| Title | RADARSAT SAR applied to ice margin mapping: The Barnes Ice Cap, Nunavut, Canada |
| |
| Author | Short, N H ;
Simms, É L; Jacobs, J D |
| Source | Canadian Journal of Remote Sensing vol. 26, no. 2, 2000 p. 91-102, https://doi.org/10.1080/07038992.2000.10874759 |
| Year | 2000 |
| Alt Series | Natural Resources Canada, Contribution Series 20181910 |
| Publisher | Informa UK Limited |
| Document | serial |
| Lang. | English |
| Media | paper; on-line; digital |
| File format | pdf |
| Province | Nunavut |
| Subjects | geophysics; remote sensing |
| Program | Canada Centre for Remote Sensing Divsion |
| Released | 2014 07 31 |
| Abstract | The ice margin or 'limit of active ice' is extracted from RADARSAT imagery using a sequence of image speckle filtering, image texture analysis, supervised image classification, image segmentation and
edge detection. The limit of active ice for a section of the Barnes Ice Cap is accurate to within 49 m. Overall classification accuracies of the ice marginal environment are between 44 and 53%. Despite misclassification of some proglacial landforms,
it is possible to detect the limit of active ice as the ice cap is fringed by supraglacial debris cover, an elevated ice cored debris ridge and perennial snowbeds. Analysis reveals that grey level cooccurence matrix texture measures mean, homogeneity
and correlation are effective variables for classification of ice margins in RADARSAT imagery. The size of the texture window should be as large as the smallest feature it is hoped to identify, in this case approximately 85 × 85 m for the mean and
homogeneity measures and 170 × 170 m for the correlation measure. Comparisons of opposing look angles reveal that 'downglacier ' illumination produces the highest classification accuracies of ice marginal features and debris-covered ice surfaces,
whereas illumination from an off-ice perspective, looking 'upglacier', is found to produce higher classification accuracies for proglacial surfaces. Comparisons of standard and fine mode imagery show that 25 m spatial resolution yields higher
classification accuracies than 8 m spatial resolution. This is primarily due to increased confusion of supraglacial debris cover with bedrock and residuum surfaces at the finer spatial resolution. |
| GEOSCAN ID | 312265 |
|
|