Laboratory study of initial sea-ice growth: properties of grease ice and nilas

Abstract. We investigate initial sea-ice growth in an ice-tank study by freezing an NaCl solution of about 29 g kg⁻¹ in three different setups: grease ice grew in experiments with waves and in experiments with a current and wind, while nilas formed in a quiescent experimental setup. In this paper we focus on the differences in bulk salinity, solid fraction and thickness between these two ice types.

The bulk salinity of the grease-ice layer in our experiments remained almost constant until the ice began to consolidate. In contrast, the initial bulk-salinity evolution of the nilas is well described by a linear decrease of about 2.1 g kg⁻¹ h⁻¹ independent of air temperature. This rapid decrease can be qualitatively understood by considering a Rayleigh number that became maximum while the nilas was still less than 1 cm thick.

Comparing three different methods to measure solid fraction in grease ice based on (a) salt conservation, (b) mass conservation and (c) energy conservation, we find that the method based on salt conservation does not give reliable results if the salinity of the interstitial water is approximated as being equal to the salinity of the underlying water. Instead the increase in salinity of the interstitial water during grease-ice formation must be taken into account. In our experiments, the solid fraction of grease ice was relatively constant with values of 0.25, whereas it increased to values as high as 0.50 as soon as the grease ice consolidated at its surface. In contrast, the solid fraction of the nilas increased continuously in the first hours of ice formation and reached an average value of 0.55 after 4.5 h.

The spatially averaged ice thickness was twice as large in the first 24 h of ice formation in the setup with a current and wind compared to the other two setups, since the wind kept parts of the water surface ice free and therefore allowed for a higher heat loss from the water. The development of the ice thickness can be reproduced well with simple, one dimensional models that only require air temperature or ice surface temperature as input.