Fast Radiative Transfer Approximating Ice Hydrometeor Orientation and Its Implication on IWP Retrievals

Round 1

Reviewer 1 Report

Review comments on “Fast radiative transfer approximating ice hydrometeor orientation and its implication on IWP retrievals” by Kaur, Eriksson et al.

In this manuscript, the authors reported studies on approximating ice hydrometeor orientation in fast radiative transfer calculation and further discussed its implications in IWP retrievals. Accurate and efficient computation of single-scattering properties as well as the radiative transfer is much more complicated than the random orientation assumption and the classical vector radiative transfer equation. To the best of my knowledge, the novelty of the present study as well as Barlakas et al. (2021) was to incorporate the non-random orientation issue in the relevant studies associated with data assimilation.  The scope of this study is well defined and the paper is well organized and clearly written.

My only concern is the method. Any assumptions or simplified treatment for fast radiative transfer equations should be carefully examined based on rigorous calculations. In that case, the uncertainties due to mathematical treatment are clearer before further applications. I understand that this work is untrivial. Since the present work is entirely based on RTTOV and a somewhat artificial tuning of the polarization parameter, one or two sentences somewhere that highlight the necessity of such studies will be nice. Very minor things: 1) The reference could be explicitly included in line 7 (abstract). 2) Fig. 6, seaice => sea ice .  

Overall, I think the manuscript can be published as is.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Overview:

The manuscript "Fast radiative transfer approximating ice hydrometeor orientation and its implication on IWP retrievals" by Inderpreet Kaur, Patrick Eriksson, Vasileios Barlakas, Simon Pfreundschuh and Stuart Fox presents a new approach dealing with ice particle orientation effect in remote sensing. Using the Global Precipitation Measurement Microwave Imager (GMI), model data and ML techniques. The authors showed that the ice water path (IWP) retrievals measurements are sensitive hydrometeors orientation. This fact was well known before but the authors suggested a reasonable and affordable correction method. Moreover, the suggested method can be beneficial for future ice particles properties classification using satellite data which is essential for data assimilation in Numerical Weather Prediction models. In this manuscript, the theoretical background and technical methods are well presented to the reader. It is overall well written and structured. In my humble opinion, the paper is too long and could be easily separated to two different manuscripts. Some topics/sections are important enough by themselves and worthy of a separate publication. But it is acceptable as is. The manuscript is recommended for publication in Remote Sensing upon addressing these minor comments:

Comment 1: The issue of small particles that are not included in your model (but are part the reality in clouds) should be discussed in a more qualitative way. I.e. what is the error cost and significant of ignoring them?

Comment 2: The particle habit classification and sizes (PSD) should be discussed and rationalized/justified in a microphysical sense. A qualitative discussion is needed here. A short discussion of why choosing a specific optical properties data set from the other.

Comment 3: You should explain why the observations and model runs were not covering the same periods and justify in more details why, nevertheless, the results are not greatly affected by this.    

Line 203: You should specify the optical properties you used for the model at 166GHz as a function of size, habit etc. For this frequency/wavelength, the size parameter should be very small for small ice particles. In figure 1 you show the ratios between ARO/TRO. It would be interesting to see the same for small ice particles which are significant part of the total ice content (if possible).

Line 206: This assumption is wrong since liquid droplets can exist in temperatures as low as ~243K. Maybe it is a typo error? (see line 198).

Line 233: More information is needed about this PSD: how it was calculated, mode size, habits, aspect ratios etc. How do you classify them correctly in the model and observations etc. As you mentioned several times throughout the manuscript, misclassification is a major source of error in your work so you need to be more detailed here (although, some is mentioned in section 5.2).

Line 344: missing ")".

Line 809-810: why is that important?

