Estimation of Land Surface Albedo from MODIS and VIIRS Data: A Multi-Sensor Strategy Based on the Direct Estimation Algorithm and Statistical-Based Temporal Filter
Round 1
Reviewer 1 Report
I do not have any other comments or suggestions, thank you for updating the article.
Author Response
Thanks for your positive evaluation and valuable comments that helped us to improve the presentation of our work. We have also carefully checked the spelling and grammar mistakes throughout the manuscript with the aid of an English editor. Please see the words labeled in blue for more details.
Reviewer 2 Report
The aim of this study is to provide an algorithm for generating spatio-temporal continuous land surface albedo datasets with high temporal resolution (e.g., daily). The accuracy is properly verified, which is effective as a basic study for datasets maintenance. I agree with the practical importance of datasets maintenance, but do not confirm any new findings.
Author Response
Thanks for your positive evaluation and valuable comments that helped us to improve the presentation of our work. We have also carefully checked the spelling and grammar mistakes throughout the manuscript with the aid of an English editor. Please see the words labeled in blue for more details.
Reviewer 3 Report
The paper, Estimation of Land Surface Albedo from MODIS and VIIRS data: A multi-sensor strategy based on the direct estimation algorithm and statistical-based temporal filter, is for the most part well written. There are awkward phrases such as on line 100: "The DEA approach enables to estimate..." which might be better as "...enables us to estimate..." or "....enables estimation of..."
Similarly, line 109 "global distributed" might be better as "globally distributed".
Line 115 "descripted" should be "described"
Line 131 "are that" should be "is that"
Line 150 it is that data that are used....so "spectral library was" should be spectral library were"
The MCD43A3 data set isn't defined until page 7 but is used in 3 prior places. It would be better to define it where it is first used.
Line 247 should be "events"
Line 258 delete "relative". Basically, you want to say that the correlation is higher with the MSS method.
Lines 278 and 279: "Although the in situ..." is not a sentence. Delete "although".
I'm not sure that I find the discussion on lines 278-282 to be satisfactory. I don't understand your surface heterogeneity argument. I agree that ground-based versus satellite pixel may see different things but I'm not sure how that explains the differences shown in Figure 4. What accounts for the temporal variability in the SURFRAD albedo shown in Figure 4b? In particular, the lower albedos near days 30, 61, 183, 204 and days between 274 and 305. I do agree, however, that more work is needed for validation.
While the MSS method provides a more complete data record, which is the point of this paper, I am troubled by the poor agreement for the Desert Rock SURFRAD site for both the MCD43A3 and MSS albedos with the MCD43A3 providing the "better" agreement.
Author Response
We would like to thank the anonymous reviewers for their valuable comments that helped us to revise and improve the presentation and the technical context of our paper. In the following, we addressed all comments and suggestions made by the reviewers, and checked the spelling and grammar mistakes throughout the manuscript with the aid of an English editor. The corrections made in this revision are highlighted in blue. All numbered items (pages, equations, figures, and references) are consistent with those in the revised manuscript, excepted if otherwise stated. For convenience, the comments of the reviewers are repeated below in italics.
Comments and Suggestions for Authors
The paper, Estimation of Land Surface Albedo from MODIS and VIIRS data: A multi-sensor strategy based on the direct estimation algorithm and statistical-based temporal filter, is for the most part well written. There are awkward phrases such as on line 100: "The DEA approach enables to estimate..." which might be better as "...enables us to estimate..." or "....enables estimation of..."
Author’s reply: We have revised it as “...enabled us to estimate...”. Please see Page 3, Line 99 for more details.
Similarly, line 109 "global distributed" might be better as "globally distributed".
Author’s reply: We have revised it as your recommendation. Please see Page 3, Lines 108-109 for more details.
Line 115 "descripted" should be "described"
Author’s reply: We have revised it as your recommendation. Please see Page 3, Line 114 for more details.
Line 131 "are that" should be "is that"
Author’s reply: We have revised it as your recommendation. Please see Page 4, Line 131 for more details.
Line 150 it is that data that are used....so "spectral library was" should be spectral library were"
Author’s reply: We have revised it as“we used 1,232 samples of spectral reflectance data from the U.S. Geological Survey (USGS) spectral library to establish the relationships…”. Please see Page 4, Line 150 for more details.
The MCD43A3 data set isn't defined until page 7 but is used in 3 prior places. It would be better to define it where it is first used.
