Please note that we found an error in how the numbers in Tables 5, 6, S5, and S7 were generated. The tables were regenerated. The numbers are not importantly different from what was in the preceding versions of the paper, and for clarity we have accepted that changes for those tables in the returned documents. Some text is slightly altered to account for the changes in numbers presented.

pointdepressedMécanique

22 févr. 2014 (il y a 3 années et 4 mois)

70 vue(s)

Please note that we found an error in how the numbers in Tables 5, 6, S5,
and S7 were generated. The tables were regenerated. The numbers are not
importantly different from what was in the preceding versions of the paper,
and for clarity we have
accepted that changes for those tables in the
returned documents. Some text is slightly altered to account for the
changes in numbers presented.


Reviewer #1 (Comments to Author):


I have read the revised manuscript and the authors' responses to my previ
ous
comments. I am satisfied that the authors have addressed my previous
comments. The revised manuscript is significantly improved over the initial
submission and I believe it is suitable for publication.


We appreciate the time the reviewer has given to

reading and commenting
on our manuscript.


Reviewer #3 (Comments to Author):


The revised version is a significant improvement from the original manuscript.
The new abstract states clearly the major findings of the study. In my opinion,
two of the findin
gs will attract the most attention of dust researches: (1) strong
dust absorption, and (2) a positive feedback of dust aerosol on dust emission.
The reason is that not only these findings address the important aspects of dust
research, but also because the
y are in conflict with past studies. My comments
below are mainly related to these two findings.


We appreciate the time and thought the reviewer has given to reading and
commenting on our manuscript.


i) The authors' conclusion about strongly absorptive
dust is mainly based on
intercomparison of modeled dust fields vs. observational data. Specifically, the
authors examine the spatial distribution of African dust aerosol in the summer
and find better agreement in the case of strongly absorptive dust. My co
ncern is
that the reader may take away a wrong take
-
home message that high dust
absorption applies to all dust types in various global dust source regions, which is
not demonstrated by the manuscript. In order to avoid confusion, I would suggest
the author
s to modify the title or the beginning of the abstract, etc. to state more
clearly that this conclusion is based on the analysis of African dust only.


We clarify in the first sentence of the abstract that we are talking about the
radiative effects of Sah
aran dust aerosols.
Additionally, i
n the conclusion
section we make the point that we will need to separately consider other
dust source
regions,

as the properties of dust may be different.

We write
(lines 1058


1062):



Our results should be understood

to pertain especially to Saharan dust,
as that was the focus of our evaluation. We recognize that dust optical
properties may exhibit considerable variability depending on source region
and mobilization processes, and that this may be determinant in the
overall
forcing of dust aerosols (Sokolik and Toon, 1996).



Comparing modeled against observed dust fields, it is important to keep in mind
that several different mechanisms affect the spatiotemporal distribution of dust,
and some can cause the compensati
ng effects. For instance, cloud processing
and wet removal (rainout) will strongly affect dust fields. Alternatively,
dust/clouds/precipitation interactions could result in the longer lifecycle and
hence the larger extent of dust fields if dust would suppr
ess precipitation. In
addition, there is growing evidence that non
-
spherical particles may have a
longer lifetime compared to spherical particles. These and other processes
combined will be controlling dust fields in real conditions. However, the authors
d
iscuss the strengthening of the summer Hadley cell circulation caused by strong
absorptive dust as the only mechanism and do not elaborate on the role of other
relevant processes. The author state that considering varying SST may lead to a
very different r
esult, but they do not elaborate on the processes either.


As the reviewer points out, there are many factors controlling the overall
distribution of dust. Our model presents only a single realization of how
those factors may be parameterized (i.e., one
realization of aerosol wet
removal processes, one realization of dry removal processes, etc.). The
focus of the paper is on a single factor, in this case the assumed input
optical properties of dust (shape and refractive index), and other factors
are held

constant. We acknowledge the reviewer’s concern i
n the final
paragraph of the conclusions

(lines 1363


1367):



Additionally, there is a need to evaluate model parameterizations,
particularly with respect to wet loss processes in governing the east
-
west

gradient of dust over the Atlantic. Particle non
-
sphericity was considered
as pertains to dust optical properties, but its impact as far as reducing
particle fall speeds was not considered in this study (e.g., Cola
rco et al.
2003; Ginoux 2003).”


