Accuracy Assessment of the ESA CCI 20M Land Cover Map: Kenya, Gabon, Ivory Coast and South Africa

Lesiv, M. ORCID: https://orcid.org/0000-0001-9846-3342, See, L. ORCID: https://orcid.org/0000-0002-2665-7065, Mora, B., Pietsch, S. ORCID: https://orcid.org/0000-0001-6431-2212, Fritz, S. ORCID: https://orcid.org/0000-0003-0420-8549, Bun, H., Sendabo, S., Kibuchi, S., Okemwa, J., Derrik, O., Oima, G., Mgoe, A., Omondi, E., Ongo, D., Abdel-Rahmon, E., Musiln, Z., Dgole, P., Wangai, J., Dariu, D., Marina, K., et al. (2019). Accuracy Assessment of the ESA CCI 20M Land Cover Map: Kenya, Gabon, Ivory Coast and South Africa. IIASA Working Paper. Laxenburg, Austria: WP-19-009

Preview

Text WP-19-009.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (2MB) | Preview

Abstract

This working paper presents the overall and spatial accuracy assessment of the European Space Agency (ESA) 20 m prototype land cover map for Africa for four countries: Kenya, Gabon, Ivory Coast and South Africa. This accuracy assessment was undertaken as part of the ESA-funded CrowdVal project. The results varied from 44% (for South Africa) to 91% (for Gabon). In the case of Kenya (56% overall accuracy) and South Africa, these values are largely caused by the confusion between grassland and shrubland. However, if a weighted confusion matrix is used, which diminishes the importance of the confusion between grassland and shrubs, the overall accuracy for Kenya increases to 79% and for South Africa, 75%. The overall accuracy for Ivory Coast (47%) is a result of a highly fragmented land cover, which makes it a difficult country to map with remote sensing. The exception was Gabon with a high overall accuracy of 91%, but this can be explained by the high amount of tree cover across the country, which is a relatively easy class to map.

Actions (login required)

View Item