A Landslide Inventory for Prince of Wales Island, Alaska from 2009 to 2023 Using Planet Imagery
| dc.contributor.advisor | Roering, Josh | |
| dc.contributor.advisor | Shoop, Casey | |
| dc.contributor.advisor | Patton, Annette | |
| dc.contributor.author | Lawrence, Eliza | |
| dc.date.accessioned | 2024-12-12T22:46:37Z | |
| dc.date.available | 2024-12-12T22:46:37Z | |
| dc.date.issued | 2024-11 | |
| dc.description | 52 pages | |
| dc.description.abstract | The combination of extreme rainfall brought on by atmospheric rivers and steep topography has made landslides a frequent and hazardous occurrence on Prince of Wales Island (PoW) in Southeast Alaska. Detecting and monitoring landslides on the ground can be costly and time-consuming, especially in remote and inaccessible locations. By using high-resolution (3-5 m) Earth observation imagery from Planet Labs, we can pinpoint the location and constrain the timing of past landslides through visual change detection of multiple images acquired every 2-3 days. We have mapped over 750 landslides across PoW (6670 kmĀ²) and the surrounding islands from 2009 to 2023. By comparing pre- and post-landslide failure images, we determined landslide timing with temporal constraints ranging from one year to less than one week. This study expands upon the USFS Tongass National Forest Landslide Inventory, which utilizes a combination of decadal aerial photos and high-resolution satellite imagery, providing multi-year constraints on landslide timing. The inventory presented in this study includes landslide polygons, corresponding initiation points, and an attribute table containing the last pre-failure date, first post-failure date, slope, elevation, forest stand age, geology, distance to the nearest road, and area. The distribution of landslides across PoW over the past decade demonstrates a strong seasonality, with most landslides occurring between August and December. We find a strong relationship between landslide location and aspect, with landslides predominantly occurring on southern slopes. Our methods also capture small landslides, revealing that most are located within 10 km of the nearest road. We find that cumulative precipitation, both monthly and annually, does not fully explain the relationship with landslide frequency and requires further exploration. Future research will use gridded precipitation data, coupled with well-characterized timing of landslides, to determine precipitation intensity thresholds for landslide triggering. By quantifying the temporal and spatial distribution of landslide occurrences, we can provide valuable information to support the development of landslide early-warning systems. | |
| dc.identifier.orcid | 0000-0002-3406-1296 | |
| dc.identifier.uri | https://hdl.handle.net/1794/30242 | |
| dc.language.iso | en_US | |
| dc.publisher | University of Oregon | en |
| dc.rights | Creative Commons BY-NC-ND 4.0-US | |
| dc.subject | landslide, remote sensing, hazard, inventory, Alaska | |
| dc.title | A Landslide Inventory for Prince of Wales Island, Alaska from 2009 to 2023 Using Planet Imagery | |
| dc.type | Thesis |