Thermal conductivity of landfast Antarctic and Arctic sea ice

Abstract

We present final results from a program to measure the thermal conductivity of sea ice with in situ thermistor arrays using an amended analysis of new and previously reported ice temperatures. Results from landfast first‐year (FY) ice near Barrow, Alaska, and McMurdo Sound, Antarctica, are consistent with predictions from effective‐medium models but 10–15% higher than values from the parameterization currently used in most sea ice models. We observe no previously reported anomalous near‐surface reduction, which is now understood to have been an artifact, nor a convective enhancement to the heat flow, although our analysis is limited to temperatures below −5°C at which brine percolation is restricted. Results for landfast multiyear (MY) ice in McMurdo Sound are also consistent with effective‐medium predictions, and emphasize the density dependence. We compare these and historical measurements with effective‐medium predictions and the representation commonly used in sea ice models, developed originally for MY Arctic ice. We propose an alternative expression derived from effective‐medium models, appropriate for both MY and FY ice that is consistent with experimental results, k = (ρ/ρi)(2.11 − 0.011 θ + 0.09 (S/θ) − (ρ − ρi)/1000), where ρi and ρ are the density of pure ice and sea ice (kg m−3), and θ (°C) and S (ppt) are sea ice temperature and salinity. For the winter and spring conditions studied here, thermal signatures of internal brine motion were observed rarely (22 times in 1957 days), and their maximum contribution to the total heat flow is estimated to be of the order of a few percent.