Making the direct link between light regime and forest biodiversity – a 3D spatially explicit modelling approach based on TLS, CNNs and ray tracing.
Final Report Abstract
Forests play a critical role in maintaining biodiversity and ecosystem services, but understanding their complex spatial and temporal structures remains a challenge. Light distribution, a key abiotic factor, influences ecological processes, making its accurate modelling essential for habitat characterization and forest growth simulations. The project aimed to model light distribution in forest ecosystems by leveraging advanced 3D modelling techniques, including terrestrial LiDAR scanning (TLS), convolutional neural networks (CNNs), and ray tracing. This required various methodological innovations. The project was successful in developing a full pipeline from a LiDAR scanning point cloud to a forest mesh model with spectral properties assigned to each element. Therefore, we implemented a segmentation of single trees, the classification to leaf and wood points, the classification of a species for every tree and the reconstruction of a mesh geometry. Leaf and wood spectra were generated using the TRY trait database and literature values. These mesh models and spectra can be directly utilized to run radiative transfer models. Within this pipeline the deep learning-based species classification developed within the project yielded an overall accuracy of 0.79 (F1 score 0.79) in the classification of 33 species. A test of the whole pipeline on a solitary tree showed very high correlations (r = 0.92) with ground truth data of measurement values of 60 photosynthetically active radiation (PAR) sensors. Additionally, various methodological improvements could be achieved for ground truth validation. A mobile PAR sensor system incorporating a precise positioning system has been developed and we conducted a study on the sensor fusion of thermal imaging and terrestrial laser scanning. The latter showed a significant relationship of surface temperature gradients along the tree stems and sap flow measurements on the same tree.
Link to the final report
https://doi.org/10.5281/zenodo.15280734
Publications
-
Detailed mapping of below canopy surface temperatures in forests reveals new perspectives on microclimatic processes. Agricultural and Forest Meteorology, 341, 109656.
Frey, Julian; Holter, Patricia; Kinzinger, Laura; Schindler, Zoe; Morhart, Christopher; Kolbe, Sven; Werner, Christiane & Seifert, Thomas
-
Sensor fusion of TLS and IR-Imaging in Forest – A new Perspective on the Microclimate. SilviLaser, London
Frey, J., Holter, P., Kinzinger, L., Schindler, Z., Morhart, C., Kolbe, S., Werner, C. & Seifert, T.
-
DetailView model for tree species classification based on single tree LiDAR data (Version v1.0.0) [Computer software]
Frey, J. & Schindler, Z.
-
Dotshadow - create a polygon reconstruction of vegetation from terrestrail LiDAR data for radiative transfer modelling (Version v0.1.0) [Computer software]
Frey, J. & Kröner, K.
-
FOR-species20K dataset
Puliti, S., Lines, E., Müllerová, J., Frey, J., Schindler, Z., Straker, A., Allen, M. J., Lukas, W., Rehush, N. & Hristova, H.
-
Single tree segmentation from terestrial LiDAR forest scans (Version v1.1.0) [Computer software]
Frey, J. & Schindler, Z.
-
Training weights to classify tree species from single tree LiDAR point clouds—The DetailView AI-model. [Computer software]
Frey, J. & Schindler, Z.
-
Benchmarking tree species classification from proximally sensed laser scanning data: Introducing the FOR ‐ species20K dataset. Methods in Ecology and Evolution, 16(4), 801-818.
Puliti, Stefano; Lines, Emily R.; Müllerová, Jana; Frey, Julian; Schindler, Zoe; Straker, Adrian; Allen, Matthew J.; Winiwarter, Lukas; Rehush, Nataliia; Hristova, Hristina; Murray, Brent; Calders, Kim; Coops, Nicholas; Höfle, Bernhard; Irwin, Liam; Junttila, Samuli; Krůček, Martin; Krok, Grzegorz; Král, Kamil ... & Astrup, Rasmus
