Reindeer Range Inventory: Use of Winter Landsat Imagery for Stratification of Digital Classification
Abstract
Range inventories using Landsat data have been conducted since 1976 on the tundra in northwest Alaska. Landsat digital classifications of large areas (> 1 million hectares) were used to support collection of ground data on plant productivity and soils. This technique was effective where the inventory team had little or no previous knowledge of the area, and a very short summer field season to collect field data. Problems were encountered where resource categories differed from Landsat spectral classes across these large areas, due to regional differences in ecological condition. A means was required to break the survey area into smaller units so that resource categories could be locally described.
Winter Landsat imagery was photo interpreted to stratify the survey areas into physiographic units. Images were selected from mid-winter, when the landscape is snow covered, and low sun angles provide enhancement of subtle topographic patterns. The physiographic units derived from winter Landsat imagery were digitized to serve as boundaries for stratification of a previously classified Landsat digital image. Spectral categories were then re-identified to a resource category within each stratum. An output image was produced and used as the base for preparing the final range inventory map.
Preliminary verification results of the inventory indicate an overall accuracy of 77% ±2.6% (.95 probability level) in comparison to a reference data set collected independently.
Date of this Version
1981