Monitoring the Ice Phenology of Qinghai Lake from 1980 to 2018 Using Multisource Remote Sensing Data and Google Earth Engine

Round 1

Reviewer 1 Report

This paper presents recent trends in lake ice phenology for Qinghai Lake from 1980-2018 using MODIS and AVHRR surface reflectance products. The paper shows expected trends, with freeze-up dates shifting later, break-up dates shifting earlier, and shortening ice cover duration. Analysis of local climatic factors is interesting with strong connections demonstrated between LIP dates and temperature/radiation. Overall, the paper contributes nicely to the current lake ice remote sensing literature, particularly with the recent interest in the Qinghai-Tibetan Plateau.

 While the paper does provide a nice analysis, there are some elements that need to be addressed. The methodology of selecting the threshold values for the red and NIR bands is unclear, and the use of Landsat images for their selection needs to be better explained. Additionally, a table defining the LIP events evaluated is needed and a figure showing the final classification using AVHRR data. There are certain parts of the introduction that need to be refined, particularly sections that simply list papers without making connections.

The results need modification. The method of extracting trends in the LIP is currently unclear and no significance values are given. Also, the air temperature index is the paper should be replaced with the more commonly used freezing degree-days as this would help clarify the results. Overall, the discussion of the result sections in the paper could be improved by connecting the findings to global patterns in recent literature. Corrections to the sentence structure and grammar are also needed. Specific line by line comments are presented below.

Line 20 – “the” is not needed in front of lake ice phenology, “technology to study lake ice phenology”.

Line 21 – “the” is not needed in front of freeze-up start.

Line 23 – “advance trend” is not the correct wording, “an advancing trend” or “break-up end shifted earlier” would be a better phrase.

Line 24 – The word “advance” should also be replaced.

Line 25 – “completely freezing duration”, should be complete freeze duration?

Line 25 – “the” is not needed in front of change rates.

Line 35-36 – This is a very broad/generalized statement, it should be modified. Alternatively, it could be removed as the importance of lake ice as a climate variable is discussed in the next statement.

Line 38-40 – The terminology in this sentence is incorrect, lake ice cover is defined as an essential climate variable (ECV) under the Global Climate Observation System (GCOS). This is important information that should be included.

Line 41 – “lake ice changes” should be changed to “lake ice change”.

Line 54-57 – This statement needs to be revised. The text refers to Laurent Lake, this is incorrect, it is the Laurentian Great Lakes.

Line 62-63 – The phrase “domestic and foreign research” should be changed, “research” itself is sufficient.

Line 65 – It appears this section is listing commonly used sensors? SAR (synthetic aperture radar) is not a sensor like MODIS, AVHRR, SMMR, AMSR-E, or SSM/R, but a type of radar data. Satellites that could be used in place include Sentinel-1, RADARSAT-2, and ASAR. There is a comma missing between AMSR-E and SMMR, these are two different sensors.

Line 68 – No citation is given for a paper that used an empirical formula method as was done with the threshold and exponential method.

Line 70 – Citation of an example(s) is needed.

Line 71 – Howell and Brown should be removed [23] alone is acceptable.

Line 71-Line 81 – This should be revised, currently it is just a list of short literature summaries. It could be better synthesized in a discussion about the various methodologies or at least shortened to focus on those that use optical remote sensing.

Line 76 – “microwave data” should be “passive microwave data”.

Line 97-100 – Similar to Line 71-81 this section lists several papers, it could be better synthesized as a discussion on recent advances in lake ice phenology research for the Qinghai-Tibetan Plateau.

Line 100 – The results from which paper? Related to the previous comment this may be a good statement to base the discussion around.

Line 105 – “more than 200 species of fauna”? Also, change surrounding lakes to surrounding area.

Figure 1 – What imagery is this? Landsat? Make sure the data is properly referenced in this figure and all others.

Table 1 – This table could be improved by including the bands and their associated wavelengths.

Line 155 – What method was used? This needs to be fully explained or at least the method needs to be properly referenced so that it could be replicated.

Line 160 – The equation for NDWI should be stated.

Line 180 – While good to provide the GEE source location, the original algorithm theoretical basis document or published article should be cited as well.

Line 183 – What are the names/wavelengths of the B1 and B2 bands?

Line 192 – “Typical methods of lake ice extract”, should be extraction?

Equation 1 – Can this be replaced with an equation embedded in-text? It is currently unclear.

Line 203-205 – The methodology for determining the thresholds is unclear. What does human-computer interaction tests mean? Was there a form of validation performed where multiple thresholds were tested? This needs to be clarified.

