Review of

National Weather Service Alaska Sea Ice Program: Gridded ice concentration maps for the Alaskan Arctic

by

Pacini, A., et al.

Summary:

This manuscript introduces gridded sea ice concentration maps available since 2007 from the National Weather Service Alaska Sea Ice Program for the Alaskan Arctic. The main content of the manuscript is the evaluation of this new product (called ASIP henceforth) by means of comparing it with independent data. These are ship-based observations and saildrone images, the MASIE ice extent product and a high-resolution sea ice concentration product. The comparison shown focusses a lot onto so-called parity plots in which the hit and false alarm rates of binary ice information provided and/or derived from the products is compared with each other. Conclusions are drawn from these plots; in addition to these the authors also look a bit into the comparison of the actual sea ice concentration values and take a look at the location o the ice edge and how this intercompares between the different products usd.

I am listing a number of general concerns first. Subsequently you find a number of specific comments which in part detail the general concerns further. I also have a few editoral comments / typos.

General Comments:

GC1: I have a major concern with the scientific rationale and motivation to evaluate a product (your product) providing more information than just binary ice / no ice mainly by means of reducing the information content to compare it with evaluation data that (only!) partly also come as binary information. This I really don't understand and find it neither convincingly explained in the manuscript nor do I find compelling evidence in the manuscript that doing the evaluation this way really adds value and provides credible and useful results.

GC2: The manuscript is not convincing with respect to the description of the steps that are undertaken to i) grid all data into one common grid and to ii) explain how data sets are reduced in their information content from sea ice concentration to binary information - including the assciated uncertainty that is involved in this conversion process.

GC3: I am not convinced that the suite of parity plots that is presented are the optimal solution to show the quality of the new data set that you are evaluating in your manuscript. While I believe 1-2 specific parity plots could stay - especially when these are used to compare data that are per se binary, i.e. the saildrone data and MASIE, I very much recommend to work more with 2-dimensional histograms such as the one shown in Figure 6 and work along the lines of computing mean and median differences (also the absolute ones) and their standard deviations. This appears to me a more quantitative way to evaluate the ASIP product in its current form.

GC4: In case the parity plots stay as a central element of the manuscript I recommend to reshape them such that they use the space given in the manuscript more efficiently - i.e. decrease the block size but increase the font size.

GC5: The Discussion section should be before the Summary section. The discussion section should furthermore discuss in substantially more depth the limitations of the data sets involved - as laid out in my respective specific comment.

GC6: I find room for improvement in the structure of the manuscript. I find that data, methodology and results are in part quite mixed and call for a better organization in that respect.

Specific Comments:

L19: "in-situ asset distribution" --> Not immediately clear what you mean with this. What do you mean by "asset" in this context?

L56: I suggest to add at least Lavergne et al., 2019,  https://doi.org/10.5194/tc-13-49-2019 to this list since it adds a novel approach. Also, you might want to point towards the Ivanova et al., 2014 10.1109/TGRS.2014.2310136 / 2015 doi:10.5194/tc-9-1797-2015 papers here since these provide a good overview of the different existing approaches.

L64: "synthetic aperture radar" --> There is a growing number of sea ice cover / sea ice concentration products based on SAR data; recent years have seen a boost in such maps thanks to more frequent coverage of the polar regions with SAR images and advanced computational tools. I was wondering whether you should not come up with a few examples of such tools / products for completeness. There is for instance the "MAGIC" tool (see Leigh et al., 2014, 10.1109/TGRS.2013.2290231 ) and there are other products, e.g. DTU_AI4Arctic.

L68-70: "Operational ... products." --> I agree only partly to this statement because operational ice charts - at least those of most ice services - use polygons to provide information of groups of dominant ice classes. In addition these only provide ice concentration ranges of, e.g., 10% resolution.

