resolution (30m). When detecting the same lead, how often do moderate resolution results over estimate (or under estimate) the lead area relative to the higher resolution detections. The claim that high resolution results detect more lead area because it can detect narrower leads is incomplete without establishing that moderate resolution lead detections do not have a bias in lead area. Because, visual inspection of the results suggest that moderate resolution lead detections have a bias in terms of over-representing lead area for leads that are wide enough to be detected. It may be true that more leads can be detected by a 30 m resolution imager, but the total lead area detected at 30 m resolution may not necessarily be higher overall if the 30 m lead detections do not have the same bias towards over-representing lead area for larger leads.
The justification for using a 3 channel approach is a little weak. There is mention of how there is ozone absorption in band 1, but the authors appear to be correlating ozone absorption with the air temperature near the surface rather than the air temperature in the ozone layer in the upper atmosphere. Also, the analysis of how the 11 and 12 micron channels (bands 2 and 3) compare in their sensitivity to ice vs water clouds is lacking; again the thermal contribution of water vapor and ice would tend to be much higher in the atmosphere than 2m.
Line 9: Use “sea ice” rather than “sea ice cover”.
Line 9: Use “between” rather than “from” because the heat exchange can go in either direction.
Line 15: The I-band on VIIRS has an IR resolution of 375m at 11 microns. So I would say that the resolution is an order of magnitude improvement rather than 2 orders of magnitude; hundreds of meters rather than kilometers.
Line 16: It does not seem appropriate to attribute 4 significant digits to the results; the results may be 96% accurate, but the precision of the 0.01% is not likely.
Line 35-36: There are examples of the heat flux occurring in either direction (see above comment about line 9)
Line 54: This overlooks the importance of clouds, cloud shadows, and the thermal contrast as ice ages.
Line 76: This 1993 paper is too old – it references AVHRR – which does not describe the detection capabilities of modern satellite imagers.
Line 78: Clarify what is meant by “resolve”. Satellite imager can “detect” sub-resolution thermal emissions – if the thermal contrast is larger enough. Does this line mean just mean that it hard to attribute a width to sub-resolution features?
Line 92: The word “parallel” can be removed, it does not add any descriptive value.
Line 111, Table 1: Is the expected noise still on the order of 0.2K at the cold end of the temperature range?
Line 131: What is level-4 data?
Line 135: The first sentence is hard to understand. Could it be rephrased as “is a two satellite constellation” rather than “is formed by two satellites”?
Line 139: Do you just mean that 560 nm is close to green on the visible spectrum? Why is that important? Is this saying that leads appear to be green in color?
Line 150 & 151: Remove “The” in front of “MODIS”.
Line 152: Why are level 3 products used instead of level 2 (or level 1)?
Line 158: While it is true that MODIS-Terra crosses the equator in the morning, this does not have any correlation with what time of day the satellite will provide coverage in the Arctic. And again, level 1 or 2 products will provide a better time-match than averaged level-3 products.
Line 160-164: What is the importance of the near-surface air temperature? If there is ozone, water vapor, or ice crystal absorption, those phenomena would be occurring much higher in the atmosphere.
Line 170-174: Without objecting to the accuracy of this section, I don’t know that it is necessary for this paper.
Figure 2: The chart shows 1 path going to Step 1 and 1 path going straight to Step 2. By what logic can Step 1 be bypassed?
Line 228: On example is presented, is this representative of what other cases look like?
Line 229: False-positive detections could be clouds, cloud shadows, or cloud edges; not just sea ice.
Line 232: “Multiple tests” needs further explanation.
Figure 7: Do not slit the figure across 2 pages. And, in (a), what do the numbers 1,2,3 mean?
Line 247: Hard to follow.
Line 248: Which temperature gradient are you talking about? Ice surface temperature, retrieved brightness temperature, surface air temperature?
Line 253: The false positives are likely clouds or cloud artifacts.
Line 256: Does “in view 1” mean Figure 7, Panel 1?
Line 257: Is this true for just this case, is it also true for other times of day or times of year when the ice and clouds may tend to have different temperatures?
Line 258: When you say “remaining”, where do you mean, Figure 7 (a)?
Line 279: “co-located” is more commonly used than “collocated”.
Line 281: Is this just saying that leads are darker than ice in the visible spectrum? I think it is already well understood that water is visually darker than ice.
Line 285: So this is not a binary mask, it could have 3 outcomes: lead, non-lead, ambiguous?
Figure 8: Why is the B1 brightness temperature (blue) so much colder than B2 & B3 (yellow)?
Table 3: How can you have a binary comparison when you are excluding brightness values between 0.007-0.7 (see not on Line 285)?
Table 5: How do these numbers change if the temperature threshold is changed?
Line 360: Plead define strip noise.
Table 6: Do not split table across two pages. Also, the ratio is 3 cases are on the order of 1, but Case 2 is more than 5 times higher. It is hard to generalize a relationship with such a big outlier yet a small sample size.
Line 441: Do you mean 30 m instead of 30 km?
Figure 12: Why is the 2m air temperature shown? If Ozone is contributing to the brightness temperature retrieval, that thermal contribution would be coming from much higher in the atmosphere.