Comparison of the Remapping Algorithms for the Advanced Technology Microwave Sounder (ATMS)

Round 1

Reviewer 1 Report

Authors have compared two remapping algorithms, the Backus-Gilbert inversion (BGI) technique and the filter algorithm (AFA), widely used in the operational data preprocessing of the Advanced Technology Microwave Sounder (ATMS). The algorithms are compared using simulations and real ATMS data. Further claim that, the use of the antenna pattern instead of a point spread function in the AFA algorithm create the persistent bias found in the AFA-remapped image.  Results show that both implemented algorithms can effectively enhance or degrade the resolution of the data. I would recommend that author's need to explore the literature further and improve the introduction with furthers state of the art. Also, the  ATMS Remapping Algorithms section including Backus-Gilbert Inversion Algorithm AAPP Filter Algorithm need to be revised again and some baisc equations need to be removed. 
Figure 2 and 3 the axis, legends, and texts ae blurred and not visible if printed. (same apply for fig 4 and 5)
"The comparison between the PSFs before and after applying the BGI correction demonstrates that the BGI can achieve some improvement for the resolution enhancement and can near-perfectly match the PSFs for the resolution degradation". The sentence need to be clarified further with proper explanation before simulations.
The impact of the irregularity of the antenna pattern on remapping need to discussed in detail.
From fig. 9 the BGI have stronger effect in remapping than the AFA, any proper reasoning for that?
Finally conclusion should be revised and limited to one parahraph with acheived observations.

 

 

 

 

 

 

Author Response

Please refer to the attached file. 

Thank you!

Author Response File: Author Response.docx

Reviewer 2 Report

The paper by Jun Zhou and Yu Hang is very well presented and argumented. I have basically no in-depth comments.

I just recommend that the authors emphasize, in their description of the algorithms, the differences between isotropic (that depends only in the distance between points) and non isotropic kernels (eq. 1.)

Isotropic kernels, by essence will behave badly in coastal areas, as observed by the authors in line2 259-260, and nothing is new here.

The literature about non isotropic kernels is pretty scarce, but I give a downlink, in this review, about a Master's student paper that was published many years ago in a Bulletin of the Int. Association of Geodesy, dealing with covariance functions of the gravity field in coastal areas. This may help or induce ideas.

A simple anisotropic model of the covariance functions of the terrestrial gravity field over coastal areas, J. Chenal, J.P Barriot, http://www.isgeoid.polimi.it/Newton/Newton_2/newton_2.html
Newton's Bulletin, Issue n. 2, December 2004

 

 

Author Response

Please see the attached file.

Thank you!

Author Response File: Author Response.docx

Reviewer 3 Report

This is an interesting paper with potential to add some understanding of the tradeoffs involved when choosing a remapping algorithm for the ATMS measurements. However, there are numerous omissions which make it incomplete for understanding and would prevent duplications of the authors results. My specific questions and comments are given below.

