What we measure and publish: open datasets, satellite observations and field notes. Each one names its sources, explains its method and says what it cannot show.
Satellite radar, Diomede Islands, Bering Strait, Nov 2024 and Mar 2025
Open datasets
Every dataset we publish is free under CC BY 4.0, with a documented method and a validation record. Arctic Open-Water Window 1.1, Western Terminal Lakes 1.1 and Storm Blackout and Restoration 1.2 each have a DOI on Zenodo. The full list, with the coverage and period of each dataset, is on the Data page.
Thirty-eight Sentinel-1 passes at a twelve-day interval record icebergs leaving the icefjord, the freeze and break-up of Disko Bay, and the seasonal advance and retreat of sea ice over fifteen months.
Short written pieces about a place, a record or a dataset. Field notes are written for some releases: what the data shows, how it compares with existing products, and its limits. Corrections are recorded on the note itself.
The National Register's mapped boundary for the Beverly Drive-In lies about 280 meters north of the theater's footprint in aerial imagery. Surveys from 2012 and 2023 record the site's demolition and redevelopment, and a listing that has not been updated.
Results are compared with independent records, such as gauge readings, outage records or published maps, and the comparison is published with the data.
Uncertainty stated
Lake areas carry an uncertainty range. Other tables carry quality flags.
Checked before release
Automated checks run on every update. An update that fails them is not published, and the last version that passed stays live.
Versioned
Each release has a version and a date. Arctic Open-Water Window 1.1, Western Terminal Lakes 1.1 and Storm Blackout and Restoration 1.2 each have a DOI on Zenodo. Earlier versions stay available to cite.
Corrected in public
Errors are fixed in a new version and recorded in the changelog and, where the release has a field note, on that note.
How we measure
Every published number comes from a written chain: the raw satellite or record data, the masking and quality rules, the measurement, an independent check, and a stated uncertainty. Each dataset's method note gives the full detail.
Sea ice
Source
NSIDC daily sea-ice concentration: the Climate Data Record (G02202 v6), with the near-real-time product (G10016 v4) for the latest days, marked provisional until the final record replaces it.
Reading
Concentration averaged over fixed gates across each Arctic passage. A day counts as open water below 15, 30 or 50 percent ice, reported separately.
Check
Agrees with NSIDC's MASIE ice charts on 96.3 percent of clear days (52 of 54). Break-up and freeze-up days are harder and are reported on their own.
Images at 80 m for the 15 gates, 2015 to 2026, for the weeks that have a usable one. Radar sees through cloud and polar night, so it fills the gaps optical imagery leaves.
Check
A radar freeze-up test on held-out days in the Bering Strait agreed with MASIE on 86.7 percent of days, above our 85 percent release threshold.
Surface water
Source
Copernicus Sentinel-2, surface reflectance.
Reading
A monthly composite of every usable scene. Cloud, shadow and cirrus are removed; ice and snow are masked from November to April and never counted as water or dry lakebed. Area is measured in an equal-area projection inside a fixed outline for each lake.
Check
Compared with USGS gauge levels and level-to-area tables. Every month carries a range, the share of the lake actually seen, and a quality flag; a month that jumps more than 10 percent from its three-month median is flagged.
Power outages
Source
U.S. Department of Energy EAGLE-I county outage records (every 15 minutes since November 2014) and NASA Black Marble nightly light (VNP46A2).
Reading
From utility records: the peak, customer-hours without power, and hours to 50 and 90 percent restored. From night light: clear nights only, a five-night median against each county's normal, and days to recovery. A county's window ends when the next storm starts.
Check
Each release includes a table of how well the satellite readings agree with the utility records, storm by storm, and says where records are missing.
Open-source tools we build on: GDAL and rasterio for reading and writing imagery, xarray and rioxarray for gridded data, GeoPandas, Shapely and PROJ for boundaries and projections, and STAC catalogues (pystac-client, odc-stac) for finding and loading scenes. The masking, compositing, validation, release checks and packaging are our own code.
Arctic. Not for navigation. Use at your own risk. Provided as is, without warranty of any kind.
Lakes. Not for water-rights, legal or engineering decisions.
Storms. Estimates from public records and satellite night light. Not for real-time emergency response or for any decision about one address or property.
