Old-growth forests (OGF) play a critical role in biodiversity conservation and climate regulation. The preservation of Europe’s remaining OGFs is therefore essential and as such addressed in the European Union’s (EU’s) biodiversity strategy 2030. In order to strictly protect all remaining EU primary and OGFs, their locations and extent need to be mapped. Remote sensing (RS) offers the possibility to assess extensive and remote areas. This study evaluates the use of Sentinel-2 satellite images and airborne laser scanning (ALS) data for the assessment of dominant stand age and development classes for beech forests in four test sites located in three different biogeographical regions across Europe. We use up to 150 spectral, textural and height features as input to a random forest (RF) regression. Elevation consistently ranks among the top eight most important features, showing the highest importance in mountainous regions and the lowest in predominantly flat terrain. Texture, on the other hand, varies in importance across the sites and appears to be inversely related to elevation, with higher importance values observed in flat areas. Regarding spectral indices, the Normalized Difference Red Edge (NDRE1) emerges as a significant feature across most sites. Near and short-wave infrared and the third red-edge band are important individual features in several sites. Training data is derived from existing age maps. Validation is done using 512 independent field measurement plots. The results show overall accuracies (OA) for five structural development classes between 53 and 81 % for Sentinel-2 data only. Where available, ALS data increases the OA by about 6 %. When considering only two classes (OGF vs. non-OGF), the OA is between 59 % for Bulgaria with Sentinel-2 data only and 94 % for Belgium, when including ALS. Our approach is constrained by the potential unavailability of high-quality reference data for various biogeographical regions, as well as the limited accessibility of LiDAR data. The comparison with existing global RS-based maps evidently shows many more details and higher accuracy of our products. In comparison with a European map of existing primary forests, we see overall congruence, but also differences: our approach spots similar spectral and structural characteristics in areas outside the known primary or old-growth forests. RS can thus provide valuable spatial insights into potential OGF locations to better target field visits and facilitate the faster identification of currently unprotected OGFs.
扫码关注我们
求助内容:
应助结果提醒方式:
