Review of: Performance Evaluation of THz Atmospheric Limb Sounder (TALIS) of China. by Wang and colleagues for publication in AMT
General comments
----------------
This paper describes the theoretical performance of a potential future microwave limb sounding atmospheric sensor, the spaceborne "TALIS" instrument. TALIS is modeled after previous limb sounding submillimeter/microwave instruments, measuring many of the same species with similar performance.
Overall I think the authors' approach is sound, and this paper will ultimately appear in AMT and provide useful information to the community. However, there are some omissions and points I think the authors should expand upon before it is ready to be published. None of these are particularly challenging.
Also the English, while clear in most places, is rather haphazzard. I urge the authors and/or the journal to employ the services of an English speaking copy editor prior to publication. I have identified a few places where wording improvement is needed in my specific comments below (in some cases suggesting changes). However, this is far from being an exhaustive list of such cases.
As I say, overall I am happy with the approach. My detailed comments are given below. Of these I think the most important are related to a need to give more information on the planned instrument design, and to better explain and explore the reasons for the differences between the TALIS and MLS performance, which I currently find hard to square with my understanding of the two instruments.
Specific comments
-----------------
---- Page 1
Line 8/9: "high precision" means different things to different people (e.g., an airborne in situ instrument would most likely have way higher precision than anything remotely sounded). I'd simply changed it to something like "vertically resolved profile observations"
Line 11: The measurements clearly do not extend to the surface. I'd say "~10-100km, or "the upper troposphere to the mesosphere" or something like that.
Line 20: "or" -> "and/or"
Line 22: Delete "the"
Line 23: I suggest "...studies. Satellites can provide dailhy global coverage of the atmosphere. Instruments such as nadir microwave and infrared sounders have been applied..."
---- Page 2
Line 5, add a citation to Barath et al., 1993, doi:10.1029/93jd00798. Also to Waters et al., 1999, doi:10.1175/1520-0469(1999)056<0194:tuaeml>2.0.co;2
---- Page 3
Line 6: Please expand on the "higher" precision. It's not clear to me how the TALIS optical layout specifically enables that. Firstly, to be clear, you do mean "precision" (i.e., random noise-type errors) versus "accuracy" (more "systematic bias" type errors), correct? How does the optical layout improve that? Is it because there are fewer beam splitters etc. required? Also, please use terms like "better" and "worse" rather than "higher" and "lower" for terms like precision, accuracy and resolution. This is because when the words are used, having "more" sounds "better". However, when these terms are referred to quantitatively (e.g., with a noise level), then "more" sounds "worse".
Line 7: Please expand on the 20km displacement. Is this in azimuth (i.e., across the line-of-sight) or in elevetion (i.e., vertically) or some combination of the two? How does this modify your statement that the instrument scans from 0-100km, is it 0-100km for some receivers and 20-120 for others for example?
Line 10: Please expand on the calibration and instrument design overall. How do the antenna feeds in figure 2 "view the hot target"? Is the box at the end of the arm tipped up to look directly at the target? Does the target fold down to cover the feed horns?
Figure 1: Please give more detail on the sketch or in the caption. Which parts of the instrument move to accomplish the scan? Is it just the antenna, just the feedhorn assembly, or both in concert. Alternatively, does the whole spacrcraft nod up and down?
Table 1: Some of the numbers in the table (and in the discussion that follows) appear contradictory. If the instrument scans from 0-100km in 36 seconds, that's a rate of 2.7km per second. However, the integration time is quoted as "0.1s (1km)". Whatever the case, I've been assuming that all the distances quoted are "vertical", but some may think you're referring to the "along-track" distances (i.e., horizontal). It would be good to be clear on this.
---- Page 5
Line 6: It is not made clear to me (here or elsewhere) how the "broader bandwidth and finer resolution" is what leads to the improved performance over MLS. I grant that broad bandwidth counts in the lower parts of the atmosphere, but a lot of your improvements come in the mid-stratosphere where the lines are only a 10-100 MHz wide, so bandwidth is not an issue (unless you are bringing in additional lines, which works for ozone, which has many lines, but not water vapor). Are you sure you're capturing the MLS characteristics correctly? As I understand it, the MLS 25-channel spectrometers cover their full bandwidth - as the channels become more widely separated away from the line centers, the channels also get wider increasing the signal to noise. There are no "gaps" in the spectral coverage, which is what the asterisks shown in Figure 3 seem to imply (it might be best to replace the asterisks in Figure 3 with horzontal lines of the correct channel-width, or something similar to make
that point). Table 2 and Figure 7 of the Waters et al. 2006 reference shows the MLS configuration, please be sure you're capturing it correctly. I suggest you contact the MLS team if you are in any doubt (they are typically fairly responsive).
Further, you may have omitted the MLS "Digital Autocorrelator Spectrometers" that I believe provide very fine spectral resolution in the line centers for some species. These contribute to improved performance for MLS in the upper altitudes.
---- Pages 6-9, Figures 4-7.
It would be really helpful if the colors for each molecule could be consistent from band to band, at least for the main ones that are common to all many plots (O3, H2O, HNO3).
---- Page 9
Line 13: Poor wording: "Scattering can usually be negelected above the upper troposphere as the atmosphere is largely cloud-free at these altitudes, and such clouds as there are (e.g., Polar Stratospheric Clouds) have particle sizes shorter than the TALIS observation wavelengths.
