Dear Editor,
We have addressed the reviewers’ comments and revised the
manuscript accordingly within the time period stated.
Reviewer 1 was very supportive of the paper and recommended
publication and asked if we would address their questions, which
we have done. The questions are mostly clarifications which we
have provided and we believe it has helped with making the paper
clearer.
Reviewer 2 raised more challenging points. The reviewer
required three modifications in order to support publication. 1)
state the new work relative to the papers cited by the reviewer, 2)
address shattering by looking at the sensitivity to small particles in
the modeling and 3) should perform 3D simulations.
We have addressed all three points. We have revised the
manuscript with respect to the papers the reviewer mentioned (and
performed another search to identify any additional papers), 2) we
provided justification that ice particle shattering did not affect our
results, and 3) we have redone all the simulations in 3D and
updated the analysis and text in the manuscript.
We hope that you find the responses and manuscript alterations to
your satisfaction and look forward to your response.
With regards,
Huiyi Yang,
Steven Dobbie,
Paul Connolly
Reviewer 1:
The introductory comments in the review were:
This work provides a study of cirrus cloud optical property and its radiative impact on
single column large eddy simulation models. Cirrus cloud is one of the unsolved
issue for energy budget on the atmosphere-ocean system. In this work, the optical
properties of cirrus cloud are calculated based on different ice crystal habits. This
provides a potential useful parameterization for modeler to handle the cirrus cloud
correctly in the climate model. The study shows that the different ice crystal habits
and the vertical structure can strongly affect the radiative flux. The difference in flux
can be up to tens W/m2 by different assumption of ice crystal habits. This result is
new to the atmospheric radiation community. I therefore recommend publishing this
paper. I hope the authors can address following questions.
1. In the last part of the introduction, better to emphasize the significance of this
work, pointing out the new results of the ice cloud optical properties calculated in
this work.
Done. The last part of the introduction has been updated to better
emphasize the significance of the work.
2. Page 3 line 17 (P3 l17), better point the ‘averaged’ means grid average, i.e the
two-dimensional average, also 4176 should 4176L.
This value is the average ice concentration from the EMERALD1 EM09
flight and so represents an average during the flight of the aircraft. I
have clarified this in the text.
3. P3 l53, better put the reference for Fu-liou model, since it appears at the first time.
Done. Text updated.
4. Figure 1, the information for habit types is not clear, should use color bars? Also
better to give a brief explanation for sph, col, ...
Done. We have updated the plots showing the distributions for each
habit as well as the total distribution.
5. P4 l30, only using 2D because the computational expensive of 3D. The LES
model should not be very expensive in computing. I think one reason for the
authors to choose 2D is that 3D input data from EMERALD1 is less reliable or
not complete for use. If this is true, please mention it.
Done. We have redone all the model simulations to be 3D and updated
the plots and analysis in the manuscript.
6. P4 l55, this paragraph is not very clear to me. Do the authors mean that they
continuously change ice water path and the ice habits during the running
process of LEM?
The manuscript has been updated with a clear explanation.
7. P5 l15, Baran and Yang’s compilation is used to calculate the optical properties
of different ice habit. I just wonder what kind method is used in this compilation,
T matrix, geometric optics approximation, ..? Could the authors add a sentence
of explanation.
Text needs adding
Calculation of the single-scattering properties for the long-wave region
is based on the T-matrix method and the Complex Angular Momentum
Approximation (CAM). The paper by Havemann and Baran (2001)
describes the implementation of the T-matrix method to the hexagonal
column.
Ref
Havemann, S., and A. J. Baran, “Extension of T-matrix to scattering of
electromagnetic plane waves by non-axisymmetric dielectric particles:
application to hexagonal ice cylinders”, J. Quant. Spectroscop. Radiat.
Trans, 70, 139-158, 2001
8. P5 l9 left column, COLAGG, I saw in Fig. 4 it is written as COL&AGG, and inP8
it is written as AGGCOL. I think they are all the same, but why it is writtenin
different ways.
We have corrected this in the manuscript.
9. P6, Eq.(5), what is useful for _βext/sca? could the authors give a simple
explanation.
Done. This has been clarified in the text.
10. As a climate modeler, I am very interested in the calculation results shown in P56. I have a few questions about them. a) what kind distribution was assumedfor
n(L)? is that from Fig. 1, please write it clearly. b) from (6-9), I can not findwhat is
variable that the ice cloud optical properties depend on. As we know theprevious
parameterization is dependent on De or Dg. If I understand it correctly,the
calculations here are all based on Fig. 1 for mixing of different habits. Writeit
clearly. c) is it possible to provide a parameterization pertaining the habit
segregated size distribution, which can be generally used in climate models. At
least,the related discussion should be provided.
