Tracking native woodland expansion in the Scottish Highlands using LiDAR

Abstract

Sustained culling of deer in parts of the Scottish Highlands is resulting in natural recolonisation of open landscapes by woodlands. Tracking these changes and furthering our understanding of the drivers of expansion is challenging due to the limited scale of field-collected data. We developed a new pipeline to automatically segment and map small regenerating trees with airborne LiDAR data (pulse density = 22 m-2) and applied it to a 569 km2 study site over Cairngorms Connect – the largest contiguous restoration partnership in the UK. The results were validated by young trees accurately geolocated and measured in the field. We then used the map of regenerating trees to build a Hurdle Model and assessed the influence of topographical and biophysical factors on regeneration success. Results show that clusters of regenerating trees are detectable once they outgrow neighbouring shrubs (up to 1.1m in height). Detection rates reached 83% and >95% for tree clusters 1.4 – 1.6 m and >1.8 m in height, respectively. The algorithm also gave good estimates of the heights (R2 = 0.94) and canopy cover (R2 = 0.90) of young trees, while keeping commission errors low (0.019%). Deer control has led to large-scale natural forest colonisation in Cairngorms Connect. The new pipeline delineated 2.65 million young trees (<5m in height) in the study area. It is estimated that 5325 ha of previously non-forested land now records sapling density >100 ha-1. Geospatially, proximity to seed sources, elevation, habitat type, and deer density all strongly affected tree establishment and sapling density. Synthesis and applications: Forest restoration projects should utilise the new LiDAR-based pipelines to effectively map young trees, assess whether targets are met, and plan further action. Notably, the Scottish Government is now commissioning a national airborne LiDAR scan, providing a unique opportunity to upscale our algorithms to create national woodland expansion databases.