Search

Automated sea ice charting from synthetic aperture radar (SAR) has been researched for more than a decade, and we are still not close to unlocking the full potential of automated solutions in terms of resolution and accuracy. The central complications arise from ground truth data not being readily available in the polar regions. In this paper, we build a data set from 20 near-coincident x-band SAR acquisitions and as many airborne laser scanner (ALS) measurements from the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC), between October and May. This data set is then used to assess the accuracy and robustness of five machine-learning-based approaches by deriving classes from the freeboard, surface roughness (standard deviation at 0.5 m correlation length) and reflectance. It is shown that there is only a weak correlation between the radar backscatter and the sea ice topography. Accuracies between 44 % and 66 % and robustness between 71 % and 83 % give a realistic insight into the performance of modern convolutional neural network architectures across a range of ice conditions over 8 months. It also marks the first time algorithms have been trained entirely with labels from coincident measurements, allowing for a probabilistic class retrieval. The results show that segmentation models able to learn from the class distribution perform significantly better than pixel-wise classification approaches by nearly 20 % accuracy on average. [ABSTRACT FROM AUTHOR]

Airborne laser scanners (ALS) are used to map the sea-ice surface at sub-meter resolution. We conducted 64 flights over the Arctic sea ice between September 2019 and September 2020 during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition to measure sea-ice surface elevation. The flights ranged from repeated, local-scale 5 × 5 km2 floe grid surveys to regional-scale transects more than 100 km long. We provide data at different processing levels: geolocated elevation point clouds and gridded segments of elevation and freeboard with a spatial resolution of 0.5 m. The latter product is corrected for atmospheric backscatter, sea-ice drift, and offset in elevation due to degraded INS/GNSS solutions > 85° N. For floe grid surveys, all data are combined to merged two-dimensional elevation maps. Other provided parameters include laser reflectance and echo width. The presented data offer a unique possibility to study the temporal evolution, spatial distribution, and variability of the snow and sea-ice surface and their properties in addition to validating satellite products. [ABSTRACT FROM AUTHOR]

The Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition took place between October 2019 and September 2020 giving the rare opportunity to monitor sea-ice properties over a full annual cycle. Here we present 24 high-resolution orthomosaics and 14 photogrammetric digital elevation models of the sea-ice surface around the icebreaker RV Polarstern between March and September 2020. The dataset is based on >34.000 images acquired by a helicopter-borne optical camera system with survey flights covering areas between 1.8 and 96.5 km2 around the vessel. Depending on the flight pattern and altitude of the helicopter, ground resolutions of the orthomosaics range between 0.03 and 0.5 m. By combining the photogrammetric products with contemporaneously acquired airborne laser scanner reflectance measurements selected orthomosaics could be corrected for cloud shadows which facilitates their usage for sea-ice and melt pond classification algorithms. The presented dataset is a valuable data source for the interdisciplinary MOSAiC community building a temporal and spatially resolved baseline to accompany various remote sensing and in situ research projects. [ABSTRACT FROM AUTHOR]