A dataset of 40'000 trees with section-wise measured stem diameter and branch volume from across Switzerland.

Estimating growing stock is one of the main objectives of forest inventories. It refers to the stem volume of individual trees which is typically derived by models as it cannot be easily measured directly. These models are thus based on measurable tree dimensions and their parameterization depends on the available empirical data. Historically, such data were collected by measurements of tree stem sizes, which is very time- and cost-intensive. Here, we present an exceptionally large dataset with section-wise stem measurements on 40'349 felled individual trees collected on plots of the Experimental Forest Management project. It is a revised and expanded version of previously unpublished data and contains the empirically derived coarse (diameter ≥7 cm) and fine branch volume of 27'297 and 18'980, respectively, individual trees. The data were collected between 1888 and 1974 across Switzerland covering a large topographic gradient and a diverse species range and can thus support estimations and verification of volume functions also outside Switzerland including the derivation of whole tree volume in a consistent manner.

Predicting Tree Species From 3D Laser Scanning Point Clouds Using Deep Learning.

Automated species classification from 3D point clouds is still a challenge. It is, however, an important task for laser scanning-based forest inventory, ecosystem models, and to support forest management. Here, we tested the performance of an image classification approach based on convolutional neural networks (CNNs) with the aim to classify 3D point clouds of seven tree species based on 2D representation in a computationally efficient way. We were particularly interested in how the approach would perform with artificially increased training data size based on image augmentation techniques. Our approach yielded a high classification accuracy (86%) and the confusion matrix revealed that despite rather small sample sizes of the training data for some tree species, classification accuracy was high. We could partly relate this to the successful application of the image augmentation technique, improving our result by 6% in total and 13, 14, and 24% for ash, oak and pine, respectively. The introduced approach is hence not only applicable to small-sized datasets, it is also computationally effective since it relies on 2D instead of 3D data to be processed in the CNN. Our approach was faster and more accurate when compared to the point cloud-based "PointNet" approach.

Classifying development stages of primeval European beech forests: is clustering a useful tool?

BACKGROUND: Old-growth and primeval forests are passing through a natural development cycle with recurring stages of forest development. Several methods for assigning patches of different structure and size to forest development stages or phases do exist. All currently existing classification methods have in common that a priori assumptions about the characteristics of certain stand structural attributes such as deadwood amount are made. We tested the hypothesis that multivariate datasets of primeval beech forest stand structure possess an inherent, aggregated configuration of data points with individual clusters representing forest development stages. From two completely mapped primeval beech forests in Albania, seven ecologically important stand structural attributes characterizing stand density, regeneration, stem diameter variation and amount of deadwood are derived at 8216 and 9666 virtual sampling points (moving window, focal filtering). K-means clustering is used to detect clusters in the datasets (number of clusters (k) between 2 and 5). The quality of the single clustering solutions is analyzed with average silhouette width as a measure for clustering quality. In a sensitivity analysis, clustering is done with datasets of four different spatial scales of observation (200, 500, 1000 and 1500 m2, circular virtual plot area around sampling points) and with two different kernels (equal weighting of all objects within a plot vs. weighting by distance to the virtual plot center). RESULTS: The clustering solutions succeeded in detecting and mapping areas with homogeneous stand structure. The areas had extensions of more than 200 m2, but differences between clusters were very small with average silhouette widths of less than 0.28. The obtained datasets had a homogeneous configuration with only very weak trends for clustering. CONCLUSIONS: Our results imply that forest development takes place on a continuous scale and that discrimination between development stages in primeval beech forests is splitting continuous datasets at selected thresholds. For the analysis of the forest development cycle, direct quantification of relevant structural features or processes might be more appropriate than classification. If, however, the study design demands classification, our results can justify the application of conventional forest development stage classification schemes rather than clustering.

Standardizing the protocol for hemispherical photographs: accuracy assessment of binarization algorithms.

Hemispherical photography is a well-established method to optically assess ecological parameters related to plant canopies; e.g. ground-level light regimes and the distribution of foliage within the crown space. Interpreting hemispherical photographs involves classifying pixels as either sky or vegetation. A wide range of automatic thresholding or binarization algorithms exists to classify the photographs. The variety in methodology hampers ability to compare results across studies. To identify an optimal threshold selection method, this study assessed the accuracy of seven binarization methods implemented in software currently available for the processing of hemispherical photographs. Therefore, binarizations obtained by the algorithms were compared to reference data generated through a manual binarization of a stratified random selection of pixels. This approach was adopted from the accuracy assessment of map classifications known from remote sensing studies. Percentage correct (Pc) and kappa-statistics (K) were calculated. The accuracy of the algorithms was assessed for photographs taken with automatic exposure settings (auto-exposure) and photographs taken with settings which avoid overexposure (histogram-exposure). In addition, gap fraction values derived from hemispherical photographs were compared with estimates derived from the manually classified reference pixels. All tested algorithms were shown to be sensitive to overexposure. Three of the algorithms showed an accuracy which was high enough to be recommended for the processing of histogram-exposed hemispherical photographs: "Minimum" (Pc 98.8%; K 0.952), "Edge Detection" (Pc 98.1%; K 0.950), and "Minimum Histogram" (Pc 98.1%; K 0.947). The Minimum algorithm overestimated gap fraction least of all (11%). The overestimation by the algorithms Edge Detection (63%) and Minimum Histogram (67%) were considerably larger. For the remaining four evaluated algorithms (IsoData, Maximum Entropy, MinError, and Otsu) an incompatibility with photographs containing overexposed pixels was detected. When applied to histogram-exposed photographs, these algorithms overestimated the gap fraction by at least 180%.