Review of
A combined Terra and Aqua MODIS land surface temperature and meteorological station data product for China from 2003-2017 - revision 1
by
Zhao, B., et al.
I reviewed the first version of this manuscript. Therefore I am referring to my first review regarding the summary of this manuscript.
First of all, I thank the authors for taking some of my concerns serious and for revising the manuscript with respect to some of the concerns brought up. I note that the authors did not answer all of my comments in the first review properly; some comments were not formulated clearly enough, it seems, so that these were misunderstood by the authors.
The topic of the paper is highly interesting and the data set behind has large potential. However, the description of data and methodology lack important and required details and needs to improved also with respect to the consistency of the wording and measures applied. The illustration of the re-construction method as a whole is very light and not satisfying. The application of the final bias-correction is neither well motivated nor does it appear to me that the final bias-corrected re-constructed LST data set is adequately evaluated because the 20% ground-observation data set aside are already used for the bias-correction. Finally, the results section lacks any example of the data and only focusses on trend analysis and interpretation of average conditions; hence it has to be restructured.
To my opinion this manuscript is not yet ready to go and requires major revisions.
General comments:
GC1: Still the paper is difficult to read and understand because the definition and naming of grid cells which are in one way or the other contaminated by clouds (or aerosols) is not consistent throughout the paper. It would be very helpful if the authors could put at one central place and for the entire paper a definition that they reconstruct the LST of grid cells they defined as missing. These grid cells might be defined as missing because a) for these simply no LST value exists [possibly due to thick clouds mostly but potentially also due to other factors which are - however - not relevant for this paper] and b) the QA analysis provided an average LST error > 1 K or an average emissivity error > 0.01 - also potentially due to clouds [here perhaps only with a partial coverage of the grid cell but again this does not matter]. What matters is that there is still a suite (or zoo) of different names for the grid cells (or pixels) for which the LST is re-constructed. This needs to be harmonized!
GC2: While the authors agreed that it would be important to illustrate and better describe steps of the LST reconstruction method in their reply to the reviewers' comments, they did not include any such information into the revised manuscript. Still the focus is on analysing the data set, showing trends and interpreting the data set. This is material for a geophysical paper. Earth System Science Data is a data set journal. It is neither like Journal of Geophysical Research nor like similar journals where geophysical interpretation is in the foreground of the interest.
Therefore, I am still missing examples of how each of the three LST reconstruction steps 1) to 3) changes a LST time series for a specific month at a specific location. The reconstruction method is not illustrated adequately enough. It is not illustrated how a single LST value in-filled from a meteorological station might impact the GWR method in case that there are only 5 or 6 daily LST values in the 5 x 5 grid cell box used. It is not illustrated how the elevation-LST method fills in missing data. No investigation is presented which illustrates a potential different skill of the reconstruction methods during different times of the year. And, most severily, I don't find any high-resolution LST map which would illustrate how the three different steps of the re-construction process successively fill a LST map with a large count of missing data. In other words: The merit of the reconstruction method over the existing level 3 MOD11C3 and MYD11C3 products is demonstrated only to some extent. None of the trend analyses shown have been derived with the original product as well and compared with the results of the new product.
Another missing element of the methodology which radiates well into the entire manuscript is the brevity with which the about 2400 meteorological stations are described. These data are key to this paper - for re-construction, for the bias correction and for the evaluation. But the readers learn amazingly little detail about these stations, about what they regularly measure, how quality assessment is assured, and which surface conditions these stations represent. I had one major comment about this in my previous review and it seems that the authors accidentally ignored this missing piece of information. The description of the representativity of the station with respect to the MODIS grid cells (5600 meters x 5600 meters) is not included; neither did I find information about the biome and/or the topography for the more complex observation sites. I have seen too many papers showing wonderful results for the wrong reason. I cannot expect a comprehensive discussion of this issue in your manuscript. But I can expect that you spend more effort on this issue, because it is directly linked with the credibility of your data set, than you spend on interpretation and analysis of the data set - and I am well aware of that I am repeating myself here - which is material for a different publication.
GC3:
In Earth System Science Data data sets are introduced to potential users. I am missing a section which informs the reader about the data format and about the variables that are included in the product.
