Three-dimensional Forest Ecosystem Monitoring and Better Understanding by Terrestrial-based Technologies - 3DForEcoTech COST Action
Martin Mokros
Czech University of Life Sciences Prague, Czech Republic
In recent years, there has been rapid development in close-range technologies for generating highly detailed three-dimensional (3D) point clouds within forest environments. LiDARs, which are part of terrestrial and mobile laser scanner technology, can be mounted on various devices and provide accurate data about individual trees. Additionally, advancements in computational power and algorithms have allowed for processing hundreds of 2D images into 3D point clouds through photogrammetry. Sensors equipped on unmanned aerial vehicles can also provide close-range above-canopy data.
Despite the technological advancements, the application of these technologies lags behind their development. Research often focuses on evaluating a particular technology's potential rather than addressing fundamental issues. For example, researchers have yet to determine how to place TLS within different forest types to fully reconstruct research sites and detect and measure biometric variables of all trees.
Processing the 3D point cloud data is also a significant challenge. Selecting the appropriate technology and data collection approach requires careful consideration, and processing the data with available algorithms is complicated. While algorithms are available from various research groups, the complexity of processing solutions varies based on the target characteristics that researchers want to measure.
To address these challenges, a network of scientists has been established under the 3DForEcoTech COST Action, with over 320 scientists from 50+ countries. The goal of the network is to synchronize knowledge, develop general protocols for data acquisition, processing, and fusion for forest inventory and ecological applications, and make novel technologies available to a broader audience. By unifying efforts, the network aims to progress the application of close-range technologies and unlock their full potential for forest assessment.
This contribution is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
Software solutions for close-range forest point clouds: What is out there?
Carlos Cabo1, Martin Mokros2, Arnadi Murtiyoso3, Arunima Singh2, Dimas Pereira5, Jaz Stoddart4
1University of Oviedo, Spain; 2Czech University of Life Sciences, Czech Republic; 3ETH Zurich, Switzerland; 4Bangor University, United Kingdom; 5University of Leon, Spain
Terrestrial point cloud technologies are increasingly used in forest inventories; however, using such data is challenging for potential users because the processing pipeline is typically long and sophisticated. Furthermore, operators are not necessarily familiar with the techniques involved. Currently, the number of available software solutions is still small, especially compared to the amount of research publications that describe new algorithms for dealing with dense point clouds in forest environments. To date, there is no repository, list, or database with a comprehensive compendium of publicly available software solutions for forest monitoring using terrestrial point clouds.
In the past few years, some benchmark projects and publications have successfully assessed the performance of a limited number of research and proprietary algorithms for forest inventory from terrestrial point clouds. These projects gave an accurate overview of the current state of the art and capabilities of terrestrial 3D technologies at plot level. However, they were not focused on the availability or suitability of the algorithms for end users, and only a few of the analysed algorithms had a public implementation. Therefore, some key aspects are still missing: a compendium of publicly available software solutions that are accessible for potential end users, and an assessment of their performance. Within the COST Action 3DForEcoTech we are compiling a comprehensive list of publicly available software solutions for processing close-range point clouds from forests. We organise different activities (meetings, public questionnaires, surveys, etc.) that allow us to identify existing solutions and to keep the list updated with new additions. From the software solution list, we are creating a database with descriptions, requirements, installation and operation instructions, benchmarks, and a web-based information system to manage and give public, user-friendly access to all that information.
