Design-based mapping of errors in remote sensing-based land use/land cover maps

“

From a practical point of view, our proposal allows LULC map users to be congruent and to fully operate in a design-based mode when assessing the quality of satellite maps. As accuracy estimates and standard errors of the familiar “summary” measures achieved from confusion matrices are traditionally used in a design-based inference setting, it is quite natural that the subsequent analysis of the spatial pattern of accuracy also stems from design-based inference without resorting to models. The ability to produce error estimates for each pixel in the map under a design-based inference setting is a novel aspect in the context of the current advances in environmental monitoring and assessment. Per-pixel uncertainty informs users about areas where the map estimates are unreliable, at the same time highlighting the areas where the information provided via the map is trustworthy; therefore, it constitutes support not only from an analytical point of view but also as a powerful communication tool.

”