GC4:
Again - despite my comment in the first review of the paper - you right-away jump into a discussion of annual-mean LST trends derived for China as a whole. At this stage the reader has still no clue how the reconstructed data set looks in comparison to the original data set. What, to my opinion, needs to be shown are maps of at least the monthly LST of the original product and of the new product. These should be supported by maps of the LST differences. Here you might want to zoom into critical areas (i.e. areas where QA of the original product indicates low-quality LST values).
All these trend analyses do not contribute to the evaluation of the product. For that they would have needed to be compared to other, independent products.
In general, I don't understand why the flow of the results and discussion sections isn't like this:
1) Presentation of the product
2) Evaluation of the product
3) A few first examples (!) of the application of the product for trend analysis [for this journal possibly 2 pages maximum]
4) Discussion of potential caveats and limitations of use [this is for the user]
GC5: You intend to use the independent 20% of the in-situ observations of the LST to evaluate the product. However, instead of discussing observed differences (which are displayed and illustrated in a very global way without going into detail) based on physical principles you rather take the observed differences to perform a bias correction. One could provocatively say that your reconstruction methods hence does not work properly. In addition, now that you have used these independent LST observations for the bias correction - what are your independent observations to evaluate the final, bias-corrected LST product?
Specific and editoral comments and remarks:
Abstract:
Lines 18/19: Please check sentence; it does not read properly.
Lines 19/21: Neither "real" nor "true" appear to be the correct expressions here. About "actual" instead of "true" and simply delete "real"?
Lines 23-25: Are these statistical measures valid / computed from the region mean LST values? Or are you referring to the statistical measured obtained for separate single station-to-LST-data-product-comparison within the regions? Please change the text accordingly so that this becomes clear.
In addition, I recommend to explain what RMSE, MAE and Rsquared is.
Line 28: Please change "slope >0.10" such that it is clear that you write about an LST change of x Kelvin over time period y.
Line 30: "exhibited" --> "exhibits"
Introduction:
Line 36: "surface temperature" --> "LST"
Line 38: You deal with LST in this paper while these "temperature changes" you are referring here to (with respect to glacier melt) are air temperatures, right? please connect these better to the LST (which should be 0degC over melting surfaces such as glaciers and snow).
Line 41: "LST data ... measured" --> Either "LST is measured" or "LST data are obtained"
"by" --> "at"
Line 42: "ground surveys" --> What is this? Are you speaking about specifically targeted campaigns or expeditions? Where can one find data of such "ground surveys" if these are not carried out in the context of regular measurements at the meteorological stations?
Line 54: "throughout the world" --> "and its global coverage"
I note that you could and should have mentioned NOAA-AVHRR data as well here - particularly since with AVHRR data one can build a substantially longer time series than with MODIS and since data of that sensor are successfully used for sea surface temperature retrieval back to 1981.
Line 58: "low quality values from undetected cloud-low-quality pixels" --> consider re-writing this part please; it is confusing.
By the way, if you intend to use this expression "cloud-low-quality pixels" you might want to define once what you mean by it.
Lines 60-62: "Although the integrity of the data has been greatly improved" --> What do you mean with this? Further: "synthetic data" ? Further: "thus contain an insufficient quantity of daily LST pixels" Again, what do you mean with this? Perhaps you wanted to write something along these lines: "The number of missing or low-quality LST retrievals particularly high in the daily LST products. However, even in the 8-daily and the monthly LST products are subject to a considerable number of data gaps - particularly in regions and at time periods with persistent cloud coverage." ???
Lines 62/63: "Cloud cover and other factors" appear to be several; therefore "cause", "reduce" and "pose".
Line 64: One "low-quality" has to be deleted.
Line 75: "only LST data" --> perhaps better: "only adjacent high-quality MODIS LST data"?
Lines 78/79: "it cannot obtain enough information for reconstruction" --> perhaps better: "data coverage is too sparse for a reliable reconstruction."?
Line 90 "improve" --> "improved"; "add" --> "added"
Lines 92/93: "... clear-sky conditions instead of cloudy conditions, which cannot fulfill the need to obtain the real situation at the land surface" --> perhaps better: "... clear-sky conditions. However, clouds reduce night-time surface cooling and day-time surface warming due to solar irradiance. These effects are not taken into account using this assumption and therefore the derived LST values are likely biased towards clear-sky conditions."
Line 94: "the real" can be deleted.
Line 95: "is capable of penetrating clouds" --> This is definitely not true for the higher frequencies (near-90 GHz) and also not completely true for even the lower frequencies. Please reformulate this statement therefore.
Study Area
You could, this is just a suggestion, provide some additional information about the average cloud cover in these six regions. With that you could underline right from the beginning where your approach potentially will have the largest added value.
Lines 114/115: "the key areas" --> "key areas"; "significantly" --> really? Or just substantially. How about stations outside these red circles? Do none of these exhibit significant / substantial changes in LST?
Data and Methods
Line 148: "near-polar orbit" --> "polar orbiting"
Line 149: "with a flying height of" --> "flying at an altitude of"
I suggest to delete "with a temporal resolution".
Lines 150-152: This is not entirely correct because the satellite switches orbit direction between the two overpass times given, i.e. descending and then ascending for Terra and ascending and then descending for Aqua. Hence you need to correct your notion of the direction. Please also include the names "ascending" and "descending". If you are in doubt I recommend to take a look here: https://www.ssec.wisc.edu/datacenter/aqua/ or https://www.ssec.wisc.edu/datacenter/terra/
Line 153: I don't get the notion of the repeat orbits every 1-2 days. What is the exact repeat cycle? Which areas of China receive complete daily coverage using one satellite (I guess you can specify a latitudinal boundary)? Which area require data of one satellite from two days for complete coverage?
Line 158/159: "another important algorithm:" can be deleted.
Line 162/158: You write of the latest LST V006 product and refer to "MOD07_L2". This puzzles me because when I check respective LP.DAAC web pages I only see that these LST products are based on MOD11_L2 / MYD11_L2 products; MOD07_L2 is a product containing air temperature and water vapor profiles without any notion of an accurate land surface temperature. It appears to me that mentioning of MOD07_L2 (you should add MYD07_L2?) is ok for the atmospheric data included in the improved retrieval but that the LST retrieval itself is based on different MODIS products which you should give here as well.
Line 172: Now I am even more puzzled: When I look at the LP.DAAC web page about the MOD11C1/MYD11C1 data set I find:
"The MOD11C1 Version 6 product provides daily Land Surface Temperature and Emissivity (LST&E) values in a 0.05 degree (5,600 meters at the equator) latitude/longitude Climate Modeling Grid (CMG). The MOD11C1 product is directly derived from the MOD11B1 product. "
and further
"The MOD11B1 Version 6 product provides daily per pixel Land Surface Temperature and Emissivity (LST&E) in a 1,200 by 1,200 kilometer (km) tile with a pixel size of 5,600 meters (m). Each MOD11B1 granule consists of the following layers for daytime and nighttime observations: LSTs, quality control assessments, observation times, view zenith angles, number of clear-sky observations, and emissivities from bands 20, 22, 23, 29, 31, and 32 (bands 31 and 32 are daytime only) along with the percentage of land in the tile. Unique to the MOD11B products are additional day and night LST layers generated from band 31 of the corresponding 1 km MOD11_L2 swath product aggregated to the 6 km grid."
Together with the information you gave in the paragraph ending in Line 165 I obtain the impression that these three paragraphs should be condensed and put into a more logical flow. Would it perhaps make sense to state upfront, i.e. in Line 158 that you use MOD11C1/MYD11C1 and MOD11C3/MYD11C3 and that this is the last generation of this V006 products which utilizes the day/night algorithm explained?
Lines 172-182: How relevant is information about that collection 6 is an improved version of collection 5 and that there is an 8-daily product as well? I recommend to substantially condense this paragraph and merge it with the two previous ones. Please check in this context also the last sentence "resampling from 5600 m spatial resolution to a resolution of 7200 columns ..." which appears to be strange given the fact that the C1 and C3 products have this number of columns and rows exactly at that grid resolution of 0.05 degrees (=5600 m).
Lines 183-188: What is the purpose of this paragraph? I don't think that you need to describe the steps required to generate the 11C1 and 11C3 data sets (which seems to be your aim here). I also have a problem with the last sentence about the conversion of "brightness temperatures". As said above the C1 product is based on the B1 product which already includes surface temperatures. Therefore, this whole paragraph could potentially deleted.
Line 190: "LST records at the hourly intervals" --> "Hourly LST observations"
Question 1: Do all these stations provide LST measurements over the entire period 2003-2017 at exactly the same location?
