Comparison of Accelerated Versions of the Iterative Gradient Method to Ameliorate the Spatial Resolution of Microwave Radiometer Products

Round 1

Reviewer 1 Report

 

The following is the review for:

Comparison of accelerated  and preconditioned iterative gradient techniques

For microwave super-resolution products in support to regional scale studies

 

 

Summary: A basic issue in the application of microwave remote sensing is the

Lower resolution. This limitation can prevent the applicability to coastal and high resolution regimes where mesoscale and submesoscale dynamics dominate. The paper focuses on a technique to improve the resolution of microwave data. The paper describes three algorithms for improving the resolution of the microwave data based on iterative approaches. One goad was to improve both the resolution of the microwave data, while minimizing the number of iterations for convergence. I will leave commentaries on the methodologies applied to others who might have more expertise on the subject. The others present statistics to show how well the algorithms are performing based on direct comparisons with the reference field.

 

Strengths: An improvement in the resolution of microwave data would be huge breakthrough.

As an example remote sensing of sea surface temperature is limited to the availability of infrared data for deriving high resolution (< 25km) SSTs. The big issue here is that such data is not available during periods of cloud cover, which makes improvements of microwave data critical for resolution of mesoscale and submesoscale features. Thus the overall goal of the

work could lead to breakthroughs in application to coastal dynamics.

 

Weaknesses: The paper itself has several key weaknesses which leads me to recommend

Rejection but with a recommendation of resubmission.

A major selling point would be to present a map showing the spatial resolution of brightness temperatures before and after the application of the iterative approach? Right now it is hard

to recommend the publication in remote sensing because the paper is focused exclusively

on the technique, with error analysis, but has no examples? The authors have done a tremendous job with their expertise in comparing the different techniques, but they

need to give an example of it’s application.  Once this is done I think the paper should be resubmiited.

 

Minor: Additionally the paper needs a thorough proofreading as it contains multiple errors.

 

Overall I think the work could lead to a huge breakthrough in the application of microwave data, but needs to take things one step further.

Author Response

Authors: We thank the AE and the reviewers for his/her useful comments/suggestions that have been incorporated in the new version of the manuscript. All the suggested major adjustments are marked in red in the new version of the manuscript.

Summary: A basic issue in the application of microwave remote sensing is the

Lower resolution. This limitation can prevent the applicability to coastal and high resolution regimes where mesoscale and submesoscale dynamics dominate. The paper focuses on a technique to improve the resolution of microwave data. The paper describes three algorithms for improving the resolution of the microwave data based on iterative approaches. One goad was to improve both the resolution of the microwave data, while minimizing the number of iterations for convergence. I will leave commentaries on the methodologies applied to others who might have more expertise on the subject. The others present statistics to show how well the algorithms are performing based on direct comparisons with the reference field.

Strengths: An improvement in the resolution of microwave data would be huge breakthrough.

As an example remote sensing of sea surface temperature is limited to the availability of infrared data for deriving high resolution (< 25km) SSTs. The big issue here is that such data is not available during periods of cloud cover, which makes improvements of microwave data critical for resolution of mesoscale and submesoscale features. Thus the overall goal of the work could lead to breakthroughs in application to coastal dynamics.

 

Weaknesses: The paper itself has several key weaknesses which leads me to recommend

Rejection but with a recommendation of resubmission.

A major selling point would be to present a map showing the spatial resolution of brightness temperatures before and after the application of the iterative approach? Right now it is hard to recommend the publication in remote sensing because the paper is focused exclusively on the technique, with error analysis, but has no examples? The authors have done a tremendous job with their expertise in comparing the different techniques, but they need to give an example of it’s application.  Once this is done I think the paper should be resubmitted.

Authors: We agree with the Reviewer’s concern and we revised the ms accordingly. 

In the new version of the ms two real cases that refer to the L-band SMAP and the multi-channel SSM-I microwave radiometer are shown to demonstrate the ability of the reconstruction methods to generate an added-value product whose spatial grid is finer than the native radiometer one. On the other side, these two showcases point out the importance of generating microwave radiometer products on a same spatial grid.

 

 

Minor: Additionally the paper needs a thorough proofreading as it contains multiple errors.

Authors: Thanks for pointing out this issue. We performed careful proofreading and corrected several grammar and spelling errors.

 

Overall I think the work could lead to a huge breakthrough in the application of microwave data, but needs to take things one step further.

Authors: Thanks.

 

Reviewer 2 Report

Research on global warming and climate change has gained a critical importance in many scientific fields. In this context, the enhancement of the spatial resolution of microwave radiometer is addressed by contrasting the accuracy of a gradient-like antenna pattern deconvolution method with its accelerated versions. The analysis points out that accelerated methods result in improved performance when dealing with spot-like discontinuities; while they perform in a similar way to the canonical gradient method in case of large discontinuities.

The main comments are as follows.

1. In the work, the authors mentioned "regional-scale climate studies" for several times. However, there is not any content about the work.

