Calibration Inter-Comparison of MODIS and VIIRS Reflective Solar Bands Using Lunar Observations
Martin Snow
Pepe Phillips
Round 1
Reviewer 1 Report
Review of:
Calibration Inter-comparison of MODIS and VIIRS Reflective Solar Bands Using Lunar Observations: X. Xiong, J. Sun, A. Angal, and T. Wilson
Summary:
This paper compares the results of calibration using the Moon and on-board solar-diffusers for two pairs of low-earth-orbit instruments. It includes the
analysis of the effect of using a single modern solar spectrum versus 3
different older spectra.
This is an important paper and summarizes many prior efforts. The authors are well-know and well-qualified to do this work. The results (Section 4) are
enlightening and useful. The special study using the same solar spectrum for all instruments is a significant advance. The conclusions are succinct and
warranted.
The authors should be given ample time to respond to General Comments 1 and 4
----------------- General comments -----------------
1) The geometry of the optics, scan-mirror and rotation axis, HAM, and the SV and SD ports are critical to this discussion. A good diagram with real angles is required. Can you point to an open-literature description of the optical design of MODIS and VIIRS? E.g., A reader might wonder how EV and SD AOI's can be the same: Lines 291, 374 and 401.
2) Use of multiple-character symbols in several equations is open to confusion. e.g., Eq 1 with 'RVS' and 'dn*'. Replace with well-defined one=character symbols.
3) The historical use of 3 different solar spectra for these instruments is
unfortunate but unavoidable. Although the difference between the spectra may be much smaller than the trends of response, it would be nice to get all the calibrations used in production of science products on the same system. It is important to know the effect of using different reference spectra. This is done semi-qualitatively in lines 450-460. However, the intent of Fig 5, to show the differences between solar model used, is not met by the figure. Need to show the ratio of the spectra to some standard so that the differences at the 1% (or less) level are clear. Ideally this standard would be the HSRS, but it could be one of the spectra used, or their average.
4) Lines 382-383: "collect lunar data in the EV data sector." This merits much
more discussion as obtaining lunar calibrations using the identical optical
configuration as science data has been a major issue for many instruments. A major attitude maneuver is implied, and how that may effect instrument
temperatures and response coefficients could be significant. Please tell us what was learned.
Note: Here ">>" means 'change to become'
73: Briefly say why day and night band not included here. Also, including it in Table 1 might indicate to readers how it is unlike the other VIIRS bands.
78: The primary reference for the stability of the Moon is: H.H. Kieffer,
"Photometric Stability of the Lunar Surface", Icarus v.130, pp 323-327 (1997) 78: Citations [30-32] refer largely to the lunar model and better placement
would be in line 81.
81-82: See above, The current placement of citation [33-37] suggest that those developed the ROLO model, which is not correct. Better to put citations [30-32] here, and add a phrase or sentence to indicate use of the Moon by citations [33-37]. Note, ref. 33 predates ROLO.
118: This hints that differences between lunar observations using the SV and EV ports will be presented. However; section 4 has no quantitative discussion of this.
169: None of the 3 citations is in the reference list.
202-203: Do you mean that Delta_SD(angles, time) is assumed to be linear with angle or cos(angle) or what?
208: Stated already at line 162-163
210: last sentence is unnecessary
222; Need some statement about the linearity of instrument response (in each band?)
258: Explicitly define meaning of "calibration coefficient". As used here it is a
divisor of the measured irradiance, and would generally increase with instrument age.
259: Eq. 5 right-hand-side is same as ROLO/I; I from Eq 4, with modified m_1.
Easier to understand.
265: Cite only [30] here for development.
271: Fig 2 shows highest radiance levels are NOT near the center of the Moon.
293-294: reword to associate the angles explicitly with the correct port.
Fig 3: Very interesting figure! The small annual oscillation in the lunar
trending is probably due to small errors in the ROLO libration terms, magnitude less than 2%, indicative of the precision of ROLO. The different trends for Terra B8 need discussion.
329:331 The description of F is convoluted; and Eq 7 makes it more difficult.
358: Unclear what "in the declination angle" means.
360 and 460: RTA is undefined.
364-367: This sentence seems unnecessary given Eq. 7.
367: Need citation for Kurucz
382: What is a "sector rotation"
Table 3 caption: difference >> Terra minus Aqua
Table 3 and 4: what is the basis for the listed uncertainty?
Table 4 caption: difference >> SNPP minus N20
Table 5 caption: difference >> Aqua MODIS minus N20 VIIRS
675-676 and 683-685: say basically the same thing
688-689: IFOV needs a careful definition. Note that it is not the response solid angle of a detector that is important, but the solid angle that the detector represents in target space. This also depends upon the Moons motion during image acquisition. There are subtle distinctions which can be important at the 0.1% level.
694-700. Some Journals do not permit discussion of future plans. Editor decision.
---------------- Minor: typos and grammar ----------------
In general, subscripts in equations appear in text only as smaller font; they
should be subscripts. E.g., 156: m subscript 1. 164 and 165: make EV subscript.
Table 1 caption: spatial resolutions >> spatial resolution at nadir
Table 1 heading: Use two lines for VIIRS heading, as done for MODIS
Table 1 : Remove the "or" from column 4 and remove the comma in Col. 6
152: remove dot before MODIS
184: Equation location
268: move citation before period
286: 'RSB', you probably mean 'EV'
312 "It is worth mentioning" is subjective. Perhaps "Note"
327: Equation placement
580: based their time-series >> based on their time-series
Table 3 and 4: The lower "Band" line should be in bold.
