Global Surface Water: Global Water-Occurrence Data for GIS, Earth Observation, and Spatial Analysis
Global Surface Water is a global Earth-observation dataset from the European Commission Joint Research Centre (JRC) that describes where surface water has appeared and how its presence has changed over time. The current download release extends the long-term record through 2024, while the widely used JRC/GSW1_4 Earth Engine asset covers 1984–2021. At 30-metre pixel size, the dataset provides raster layers for occurrence, change, seasonality, recurrence, transitions, maximum water extent, monthly history, and yearly history. It is useful for water-body screening, environmental change analysis, floodplain and wetland studies, regional statistics, map production, and combining surface-water evidence with other geographic layers. It is not a live flood-monitoring service, a hydrologic model, a legal wetland inventory, or a substitute for locally authoritative water data.
Global Surface Water turns a long archive of Landsat observations into a consistent geographic description of surface-water presence. Rather than representing only named lakes or rivers, it records whether individual 30-metre pixels were classified as water across repeated observations. This makes the dataset especially valuable when the question is temporal: where water is persistent, where it is seasonal, where it has appeared or disappeared, and how its observed extent differs between time periods. The dataset is global in geographic scope and is distributed through the Global Surface Water Explorer, downloadable raster files, Google Earth Engine, and web map services. The current JRC download page identifies a 1984–2024 release updated on August 26, 2026, but users should not assume that every layer covers the entire period. Some newer historical layers must be merged with earlier assets, and the project documents known differences between Landsat processing collections. (global-surface-water.appspot.com)
How EDK fits
For: GIS analysts, Earth-observation researchers, environmental planners, hydrologists, conservation organizations, map developers, public agencies, and data engineers building spatial workflows around surface-water extent and change.
Global Surface Water is a focused environmental layer rather than a complete representation of the physical world. Its value in a broader world model is that it supplies time-aware evidence about one class of physical feature: observed surface water. When combined with places, addresses, roads, buildings, infrastructure, administrative areas, terrain, imagery, and public records, the layer can help answer geographic questions such as which roads or buildings lie near persistent water, which neighborhoods intersect changing water extents, or how water presence differs across regions. The important distinction is that Global Surface Water contributes raster observations and derived summaries; it does not by itself provide named feature identity, ownership, navigable waterways, flood forecasts, or a complete hydrologic network.
- •Use Global Surface Water as an environmental evidence layer when searching or filtering a broader physical-world representation for places, infrastructure, roads, or buildings near persistent, seasonal, or recently changing surface water.
- •Compute geographic summaries by intersecting water-occurrence or maximum-extent rasters with regions, watersheds, road corridors, parcels, facilities, or other physical features, while preserving the dataset's observation period and classification limitations.
- •Compare water-related context around a location over time by combining the dataset's occurrence, recurrence, seasonality, and transition measures with other location data such as terrain, land use, public assets, or imagery.
Dataset profile
Formats
GeoTIFF raster files in 10° by 10° download tiles
Google Earth Engine images and image collections
Web Map Tile Service and web-tiled map layers for visualization
ISO 19139 metadata files
QGIS symbology files
Access methods
Global Surface Water Explorer for interactive viewing
Official JRC data-download page and tile downloads
Official Python 3 download script for full datasets
JRC FTP access for historical data through 2021
Google Earth Engine catalog and project assets
WMTS and web-tiled services for desktop GIS and web maps
Quality and limits
Thirty-metre pixels are not sufficient for every small or narrow water feature.
Optical Landsat observations can be affected by clouds, shadows, snow, vegetation, sensor availability, and observation frequency.
Observed water presence is not equivalent to legal wetland status, flood hazard, groundwater, channel connectivity, or navigability.
The 2022–2024 extension uses Landsat Collection 2 after earlier Collection 1 processing, and the JRC documents residual co-registration offsets.
The current release does not provide one uniformly complete timeline for every derived layer without merging assets.
The project identifies a Monthly Recurrence issue in v1.4 and v1.5 that may lower long-term monthly values where observations are sparse; the stated correction is foreseen rather than completed.
Areas where water has never been detected may be masked, so absence of a displayed value requires careful interpretation.
Analysis workflows
Select the product version and observation period, then document the exact asset, download date, and processing collection.
Choose a variable that matches the question: occurrence for frequency, seasonality for months present, recurrence for year-to-year return, maximum extent for any observed water, and transition or change for long-term comparison.
Align projections, raster grids, masks, and no-data handling before calculating zonal statistics or comparing periods.
Use raster-based thresholding and distance analysis for large areas; vectorize only when polygon features are required for a specific downstream operation.
