Making trees visible : legislation, policy and LiDAR in spatial planning

Sammanfattning

Trees provide irreplaceable ecosystem services, yet they are increasingly threatened by urban development. In response, the legally binding EU Nature Restoration Law, adopted in 2024, mandates no net loss of urban tree canopy cover at the national level by 2030. To comply with these directives and preserve existing trees, a new Swedish standard (SS 990002:2025) requires a physical root protection zone extending four meters from the tree’s outermost branches during construction work. Securing this substantial physical space requires establishing tree protection plans and actively moving infrastructure footprints away from trees during the earliest planning stages. However, today’s planning, design, and construction drawings often rely on base maps that represent trees as dimensionless points. This causes an architectural lock-in effect, where developments are finalized before a tree's actual physical dimensions are established, leading to inevitable canopy loss. To address this, this thesis explores how landscape architects and urban planners can efficiently detect legally protected trees before this lock-in occurs. This thesis investigates remote sensing as a solution, combining interviews, literature review and a technical case study in Helsingborg, Sweden. The study evaluates the capacity of high-resolution Airborne Laser Scanning point clouds to act as an early spatial filter. It assesses raster-based vwf() and point-cloud-based li2012() detection algorithms, alongside a method to estimate diameter at breast height (DBH) from LiDAR-derived crown areas to screen for legally protected trees. The interview study revealed that despite legal frameworks, trees are systematically deprioritized due to the protracted planning process, conflicting utility interests, and a lack of early 3D spatial representation. The technical application demonstrated that LiDAR is a viable large-scale screening tool. The vwf() method achieved an F-score of 0.813 (86.0% recall, 77.1% precision), while the li2012() method achieved an F-score of 0.804 (95.3% recall, 69.5% precision), though both methods struggled with overand under-segmenting complex broadleaved canopies. The tool proved robust at identifying larger trees (DBH >40 cm) but detected no Trees Worthy of Special Protection (DBH >100 cm), despite field-validated specimens on site. These results point to the oversegmentation of multi-stemmed individuals and the limitations of crown-area-based allometric proxies for mature broadleaved species. This highlights the need for site- and species-specific models or deep learning approaches capable of handling irregular, multi-peaked canopies. Automated LiDAR-based detection cannot replace traditional field inventories, but it can function as a strategic early warning system. By translating point clouds into objective spatial constraints at the beginning of a project, planners can proactively secure the mandatory four-meter protection zones required. Ultimately, this facilitates a spatial planning process where mature trees are treated as foundational green infrastructure, prioritized as both ecological and economic assets.