Radiative closure tests of collocated hyperspectral microwave and

infrared radiometers, by Lei Liu et al.

This is an interesting work which describes a radiative closure experiment in clear sky conditions performed by using collocated measurements collected with the HiSRAMS radiometers and the AERI Fourier transform spectroradiometer. Measurements were performed during three field campaigns: in 2021 and 2022 from ground, while in 2023 both from airborne and from ground. HiSRAMS is composed of two FTS radiometers operating in the oxygen band between 49 – 58 GHz and the water vapor band between 176 – 183 GHz; AERI is a well known Fourier spectroradiometer operating in the far and mid infrared portion of the Earth’s emission spectrum, between 500 and 1800 cm -1 . The observations took place in zenith-pointing view with AERI from ground, and both in zenith- and nadir-pointing view with HiSRAMS. Simultaneous collocated radiosoundings were launched during the three campaigns to measure the vertical profiles of

temperature and water vapor. Simulations were performed by the authors to mimic the measurements of HiSMRAMS radiometers by using a validate code (Bliankinshtein et al. 2019) that uses the Rosenkranz gas absorption parameterization, and the well known code LBLRTM (Claughet al. 2005) to mimic the AERI measurements. To estimate possible biases, the differences between the measurements and simulations were calculated with the associate standard deviations by propagating, respectively, the spectral measurements uncertainties and the errors on the radiosondes, accounting also for the spatial inhomogeneity. The closure is well reached in the

spectral bands below 800 cm -1 and above 1200 cm -1 simulating the AERI observations and in the weak absorption bands of Oxygen (50-54 GHz) and Water Vapor (176-181 GHz), even though for water vapor the differences remains inside the 3-σ error. The bias with AERI comparison is attributed to the presence of very optically thin cirrus clouds or aerosols, the latter is attributed to

the calibration procedure in zenith-pointing observations.

The paper is well structured and written, even though I suggest few corrections to enhance the clarity:

1) Lines 92-94. I would change the sentences in : “The small fluctuations in the temperature and water vapor profiles have a negligible effect in AERI and HiSRAMS detected radiances”.

2) Lines 95-99. I would improve the discussion about the temperature inversions shown in the inset of Fig. 2a, if the authors want to mentioned it, it is fine but I think the description should be more accurate. For instance, from FC2021 (blu profiles) to me seems there are more than two temperature inversion, at least one at around 2.5 km (T starts to increase again) stronger than the one at 1.2 km. Also, for FC2022 I can see at least two inversions, one at 0.5 km as pointed out, but also one at 2.5 km, similar to 2021, etc.. Maybe, if you could increase the grid vertical resolution on the y axes in both figures 2a and 2b would be helpful.

3) Line 119, “against” → “pointing”

4) I think it would be clearer indicating the actual radiance unit in all figures for AERI, I assume they are mW/(m2 sr cm-1), but it would be clearer if it would specified, in particular for me it is more helpful when I have to quantify the biases.

5) Section 2.2.1. Just a curiosity, I was wandering why you did not used the latest version 12.15 of LBLRTM, with the updated continuum 4.1 version?

6) Line 103, please write the coordinates in the standard form such as: [45.32° N, 75.66° W]7) Line 112, “..200 level are inputs..” → “..200 levels are provided in inputs..”

8) Line 220, “errors” → “those” , “..inputs..”, → “.. input profiles..”

9) Equations (1) and (2), I think it would be clearer if you could replace the x parameter with Rν to indicate the vector of the radiances or brightness temperatures. Also, please shift the x → Rν=Radiance of BT to the right side. Since the uncertainty of the measurements play a key role in the study and discussion of a radiative closure experiment, I suggest to explain in detail the components of the errors, for instance, from lines 220-222 I assume that the standard deviation on the model is obtained by summing in quadrature the 1-σ error on the radiosondes profiles due to the instrumental error and that one due to spatial variability, is it so? In this is the case it would be really helpfull to follow if the authors could write the formulas for the errors propagation used.

10) I suggest to indicate the average biases arising for both instruments.

11) Figure 9, I suggest to exchange figure 9c with 9b and viceversa to be coherent with the next figure 8. Also it would be really helpfull to indicate O 2 and WV band at the top of the two columns and zenit and nadir views horizontally.

12) Line 301, do not indicate Fig. 10a because I think is misleading here.

13) Line 313, “, the measurements uncertainty ..” → “, both the contribution of the simulation and measurement uncertainty is not negligible..”

14) In Figure 12, please add the correlation coefficients.

References: 

Bliankinshtein, N., Gabriel, P., Huang, Y., Wolde, M., Olvhammar, S., Emrich, A., Kores, M., and Midthassel, R.: Airborne

Measurements of Polarized Hyperspectral Microwave Radiances to Increase the Accuracy of Temperature and Water

Vapor Retrievals: an Information Content Analysis, AGU Fall Meeting Abstracts, A13K-2959, 2019

Clough, S. A., Shephard, M. W., Mlawer, E. J., Delamere, J. S., Iacono, M. J., Cady-Pereira, K., Boukabara, S., and Brown, P. D.: Atmospheric radiative transfer modeling: a summary of the AER codes, Journal of Quantitative Spectroscopy and Radiative Transfer, 91, 233-244,

10.1016/j.jqsrt.2004.05.058, 2005.