Line 906: Did you consider using other parametrizations/data sets? (i.e. Yang, P., L. Bi, B.A. Baum, K. Liou, G.W. Kattawar, M.I. Mishchenko, and B. Cole, 2013: Spectrally Consistent Scattering, Absorption, and Polarization Properties of Atmospheric Ice Crystals at Wavelengths from 0.2 to 100 μm. J. Atmos. Sci., 70, 330–347)

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This study is concerned with microwave radiances emitted/reflected from ice-bearing clouds that typically show H-V polarization differences presumably due to particle orientation. The authors use synthetic data from CloudSat, an idealized set of particle size distributions (including two shapes assumed for frozen hydrometeors), and radiative transfer simulations adjusted to incorporate orientation-mimicking polarization effects to reproduce the distribution of measured microwave radiances. For best reproduction, they determine that orientation needs to be treated stochastically. The retrieval method they develop obtains similar ice water paths compared to existing reference data sets. Lastly, the authors demonstrate the impact of neglecting particle orientation which reflects the current retrieval standard.

The article is generally well written and contains important findings. I recommend this article for publication after the authors resolve a few major issues.

 

Major points

The randomly generated scaling factor produces suitable polarization differences but evades the burning question of what explains preferred orientations. Can one really assume uniform distributions in all circumstances? Perhaps a composite analysis (splitting scenes by suspected candidate variables) could solve this? The authors could show ways forward in their discussion. What else is needed for a future retrieval that includes azimuthal orientation?

Frozen hydrometeors naturally appear in many shapes with individual size-mass relations. It is unclear whether the idealizations in this study (using two shapes with some assumed density and one particle size distribution) are representative for the collection of shapes (and their size-mass relations); the authors should discuss whether they expect similar findings on other hydrometeor shapes.

The authors should include uncertainty/error estimates in their figures to demonstrate how significant the findings are; ideally considering stochastic uncertainty (i.e., from stochastically sampled orientation) in retrieval errors as well as uncertainty (or standard errors) of references products.

The relatively lengthy study contains many new items, and it is hard to quickly assess the scientific advance. The authors should try to improve their abstract and conclusions to better highlight the objective of their work and more clearly distinguish new from existing methods that helped to arrive at their findings.

When carefully reading the paper, the use of various years (2009, 2013, 2017) confused and should ideally be consolidated to one set of samples (i.e., from one particular month in one year) or be better motivated in the results to help the reader.

 

Minor points

l. 27 rephrase “ice” to “frozen”

ll. 78-79 I don’t follow this sentence. Would it be simpler to say that flattened drops do not have a preferred azimuthal orientation?

l. 121 What is the native resolution?

l. 152 (16 km)^2 ?

l. 230 Please add a citation for “Evans snow aggregate”.

ll. 205-206 Is that a realistic assumption? What about supercooled liquid?

ll. 228-230 This seems to repeat the same structural error that was criticized earlier in the paper.

l. 275 Why “frozen water path” and not “ice water path”?

ll. 348-349 Why are different scenes used here? Is it part of a validation strategy?

ll. 383ff. Does rho change from footprint to footprint?

l. 564 What is “Q-IWP”? Not introduced, I believe.

ll. 685-686 The authors raise an excellent point. Perhaps samples should ideally coincide with in-situ sampled clouds from various aircraft campaigns?

ll. 689-702 This portion seem best suited for the introduction.

Table 3 Perhaps highlight best performers in bold font.

Fig. 3 Perhaps add contours of 5^th and 95^th percentiles of observations above all points to better show observed distribution.

Fig. 10 How are values for the x-axis determined? Do they stem from uniformly sampled values or are they from simulations with constant values (which I thought were only done for rho in [1.0, 1.1, … 1.4] but here appear in smaller increments?)

Fig. 11 It is unclear which “reference” was used. Perhaps better clearly define in text and use consistently.

Fig. 1,7,8,9,10,11,14, Ideally show uncertainties of all lines (see above major point on uncertainty).

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Abstract: In their response the authors express that the abstract has been updated but the tracked version doesn’t show any updates. Could the authors please highlight changes to the abstract?

Fig. 3 (and likewise Fig. A1): It is hard to assess the full range of observed values as they are obscured by the simulated points. As suggested in the initial review, I would overlay lines of 5^th and 95^th percentiles of the observations (computed per increment in TB) to better highlight the observed range.

l. 95 Perhaps better “liquid particles (i.e., rain drops and cloud droplets)” instead of “liquid (and clouds) droplets”.

l. 138 Perhaps better “(i.e., a pixel size of ~6 km at nadir)” instead of “(6 km)”.

 

The authors addressed all other points adequately, thank you!

Author Response

Please see the attachment

Author Response File: Author Response.pdf