Skip to main content
ARS Home » Research » Publications at this Location » Publication #267911
Title: Use of Kendall's coefficient of concordance to assess agreement among observers of very high resolution imagery

Author
item GEARHART, AMANDA - North Dakota State University
item Booth, D
item SEDIVEC, K - North Dakota State University
item SCHAUER, S - North Dakota State University

Submitted to: Geocarto International
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/28/2012
Publication Date: N/A
Citation: N/A

Interpretive Summary: The use of aerial photography can reduce the time needed to collect vegetation survey data but there can interpretation differences among users (observers). To measure differences among observers of the same set of high-resolution aerial photography, we asked 6 females, ages 18 to 55 years, having 7 days to 8 years of field experience, to use SamplePoint software to analyze 146 images from the Grand River National Grasslands. The observers generally agreed in their analysis. There was general disagreement with respect to grassland shrubs, and the strength of agreement varied among the other categories of vegetation classification. We found that older observers with little or no local experience on the ground at the study site differed with younger observers who had local experience. We believe that individual differences in eye health, and on-the-ground experience with the vegetation of the area photographed, account for some of the among-people differences in vegetation survey data developed from high resolution aerial photographs.

Technical Abstract: Use of remotely-sensed imagery can replace expensive and time-consuming field collection of vegetation survey data. There are often discrepancies, however, in how different observers characterize the same objects within images. The goal of this study was to determine if different observers of the same set of very large scale aerial (VLSA) imagery characterized ground-cover categories differently. Six female observers ranging in age from 18 to 55 years, with different amounts of field and image-analysis experience, participated. The observers used SamplePoint software to analyze 146 images from the Grand River National Grasslands for 19 independent ground-cover categories. Categories included bare ground, litter, rock, shadow, life form and several individual species including crested wheatgrass, western wheatgrass, blue grama, and needle-and-thread as well as bluegrass species and brome species. Kendall’s coefficient of concordance (W) was used to measure agreement for observations of 19 ground-cover categories among observers. Overall, observers were generally concordant in all categories except shrub species. Strength of concordance varied among the other categories. The older observers had little to no experience with the target ecosystem and showed decreasing concordance for ground-cover categories that lacked easily identifiable characteristics. We believe lack of concordance among observers may be related to individual differences in ophthalmological health and experience with the target ecosystem.