Section 3.1 discusses using n x n adjacent original Tas in the BGI algorithm. The authors should clarify whether the intent is that the adjacent samples are an n x n grid of Tas in the along-scan and along-track directions. If that is the intent then please explain why this choice is appropriate at the edge of the scan where the footprints are elliptical so that there are more footprints that have a significant antenna gain at a given point in the along-scan direction? Lines 154-155 state that "The noise suppression term is not necessary for the resolution degradation, as the process does not amplify the noise as the resolution enhancement does." While it is true that the resolution degradation will usually not result in noise amplification the gamma factor is still a tradeoff factor to control the level of noise suppression versus the accuracy of matching the target footprint. Further justification should be given for the choice of γ = 0. Lines 156-157: What criteria was used to choose the window sizes of 3 and 5? These choices should be justified in the paper. This is also related to my first comment above. Lines 162-167: A description of the antenna patterns or an appropriate reference should be included for the real antenna patterns. This should include the resolution of the antenna patterns and how they were measured or modeled. How do the real antenna patterns compare to the Gaussian approximations used for the comparison in Section 5.2? Section 3.2: My reading of this section is that the real antenna patterns are not used for the AAPP filter algorithm. Please clarify whether the form of 7a is used to approximate the real antenna pattern. Is the approximation the same as that used in Section 5.2? Is the term point spread function (PSF) intended to denote the effective antenna gain pattern projected onto the earth surface or is there some additional processing involved? Please clarify this. Section 5: What value is used for the cutoff factor in the AFA algorithm? Section 5: The value or values used for γ in the BGI (for resulution enhancement) should be given and justfied. Was this optimized based on some criteria? Line 263-264: Were a range of values for the cutoff factor c used to "optimize" the AFA algorithm in some sense? Could further tuning reduce the biases around the coastlines? Figure 4e-f: Do the patterns spatial extent of the biases make sense when compared to the number of samples used to construct the remapped values and the resolution of the source and target footprints? Some discussion of this should be included in the paper. Some readers may also benefit from an explanation for the sign of the biases over land and water. Lines 265-269: It is not obvious to me that Figure 4e shows a larger bias near the edge of scan for the BGI. Please explain further. Section 5.2: The remapped pattern on the surface from the BGI algorithm can be calculated. Providing a quantitative comparison of the matching between the remapped and target footprints would aid interpretation of the results in this section. Section 6:How was the ATMS data used obtained? Please provide appropriate documentation and references for the source of the data.

Author Response

Please see the attached file.

Thank you!

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thanks for the revisions.

All my concerns are met.

Author Response

Thank you very much!

Reviewer 3 Report

The authors have adequately addressed my comments on the original manuscript with
one exception.

My comment regarding the sign of the biases near coastlines was intended to refer to what is now Fiugre 9a (I apologize for not having made this clear). My interpretation is that the sign of the biases is due to the beamwidth fo the remapped PSF being larger than the simulated PSF. The remapped PSF therefore has higher gain from the ocean (lower BT) when the beam center is over land and higher gain from land (higher BT) when the beam center is over water. Is this correct? Please explain in the paper.

I am also surprised that the biases extend as far as they do from the coastline. Do the distances follow from the 3x3 grid of measurements and the beamwidths?

Author Response

Thank you for your in-depth comments. The response to each of them is as follows. The manuscript has been revised accordingly, and the changes made to it have been marked by using the "Track Changes" function in Microsoft Word.

My comment regarding the sign of the biases near coastlines was intended to refer to what is now Figure 9a (I apologize for not having made this clear). My interpretation is that the sign of the biases is due to the beam width of the remapped PSF being larger than the simulated PSF. The remapped PSF therefore has higher gain from the ocean (lower BT) when the beam center is over land and higher gain from land (higher BT) when the beam center is over water. Is this correct? Please explain in the paper.

Reply: Thank you very much for pointing out the essential reason of this problem. It is true that the special pattern of the bias around coastlines showed in Figure 5(e) and Figure 9(a) is due to the fact that the beam width of the synthetic PSF is larger than that of the expected PSF. The remapped PSF therefore has higher gain from the ocean (lower BT) than expected when the beam center is over land and higher gain from land (higher BT) when the beam center is over water. As this special pattern is first found in Figure 5(e), the following statement is added to the second paragraph of Section 3.2.1 from Line 314-319:

“Specifically, negative biases appear over land and positive biases over water, indicating that the resolution enhancement for the sharp change in signal is insufficient. This is due to the fact that the beam width of the synthetic PSF is larger than that of the expected one, which is consistent with Figure 4(a) and Table 1. The remapped PSF therefore has higher gain from the ocean (lower Tbs) than expected when the beam center is over land and higher gain from land (higher Tbs) when the beam center is over water. As discussed in Section 3.1, the resolution enhancement has to be sacrificed to suppress the amplification of noise and that leads to the imperfect matching of the PSFs and consequently the insufficient enhancement for the sharp change in signal.”

I am also surprised that the biases extend as far as they do from the coastline. Do the distances follow from the 3x3 grid of measurements and the beam widths?

Reply: Yes. For resolution enhancement, the window size is set to 3x3 and the beam width is expected to be narrowed from 5.2° to 3.3° for all 96 FOV positions. This has been clarified in Section 2.1 and 2.2.1.