Forests are among the most widespread and diverse ecosystems on Earth, providing essential ecosystem services at local and global scales. However, they are facing major challenges due to climate change, economic pressures and human population growth. Digital twins of forests could help address these challenges by enabling comprehensive forest monitoring and supporting management decisions. In this publication, we describe how digital twins differ from other digital tools in the forest domain and explore concepts and technologies that can serve as the basis for implementing forest digital twins. We outline the underlying data model of the digital twins, which includes trees as the core forest elements, as well as their environment. We explain how a wide range of data collection approaches can be combined for comprehensive data collection and how the data can be integrated into a spatio-temporal forest data space. We describe data processing approaches to enrich raw data with semantic information and address how digital twins can support decision making through modeling and simulation. We explain the role of web-based visualization in interacting with forest digital twins. Overall, our concept lays the foundation for the technical implementation of forest digital twins that integrate, process, analyze and visualize forest data from a variety of sources. The implementation of forest digital twins in practice would enrich our understanding of forest ecosystems and enable targeted management of forests and their ecosystem services.

3D point clouds acquired with terrestrial or mobile LiDAR sensors are increasingly used to map urban forests. The segmentation of separate tree instances, i.e., subsets of points representing individual trees, is a relevant step in automatically extracting tree inventory data from 3D point clouds. Various algorithms have been proposed for tree instance segmentation, offering different trade-offs between accuracy, runtime, and robustness against data incompleteness and noise. In this work, we propose a coarse-to-fine algorithm for segmenting tree instances in urban 3D point clouds from terrestrial or mobile LiDAR scanning that combines two existing techniques: (1) the computationally efficient marker-controlled Watershed algorithm and (2) a more accurate 3D region growing algorithm. Initially, the marker-controlled Watershed algorithm generates a coarse segmentation, which is further improved by a Voronoi segmentation-based error removal. Subsequently, the coarse segmentation is refined by the 3D region growing algorithm in areas with overlapping tree crowns and sufficient data quality. In both steps, our algorithm uses the results of a prior semantic segmentation to select suitable markers and seed points. We evaluated our coarse-to-fine algorithm in an ablation study using two mobile LiDAR datasets and one terrestrial LiDAR dataset from three German cities. Our results show that our algorithm outperforms the standard marker-controlled Watershed algorithm in terms of panoptic quality by 3.7, 25.5, and 29.6 percentage points, respectively, while being computationally more efficient than an approach purely based on 3D region growing.

Forest inventories are essential for sustainable forest management and health monitoring. Image-based surveys using unmanned aerial vehicles (UAVs) are increasingly adopted for this purpose. DeepForest, a deep learning model pre-trained on large annotated datasets, enables scalable and cost-efficient tree detection in the resulting imagery. Although it is known that fine-tuning DeepForest to specific target sites improves its performance, optimal fine-tuning strategies remain unclear. In this study, we investigate the fine-tuning of DeepForest on a small dataset of UAV imagery from a temperate mixed forest. We examine the influence of four parameters on model performance: (1) the method used to label training data (manual labeling, manual correction of DeepForest predictions, and automatic labeling using 3D point cloud segmentation), (2) input image resolution, (3) the number of fine-tuning epochs, and (4) the size of the training dataset. Our results show that the out-of-the-box DeepForest model performs poorly on the target dataset but substantially improves with fine-tuning, even when using limited data and a small number of training epochs. Manual and automatic labeling achieve similarly high performance, both outperforming the manual correction approach. Although the model initially performs better on lower-resolution images, this difference diminishes after fine-tuning on higher-resolution inputs. We observe no clear benefit from increasing the training set size, possibly due to the low diversity of our test dataset. Overall, our findings confirm that DeepForest is a suitable base model for forest tree detection in high-resolution UAV imagery and can be adapted to new environments with low effort.

Woody biomass crops are increasingly considered a promising alternative to conventional agricultural systems due to their potential for sustained carbon sequestration under accelerating climate change. Optimizing management practices in such systems is therefore critical to enhance biomass production and carbon storage. In this study, we investigated how management influences biomass allocation in four poplar plots differing in planting density, variety, and harvest-rotation design during their 6th and 7th year of growth. Biomass stocks were quantified for crown, stem, coarse roots, and fine roots. Management effects were most pronounced in aboveground biomass, whereas belowground responses were less consistent. The highest aboveground biomass was observed in the high-density system within the first rotation (MxHD1), reaching 55.32 Mg ha⁻¹ in 2024 and 94.91 Mg ha⁻¹ in 2025. Belowground biomass ranged from 8.12 to 18.35 Mg ha⁻¹ across plots and years. The root:shoot ratio declined with increasing shoot basal diameter and was highest in the year following harvest. Based on these data, we developed general and management-specific allometric models to predict aboveground and belowground biomass from diameter at breast height. Including management factors improved prediction accuracy, supporting more precise quantification of biomass allocation under different cultivation strategies.

Geospatial objects are embedded in geographic space and simultaneously described by multiple thematic attributes. Conventional maps preserve spatial context but offer limited support for exploring thematic relations and multivariate similarity structures beyond geographic proximity. We present an interactive feature-space visualization approach that maps each geospatial object to a multidimensional feature vector and represents it as a particle in a two-dimensional reference plane. Particle positions are determined by thematic, analytical, and spatial magnets whose attraction strength depends on selected feature values or spatial relations. Starting from barycentric target positions, particles are iteratively arranged with collision-aware placement, revealing thematic neighborhoods, outliers, and latent groups independently of the original geographic layout. By configuring magnets interactively, analysts can formulate and test hypotheses about geospatial datasets. We demonstrate the approach using an urban tree inventory derived from mobile mapping LiDAR data and show how it supports the analysis of structural relations, isolated trees, similarity patterns, and terrain-dependent distributions. The results indicate that interactive feature-space visualization provides a useful complementary perspective for exploratory geovisualization and visual analytics.