Sensitivity of Arctic Sea Ice Thickness to Intermodel Variations in the Surface Energy Budget - Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications - page 79

Figure 3. (a) Mean September to April thickness for grid points with perennial Arctic sea ice from model output (light shad-
ing) and from diagnostic calculation (dark shading), described in section 4.1. Model output is obtained from the Program
for climate Model diagnosis and Intercomparison, and data are from run 1 of each model. (b) Scatter of Ŝ (denominator
in (4)) and diagnostic thickness h D from 20c3M model output (asterisks), h D calculated using the ensemble mean value of
the numerator in (4)(plus signs), and h D as a continuous function of Ŝ using the ensemble mean value of the numerator in
(4) (solid line).

Model diagnosis and Intercomparison, and data are from
run 1 of each model (models are listed in table 2). for most
models, the diagnostic calculation produces a thickness esti-
mate which is loosely consistent with the actual model out-
put, the most obvious exception being model 7. there are, of

course, a variety of factors neglected in (4) which could ac-
count for discrepancies between h and h D , like the insulating
effect of snow cover, the linearization of the SB equation,
and the use of the ratio seasonal-mean h and T values instead
of the seasonal mean of ( T / h ) in (1).