Open spatial data used to map old-growth structures in small German private forests

Researchers from the Northwest German Forest Research Institute and partners modelled old-growth habitat structures in small private woodlands across the 14,228 km² Lower Saxon Hills. Published in Annals of Forest Science, the study built a parcel-level conservation index from field surveys of 126 woodlands and trained gradient boosting models on five free predictors: terrain slope, CORINE broadleaf cover, distance to paved roads from OpenStreetMap, and nearby open land.

Tested on 48 unseen parcels, the model reached an R² of 0.427. The final model, trained on all 126 parcels, was applied to 38,085 100 m cells of small private woodland. The highest-scoring parcels sat in forest blocks averaging 272 ha, against 1,223 ha for lower-scoring ones, so valuable habitat clusters in small patches within farmland.

Why it matters: The method locates unprotected, unmapped habitat from open datasets alone, avoiding costly inventories. Practitioners elsewhere with similar fragmented tenure and free elevation, land cover and road data could adapt it to target conservation incentives.