L76: You refer to "many ... techniques used" but you do not further refer to them. Is this on purpose? Because, in what follows you rather report on the results of evaluation studies dealing with two such different products. And in contrast to the CIS sea ice charts the MASIE product is not an operational sea ice product that can be used for navigation but is simply another form of deriving the sea ice extent. I was therefore wondering whether first mentioning a few more "real" ice charts, such as from NIC, AARI, and the Norwegian, Danish and Finish ice services would not make sense.

I note that it would be helpful to provide the period (i.e. number of years) that were used in the two evaluation studies mentioned.

L82/83: It might make sense to emphasize that this larger sea ice extent reported for MASIE by Meier et al. (2015) is particularly large / pronounced during summer melt, right?

L147-153: Have these maps ever been compared to AARI or NIC charts? If not why not?

L175-177: "Polygons ... a larger polygon." --> This I don't understand ... Does this mean that if there is a large polygon containing 70-90% sea ice concentration within which there is a smaller polygon with 10-30% sea ice concentration will result in the entire area (small + large polygon) to be displayed as 10-30% sea ice concentration? Please modify your writing such that it becomes more clear.

Table 1: There is no SIC value in the last row. Does this mean that a value of 100% is never given - also not for landfast sea ice? This reads a bit strange I have to admit.

249-252: While details of the respective data analysis can be found in the Chiodi et al paper I would like to see a more balanced approach (when compared to the ship-based observations) and ask for some basic description about the spatial and temporal resolution of these saildrone data, the observations height and approximate "footprint" and information like this.

L223/224: Worby and Comiso (2004) studied Antarctic sea ice and hence "evaluated" the ASPeCt observations; ASSIST is something which was combined with ASPeCt substantially later, kind of in parallel to the ASPeCt / ASSIST data set that is available, e.g. here: https://www.cen.uni-hamburg.de/en/icdc/data/cryosphere/seaiceparameter-shipobs.html

L239/240: How is this conversion done? Please give a description here or refer to the place in the paper where the respective information is given.

L241/242: I am not sure the mentioned "subjectivity" is something you need to remove - for two reasons. First of all, also the ice charts contain a certain degree of subjectivity. Secondly, the ASPeCt / ASSIST sea ice observations have a reported uncertainty which is similar to the one you reported in the previous section about the ASIP data set; it is around 5-10%. So the uncertainties are the same and I do not see added value to assess ASIP with binary ice/no-ice values.

L260-263: "It utilizes ... stereographic grids" --> This needs to be rewritten. The framing information is: 

- AMSR2 is a multi-frequency passive microwave sennsor that provided brightness temperatures at a number of different frequencies; one of these is 89GHz.

- The ARTIST algorithm has been developed for SSM/I data (Kaleschke et al., 2001), adopted to AMSR-E data (Spreen et al.,  2008) and then applied to AMSR2 data - without further tie point modification as far as I know.

- Sea ice concentration data are derived using the brightness temperature polarization difference of the 89 GHz channels (not from "swath brightness data").

- I invite you to check whether the brightness temperatures aren't first gridded into the polarstereographic grid before the SIC is computed. You might want to check the documentation.

And as a comment: You use this ASI algorithm SIC data for kind of an "evaluation". While this is of course fine I was wondering whether you can report about any validation studies that report about the accuracy of the AMSR2 SIC product provided by the University of Bremen. How reliable is this data set? It is credible to use this data for an evaluation?

L275: So you regrid the MASIE data but you do not regrid the AMSR2 SIC data? At least you did not comment on that in the previous paragraph.

L280-282: Again my question why? Why did you not use the concentration values as provided?

And: How did you do the ice / no ice conversion for these data sets?

L294/295: While it is true that historically 15% has been used as the SIC threshold to define where there is ice, I find your approach not well motivated. ASIP provides non-binary observations (see Table 1) and these should be evaluated - not the binary value.

In addition, seeing that you included MASIE which uses a 40% threshold to define between ice and no-ice, I get confused about the credibility of your results. This does not look like a well-thought through intercomparison approach, I am sorry.

L296-298: "We note ... with ice" --> I don't understand this sentence.

