Improved characterisation of sea ice using simultaneous aerial photography and sea ice thickness measurements

We present a semi-automatic classification algorithm for processing of aerial photographs of sea ice in combination with total (sea ice plus snow) sea ice thickness measurements from a helicopter-borne electromagnetic induction device. The algorithm is based on discriminant analysis of spectral and textural features in normalised images using a training set and includes five ice classes. Supervision and optional additional training sets can be used to further reduce misclassifications. The photographs were taken simultaneously with the sea ice thickness measurements and the results from the classification and the thickness data have been merged for more detailed analysis. The algorithm was tested using two case studies from flights over the summer pack ice north of Svalbard and in Fram Strait. The observed ice thickness distribution was very similar in both regions but the results from the classified images revealed considerable differences. The sea ice cover north of Svalbard contained a high fraction of melt ponds and only very little thin or brash ice. In Fram Strait, however, with predominantly adverted ice, the ice cover was subject to more ageing and deterioration and included significantly higher fractions of brash ice and submerged ice. The results from the algorithm in combination with the ice thickness provide new information about the state of the observed sea ice and the relationship of present ice types to ice thickness. (C) 2013 Elsevier B.V. All rights reserved.