An Uncertainty Quantified Fundamental Climate Data Record for Microwave Humidity Sounders
, Marc Prange 1, Theresa Lang 1 and Viju O. John 2Round 1
Reviewer 1 Report
Formally, the paper is suitable for publication in the Remote Sensing journal.
The topic of the paper is related to the refinement of satellite sensing data (product 1b). The problem of satellite data calibration is very actual. Methods for remote sensing data calibration have been developed since the middle of the last century.
The work is largely the applied one. It is aimed at recording and reducing the errors of satellite data of microwave humidity sounders obtained due to different satellites from 1994 up to present that allows the obtained data to be compared and the fundamental climate data record to be compiled. The topic of the paper is important because product 1b is used for the creation of higher level products.
Remarks
1. In paper, a figure (fig. 1) taken from article [12] is used. Some values from fig. 1 are not explained. This makes it difficult to analyze the material, in particular, formulas (1) and (2), which are the basis of the study. Reference [12] is a technical report (Woolliams, E.; Mittaz, J.; Merchant, CJ; Harris, P. D2.2a: Principles behind the FCDR effects tables. Technical report, National Physical Laboratory and University of Reading, 2017). It is not clear how this source is available to a wide circle of readers.
2. There are 35 references in the work, of which 11 are links to technical reports, 6 are references to own works, including two papers that have not been published yet. How relevant is it to make links to unpublished works?
Comments for author File: 
Comments.pdf
Author Response
Response to Reviewer 1 Comments
We would like to thank you for your valuable comments and suggestions which helped to improve the manuscript. In the new version of the manuscript, we implemented the according changes and highlighted these (the “diff” function did not work for the list of references unfortunately).
Please find detailed answers below.
Point 1: In paper, a figure (fig. 1) taken from article [12] is used. Some values from fig. 1 are not explained. This makes it difficult to analyze the material, in particular, formulas (1) and (2), which are the basis of the study. Reference [12] is a technical report (Woolliams, E.; Mittaz, J.; Merchant, CJ; Harris, P. D2.2a: Principles behind the FCDR effects tables. Technical report, National Physical Laboratory and University of Reading, 2017). It is not clear how this source is available to a wide circle of readers.
Response 1: Figure 1 shows the full measurement equation for MW sounders which is basically repeated in formula (1). You are right; some quantities appearing in Fig.1 were not explained. We changed the figure to match the naming of quantities in formula (1), which are explained in the other paragraphs. Further quantities not explained are now explained in the caption of Fig. 1.
Concerning the reports, both [12] and [21] (now [16] and [28]) are available online (open access) via the FIDUCEO website: fiduceo.eu/deliverables. This note also appears now in the reference list.
Point 2: There are 35 references in the work, of which 11 are links to technical reports, 6 are references to own works, including two papers that have not been published yet. How relevant is it to make links to unpublished works?
Response 2: As the works is a rather technical one relating to calibration of instruments, it is inevitable to refer to technical documents. Some of them are publicly available, such as the NOAA KLM User guide and others. Some of them are available at request from the authors. We believe that we still have to cite the works, although they are not yet published. Otherwise, there is no information on where the results relate to. We see that this looks unsatisfying at the moment, but we hope that we can update the status on these works soon.
Reviewer 2 Report
A very good and well written manuscript. Please refer to my detailed comments in the attached review.
Comments for author File: Comments.pdf
Author Response
Response to Reviewer 2 Comments
We would like to thank you for your valuable comments and suggestions which helped to improve the manuscript. In the new version of the manuscript, we implemented the according changes and highlighted these (the “diff” function did not work for the list of references unfortunately).
Please find detailed answers below.
Point 1: Page 3, line 89: I believe the author’s meant to state that MHS on MetOp-C is the fifth MHS instrument, not the fourth.
Response 1: You are right, of course. Thank you for pointing us to this mistake.
Point 2: Page 7, last paragraph: How large is the presumably pre-launch measured nonlinearity? What were the specifications for the maximum allowed AMSU-B and MHS nonlinearity?
Response 2: The non-linearity coefficient varies for the channels and instruments. To give rough impressions: Channel 1 and 2 have about -200 to -100 cm^-1m^2/W, channel 3-5 of MHS (zero for AMSUB) have about -50 to -10 cm^-1m^2/W, though Noaa19-MHS gets positive values (note also that we used W as unit here, not mW). The actual specifications and traceable derivation of the nonlinearity coefficients are not known to us. In Saunders1995, [20] in the manuscript, it was shown that only channel 1 showed a deviation from linearity that should be corrected for.
