Winter sea ice characteristics in the central Arctic from thermal infrared imaging

Abstract

This dissertation is based on helicopter-borne thermal infrared (TIR) imaging performed during the MOSAiC expedition in the central Arctic in winter. The aim is to increase the understanding of processes influencing the Arctic heat budget. TIR imaging benefits from the large temperature differences between the colder thick sea ice and the warmer leads of open water or covered with thin ice. Between October 2019 and April 2020, 35 helicopter flights were used for investigation on the local (5-10 km) and regional scale (20-40 km). From several thousands of images, surface temperature maps at 1 m resolution were created for each flight. This thesis focuses on (i) melt pond pre-conditioning based on warm temperature anomalies, (ii) lead classification resulting in area fraction, width distribution, and orientations down to the spatial scale of meters, and (iii) the relevance of the satellite sub-footprint scale variability of surface temperatures for the heat exchange. Melt ponds are a crucial part of the summer heat budget as they lower the surface albedo and contribute to the positive ice-albedo feedback. The comparison of winter temperature and summer images from the same ice revealed the presence of warm anomalies of 0.3-2.5 K at the melt pond locations of the subsequent summer. This pre-conditioning enabled a correct seasonal prediction of 41% of the melt ponds. Leads are relevant for the winter heat budget as they allow for an increased heat transfer from the warmer ocean to the colder atmosphere. The determined lead area fraction is 1.2% on average and up to 4% on 24 December 2019, with stronger variability on the local than on the regional scale. The power law distribution of lead width has a negative exponent of 2.63 that is valid to a width of 3 m. Consequently, there is an exponentially larger number of narrow than wider leads. The numerous small-scale leads are not resolved in operationally used satellite products. Due to the lack of sub-footprint scale variability the sensible heat flux derived from these products is underestimated. The decrease of the average sensible heat flux of up to 0.69 W m⁻² between the 1 m and re-gridded 1 km resolution helicopter data has a linear relation with the lead area fraction. A comparison between the overlapping helicopter and satellite surface temperatures of MODIS showed no alignment in spatial variability. The thesis’ findings provide results to improve satellite retrievals and model parameterizations.