Relative Radiometric Correction of Pushbroom Satellites Using the Yaw Maneuver

Round 1

Reviewer 1 Report

The focus of this manuscript is the relative radiometric correction of push-broom satellites using the yaw maneuver. The method to determine both FPM and detector level relative gain using the side slither data was detailed presented, and the results & discussion was also detailed introduced. In general, the manuscript is well written, and it can be a valuable reference to the related community.

The comments and suggestions are shown as follows.

(1) In line 47, “FPM” should be explained in the main text, when it appeared for the first time.

(2) For the relative radiometric correction using the yaw maneuver, the data process is very important to ensure the DN values of all detectors were correctly extracted when imaging the same target. So, the detailed pixel shift method is suggested to introduced in section 3.3.1.

(3) The data processing for Landsat 8 OLI and TIRS in section 3.2 and section 3.4 respectively are very similar, which makes the third section of the manuscript longer. So, I suggest merging the data processing description of the two sensors to simplify the section 3 “Methodology”.

(4) There are some sign errors for the explain of equation 4 in line 284 – 286, i.e., ““ should be , and ”Var“ should be .

(5) Please explain why using the periodic model to calculate FPM relative gain can be more justified from a consistent calibration perspective. Is there any relevant analysis?

Author Response

Thank you for your great feedback. I have implemented all the suggestions, explicitly: 

(1)  Agreed, the abbreviation explanation was added. 

(2) A more detailed process about the circular pixel shift was added to section 3.3.1.

(3) The data processing sections for OLI and TIRS were combined into one. 

(4) The explanation of equation 4 now matches the italicized portion of equation 4. 

(5) A small explanation as to how the periodic model is justifiable compared to other methods was added to section 3.5.2. Relevant analysis of the periodic models applicability for justifying its usage is shown in the results section 4.2.

Reviewer 2 Report

This paper gave a novel technique of relative gain estimation that employs an optimized modified SNR through side slither manuver to correct the differences in Landsat 8 OLI  sensor measurement. It presented that allows for both FPM and detector level relative gain calculation. A periodic model based on in-scene FPM corrections was designed as a go-to model for all bands aboard Landsat 8. Relative gains derived from the SS technique and applied to imagery provide a visual and statistical reduction of detector level and FPM level striping and banding in Landsat 8 imagery. Both reflective and thermal wavelengths are corrected to a level that rivals current operational methods.   The paper can be published after minor revision. But the abstract should be shortened especially in the first 11 lines because there is too much introduction description.

Author Response

Thank you for your great feedback. Your comments about the abstract containing too much introductory information have been taken into account and changed. Specifically, multiple lines before line 11 were deleted or altered in the abstract to remove non-necessary information. 

Reviewer 3 Report

The authors present a technique to correct for differences in relative gain estimation sensor measurement that employs a modified signal-to-noise ratio optimized through a 90◦ yaw maneuver, also known as sideslip, which enables both FPMs. and calculating the relative gain of the detector level. A periodic model based on FPM corrections at the scene was designed as a reference model for all bands on board Landsat 8. Relative gains derived from the sideslip technique and applied to the imagery provide a visual and statistical reduction of the level of the detector and the FPM level. and banding in Landsat 8 imagery. Both reflective and thermal wavelengths are corrected to a level that rivals current operational methods.

Although Landsat 8 is used as an example, the methodology is applicable to all linear array sensors that can perform a 90° yaw maneuver. However, it would be interesting to know the discussion about the precision in the use with other sensors that do not have moving parts to determine if it is a new and more reliable technique.

The presentation of the manuscript requires adapting to the instructions for authors of the journal. It does not contain a discussion of the results. I consider that it has more profile for a technical note: These are short articles (less than 18 pages) on new developments, significant advances and novel aspects of experimental and theoretical methods and techniques relevant to the scope of Remote Sensing.

Author Response

Thank you for your review. Your response is greatly appreciated. Although the side slither maneuver has been around for quite a while, the full potential of the method was never fully demonstrated. There has never been a thorough analysis done of how to process side slither data from beginning to end, including all operational scenarios.  This paper was an attempt to provide this exhaustive discussion/analysis and thus should serve as something of a reference for others who are working with similar types of whiskbroom scanning instruments.  This manuscript does not require adapting to our instructions but presents a tried and tested method for deriving the best relative gains, both FPM and detector, for all wavelengths of a push broom scanning satellite. Furthermore, a comparison between other methods and our method is portrayed and discussed within the results & discussion section. In said results & discussion section, a discussion over the statistical and visual testing of the data occurs, and from this discussion, conclusions are drawn.  The discussion is embedded in section 4.Results & Discussion. Thus, lack of a formal discussion section should not be a reason for changing to a technical note. This is an original research topic based on the lacking exploration of side slithers past usage. All methodology and data were collected, analyzed, and interpreted separately from previous studies, pointing towards a research paper as opposed to a technical note. MDPI's definition of the technical note states "These should describe important modifications or unique applications for the described method.", as opposed to this research paper where we describe how to use the side slither method fully and comprehensively for all push broom scanners and that it is not a unique application for Landsat 8. I ask you to reconsider your position on this being a technical note and again thank you for taking time out of your day to review this. 

