Estimating land cover class area from remote sensing classification

Accurate estimation of areas covered by land use land cover classes is central to many resource management and monitoring programs, crop yield forecasting, forest and environmental management. In this paper, various techniques of estimating areas of land cover classes derived from remote sensing image classification have been discussed. A comparative study has been conducted to examine the accuracy and consistency of the area estimated from five error matrix based techniques. The results show that 'direct' and 'additive' estimators produce the most accurate and consistent results. The 'map marginal proportion based estimator' and 'inverse estimator' produce accurate results when the testing sample size is large.