: Comparisons of bispectral and polarimetric retrievals of marine boundary layer cloud

Abstract. Many passive remote-sensing techniques have been
developed to retrieve cloud microphysical properties from satellite-based
sensors, with the most common approaches being the bispectral and
polarimetric techniques. These two vastly different retrieval techniques
have been implemented for a variety of polar-orbiting and geostationary
satellite platforms, providing global climatological data sets. Prior
instrument comparison studies have shown that there are systematic
differences between the droplet size retrieval products (effective radius)
of bispectral (e.g., MODIS, Moderate Resolution Imaging Spectroradiometer)
and polarimetric (e.g., POLDER, Polarization and Directionality of Earth's
Reflectances) instruments. However, intercomparisons of airborne bispectral
and polarimetric instruments have yielded results that do not appear to be
systematically biased relative to one another. Diagnosing this discrepancy
is complicated, because it is often difficult for instrument intercomparison
studies to isolate differences between retrieval technique sensitivities and
specific instrumental differences such as calibration and atmospheric
correction. In addition to these technical differences the polarimetric
retrieval is also sensitive to the dispersion of the droplet size
distribution (effective variance), which could influence the interpretation
of the droplet size retrieval. To avoid these instrument-dependent
complications, this study makes use of a cloud remote-sensing retrieval
simulator. Created by coupling a large-eddy simulation (LES) cloud model
with a 1-D radiative transfer model, the simulator serves as a test bed for
understanding differences between bispectral and polarimetric retrievals.
With the help of this simulator we can not only compare the two techniques
to one another (retrieval intercomparison) but also validate retrievals
directly against the LES cloud properties. Using the satellite retrieval
simulator, we are able to verify that at high spatial resolution (50 m) the
bispectral and polarimetric retrievals are highly correlated with one
another within expected observational uncertainties. The relatively small
systematic biases at high spatial resolution can be attributed to different
sensitivity limitations of the two retrievals. In contrast, a systematic
difference between the two retrievals emerges at coarser resolution. This
bias largely stems from differences related to sensitivity of the two
retrievals to unresolved inhomogeneities in effective variance and optical
thickness. The influence of coarse angular resolution is found to increase
uncertainty in the polarimetric retrieval but generally maintains a
constant mean value.


As polarimetric instrument development continues, papers like these help track the state of the field. They will be used to inform new instrument designs, missions, and algorithms. Therefore, I recommend this exciting paper for publication in AMT. However, I suggest a mix of specific and minor revisions.

Specific Comments
(1) The opening paragraph needs more discussion on the current state of cloud and climate science: their non-uniform global distribution, both vertically and horizontally, thermodynamic phase differences (ice vs. water), and current challenges in comparing remote sensing/in-situ measurements/retrievals, cloud simulations, and global climate models. These are just examples, but please add a few extra sentences that put the paper in stronger context with the field.
The connection between aerosols and clouds and the potential benefit of using polarmetric measurements in aerosol-cloud studies is missing in the Introduction. This is a hot field of current cloud research and polarimeters like specMACS are highly relevant to this topic. Please add a few sentences about this.
(2) One of the major results of this paper is the ability to retrieve reff ~40um from specMACS cloud data. This capability is highly attractive for future climate applications and missions. This can be a challenge for some polarimeters and retrievals done on their data. For this kind of retrieval, some polarimetric instruments are capped in the upper bound of retrieved reff due to limitations in view zenith angle density (Miller et al. 2018, citation below). specMACS may get around this limitation with its dual-camera design and access to a second set of retrievable pixels and geometry for the same cloud target. This is important to mention. Also, more details about how the specMACS design and sampling directly compares to other, similar cloud measuring instruments (specifically AirHARP and RSP) would be valuable.
Miller, D. J., Zhang, Z., Platnick, S., Ackerman, A. S., Werner, F., Cornet, C., and Knobelspiesse, K.: Comparisons of bispectral and polarimetric retrievals of marine boundary layer cloud microphysics: case studies using a LES-satellite retrieval simulator, Atmos. Meas. Tech., 11, 3689-3715, https://doi.org/10.5194/amt-11-3689-2018, 2018 This paper would also benefit from a short, quantitative discussion of retrieval uncertainty (Qual and RMSE). How well can one use specMACS data to reliably retrieve large reff and veff? An accompanying sub-figure similar to Figure 7a and b that shows the spatial distribution of the bestfit RMSE may support this discussion.
(3) In many areas, this paper has colons (:) when a period would be more effective. Please look through the document and make revisions as needed.

Abstract
Please put the conclusion of this paper at the end of the abstract.