Furthermore: Do all these stations provide LST measurements or are some of these only measuring atmospheric parameters and as such measure the 2m air temperature but not necessarily the LST? Please provide an adequate sentence / statement into your manuscript.
Question 2: What is the surface type of the stations listed in Table 1? By surface type I mean the "biome" represented by the station. Is it urban, grassland, forest, barren, whatsoever? Please include this information into Table 1.
Question 3: You write hourly. Is that true? Do really all these stations provide these measurements with hourly temporal resolution? I am asking because typically such stations report either every 6 hours (0,6,12,18UTC) or, if it is climate stations, at three different times per day. I just want to be sure that you stress to the reader the apparently amazing consistency and temporal resolution of this data set of in-situ observations.
Line 196: "overpass times" --> "local overpass times";
Furthermore: How is the "extraction of meteorological station data" done? Did you interpolate between two hourly temperature readings? I note that the MOD11C1 files contain the local overpass time so that you have a very exact measure of it; did you use this information for the co-location in time with the in-situ observations?
Line 197: "warming/negative trends" --> "positive/negative trends" or "warming/cooling"
Line 198: Which LST data are meant here?
Lines 205/206: It appears some information is missing here?
Lines 208/209: Please check this sentence. It is not complete.
Line 209: Why do you want to reconstruct a cloudless [please write "clear-sky"] LST data set? I suggest to delete the first part of this sentence and write seomething along the lines: "It is difficult to fill data gaps caused by clouds in LST data products based on satellite infrared imagery with data of the same quality as the clear-sky LST observations."
Line 212: "that can reflect the true LST under cloud coverage." --> perhaps better: "that takes into account the actual LST under both clear-sky and cloudy conditions."
Line 213: Aren't the highly accurate pixels preserved in both, the daily AND the monthly data?
Line 217: "filtering(see ... details) . " --> "filtering (see ... details). "
Line 218: "low qualityFor" ???
Line 219: "at the corresponding time" --> perhaps better "for all days of the respective month" ?
Lines 221-224: wrongly placed spaces
Also: "substeps" --> "steps"
Furthermore : "1) in situ ... " --> "1) Where possible we filled invalid grid cells with co-located in situ observations of the LST."
And: "2) In case in situ observations are lacking, we employed the ... method to interpolate ..."
And: "3) The remaining invalid grid cells we filled with LST values reconstructed based on regression ..."
Lines 224/225: "Finally, the averages ..." --> I don't understand. I would have thought that after steps 1) to 3) all grid cells with invalid LST values in the daily data were filled so that one only needs to average over the full monthly record of daily LST values. If this is not the case then please say so in the manuscript, e.g.: "Even after these three steps, some grid cells still contain invalid values. These we filled by "... doing what? However, if after steps 1 to 3 all daily grid cells contain valid data, then : "Finally, we averaged over all daily data of the respective month and replace the invalid data in the original monthly LST product with the new monthly LST value based on the reconstructed LST time series of that month."
Line 231/234: "other atmospheric disturbances" --> a very vague expression. Would it make sense to simply only write "aerosols" and leave it with that?
Lines 232/233: "temperature values in the MODIS LST data" --> "LST"
Liens 235-237: Is this general statement required? I mean, the MODIS LST data contain a suite of quality flags which give information about the cloud cover as well as an uncertainty value for the LST between < 1 K and > 3K. So, I'd say the producers are well aware of that and put adequate measures into the existing data sets.
Line 238: "Statistical calculations were performed ..." --> What exactly did you do? Did you use the quality information (which and how) included in the MODIS LST product? Please be more specific and write explicitly in the manuscript which data you used for which step. Yes, I note that you write further down how you defined invalid grid cells in the monthly LST data but your description how you "created" Figure 3 is light.
Lines 243/244: "to reconstruct a high-precision daily LST images that represents real values under clouds using ... " --> "to reconstruct high-precision LST under clouds using ..."
Lines 250/251: "It is necessary ..." --> We know this already from the introduction and hence this sentence can be removed.
Line 255: "Finally, pixels ..." --> "Grid cells ..."
Line 257: "Finally, we reconstructed all the invalid pixels in ..." --> "Our aim is to reconstruct the LST for all these grid cells with invalid data."
Line 257/258: "Quality ... pixels (Benali et al., 2012)." --> What do you want to state here? I don't understand.
