Tehnik Mendeteksi Kebakaran Hutan Melalui Citra Satelit Multiwaktu: Studi Kasus di Propinsi Sumatera Selatan dan Riau

Abstract

This paper describes the techniques for detecting forest and land fire. Of the three change detection techniques evaluated, i.e., post-classification comparison, multitemporal principal component and direct multitemporal classification, the multitemporal principal component, specifically that was derived from variance-covariance matrix (unstandardized principal component) was recognized to be suitable in detecting changes due to forest and land fires. The study found that the delta brightness, delta greenness, stable brightness and stable greenness indices derived from unstandardized multitemporal principal component analysis effectively summarized burnt-forest information. In this study, it was shown that Landsat TM provides information of totally and moderatelly burnt logged-over forest as well as burnt bush/shrub.