Section 1
Line 21: What is "extreme precipitation"? Line 22: "future temperature changes" is vague. The IPCC AR6 is specific in defining how future climates may be influenced by changes in sea surface temperature, global mean air surface temperature, or other similar measures. Please be specific.
Line 39: Large phrases like "droplet size distribution" can be reduced to conventional acronyms, like DSD. Also, since effective radius and effective variance are defined as reff and veff in Lines 43/44, please can use these abbreviations going forward in the paper. Other phrases can be simplified too -like "degree of linear polarization" to DOLP.
Lines 42 and 97: Remove the "e.g." Line 51: "it has some difficulties which are mainly related to 3-D effects occurring especially in inhomogeneous cumulus cloud fields" should change to "has known biases in the presence of 3-D effects and spatial inhomogeneity." Spatial inhomogeneity impacts the retrieval at some level for all cloud types, and those biases aren't always related to 3-D effects.
Line 53: "Furthermore, retrieving the effective variance of the cloud droplet size distribution is not possible." Please add a citation and/or elaborate.
Line 58, 134, 239, 292, 345: please remove "so-called" in all instances. In most cases, these techniques actually go by these names (i.e. scattering matrix is the name of that matrix, wire-grid polarizer is the official term for that kind of polarizer). In Line 59, it would be a stronger sentence as "Based on polarized observations of the cloudbow, a new kind of DSD retrieval was developed." Line 73: The following two sentences are somewhat redundant. Something like "Furthermore, the veff of the cloud DSD is derived in the polarimetric retrieval. This parameter may be directly linked to entrainment and mixing processes at the cloud top." would be stronger.
Line 91: The ~2um reff bias between the MODIS bi-spectral and polarimetric DSD techniques was found in other earlier studies as well, largely due to information content differences in location of the cloud DSD creating the signal, retrieval resolution, inhomogeneity in the pixel, and choice of SWIR band used in the bi-spectral retrieval. Please cite as necessary: Line 97: Please re-word this sentence for clarity, something like "RSP data can provide a bispectral and a polarimetric reff from the same cloud target, due to spectral coverage from VIS to SWIR and along-track, co-located multi-angle sampling."

Section 2
Line 133 (and Figure 1). I appreciate the relative spectral response figure added to the manuscript. It would be helpful to also include the center wavelength and bandwidth (FWHM) of the three filters in the text itself. The typical convection is to write it as wavelength (bandwidth), like 440 (15) nm.
Line 142: Please reword this statement -the degree of linear polarization (DOLP) describes the fraction of the incoming light that is linearly polarized. Q and U also quantitatively describe the linear polarization.
Line 152: Remove the comma between "advantage, that" Line 156: "For further analysis, each measured Stokes vector is rotated into the scattering plane (Hansen and Travis, 1974) and we only evaluate Q." Although this work focuses on solar principal plane (SPP) geometries (typically a narrow line of pixels in an observation), there will likely be non-zero U values for cloud targets located off the SPP in the spatial field retrieval. Were there cloud targets with non-negligible U values? If so, how does that contribute to the interpretation of the retrieval results? Figure 2 caption: Please reword for clarity -"The primary bow of the cloudbow is visible in the degree of linear polarization as a bright ring at a scattering angle of about 140°"

Section 3
Line 173: "This method can easily be applied to any cloudbow observations, including those from commercial cameras, but it requires averaging over a large area." Please describe why is this less desirable than the co-located, along-track method presented in this paper.
Line 175: Please change "we fly over it" to "specMACS images the scene".
Line 181: "In a final step, a look-up table (LUT) of cloudbow signals for different cloud droplet size distributions.." Please be specific that a simulated Mie LUT is used for this comparison.  Line 241: The polarized phase function is directly proportional to the measured polarized radiance Q, under the assumption of single scattering. P12 is not an approximation of Q. Please revise.
Line 245: Effective variance determines the amplitude of the secondary maxima/minima (see Figure 5b), which is where the information content lies (not number of peaks). Please revise this sentence.
Line 254: Please add the word unimodal or monomodal before "gamma distribution".
Line 280 and 287: Thanks for changing Eq. (6) and (8)   The presentation of the "number of measurements" is a little confusing. In the plots, it almost looks like an uncertainty but it's a bit more complex than that and it blurs the overall message in my view. When the slope of the Q_fit signal is large in a small scattering angle range (see 6a between 142-147 degrees), it is hard to differentiate the box values from the measurement/fit lines. I recommend to replace the "number of measurements" pixels in the plots with errorbars that correspond to the angular measurement. This goes for Figures 9 and 15 as well.

Section 4
Line 376: Please remove the comma between "example, to" Line 378: Please remove the e.g. and commas.
Line 380: Please change to "In the case of small cumulus clouds, a precise geolocalization is important for image-to-image tracking." Line 393: Please define 100m as a "target unit" here. This is definition referenced later in Line 401, but in an indirect way and was confusing on a blind read.  Line 423: If the error due to incorrect geolocalization is yet be estimated quantitatively, how is it that the cloudbow retrieval is "affected [by it] to a much lesser degree"? This is confusing. Please elaborate.
Line 433: It is clear in 12b that the majority of 0.3+ veff values are "tracing" the lower boundary of the overlap between polA and polB cameras. 12f shows a big spike at the 0.32 veff bin, which is the very edge of the Mie LUT described on Line 269. This together suggests that these retrievals are artificially converging (either due to noise or errors in geolocalization) and may not be valid, even if they pass the Qual and RMSE. I suggest removing this veff bin from the Figure 7f and 12f plots (and corresponding data points from the 7b and 12b maps) to focus the discussion on the valid larger veff that do exist in the data.
Line 458: McBride et al. (2020) was also careful to note that the subpixel reff and veff distribution that contributed to their larger-scale wide DSD result could also be impacted by cloud height georegistration (they used a granule-wide average, not pixel-by-pixel as in this study). I suggest to change the "does" to "may" in this sentence.

Section 5
Line 468: Remove the comma between "cloud, that" and change "are" to "may be". This has not been shown in this study, only suggested.
Line 470: Add a comma between "layer very"