2. It should be a good work to enhance the spatial resolution of the microwave radiometer data, but it is hard to understand the work by the figures provided in the paper. I suggested to introduce the work by using the obtained microwave radiometer data. Actually,  the enhancement of the spatial resolution of microwave radiometer data is important to apply the data.

3. In conclusion, I don't know how to obtain the conclusion "Experimental results show that a climate added value product represented by enhanced spatial-resolution microwave radiometer measurements can be achieved by means of thw ILWand LW-P reconstruction methods."

4. There are multiple grammatic or expression errors, e.g.

(1)Line 1: Change “the the” to “the”

(2)Line 50: The expression is problematic.

(3)Line 217: Change "thw" to "the".

Please carefully check the paper.

Author Response

Authors: We thank the AE and the reviewers for his/her useful comments/suggestions that have been incorporated in the new version of the manuscript. All the suggested major adjustments are marked in red in the new version of the manuscript.

Research on global warming and climate change has gained a critical importance in many scientific fields. In this context, the enhancement of the spatial resolution of microwave radiometer is addressed by contrasting the accuracy of a gradient-like antenna pattern deconvolution method with its accelerated versions. The analysis points out that accelerated methods result in improved performance when dealing with spot-like discontinuities; while they perform in a similar way to the canonical gradient method in case of large discontinuities.

The main comments are as follows.

1. In the work, the authors mentioned "regional-scale climate studies" for several times. However, there is not any content about the work.
2. It should be a good work to enhance the spatial resolution of the microwave radiometer data, but it is hard to understand the work by the figures provided in the paper. I suggested to introduce the work by using the obtained microwave radiometer data. Actually,  the enhancement of the spatial resolution of microwave radiometer data is important to apply the data.

Authors: Space-borne microwave radiometers (MWRs) are at the forefront of climate applications of satellite data, but due to their intrinsic limited spatial resolution the products derived from their measurements are limited to the global/large-scale applications. However  there are several climate-related regional applications which could benefit from a super-resolution MWR product, such as the complex spatial pattern of a sea-ice scenario or coastal area management (where sea/land mixed pixels could be present in low-resolution measurements). 

All this matter is discussed in the new version of the ms where two real cases that refer to the L-band SMAP and the multi-channel SSM-I microwave radiometer are shown to demonstrate the ability of the reconstruction methods to generate an added-value product whose spatial grid is finer than the native radiometer one. 

 

3. In conclusion, I don't know how to obtain the conclusion "Experimental results show that a climate added value product represented by enhanced spatial-resolution microwave radiometer measurements can be achieved by means of the ILW and LW-P reconstruction methods."

Authors: We agree with this Reviewer and we modified the new version of the ms by: a)changing the conclusions section by first discussing the outcomes of the accelerated methods and then providing some hints related to one of the possible applications of these accelerated methods, namely the enhancement of the spatial resolution of long time series of measurements that arise from climatological studies. On the other side, we also introduced better the problem by providing a fist subsection in the Experiment section that is devoted to discuss the benefits of spatial resolution methods based on the gradient iterative algorithm.

 

4. There are multiple grammatic or expression errors, e.g.

(1)Line 1: Change “the the” to “the”

(2)Line 50: The expression is problematic.

(3)Line 217: Change "thw" to "the".

Please carefully check the paper.

Authors: Thanks for pointing out the errors that have been adjusted in the new version of the manuscript. We also performed careful proofreading correcting several other errors.

 

Reviewer 3 Report

Dear Authors,

A summary of my revisions to the received manuscript is attached below.

Comments for author File: Comments.pdf

Author Response

Authors: We thank the AE and the reviewers for his/her useful comments/suggestions that have been incorporated in the new version of the manuscript. All the suggested major adjustments are marked in red in the new version of the manuscript.

1. In lines 143-162, the procedure of the first experiment is described. Some places may not be very clear:

(1) In lines 144-145: “Hence, a non-realistic reference profile (RP) that consists of 1- pixel Kronecker function calling for a 106 K brightness temperature is used.” what is 1- pixel Kronecker function? Why did you choose such a parameter setting?

Authors: The goal of this experiment is to test in a simple and effective way the center-swath resolution enhancement performance by analyzing the system spreading function. We need to approximate the Dirac delta function (the Kroneker discrete one) and that is why we decided to use an “un-realistic” reference field which consists of a Kronecker Delta Function, i.e., function with value 1 only in the central pixel of the SMAP swath and 0 otherwise. The brightness temperature level (TB = 106 K) has been chosen large enough to allow a fairly good signal level in the measured field after the low-pass MWR system function filtering.

(2) In lines 152-153: “Note that the RP (shown as cyan line) is interpolated onto the finer spatial grid using the conventional linear interpolation.” What is the relationship between RP and other methods?

Authors: The acronym RP stands for Reference Profile, i.e., the signal used as a reference in the direct problem A x = b to obtain the simulated radiometer measurements. Hence, it is the benchmark to test the performances of the methods in simulated case.