This paper conducts a radiative closure exercise, using radiosondes observations to drive radiative transfer models which are then compared against both ground-based mid-infrared observations from the AERI and both ground-based and airborne observations from the microwave instrument HiSRAMS.  Closure is considered achieved if the difference between the observation and radiative transfer calculation is within the uncertainties of both.  Three different field campaigns were conducted, which had varying temperature and humidity conditions.  The results demonstrated that there are times when closure was achieved, and other times when it was not.  The authors then demonstrated how to compare the two instruments’ closure using optical depth, arguing that the AERI and HiSRAMS agree when the HiSRAMS was airborne and looking downward, but that the did not agree when the HiSRAMs was on the surface looking upward.

 Generally speaking, these closure tests are highly needed and should be conducted before thermodynamic profiles are retrieved from them (which the authors indicated is future work), so I commend them for this.  However, there are some details that are missing and some concerns I have that should be addressed before this paper could be considered for publication.

One of my primary concerns is that this entire paper is focused upon three campaigns, where each one had a single radiosonde launch (Table 1).  Thus, N=3.  While careful quality control and processing of the radiosonde data can be done, the sampling uncertainty is still quite large.

Another is that radiosondes, by their nature and how they are calibrated at the factory, have both systematic and random errors combined together when providing their uncertainty values (which the authors reported on line 86-88).  Later, they tried to estimate the forward calculation uncertainty by Monte Carlo sampling the radiosonde data (line 225), attributing all of the uncertainties to random errors; it is not surprising to me that this did not result in much spread.  I would highly recommend the authors consider apportioning the stated uncertainties into random and systematic errors as was done in Blumberg et al. JAMC 2017 (in the appendix), as correlated error will lead to larger impacts on the radiative transfer calculations.

The authors have identified an apparent bias in the infrared window channels of the AERI.  I recommend they also read/reference the paper by Delamere et al. JGR 2010, which also noted a similar sized bias and did some exploratory work to understand it (albeit to no avail).  One mechanism the authors should consider is that the MCT detector used in the AERI has a non-linearity correction applied, and perhaps this correction being applied is not correct. 

More importantly, however, is that this apparent calibration error likely changes magnitude based upon the scene temperature; subtracting a bias offset is not appropriate for all scenes.  A radiometric error is almost certainly associated with the slope relating detected signals to radiance (i.e., which is determined by the blackbody views), and as these are very cold scenes a small error in the slope will become larger as the sky temperature becomes colder (i.e., farther away from the ambient blackbody temperature).  Thus, subtracting a bias only works for these cases, where the bias was determined directly from the radiosondes and the associated forward calculation, and is not something “easily removed” as suggested on line 270.

Figure 10 and text around line 333: I am not convinced by the discussion associated with the bias seen in leg 1 on this flight.  The environment at 6.8 km (leg 1) is colder than at 5.3 km (leg 2), but leg 3 at 4.3 km is also colder – yet the bias between legs 2 and leg 3 are very similar whereas the bias in leg 1 is a clear outlier.  I believe something else is likely the problem, and that you should not try to simply state (as on line 333) that the issue “may be due to poor calibration in a cold environment.”

The comparison of the two instruments using optical depth is interesting.  However, there are many trace gases that will be impacting the downwelling radiance observed by the AERI.  Indeed, I can see in Figure 3 spectral structure associated with CFCs, which the authors indicated are not included in their downwelling calculations.  I would highly recommend that the authors only use spectral channels from the AERI that are associated only with water vapor and carbon dioxide; spectral elements that have a contribution from other gases (CFCs, O3, CH4, N2O) should not be used in their optical depth plots.  I suspect that it won’t change the results too much, but makes the messaging much cleaner.

Another large concern I have is associated with the surface contributions to the HiSRAMS observations when the radiometer is aloft and looking downward.  You indicated that you are using the lowest flight level to effectively constrain the emission from the surface with observations; however, this was not discussed in any detail and thus we don’t know if the effective skin temperature is realistic or not, and similar for the surface emissivity.  This should be discussed in more detail. 

Furthermore, you pointed out that the nadir-pointing HiSRAMS seems to agree well with the AERI but that the zenith pointing HiSRAMS does not.  How much is the former results due to the way that the surface was constrained in the forward calculations?  And in optically thin channels (e.g., 52-54 GHz), most of the observed radiance when airborne will be from the surface, and so are you really able to extract out the atmospheric optical depth well?  What are uncertainties here?

Minor details:

The authors need to mention the location of the intercomparison in the abstract and the paper.

Line 102: using 9 adjacent gridboxes (i.e., a 3x3 grid around the desired point) in a model is not a good measure of the heterogeneity of a region, as virtually all models are diffusive and thus there will be a reasonable amount of correlation associated with model smoothing.  Most models have significant model-induced correlations out to 5 times to 8 times the horizontal model grid spacing.

Figure 6: it does not seem that the uncertainties from the observations and simulations are being combined properly in quadrature to get the total uncertainty.  For example: Fig 6a at 850 cm-1 looks like the total uncertainty is the same as the measurement uncertainty; but as the simulated uncertainty is the same magnitude, the total uncertainty should be larger.  

Line 299: I think this phrasing for this sentence sounds better: “The primary contribution to the radiative closure uncertainty in the zenith…”

Line 315: this sentence seems very obvious.   

Fig 13c: please explain the second cloud of points here (which has a bias of -3 to -8 K)?