Author’s reply: We have moved the definition of MCD43A3 product to its first appearance. Please see Page 6, Lines 180-181 for more details.
Line 247 should be "events"
Author’s reply: We have revised it as your recommendation. Please see Page 8, Line 245 for more details.
Line 258 delete "relative". Basically, you want to say that the correlation is higher with the MSS method.
Author’s reply: We have revised it as your recommendation. Please see Page 10, Line 256 for more details.
Lines 278 and 279: "Although the in situ..." is not a sentence. Delete "although".
Author’s reply: We have revised it as your recommendation. Please see Page 10, Line 279 for more details.
I'm not sure that I find the discussion on lines 278-282 to be satisfactory. I don't understand your surface heterogeneity argument. I agree that ground-based versus satellite pixel may see different things but I'm not sure how that explains the differences shown in Figure 4. What accounts for the temporal variability in the SURFRAD albedo shown in Figure 4b? In particular, the lower albedos near days 30, 61, 183, 204 and days between 274 and 305. I do agree, however, that more work is needed for validation.
While the MSS method provides a more complete data record, which is the point of this paper, I am troubled by the poor agreement for the Desert Rock SURFRAD site for both the MCD43A3 and MSS albedos with the MCD43A3 providing the "better" agreement.
Author’s reply: We checked the site information of SURFRAD Desert Rock (DRA, 36.62373°N, 116.01947°W, Nevada, United States of America). It located on the Nevada Test Site, 65 miles northwest of Las Vegas, which is a warm semi-arid site covered with bright mineral soil and sparse grass. Thus, there are less seasonal variations of land surface albedo and few snowfall events. The upwelling pyranometer was mounted on a ten-meter tower (see the attached file Figures 1 and 2), which had a different footprint compared with the satellite remote sensing data. We conjectured that the lower albedo values near days 30, 61, 183, 204 and days between 274 and 305 may be caused by precipitation, cloud shadow, or human activities. However, we did not have sufficient information to infer what happened at these days. We have checked the meteorology records of Desert Rock site, which showed lower downward solar radiation, and higher relative humidity at these days (see the attached file Figure 3), which may support our hypothesis. Unfortunately, the lower albedo values at Desert Rock site cannot be detected and represented by neither of the MSS approach and MCD43A3 product. We have revised sentences in this discussion: “In addition, we found that the lower albedo values at FPK site (Figure 4b) cannot be well represented by neither of the MCD43A3 product and MSS approach…, which suggested that more fieldwork and investigations are needed to improve the estimation and validation of land surface albedo.” Please see Page 10 , Lines 275-284 for more details.
We acknowledged that the estimation accuracies between the MSS approach and MCD43A3 product under snow-free conditions are quite similar. However, we emphasized that the main improvements for the estimation accuracy are mainly occurred during the snow-covered period, as the temporal variation of albedo could be well captured by the MSS approach. We have revised the sentences in this discussion, and Sections of abstract and conclusion. Please see Page 10, Line 275; Page 1, Line 27; and Page 16, Line 398 for more details.
Thanks for your positive evaluation and valuable comments that helped us to improve the presentation of our work. We have also carefully checked the spelling and grammar mistakes throughout the manuscript with the aid of an English editor. Please see the words labeled in blue for more details.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors did not improve anything in response to my first points.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
This paper seeks to address a current open problem in the remote sensing community: merging data from multiple sensors. In particular the authors are examining the application to land surface albedo with data derived from MODIS and VIIRS. Overall the results presented in this paper are underwhelming, I was hoping for much more progress than what is presented.
Science comments:
1) It would be helpful to include the central wavelengths of MODIS bands as compared to VIIRS bands, perhaps in Table 1. This would help with interpreting the coefficient values.
2) There is no discussion about errors that are introduced through all the work-flow steps, especially the conversion from VIIRS to MODIS bands. It needs to be addressed.
3) The weights described in Equations 6 and 7, and illustrated in Figure 2 are not motivated. Why are there seasonal dependencies? Why is VIIRS data weighted less in the winter? Why does 50/50 not work?
4) The largest scientific deficiency in this paper is in the datasets being applied. Identical spatial and temporal footprints are selected for MODIS and VIIRS. Why are the authors not using the finest available? Why are you not addressing how to blend data with different spatial and temporal footprints? There have been studies in the statistics community that have addressed these issues (e.g. (2014) Spatio-Temporal Data Fusion for Very Large Remote Sensing Datasets, Technometrics, 56:2, 174-185, DOI: 10.1080/00401706.2013.831774)
5) MCD43A3 is being proposed as an independent validation dataset. But how independent is it really, as it also is derived from MODIS? More discussion needs to be made comparing the MSS algorithm to the MCD43A3 algorithm and why the authors are touting the current product as an improvement.