In this work, to calculate the optical properties of the cirrus layers we do
not assume a size distribution, as we directly use the observed size
distribution from the EMERALD1 EM09 observations.
We believe the reviewer expects the optical properties to be in a
parameterized format and this is why they are seeking the variable
dependence of De or Dg. We have not derived a parameterization in
this work. The optical properties are derived specifically for the
observations of EMERALD1 EM09 only.
To clarify, we have renamed the section 4.2 as “Calculation of optical
properties specific for the EMERALD1 observations”
Also, we have included in the text that our optical properties are specific
to EMERALD1 and do not constitute a parameterization. This appears in
the manuscript at the end of the optics calculations, just after the band
averaging equations.
11. P7, is Table 11 necessary?
It is not essential and so Table II has been removed.
12 P9 Fig. 6, not clear to what is time 1.5 and 2 hours. Do the authors mean that
difference in flux after 1.5 sec run? Please write it clearly. I just wonder why don’t
compare a steady sate case, just the pure radiation calculation for different ice
cloud input. Then the readers can have more clear picture for the radiative
impact of habit segregated size distribution.
In the original manuscript the two times (1.5 and 2 hrs) where chosen to
be around the peak in IWP.
We have updated these analysis times to
be 120 min and 165 min. The time of 120 min is when the model run ice
concentration is equal to the average ice concentration from the
observations for the whole of flight duration of EM09 and the second
time of 165 min is chosen because the ice concentration from the model
agrees with the average ice concentration from the observations during
the flight segment in which the four layer vertical structure was
observed. We trust this makes the choice of times completely justified
and have updated the manuscript.
Reviewer 2:
The introductory comments in the review were:
This study looks at the sensitivity of radiative fluxes to microphysical assumptions
based on a Cirrus observations study. Changes are made to ice crystal habit and ice
crystal size distribution in 2D model simulations.
The authors claim that this is the first study to assess the importance of vertical
structure, size distributions and habit. The paper by Liu et al. (Liu, Hui-Chun, Pao K.
Wang, Robert E. Schlesinger, 2003: A Numerical Study of Cirrus Clouds. Part II:
Effects of Ambient Temperature, Stability, Radiation, Ice Microphysics, and
Microdynamics on Cirrus Evolution. J. Atmos. Sci., 60, 1097–1119.
doi: 10.1175/1520-0469(2003)060<1097:ANSOCC>2.0.CO;2 ) looked at the effect of
changing crystal habit and vertical structure. There are also many papers by P. Yang
looking at similar effects (e.g. Sensitivity of cirrus bidirectional reflectance to vertical
inhomogeneity of ice crystal habits and size distributions for two ModerateResolution Imaging Spectroradiometer (MODIS) bands
Author(s): Yang P, Gao BC, Baum BA, et al.
Source: JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES Volume:
106 Issue: D15 Pages: 17267-17291 Published: AUG 16 2001 )
Unfortunately, recent findings about the effects of shattering of ice crystals on the
instruments used to measure them mean that doubt is cast on many of the
measurements. This is especially true where small particles are important as they
are for this paper and when using the CPI that has a small inlet. It is now the
responsibility of the authors to make a case for why their observations are useable.
The issues are
i) does this paper offer new insights that were not dealt with by other authors such as
Wang and Yang?
ii) Would using observations not affected by shattering change the sensitivity
significantly.
The reviewer 2 states three required modifications.
Required modifications:
1. Review previous work and decide if new insight is being offered.
2. Test sensitivity when fewer small particles are contained in the ice crystal size
distribution (e.g. following Mitchell 2008, GRL). This could be done by applying a
function to ramp down the concentrations by a large factor for sizes smaller than 200
microns, for example. When the concentrations of small particles are reduced are
the sensitivities seen in the 7 cases still as large, or are they diminished?
3. 3D simulations need to be carried out. It has to be demonstrated for this case that
the results obtained for 2d are the same as 3d. The text talks about how it is
important to spin up the correct turbulence. There are numerous studies that show
that cloud structure developed from 2d simulations is inferior to that developed from
3d. This would require a minimum of a control and one of the experiments to be
simulated in 3d and show that the radiative profiles are the same for 3d and 2d
cases.
We have addressed all three of these below as well as the minor other
points.
1. Review previous work and decide if new insight is being offered.
The reviewer has identified two papers that published work on the
importance of vertical structure, size distributions and habit. The two
papers were Liu et al 2003 and Yang et al 2001. We have included
these papers in the discussion of the introduction and clearly delineate
our work from theirs.