L313-319 ... what is given here is essentially a description of the methodology. I suggest to have a more clear structure in the paper, with a Data section, a Methods section and then a Results section.

L319-323: This information actually belongs to the section where you described the saildrone observations / data.

L327/328: I don't understand how data products (e.g. ASIP or MASIE) can "report" an accuracy. Please re-consider your writing.

L328/330:  I don't think it is a credible approach to refer to over- or under-prediction of ice when the respective SIC ranges that you are considering here are as large as 40% or 60%.

What happens to these n=13 or n=12 (Q3) and n=23 (Q2) values if you would change the threshold value of 40% used by 5% or 10%, i.e. the uncertainty of the involved products?

Figure 5: 

- The font size used is quite small.

- It would be helpful to have Q1 to Q4 denoted again in at least one of the panels.

- In the caption you write "in-situ observation ... (non-binary)". I am confused ... so here you binned the ice products but not the evaluation data? Why? This is inconsistent.

L341/342: "but at this point framework" ---> Why? I doubt that this is a useful comparsison and that it provides a credible result.

L348: "binned at ..." --> To me this looks as if this would result in 11 bins but Figure 6 contains 12 bins at both the x and the y-axis. Also the annotation with 10, 20, 30 ... does not fit well with the respective bin boundaries of 5, 15, 25, 35, 45% et cet. Please check and if need be correct.

L353-359: "Subsequently ... by AMSR2" --> I don't think that this step, particularly in this over-simplified fashion, adds value to what is shown in Fig. 6. I suggest you compute the overall difference and its standard deviation (or the RMSE) and to also compute the mean absolute difference. Both you can report in a separate table or in the text.

L370+ / Figure 7: I don't find this additional parity plot useful. The information one can take from this figure one can as well simply take from Figure 6.

L404: "where the products most strongly disagree" --> Which you could again nicely derive from Fig. 6 by computing the mean SIC difference and the mean absolute SIC difference using the in-situ SIC range of 15-80%.

L406: I don't understand what this "accuracy rate" is. Are you computing the SIC difference? Possibly not because you seem to refer to the ice edge only. So what are you looking at here? The accuracy of which geophysical parameter? And why "rate"

L410-413: Sorry, but I don't understand what you did here. I see an accuracy rate given in percent at the y-axis (in %) but I don't know of which parameter and I see a distance from the ice edge in km (possibly the center of the grid cells are taken - even though I recall that you were reprojecting data onto a 0.05 degree grid ...).  But what do the curves tell me?

L422/423: I don't understand the purpose of this 3x3 pixel window smoothing. Why do you want to remove small-scale features? What is the motivation / scientific rationale behind this step?

L423-426: Please check the scientific literature with respect to the ice edge delination as carried out by you. There should be several papers published that have done this (e.g. Cortenay Strong et al. "On the definition of marginal ice zone width", Journal of Atmospheric and Oceanic Technology, 34, 2017). You might want to check whether your idea is similar to their's and cite and/or check the existing literature for more examples to back up your approach better.

L450/451++: "This is likely ..." --> maybe yes, but not necessarily because at the ice concentration ranges (around 15% and around 40%) you are considering here, the melt pond fraction on the sea ice should be rather small because ice floes have disintegrated and quite some amount of the ice encountered might be brash ice.

I invide the authors to check the available literature about other possibilities to explain the observed discrepancies. There has been a study about why MASIE shows ice while other products don't, for instance.

In general, what should follow here is a discussion into the direction of the credibility of the approaches compared. Influencing factors are the grid resolution and/or the resolution of the input data. This applies to ASIP, MASIE and AMSR2-ASI. Please carefully check how ASI treats potential spurious ice along the ice edge due to the elevated weather effect one has to deal with at 89 GHz. If I am not mistaken, then the ASI algorithm is actually kind of a hybrid product where "bad" sea ice is filtered away by using other, coarser resolution SIC data.

