News-reported landslide impacts in Southeast Alaska, 1990-2024

1. Abstract
Landslides are common across Southeast Alaska (SEAK) and have caused damages and fatalities in populated areas. Better understanding of landslide triggering conditions and impacts can inform risk reduction strategies and alerting efforts. Landslide inventory data with weekly to hourly time resolution are needed to quantify these conditions. Local news sources are a source of landslide information that often report both timing and effects. Here, we present an inventory of 162 news-reported landslides in SEAK between 1990 and 2024. This data release focuses on landslides reported near seven population centers in Alaska with long-term meteorological records: Juneau, Haines, Sitka, Petersburg, Wrangell, Prince of Wales Island, and Ketchikan. The inventory was created by searching the internet for the keyword “landslide” in 13 different local news sources. The dataset is a comma separated value table (.csv) with columns representing the minimum and maximum date and time of occurrence for each landslide, the closest population center, latitude and longitude, an estimated location accuracy, a description of the location, the landslide type, the potential trigger, and a description of the size and initiation point of each landslide. Where available, the number of fatalities and buildings damaged and a description of the effects were documented. The primary news source, publication date, and links to the news article(s) describing each event are also included. 
This data release complements existing sources of landslide information in SEAK, for example the Alaska Department of Transportation and Public Facilities Geo Event Tracker (Alaska Department of Transportation and Public Facilities, 2022), the U.S. Forest Service Tongass National Forest Landslide Inventory (U.S. Forest Service, 2024), and the National Oceanic and Atmospheric Administration Storm Events Database (National Centers for Environmental Information, 2020), by providing high temporal resolution and impact descriptions for many of the reported events.

This data release includes: (1) a comma separated variable (csv) table documenting each landslide with its attributes (SEAK_News_Reported_Landslides.csv) (2) a table with data field definitions and descriptions (SEAK_News_Reported_Landslides_Definitions.csv)

Disclaimer: Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
2. Data processing

We compiled local news sources for seven populated areas with long-term meteorological records (Juneau, Haines, Sitka, Petersburg, Wrangell, Prince of Wales Island, and Ketchikan) by searching by name and news sources on the internet, for example, “Ketchikan news sources.” We verified that the resulting list of 13 news sources was complete by consulting with the Alaska State Library. We targeted the date range 1990 to 2024 but note that not all sources covered this entire time period and that more news reports were available for more recent years. The record length and links to each source can be found in the metadata entity and attribute explanation.
We used the search bar on the website of each news source and entered the keyword “landslide” to narrow down the articles. When applicable, the results were filtered to only show articles from 1990 to present. 
We read through the articles returned from the search results and determined if the article mentioned the occurrence of a landslide in proximity to one of the focus population centers. We did not apply a distance criteria to include or exclude landslide reports; if a focus population center was mentioned in a news article in conjunction with a landslide, the landslide was included. Landslides that occurred in other areas were excluded.
If the article described a landslide that had not previously been identified in a reviewed news source, it was entered as a new event. If an article contained new information about a landslide that was already identified in another news source, it was added as an additional source and extra information was added to the corresponding columns. If an article reported multiple landslides during one event each landslide was recorded as a separate entry in the dataset and an indicator added to the location description (Loc_des) column (e.g. Slide 1 of 3).  This count only includes landslides reported in the news articles.
For each landslide, the exact or minimum and maximum date of occurrence, and if reported, the exact, approximate, or minimum and maximum time of occurrence were recorded.
To determine the latitude and longitude, Google Maps was primarily used, using the description found in the location description (Loc_des) column. If it was not possible to locate the landslide using Google Maps, Google Earth Pro was used. For example, for any landslide where the location description was a mile post number on a road, the latitude and longitude were found by tracing the road on Google Earth Pro and using the “show elevation profile” functionality to travel to the milepost, where the latitude and longitude were taken from. In many cases, the latitude and longitude of the landslide had to be estimated based on the description. The estimated accuracy of the latitude and longitude is reflected in the location accuracy (Loc_acc) column.
If the article used the words “rockfall”, “landslide”, or “mudslide” to describe the movement, these terms were included in the landslide type (LS_type) field. Based on the description of the failure and/or photos included with the article, we interpreted the failure type as rockfall, deep-seated, shallow, debris flow, or unknown, and recorded this in the landslide type interpretation (LS_type_int) field. 
We assigned a potential trigger of rain (the article discussed recent rainfall), rain/snow (the article discussed recent rainfall/snowfall), other (the article did not discuss rain, rain on snow, or wind, and may be human triggered), unknown (the article did not describe any trigger), or wind (the article discussed recent wind conditions and the article did not mention rain).  These are documented in the potential trigger (Pot_trigg) field. 
For each landslide entry, all additional fields were filled out as completely as possible based on information in news reports. These include the number of fatalities, buildings damaged, other effects, size and initiation site description. The amount of detail on each landslide available from the news sources reviewed was highly variable, such that not all fields could be completed for each landslide.

3. References
Alaska Department of Transportation and Public Facilities, 2022, Geo Event Tracker, accessed April 15, 2025, at https://www.arcgis.com/home/item.html?id=7d34a6dc51194296be7a38fd965c22…;

National Centers for Environmental Information, 2020, Storm Events Database (Version 3.1), accessed April 15, 2025, at https://www.ncdc.noaa.gov/stormevents/

U.S. Forest Service, 2024, Tongass Landslide Initiation and Areas, accessed April 15, 2025, at https://data-usfs.hub.arcgis.com/datasets/usfs::tongass-landslide-areas…