CHANGE DETECTION OF LAND COVER IN SEMARANG CITY, FROM 1994 TO 2007 USING THE POST-CLASSIFICATION COMPARISON, NDVI, AND URBAN INDEX DIFFERENCING.

• Main Authors: , Andrew Mulabbi, , Dr. Projo Danoedoro
• Format: Theses and Dissertations NonPeerReviewed
• Published: [Yogyakarta] : Universitas Gadjah Mada 2011
• Subjects: ETD
• Online Access: https://repository.ugm.ac.id/97302/; http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=52504
• Institution: Universitas Gadjah Mada

This research examined land cover change in Semarang city with the aims of 1) carrying out an inventory of land cover change in Semarang, 2) examine the effectiveness of NDVI, UI and Post-classification comparison as methods of land cover change detection, 3) gain an understanding of the location, nature, and magnitude of land cover change and its causes in Semarang city between 1994 and 2007. Therefore Postclassification comparison and combination of NDVI and Urban index image differencing were used on Landsat TM images of 1994 and Aster image of 2007 to determine and quantify the land cover changes that have occurred in Semarang between 1994 to present. In the case of NDVI and UI differencing, vegetation and building density and not pure NDVI and UI values were used. The density maps were derived using the density data collected from the field. Further the methodology also involved combining density data with land cover data so as to add detail to the land cover classes. A comparison between NDVI and UI data also indicated that the higher the UI values, the lower the NDVI and vice versa and also indicated that UI derived map showed some discrepancies as compared to the NDVI derived map. Post-classification comparison change detection was conducted to reveal the areas that have changed over the 13 years period. In this method, the from-to-change informational classes were available and the results revealed drastic growth of built up areas and reduction of agriculture and forest lands over the decade. The image transformation method also indicated an increase in built up land and a reduction in vegetative cover. Post-classification comparison accuracies showed good agreement between land cover maps and the groundtruth data, with kappa values of 0.76 and 0.79 and the overall accuracy of 80% and 83.9 for the maps of 1994 and 2007 respectively. For the change detection accuracy, overall and kappa accuracy of 64.4% and 0.6 respectively portrayed strong agreement between land cover maps and the groundtruth. There occurred tremendous change in land cover both positive (increase in coverage) and negative/reduction in area coverage with hardened clay surfaces and plastic steel and fiberglass (built-up land) experiencing almost 100% increase. This was attributed to the swift expansion of Semarang city. On the other hand, the land cover that experienced reduction in area coverage are bare dry soil, mud and wet surfaces which all attributed to land reclamation and industrial expansion.