Another issue you might want to discuss is the tendency for ice analysts to, as a first guess, take the conditions of the previous day - especially if there are not enough (high-resolution) satellite data of the day in question at hand. How often is the information given in the ASIP or MASIE product actually based on coarse resolution satellite data from passive microwave sensors (e.g. 25 km)? 

Another issue not touched by you is the fact that ASIP uses polygons and that you are dealing with a sea ice concentration range. Neither the location and extent of the polygon nor the sea ice concentration range in these are overly well defined or FAIR in the sense that repeated analysis would result in exactly the same result; it is not transparent.

Finally, how much are ASIP maps generated in the sense to provide maximum safety for navigation and therefore - similarly to the various ice charts available - come up with a rather conservative estimate, likely tending to overestimate the true ice conditions for the sake of maritime safety?

L485-486: "Since the ... " --> Ok, but how much "hand-waving" is involved into drawing the polygons' boundaries in comparison to a well-defined 3.125 km gridded SIC product as provided from AMSR2 using ASI?

L488/489: "...where they have been observed" --> exactly. So what is with, e.g., the next day, when there is no high-resolution information available but only a AMSR2 6 GHz 50 km footprint-based SST estimate because there are clouds? Such a day-to-day hetereogeneity is not helpful and combining different spatial scales of information requires particular care when it comes to assess uncertainties. I am pretty sure that the ASIP and to some degree also the MASIE product stitch different scale-observations together and the credibility of the data product can change quickly from one pixel to the next and from one day to the next.

L491: See my earlier comment about the work Worby and Comiso did. It is the Antarctic and it is ASPeCt. You must not use it to refer to ASSIST.

L491/492: "recall that ... at that time" --> While this is true, the ship is moving during the 10-minutes observation time, hence elongating the observed area towards an elliptically shaped region centered along the ship's track. In addition, if I am not mistaken, you did not compare single ASSIST observations but looked into daily averages?!?

L492-502: All true and possibly also discussed to some extent in Kern et al. (2019), right?

L503-511: What I would strongly recommend is to suggest further evaluation of ASIP with independent observations of the sea ice conditions from Sentinel satellites (Sentinel-1 SAR and Sentinel-2 MSI). These provide a spatial representation of the conditions at the ice edge / in the MIZ and potentially would be a more solid basis for any further evaluation.

Editoral Comments / Typos:

L60: "Steffan" needs to be "Steffen"

L64: "imagery" --> "imagers"

Figure 1: I suggest to increase the size of the panels a bit to enhance readability. Alternatively, increasing the font size would help as well.

L130: I guess this was August 21 and not August 12? Where was the image taken? Could you indicate that in one of the maps?

L221: As far as I know the two Kern et al. papers are dealing with both the Antarctic and the Arctic - especially the one from 2020.

Table 3: The way to specify inclusivity in values ranges would be [15% to 80%] or [0-40%[ or ]80 to 100%]

L345: "double triple" --> typo

Figure 6:

Fonts at the legend should be larger.

I suggest to change "% of time" to "count"

I also suggest to write "sea ice concentration" instead of just "ice" when denoting the axes.

Figure 8:

Please increase the font sizes.

Figure 10: Please provide the unit of the distances.

General comments

GC1: The authors compare how well the NOAA Alaska Sea Ice Program (ASIP) daily ice charts, along with another operational ice extent map and a passive microwave sea ice concentration product, match up with sea ice concentration estimates from visual shipboard observations.  I have some concern that a casual reader of the paper will see a statement like “ASIP’s overall accuracy rate of 95.7%...” and use it without reference to the limitations of the validation method that the authors are aware of.   The authors will improve the manuscript by tightening and clarifying the presentation, especially when describing the ASIP and MASIE products and how they are “parsed”.

GC2: Sea ice charts are often the best information available to researchers as well as to those operating in polar waters. Yet, charts are underused by the research community, because researchers are often unfamiliar with them and have no way to evaluate their quality. Research papers that attempt to quantify the accuracy of operational products are few. That makes this one especially valuable, if the presentation is improved.

