Could CORINE Land Cover (CLC) Data Be Used for Downscaling Analysis? A Case Study from NE Romania

CORINE Land Cover (CLC) is one of the most used land databases intended for pan-European-scale analysis. Despite periodic updates, the data suffer from generalization, subjectivity, and inconsistent local knowledge, leading to a distorted representation of reality. In this study, we raise awareness of the use of CLC data in local analysis, for which it was never intended. Our study investigates whether incorporating auxiliary spatial data and local geographical knowledge can yield a higher-accuracy CLC product, without departing from the official CLC definitions and standards. We critically remapped polygon by polygon the 1990, 2000, and 2006 CLC layers, for Iași County (NE Romania), using topographic maps (1972–1989), aerial imagery (1978–2008), and satellite data (1980–2006), and compared it to the original CLC, through change detection analysis. The new maps revealed several issues imposed by (a) generalization—cartographical omissions among settlements, due to the application of a 25 ha minimum mapping unit and a 100 m minimum mapping width; (b) confusions between land cover and land-use classes, such as pasture and wetlands, especially under varying climatic conditions, or imposed by landforms, where we suggest the use of complementary data, such as a Digital Elevation Model (DEM); and (c) the inconsistencies of mapping between successive CLC editions. Our results indicate that the CLC should not be used for downscaling analysis. Therefore, the authors advocate integrating multiple temporal remote sensing layers to achieve a more accurate assessment of land cover classes, thereby compensating for the data’s top-down character. Based on these findings, we propose an error classification approach to serve as a reference for risk mitigation in downscaled spatial analysis. We emphasize the need for CLC data users to validate their data against ground truth to mitigate analytical uncertainties.