This paper presents the use of satellite data (i.e., Landsat-5 & Landsat-8) to interpret the change of land cover from 1997 to 2020. The study area covers the administrative boundary of Lumajang Regency. The land-cover map of the year 1997 derived from Landsat-5. The Land-cover map of the year 2020 interpreted from Landsat-8. This study uses two methods of image classifications (i.e., unsupervised and supervised). The procedure includes image enhancement, registration, and classification. Then, classification results evaluated by confusion-matrix (overall and kappa accuracy). The supervised classification produces 7 classes of Land cover (i.e., forest, pavement/urban area), paddy field, plantation, rural area, water body and sand mining area. Unsupervised classification produced four 5 class i.e., forest, built-area, paddy field, rural area, and plantation. Supervised classification done the overall and kappa accuracy = 86% and 82%, while unsupervised classification = 73% and 64% for 1997 imagery. Furthermore, for 2020 image, the Supervised classification reaches the overall and kappa accuracy = 93% and 90%, while unsupervised classification done 81% and 72%. The supervised classification method gives a better result than un-supervised. Comparison of 1997 to 2020, it also shows the increase in pavement or build-area, followed by paddy field, rural area, and sand-mining. The change also appears as the decrease in forest and plantation areas.Keywords: Landsat-5, Landsat-8, Unsupervised, Supervised, Lumajang
Sentinel images are widely used for monitoring and mapping our environment phenomenon. The imagery applies in land-use and land cover mapping using pixel-based classification, image segmentation, or other image interpretation algorithms. One type of algorithm may be more suitable for a specific area, depending on many factors. This study aims to analyse and compare two classification algorithms for land cover (LC) mapping in the region characterised by a small scale type of agricultural land occupation. The primary input for this study is the Sentinel 2A image. Two well-known pixel-based classification algorithms, i.e., Maximum Likelihood classifier (MLC) and ECHO (Extraction and Classification of Homogeneous Objects), are used and are compared. The study covers an area of 3320.3 km2. The classification result produces nine (9) land cover classes, i.e., (1) pavement or urban area, (2) heterogeneous agricultural land, (3) irrigated paddy, (4) open water body, (5) dense vegetation or forest, (6) sparse vegetation or plantation, (7) shrubland or dry-land, (8) wetlands, and (9) sand-clay-rock. Classification using MLC and ECHO produced kappa and overall accuracies of more than 90%. In general, both algorithms can produce a relatively similar area extend for each class. However, two classes, i.e., (2) heterogeneous agricultural land and (6) sparse vegetation, are still tricky to distinguish.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.