Institute for Geoinformatics
Spatio-Temporal Modelling Lab
- Frederick Bruch: [email protected]
- Jakob Danel: [email protected]
Time frame: Winter Term 2023/2024
- Dr. Christian Knoth
- Johannes Heisig
- Prof. Dr. Edzer Pebesma
The combination of tree species is important for the functionality and biodiversity of a forest [1]. Depending on the particular kind of tree, there are various threats to both the environment and people. An example therefore is the oak processionary moth [3] or the bark beetle [2]. Acquiring this knowledge is also necessary to check a forest’s quality in terms of biodiversity, monocultures, and habitats for specific species. It is evident that gaining a thorough understanding of forests is crucial, particularly with regard to the (spatial) distribution of different tree species. The goal of this project, is to research different parameters of monocultural forest areas and analyze their potential as a parameter for differentiation between tree types. Within the four species beech, oak, spruce and pine we want to research different parameters, such as height, distance between trees or canopy density in example areas. The project is dividable into three different topics:
- Data acquisition: Use data from waldmnoitor.de^1 to identify areas which has which are populated by the researched species. This data are generated by analyses of Sentinel^2 from 2017 [4]. We will identify such areas for NRW automatically if we are provided with the data from the creator of Waldmonitor or by hand if we can not get access to this data, following the markings on the map.
- Data Preprocessing: To preprocess the data for their usage, we will down- load and retile the datasets, to ensure that the data can easily be handled in memory. We want to detect human made artifacts (houses, perches, etc.) and remove them from the point cloud. Lastly, we perform a tree detection and a segmentation by individual trees.
- Analyze and differentiation: The following questions we want to answer for each species and analyze the differences between these answers.
- Are the point patterns of the detected trees randomly distributed, or are their clusters/anti-clusters?
- How are the tree heights distributed?
- How much distance have a tree to his direct neighbors?
- How is the intensity of the signals distributed for each tree along the z-axis?
- How are the number of return points distributed within a species?
For solving this problem, we would use R^3 with the package lidR^4. We would sup- ply R markdown files with the obtained results. We want to ask the owner of the waldmonitor.de data to gain us access to their data for an exact differentia- tion between the species areas. As LiDAR data, we want to use data from the OpenGeodata.NRW^5 portal. In the end, we want to produce two main results in the project: Firstly, we want to report and visualize the results of the analysis in a meaningful way, as well as providing context and limitations for the results. Also, we write an R-Project which
is able to reproduce the analysis within different areas.
- (^1) https://waldmonitor-deutschland.de/ last visited: November 7, 2023
- (^2) https://sentinels.copernicus.eu/web/sentinel/homelast visited: November 7, 2023
- (^3) https://www.r-project.org/last visited: November 7, 2023
- (^4) https://r-lidar.github.io/lidRbook/last visited: November 7, 2023
- (^5) https://www.opengeodata.nrw.de/produkte/geobasis/hm/3dm_l_las/3dm_l_las/ last visited: November 7, 2023
- [1] Akira S. Mori, Kenneth P. Lertzman, and Lena Gustafsson. Biodiversity and ecosystem services in forest ecosystems: a research agenda for applied forest ecology.Journal of Applied Ecology, 54(1):12–27, 2017.
- [2] Martin M ̈uller and Nadja Imhof. K ̈aferk ̈ampfe: Borkenk ̈afer und landschaftskonflikte im nationalpark bayerischer wald. Landschaftskonflikte, pages 313–329, 2019.
- [3] Thomas Sobczyk. Der Eichenprozessionsspinner in Deutschland: Historie, Biologie, Gefahren, Bekämpfung. Deutschland/Bundesamt f ̈ur Naturschutz, 2014.
- [4] Torsten Welle, Lukas Aschenbrenner, Kevin Kuonath, Stefan Kirmaier, and Jonas Franke. Mapping dominant tree species of german forests. Remote Sensing, 14(14), 2022.
git clone https://github.com/jakobdanel/lidar-forest-analysison your local machine- Install R (see R-Project)
- Install the dependencies listed below
- Run the functions from the root of this project.
The results of the course analysis can be found on Github Pages
In the following sections the maintaining, building and deployment of the report is documented.
The report is saved inside the results directory. The report is organized in a directory type style, means that each section becomes an own directory (e.g
results/methods). The .qmd files needed to be inserted into results.report.qmd to be displayed. This can be achieved by:
{{< include path/to/subsection/file.qmd >}}
Please make sure that only the *.qmd files, helper files for building and computations (e.g. the results/_freeze directory needs to be registered) are tracked by git.
For referencing add a .bibtex entry into the file results/references.bib. You can cite by using @citeKey as inline reference or normal via [@citeKey] (More information on quarto website)
For building the reports pdf:
./build_report.sh
For deploying the pages to GitHub Pages:
quarto publish gh-pages
GEDIcalibratoR
future
lidR
sf
dplyr
- first ideas: train model to detect single tree species
- dive into R scripts to gather LIDAR data
- setup local working environment
- present idea
- look for training data (dataset with categorised tree species)
- cancel initial plan because we didnt find any data
- new plan: compare forest with (dominant) tree species
- implement data gathering, retiling, detection and segmentation