This publication is a byte-for-byte mirror of the JRC (European Commission Joint Research Centre) Global Surface Water Monthly History tiles for 2017 to 2024, in two editions: VER5-0 under tiles/ for 2017 to 2021, and V1.5 under v1_5/ for 2022 to 2024. This republication does not include the full 1984 to present history.
This publication is a byte-for-byte mirror of the JRC Global Surface Water Monthly History tiles for 2017 to 2024, in two editions: VER5-0 under tiles/ for 2017 to 2021, and V1.5 under v1_5/ for 2022 to 2024. The dataset describes surface water cover at 30 m resolution. We added a STAC index for this dataset. We mirror VER5-0 from the JEODPP FTP server and V1.5 from CloudFerro, both to a modern object storage solution.
This republication does not include the full 1984 to present history, but if you're interested in a full republication, contact jake@epoch.blue.
ftp/jrc-opendata/GSWE/MonthlyHistory/VER5-0/tiles/<YYYY>/<YYYY>_<MM>/.s3.waw4-1.cloudferro.com/swift/v1/global-surface-water/download2024/monthlyhistory/VER1-5/<YYYY>/<YYYY>_<MM>/.The data is produced under the Copernicus Programme and is free of charge, without restriction of use. Attribute as "Source: EC JRC/Google". The mirror does not imply endorsement by the European Commission or the JRC.
Nothing. We mirror the tiles unchanged.
The VER5-0 source files are tiled GeoTIFFs with 256 × 256 blocks and no overviews. JEODPP gives no MD5 hash, so we verify integrity by a full TIFF header re-parse of each tile.
The V1.5 source files are already Cloud-Optimized GeoTIFFs: 1024 × 1024 ZSTD blocks with 6 overview levels. CloudFerro reports a size for every file and an MD5 for most, so we verify the size of all 18,144 files and the MD5 of the 14,638 that have one.
In tiles/, y_off and x_off are the pixel offsets of the zone in the global grid, zero-padded to 10 digits. In v1_5/, lon and lat are the north-west corner of the zone, for example 0E_0N is the zone from 0° to 10° east and 0° to 10° south. Both name the same 504 zones.
V1.5 is a newer JRC edition of the same product, released in 2026 and computed from Landsat Collection 2. It covers 1984 to 2024; we mirror only 2022 to 2024, the months that tiles/ does not have.
There is a processing seam between 2021-12 and 2022-01. tiles/ (VER5-0) is computed from Landsat Collection 1, v1_5/ from Landsat Collection 2, which is a reprocessing of the Landsat archive with different radiometric and geometric calibration. Water classified in one edition is not guaranteed to be classified the same way in the other. Treat a comparison across that boundary as a comparison between two products, not as a time series.
The zone grid, resolution, CRS and band values are the same in both. V1.5 files are Cloud-Optimized, so unlike tiles/ they have overviews.
The STAC index covers tiles/ only. Read a v1_5/ tile by its URL, built from the pattern in Files, for example https://data.source.coop/epoch/jrc-global-surface-water-monthly/v1_5/2022/2022_01/monthlyhistory_0E_0N_v1_5_2022_01.tif.
Example of an anonymous read through the source.coop proxy. Find tiles for an area and a month with the STAC index:
tiles/ files have no overviews. A low-zoom view will read the full-resolution data and take up significant memory. The v1_5/ files have 6 overview levels.tiles/ only.mask_and_scale does not convert 0 to NaN. It only promotes the data to float32. 0 means no observation and is not a measurement of dry land: mask it yourself, for example da.where(da != 0).