Extended Cross-Calibration Analysis Using Data from the Landsat 8 and 9 Underfly Event

Round 1

Reviewer 1 Report

I think this paper is appropriate to be published. The following comments may be considered for betterment.

1) The basic characteristics of each band should be described for anyone who is not familiar with in details of Landsat like me.
   For example ...
    Band 1-10?, 11? <-> CA? Blue, Green, Red, NIR, SWIR1?, SWIR2?, Pan?
    wavelength, bandwidth, ground resolution, noise, dynamic range.

2) The objective of this paper is cross calibration between L8 and L9. The popular uncertainty factors are discussed. The main topic is pixel classification. The classification accuracy should be discussed more clearly regarding to both spectral and radiometric perspective.

3) The classification is based on MODIS data. Because the ground resolution is so different between MODIS and Landsat, its effect should be discussed. The same discussion should be described for SCHIMACY resolution, also.

4) Even though the same filter lot is used, the spectral difference still exists. It is the main source of Table 6. The test results of filter spectral characteristics before launch should be described.

5) The methodology for TIR bands and Pa bands are insufficient.

Author Response

1) The basic characteristics of each band should be described for anyone who is not familiar with in details of Landsat like me.
   For example ...
    Band 1-10?, 11? <-> CA? Blue, Green, Red, NIR, SWIR1?, SWIR2?, Pan?
    wavelength, bandwidth, ground resolution, noise, dynamic range.

1) Table has been added to introduction

2) The objective of this paper is cross calibration between L8 and L9. The popular uncertainty factors are discussed. The main topic is pixel classification. The classification accuracy should be discussed more clearly regarding to both spectral and radiometric perspective.

2) Several of the papers referenced delve much deeper into the accuracy of the EPICS cluster classification, we’re merely just applying it. Clarification has been added

3) The classification is based on MODIS data. Because the ground resolution is so different between MODIS and Landsat, its effect should be discussed. The same discussion should be described for SCHIMACY resolution, also.

3) The point of using the EPICS clusters is so we could use the ground resolution of Landsat without worrying about the MODIS resolution. We used MODIS to create spectral profiles of the land cover types in November, then SBAF-corrected them so they look like Landsat, and then used the EPICS clusters for classification, which is all in Landsat resolution. Clarification is added to reinforce this

4) Even though the same filter lot is used, the spectral difference still exists. It is the main source of Table 6. The test results of filter spectral characteristics before launch should be described.

4) Not quite sure what this means. We applied SBAF corrections so L9 would look spectrally like L8. These results are based on the spectral band passes before launch. Test results indicated a 0.3% difference in the green band between L8 and L9 when viewing vegetation. This reinforcement in the paper

5) The methodology for TIR bands and Pa bands are insufficient.

5) The methodologies are nearly identical to the other bands, and any deviations are noted and clarified further

Thank you again for taking the time to read through our manuscript and providing very helpful feedback!

Reviewer 2 Report

This is a very good and detailed documentation of the L8 and L9 cross-calibration results with the data collected during the Underfly period in November 2021.  It includes the retrospection of the method, results and issues of an early time-limited inter-calibration effort, and the thorough description of the recent data analysis results with improved algorithms and better understanding of the underfly data.  Documentation like this paper is important for a satellite program. 

Also this paper is well written and organized, although it is a very long paper.

I would recommend considering to accept this paper for publication purpose.

Minor suggestions to the authors:

1) How about add one table to show the central wavelengths and their main applications for the OLI sensors?

2). Please consider adding several sentences in the Introduction section to  briefly describe how the collocation areas are defined.  This will help readers to understand why the viewing/illumination error is considered as a geometric uncertainty in Section 1.21.

 

Author Response

1) How about add one table to show the central wavelengths and their main applications for the OLI sensors?

1) Table has been added to the introduction

2). Please consider adding several sentences in the Introduction section to  briefly describe how the collocation areas are defined.  This will help readers to understand why the viewing/illumination error is considered as a geometric uncertainty in Section 1.21.

2) Clarification added

Thank you again for taking the time to read through our manuscript and providing very helpful feedback.  Your suggestions show you understood the material and your constructive criticisms are much appreciated!

Reviewer 3 Report

This manuscript demonstrates the work on the extended cross-calibration analysis using data from the Landsat 8 and 9 underfly events. The uncertainty of cross-calibration has been decreased greatly, which is in part due to the similarities between the instruments, but also from the approach described in this paper which can give us some insight. Overall, the method description and result analysis in this article are complete and detailed, but some minor issues need to be explained and modified.

(1) In section “3.3.1 Spectral Uncertainty”, the text is “The resulting SBAF ratio means and standard deviations are shown in Table 9…”, but the standard deviations are not shown in Table 9. In addition, I think the calculation description of spectral uncertainty from these 26,460 simulations is necessary.

(2) Similarly, the BRDF uncertainty calculation hasn’t been introduced based on 4.2 million samples, and what are the corresponding conditions for uncertainty given in the Table.10? For example, what is VZAD?

(3) For Equation (10), geometry uncertainty is considered a bias, and additional explanations and evidence for this determination need to be supplemented.

(4) It can be suggested to explore what other requirements exist to reduce the uncertainty if the two satellites are not loaded with nearly identical spectra. For example, the reference satellite is a CLARERO-like mission. In this way, the final conclusion of this manuscript will be more convincing.

(5) Small labels (a) to (e) in Figure 1 are missing, and the last sub-figure is not very clear.

Author Response

(1) In section “3.3.1 Spectral Uncertainty”, the text is “The resulting SBAF ratio means and standard deviations are shown in Table 9…”, but the standard deviations are not shown in Table 9. In addition, I think the calculation description of spectral uncertainty from these 26,460 simulations is necessary.

1) Clarification added, the calculation describes that the standard deviation can be considered the spectral uncertainty across FPMs, IGBP types, and atmospheres 

(2) Similarly, the BRDF uncertainty calculation hasn’t been introduced based on 4.2 million samples, and what are the corresponding conditions for uncertainty given in the Table.10? For example, what is VZAD?

2) Clarification added, the calculation is actually an interpolation based on the linear trend of each VZAD uncertainty

(3) For Equation (10), geometry uncertainty is considered a bias, and additional explanations and evidence for this determination need to be supplemented.

3) Clarification added

(4) It can be suggested to explore what other requirements exist to reduce the uncertainty if the two satellites are not loaded with nearly identical spectra. For example, the reference satellite is a CLARERO-like mission. In this way, the final conclusion of this manuscript will be more convincing.

4) More examples given

(5) Small labels (a) to (e) in Figure 1 are missing, and the last sub-figure is not very clear.

5) Labels added and figure increased in size

Thank you again for taking the time to read through our manuscript and providing very helpful feedback.  Your suggestions show you understood the material and your constructive criticisms are much appreciated!