Large-Scale Mapping of Individual Standing Dead Trees Using Deep Learning and Open Data

Climate change is causing drought stress in temperate forests and promoting the spread of pathogens and pests. The resulting increase in deadwood in numerous forests has highlighted the importance of deadwood mapping at the individual tree level to enable the early implementation of measures aimed at limiting the infestations. While successful deadwood mapping at the individual tree level has been reported, whether these models exhibit the high transferability and speed required for nationwide mapping has yet to be determined. This paper describes large-scale, AI-based deadwood mapping at the individual tree level developed using open data. Specifically, based on color infrared orthophotos with a ground sampling distance of 20 cm, YOLO11 was adapted for instance segmentation of standing deadwood at the state scale. To ensure a high degree of model robustness, acquisition time, radiometry, and forest types were taken into account during the manual annotation process. The dataset consisted of 23,000 instances of deadwood. The best model configuration, which included a postprocessing filter employing the canopy height model, achieved an F1 score of 78% in independent test areas. During the application phase the model was used in five large-scale areas: Arnsberg Forest, Bavaria, Harz, Lake Constance district and Graz. For example, the entire area of Bavaria (71,979 km 2 ) was fully processed in 41 h (2 s/km 2 ) on a NVIDIA Quadro RTX 5000 (16 GB-RAM). Overall, our end-to-end pipeline facilitates large-scale, object-based deadwood mapping while delivering tree coordinates, confidence scores, tree height, crown radius, and acquisition time. As the open data used in this study are updated every 2 years, standing deadwood can be regularly monitored.

Authors

Institutions

Publication Details

Journal
PFG – Journal of Photogrammetry Remote Sensing and Geoinformation Science
Published
2026-11-20
DOI
https://doi.org/10.1007/s41064-026-00428-4
Primary Topic
Remote Sensing and LiDAR Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Large-Scale Mapping of Individual Standing Dead Trees Using Deep Learning and Open Data

Sebastian Briechle, Michael Schmitt, Kilian Sperber
4 links
PFG – Journal of Photogrammetry Remote Sensing and Geoinformation Science
Remote Sensing and LiDAR Applications
article

Large-Scale Mapping of Individual Standing Dead Trees Using Deep Learning and Open Data

Sebastian Briechle, Michael Schmitt, Kilian Sperber
article en

Abstract

Climate change is causing drought stress in temperate forests and promoting the spread of pathogens and pests. The resulting increase in deadwood in numerous forests has highlighted the importance of deadwood mapping at the individual tree level to enable the early implementation of measures aimed at limiting the infestations. While successful deadwood mapping at the individual tree level has been reported, whether these models exhibit the high transferability and speed required for nationwide mapping has yet to be determined. This paper describes large-scale, AI-based deadwood mapping at the individual tree level developed using open data. Specifically, based on color infrared orthophotos with a ground sampling distance of 20 cm, YOLO11 was adapted for instance segmentation of standing deadwood at the state scale. To ensure a high degree of model robustness, acquisition time, radiometry, and forest types were taken into account during the manual annotation process. The dataset consisted of 23,000 instances of deadwood. The best model configuration, which included a postprocessing filter employing the canopy height model, achieved an F1 score of 78% in independent test areas. During the application phase the model was used in five large-scale areas: Arnsberg Forest, Bavaria, Harz, Lake Constance district and Graz. For example, the entire area of Bavaria (71,979 km 2 ) was fully processed in 41 h (2 s/km 2 ) on a NVIDIA Quadro RTX 5000 (16 GB-RAM). Overall, our end-to-end pipeline facilitates large-scale, object-based deadwood mapping while delivering tree coordinates, confidence scores, tree height, crown radius, and acquisition time. As the open data used in this study are updated every 2 years, standing deadwood can be regularly monitored.

PFG – Journal of Photogrammetry Remote Sensing and Geoinformation Science
Munich University of Applied Sciences (DE), Universität der Bundeswehr München (DE)
Life on land, Climate action
Remote Sensing and LiDAR Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon