Sensitivity of Arctic Sea Ice Thickness to Intermodel Variations in the Surface Energy Budget - Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications - page 80

without considering variations in growth season temperature
and energy fluxes.

Table 2. Models Enumerated as in figures 3a and 5
Model center country

1, cccma_cgcm3_1 canadian centre for

climate Modeling and
Analysis

canada

4.2. Realism of 20C3M Surface Energy Fluxes

As noted in section 3, the ensemble mean value of Ŝ is
the critical factor in relating the spread in h D to the spread
in surface fluxes. If the ensemble mean Ŝ is close to its real-
world counterpart, then we have some confidence that the
ensemble spread in h D correctly represents the uncertainty in
h D which results from errors in Ŝ .
to assess the accuracy of Ŝ as simulated by the climate
models, table 3 compares the values of MJJA mean surface
fluxes from L98's analysis of Russian ice station data (21 sta-
tions over 45 annual cycles), Persson et al. 's [2002] analysis
of data from the 1997-1998 Surface Heat Budget of the Arc-
tic (SHEBA) field campaign, and the mean of the 17 mod-
els in figure 3. While discrepancies exist in the individual
fluxes, the ensemble mean S agrees quite closely in all three
data sets, 36 (SHEBA), 35 (l98), and 37 (20c3M ensemble
mean) W m -2 . It should be noted that the l98 albedo is high
(0.74) because the value does not include ponds, thin ice, or
leads. their high albedo value may lead to their higher value
of F SW ¯ (234 compared to 214 W m -2 for Persson et al.), since
albedo is used in the formula from which they calculate F SW ¯ .
to account for melt ponds and leads in the formula for F SW ¯ ,
they reduce the albedo value in the formula by 0.05 for July
and August, which may be an insufficient reduction. We
note, however, that the net shortwave F SW agrees closely
between l98 (59 W m -2 ) and Persson et al. (60 W m -2 ), so

2, cccma_cgcm3_1_t63 canadian centre for

climate Modeling and
Analysis

canada

3, cnrm_cm3 centre national de

recherches
Météorologiques

france

4, gfdl_cm2_0 Geophysical fluid

dynamics laboratory

united
States

5, gfdl_cm2_1 Geophysical fluid

dynamics laboratory

united
States

6, giss_aom Goddard Institute for

Space Studies

united
States

7, iap_fgoals1_0_g Institute of

Atmospheric Physics

china

8, inmcm3_0 Institute for numerical

Mathematics

russia

9, ipsl_cm4 Institute Pierre Simon

laplace

france

10, miroc3_2_hires center for climate

System research

Japan

11, miroc3_2_medres center for climate

System research

Japan

12, miub_echo_g Meteorological

Institute of the
university of Bonn

Germany

13, mpi_echam5 Max Plank Institute for

Meteorology

Germany

14, mri_cgcm2_3_2a Meteorology research

Institute

Japan

Table 3. comparison of Surface Energy flux Quantities for MJJA
Between observations from SHEBA [ Persson et al. , 2002], Lind-
say [1998], and the 20c3M Ensemble Mean, Plus Maximum and
Minimum Quantities from 20c3M a
F SW ¯

15, ncar_ccsm3_0 national center for

Atmospheric research

united
States

16, ukmo_hadcm3 Hadley center for

climate Prediction and
research

united
kingdom

F LW ¯

F SW F LW F RAD S cloud

(%)

a
(%)

17, ukmo_hadgem1 Hadley center for

climate Prediction and
research

united
kingdom

SHEBA 214 282 60 -20 40 36 - 61
lindsay 234 284 59 -16 43 35 85 74
20c3M 202 273 76 -28 47 37 83 61
Max 248 289 106 -17 72 61 93 72
Min 176 250 50 -38 12 13 66 46
a F RAD is the net surface radiation and S = ( F RAD + F SL ) M . Maximum
(Max) and minimum (Min) values for F Sw and F LW are not con-
strained to sum to the Max and Min F RAD values, since the extreme
values for F SW and F LW will not, in general, come from the same
simulation. Albedo from 20c3M is the ratio of monthly mean F SW
and F SW ¯ . Albedo from SHEBA is from the albedo line measure-
ments. Albedo from lindsay does not include leads or melt ponds.
fluxes are positive down, i.e., positive when the surface gains en-
ergy from above. Values in F LW and F SW columns are net fluxes.

to relate the ensemble spread of h D to the spread of Ŝ , we
scatter h D against Ŝ in figure 3b (asterisks). As expected,
there is a strong inverse relationship between Ŝ and h D . the
plus signs show the values of h D obtained using the ensem-
ble mean value of - kT G (t G /t M ) in (4) so that h D becomes a
unique function of Ŝ . According to figure 3b, the ensemble
spread of h expected from the flux variations is between 1
and 5 m (asterisks), of which a spread of 1 to 4 m in thick-
ness can be explained by Ŝ variations alone (plus signs),