Replication data and R scripts for: Optimising rare tree species detection for large-area mapping: A case study on European aspen (Populus tremula)

The material was collected to support the development, evaluation, and reuse of methods for detecting and classifying European aspen (Populus tremula L.) in boreonemoral forests using high-density airborne laser scanning (ALS) and high-resolution multispectral aerial imagery. The primary aim was to assess how well rare tree species can be identified at the individual tree level and to evaluate classifier performance under realistic base rate conditions. The material is suitable for research on individual tree detection, tree species classification, feature engineering from ALS and multispectral data, and evaluation of classification uncertainty for rare species. The material consists of (i) field inventory data from 253 circular plots (400 m² each) including spatially explicit measurements of all living trees with DBH ≥ 4 cm, (ii) wall-to-wall ALS point clouds with very high point density and associated spectral information derived from multispectral imagery, (iii) individual tree segments derived from ALS data, (iv) engineered tree- and plot-level features, and (v) classification outputs and performance estimates. The research object is standing living trees in boreonemoral forests, and the primary unit of observation is the individual tree, with additional aggregation to plot level for cross-validation and performance assessment. Each tree is represented by field-measured attributes, remotely sensed features, and (where applicable) a matched ALS-derived tree segment, enabling reuse of the material for method development, benchmarking, and comparative studies. Please note, however, that the raw point clouds and coordinates of the segments have not been shared. As such, the script is functional starting from 02_smote_optimisation.R.