20

15

10

S w = 34 ‰ (Amundsen and Bellingshausen Seas)

5

S w = 31.5 ‰ (Beaufort Sea)

*Figure 12. Sea-ice bulk salinity vs. floe thickness*

*as calculated with Ryvlin's equation for 31.5 and*

0

50

100

150

200

*34‰ seawater.*

TF , Floe Thickness (cm)

20

0.220 = a

15

0.320

10

0.375

0.400

a = Varied

0.450

*Figure 13. Ryvlin's equation for calculating sea-*

a = 0.5

*ice bulk salinity vs. floe thickness where param-*

5

*eter *a is constant (lower curve) and varied (up-

*per curve) to match the regression curve through*

0

50

100

150

200

*the Beaufort Sea ice data in Figure 11.*

T F , Floe Thickness (cm)

Arctic (Fig. 11) and Antarctic (Fig. 14) are not

*S*BF = 5.41‰ at *T*F = 200 and dividing this salinity

sufficient to fully define the *S*B vs. *T*F trends or

by *S*w = 34‰ gives an *S*R of 0.159. Using this

that the effect of the seawater salinity differ-

value in eq 1 gives the lower curve in Figure 15.

Varying coefficient *a *at the *T*F increments shown

ence on *S*B is masked by the scatter in the data

and/or ice structure differences. Perhaps as im-

in Figure 15 causes the Ryvlin equation to give *S*B

portant is the regression curve selected to repre-

vs. *T*F values that match the regression curve

sent the data. Statistically, two regression curves

through the Antarctic data (Fig. 14). Both the modi-

may fit the data with nearly the same correlation

fied Ryvlin and regression curves are represented

coefficient *r*2 value, but they give different *S*B val-

by the upper curve in Figure 15.

ues at 200 cm. In any event, since the regression

The fact that growth rate coefficient *a *needs

curves (Fig. 11 and 14) lie well within each other 's

to vary with *T*F is not surprising. Ryvlin indicated

standard deviation, there is no statistically sig-

as much in his description of *a *in eq 1. A further

nificant difference between them. Given this rea-

indication of the need to vary *a *can be seen in

soning, it would seem that the Arctic and Antarc-

Figure 8. Here the time to grow successive incre-

tic sea ice *S*B vs. *T*F data can be combined.

ments of ice is shown to increase with thickness.

Also shown in Figure 14 is the curve derived

In short, the heat flux through growing sea ice

from unmodified eq 1 (Fig. 12). As with the Beau-

decreases exponentially with thickness and like-

fort Sea data, the curve does not fit the data.

wise reduces the rate of ice growth (Makshtas

Taking the regression curve (Fig. 14) value for