The efficiency of sampling very high resolution images for area estimation in the European Union

Abstract

Very high resolution (VHR) images are a valuable information source to estimate land cover area and land cover change. When full coverage of a region with VHR images is not affordable, a sample of images can be considered. Square grids provide a practical sampling frame for VHR images. When using a land cover map as pseudo-truth, the sampling variance is easily assessed but may be overestimated if the land cover map has a coarse resolution. To estimate the potential sampling variance of a cluster sampling scheme, we propose a method based on intra-cluster correlation (ICC) computed from a correlogram. The ‘equivalent number of points’ is a useful indicator to quantify cost-efficiency of sites of a given size. We obtained poor efficiency results for area estimation of major land cover types in the European Union (EU) with a sample of 10 km × 10 km sites, but results are more encouraging for classes with a more scattered layout or for land cover change.

Acknowledgements

The author is grateful to Elisabetta Carfagna for the fruitful discussions on using correlograms in previous studies, to the group of researchers of the Geoland2-Seasonal and Annual CHange MOnitoring (SATCHMO) project for having stimulated this study, to the European Environment Agency (EEA) that distributes CORINE Land Cover and to Marjo Kasanko, Laura Martino, Alessandra Palmieri and Pascal Jacques (Eurostat) for providing the LUCAS data. Sarah Mubareka made useful suggestions to improve this article. Two anonymous reviewers gave a major contribution to improve this article.