The dynamics of a surge is manifested in the crevasse patterns: literally, deformation state frozen in ice. This basic observation is utilized as the concept of an automated approach to map and analyze deformation stages and progression of surge kinematics. The classification method allows imagery to be used as geophysical data and is applied to aerial observations (photographic and video imagery, GPS data) collected in September 2011 during the surge of the Bering Glacier–Bagley Ice Valley system, Alaska, USA. As the third dimension that complements two-dimensional imagery, ice-surface elevation is observed using aerial laser altimetry. The classification method builds on concepts from signal processing, geostatistical data analysis and neural networks. Steps include calculation of generalized directional vario functions from image data and composition into feature vectors. The vario function operates as an information filter that retains spatial characteristics at an intermediate scale that captures crevasse spacing, anisotropy and other generalized roughness properties. Association of feature vectors to crevasse classes and hence deformation types employs a connectionist algorithm. In general, the connectionist–geostatistical classification allows the mapping of kinematic changes in crevassed glaciers.Dedicated to the memory of Austin Post

ABSTRACT. Our objective is to map dynamic provinces and investigate dynamic changes in Jakobshavn Isbrae, Greenland. We use an approach that combines structural glaciology and remote-sensing data analysis, facilitated by mathematical characterization of generalized spatial surface roughness that provides parameters related to ice dynamics, deformation and interaction of the ice with bed topography. The approach is applied to derive time series of elevation and roughness changes and to attribute changes during rapid retreat.