This contribution is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
A web platform for forest point cloud processing algorithms
Martin Mokros1, Nataliia Rehush2, Arnadi Murtiyoso3, Carlos Cabo4, Arunima Singh1, Wout Cherlet5, Mirela Beloiu3
1Czech University of Life Sciences, Czech Republic; 2Swiss Federal Research Institute WSL, Switzerland; 3ETH Zurich, Switzerland; 4University of Oviedo, Spain; 5Ghent University, Belgium
In recent years, the use of 3D point clouds in the forestry sector has seen a significant increase in interest. With the development of novel technologies such as laser scanning, an increasing number of algorithms have also been developed in parallel to process 3D data into more tangible results for forestry applications. From this variety of algorithms available, it might be challenging for users to decide which one to choose to fulfill their goals best. Within the framework of 3DForEcoTech COST Action, a comprehensive database was compiled to collect information about existing forest point cloud processing algorithms in one place. The database currently includes 25 algorithms. The database will be published as a web-based platform, in which users may consult it easily using a query system (Fig. 1). For each of the algorithms, metadata was collected while installation and test runs to assess their applicability for forestry. Moreover, technical guides on installation and general use were written and will be included in the web platform. Additionally, a hackathon will be organised to benchmark the algorithms in identical settings using datasets representing various forest stands. The performance of the algorithms will be evaluated in terms of accuracy, time and memory consumption, level of user interaction needed, among others. The results will be reflected on the platform and should help users in choosing which solution suits their needs best and meets the required accuracies. In this way, the web platform may serve the community as a single source of information to select a specific software/algorithm that works for their requirements. Furthermore, the online nature of the platform means that it will evolve in time with regular updates of new algorithms and features.
This contribution is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
Challenges and Opportunities in Retrieving Forest Allometric Traits Using Close Range Sensing for Allometric Modeling
Meinrad Abegg1, Barbara D’hont2, Aline Bornand1, Kim Calders2, Hossein Chavoshi3, Ana Novo4, Henri Riihimäki5, Ninni Saarinen6, Enrico Tomelleri7, Tom Verhelst2, María Menéndez-Miguélez8
1Swiss Federal Institutefor Forest, Snow andLandscape Research WSL; 2Ghent University; 3Norwegian University of Life Sciences; 4Geotech Group, CINTECX, Department of Natural Resources and Environmental Engineering, School of Forestry Engineering, University of Vigo; 5Norwegian University of Life Sciences; 6University of Helsinki; 7Free University of Bolzano; 8Instituto de Ciencias Forestales (ICIFOR-INIA), CSIC
Allometric models play a crucial role in forest monitoring, e.g., converting diameter measurements into biomass. Many studies show the potential of close-range remote sensing (CRS) technologies in forest ecosystem assessment. However, allometric models are still based on traditional measurements, and there is no established procedure to include CRS in developing new allometric models. Within the Cost Action 3DForEcoTech, a wiki will be developed to document the challenges and opportunities in collecting and analyzing forest allometric data using different CRS techniques for comparing with classical approaches and develop new advanced models.
The wiki's main objective is to provide practitioners with a hands-on guideline on the steps to be taken combining existing allometric models with CRS technologies. The wiki contains a general introduction to approaching a CRS survey. Additionally, an overview of data, collections and models for sharing are compiled to facilitate collaboration. Specific information about scanners, including TLS, MLS, photogrammetric, and UAVLS point clouds, is documented and reviewed in detail.
The link between CRS features and scanner types provides a comprehensive overview of the potential connections between the two. Firstly, traditional allometries are evaluated regarding input datasets, tree components or statistics considerations and goodness of fit statistics. Secondly, the way forward with CRS data is discussed in terms of sample types, ground truth, and new CRS input data, emphasizing the importance of explanatory variables and their connection with the acquisition approach.
Validation considerations are also discussed in terms of the conditional use case. This open document will support future internationally standardized forest data collection and model development for a broad range of purposes.
Scan4est: Measuring Spatiotemporal Changes in Finnish Forest Ecosystem
Ninni Saarinen1, Juha Hyyppä2, Mikko Vastaranta1
1School of Forest Sciences, University of Eastern Finland, Finland; 2Finnish Geospatial Research Institute, National Land Survey of Finland, Finland
Currently, the lack of open science-driven long-term forest monitoring experiments and the lack of purpose-specific automatic and digital technologies have limited our understanding of natural and anthropogenic processes changing forested ecosystems. We present Scan4est Research Infrastructure (RI) that is a living laboratory in a Finnish boreal forest ecosystem established for exploring forest structure and developing new technologies for forest mensuration.
The foundations of the Scan4est RI are annually acquired laser scanning data sets capturing changes in trees and forests, linked with microdensitometer measurement characterizing internal wood properties. The RI includes four elements, supporting each other: i) laser scanning change monitoring infrastructure, including terrestrial and airborne laser scanning instruments; ii) experimental and automatized laser scanning infrastructure (e.g. above- and under-canopy drones, backpack and hand-held laser scanning, hourly terrestrial laser scanning monitoring system) for tree characterization; iii) X-Ray microdensitometer for measuring wood properties; and iv) test site and key tree characteristics derived from a variety of laser scanning sensors from more than 10 000 trees growing on 120 sample plots.