Liu et al 2003 had a small final section in their paper focusing on habit
effects on the development of cirrus. Their work is valuable as an
idealised assessment of the bounding effects of switching whole-cloud
habits from one habit type to another. Our paper is distinct from their
paper by assessing the habit effect on the radiative fluxes based on
using actual observed habit mixtures. Furthermore, we consider
vertical structure of habits and size distributions which that paper does
not address.
The second paper is Yang et al 2001. This was an oversight on our part
as their paper does have some similar material with regard to observed
vertical structure however with an emphasis on informing bi-directional
satellite remote sensing retrievals. Our paper has significant
differences with Yang et al 2001 in that a) our paper addresses fluxes
and both at TOA and the surface (our focus on fluxes makes the work
more attractive to the broader audience such as weather and climate
modelers), whereas their paper has a focus on TOA radiances; b) their
research is based on a case with a significantly larger Deff, and we
illustrate a strong sensitivity of cloud evolution to smaller sizes; c) our
study has a sensitivity analysis that illustrates the effects of making
progressively more and more simplified assumptions which gives
modelers a good idea about the importance of, for instance, treating
vertical structure compared to habit changes and more simplified
common assumptions, and other differences such as d) our cloud is at a
significantly different temperature and e) we address both solar and
infrared effects on the radiative and cloud evolution.
We have revised the paper, especially the introduction, clearly stating
our contributions in the context of these other works. Also, we have
removed our claim to be the first study to assess the importance of vertical
structure, size distributions and habit.
2. Test sensitivity when fewer small particles are contained in the ice crystal size
distribution (e.g. following Mitchell 2008, GRL). This could be done by applying a
function to ramp down the concentrations by a large factor for sizes smaller than 200
microns, for example. When the concentrations of small particles are reduced are
the sensitivities seen in the 7 cases still as large, or are they diminished?
We understand that the reviewer is requesting this sensitivity because of a
concern that about shattering of ice into smaller ice particles at the probe inlet
during the observations. This led to the stated issue: “ii) Would using
observations not affected by shattering change the sensitivity significantly”.
We have clarified in the manuscript that the ice crystal observations were not
significantly affected by shattering. We base this on two points. First, there
is very good agreement in ice crystal number concentrations between the two
probes (CPI and FSSP). This fact alone is evidence for lack of shattering since
the response of both probes to ice smashing on the inlet would have to be the
same: highly unlikely. Second, the CPI can tell you how many particles there
were per image so we plotted a histogram of that. It is clear that almost all
images have only one particle in them, but few have two and even fewer have
3. We can compare this to a Poisson distribution to assess the randomness of
the particles. If randomness is observed then it points to shattering not being
important as this would produce some bias in arrival times. The parameter
for the Poisson distribution was calculated from the mean number of particles
per image. The mean concentration for that period was 280 L^-1 and the
sample volume of the CPI is ~2.3e-3.^2 (see Connolly et al 2007 in JTECH).
Therefore, the mean number of particles per image is 0.035 (which is the
lambda parameter for the Poisson distribution). Plotting a poisson distribution
on top of the CPI histogram reveals they are almost the same. Included in the
plot is a curve for Lamda=0.70 just to illustrate how sensitive it is.
There could be a very small amount of shattering occurring since the number
of frames with 3 particles in it (orders of magnitude less than frames with 1 or
2 particles) is slightly more than that expected by randomness, but this is very
slight. The histogram is attached and discussion are now included in the
manuscript.
3. 3D simulations need to be carried out...
We have ported the model to a faster computer and performed all the
model simulations in 3D. All of the plots, analysis, and text are
updated in the revised manuscript. The 3D results show slightly
stronger effects.
The reviewers ‘other points’ are addressed below:
Key: page,column,line
2,2,3. SID-3. Some more information and references are needed here. Are the
observations from similar cases?
There is limited literature and we are unsure if the conditions that give
rise to roughness are present in our case. We are rather using
roughness to illustrate a comparison of the relative importance of
vertical structure, habits, and size distributions. We have clarified this
point in the manuscript.
3,1,4. Cloud 'particle' size distributions 3,1,19. Mean mass-, number-, areaweighted size?
We have updated 3,1,4 with “cloud particle size distributions”. We have
clarified the determination of the mean particle size. It is the directly
observed mean projected size by the CPI.
3,1,18-19. Very high concentrations of 0.5-5 per cubic centimetre.Possible at
homogeneous freezing altitudes, but less likely at warmer temperatures. See
arguments above regarding shattering.
