Does the implementation of Automatic Individual Tree Crown Delineation (ITCD) impact the early detection of bark beetle (BB) infestation in Norway spruce forests?
Early detection of bark beetle (BB) infestations in Norway spruce forests is essential for effective forest management. While UAV imagery offers high-resolution data, selecting appropriate crown pixels to detect subtle spectral changes during early BB infestation stages is still a challenge that nee...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2025-07-01
|
| Series: | The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| Online Access: | https://isprs-archives.copernicus.org/articles/XLVIII-G-2025/205/2025/isprs-archives-XLVIII-G-2025-205-2025.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Early detection of bark beetle (BB) infestations in Norway spruce forests is essential for effective forest management. While UAV imagery offers high-resolution data, selecting appropriate crown pixels to detect subtle spectral changes during early BB infestation stages is still a challenge that needs further investigation. This study examines the impact of automatic Individual Tree Crown Delineation (ITCD) methods in detecting early-stage BB infestations, particularly during the green-to-yellow stage. On July 19, 2022, high-resolution multispectral UAV imagery (2 cm) was acquired using the DJI Phantom 4 Multispectral sensor over a 4-hectare forest plot in Krkonoše National Park, Czech Republic. Treetop detection was performed using local maxima filtering, while four ITCD algorithms: Buffer, Marker-Controlled Watershed Segmentation, thiessen polygons, and seeded region growing, were used for crown delineation. Spectral data from five bands and five vegetation indices were extracted for each automatic ITCD method, as well as for manually delineated crowns, across 11 infested and 11 healthy trees. Spectral separability was assessed using the Mann-Whitney test. The findings revealed that the 3-meter fixed window filter effectively detected treetops but encountered challenges with double detections and missing smaller trees. Seeded region growing proved the most accurate for crown delineation. Statistical analysis showed that red-edge and near-infrared spectral bands, along with vegetation indices (NDVI, GNDVI, OSAVI, and RENDVI), successfully separated healthy from infested trees using both automatic ITCD and manual delineation. However, manually delineated crowns exhibited greater sensitivity to spectral variations, especially in the red band, making manual delineation more effective for early-stage BB detection. While automatic ITCD methods excelled in detecting Excess Green Index (ExG) differences. Though, automatic ITCD methods are computationally efficient, manual delineation or refinement of automatic ITCD is needed for accurate monitoring of subtle spectral changes during BB infestations (green-to-yellow transition). Precise crown delineation and early BB detection rely on high-quality pre-processing, expert knowledge (of infestation stages by foresters), and field observations (e.g., tree positioning using GPS or total station and BB symptoms), with multitemporal imagery aiding in tracking infestation progression within the tree crowns. |
|---|---|
| ISSN: | 1682-1750 2194-9034 |