Reviewer 4 Report

The authors discuss a relative calibration approach using the side slither technique for both reflective solar and thermal emissive bands that accounts for the relative detector effects within a focal plane module and across modules. Please address the following comments:

General

• The relationships between the signal levels/SNRs and the streaking metric could be expanded. It is not clear to me if this is the driver in deriving a relative calibration that reduces streaking. I would recommend adding a plot of SNR/radiance vs streaking metric.
• I think it might be helpful to have a figure with relative calibration coefficient per detector, to give the reader an idea of the magnitude of the relative calibration corrections and any spatial shape that is apparent.
• The paper seems a light on references for previous studies on side slither. It’s important to be comprehensive, since the authors claim that the work improves the established approaches and mentions that the approach has been around for decades.
• As evident in some of the comments below, the description of the origin of detector non-uniformities could be clarified.

Abstract:

• Line 33: The differences in relative spectral response functions between detectors can also contribute to striping.
• Line: 136 I think more explanation is needed on why a radiance decrease would affect the relative calibration. The decrease in radiance is not really the issue – the effect could be a consequence differences in detector non-linearity, changes in the source spectra (which would give different radiance due to their variation in detector-level spectral responses), or perhaps a change in the illumination uniformity.
• 2.1: An illustration of this quantization issue would be helpful. What does the data look like with 12 bits? How does the quantization noise compare with the NEDL?
• Line 224: It may be clearer if the authors explain that the bias is the dark counts plus the background counts with the background being negligible for the reflective solar bands.
• Line 241: Are these parameters derived pre-launch? Please include their origin.
• Line 272: You mention that the “noise does not increase at a greater rate than linearly when compared to the signal.” Is this dependence go as the square root of the signal, i.e. dominated by shot noise?
• Fig 14: Were the relative gains from the side slither back propagated in producing the blue line?
• 571: The origin of the CPF gains should be introduced earlier.
• Conclusion: Line 636: Please be more specific about the current issues with side slither here (and in the introduction).
• Conclusion, 647: Please be more specific about the issues with solar diffusers. For instance, the problem isn’t primarily degradation per say, it’s more that they degrade non-uniformly (or maybe if they degrade so much that the calibration derived is dominated by noise).

Author Response

Thank you for your review. Your comments and tips are greatly appreciated. 

• Line 33: Agreed, the difference in relative spectral response functions would contribute to differences between FPMs and potentially ever so slightly between detectors if the spectral filter is not uniform across detector arrays. 
• Section 2.1: More explanation was added on the degradation of both the diffuser panel and on-board lamp. 
• Section 3.2.1:  The major source of error with this quantization issue is that all the bias values and CPF values are meant for the 14 bit sensor and the data is received in 12 bit form. The raw data in 12 bit form does not change significantly between the 12 and 14 bit alteration, however, without the change the bias removed would be incorrect. The quantization noise is much smaller than the NEDL of the instrument as the NEDL is around 0.002 while the quantization noise is on a level of 1/6000. 
• Section 3.2.2: A further explanation of the bias was added per your suggestion. 
• Section 3.2.3: Yes, the parameters were derived pre-launch. A further explanation was added alongside the reference. 
• Section 3.3.2: Yes, the dependence was around shot noise. Another reference to shot noise and an explanation of it was added. 
• Fig 15: Yes, a single side slithers relative gains were used for all test images across the timeline. 
• 571: The origin of CPFs is now introduced in the section about Diffuser Panels at the beginning, within section 2.1.1. 
• Conclusion: Current issues for side slither are now expounded upon in Section 2.1.5 as well as the conclusion.
• Conclusion: Further explanation of the diffuser panel was added in section 2.1.1 and again concluded at this point. 
• The relationship between signal level and streaking metric was further explored in Section 3.7.1 and in figure 13. An explanation of it was expounded upon. 
• A graph of the detector relative gain vs detector number as well as an explanation was added. 
• Further sources on the previous side slither works were added in the background section on side slither.

Round 2

Reviewer 4 Report

Thank you for fully addressing all of the comments. There may be a couple of remaining typos in the added material, and you may want to add another reference or two to highlight past work on past slide slither approaches (but not required).