Specific comments

SC1: In the abstract, the authors write “…we show that the similarity in performance among products is due to the in-situ asset distribution, as most in-situ observations are far from the ice edge in 20 locations where all products agree.”   This statement would seem to discount their results.  It illustrates why I think the manuscript needs at least a paragraph in section 2.1 describing how analysts make the charts, and a section with at least a few sentences describing how “ice edge” is defined and drawn, if it its drawn, in or using the ASIP, AMSR2, and MASIE products for the purposes of this study.

SC2: This discussion of how “ice edge” is defined, drawn, and used in the three products should come ahead of the Results section.  Section 5 Discussion has some of this, but having the information in a stand-alone section and moving it forward will help readers understand how the differences in where products put an ice edge may arise. As it is, the authors begin using “ice edge” without an explanation.  I think of the ice edge as a contour line. Do the authors create a contour line in 0.05° gridded versions of AMSR2 SIC fields, MASIE 1 km binary ice/not ice fields, and ASIP polygons containing ice concentration ranges?

SC3: Each product sets a different-sized area within which SIC is estimated. The AMSR2 grid cell size may be 3.125 km but the SIC algorithm integrates brightness temperature information from frequency channels that have different footprint sizes and shapes.  The ASIP analyst looks at satellite imagery and draws a polygon around ice that looks homogenous or ice floes that are fairly evenly distributed, and labels it with an ice concentration range. Each polygon is different. The USNIC analyst that draws the IMS product used by MASIE estimates which 1-km grid cells cover areas with more than about 40% ice and labels them “ice”, using a variety of satellite and other data sources. Finally, the ASSIST observation is for an area within 1 nm of a ship, although visibility may limit this, as the authors note. Describing all this in one place will help the reader have a fuller picture of how differences in ice edge position arise.

SC4: I don’t think it would be particularly useful even if it were possible to come up with a rigorous accuracy estimate for these products.  I think it’s more important to understand how they are made and the strengths and limitations of each.  The authors note that the ASIP product puts the ice edge further south than MASIE or AMSR2, and is more accurate when it does so, judging by shipboard obs.   If you are a researcher that needs to know how likely it is that ice at any concentration will be present at some location off the coast of Alaska, then the ASIP product is your best choice.    A tighter, more carefully written Discussion section up front will help more researchers understand that choice.

SC5: The Discussion section also needs something on why MASIE and the hi-res AMSR2 from Bremen were chosen. Note that MASIE is not itself an operational product but is a reformatting of the USNIC IMS operational product. I assume MASIE was chosen because it is easier to work with than IMS and offers a unique daily high-resolution map of ice extent. 

SC6: The USNIC MIZ product (U.S. National Ice Center, 2020) is another daily product that shows a 10% and 80% SIC contour. The authors could consider working with it as an alternative or in addition to MASIE.

U.S. National Ice Center (2020). U.S. National Ice Center Daily Marginal Ice Zone Products, Version 1 [Data Set]. Boulder, Colorado USA. National Snow and Ice Data Center. https://doi.org/10.7265/ggcq-1m67.

SC7: It would be helpful to mention that USNIC charts also cover the region covered by the ASIP charts, and have some words about how they compare, as RC1 noted.

SC8: After years of working with ice chart products along with satellite data, I strongly agree with the authors concluding statements about the value of ASIP products for scientific studies.

More specific comments follow.

L9: “….we present a new SIC product…”  Please clarify exactly what the new product is and how it differs from the ice charts that are available on https://www.weather.gov/afc/ice. The text isn’t clear on this.

L11: Does EGU prefer “in-situ” to “in situ”?

L16:   Consider rewriting as “ … and (iii) a product available from the National Snow and Ice Data Center (MASIE) that originates with the US National Ice Center (USNIC) operational IMS product.” 

