eq 1 does not fit the data. This is surpris-

40

ing, given the good agreement shown in

Figure 10.

Weeks and Anderson

Ryvlin

At 200 cm the regression curve in Fig-

30

ure 11 gives a value of 5.50‰ for *S*B. Divid-

ing this salinity by *S*w = 31.5‰ gives a

value for *S*R of 0.175, not 0.13 as Ryvlin

20

determined from his data analysis. Using

this new *S*R value along with his suggested

growth rate coefficient of 0.5 in eq 1 brings

10

the Ryvlin curve into good agreement with

the regression curve between *T*F = 125 and

200 cm. However, there is still a significant

difference between eq 1 and the regression

0

40

80

120

20

60

100

curve where *T*F is less than about 50 cm. To

T F , Floe Thickness (cm)

bring the two curves into agreement, coef-

*Figure 10. Comparison of Ryvlin's empirical equation for esti- * ficient *a *has to vary with *T *. This is shown

F

*mating sea-ice bulk salinity vs. floe thickness, with field data * in Figure 13, where the lower curve repre-

*(after Ryvlin 1974).*

sents eq 1 with only *S*R changed from 0.13

to 0.175 and the upper curve represents

both the regression line through the data

20

in Figure 11 and eq 1 with *S*R = 0.175 and

2

S B = (2.242 + 20.556/TF)

coefficient *a *varied as shown.

2 = 0.840

r

To verify the need to vary *S*R and *a *in eq

S D = 1.45

1, I analyzed the *S*B vs. *T*F data collected by

15

Jeffries (1994) in the Amundsen and

Bellingshausen Seas, Antarctica, in late

winter. Unlike the predominantly congela-

10

tion sea-ice data from the Arctic (Beaufort

Sea) the Antarctic sea-ice data are from ice

sheets having a predominantly frazil, large

interlaced platelet and infiltrated snow ice

5

structure. Core data on ice up to 200 cm

thick with a mean temperature below 3C

S w = 31.5 ‰ (Beaufort Sea)

were selected. The latter ensured that the

0

50

100

150

200

T F, Floe Thickness (cm)

ice was "cold" and not melting. Their data

from 74 ice cores are shown in Figure 14.

*Figure 11. Beaufort Sea first-year ice bulk salinity vs. floe thick-*

T h e re p r e s e n t a t i v e re g r e s s i o n c u r v e

*ness. The lower curve represents the results of Ryvlin's equa-*

through the data is of the same form as the

*tion for 31.5‰ seawater.*

one through the Beaufort Sea ice data in

Figure 11. It is important to note that the re-

linity difference caused an 8% difference in the

gression curve at *T*F = 200 cm gives *S*B = 5.41

bulk sea-ice salinity.

The regression curve through the Beaufort Sea

vs. 5.50‰ for the Beaufort Sea data. Given the

*S*B vs. *T*F data (Fig. 11) is, for all intents and pur-

fact that the Antarctic seawater salinity is 34‰*

poses, horizontal after an ice thickness of about

vs. 31.5‰ for the southern Beaufort Sea, one

150 cm. Therefore, *S*BF may be determined from

would expect these *S*B values to be reversed for

the regression curve at any *T*F after 150 cm with

reasons previously discussed. This unexpected

minor effect on the results of eq 1. For consis-

result suggests that the sea-ice data for the

tency, *S*BF at *T*F = 200 cm is used in this report.

Running below the regression line, and most

* M. Jeffries (1995), Geophysical Institute, Univer-

of the data, in Figure 11 is the curve derived using

sity of Alaska, Fairbanks, and S. Jacobs (1995),

eq 1 and an *S*w of 31.5‰ (Fig. 12). Even though

Lamont-Doherty Geological Observatory of Colum-

the number of data points is limited (and non-

bia University, Palisades, New York, personal com-

existent for a *T*F of 25 to 50 cm) it is apparent that

munications.

7