An Ecological Framework for Evaluating Map Errors Using Fuzzy Sets

Abstract

The use of fuzzy sets to assess uncertainty in land-use/cover maps provides a robust conceptual framework for examining unique characteristics of map error. By recognizing the possibility of gradations of error, fuzzy sets can be used to assess errors due to class similarity, or the sensitivity of the map legend to class boundaries. Building on the theoretical work of Gopal and Woodcock (1994), we present a practical methodology for assessing map errors using fuzzy sets. A key component of our methodology focuses on improving the decision-making process map experts assume when conducting a fuzzy set assessment of map errors. Using an ecological context to define varying levels of land-cover class similarity, we demonstrate how a decision framework guides the map experts’ decisions and provides a more meaningful assessment of map errors. Our methodology differs from traditional fuzzy set error assessment methods in that the map expert evaluates misclassifications within the error matrix (off-diagonal cells) rather than individual reference sites. Advantages to a matrixbased approach include a reduction in the time required by map experts to evaluate map errors, and a relatively simple means of conveying map error information to the map user. We conclude that establishing criteria for determining multiple set memberships in a fuzzy set error assessment is an important methodological procedure that is commonly overlooked. Our methodology, designed to explicitly identify land-cover class similarities based on ecological criteria, serves as a practical example of how to address this issue.