Expert Classification of Land Use and Land Cover in Markham, Morobe Province, Papua New Guinea

Abstract

The expert classification approach useddata other than satellite images to improve and increase the accuracyof classificationof land use and land cover. This study was focused on understanding expert classification and formulating the rules for extracting LULC classes using 1992 and 2001 Landsat Thematic Mapper datasets. Auxiliary data of Principal Components Analysis (PCA) and Normalized Difference Vegetative Index (NDVI) were applied in order to improve the accuracy of classification using a decision tree approach for classification. A total of 7 classes, which include Dense Forest, Less Dense Forest, Agriculture, Grassland, Bareland, Built-Up Landand Water were identified and extracted with overall accuracies of 90% for 1992 and 92% for 2001 with Kappa statistics of 0.8786 and 0.8935. However, there were still issues faced with spectral similarities between classes for Landsat TM images. Despite this, the study has shown that expert classification technique can be used for acquiring useful and reliable LULC information for natural resource management in Papua New Guinea.