Are remote sensing-based crop type classifications suitable for calculating a landscape heterogeneity metric? A data-fitness-for-purpose assessment

Remote-sensing products are widely used to assess biodiversity in German agricultural landscapes. Evaluating their suitability for specific purposes is crucial for reliable decision-making. This study develops and applies a problem-oriented data-fitness-for-purpose (DFFP) perspective on satellite-based crop type classifications (CTC) to derive land use heterogeneity metrics (LHM) as proxies for biodiversity. Two nationwide CTC (SWD, PRE) products were compared with detailed administrative reference data (IACS) for Lower Saxony and Brandenburg in Germany (2017–2019). Class schemes and geometries were harmonized, and the Shannon Evenness Index (SEI) was calculated on 1 km 2 hexagons as an illustrative LHM. Regional accuracy metrics differed substantially between classification products and between federal states, with systematically lower accuracies in Brandenburg. High overall accuracies masked pronounced class imbalances and regional variations in performance. A simple, spatially explicit classification uncertainty metric was computed at the field level and aggregated to the hexagon scale, demonstrating that local uncertainty information can be carried through a multi-step workflow. The resulting patterns revealed systematic differences between federal states, classification products, and reference data. Our findings indicate that nationally reported accuracy measures are insufficient to evaluate local or regional suitability of CTCs for biodiversity-relevant LHMs. Spatially explicit, domain-specific uncertainty layers should become standard components of remote-sensing-based biodiversity indicators. Integrating such layers within ISO 19157-1 data quality concepts, including “spatial uncertainty” and provenance-rich, FAIR geospatial workflows, will enhance the robustness and interpretability of landscape heterogeneity assessments.