This paper presents a protocol for assessing the predictive accuracy of thematic maps. Although developed during the Cariboo Predictive Ecosystem Mapping (PEM) pilot, the Canim Lake PEM operational prototype, and the Quesnel PEM to test the accuracy of predictive ecosystem maps, this protocol is applicable to many thematic maps. The Protocol for Accuracy Assessment of Ecosystem Maps (Meidinger 2003) incorporates many of the concepts presented here, but recent work done on the Canim Lake and Quesnel PEMS suggests that a more detailed operational protocol is needed to ensure the integrity of the data collected. In addition to detailed instructions and specifications for data collection, this paper describes a staged sampling protocol designed to minimize the sampling cost required to attain a specified level of confidence in the assessed accuracy.
We recommend a sampling strategy that is independent of the final map polygons delineated. The sample entity-a closed, three-side traverse, 500 m on a side-estimates the ability of a map to predict the thematic (e.g., site series) composition of a small area. However, it does not estimate the accuracy of predicting the exact location of site series within the small area. A 60-m corridor centred on the traverse is about 9 ha and estimates the ability of the map to predict the proportions of thematic units (e.g., site series) within areas tens of hectares and larger. The accuracy of prediction for areas smaller than tens of hectares will be lower.
This paper describes the following procedures:
1. A rationale for a small-area sampling protocol versus polygon sampling.
2. Detailed field procedures for the collection of traverse data, which will be the basis for estimating the composition of the small area sampled.
3. Methods for estimating measurement error, which can be used to estimate the upper limit of accuracy possible.
4. Calculations of percentage overlap as a measure of map accuracy.
5. A staged sampling procedure to minimize the number of samples required to meet a specified confidence level.
This protocol is not meant to replace the Protocol for Accuracy Assessment of Ecosystem Maps, but rather to provide more detailed procedures for the collection of traverse data and the calculation of sample size. This protocol is designed for use in training and in the specification of contracts for accuracy assessment.
Download Technical Report 023 PDF file (769 KB)
To view this document you need the current version of
Adobe Acrobat Reader, available free from the
Adobe Web Site.
Updated May 11, 2007