Statistical properties of measures of association and the Kappa statistic for assessing the accuracy of remotely sensed data using double sampling

Abstract

A Monte Carlo simulation was used to evaluate the statistical properties of measures of association and the Kappa statistic under double sampling with replacement. Three error matrices representing three levels of classification accuracy of Landsat TM Data consisting of four forest cover types in North Carolina. The overall accuracy of the five indices ranged from 0.35% to 82.1% depending on the number of classes, the level of classification accuracy of satellite imagery, and the simulated sample sizes of reference plots. Statistical criteria used in the evaluation included: percent bias, mean squared error, relative error, ratio of the mean variance to the simulation variance, and 95% confidence coverage rates. Results of the simulation indicated that double sampling provided unbiased estimates of the overall accuracy of remotely sensed imagery irrespective of the number of classes in the image being analyzed, or sample size. While no one index was superior for all levels of accuracy, number of classes, or sample size, the Kappa statistic and Pearson's P provided the best estimates of the overall accuracy of the remotely sensed images. Results from previous studies suggest that increasing the sample size, or reducing the number of classification in the remotely sensed image may increase the accuracy and precision of the estimates. However, this was not the case in this study because the low accuracy of the aerial photos to ground data.

Parent Publication

• Spatial accuracy assessment in natural resources and environmental sciences: Second International Symposium

Citation

Kalkhan, Mohammed A.; Reich, Robin M.; Czaplewski, Raymond L. 1996. Statistical properties of measures of association and the Kappa statistic for assessing the accuracy of remotely sensed data using double sampling. In: Mowrer, H. Todd; Czaplewski, Raymond L.; Hamre, R. H., tech coords. Spatial Accuracy Assessment in Natural Resources and Environmental Sciences: Second International Symposium. May 21-23, 1996. General Technical Report RM-GTR-277. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station. p. 467-476.