The Salford Advanced Laser Canopy Analyser was first deployed in 2010 and was the first of two full-waveform dual-wavelength terrestrial laser scanners specifically designed to measure the three-dimensional structure and composition of forest canopies. In the last decade the instrument has been deployed in experiments in England, Scotland, Spain, Australia, and the United States. SALCA has been used in experiments to estimate leaf and canopy water content, plant and leaf area index, and LAI profiles. Ratios of the calibrated reflectance from SALCA have been used to separate leaf and woody material in tree canopies, and most recently in 2023 to estimate the fuel moisture content in fire-prone Mediterranean woodlands. SALCA data has also been used in an art project where 3D scanner data have been used in works to generate ambient music.

This paper aims to critically evaluate the research undertaken with SALCA over the last decade. It highlights key successes, the obstacles overcome, and the remaining challenges for future development of such instruments. Examples are drawn from previously published research, from unpublished experimental results, and through critical scientific debate amongst the authors of this paper. Three main challenges are identified discussed: (i) instrument design and operation, (ii) laser reflectance calibration, and (iii) data processing and analysis. These together provide a roadmap for future research and a blue-print for the development of the next generation of full-waveform, multi-wavelength terrestrial laser scanners for forest applications.

During SilviLaser 2021, a benchmark was organized to demonstrate the capability of different 3D data acquisition techniques for capturing various forest parameters. In total, 9 groups of participants revealed their advanced techniques and setups to acquire point clouds in the designated sites. Based on the applied equipment and platform, all approaches are categorized into 3 groups: mobile laser scanning, terrestrial laser scanning, and photogrammetry (plus others). In order to efficiently and accurately extract basic forest parameters (e.g., stem position, tree species, DBH) from this data set, understanding the data behavior of different approaches is the main key to maximizing their strength. This study aims to characterize each method in terms of spatial distribution and coverage of point clouds, extra attributes, as well as pros and cons for different usage purposes.
The output of this study is valuable for selecting an adequate method to fulfill the requirements of user-specific forestry applications. Also, it is beneficial for solving the existing or future problems of multi-source point cloud processing in forests, e.g., co-registration of various data sources. This study is done within the framework of the project 4Map4Health.

Quantitative structure models (QSMs) of trees using terrestrial laser scanning (TLS) data are widely used to estimate tree metrics such as aboveground biomass (AGB), which can be estimated through volumetric modelling. However, reconstructing QSMs to estimate AGB has been mainly developed using TLS data of leaf-off trees, while using leaf-on data introduces additional uncertainties. Classifying leaf-on point clouds into leaf and wood parts of a tree and using the woody part for the QSM reconstruction is one approach to estimate AGB using leaf-on trees. In this study, we aim to compare different leaf-wood separation algorithms and explore how their performances influence the accuracy of AGB estimation. Four open-source leaf-wood separation algorithms are selected in this study: LeWoS, TLSeparation, RF model, and GB separation. Here we use a leaf-on and leaf-off dataset from Wytham Woods (876 trees, UK) and implement the four leaf-wood separation algorithms on each tree and use the wood-only part to generate a QSM for each leaf-wood separation model. We will use the QSM generated from the point cloud obtained in leaf-off conditions as the benchmark to evaluate the four different leaf-removal algorithms and their impact on AGB estimation and associated uncertainties.

Separating leaf from wood returns in forest lidar point cloud is a commonly required pre-processing step to estimate leaf area or wood volume. This key step is particularly challenging in tall dense hyper-diverse evergreen forests.

Unmanned aerial vehicle laser scanning (ULS) has emerged as a premium solution to collect high density point clouds (~10^3/m2) rapidly over several hectares of forest. However, the level of detail remains much lower than what is typically achieved using multiple TLS positions (~10^5/m2). In particular in the lower canopy.

In this study, we compared three semantic segmentation procedures applied to TLS and DLS point cloud over one-ha of moist tropical forest in French Guiana. Both point clouds were acquired in October 2021. A subset of 418 trees were manually isolated in the TLS point cloud and a semi-automatic segmentation of leaf and wood returns was conducted on each isolated tree. Labels were then transferred to the co-registered ULS point cloud based on nearest neighbours. These two datasets (TLS and ULS) were used as a reference to evaluate the three algorithms.

The three procedures to evaluate were 1) A geometric leaf-wood classification method (Lewos) 2) a deep learning (DL) approach (Forest Structural Complexity Tool: FSCT) and 3) A DL algorithm we developed. The latter was based on PointNet++ that combines convolution networks on 3D point clouds, quantitative information attached to each point and prior spatial and geometrical and topological information related to the neighbourhood of each point.

We compare the three models in terms of their performance in separating leaf from wood, in TLS and DLS point clouds. We further discuss which features of our DL approach allow it to outperform the other two methods for the designated task.

Airborne laser scanning (ALS) data enable the wall-to-wall estimation of forest aboveground biomass. However, the use of ALS data for biomass estimation is often limited by the lack of biomass reference data that are required to build prediction models, because the field work to collect these data is time-consuming and therefore costly. One approach to deal with missing in-situ reference data is a spatial model transfer: models trained with ALS and biomass reference data collected from other sites are applied to the ALS data from the site of interest, for which no reference data are available. Another approach to overcoming the need to collect reference data is to train models with computer-generated data. We investigated the performance of biomass models trained with simulated forest and ALS data in comparison to spatially transferred models. Simulated data were generated by combining a forest generator, real laser scanning point clouds of individual trees, and a laser scanning simulator (HELIOS++). Real datasets collected from forest sites in Poland, the Czech Republic, and Canada were used for training and testing the models.

We found that models trained with simulated data did not perform as well as models trained with real data collected from the same site the models were applied to. However, using simulated data as additional training data could improve model accuracies in terms of RMSE and r² when only a limited number of real training samples (12 – 346, depending on the study site) are available (Figure 1). For three of the four test datasets, training models with data collected from other sites resulted in higher model accuracies than when models were trained with simulated data.

Although the simulated data we generated cannot compete with real data, our study showed promising results for using simulated laser scanning data to train biomass models.