Land-cover mapping using Landsat for sustainable green house gas (GHG) inventory development

Abstract

Information on land-cover is important for verification of land use, land-use change and forestry (LULUCF). LULUCF is significant in assessing anthropogenic Green House Gas (GHG) emissions. This study aimed at developing a simple and computationally efficient yet accurate methodology for national land-cover mapping. The longstanding Landsat data freely available with a renewed and sustainable future archive after the launch of Landsat 8 was used. Data of two epochs namely 2000 and 2010 were selected. A chain classification approach using maximum likelihood classification (MLC) coupled with decision tree was used. Chain classification approach was significant in classifying images of different seasons given that in national land-cover projects it is rare to obtain images of the same date. Six classes recommended by IPCC were adopted. The developed approach attained an accuracy of 86% with a kappa coefficient of 0.8. The study concluded the freely available Landsat data, computational efficiency of MLC and decision tree can be tapped for sustainable land-cover mapping for GHG. This method is replicable and therefore can be used to produce complete and comparable national land-cover products.