Regarding the forcing of the Hadley cell circulation, the variable in our
experiments is the dust radiative forcing, and so the differences we
illustrate between the baseline No Forcing case and subsequent model
cases are those caused by aerosol forcing.
We have somewhat toned
down our appeal to interactive SSTs in the conclusion of the paper,
although we still point out the on
-
shore flow pattern illustrated in Figure 16.
This is a result of
the

dust radiative forcing

realized in our model
, which
cools
the surface over the continent

but because it cannot cool the surface
over the ocean sets up an onshore flow pattern to try to reduce the

east
-
west temperature gradient (lines 952


956):



The wind vectors shown in Figure 16 indicate this enhanced wind sp
eed
across the southern Sahara is generally an onshore flow pattern, resulting
from enhanced heating over land due to dust relative to the fixed (cooler)
sea surface temperature immediately offshore. This is consistent with the
previous study of Lau et al
. (2009).



A related issue is that the authors use interchangeably single scattering albedo
(SSA) and the imaginary part of the refractive index while discussing dust
absorption throughout the manuscript. They even state that OPAC and SF
refractive indice
s, which have highest imaginary part of the refractive index, will
be preferable. However, SSA depends not only on the refractive index but also
on the particle size distribution. In particular, the presence the coarse size mode
leads to lower SSA values f
or the same refractive index. GEOS
-
5 treats dust size
recent publications from SAMUM, AMMA and other field campaigns on the size
distribution of African dust, e.g. Otto et al 20
07, 2009). Using the full size
distribution, with fine and coarse modes considered, and applying OPAC (or SF)
refractive index will give much lower SSA values than those discussed in the
manuscript. So my argument is that the truncated size distribution in

GEOS
-
5 (or
other GCMs) may lead to an erroneous conclusion on high values of the
imaginary part of the refractive index.


The truncated size distribution also affects radiative fluxes in the IR, and thus the
assessment of net (SW+LW) dust radiative effec
ts.


In my opinion, the above arguments question the validity of the conclusion of
strongly absorptive dust with large values of the refractive index. In any case, I
would encourage the author to provide a more in
-
depth discussion addressing
the major rel
evant factors and processes to support the realism of their
conclusion.


The reviewer raises interesting and important points here about the dust
absorption and its sensitivity to particle size. We have largely folded these
points into our revision of th
e paper’s conclusion section, and please see
also the discussion in point (ii) below, which addresses mechanisms for
the positive feedback on emissions we find. The issue raised here is
entangled with that, so for efficiency we have combined our response
accordingly.


ii) The authors conclude that dust causes a positive feedback : (see Abstract)
"...Dust atmospheric heating enhances surface winds over important dust source
regions, increasing emissions, in contrast with previous modeling studies that
inclu
ded a coupled ocean..." citing work by Perlwitz and Miller (2001) and Miller
et al. (2004). There have been several other studies that used either GMCs or
regional models with radiatively interactive dust. To the best of my knowledge, all
past studies have

demonstrated that dust caused a negative feedback on the
dust emission.


Although varying SST (coupled ocean) can play a role as the authors suggest,
the negative feedback is mainly caused by the response of PBL to a decrease in
SW radiation reaching the

land surface under dust
-
laden conditions. See for
instance, Miller et al. (2004) and Perez et al. 2006 (JGR) for the discussion of the
mechanism involved. In particular, Perez et al. demonstrated a strong dust
negative feedback upon dust emission caused b
y radiatively coupled
atmospheric dust, leading to a strong reduction in AOD (by 35
-
45%). They used
a mesoscale model in a case study so the SST is not a factor.


Since the manuscript's finding is in direct conflict with all past studies, I would
suggest
the authors to provide a process
-
level explanation (or at least
hypothesis) in order to make their results credible.


This comment and the final part of comment (i) are we think intimately
related. In addressing these comments we performed
an
additional
experiment with our model

(the No
-
IR run)
.

