The retrieval of forest above-ground biomass (AGB, or biomass hereafter) using remote sensing data has received increasing attention in the last decades for several reasons: primarily, for the ability of remote sensing to spatially extrapolate point field information on forest parameters, enabling AGB mapping over large areas; for the increased availability of different Earth Observation data types; and, in a global scenario of climate change and biodiversity loss, for the relevance of forest biomass with respect to global carbon accounting and reporting, conservation and management of natural resources, and environmental modeling.¹ 2. 3. 4. 5. 6. ^{– 7}

The development of laser scanning technologies has gradually modified methods for forest mensuration and inventory. The main objective of this study is to assess the potential of integrating ALS and TLS data in a complex mixed Mediterranean forest for assessing a set of five single-tree attributes: tree position (TP), stem diameter at breast height (DBH), tree height (TH), crown base height (CBH) and crown projection area radii (CPAR). Four different point clouds were used: from ZEB1, a hand-held mobile laser scanner (HMLS), and from FARO® FOCUS 3D, a static terrestrial laser scanner (TLS), both alone or in combination with ALS. The precision of single-tree predictions, in terms of bias and root mean square error, was evaluated against data recorded manually in the field with traditional instruments. We found that: (i) TLS and HMLS have excellent comparable performances for the estimation of TP, DBH and CPAR; (ii) TH was correctly assessed by TLS, while the accuracy by HMLS was lower; (iii) CBH was the most difficult attribute to be reliably assessed and (iv) the integration with ALS increased the performance of the assessment of TH and CPAR with both HMLS and TLS.

Proximally sensed laser scanning presents new opportunities for automated forest ecosystem data capture. However, a gap remains in deriving ecologically pertinent information, such as tree species, without additional ground data. Artificial intelligence approaches, particularly deep learning (DL), have shown promise towards automation. Progress has been limited by the lack of large, diverse, and, most importantly, openly available labelled single‐tree point cloud datasets. This has hindered both (1) the robustness of the DL models across varying data types (platforms and sensors) and (2) the ability to effectively track progress, thereby slowing the convergence towards best practice for species classification. To address the above limitations, we compiled the FOR‐species20K benchmark dataset, consisting of individual tree point clouds captured using proximally sensed laser scanning data from terrestrial (TLS), mobile (MLS) and drone laser scanning (ULS). Compiled collaboratively, the dataset includes data collected in forests mainly across Europe, covering Mediterranean, temperate and boreal biogeographic regions. It includes scattered tree data from other continents, totaling over 20,000 trees of 33 species and covering a wide range of tree sizes and forms. Alongside the release of FOR‐species20K, we benchmarked seven leading DL models for individual tree species classification, including both point cloud (PointNet++, MinkNet, MLP‐Mixer, DGCNNs) and multi‐view 2D‐based methods (SimpleView, DetailView, YOLOv5). 2D Image‐based models had, on average, higher overall accuracy (0.77) than 3D point cloud‐based models (0.72). Notably, the performance was consistently >0.8 across scanning platforms and sensors, offering versatility in deployment. The top‐scoring model, DetailView, demonstrated robustness to training data imbalances and effectively generalized across tree sizes. The FOR‐species20K dataset represents an important asset for developing and benchmarking DL models for individual tree species classification using proximally sensed laser scanning data. As such, it serves as a crucial foundation for future efforts to classify accurately and map tree species at various scales using laser scanning technology, as it provides the complete code base, dataset, and an initial baseline representative of the current state‐of‐the‐art of point cloud tree species classification methods.

Spatially-explicit information on forest composition provides valuable information to fulfil scientific, ecological and management objectives and to monitor multiple changes in forest ecosystems. The recently developed Sentinel-2 (S2) satellite imagery holds great potential for improving the classification of forest types at medium-large scales due to the concurrent availability of multispectral bands with high spatial resolution and quick revisit time. In this study, we tested the ability of S2 for forest type mapping in a Mediterranean environment. Three operational S2 images covering different phenological periods (winter, spring, summer) were processed and analyzed. Ten 10 m and 20 m bands available from S2 and four vegetation indices (VIs) were used to evaluate the ability of S2 to discriminate forest categories (conifer, broadleaved and mixed forests) and four forest types (beech forests; mixed spruce-fir forests; chestnut forests; mixed oak forests). We found that a single S2 image acquired in summer cannot discriminate neither the considered forest categories nor the forest types and therefore multitemporal images collected at different phenological periods are required. The best configuration yielded an accuracy > 83% in all considered forest types. We conclude that S2 can represent an effective option for repeated forest monitoring and mapping.