Operational large-area land-cover mapping: Ethiopia case study

Authors: Reza Khatami*, Department of Geography, University of Florida, Jane Southworth, Department of Geography, University of Florida, Carly Muir, Department of Geography, University of Florida, Trevor Caughlin, Department of Biological Sciences, Boise State University, Alemayehu Ayana, Ethiopian Environment and Forest Research Institute
Topics: Remote Sensing, Land Use and Land Cover Change, Africa
Keywords: Land Cover, Classification, Change Analysis, Accuracy Assessment, Environmental Sampling
Session Type: Paper
Day: 4/11/2018
Start / End Time: 8:00 AM / 9:40 AM
Room: Maurepas, Sheraton, 3rd Floor
Presentation File: No File Uploaded

Large-scale land transactions over the last couple of decades has affected agriculture, ecosystem services, and food and energy security in Ethiopia. In this research, to analyze agricultural activities and vegetation cover we produced a recent 30 meter land-cover map of Ethiopia using Landsat 8 images. The entire area of Ethiopia was covered by 67 Landsat scenes. Two existing land-cover maps including “Copernicus Global Land Operations” and “GlobeLand30” were used for spatial stratification of training data location selection and high resolution images were used to manually digitize the label of the selected training pixels. Different classification algorithms were tested. In addition, application of textural layers, spectral indices, and elevation data for classification accuracy improvement were evaluated. Accuracy of the final product and land-cover classes’ area were estimated using an independent probability sample. Because of the large area of Ethiopia, which is greater than 1.1 million square kilometers, the classification task can be identified as an “operational” large-area land-cover mapping project. Therefore, classification results reported by this research can be utilized to enhance actionable decision-making when aiming to improve accuracy of land-cover mapping for large-area studies.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login