Ekaterina Kim

There are extreme challenges unique to the Arctic, from human activities and impacts in remote Arctic locations to Arctic data acquisition, sharing, and quality. The amount of Arctic data is growing. In fact, this growth is faster than the capacity of experts to process, adequately validate, and evaluate all uncertainties in the data. Ocean sciences and maritime industry-oriented applications in the Arctic regions can benefit from learning on previous data and from cross-disciplinary expert knowledge. Despite rapid progress in artificial intelligence (AI) and data science, AI applications to Arctic science, engineering, and technology (e.g., automated validation of remote sensing data from the Arctic, learning from past humans’ activities in the ice infested waters) have received less scientific attention in comparison to the fields of finance, logistics, medicine, advertisement, etc. This worldwide trend may be attributed to the data quality, availability, and the expertise that is needed to process these data. There is a strong need to direct AI applications towards solving Arctic challenges. In this talk I would like to present two examples on how artificial intelligence can help in automated interpretation of ice imagery from ground operations (e.g., imagery from surface vessels, shore stations, in-situ campaigns) and how it can help us in learning from historical AIS data (e.g., from the Kara Sea region). Data processing algorithms underlying the presented examples have been made publicly available on GitHub (i.e., ice image classification, ice image segmentation, and AIS data processing).