6) Consider showing in situ sites on a map. Also include some detail and discussion about land cover at the sites.
7) In figures 5 and 6 it sees like MSS has a lower threshold given the horizontal spread in the scatterplots. Please provide an explanation.
8) How are pixels determined to be snow-free or snow-covered? I couldn't find in the text.
9) The conclusions in Section 3.2 are overstated. IT seems like it can not be concluded that MSS is better than MCD43, but that simply they are similar in snow-free conditions.
10) Why is the snow-free R values so much higher for FLUXNET than SURFRAD comparisons? Please provide explanation.
11) The plots in Figure 10 are very hard to read. Does MSS always have 365 days of data? It could be helpful to show a table about number of days per year data is available for other data combinations?
12) Lines 296-298: There is nothing in this paper describing efficiencies of this method for massive data, so remove sentence.
13) No details on the accuracy of gap-filled data are given. Just because gaps are filled does not mean the data is of good quality. The authors need to do more comparisons with in situ data both with original MSS and then gap-filled MSS data.
14) Lines 306-309: It is not at all clear to the reader how the change of input would cause a reduction in computations. Please explain.
Editorial comments:
1) Please have the paper edited for English grammar, there are a number of readability problems throughout.
2) Terra is mis-spelled as "Tera" throughout, please correct.
3) Be consistent in italicizing (or not) the term "in situ".
4) Typo in title of Section 3.1
5) Be sure to include description of subplots in all figure captions.
6) Line 233: appears to be a copy-paste error for snow-free R value.
7) Line 244: Do you mean "snow-free" instead of "snow-covered"?
Reviewer 2 Report
Brief summary :
This article offers to merge MODIS and VIIRS data to produce an global albedo (multi-sensor). VIIRS data is converted to MODIS-like data using a linear model that has been build a database of simulated MODIS and VIIRS observation (using a BRDF database). After mixing the data, a temporal filter is applied to remove gaps in the final data. Evaluations with in situ data show some improvements compared to mono-sensor albedo.
Broad comment :
The main idea of offering a dataset based on multiple sensors is very welcome. The structure of the article is clear and easy to follow. The novelty of the DEA approach part is low, especially because the interest of using atmospheric corrected reflectance remains unclear (see comments). Overall, this article offers a good extension of previously published material and simple proof-of-concept on multi sensor merging.
Detailed comments :
L 28 : Line ending with “-” should stick to the number “0.0034”
L 43 : “heterology”. Do you mean “heterogeneity”
L 59 : Citing [21] may not be the most relevant here.
L 60 : Citing [35] may not be the most relevant here.
L 76 : typo “senor” instead of “sensor”.
L 79 : typo “in situ” should be italics.
L 102, l 104 : “adopted”. I am unsure about this. Do you mean “adapted” ?
L 106, L140-144 : The BRDF database could be better described. It seems that is does not take into account soil wetness or snow.
L 122: “It is noted the main difference between this study and former studies are that the input variable was the atmospheric corrected land surface reflectance instead of the top of atmosphere reflectance.” The atmospheric correction processes seem to be different for the MODIS and VIIRS sensors, an analysis or a description of these differences is required, especially because this is the novelty of the presented DEA approach.
L 131 : “the most efficient way is to convert” : not based on evidence. Why is it so ?
Formula (4) : Unclear : the index “i” in the LHS is the same as the index inside the sums in the RHS.
L140 : “ground-measured spectral data at wavelength λ_i”, Does this quantity depends on the solar spectrum. A clarification is needed : should the solar spectrum not be used in the formula (4) ? If so, a description which solar spectrum (Top of atmosphere, or after atmospherically filtered spectrum).
Formula 5, L 262 : Please clarify why a capital “A” is used when the albedos are “alpha” everywhere else in the article.
L 168 : replace “on average” by “on an average” ?
L 195, L 197 : Clarify why using “clear-sky” and “blue-sky” apparently for the same value.
L 233 : typo : 0.6390 instead of 0.53
L 318 : typo : “various of global” should be “various global”
L 321 : typo : “estimation the land” should be “estimation of the land”