This procedure is a little confusing as it appears that the Landsat imagery was used to develop these thresholds, which were then applied to different satellites. Were the products compared to ensure that reflectance values were similar between corrected products? Could you clarify some of the details about this portion of the methodology?

Line 213 – “was stably less” could be changed to “was stable at less”.

Line 216 – It would be helpful to the reader if somewhere in the text (this section or the introduction) a table was included that listed the different ice phenology timings and an associated definition.

Line 220 – “domestic and overseas”, this should just state existing research.

Line 223 – This is the first usage of the term AD, it needs to be defined.

Section 3.3 – This section discusses the extraction of LIP events from AVHRR images; however, no image representation of the results is shown. It would be good if the authors included the presentation of Qinghai Lake in a figure and demonstrated the appearance of ice and water in the classified image.

Line 227 – Abscissa typically refers to the x-coordinate, while ordinate refers to the y-coordinate (which is the day of the year in Figure 3. This statement would fit better in the figure caption for Figure 3.

Line 229 – The text lists Figure 5, it should be Figure 3.

Line 252 – Here the authors have mentioned the interference of clouds for MODIS data but did not highlight this same issue with AVHRR, would the same issue not be present as they are both optical data? How was the issue of clouds addressed for AVHRR?

Section 4.3 – What method was used to extract trends for ice phenology? Was a simple linear regression employed? Because the data deals with a time series typically something such as the Zhang trend method (Wang and Swail, 2001 and Zhang et al., 2000) would be used. This is a common method of analyzing trends in physical properties such as lake ice (see Murfitt and Brown, 2017). Also, no reference is made to the significance of the trends. Are the changes in LIP observed over the 28 years significant?

Additionally, this section would be improved through some connection to international trends. Do the results for Qinghai Lake match with expected northern hemisphere trends in lake ice phenology? This would strengthen the discussion.

Section 4.4 – This section seems out of place in the paper, it does not add much to the analysis. Is the section proposing a model? Or is it just highlighting the pattern in the spatial development of ice cover? This part is unclear. This section could be shortened and added to the initial introduction/methodology of the paper to provide context on the ice cover of Qinghai Lake.

Line 385 – Why was this method selected as opposed to freezing degree-days? This is a more common metric for lake ice and would show similar patterns. The analysis of NAT on line 394 is confusing because NAT is negative whereas FDD is positive. If you used FDD, a smaller AFDD would show delayed FUS, earlier BUS, and shorter ICD.

Section 4.5 – This section is good; the analysis of the different climate variables is interesting. The radiation variable needs to be clarified in terms of the energy budget, is it shortwave, longwave, total? It would also be good to place these results in the context of other studies that have evaluated ice phenology dates in connection with similar variables to see if similar patterns are observed.

 

 

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper discuss a very interesting research on long-term (38 years) monitoring of ice phenology of Lake Qinghai from 1980 to 2018, primarily using NOAA AVHRR data. From the long time-series of data, the trends in lake freezing and thawing and their correlations with various climate variables are explored. MODIS data is also used to verify the accuracy of the results generated using a threshold-based method. The data is processed in Google Earth Engine platform. Overall the paper is well written. Experimentation part looks sufficient, however there are some issues which need to be rectified.

Major issues:

• GCOS recommendation for ice-on (FUS,FUE) and ice-off (BUS,BUE) dates is +/-2 days. The authors doesn't validate the absolute accuracy of these 4 dates against any reliable ground truth. Validating 38 years of data against ground truth is impractical, but for 1 or 2 years is doable. This will be helpful to quantitatively check how accurate the threshold-based method is. Instead, the authors perform a relative validation by comparing AVHRR results with MODIS results (Table 2). Still, the MAE is >2 days (violating GCOS requirements) for FUS (2.66 days), FUE (2.38 days) and BUS (2.05 days). In nutshell, the relative accuracy itself is >2 days while the absolute accuracy should be <2 days. Hence, it is not possible to agree to the statement in line 273 :"In summary, we believe that the LIP extracted based on AVHRR and MODIS data has high consistency"
  
  
• Absolute geolocation accuracy: It is specified in the paper that the lake boundary is extracted using the relatively high spatial resolution (30m) Landsat image. However, there can be absolute error in geolocating the lake. What about absolute geolocation shift from Landsat to AVHRR data? Are these errors corrected by Google Earth Engine (like a pre-processing step)? if not, this can be a source of error in the final results. Though the authors use the word 'corrected' on lines 174 and 182, it is not very clear. Probably they are not talking about the absolute geolocation correction. Given the size of the lake, this error is not super critical, however cannot be neglected.
  