Lines 272-276: This paragraph can be deleted.
Lines 277-282: What is the intention of this paragraph? Are you making the statement that even if there are, say, 20 accurate daily LST data in a given month the data quality of the respective monthly LST value (if retrieved at all) is reduced? Do you have a reference for this statement? In other words: An interesting information you could pass on to the reader is: What is the minimum number of days with valid daily LST values (MOD11C1) which is required for a monthly LST value (MOD11C3)?
Line 283: "filter the monthly image" --> "filter each monthly image"
Line 284: "Then, we filter all the daily pixels from the month in which the cloud-low ... " --> Then, for each month, we filter all daily images of the respective month by determining all missing and low-quality grid cells."
You use the term "cloud-low-quality pixels" in the sense that this is the group of pixels which are either missing completely or which are defined low-quality using the above mentioned QA information. If this is the case then also Line 285: "low-quality daily data" and Line 286: "low-quality pixels" should have this term used; in both case you aim to replace both kind of missing pixels: those which are missing anyways because they have been completely flagged as useless because of clouds and those which are of low quality.
Line 286: "with the corresponding averaged pixel values from the daily data" --> "with the average LST derived from the gap-filled daily LST time series of the corresponding month."
Line 289: "real ground LST" --> "actual LST"
Line 290: "Many influencing ... reconstruction" could be deleted. Also, I would not group in-situ observations of the LST into "influencing factors". Therefore, please rewrite L291/292 as well.
Line 292/293: "Therefore, the latitude ... same location" --> I would simplify this sentence towards: "Grid cells with invalud LST values were co-located with meteorological stations". Please provide more information (this is one of the things which were not properly answered in your reply to the review comments): How is the co-location done? Were you co-locating distances in degrees or in kilometers? Were you computing distances to the grid cell centers? Was there more than one station in the grid cells? If there were more than one station in the grid cell, did you average over the values?
Line 293: "Invalidt" --> "Invalid"
Line 294: "ground-based" --> "in situ"
Lines 298-309: Is the degree of detail with which you describe this method really needed for your purposes? I note that for instance seasonal changes are possibly not important because you are using daily LST images here. I find the description overly complex and not to the point with respect to your specific problem.
Line 306: "temporally closest at the same overpass time" --> What does this mean? In the ideal case this is a pair of partly overlapping consecutive overpasses (~100 minutes time difference and the closest descending overpass of the same day (within 12 hours).
Also, you refer to "reference images" but then in Line 307 to "a adaptive threshold is then used in the reference image ..." --> So, it is first several reference images, then it is just one.
Line 311: seaons? See above.
Line 312: Check for missing " "
Line 313/314: "calculated from the standard deviation" --> of what?
Line 315: "closer" --> Are you referring to a distance here?
Line 325: "greater than 4" --> So of the 5 x 5 = 25 pixels minus the central one = 24 pixels only 5 valid pixels are requuired for application of this method. Since, as you said, this is not done in a sliding window approach this reduces the grid resolution of the LST product effectively to 5 x 5600 m = about 25 km ... because to take into account the information of a 25 km x 25 km area to replace a value at the center grid cell.
Line 327: "... the LST values of the pixels low-quality by clouds are determined through GWR." --> I suggest to use "reconstructed" or "filled" instead of "determined". In addition, I thought so far that you aim to replace any missing pixels and not just those which are potentially low-quality by clouds. This is confusing. I am still missing a clear, consistent and stringent definition and usage of which pixels you replace. It is tedious work to try to understand what you did.
Line 330: "The" -->"the"
Line 348: After all this discussion I still don't get whether grid cells for which the LST has been "reconstructed" using in situ observations (actually in this case a missing value is simply replaced by the respective value, so we cannot speak of a reconstruction) are getting a higher weight in the GWR than grid cells which contain valid high-quality clear-sky LST values. Please provide this information in the manuscript in a direct, non-hidden way. In case my assumption is wrong, then I don't understand the reasoning given in Lines 327-330.
Line 350: "the elevation factor" --> "elevation"
Line 351: "variation" --> "variation of the LST"
Equation 7: An interesting question would be whether alpha is in the range between moist and dry adiabatic lapse rate of the air temperature? Is this the case? If not, are there enough physically reasonable explanations about why this is not the case?
Line 371: This is the end of the methodology section. I have one major concern. See GC2.