It is definitely desirable to have traceable pre-launch information in future.
Point 3: Page 8, APC section: The authors accurately discuss the issue of the antenna pattern and the effect of radiation from the satellite and from space on the side lobes, but they don’t mention the potential impact of uncertainties in the pre-launch measured antenna pattern. The fact that they had to correct for erroneous values in several of the channels raises questions regarding the residual uncertainties in the measured values and their impact on the brightness temperatures. Although it is a conically-scanning radiometer, errors in the pre-launch measured antenna patterns for the GPM GMI instrument were found to be by far the largest source of error in the brightness temperatures for similar channels (i.e. 89 to 183 GHz), which were subsequently refined on orbit [Wentz and Draper, 2016, J. Atmos. Oceanic Technol.]. As the authors note, this information isn’t even available for SSMT-2, however, even for AMSU-B and MHS measuring the antenna pattern in a lab is difficult. What are the beam efficiencies for these channels, since the lower the beam efficiency, the larger the impact of uncertainties in the antenna pattern are on the computed brightness temperatures. I believe the beam efficiencies are in the range from 95 to 99% or so, but it would be useful if the measured values were included in Table 2.
Response 3: The inaccurate antenna pattern correction includes both aspects of the antenna pattern itself and the geometry of the surroundings, you are right. We now stress more the possibility of both being inaccurate which leads to uncertainties, especially for lower beam efficiencies (section 2.2.1 paragraph on antenna pattern). We also cite reference [14] for the measured AMSUB antenna efficiencies. We did not include the values in table 2, though, since the structured of the table does not fit (no individual instruments) and we cannot provide all relevant values. Hence, we give the reference to [14] which deals with the antenna pattern.
Point 4: Page 8, Band correction: It is a bit unclear what the band correction factor coefficients do within the AAPP. Do they account at all for the spectral response within the pass bands, both for the single and dual passband channels? Near the 183.31 GHz water vapor line, non-linearity in the atmospheric transmittance will lead to a bias between the observed Tb (1 GHz passband width according to Table 2) and the simulated Tb if only 182.31 and 184.31 GHz central frequency bands are used in the calculated Tb. How is this handled and does it assume a boxcar shape (i.e. center frequency plus/minus half of the passband width), or does either AMSU-B and/or MHS have pre-launch measured spectral response functions similar to those from ATMS (https://www.star.nesdis.noaa.gov/jpss/ATMS.php)? Accurate knowledge of the spectral
response is important not only for the bias correction etc., but also presumably for CDR developers using the FCDR data.
Response 4: The band correction coefficients of AAPP are a parametrisation to account for the deviation that emerges when computing the Planck function at the frequency of 183 GHz compared to the average of two Planck functions evaluated at center of the lower and upper passband (the latter is closer to what the instrument does). The spectral response function (SRF) of the instrument is not involved here. The SRF is not important for the photometric calibration, since both the target (earth) and the calibration targets (IWCT and space) are recorded with the same spectral response function. But you are right, for higher level products, or, as soon as radiative transfer models are used, the SRF is important. It was shown, however, that the impact is below 0.1K if using a box-car function compared to a Gaussian for example [John, Analysis of upper tropospheric humidity measurements by microwave sounders and radiosondes 2005, Dissertation, http://nbn-resolving.de/urn:nbn:de:gbv:46-diss000013503]. Of course, using the true SRFs is desirables for such applications. In our new FIDUCEO FCDR we can provide the SRF for MHS on NOAA18 and NOAA19, but for no further instruments, unfortunately.
Point 5: Page 9, CMB: Doesn’t the Planck equivalent temperature for cold space vary with frequency [Wentz and Draper, 2016, Table 1; Microwave Radiometry by Michael Janssen, 1993]? I believe this effect is significant for the frequencies considered here.
Response 5: The temperature of the CMB itself is not depending on frequency. However, you are right in so far that the linear relation of brightness temperature and radiance from the Rayleigh-Jeans approximation is not valid at the frequencies of the MW instruments. That is why Wentz and Draper provide a ”Planck equivalent temperature” of the CMB in their study (as modification to the actual CMB). This temperature is then depending on the frequency and hence accounting for the failure of the Rayleigh-Jeans-law at these frequencies. In our study, we use the Planck function itself, not Rayleigh-Jeans.