T
his
experiment

is identical to
our OPAC
-
Spheres model run, except that we turned off the IR forcing of
dust aerosols by setting the extinction efficiency to zero in the IR bands
.


Miller et al. (2004
a
) and P
é
rez

et al. (2006) find that dust emissions are
reduced because of strong shortwave forcing caused by the dust and a
corresponding reduction in the sensible heat flux from the surface
,

leading
to an overall reduction in mixing within the planetary boundary lay
er (PBL)

and hence a weakening of the surface winds responsible for emissions.


A significant difference between their results and ours is the relative
importance of longwave forcing.

Miller et al. (2004a,b) use the same dust
refractive indices as in our
OPAC
-
Spheres case, but find a very small
longwave radiative forcing in the atmosphere and at the surface (see Table
3 in Miller et al. (2004b)).
We c
ompare their results to o
urs (Table 5 in our
manuscript):

They have shortwave surface forcing of
-
1.82 W
m
-
2 and
longwave surface forcing of 0.18 W m
-
2. We have a similar shortwave
forcing
,
-
1.62 W m
-
2
,
but a much greater longwave forcing of 0.67 W m
-
2,
leading to an overall much smaller net surface forcing of
-
0.95 W m
-
2
compared to their
-
1.64 W m
-
2. The
explanation for this can be found in our
different particle size distributions simulated. Figure R1 shows the global
annual mean particle size distribution from our study (black line) compared
to that from Miller et al. (2004b, solid grey line). Note tha
t their size
distribution is peaked at smaller particle sizes.

Comparison of our particle
size distribution to AERONET retrievals (Figure 12) shows reasonable
agreement, although we do not maintain the largest particles
AERONET
observes
. The Miller et al
. (2004a,b) results would presumably compare
less well.
Indeed, this deficiency was recognized, and in Miller et al. (2006)
a different formulation of the emitted particle size distribution was
incorporated, resulting in a similar particle size distributio
n to our own
(dashed grey line in Figure R1) and a relatively stronger longwave surface
forcing (0.49 W m
-
2, Figure 2c

in Miller et al. 2006
) and weaker net surface
forcing (
-
0.82 W m
-
2). Miller et al. (2006) did not revisit the question of dust
radiative

feedback on emissions, so we don’t know how that would be
different under the stronger longwave forcing

in their model
.


I
n our simulation we find a relatively stronger longwave forcing,
attributable to our simulating a particle size distribution shifted toward
larger particles than Miller et al. (2004a,b).
We do not have enough
information to evaluate the P
é
rez et al.
(2006) paper in quite the same way,
although we note they find an overall weak longwave forcing of dust (their
Figure 6) despite a very intense dust event simulated, perhaps indicating
they too are simulating small particle sizes.


In Figure R2 we show the

climatological JJA diurnal cycle of (a)
net dust
surface forcing
, (b)
sensible heat flux
, (c) friction speed, and (d) dust
emissions. This is given for the region 15 W


5 E, 16 N


24 N (roughly the
region of strongest enhancement in dust emissions for
our OPAC
-
Spheres
model run versus the
No Forcing

run, showing in Figure 16 in our
manuscript).
Three
model experiments are shown: No Forcing, OPAC
-
Spheres,
and
No
-
IR, with surface forcing only shown for
the difference of
the latter
two
against the No Forc
ing run. What this shows is that the
strong net dust surface forcing at mid
-
day is compensated by a partial
reduction in the sensible heat flux

as is shown in Miller et al. (2004a)

but
that because of the longwave effects in our model run there is a relat
ively
strong positive surface warming at night in the OPAC
-
Spheres model run.
This forcing is absent, as expected, for the No
-
IR run.

The effect of this
additional en
ergy in the boundary layer is a reduction in the amplitude of
the diurnal cycle in the f
riction speed. Peak friction speed is similar at mid
-
day for all model runs, but for
the
OPAC
-
Spheres model run
it is relatively
higher at nighttime. Correspondingly
,

there are much higher nighttime
dust emissions in the OPAC
-
Spheres model run than for e
ither the No
Forcing or No
-
IR runs. The No
-
IR model run has very similar emissions to
the No Forcing run
, and the climatological, global annual mean for that
model run is similar to the No Forcing model run (2116 Tg yr
-
1 versus 2107
Tg yr
-
1).