Intersect water layers with administrative areas, watersheds, roads, buildings, facilities, land cover, terrain, or other location data while retaining the source period and threshold.
Validate local or high-stakes results against recent imagery, local hydrography, field observations, gauges, or other authoritative sources.
For current flood response or engineering analysis, supplement Global Surface Water with near-real-time observations and hydrologic or hydraulic data rather than using it alone.
Value inside EDK
What Global Surface Water contains
Global Surface Water was developed by the JRC from Landsat imagery. The original mapping work classified individual pixels as water or non-water and assembled the observations into a long-term record. The JRC's Earth Engine documentation describes the 1984–2021 version as using 4,716,475 scenes from Landsat 5, 7, and 8 acquired between March 16, 1984, and December 31, 2021. The product is designed to describe the spatial and temporal distribution of surface water, not merely to draw outlines around permanent lakes. (developers.google.com)
The principal map-layer products include occurrence, which expresses how frequently water was present; seasonality, which records the number of months water was present; recurrence, which measures how often water returned from year to year; maximum extent, which identifies locations where water was detected at least once; and transition classes describing changes such as permanent-to-seasonal, seasonal-to-permanent, new permanent, lost permanent, and related categories. Change layers compare water occurrence between defined time epochs. The older Earth Engine documentation identifies seven 30-metre bands for the v1.4 map-layer product. (developers.google.com)
- •Spatial resolution: 30 metres for the primary mapped layers.
- •Geographic coverage: global, between approximately 90° south and 90° north.
- •Temporal coverage: current download release through 2024; legacy Earth Engine v1.4 asset through 2021.
- •Typical variables: occurrence, change, seasonality, recurrence, transition, maximum extent, monthly history, and yearly history.
Geography, temporal coverage, and update cadence
Global Surface Water is intended for worldwide analysis, including areas without consistently maintained national water-body inventories. Its geographic scope makes it useful for comparative studies across countries and regions, but global coverage does not mean uniform observation quality everywhere. Cloud cover, snow, shadows, vegetation, narrow channels, shoreline complexity, sensor gaps, and the availability of usable Landsat observations can affect individual pixels and places.
The dataset is versioned rather than continuously updated in real time. The JRC data-access page currently identifies a 1984–2024 release and gives a last-updated date of August 26, 2026. It also explains that the extension beyond 2021 moved from Landsat Collection 1 to Collection 2. Because the two collections have different geolocation characteristics, the project warns that combining them can introduce residual co-registration offsets, usually below one pixel but sometimes reaching or exceeding one 30-metre pixel in particular Landsat path/rows. A complete backprocessing for full archive consistency is described as planned, not as already completed. (global-surface-water.appspot.com)
- •Use the same release and processing generation across a comparative study whenever possible.
- •Treat apparent shoreline movement near the 30-metre pixel scale cautiously.
- •Document whether a result uses the 1984–2021 legacy asset, the 1984–2024 release, or merged subperiod assets.
- •Do not interpret a missing detection as proof that no water existed; it may reflect observation, classification, or masking conditions.
Formats and access methods
The official data-access page provides several ways to work with Global Surface Water. Individual downloadable files are available as GeoTIFF tiles covering 10° by 10°. The page also describes a Python 3 script for downloading the full global v1.4 or v1.5 datasets, along with a file listing for v1.5. Earlier data remain available through the JRC FTP path. Downloadable files include embedded color maps for desktop GIS display, but the project states that the files do not contain all metadata, so users should obtain the associated ISO 19139 metadata files and consult the Data Users Guide. (global-surface-water.appspot.com)
Google Earth Engine provides a convenient analysis environment. The established v1.4 asset is available as JRC/GSW1_4/GlobalSurfaceWater, with related assets for yearly history, monthly history, monthly recurrence, and metadata. The newer 2024 assets use project-specific asset identifiers and, in some cases, require merging image collections across periods. Earth Engine is useful for clipping, masking, reducing, compositing, and exporting results without first downloading the entire global archive. (global-surface-water.appspot.com)
- •Raster download: GeoTIFF tiles, generally organized as 10° by 10° files.
- •Cloud analysis: Google Earth Engine image and image-collection assets.
- •Visualization: Global Surface Water Explorer and web map or tiled services.
- •Desktop GIS: QGIS, ArcGIS, and other software that supports GeoTIFF and web map services.
- •Supporting material: Data Users Guide, ISO 19139 metadata, and QGIS symbology files.