The Scan4est RI offers possibilities to test new sensors, use RI’s quality-controlled data for algorithm development, as well as provides various data products ranging from raw data to pre- or fully analyzed products and open-source algorithms for scientists, R&D personnel, and decision makers.
Finally, the Scan4est RI will be included in the international network StructNet for global vegetation structure monitoring enhancing. Similar collaborations are actively sought and very welcomed.
This publications is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
StrucNet: A global network for automated vegetation structure monitoring
Kim Calders1,2, Benjamin Brede3, Glenn Newnham4, Darius Culvenor5, John Armston6, Harm Bartholomeus7, Anne Griebel8, Jodie Hayward9, Samuli Junttila2, Alvaro Lau7, Shaun Levick9, Rosalinda Morrone10, Niall Origo10, Marion Pfeifer11, Jan Verbesselt7, Martin Herold3
1CAVElab - Computational & Applied Vegetation Ecology, Department of Environment, Ghent University, Coupure links 653, 9000 Gent, Belgium; *contact: kim.calders@ugent.be; 2School of Forest Sciences, University of Eastern Finland, 80101 Joensuu, Finland; 3Helmholtz Center Potsdam GFZ German Research Centre for Geosciences, Section 1.4 Remote Sensing and Geoinformatics, Telegrafenberg, Potsdam, 14473, Germany; 4CSIRO, Research Way, Clayton Vic 3168, Australia; 5Environmental Sensing Systems, Bentleigh East, Vic 3165, Australia; 6Department of Geographical Sciences, University of Maryland, College Park, USA; 7Laboratory of Geo-Information Science and Remote Sensing, Wageningen University, 6708 PB, Wageningen, the Netherlands; 8Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, New South Wales 2751, Australia; 9CSIRO, 564 Vanderlin Drive, Berrimah, NT 0828, Australia; 10Climate and Earth Observation group, National Physical Laboratory, Hampton Road, Teddington, London, UK; 11School of Natural and Environmental Sciences, Newcastle University, NE1 7RU, Newcastle Upon Tyne, UK
Climate change and increasing human activities are impacting ecosystems and their biodiversity. Quantitative measurements of Essential Biodiversity and Climate Variables (EBVs/ECVs) are used to monitor biodiversity and carbon dynamics and evaluate policy and management interventions. Ecosystem structure is at the core of EBVs and carbon stock estimation and can help to inform assessments of species and species diversity. Ecosystem structure is also used as an indirect indicator of habitat quality and expected species richness or species community composition. Spaceborne measurements can provide large-scale insight into monitoring the structural dynamics of ecosystems, but existing sensors have not been designed for vegetation monitoring and do not have sufficient spatiotemporal coverage to quantify their full three-dimensional vegetation structure at local scales. Here we demonstrate the potential of high-frequency ground-based laser scanning to systematically monitor structural changes in vegetation. We present a proof-of-concept high-temporal ecosystem structure time series of five years in a temperate forest using terrestrial laser scanning (TLS). We also present data from automated high-temporal laser scanning that can allow upscaling of vegetation structure scanning, overcoming the limitations of a typically opportunistic TLS measurement approach. Automated monitoring will be a critical component to build a network of field monitoring sites that can provide the required calibration data for satellite missions to effectively monitor the structural dynamics of vegetation over large areas. We will discuss our perspective and reflect on how this network could be designed and discuss implementation pathways. This contribution is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
Occlusion mapping tools for point cloud quality assessment in forest laser scanning
Benjamin Brede1, Daniel Kükenbrink2, Bernhard Höfle3, Teja Kattenborn4, Lennart Klinger5, Timo Pitkänen6, Arunima Singh7, Hannah Weiser3
1Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, Section 1.4 Remote Sensing and Geoinformatics, Telegrafenberg, 14473 Potsdam, Germany; 2Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Zürcherstrasse 111, CH-8903 Birmensdorf, Switzerland; 33DGeo Research Group, Institute of Geography, Heidelberg University, Im Neuenheimer Feld 368, 69120 Heidelberg, Germany; 4Remote Sensing Centre for Earth System Research (RSC4Earth), Leipzig University and Helmholtz Centre for Environmental Research, Talstr. 35, 04103 Leipzig, Germany; 5Laboratory of Geo-Information Science and Remote Sensing, Wageningen University, 6708 PB, Wageningen, the Netherlands; 6Natural Resources Institute Finland (Luke), Latokartanonkaari 9, FI-00790 Helsinki, Finland; 7Faculty of Forestry and Wood Sciences, Czech University of Life Sciences Prague, Prague, 16500, Czech Republic