Our suggest
ion is thus that the difference between our results

and those

in
Miller et al. (2004a) is due to our relatively stronger longwave forcing,
which impacts the diurnal cycle of winds in the boundary layer and leads to
relatively greater dust emissions at nigh
t when dust radiative effects are
included. We suspect that if, as suggested by the reviewer above, we still
have too small of a particle mode relative to recent observations (e.g.,
those from
SAMUM and
Fennec) this effect may be further enhanced. It
may

be, however, that enhanced large particles to meet the Fennec
observations will yield adequate forcing for our less absorbing dust optics
(e.g., OBS) to give desirable transport patterns. This is a subject for future
research.


We have substantially revi
sed our conclusion section to make these points,
and Figure R2 is included in the paper as Figure 19.



Figure R1.

Climatological, global, annual column integrated particle size
distribution from our OPAC
-
Spheres model run (solid black line) compared
to those from Miller et al. (2004a,b, solid grey line) and Miller et al. (2006,
dashed grey line).



Figure R2
.
Diur
nal, climatological JJA regionally averaged net surface
forcing, sensible heat flux, friction speed, and dust emissions. Values are
the regional average over the area bounded by 15° W


5° E longitude, 16°
N


24° N latitude, the region of strongest enhan
cement in dust emissions
over the western Sahara shown in Figure 16. Three model simulations are
considered: our No Forcing case, our OPAC
-
Spheres case, and a
simulation identical to OPAC
-
Spheres except that the longwave dust
forcing is turned off.


Other comments:

1) Abstract, lines 32
-
33: "Absorption is thus the primary driver of interannual
variability in the dust aerosol lifecycle...."

This is a very general and misleading statement. One can argue that
atmospheric dynamics and its interannual va
riability will be the key factor
controlling dust emission, transport and removal (via rainout) and thus the
variability of dust lifecycle. I would suggest to revise this statement to clarify the
authors' point.


Please see the revised abstract, which inc
orporates previous comments.
We have removed this line.


2) p.11, lines 231
-
233: "...Dust is removed by sedimentation (calculated in the
CARMA module), or else by turbulent dry deposition and large
-
scale and
convective
-
scale wet removal, using algorithms
that were developed for other
aerosol packages implemented in GEOS
-
5 (see Colarco et al., 2010)..."

It will be helpful to provide better explanation of wet removal treatment. .Dust
differs from other aerosols and additional assumptions are required regard
ing the
dust solubility. Treatment of wet removal in GEOS
-
5 will affect the transport and
spatial distribution of dust.


We provide additional details about aerosol loss processes and how they
are formula
ted in our model in section 2.3 (lines 260


277):



The sedimentation fall speed is computed for each size bin following the
formulation for Stokes’ Law and accounting for noncontinuum effects at
low Reynold’s number flow (i.e., low air density or small particle sizes) after
Pruppacher and Klett (1997). T
his formulation assumes spherical particle
shapes
. We have neglected consideration of particle non
-
sphericity on this
term, which would have the effect of reducing the fall speed (see, e.g.,
Colarco et al., 2003; Ginoux 2003). Dry deposition follows the
resistance
-
in
-
series approach and accounts for sedimentation, surface type, and
meteorology following Wesely (1989). Wet removal is based on Liu et al.
(2001), which parameterizes large
-
scale in
-
cloud and below
-
cloud
scavenging as a function of rainfall p
roduction rate and precipitation
fluxes, respectively, and scavenging in convective updrafts as a function of
the updraft mass flux. As has been done in previous studies (e.g., Colarco
et al., 2010), for both large
-
scale and convective
-
scale wet removal
p
rocesses we tune the scavenging efficiency to be 0.3 for dust (it is
typically 1.0 for hygroscopic aerosols, as it is for the sea salt aerosols in
this study).



Additionally, as noted previously, we have acknowledged sensitivity of our
results to paramete
rizations of dust lifecycle processes in the conclusions
section.