Mapping the uncertainty of changes in vegetation cover in and around the brussels capital region

Abstract

Images from medium-resolution satellites are frequently used to study changes in urban vegetation or impervious surface cover over relatively long time periods. In this paper, we applied both “hard” and “soft” mapping methods on two images covering the Brussels Capital Region in a time span of roughly 20 years. Although soft approaches are more accurate because they take the occurrence of mixed pixels into account, errors that result from the application of sub-pixel proportion estimation methods nevertheless propagate if the maps are used for change analysis. In this paper, we propose a method to take uncertainty in sub-pixel classification into account when producing change maps.