Assessment of the “Zero-Bias Line” Homogenization Method for Microwave Radiometers Using Sentinel-3A and Sentinel-3B Tandem Phase
, Ralf Bennartz 2,3, Frank Fell 4, Marie-Laure Frery 5, Mathilde Siméon 5 and Frank Bordes 6Round 1
Reviewer 1 Report
The authors revised the paper according to the points raised during the first review cycle, therefore the paper can be published in its present form.
Reviewer 2 Report
Authors have modified the manuscript following reviewer suggestions.
Reviewer 3 Report
The authors have addressed satisfactorily all the concerns stated in my previous revision. I have no more comments. I think it is a good paper and should be published
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The paper presents an evaluation of homogenization and harmonization processes followed for altimeters on-board EO satellites. It follows well-established methodologies, but is applied in a new and unique case, i.e., during the tandem mission of S3A and S3B, which provide a unique outlook of the two missions being separated by a very short time lapse.
I would like to congratulate the authors on submitting a very well written manuscript, which is clear in its general scope presentation, methodology, experiment conducted and results achieved. Being working for more than 25 years on altimetry, it was a real pleasure to review the paper. I only have some minor points that the authors need to address, which I think will improve the final version paper.
line 33: ...on the quality of the GSMSL.
line 39: Why split the paragraph here?
line 42: ...different signals...
line 44: Please give some references here on the historical aspect of the MWR harmonization for past missions.
line 54: ...to each instrument.
line 60: ...to be applied to the TB at the input based on a....
line 62: As illustrated...
line 63: Maybe here write both TBhA and TBhB instead of TBh. Or you can write TBhi to signal that it can be some MWR from a selection of missions.
line 65: ...of the instrument sensitivity.
line 77: So, differences existed(?)...
line 84: ...of a homogenization method in the lack of a harmonization process.
Line 87: Why split the paragraph here?
Line 92: Why split the paragraph here?
line 117: ...only within ±60o in latitude...
lines 141, 144, 146: Same here. I think that the 1-sentence paragraphs should be merged.
line 141: The beginning of the sentence is not very clear. Maybe something like "The operational algorithms used for satellites belonging to the same series of instruments (like Jason), actually assume already ....."
line 165: The parameters of Eq. 2 should be described, i.e., say that ai are the polynomial unknown coefficients and u10m is the wind speed at 10m.
Lines 175, 177: Why split the paragraph here?
Line 183: Omit "The" at the beginning.
line 187: When the authors mention some reasons in the processing, what is meant exactly? I do not mean to provide in full-details the analysis, but to pin point the reasons why the processing has failed and the orbits have not been included. This is useful information of the readers and adds to the repetition of the results.
Figure 4: The way the results are presented in this Figure is a little bit confusing as it has too much information. First, the background of the image is blurring with the homogenized 2-p scenario. Maybe a white background would be better. Then, the authors mention a variability on the 2-p on the 23.8 GHz channel (line 196) as being equal to 0.8K, but in the text they mention (same line) h-1p.
line 199. Omit "the" before Figure 5.
line 205: ...also varies....
line 208: ...a window channel is less...
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors propose a homogenization approach based on a polynomial correction of the brightness temperature TB applied to the baseline values collected by Sentinel-3A and Sentinel-3B, which are an ideal configuration for this kind of process.
This simplified method let the authors perform a "zero-bais line" approach in a simplified form that leads to a reduced bias and to a refinement of the variability.
Unfortunately, this method, if applied to data used as input to L2 products results in worse results.
That is why the authors suggest to apply this algorithm to build a reference data record.
The approach gives pretty goos results and it help in usage of data acquired by sensors that are no more active.
The manuscript is well structured and written. It would be appreciate a brief discussion about the possible improvement that the simplified method offer to remote sensing scientists.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Revision of manuscript "ASSESSMENT OF THE "ZERO-BIAS LINE" HOMOGENIZATION METHOD FOR MICROWAVE RADIOMETERS USING SENTINEL-3A AND SENTINEL-3B TANDEM PHASE" by Bruno Picard et. al.
The paper is about intercalibration of microwave radiometers. A previously published approach (so-called "zero-bias line") is applied to Sentinel 3A and 3B radiometers with good results in terms of reduction of the differences between both. The paper is somewhat obscure and unclear in the definitions of methods used. For example, it introduces concepts as "harmonization" and "homogenization", but they are not directly used later. In the end, the correction proposed is equivalent to adjusting the gain and offset of two radiometers in order to make them measure closer TB for the same target. They try to introduce the wind speed on the correction but with worse results.
Some specific comments follow:
line 41: replace "observed" by "observe"
line 48 and figure 2: There must also be an offset in what you call "radiometric model": f(G,O,CA)
line 79: acronym CLS not defined
line 86: What is the "baseline TB"? According to figure 2, the radiometric model is applied to counts to get harmonized TB's (TBh). There is not a "baseline TB" between counts and TBh. Do you mean that homogenization is applied to counts?
line 87: Something wrong with the sentence "Applying the homogenization to the harmonized TB could appear pointless". Why is it pointless? It is exactly what is proposed in Figure 2.
line 89: How could be possible that the "radiometric model" is not available?. A radiometer is a useless instrument without it. It might be very simple (TBh=G*CA+O) but it must exist.
line 92: Why you start by "For instance"?. The rest of the paragraph deals with the use of both harmonization and homogenization (as in Fig 2) whereas tho two previous paragraph suggest using only homogenization. I guess this paragraph is just introducing the next section. If this is the case, better include it at the beginning of next section.
Line 137: It is dangerous to adjust an instrument by minimizing the difference between observed and simulated TB. Risk of "validating" a wrong model. (as a matter of fact, the same can be said for the three bullets: all methods minimize observed-simulated TB)
Line 140: It is weird to define "transfer function" between simulation and observation. It is normally applied to the ratio between input and output of a given system, either a simulation or a measurement. But mixing them is strange: There is no system having simulation as input and observation as output (or the other way round)
Line 141: If "instruments actually assumed already homogenized observations", what exactly are you doing?. I thought you were proposing a homogenization algorithm.
line 144: What step?.
line 163: "... does not necessarily need to separate the two". I do not agree. I think these two sources of errors are distinct and must always be considered separately. When comparing two different instruments, the errors of the second bullet can be made equal for both. Changing both the instrument and the forward model at the same time may be a mess.
(Line 165) Equation (1): If I understand well, this equation is not consistent with the naming convention of Figure 2. It should be: T_B^H=T_B^h-\delta T_B
Equation (2): symbols are not defined.
Still about equation (2): Is TB the same as in equation (1)?. Then should be T_B^h. Furthermore, in this case you could rewrite equation (1) as a standard slope/intercept formula: T_B^H= A*T_B^h+ B with A=1-a_1 and B=-a_0-a_2u_10m-a_3 u_10m^2
line 191: I assume that parameters of table 2 (using to get these results) are not derived from the same dataset used in figure 4. It says nothing about it in the paper. Could you be more specific in specifying how the parameters are computed (besides giving just a reference)?
Author Response
Please see the attachment.
