CONDITIONAL TAU-COEFFICIENT FOR ASSESSMENT OF PRODUCERS ACCURACY OF CLASSIFIED REMOTELY-SENSED DATA

The error matrix is frequently used for accuracy assessment of the cla ssification of remotely sensed data. The classification accuracy for e ach individual category is often expressed by the percentage correct c lassified and/or the conditional kappa coefficient. In the classificat ion of remotely sensed data, the distribution of the reference data ov er various categories is often unknown beforehand. In such cases, cond itional kappa may give erroneous estimates of classification accuracy. An alternative measure for classification accuracy of individual cate gories is proposed. It is denoted as conditional tau, and it expresses the agreement obtained after removal of the random agreement expected by chance. Formulae for computation of conditional tau appropriate fo r classifications based on equal and unequal probabilities of category membership are presented. Three numerical examples are used to demons trate the calculation and interpretation of the measure. In the exampl es, conditional tau correctly estimates the classification accuracy, w hilst the results show that conditional kappa may overestimate as well as underestimate the classification accuracy.