Automated Mapping of Rapid Arctic Ocean Coastal Change Over Large Spans of Time and Geography EP33B-0849 1379672 While climate change is global in scope, its impacts vary greatly from region to region. The dynamic Arctic Ocean coastline often shows greater sensitivity to climate change and more obvious impacts. Current longer ice-free conditions, rising sea level, thawing permafrost, and melting of larger ice bodies combine to produce extremely rapid coastal change and erosion. Anderson et al. (2009; Geology News) have measured erosion rates at sites along the Alaskan Arctic Ocean coast of 15 m per year and greater. Completely understanding coastal change in the Arctic requires mapping both current erosional regimes as well as changes in erosional rates over several decades. Studying coastal change and trends in the Arctic, however, presents several significant difficulties. The study area is enormous, with over 45,000 km of coastline; it is also one of the most remote, inaccessible, and hostile environments on Earth. Moreover, the region has little to no historical data from which to start. Thus, any study of the area must be able to construct its own baseline. Remote sensing offers the best solution given these difficulties. Spaceborne platforms allow for regular global coverage at temporal and spatial scales sufficient for mapping coastal erosion and deposition. The Landsat family of instruments (MSS, TM, and ETM) has data available as frequently as every 16 days and starting as early as 1972. The data are freely available from the USGS through earthexplorer.usgs.gov and are well calibrated both geometrically and spectrally, eliminating expensive pre-processing steps and making them analysis-ready. Finally, because manual coastline delineation of the quantity of data involved would be prohibitive in both budget and labor, an automated processing chain must be used. ENVI Feature Extraction can provide results in line with those generated by expert analysts (Hulslander, et al., 2008; GEOBIA 2008 Proceedings). Previous studies near Drew Point, Alaska have shown that feature extraction software can provide the quality and consistency necessary for good results (Hulslander, 2010; AGU Fall Meeting) and can be integrated with a GIS for automated training and classification (Hulslander, 2011; AGU Fall Meeting). Here, our preliminary results show that automated feature extraction can be used to map coastlines as well as coastal ice margins over decades and across the region.
David Hulslander Exelis VIS, Boulder, CO, USA