 

(3) In lines 154-156: “It can be noted that, although all three methods improve the spatial resolution with respect to the radiometer measurements, LW-P performs best since it results in a brightness level which is the closer to the RP one.” How to conclude that LW-P is closer to RP according Figure 2(a)?

Authors: Thanks for commenting on this point that needs to be clarified. The reference profile RP is not shown in Fig. 2 (a) (Fig. 3 (a) in the new version of the manuscript) since it calls for a TB value significantly larger than the measured and reconstructed ones and, therefore, would have hampered the image readability. We added a comment on this in the caption. The LW-P reconstruction is closer to RP simply because it allows reaching a TB level that is the closest to the RP one.

 

2. To display the experimental results more intuitively, in the second and third experiments, the brightness temperature images should be added to compare the resolution enhancement effect of the three methods.

Authors: An additional experiment showing the resolution enhancement capabilities of the gradient iterative methods have been added in the experimental section. All the methods results in the same reconstructions, the only difference is their convergence speed.

 

Round 2

Reviewer 1 Report

I believe the authors have addressed my major concern, but there are still some suggestions. The authors now have Figures 1 and 2 which show the SMAP L -Band before and after the reconstruction. However they are somewhat hard to interpret and it is difficult to see the differences. Additionally the authors need to add units. These are critical images for the paper. Can the authors zoom in on a section to highlight differences? Also title should be "in support of" instead of "in support to".

Author Response

Authors: We thank the AE and the reviewers for his/her useful comments/suggestions that have been incorporated in the new version of the manuscript. All the suggested major adjustments are marked in red in the new version of the manuscript.

I believe the authors have addressed my major concern, but there are still some suggestions. The authors now have Figures 1 and 2 which show the SMAP L -Band before and after the reconstruction. However they are somewhat hard to interpret and it is difficult to see the differences. Additionally the authors need to add units. These are critical images for the paper. Can the authors zoom in on a section to highlight differences? Also title should be "in support of" instead of "in support to".

Authors: Thanks for the suggestion. Two additional boxes have been added to each plot showing an enlarged version of the selected region in order to better appreciate the effects of resolution enhancement. 

With respect to the title, we agree with your comment and we changed the title to “Comparison of accelerated versions of the iterative gradient method to ameliorate the spatial resolution of microwave radiometer products” in order to be more consistent with the content of the paper.

Reviewer 2 Report

The authors gave good responses to my comments. However, there are still some suggestions.

1. The title is not supported by the contents of the paper. From the title, I think the work is to obtain "super-resolution products" in support to "regional-scale climate studies", but the content about the regional-scale climate is little, and the work is not related to the regional-scale climate studies.

2.The design of the presentation should be improved. The authors indicated Figures 1 and 2 which show the SMAP L -Band before and after the reconstruction, which is then used to indicate the potential application in regional-scale climate. However, the relationship between the climate and the spatial resolution of the data is not addressed.

Author Response

Authors: We thank the AE and the reviewers for his/her useful comments/suggestions that have been incorporated in the new version of the manuscript. All the suggested major adjustments are marked in red in the new version of the manuscript.

1. The title is not supported by the contents of the paper. From the title, I think the work is to obtain "super-resolution products" in support to "regional-scale climate studies", but the content about the regional-scale climate is little, and the work is not related to the regional-scale climate studies.

Authors: Thanks for your comment. We changed the title to “Comparison of accelerated versions of the iterative gradient method to ameliorate the spatial resolution of microwave radiometer products” in order to be more consistent with the content of the paper.

 

2. The design of the presentation should be improved. The authors indicated Figures 1 and 2 which show the SMAP L -Band before and after the reconstruction, which is then used to indicate the potential application in regional-scale climate. However, the relationship between the climate and the spatial resolution of the data is not addressed.

Authors: We addressed this issue in the new version of the manuscript where: firstly we changed the title to reflect the real scientific objective of the ms; secondly we provide a better description of the experiments included in Figs. 1 and 2 result in a twofold importance: on one side, they show the ability of LW (and, therefore, of its child accelerated versions) to reconstruct the brightness field at enhanced spatial resolution. This can be observed also by visually contrasting the measured field with the reconstructed one. We also included a zoomed-in version of an area that allows clearly observing the improvement of the spatial resolution in terms of sharper edges and better defined coastlines (see panels c and d in Figs.1 and 2). On the other side, the gradient method allows reconstructing the brightness field on a given spatial grid. This is of paramount importance to construct reliable time-series of measurements collected by microwave radiometers calling for different native spatial resolutions. It is also important to make a data-fusion between different channels of the same multi-frequency microwave radiometer. 

The key issue resulting from the LW reconstructions relies on the processing time that is not so fast due to the low LW convergence rate . This is not a big issue when processing isolated data sets; while it is a very limiting factor in applications, e.g.; climatological studies; when a large time series of measurements is to be processed. Within this context, the two child version of the LW are of paramount importance since they accelerate its convergence rate while keeping inalterated the reconstruction’s accuracy. 

All this matter was detailed in the new version of the ms.