Mapping old-growth forests using airborne lidar data and satellite images: how do plot size and rarity affect accuracy?

The study "Mapping old-growth forests using airborne lidar data and satellite images: how do plot size and rarity affect accuracy?" published in the Canadian Journal of Forest Research emphasizes that old-growth forests in the boreal biome have become increasingly rare and fragmented, making their identification a priority for effective conservation and forest management. However, their precise locations are often insufficiently known, limiting the ability of planners and policymakers to safeguard these ecologically valuable stands. In this study, the researchers examined the potential of combining airborne lidar data and satellite imagery to map old-growth forests within Finland’s coniferous forest zone. A particular focus was placed on understanding how plot size and the rarity of old-growth forests influence detection accuracy.

Using a Gaussian process classifier, the researchers distinguished old-growth forests from managed stands based on a field dataset of 176 old-growth and 1,082 managed plots. Their results demonstrate that larger plot sizes markedly improve classification performance. When plot dimensions increased from 20 m × 20 m to 60 m × 60 m, the classifier more effectively captured the spatial tree patterns and crown structures characteristic of natural forest dynamics. The highest F1-score (0.74) was achieved through data augmentation techniques that generated additional training plots within forest boundaries, thereby enriching the representation of old-growth features.

The group also found that detection accuracy declines as old-growth forests become rarer in the landscape, an important consideration given regional differences in land-use pressures and conservation history. Understanding this rarity effect is essential for interpreting mapping results and setting realistic expectations for large-scale inventories.

Overall, the mapping framework presented offers a valuable tool for guiding field surveys and improving the spatial understanding of old-growth forest distribution, ultimately supporting more informed conservation strategies.

The paper was published as part of the FORWARDS project.