Methods and limits
How the publication is made
Configuration precedes analysis. Provider outputs are normalized into small static files, verified, assigned checksums and then bundled into the public site.
Study boundary
The project uses a compact rectangular study area around Cambridge Bay. It is an analytical boundary, not an official port, municipal or administrative boundary.
Surface change
Cloud-, shadow-, snow- and saturation-masked Landsat 8 surface-reflectance scenes are reduced to summer median composites. Euclidean spectral distance across six optical bands produces a change magnitude. Pixels above the configured threshold are vectorized as candidates. The six bands are Landsat 8 OLI bands 2 to 7 (blue, green, red, near-infrared, shortwave infrared 1 and 2). The thermal bands (10, 11) and the panchromatic band (8) are not used: they measure a different quantity or carry no spectral information to compare. See the Surface change layer page for the wavelengths.
Sea-ice classification
The pipeline inventories Sentinel-1 acquisition combinations, selects the most available consistent mode and polarization, and calculates monthly median backscatter. It applies a fixed threshold over a static water mask and repeats the area calculation at plus and minus the sensitivity interval. NOAA OISST supplies a coarse seasonal reference.
Operational sea-ice monitoring in the Canadian Arctic (Canadian Ice Service charts) is produced by analysts combining several sensors and reporting ice with the WMO egg code and MANICE procedures: concentration, stage of development and floe size. This project's single-threshold C-band SAR classification is a screening baseline for ice versus open water. It does not classify stage of development, floe size or ridging, and it is not a substitute for an ice chart.
Calibration status: a 2024 check of single HH and HV thresholds against NSIDC concentration did not agree
with the seasonal ice cycle (best correlation 0.40), so no ice threshold is validated and the ice layer is
published as a demonstration only. The test, its numbers and its limits are in the calibration note
(docs/CALIBRATION.md) in the project files.
Vessel events
The official Global Fishing Watch Python client requests AIS gap events of at least twelve hours and an aggregated SAR-presence report inside the same geometry. Geometry and selected descriptive fields are published. No intent score is added.
Projection and area
The website displays geographic coordinates. Earth Engine’s pixel-area image is used for area statistics so square degrees are never treated as square metres. The configured polar projection is available for exported cartographic work.
The interactive map is drawn in Web Mercator for tile compatibility. Web Mercator stretches lengths by about 1 ÷ cos(latitude), so at Cambridge Bay (69°N) lengths look about 2.8 times larger than they are, and areas about 7.8 times. Areas in this project are not measured on the map: they come from Earth Engine's pixel-area image, which gives ground area on the WGS 84 ellipsoid, and the study-area dimensions use geodesic distance. For desktop GIS work with the exported GeoJSON (EPSG:4326), reproject to EPSG:3413 (NSIDC Sea Ice Polar Stereographic North) or EPSG:3978 (Canada Atlas Lambert) before measuring.
Reproduce this
- Create and activate a Python 3.11 or newer virtual environment.
- Install with
python -m pip install -e . -
Run
python -m arctic_passage_atlas demofor the credential-free demonstration. -
Copy
.env.exampleto.env, register and authenticate an Earth Engine project, and add a Global Fishing Watch token. - Run
python -m arctic_passage_atlas live. - Run
python -m arctic_passage_atlas verify. -
Preview with
python -m arctic_passage_atlas serve.
Provenance
The generated manifest records the mode, configuration hash, Git
commit when available, dataset identifiers, file sizes and SHA-256
checksums. Dataset aliases ending in :latest are
mutable; reruns can therefore differ as providers update their
records.
How to cite this work
Chowdhary, M. (2026). Arctic Passage Atlas: an open-source method for observing Arctic surface, sea-ice and vessel-event change from public data [Software]. Website: https://arcticpassageatlas.com.
Known limitations
- The AOI was selected for a portfolio-scale method and is not representative of the whole Arctic.
- There is no field validation or operational ice-chart comparison at local resolution.
- Thresholds are sensitivity-tested but not statistically calibrated.
- Provider coverage and corrections can change after publication.
- The static site is an explanatory artifact, not an operational monitoring service.