  
• One main requirement for monitoring lake ice is high temporal resolution (ideally 1 day), but at least every 2 days according to GCOS, to precisely identify the phenological events like the FUS, FUE, BUS and BUE. The authors did not specify anything about the temporal resolution of AVHRR. Additionally, how much AVHRR data is lost due to the presence of clouds? This is critical also for MODIS. Approximately how much data is lost due to clouds during the critical freezing and thawing periods? This information is missing and need to be specified as this will affect the accuracy of locating the phenological events.
  
  
• In section 3.2, more information on how the threshold values were fixed need to be provided since this is the core of the whole paper.
  
  
• Lines 262-267: The explanation on why FUE date appeared earlier, is not convincing.

 

Minor issues:

• Explanation on how relative error is calculated is missing (section 4.2)
• Description about related works need to be improved. Latest research in lake ice monitoring is not well addressed in the paper (at least the ones that used MODIS and AVHRR data need to be discussed in more detail). The paper discusses a relatively outdated threshold-based method whil the state-of-the-art is now based on machine and deep learning based methods. At least, those works must be cited.

Very Minor issues:

• Equation 1 appears very blurry (line 198)
• Figure 1 title is not self explanatory. More details such as: when was the satellite image captured? which satellite? which band or a combination of bands is shown in the figure?
• Two reference sections appear in the paper (one is blank). 
• References 9 and 25 cite the same article
• References 43 and 44 uses 'pp' for page number while reference 2 uses 'p', others nothing. Use of page numbers must be consistent
• In consistency in indexing figures: eg. indexed as 'Figure5' in line 229 while 'Figure 1' in line 119

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have done a good job of responding to all the comments provided. Significant improvements have been made to the introduction and the discussion of the process used to select threshold values. However, there are still a couple of minor points that I believe need to be addressed and have listed them below. In regard to referencing, the last name of authors should not be included, only the number for the reference should be listed.

Line 25-26: “significant decreasing trends with shortening by…”, this would be better as “decreasing trends of…”. Also “which was shortening”, would be better as “and overall decreased by … between 1980 and 2018”.

Line 27: “freeze-thaw space mode” should be changed to “The spatial pattern in free-thaw of Qinghai Lake can be divided into two areas”. Similarly, Line 29: “freezing and ablation patterns are opposite”.

Line 55-56: This should be written more concisely. It can be changed to state “the warmer water surface and the lakeshore results in the formation of lake-effect snow systems that can result in significant snowfall events”.

Line 59: Apologies if I did not notice this before – 1970s or 1870s?

Line 60: Inversion should be replaced with extraction.

Line 62-63: All datasets listed are from satellite remote sensing platforms, it would be better to list by type of data. Active Microwave (Sentinel-1, RADARSAT-2, and ASAR), Passive Microwave (careful not multispectral!) (AMSR-E, SMMR, and SSM/R), and multispectral (MODIS and AVHRR) datasets.

Line 83: ‘The’ is not needed at the start of the sentence.

Line 87: ‘the predicted’, the is not needed.

Line 136: In regards to referencing images in figures, because you are using a Landsat composite you should follow these guidelines: https://www.usgs.gov/centers/eros/data-citation?qt-science_support_page_related_con=0#qt-science_support_page_related_con. This is a very minor correction.

Line 179: Apologies I should have been more specific. The ATBD that should be referred to is the MOD09GQ document. You should be able to find the full citation information at https://lpdaac.usgs.gov/products/mod09gqv006/ . Also, ensure that you include the DOI for the data.

Line 203: This section is significantly improved. Minor points for clarification – to get the maximum coincidence, this was done by visual analysis? The expression used in your reviewer response (not shown in the manuscript) has “b1 gt 0.05”, but should be “b1 lt 0.05”. This may be just a small error when taking a screenshot of the band math tool as the method appears to be working correctly in the sample images. This is also stated on line 202, did you use this method or the one listed in the equation? This should be double-checked.

Line 214: Thank you for your clarification regarding AVHRR images. The inclusion of a brief statement discussing why no visualizations of AVHRR images were included would be good. Particularly because these are the main images being used to extract the LIP dates.

Section 4.4: Thank you for your clarification regarding this section. The details you provided were very interesting. This section should be retitled: “Spatial Patterns in the Freeze-Thaw Cycle of Qinghai Lake”, I think this would clarify for readers that you are not referring to an actual model, instead of providing observations in the spatial pattern of the ice. Line 356 should be similarly changed to state: “Patterns in lake freeze-ablation reflect differences..”. Line 407 should also be modified, “freeze-thaw patterns and processes of the lake”.

Section 4.5: Line 435, should it be smaller AFDD? Smaller values of AFDD indicate warmer temperatures which would lead to the changes in LIP described.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf