COMPARISON OF LAND COVER METHODS INCORPORATING LANDSAT-8 IMAGING AND ANCILLARY DATA

Abstract

Dramatic land-cover modifications have been documented across Malaysia over the past few centuries. Previously forested regions converted primarily into rubber, oil palms and agricultural regions. A present land-cover data required owing to the ongoing land-cover modifications that researchers, planners, and decision-makers will be using. Landsat data is an excellent source of effective land-cover maps creation and updating. The aim of this research is to establish a low-cost method together with ancillary data to enhance Landsat 8 satellite information to generate a relatively accurate and existing land-cover map for the Kota Bharu district. The comparison was made between supervised, unsupervised and merging both as hybrid classification techniques from Landsat 8 information for land-cover classification. Furthermore, land-use map and land cover masking were used as ancillary data in order to enhance the precision of the Landsat 8 classification within the same GIS system.  It has been discovered that using a combination of supervised and unsupervised training programs generates a product that is more accurate instead of using either of them individually. It was also discovered that mapping this item utilizing ancillary GIS information could enhance product precision by up to 4%. The general precision of the final result was 85%. It is proposed that implementing the method described for more remote sensing pictures taken at distinct moments can make it easier to create a database for land cover modifications.