Lighthouse Global Coastline Dataset: 10 m Land/Water Tiles and Coastal-Point Ball Trees for Distance-to-Shoreline Queries (Ai2)
- Provider
- Allen Institute for AI (Ai2)
- License
- ODbL-1.0 (source)
- Stored at
- Google Cloud Storage, gs://ai2-coastlines/v1/data (us)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Earth Observation & Geospatial
- Source
- https://arxiv.org/abs/2506.18842
About this dataset
Lighthouse is Ai2's global dataset and service for computing the distance from any point on Earth to the nearest shoreline, including inland water such as rivers, lakes and bays. The land/water classification is built at 10 m resolution from a hybrid of ESA WorldCover v2 (2024) land cover and OpenStreetMap land-water polygons. The data are cut into 1 by 1 degree tiles named by their upper-left corner (for example N00E006). v1/data/resampled_h5s/ holds one HDF5 raster per tile (Ai2_WorldCover_10m_2024_v1_ _Map.h5) with the resampled land cover classes, and v1/data/ball_trees/ holds one scikit-learn ball tree per tile (.joblib) built over the extracted coastal points, which the Lighthouse algorithm combines with a precomputed spherical Voronoi tessellation of the planet's coastlines to answer nearest-coast queries in milliseconds. The full set is about 500 GB. Read the .h5 tiles with h5py and the ball trees with joblib, or mount the folder into the ghcr.io/allenai/lighthouse Docker image, which returns distance to coast in meters, the land cover class and the nearest coastal point for a latitude and longitude. Useful for coastal proximity features, maritime vessel analytics and shoreline studies. License: dataset under the Open Database License (ODbL) v1.0, code under Apache 2.0. Cite the Lighthouse paper (arXiv:2506.18842), Ai2.
How to access this dataset in Python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('ai2-coastlines/v1/data/', detail=False)[:10]Lists the files with anonymous access. Runs on any machine, with no cloud account.
Data structure
Files
Example file: v1/data/ball_trees/Ai2_WorldCover_10m_2024_v1_N00E006_Map_coastal_points_ball_tree.joblib
Analyze this dataset without downloading it
On Scigantic, Lighthouse Global Coastline Dataset: 10 m Land/Water Tiles and Coastal-Point Ball Trees for Distance-to-Shoreline Queries (Ai2) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Earth Observation & Geospatial datasets
- DEA Water Observations (Sentinel-2 Near Real-Time, Provisional, Australia)
- GIS Data by Karnataka Geographic Information System
- Multi-scale Topographic Position Image of Australia (Geoscience Australia)
- New Zealand Imagery
- Tasselled Cap Brightness
- USFS Landscape Change Monitoring System (LCMS) Change Product, 2025.11 Release
- East Africa Hydromet Observations
- Euclid Quick Release 1 (Q1)