---- Page 10
Line 6: I don't think that's what the definition of "spectroscopy" is. How about "Spectroscopy models and databases allow us to compute the absorption coefficient..."
Line 15: Delete "needs" and change "convert" to "converts"
Line 16: Change to "... intermediate frequency, folding the upper and lower sideband signals together in consequence".
---- Page 11
Line 9: I'm not sure about the sqrt(2). Surely it is only needed if the Tsys is defined in terms of single sideband for a double sideband receiver.
Line 30, Equation 14. Which of the covariance matrices are shown summarized as "Error" in the figures that follow? Looking at the plots I actually think it's the total of Eq 14 and Eq 15, but I'm not sure.
---- Page 12
Line 10: I'm worried about the authors' choice to ignore the finite size of the field of view in the discussion that follows. It really is no additional hardship with ARTS to add that step, and would make for a far more meaningful study. If nothing else, I urge them to quantify the impact it has, which will not be on the precision so much as the vertical resolution.
Line 14: Are they refering to the vertical resolution of the radiance obsevations or of the state vector vertical grid here? In either case, why is the FOV for TALIS or MLS germain to that choice? The tradeoff they are referring to with the Livesey and Snyder paper is not really related to the resolution of the reporting grid, rather to the resolution of the information yield (e.g., as represented by the width of the averaging kernels). Also, if the spectra are obtained avery 2.5km what is the integration time assumed, and the total time taken for the vertical scan. Again, as with the numbers in Table 1, there is an inconsistency here. Fundamentally there are multiple "resolutions" at play here: The spacing of the retrieval grid, the spacing of the radiances, and the degree of information content in the Level 2 data products. Please try to be clear which you are referring to in each case.
Paragraph starting at line 14, and through to end of 4.1: I think more information is needed in this section. Firstly, from where did the various atmospheric profiles assumed as "truth" and "apriori" come from? It just says "mid-latitude summer conditions". Are MLS data the source of the profiles? If so, please be explicit about the version of MLS data used, latitude/date range used, application of quality screening, assmptions made to go from pressure to altitude as the vertical coordiante etc. What about the species that MLS doesn't measure, ro doesn't measure well? If the profiles instead come from a model, again give all the information needed to suppor reproducibility. Secondly, in the retrieval calcaultions that follow, are the various molecules retrieved simultaneously from all the bands together, or are some subset (which) of the molecules retrieved independently from each band? Are there also spectrally flat (or simple linear or quadratic form) "Baseline" or "Extinction" terms included in the retrievals. These are needed to account for instrumental/forward model issues, and can impact the precision/resolution trade.
---- Page 13
Line 3: Again, where does the "typical profile" come from
Line 4: In my experience the Gaussian form is a poor choice for the off-diagnoal terms, as it gives a normal equation matrix that is close to singularity. Instead an exponential form is generally preferred (and corresponds to a 1st-order Tikhanov smoothing).
Line 10: How much averaging is the factor of 10 equivalent to? Presumably 100x as many measurements, correct? Please put that in terms a reader will readily understand (e.g., "equivalent to a 10-degree latitude weekly zonal mean" or whatever it is).
Line 16: As stated above, I find it hard to believe the spectral resolution and bandwidth alone accounts for the MLS/TELIS differences shown, especially above ~20-30km. Are the authors sure they have correctly captured the MLS spectral coverage (see above)? The effective bandwidth (i.e., the width of the spectral regions that actually contain information) are essentially the same between the two instruments. Are the authors using the same system temperatures for both instruments (as strongly implied by line 16) or are they in fact changing those numbers too, if so that's the likely cause. Note the discussion of sqrt(2) above, I'm pretty sure the MLS team quotes thir system temperatures without invoking that factor. This is an important issue that warrants greater investigation.
Figure 9 (and others): Please make it clear what "error" means in this context. Is it the sqrt-diagonal of one of the covariance matrices quoted above (or the sum of several)? Please be clear. I suggest the authors add some clear metric of vertical resolution to the right-hand figure (e.g., full width at half maximum, or degrees of freedom per km), with a separate top x-axis.
---- Page 15
Line 1: Again, which covariance matrix!
Figure 12 and its siblings: I really don't see the point of showing a single retrieved noisy profile here, it's but one representative retrieval exhibiting noise and risks giving the uneducated reader the impression of some kind of systematic biases that cannot be reduced. Related to that, is the "Error" term again the sqrt of some covariance diagonal, or is it the difference between retrieved and true (I don't think so, looking at the results, but some readers may get confused). I'd just simply show the retrieved precision, and the typical profile (combining the first two panels into one with only two lines).
---- Page 16
Line 8: "less" -> "poorer"
---- Page 24
Line 5: As discussed, it's essentially impossible for the reader to "see" the 3km resolution from your figures. Please add some suitable line/plot to make that clerer.
Line 10: "Significantly average" -> "Significant averaging"
---Page 29
Line 5: Change first sentence to: "Seven species show high sensitivity, sufficient for scientifically useful single profile retrievals".
Line 23/24: Just a note, the Zeeman effect will indeed be critical to the correct retrieval of mesospheric temperature, though it will probably not impact the precision/resolution results shown here. |