Assessment of warm bias in ERA5 surface skin temperature over Arctic landfast sea ice

ERA5 is one of the most widely used gridded reanalysis datasets; however, its 0.25° spatial resolution may not be sufficient to resolve the narrow band of landfast sea ice near coastlines, possibly leading to the misrepresentation of the thermodynamic forcing of sea ice growth and melt. In this study, the performance of ERA5 surface skin temperature (SKTERA5) over landfast sea ice was assessed through comparison with in situ observations of surface skin temperature (SKTOBS) that were collected during winter, spring, and early summer as a part of nine field campaigns at seven different locations in the Canadian Arctic. The diurnal variability and seasonal trends were captured by ERA5, yet SKTERA5 exhibited a warm bias for most of the study sites. This bias varied by site, ranging from 0.06°C to 5°C in the pre-melt period (i.e., before SKTERA5 first reached 0°C) and from −0.06°C to 2.7°C during the melt period (after SKTERA5 first exceeded 0°C). ERA5 warm biases resulted in underestimation of sea ice thickness by up to 9% estimated using an analytical sea ice growth model. During the melt period, before the occurrence of liquid water on the sea ice surface, the ERA5 ocean grid points provided a good representation of SKTOBS. When there was liquid water on the sea ice surface and SKTOBS remained consistently near or at 0°C, SKTERA5 became unreliable. ERA5 warm bias for landfast sea ice could be reduced by addressing the overestimation of the ice-free area fraction and omission of the land cover area fraction within the grid box. These findings highlight the need to conduct a careful evaluation of the environmental context before using ERA5 data. They also provide a foundation for determining the most accurate SKTERA5 for thermodynamics studies of landfast sea ice.