Yes, we find that the dominant nucleation is homogeneous in the
simulations and this is stated in the paper. As stated above the
3,1,41. Both CPI and FSSP suffer from the effects of shattering.
We believe they didn't in this cloud. See arguments above.
4,1,30-33. But 2d simulations do not capture structures developed from 3d
turbulence. Is the Yang 2009 study is of a different case? If it is then the exercise to
compare 3d and 2d needs to be repeated.
All simulations have been replaced by 3D runs. All of the plots,
analysis, and text are updated in the revised manuscript. See Above.
4,1,43. Realistic turbulence. If this is important then 3d simulations are required.
All simulations have been replaced by 3D runs. All of the plots,
analysis, and text are updated in the revised manuscript. See Above.
4,2,25. This case is a tropical case.
The reviewer is perhaps thinking of EMERALD2. EMERALD 2 is a
tropical case whereas EMERALD 1 is mid-latitude case.
5,2,36-60. Are these relations for columns or aggregates? What is used for the other
habit?
Update this with the relations or references.
8,2,10-12. Some references are needed here to support this statement.
The statement is refined and references provided.
With the complex growth mechanism of ice crystals in the atmosphere,
that depend strongly on temp and humidity as well as fall speeds of a
particle, perfect hexagonal symmetric particle shapes may be the
exception rather than the rule. Macke, A, J. Mueller, and E. Raschke,
Jas 53, no 19, 2813-2825. There have been several studies looking at
the implications of roughness (e.g. Yang and Liou, 1998; Sun et al., 2004;
Yang et al 2008).
“The unexpectedly low value of the
asymmetry parameter would then have far-reaching consequences for
the magnitude of cirrus cloud radiative
forcing. This area clearly deserves further study.” Ulanowski Z., Hesse
E., Kaye P. H., and Baran A. J., 2006, Light scattering by complex iceanalogue crystals., J. Quant. Spectrosc. Radiat. Transfer, V100, P382392.
Recent in situ cloud data from the SID-3 probe is shown which is consistent
with ice particles with rough surfaces being dominant.
Ulanowski Z, Kaye P.H., Hirst E., and Greenaway R.S., 2010, Light
scattering by ice particles in the Earth’s atmosphere and related
laboratory measurements, Electromagnetic and Light Scattering XII
Conference, Finland.
Ping Yang, George W. Kattawar, Gang Hong, Patrick Minnis, and
Yongxiang Hu
Uncertainties Associated With the Surface Texture of
Ice Particles in Satellite-Based Retrieval of Cirrus
Clouds—Part I: Single-Scattering Properties of
Ice Crystals With Surface Roughness
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL.
46, NO. 7, JULY 2008
W. B. Sun, N. Loeb, G. Videen, and Q. Fu, “Examination of surface
roughness on light scattering by long ice columns by use of a two
dimensional finite-difference time-domain algorithm,” Appl. Opt., vol. 43,
no. 9, pp. 1957–1964, Mar. 2004.
P. Yang and K. N. Liou, “Single-scattering properties of complex ice
crystals in terrestrial atmosphere,” Contrib. Atmos. Phys., vol. 71, no. 2,
pp. 223–248, 1998.
Ice particles are believed have Roughness of ice particles has been
studied by
Ans. Three references have been added to support the statement.
Ref 1: Yang P., Kattawar W.K., Hong G., Minnis P., and Hu Y.X., 2008,
Uncertainties Associated With the Surface Texture of Ice Particles in
Satellite-Based Retrieval of Cirrus CloudsPart I: Single-Scattering
Properties of Ice Crystals With Surface Roughness, IEEE Trans. Geosci.
Remote Sens., V46, NO. 7, P1940-1947.
Ref 2: Ulanowski Z., Hesse E., Kaye P. H., and Baran A. J., 2006, Light
scattering by complex ice-analogue crystals., J. Quant. Spectrosc.
Radiat. Transfer, V100, P382-392.
Ref 3: Ulanowski Z, Kaye P.H., Hirst E., and Greenaway R.S., 2010, Light
scattering by ice particles in the Earth’s atmosphere and related
laboratory measurements, Electromagnetic and Light Scattering XII
Conference, Finland.
8,2,36. a 4 hour simulation time needs to mentioned earlier in the paper (section 3).
Done. “The total model simulation time is 4 hours.” has added in
section 3 in the paragraph when the model setup is discussed.
8,2,56 - General - what if more layers were used than 4. What number is required to
begin to converge the results?
Interpolating to more layers is possible but not our intention without
reliable observed ascending profiles through the cloud.