Comparison of satellite imagery and infrared aerial photography as vegetation mapping methods in an Arctic study area; Jameson Land, East Greenland

Abstract

The two methods suited for mapping the vegetation of remote inaccessible terrain were compared in terms of their ability to reproduce and distinguish predefined vegetation classes, the appropriateness of their different spatial resolutions to the same end, and their cost-effectiveness. The study was carried out in Jameson Land, East Greenland (70° 30?N to 72°N). The vegetation had previously been classified and mapped by the use of false colour infrared aerial photos in the period 1982–86. The resultant vegetation maps had been drawn on aerial photo overlays at a scale of 1:25, 000. SPOT data from 8 September 1986 and Landsat TM data from 10 August 1987 were used. A supervised classification of the combined satellite data was made using a maximum likelihood algorithm. A vegetation map was produced which accurately plotted 9 of the dominant vegetation classes and 4 non-vegetation classes. The two mapping methods were then compared. The satellite-based method was found to be inadequate for the mapping of vegetation classes which occur in very small vegetation patches. The photo-based mapping had better potential to map these vegetation classes because aerial photos have a higher spatial resolution than the satellite data. Small patches of herb slopes were selectively pinpointed in the photo-based mapping process. As the dominant vegetation types were, however, plotted in less detail, the photo-based mapping generally produced larger vegetation units with a more homogeneous and less complex texture than the satellite-based method. Of the two methods, the satellite-based method was by far the more cost-effective. Satellite data were less expensive to obtain and process than aerial photos and time-consuming interpretation was eliminated. Both methods require detailed ground-truthing, but the satellite-based method requires less so due to the objective classification procedure. Overall, the satellite-based method seems advantageous for vegetation mapping in large remote areas. However, the satellite data still lacks the resolution to map all the important cover classes of the fine-scaled arctic tundra vegetation mosaic. Therefore purpose-specific adaptation of cover classes and manual augmentation may be necessary.