Evaluating Historical Open Geospatial Databases for Spatially Explicit LULUCF Land-Use Reconstruction

Résumé fourni par la source

Accurate retrospective, spatially explicit land-use reconstruction is essential for Land Use, Land-Use Change and Forestry (LULUCF) greenhouse gas accounting. However, the suitability of historical geospatial databases as information sources for such reconstruction has rarely been evaluated systematically. This study proposes an objective framework for assessing their correspondence with land-use observations from the Lithuanian National Forest Inventory (NFI). The analysis was based on a reference set of 16,351 systematically distributed NFI sample points and 19 database-year datasets covering the period 1990–2022. Original database classes were harmonised with the national hierarchical LULUCF classification, and correspondence was evaluated using overall accuracy, Cramér’s V and Normalized Mutual Information (NMI), complemented by category-specific representation, precision, recall and F1 score. Correspondence varied substantially among databases according to thematic scope, spatial completeness, mapping characteristics and land-use category. Among the multi-category databases, the Georeferenced Base Cadastre (GRPK) showed the strongest overall correspondence with the NFI reference data, whereas the CORINE Land Cover series provided the longest consistent multi-temporal record extending back to 1990. Forest land and settlements, as well as particularly water-related wetland classes, were represented comparatively reliably, while grassland remained the most difficult major land-use category to identify consistently. Temporal analysis showed that database performance also varied between database versions, while boundary sensitivity analysis demonstrated that observations near mapped polygon boundaries contributed to disagreement without changing the relative advantage of GRPK over CORINE. The results demonstrate that the evaluated historical databases provide substantial and complementary information for spatially explicit LULUCF land-use reconstruction and that the proposed framework provides a transparent basis for identifying and selecting suitable information sources according to land-use category and historical period.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.