Accuracy and inaccuracy assessments in land-cover classification

Several measures assessing accuracy of land-cover classification are availa ble, e,g,, overall and class-averaged accuracies, Also kappa statistic is w idely used for this purpose. In this article, we will discuss properties of these criteria and point out that the kappa statistic has an unfavorable f eature, We will propose an alternative coefficient based on Kullback-Leible r information. A test statistic for significance difference between coeffic ients is also established. Further, the Bayes risk, which takes types of mi sclassifications into account, is discussed. Our assessment measures are ex amined through actual error matrices by Ma and Redmond [3].