Many factors influence the point cloud quality in forest applications including laser ranging error, registration quality between view positions and presence of noisy points. All of these factors have an influence on feature extraction. Occlusion of canopy volume by present vegetation is routinely discussed as a negative influence. However, quantitative assessment of occlusion is rarely done and there are no standardized, user-friendly workflows that help in the quantification. Additionally, different applications have different requirements on the level of occlusion that is acceptable to still achieve reliable results. First, this contribution will review mechanisms that lead to occlusion in point clouds from the perspective of ray tracing, i.e., by knowing the sources, trajectories and reflection points of single beams. Major factors will be identified and recommendations will be given to prevent occlusion. Furthermore, potential use cases will be discussed that make use of geo-located occlusion data, e.g. by quantifying uncertainty of derived metrics. Second, existing tools with the potential for occlusion mapping will be assessed according to predefined criteria and a benchmark exercise. Criteria include requirements on the operating system and operating environment, ease of setup, computation speed, input file formats, and outputs. The most promising tool will be identified and discussed with respect to missing features that make it relevant for end users. Overall, this contribution will raise awareness of how occlusion is produced in scanning campaigns, how it can be quantified, integrated in analysis as input information, and how it can be counteracted. This contribution is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
The spaces between: volume-based estimation of gap fraction in tree crowns
Yunfeng Zhu1, Dongni Li1, Jiangchuan Fan2, Huaiqing Zhang3, Markus Peter Eichhorn4,5, Xiangjun Wang6, Ting Yun7
1School of Information Science and Technology, Nanjing Forestry University, Nanjing, China; 2National Engineering Research Center for Information Technology in Agriculture, Beijing, China; 3Research Institute of Forestry Resource Information Techniques, Chinese Academy of Forestry, Beijing, China; 4School of Biological, Earth and Environmental Sciences, University College Cork, Cork, Ireland; 5Environmental Research Institute, University College Cork, Cork, Ireland; 6Rubber Research Institute, Chinese Academy of Tropical Agricultural Sciences, Haikou, China; 7Forestry College, Nanjing Forestry University, Nanjing, China
Tree crowns, despite their apparent size, are mostly composed of empty space. Gap fraction, or porosity, is a critical variable for understanding the movement of light, air and water through forest canopies. Current estimation methods rely on 2D projections from hemispherical photography or equivalent scanning approaches, or optical transmission through voxel-based reconstructions. In this work we attempt a complete reconstruction of canopies from terrestrial laser scanning using computer graphics and apply porous media theory to directly estimate the void fraction. Woody volume was estimated by cylinder models which were fit to branch segments. To capture effective leaf volume we constructed hexagonal prisms which account for the curl and drooping of individual leaves and enclose all scanned points. These enabled calculation of total woody and foliage volume. Five individual trees of different species and a complete stand of 24 rubber trees with a closed canopy were scanned using a Leica C10. Gap fraction for single tree crowns ranged from 0.953–0.985, and the plot-level estimate for the full canopy was 0.970. These values substantially exceed those obtained through 2D projections and are moderately higher than from voxel-based estimators. By using an equivalent approach to porous media theory we open the potential to draw from a wider body of interdisciplinary work, with tree canopies demonstrating a gap fraction comparable to materials such as sponges or expanded polystyrene. A complete representation of the distribution of elements will enable more sophisticated modelling of their physical properties and behaviour beyond simplifications such as LAI. This presentation is based upon work from COST Action CA20118, supported by COST (European Cooperation in Science and Technology).
|