NSIDC archives both products, and both should be cited correctly, and listed in the References section.  Here are the MASIE and IMS citations in APA style:

U.S. National Ice Center, Fetterer, F., Savoie, M., Helfrich, S. & Clemente-Colón, P. (2010). Multisensor Analyzed Sea Ice Extent - Northern Hemisphere (MASIE-NH), Version 1 [Data Set]. Boulder, Colorado USA. National Snow and Ice Data Center. https://doi.org/10.7265/N5GT5K3K.

U.S. National Ice Center (2008). IMS Daily Northern Hemisphere Snow and Ice Analysis at 1 km, 4 km, and 24 km Resolutions, Version 1 [Data Set]. Boulder, Colorado USA. National Snow and Ice Data Center. https://doi.org/10.7265/N52R3PMC.

It’s important that readers understand that NSIDC is not an operational center, and MASIE is not an operational product, in contrast to USNIC and IMS.  It would be helpful to say what is meant by the term “operational” as used in this paper.  

L31: “MASIE has by definition no information at SIC < 40%.”  That’s not entirely true. One could regrid MASIE to some larger grid-cell size, and end up with larger grid cells with less than 40% SIC.  

L35-131 The Introduction section could be shortened and tightened up a lot. Omit needless words.

L58: Lohanick is misspelled.

L59: It would help users understand better if written “This leads to an underestimation of sea ice concentration, which in turn results in an underestimation of sea ice extent…”

L63-67: While the first method describes using a processing algorithm on satellite data, the second method describes how a human might draw a chart. Different word choices might better get across the manual nature of drawing operational charts, e.g “an analyst manually synthesizes the information in satellite imagery …” ; ”Operational maps as drawn…”

L77:  Consider citing the CIS data so that others can easily find it:

Canadian Ice Service (2009). Canadian Ice Service Arctic Regional Sea Ice Charts in SIGRID-3 Format, Version 1 [Data Set]. Boulder, Colorado USA. National Snow and Ice Data Center. https://doi.org/10.7265/N51V5BW9.

L80-81: here, please make the clarifications and add the citations that I noted with respect to the abstract. Also, I hear that USNIC prefers USNIC to NIC these days.

L87-88: This is the first mention of color code, egg code, and WMO standards. A few things are incorrect. “Egg code” is not a WMO standard, rather, it is a shorthand way, taken from the egg shape of the labeling symbol, that analysts use to refer to how a polygon in an ice chart is labeled. The ice information inside the egg symbol would be in SIGRID, which is a WMO format. While SIGRID is used by ASIP, USNIC, and other ice services to describe the ice within each polygon, egg codes are not used much anymore. (Danish Meteorological Institute charts are an exception. See https://www.bsis-ice.de/IcePortal/)

I would avoid using “egg code” entirely.   Instead, just briefly mention that you are following ASIP and international ice chart convention in using the WMO color code and descriptors for characterizing ice concentration ranges in your presentation of AMSR2 and MASIE sea ice concentration as well as for ASIP.    You can cite WMO Sea-Ice Nomenclature (WMO, 2014):

World Meteorological Organization (WMO). 2014. WMO Sea-Ice Nomenclature. Volume 1 -Terminology and Codes, Volume II - Illustrated Glossary, Volume III - International System of Sea-Ice Symbols. Fifth Session of Joint Commission on Marine Meteorology (JCOMM) Expert Team on Sea Ice. WMO Publication No. 259.

and perhaps Manual of Standard Procedures for Observing and Reporting Ice Conditions (MANICE) (Env. Canada, 2005)

Environment Canada. 2005. Manual of Standard Procedures for Observing and Reporting Ice Conditions (MANICE). Issuing authority: Assistant Deputy Minister, Meteorological Service of Canada.

L89: Include acronym here (SASSIE) if correct to do so.

L120: Remove “operational” here.

L123-131 and Figure 1:  The images here need to be MUCH bigger. Delete the photo (e) if necessary in order to enlarge the rest of the figure.  Also please clean up the text about WMO and eggs, and provide the information on how AMSR2 data are binned somewhere in the main text, not the caption.

