Although the mapping was at a large scale, most map units were compound, i.e., containing more than one association. As such, map accuracy assessment procedures documented in the literature were not appropriate for use. Various field sampling methods and summary statistics were evaluated for their practicality in application and interpretation of the results.
Results/Conclusions The protocol presents a statistically unbiased approach to evaluate the accuracy of large-scale mapping. The thematic content of randomly selected map areas (polygons or groupings of pixels) is assessed by various means – the methods varying in precision and objectivity. A new map accuracy statistic, termed the ‘percent overlap’ was developed to assess maps with compound map units.
The protocol requires the development of a sampling plan that articulates decisions about protocol level, sample size, assessment method, target error, etc. that is appropriate for the type of map – polygon vs pixel – and the intended use of the map. Information on how to decide each of these is presented.
The protocol has been applied in over 20 mapping projects, differing in mapping approach and resulting map polygons. Issues with application in various situations are discussed.
Although an additional overall cost, assessing the accuracy of complex thematic maps is critically important to determining appropriate uses for vegetation mapping. The approach outlined provides a means of obtaining some overall, statistically valid scores to rate the accuracy of vegetation or other resource maps. The results can be used as a component of quality assurance or for presenting statistics on the accuracy of mapping.
