CIRFA welcomes you to a seminar where visiting scientist Cecilia Peralta-Ferriz will talk about her work using data from the GRACE satellite and CIRFA PhD student Johannes Lohse will talk about his work on feature selection for ice classification.
Date: Thursday 9 November at 14h00-15h00
Venue: CIRFA, SIVA Innovation Centre (Forskningsparken), Tromsø [map]
Online: We will try to set up live streaming of the seminar. Access it here: [link].
Arctic Ocean bottom pressure from GRACE: Lessons about ocean circulation and the flow variability through Bering Strait
Cecilia Peralta-Ferriz, Polar Science Center, Applied Physics Lab., University of Washington (Currently at NPI for 9 months, until May 2018).
Satellite observations of ocean bottom pressure (OBP) from the Gravity Recovery and Climate Experiment (GRACE), available since 2002, have become an invaluable tool for investigating Arctic environmental changes. This presentation provides an overview of what we have learned about the Arctic Ocean using GRACE, including the patterns of variability in ocean circulation and their associated atmospheric forcing; the combination of GRACE OBP with altimetry-derived sea surface height variability, resulting in changes and distribution of the freshwater content of the Arctic Ocean; and more recently, the combination of GRACE OBP with mooring data from Bering Strait from 2002 to 2016, to investigate the driving mechanisms of the oceanic flow into the Arctic through the strait. The Bering Strait flow is known to be driven by local winds and a (poorly defined) far-field “pressure-head” forcing, typically related to sea-surface-height differences between the Pacific and the Arctic. We identify the spatial structure of this pressure-head forcing, finding that the variability of the Bering Strait throughflow is predominantly driven from the Arctic (not the Pacific), specifically by sea-level changes in the East Siberian Sea, and related to westward winds along the Siberian shelves. This pressure-head forcing explains ~70% of the Bering Strait flow variability during the summer, when the local winds are weak. During the winter, however, both the local winds and the pressure-head forcing are important.
SAR Feature Selection for (Sea Ice) Classification
Johannes Lohse, PhD student, UiT – The Arctic University of Norway
SAR data provide an excellent tool for sea ice observations and are the major data source for ice chart production. With more polarisations we obtain more information about the ice and can extract many features from the data. When trying to automate sea ice classification from SAR data, it therefore becomes necessary to reduce the dimensionality of the data and select the best features only.
In this study we look into automatic selection of feature sets based on classification accuracy. We then suggest to split up a multi-class classification problem into many two-class problems and investigate the benefit of selecting different feature sets for each binary decision. The method is developed on synthetic datasets and then applied to airborne SAR data.