L143-144: Suggest you include just one WMO reference, this one, at the end of the second sentence:

World Meteorological Organization (WMO). 2010. SIGRID-3 : A Vector Archive Format for Sea Ice Charts. Intergovernmental Oceanographic Commission. First edition: 2004. JCOMM Technical Report No. 23, WMO/TD-No. 1214: Https://library.wmo.int/index.php

L148: Suggest changing “but is analyzed from imagery over the preceding 24 hours.” To “based on imagery acquired over the preceding 24 hours.”

L154-180:  Please consider my general comments when editing this section, and describe how a person draws polygons. The word “implement” is misleading. Also, it’s not always clear what is meant when the word “parse” is used.  Consider choosing other words to describe the process of gridding the ASIP polygon SIC information onto a grid in some projection.

I suggest you reference the following data set somewhere in Section 2.1.  As with the ASIP product, when we made it, we needed to convert SIC information in shapefile polygons to gridded fields of SIC.  The User Guide for the product describes the process we used.  There are so few products of this kind that it would be helpful for readers to know about this one as well:

U.S. National Ice Center. 2020. U.S. National Ice Center Arctic and Antarctic Sea Ice Concentration and Climatologies in Gridded Format, Version 1.  Boulder, Colorado USA. NSIDC: National Snow and Ice Data Center. https://doi.org/10.7265/46cc-3952.

Figure 2 (c) and (f): Suggest using a step color bar.

L255: Remove “operational”.

L260: Clarify what “it” refers to.

L267-276: Please edit, taking my general comments and other specific suggestions into account. 

L278: “We compare satellite SIC to in-situ observations….” What does “satellite SIC” refer to? Is it just the AMSR2 25km and 3.125 km products?  Please specify.  MASIE and the ASIP products should not be referred to simply as “satellite SIC”.  Perhaps “gridded SIC fields” is a better choice?

L279: Further confusion about satellite products here. Line 280 refers to the “nearest satellite pass”, but earlier, AMSR data were described as a daily field and not as swath data. Is the time of the nearest satellite pass known?  “Satellite grid cell”: does this mean the AMSR2 SIC grid cell?

L285: Here, “three satellite products” implies that “satellite products” is being used to refer to the MASIE and ASIP fields and not just the AMSR2 data.  Please choose words other than “satellite products” and clarify the reference to  “nearest satellite pass”.

L292: Section 3.1 is titled “Satellite products compared with in-situ observations”.  Change “satellite products”.

L295: Suggest addition to read “…defining the ice edge in passive microwave products…”

L340 and section 3.1.2: It’s not surprising that AMSR2 under-predicts the presence of sea ice, and it’s helpful to see it demonstrated by the comparison with shipboard obs and with the ASIP operational charts here.

Figure 7:  I agree with RC1 that these figures can be much smaller and font bigger.

Figure 8: Suggest including only one (bigger) legend and making the MASIE edge a different color.

L434: Does the word “difference” belong after “distance”?

Figure 10: the Y axis needs units.

L468: The phrase “lowest common denominator” doesn’t work well here.

L479-486: This is important information that should come earlier.

L 485: Suggest addition to read “…ASIP product is a vector shapefile that is not provided…”

L494:  “grid cell” would be better than “pixel”

L496: Suggest instead of “the broader pixel” using “are reflected in the SIC grid cell”

L520:  IMPORTANT- Should read “…grateful to all the analysts at ASIP and at USNIC”.

L558: This should read:

U.S. National Ice Center, Fetterer, F., Savoie, M., Helfrich, S. & Clemente-Colón, P. (2010). Multisensor Analyzed Sea Ice Extent - Northern Hemisphere (MASIE-NH), Version 1 [Data Set]. Boulder, Colorado USA. National Snow and Ice Data Center. https://doi.org/10.7265/N5GT5K3K.

L585: Please include DOI or link for this report.

L598: Please include DOI or link for this report.

L602: Please include DOI or link for this report.