Author Response
Please see the attachment
Author Response File: 
Author Response.docx
Reviewer 2 Report
Overall this is an excellent article. The description of the lunar calibration algorithms are well presented and the text is well written. My comments below are all minor, mostly just asking for a clarification of additional comment by the authors.
Line 155: Should L be L_sun? If you are using the 1-au solar spectral irradiance, don't you get an incident irradiance also corrected to 1-au?
Line 168: Would it not be simpler to use the TSIS-1 reference spectrum that you mention later (reference 57) rather than the composite of the three spectra?
Line 202: You mention that you assume the SD degrades uniformly with angle. Is there a possible calibration observation that could validate that assumption? (that may be beyond the scope of this article)
LIne 230: You mention that lunar reflectance stability is "well known." Is there a reference? I think that some of the ROLO papers might include such a reference.
Line 244: Could you explain what "oversampling effects" are?
Line 265: The 5% systematic uncertainty in the ROLO model is indeed larger than your calibration uncertainty. Another source of uncertainty that is not mentioned is the variation in the solar spectrum. At these wavelengths SSI variability is small, so it will probably not change your error estimates. I am not suggesting that you need to include solar variation in this article, but perhaps a future version of the algorithm could include a time-varying solar spectrum.
Line 273: Could you have chosen a different phase angle to reduce the signal and thus avoid saturation? Can you change the integration time to reduce the effect of the saturation?
Line 286: Is the variation in view direction reproducible? How large of an effect on your uncertainty does view direction have?
Line 367: More recent measurements (reference [57]) shows that the Thuillier spectrum is systematically off in the infra-red by 5-10%. If N20 uses this spectrum while the others do not, that could produce a systematic error. In the conclusion (line 676) you mention using a single solar spectrum for all analysis. I agree that future versions of the analysis should do that.
Line 432: When calculating the lunar geometry, I assume you use the spacecraft location to accurately calculate phase angle. That will be different for each spacecraft. Does that help account for differences in inter-comparison?
Tables 3-7: The caption should mention that the uncertainties are in percent.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 3 Report
This article describes how years of lunar calibration data are used to inter-calibrate MODIS and VIIRS with direct relevance to other similar missions. It demonstrates the strengths of lunar calibration compared to other methods for inter-comparison of Earth-observing sensors which is critical for climate data records. In particular, the depth of knowledge of the calibration of the individual sensors concerned, combined with length of calibration data records is a real asset and the methodology can be extended to JPS2S-2 and other sensors in the future. Some minor editorial comments for improvement are suggested below:
Line 196: reference for SD characterisation using yaw manoeuvre could be included.
Line 199: reference for on-orbit evaluation of VF could be included.
Figure 3: What causes the gain for band 1 to increase over time (normally they decrease)? Are these SD+SDSM corrected gains rather than just SD as caption indicates?
General comment concerning the inter-comparisons for clarification: Are the MODIS-VIIRS inter-comparisons affected by the difference in lunar phase angle compared with the MODIS-MODIS, VIIRS-VIIRS inter-comparisons which have the same lunar phase angle? Or is this adequately corrected for?
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 4 Report
I have no major concerns with this paper and some suggestions for improvement. The authors are experts in MODIS/VIIRS calibration, and the subject is certainly important given the long record of MODIS in orbit, the future for VIIRS, and the use of lunar calibration as a general method. The citations and introduction are appropriate. Section 2 provides a summary of lunar calibration for MODIS and VIIRS in a clear way.
Some comments and suggestions for improving the article:
1. Typesetting in Table 1 has commas in the last column that are poorly placed. I don’t think they are even needed. Some equations could be made more legible with larger fonts and dots added between factors to show multiplication. The latter is particularly useful with multicharacter symbols such as “RSV”. This is done sometimes but not always. The authors or the journal can fix other typesetting issues that are also in the draft.
2. On line 230, we find a sentence beginning with “It is well known that…”. This statement is inappropriate for a journal article, and the authors would do better by summarizing why the moon is a stable calibration target (refs. 30-32 may be appropriate citations).
3. The reference solar spectrum is different for different sensors. At this point in the manuscript, one would think that a single common reference spectrum would be preferred in the general case. The authors might address the question of why the reference spectra were not harmonized in MODIS C7 or one of the VIIRS collections. Maybe this is not possible or otherwise undesirable. There is discussion in the paragraph beginning line 473 about a common reference spectrum as a recommendation but this doesn’t link to what has been done with NASA’s processing of the various collections. One is left to wonder why.
4. Line 238 calculates the subtense of a pixel. This should be an invariant and its value could just be stated. Same on line 372.
5. Why not show more RSBs in Fig. 3 in addition to 1 and 8? Same for Figs. 4, 7, 8.
6. What is the meaning of “normalized” in Fig. 5?
7. The phrase “a number of references” is used in line 613 to describe two. Similar formulations are elsewhere. It would be better to change this to either something that doesn’t quantify the number or is more exact.
8. The discussion of the meanings and derivations of the uncertainties U1, U2, U3 could be improved (para. beginning line 608). The text is very terse on this subject. Maybe a little more on how they are derived from Eqs. 4, 8? Is this a Taylor series approach? This a major subject of the Results and Discussions. Maybe going to the references would answer the reader’s questions, but some more explanation here would be welcome.
9. It is not immediately clear whether the TSIS spectrum improved the intercomparison results, only that it changed them a little (line 670). Later it is stated that MODIS 17-19 became more in family with the other bands, but I could not find Table 9. Table 7 is the last one.
Author Response
Please see the attachment
Author Response File: Author Response.docx
