Intercomparison of Resampling Algorithms for Advanced Technology Microwave Sounder (ATMS)

Round 1

Reviewer 1 Report

This manuscript compared three resampling algorithms for Advanced Technology Microwave Sounder (ATMS). Two of the algorithms, i.e., BGI, and AAPP methods, are widely used in observations by satellite microwave sounding instruments, while the third one is a modified AAPP method. This inter-comparison is valuable particularly for satellite data calibration scientists, environmental data record product developers, and numerical weather prediction (NWP) users. In addition, the modified algorithm demonstrates certain improvements over the two existing BGI and AAPP methods. The manuscript is well written too.  Therefore, the manuscript is recommended for publication after the following minor revisions are made.

1)      The authors carried out a quantitative assessment for the performance of the three algorithms by using an RTM simulation approach for the typhon Lekima on August 8, 2019 and a qualitative assessment by using ATMS observations on September 16, 2021. It would be very interesting to apply the simulation approach to the same date for ATMS observations. Similarly, the authors can analyze the ATMS data on August 8, 2019. This cross-analysis might help better understand any inconsistencies in the conclusion from the two approaches.

2)      The authors only assessed the performance of the three algorithms for two limited cases. The conclusions in the manuscript might not be general. So, it is necessary to point out this limitation in the “discussion and conclusion”.

3)      Suggest to change ‘2. Materials and Methods’ to ‘2. Descriptions of ATMS Instrument and Three Resampling Methods’

4)      Re-write the following sentence in lines from # 201 to # 203 in page 7 since it is too long and has grammar errors: “But it does not represent the relationship between resolution and noise very well, nor can reveal the noise amplification effect caused by resolution enhancement and the smoothing effect caused by resolution reduction.”

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript is concerned with the problem of adjusting images obtained with different antenna beam widths to a common spatial resolution. The authors propose a new adjustment algorithm, and show that it improves on both the ATOVS-AVHRR algorithm and the Backus-Gilbert algorithm. The new images look good; however, some additional discussion as noted below would make the presentation more complete.

1. The NEDT values in Table 1 are from the ATMS specification. Measured NEDT's on the built instruments are appreciably lower (Kim et al., 2014, 2022). Using the specification in the simulations illustrates differences between algorithms, but it would be worth pointing out that the lower instrument noise would permit more complete restoration of images.

2. Fig. 9 shows some edge effects with the Backus-Gilbert algorithm. However, Fourier-transforming an image can also introduce edge effects if one edge has a different level than the opposite edge. How do the AAPP and the modified-AAPP algorithms handle edges?

3. It would aid readers' understanding to have some discussion of how eqs. 6 and 7 were arrived at. For example, line 109 notes that on the right side of eq. 5, the ratio of target to source Fourier-transformed antenna beams is an adjustment factor. Let that ratio be denoted by
 M(zeta) = G_target(zeta)/G_source(zeta),
where zeta is the two-dimensional spatial frequency vector. For the Gaussian model beams, G(zeta) is real.
Noise amplification occurs when M(zeta)>1 (restoration). So two possible regularizations M'(zeta) are defined by eqs. 6 or 7. To reduce noise amplification, one would want
 M'(zeta)<M(zeta),
but it's not clear that either eq.6 or eq.7 necessarily satisfy this condition, because in both of them, the factor that multiplies M(zeta) on their right side is a function of G_target(zeta) rather than M(zeta).
On the other hand, in the case of smoothing, where M(zeta)<1, the noise is not amplified by eq.5. The simplest choice then would be M'(zeta)=M(zeta), but again eqs. 6 and 7 appear to do something different.

4. Other comments:
Line 18: change "of at least" to "by at least".
Lines 97-99: The sentence beginning "In this study..." is not clearly written.
Lines 138-139: The sentence beginning "For a 5.2..." is awkward.
Line 365: change "Bucks" to "Backus".

References
Kim et al. (2014), J. Geophys. Res. Atmos., vol.119, pp. 5653-5670, doi:10.1002/2013JD020483
Kim et al. (2022), IEEE Trans. Geosci. Rem. Sens. vol.60, 5302813, doi:10.1109/TGRS.2022.3148663

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Spaceborne passive microwave observations are proven very valuable for improving the numerical weather prediction and environmental monitoring. The microwave observation data resolution varies with frequency and satellite viewing angle, it is normally required that the measurements at each frequency be resampled to obtain a uniform resolution prior to various applications. This is a key issues for microwave data quantitative applications. This paper compared three different resampling methods, but there are two main suggestions for authors:

1.  The introduction of the three resampling methods is too simple, especially the third method. The reasons for the improvement of the AAPP algorithm and its physical basis should be clearly explained in the article.

2. The comparisons only applied to 2 channels of ATMS, this is too weak for supporting the main conclusion of this paper. These should add a table to detailed show the comparison results applied to all channels of ATMS at least. At the same time, it is suggested to apply these comparative analysis to China's Fengyun satellites' microwave sensors, which will have greater guiding significance for readers.

There are some detailed suggestions:

The expressions of author 2's affiliations is not correct.

All figures captions are too simple for readers and missing units for the z-axis, need to be added.

The MAE, RMSE and BIAS in table 2 are also missing units. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

I still think a whole table listed all 22 channels of ATMS comparison results should be added in the manuscript.  Others look good.

Author Response

Thanks a lot for your suggestion, the results for channels 1-16 are already listed in the article.

But for channels 17-22, their FOV size is 1.1 degrees, and no resampling experiments are performed in this paper. As shown in Figure 2 in the paper, the BGI algorithm uses a 3x3 window for channels 1-2 and a 5x5 window for channels 3-16. 

There are two reasons for not experimenting with channels 17-22.

Reason 1: On AMUS-A, there is no corresponding channel for ATMS channels 17-22. 

Reason 2: In fact, the 3×3 window averaging process is considered to be a common method to convert 1.1°FOV to 3.3°FOV. The BGI algorithm and the AAPP algorithm have not yet been implemented on channels 17-22.