|Based on the replies to reviewers and the revised version, here’s how I see the main questions on which a publication decision should be based:|
Is there an application to which the approach demonstrated in this paper can be applied?
The answer is arguably ‘yes’.
Let’s say FORUM has been launched, and a study is produced that shows that its far-IR radiances can be predicted reasonably accurately by FORUM radiances measured in other regions. That demonstration will suggest that a similar approach would work using the large historical record of IR observations (e.g. AIRS, IASI), thereby allowing IR measurements over decades to be “extended” to the potent far-IR. Therefore, if this paper had been able to use FORUM measurements instead of ground-based REFIR-PAD measurements, then the answer would clearly be ‘yes’. So the argument for the paper under review being publishable hinges on whether the method demonstrated for the REFIR-PAD extension from the IR to FIR is germane to a similar possible future extension for satellite instruments.
Note that the application of the method in this paper to other ground-based instruments is not worthy of publication; the authors seem to agree with this point. There are only a couple of ground-based spectral IR instruments (i.e. does not measure in the far-IR, e.g. AERI) deployed in locations in which the far-IR is not always opaque, so there would be minimal need for a method to extend these data records.
How similar are the ground-based and satellite-based situations?
To see how germane the REFIR-PAD extension presented in this paper would be to an extension for a satellite-based instrument, the similarities and differences between the two different viewing geometries with respect to the relationship between far-IR and IR radiances must be explored.
- What would the satellite see? In the far-IR, radiances would depend on the water vapor and temperature profiles – a very rough rule of thumb is the observed brightness temperature in an instrument channel would be the temperature at the height at which the integrated optical depth (primarily due to water vapor in this spectral region) from the top of the atmosphere to that height is about 1. There would clearly be similar channels in the nu2 band of water vapor, so there is every reason to expect that a good extension to the far-IR from the nu2 band could be developed. The nu2 band clearly has all information about the tropospheric water vapor field (since it is used for water vapor retrievals), and could also presumably be used for temperature retrievals. There are two important caveats to this. First, since the far-IR water vapor band is stronger than the nu2 band, the dependences on water vapor concentrations higher up (e.g. stratosphere) than the vertical region that the nu2 band is sensitive to might impose some limitations to the extension. Second, radiances in spectral channels in carbon dioxide bands would have little correlation with the far-IR radiances and wouldn’t be used in a single-channel satellite-based extension. (This paper considers only single-channel extensions.)
- What does the ground-based REFIR-PAD see in a location like Antarctica? Qualitatively, there are three types of far-IR channels (x-axis on Figure 2):
1) Purely opaque channels – These are everywhere < 200 cm-1 and where there are strong water vapor lines throughout the rest of the far-IR. Radiances in these channels are sensitive only to the temperature very near the instrument. These cases are not similar to any satellite-based channels.
2) Mostly opaque channels – Microwindow regions from 200-400 cm-1 and near some relatively strong lines from 400-600 cm-1. Radiances are sensitive to the temperature and water vapor profiles, in particular those values closer to the surface. This category is somewhat similar to satellite-based channels, although the vertical range of the profile that matters is probably somewhat smaller than for corresponding satellite channels.
3) Semi-transparent channels – Everywhere in the 400-600 cm-1 that is not near a relatively strong line. Radiances are sensitive to the water vapor and temperature profiles, with the sensitivity ranging higher than in the category above (possibly including the water vapor column). This category is pretty similar to corresponding satellite channels.
The answer to the question in this section (“How similar are the ground-based and satellite-based situations?“) is “in theory, partly similar”.
In actuality, how much of an analogue to a potential satellite extension is the ground-based extension presented in this paper?
Since the extension for the opaque channels have no dependence on water vapor, they have no analogue in the satellite case and, therefore, the results shown in the paper are not germane to the satellite case. (For the surface case, any opaque channel, whether co2- or h2o-dominated, will be able to predict the radiance.) The ‘mostly opaque’ channels do have analogues in the satellite case -- there are regions of the y-axis of Figure 2 that have similar optical depth dependences (e.g. in 1300-1400 cm-1), which would be the case for both the ground-based and satellite-based perspectives. So channels in this category have the potential to be good analogues for the satellite case, and are potentially germane to the main question. However, the high noise of the instrument from 1300-1400 cm-1 makes this region not sufficiently predictive for the corresponding far-IR points, which therefore get “matched” with spectral points in the CO2 band, i.e. with virtually no sensitivity to water vapor. This results in reasonable accuracy for the extension, but that is irrelevant to the question of whether this result is germane to the more important question at hand. Using a temperature channel from a satellite instrument to predict far-IR radiances from that instrument clearly wouldn’t have the proper sensitivity, so the results from this category are clearly not germane. The third category (semi-transparent) is most similar to the 1300-1400 cm-1 and 760-800 cm-1 regions on the y-axis. Again, the 1300-1400 cm-1 does not work well due to noise, so the best match is indeed from spectral channels in a region with similar dependences on water vapor and temperature. If this were a satellite-based exercise, I would expect these same far-IR channels to also be fairly well modeled by the same 760-800 cm-1 channels.
The answer to the question posed in this section is “a very limited analogue”.
Is the study germane enough to the satellite case to justify publication?
The entire argument that the extension to the far-IR shown in this paper is germane to the satellite case rests on the somewhat limited number of channels from ~500-620 cm-1 that are ‘matched’ with IR channels from 760-800 cm-1. That is a very limited result.
The answer is ‘no’.
For the sake of argument, assuming that this result is sufficiently germane, is the methodology presented something that someone might consider using for a satellite-based extension to the far-IR?
This study determines a single IR channel to match each FIR channel. The arguments presented above about which region has channels that best match categories of FIR channels are very qualitative. In actuality, a channel in one region will not perfectly match the dependences on water vapor and temperature in a different region. In Huang et al., a multivariate fit of IR measurements is used to simulate the far-IR region. In the generalized training for OSS, a number of monochromatic calculations in different spectral regions are needed to match channel radiances. If one were developing an extension of (say) IASI to the far-IR, it would be limiting and foolish to use a single channel.
The answer to this question is ‘no’.
My perspective on this paper has not changed since the first review, in which I wrote that this “paper suffers from significant motivational and methodological issues.” I do not think it should be published.