Geoscience Reference
In-Depth Information
the combination of high albedo and high downwelling short-
wave radiation apparently yields a net shortwave estimate in
agreement with the SHEBA data.
A more detailed comparison of the 20c3M surface energy
fluxes is presented in Figures 4a-4e, which shows the an-
nual cycle of surface radiative fluxes and turbulent (latent
and sensible) fluxes from L98 and the 20C3M ensemble.
All fluxes have relatively large ensemble spread, and down-
welling longwave and shortwave fluxes are underestimated
in the ensemble mean compared to l98. l98 computes
downwelling shortwave radiation using a formula which
accounts for the dependence of
F
SW
¯
on surface albedo, as
higher albedo leads to increased multiple scattering and
hence increased downwelling shortwave radiation [
Shine
,
1984; see also dHH]. As noted above, the l98
F
SW
¯
val-
ues may be overestimates because of the treatment of melt
ponds, and are high compared to Persson et al.'s SHEBA-
derived values.
for net radiation (
F
SW
+
F
LW
, figure 4c), the ensemble
mean cold season (SONDJFMA) flux is too low by 6 to 9 W
m
-2
. The ensemble mean overestimates the net flux in June
by 13 W m
-2
, but there is relatively close agreement between
the ensemble mean and l98 in the remainder of the warm
season. June is also the month of largest ensemble spread (96
W m
-2
), possibly because of intermodel differences in melt
onset date. For surface turbulent flux, the ensemble mean is
in reasonable agreement with l98, capturing the minima in
May-June and in August. the ensemble mean is below the
l98 value in most months, with a maximum underestimate
of 6 W m
-2
in January. Radiative and turbulent fluxes are
combined into a net surface flux in Figure 4e, which closely
resembles the net radiative fluxes. The ensemble spread in
turbulent fluxes is of the same order as the annual cycle.
Since the ensemble mean underestimates both radiative and
turbulent fluxes in the cold season, the maximum underes-
timate in figure 4e is greater than that in figure 4c at 13 W
m
-2
in January. the June overestimate is 11 W m
-2
, accom-
panied by the largest spread, 81 W m
-2
.
Figures 4f-4h compare nonflux quantities of relevance to
the surface energy budget. the ensemble spread in surface
albedo (figure 4f) is quite high, with melt season values
ranging from 0.56 to 0.81 in May, 0.45 to 0.8 in June, and
0.4 to 0.65 in July and August. the largest spread is in June,
also the month of maximum spread in net radiation. l98's
albedo values are considerably higher, presumably because
his values do not include leads and melt ponds. cloud frac-
tion (figure 4g) also shows large ensemble spread, par-
ticularly in winter, with values ranging from 34 to 95% in
January. Arctic cloud biases are a well-known shortcoming
of climate models, as noted by
Vavrus
[2004],
Walsh et al.
[2002], and
Beesley and Moritz
[1999] [see also
Gorodets-
kaya and Tremblay
, this volume]. the spread in cloud frac-
tion is smallest in the melt season, with values between 65
and 95% in JJA and good agreement between the ensemble
mean and l98. from the simple model perspective, the fact
that the best agreement occurs over the melt season is benefi-
cial, since the impact of cloud biases on
Ŝ
and the spread of
h
D
is thereby minimized. However, cloud radiative forcing is
more dependent on cloud microphysics than on cloud frac-
tion [e.g.,
Gorodetskaya et al.
, 2006; see also
Gorodetskaya
and Tremblay
, this volume], so the consequences of cloud
biases for the radiation budget cannot be assessed from
cloud fraction alone.
finally, the surface air temperature from l98 and the
20c3M ensemble are compared in panel
h
. the comparison is
motivated by the close association between surface tempera-
ture and surface flux, and earlier findings of cold biases in cli-
mate models [e.g.,
Randall et al.
, 1998]. the ensemble mean
is close to l98, with differences between ±2°c and no clear
preference for colder temperature. from these comparisons we
conclude that while monthly flux values for individual models
can disagree dramatically with l98, the ensemble mean net
fluxes are in reasonable agreement with observations.
the budget residual
R
, which includes ocean heat flux con-
vergence and latent heat transported in sea ice motion, plotted
in figure 5, has values that are generally between 0 and 10 W
m
-2
. observations are not available to verify
R
, but the con-
tribution of ocean heat flux could, in principle, be calculated
from the 20c3M ocean model output. for ccSM3, the heat
flux due to ice motion was found to be about 5 W m
-2
in a long
present-day control run (run b30.009, for which thickness
tendency due to dynamics is available), and dHH obtained a
value of 4.4 W m
-2
in a version of ccSM3 with a slab ocean
but full ice dynamics. t92 gives a range of reasonable ocean
heat flux convergence values of 2 to 10 W m
-2
. notable fea-
tures in figure 5 are the extreme value for model 9 (about 15
W m
-2
) and two models with negative
R
values, presumably
indicating that the ocean is exporting heat from the Arctic.
following
Maykut and Untersteiner
[1971], EuW and
t92 assume that shortwave and longwave radiation domi-
nate the surface energy budget, and changes in
S
are com-
pensated exclusively by changes in
W
as shown in figure
2. the extent to which the surface energy budget and its
intermodel variation in the 20c3M ensemble is dominated
by radiative fluxes is examined in Figure 6a, in which the
annual mean net shortwave and longwave fluxes are scat-
tered against each other. With the exception of models 1
and 2, the assumption holds to a reasonable degree, and the
correlation between shortwave and longwave fluxes exclud-
ing these two models is 0.75 (note also the large spread in
radiative fluxes, from 20 W m
-2
to over 40 W m
-2
). for the
outlier models (models 1 and 2, which are actually the same
Search WWH ::
Custom Search