#######################
Results
I have one general concern with the product. See GC3.
I have one general concern with respect to the structure of the results section. See GC4.
Lines 380/381: I don't think these first two sentences are an adequate introduction for a subsection about trend analysis of average Chine annual-mean LST data.
Lines 381: "To obtain ..." --> How did you do this? Did you first compute the monthly all-China average LST and then the mean over every year? Or did you first compute for each grid cell an annual mean LST and averaged the values over China as a whole? What kind of a trend analysis did you perform? A linear regression analysis? Any weights? Any significance tests? Please refer back to the respective subsections in Section 3.
Lines 388/389: What does abnormal rainfall have to do with LST? Please add the physical reasoning for this observation.
Lines 380-392: When you write "significantly" in this paragraph, is this statement based on a significance test result?
Line 408: "approximately 20.80%" --> "20.8 %"
Line 414: This is the second time you refer to the reported hiatus in global warming. Would you think it makes sense to leave these statements to the discussion / conclusions section? There it would be much more visible than somewhere inbetween the description of the results.
Lines 433-445: Please check this paragraph for superfluous blanks " ".
Line 479: ")f" --> ")"
Lines 502-511: See GC5
Line 511: "as been resolved" --> perhaps better "reduced"?
Line 520/521: I am sorry but Figure 3 and 4 are only snap-shots, showing TWO days and TWO months of the respective year. This does not support the global statement made in this sentence.
Line 537+: Please note whether you used the re-constructed bias-corrected LST data in Table 2. It is not clear which of your two LST products you compare here.
Line 543/544: Here you write that you use the original MODIS LST product directly without filter using the QA flags. I have two comments here:
1) Your product merges over data from MOD11C3 and MYD11C3. Hence you need to merge these two products for your comparison as well and/or perform the comparison with both these products.
2) Doing the intercomparison without taking the quality flags into account is ok - as long as you also present in addition an intercomparison where you applied the same QA filtering as you used for the re-construction to the original LST data. For this comparison, the improvement of your approach should be - if at all measurable close to zero. Why? Because with your re-construction you are supposed to only correct the LST for the missing and low QA grid cells. You could also split your evaluation and in addition to what I just asked for show a comparison of the original low QA data and the reconstructed high-precision LST. If I understood your approach correctly then this latter LST difference should be pointing to the achieved improved even more clearly. I am looking forward to this result of your quality check in the revised version.
Line 553/554/562/567: "The main reason for this difference may be the complex and diverse terrain" --> Don't you think that this is exactly one of these cases where a better illustration of your approach would allow you to be more to the point here? You have all the data at hand and instead of speculating what the reasons might be you could come up with a clear statement about... "as we see in figure xy in the complex terrain of region xyz the 19x19 sliding window used for the LST reconstruction based on the relationship between LST and elevation results in unrealistic LST values" ...?
Line 577:
I agree that a better spatial coverage is achieved with this re-constructed LST data set. While the overall accuracy has improved compared to the in situ LST observations it remains unspecific whether this improvement applies indeed only for those grid cells which have not been defined a lower quality because a separate uncertainty investigation has not been carried out, i.e. an answer to i) What is the difference between the original and the (bias-corrected) re-constructed LST data for QA of the original LST of < 1K? ii) What is the respective difference between the LST data sets for original LST QA > 1K?
Figure 2:
- Why does the first step include a "re-projection"? The MOD11/MYD11 data are on a regular 0.05 degree x 0.05 degree grid. Hence for which step of your processing do you require the re-projection?
Figure 3 / Figure 4:
- I am sorry, but didn't get your argument why in panel a) you kept blue for warmer LST values instead of reversing the color table and use blue for cold and yellowish / red for warm. This would be far more intuitive. I know, these figures is shown mainly to illustrate the extension of the cloud influence / number of data gaps but this does not prevent a reader from being puzzled by a color table that is reversed to what is usually used. Therefore please reverse the color table in panel a).
- Figure 3: How is "valid" defined? Did you use QA information as well? If not, why not?
- Figure 4: Reading the bit-wise encoded QA information is not straightfoward. Did you check how the number of invalid data changed when using a larger average LST error of, e.g. > 3 K?
Table 1:
Columns latitude and longitude miss the direction, i.e. degrees North and degrees East. |
- Printer-friendly version
- Supplement