Suitable GIS and spatial-analysis workflows
A straightforward GIS workflow begins by choosing the question and the appropriate layer. Use occurrence to identify persistent or frequently observed water, seasonality to distinguish water present for many months from short-lived water, recurrence to study interannual return, and transition or change layers to screen for areas of apparent gain or loss. For an administrative or watershed summary, reproject only when appropriate, align the raster grid, clip to the area of interest, and calculate zonal statistics with the dataset's valid-data mask handled explicitly.
For proximity analysis, convert a carefully selected water mask or thresholded occurrence layer into a vector representation only when necessary. Raster-based distance calculations are often more efficient and avoid introducing extra polygon complexity. Intersections with roads, buildings, facilities, or parcels should preserve the source date range and threshold used. A statement such as 'near persistent water' needs a defined occurrence threshold, distance, pixel treatment, and analysis period.
- •Regional water accounting: calculate zonal summaries of occurrence, maximum extent, or seasonality by watershed, district, or planning area.
- •Infrastructure screening: measure the proximity or overlap of roads, buildings, utilities, or public assets with persistent or changing water observations.
- •Environmental change: compare transition and change layers with land cover, terrain, protected areas, or imagery.
- •Cartography: use the Explorer, WMTS, or QGIS symbology for visual communication, while retaining raw raster values for analysis.
- •Quality control: inspect cloud-prone, snow-prone, vegetated, narrow-channel, and shoreline areas before drawing local conclusions.
License, strengths, and tradeoffs
The JRC states that Global Surface Water data are produced under the Copernicus Programme and provided free of charge without restriction of use, subject to the full Copernicus terms. Publications, models, and data products must include proper acknowledgement. For a published map layer, the project requests the attribution text 'Source: EC JRC/Google' and asks users to cite the Pekel et al. 2016 Nature paper. (global-surface-water.appspot.com)
- •Strength: global geographic coverage supports cross-border and regional comparison.
- •Strength: the long Landsat record supports historical surface-water occurrence and change analysis rather than only a single-date map.
- •Strength: multiple derived layers separate persistence, seasonality, recurrence, maximum extent, and transition behavior.
- •Strength: open downloads, Earth Engine access, and web services support both local GIS work and large-area computation.
- •Tradeoff: 30-metre pixels can miss narrow channels, small ponds, fragmented shorelines, and subpixel water.
- •Tradeoff: it is not real time; the current published extension reaches 2024, so it should not be used alone for current flood response.
- •Tradeoff: Collection 1 and Collection 2 differences can create co-registration or consistency issues when extending a historical series.
- •Tradeoff: classifications represent observed surface-water conditions and should not be treated automatically as legal wetlands, flood hazards, groundwater, stream networks, or navigable waterways.
Questions teams ask
What years does Global Surface Water cover?
The JRC data-access page currently describes a new 1984–2024 release. The widely used JRC/GSW1_4 Earth Engine map-layer asset covers 1984–2021. Coverage is layer-specific: some 2022–2024 history assets must be merged with earlier assets to reconstruct a full timeline. Always record the exact version and asset used. (global-surface-water.appspot.com)
Is Global Surface Water free to use?
Yes. The JRC states that the data are produced under the Copernicus Programme and provided free of charge without restriction of use, subject to the applicable Copernicus terms. Users must provide proper acknowledgement, cite the dataset and associated paper where appropriate, and use 'Source: EC JRC/Google' for a published map layer. (global-surface-water.appspot.com)
Can Global Surface Water be used for flood mapping?
It can support historical water-extent screening and floodplain or change analysis, but it is not a real-time flood-monitoring service or a flood forecast. For emergency or engineering decisions, combine it with current satellite observations, precipitation, gauges, terrain, hydraulic or hydrologic models, and locally authoritative information.
How can I use Global Surface Water in QGIS or ArcGIS?
For analysis, download the GeoTIFF tiles or export raster results from Earth Engine. The official page also provides symbology files and web map services for visualization. Web map services are cached RGB images and are intended for mapping rather than quantitative analysis, so use the underlying raster values for measurements and spatial statistics. (global-surface-water.appspot.com)
Does the dataset show rivers, lakes, and wetlands as named vector features?
No. Global Surface Water is primarily a raster observation and summary product. It indicates where surface water was detected and how frequently or seasonally it appeared. It does not by itself provide authoritative names, legal boundaries, ownership, navigability, channel topology, groundwater conditions, or a complete vector hydrography network.
Keep exploring
Sources
Global Surface Water - Data Access
Official JRC access page covering the 1984–2024 release, licensing, downloads, formats, Earth Engine assets, web services, update notes, and limitations.
Open sourceJRC Global Surface Water Mapping Layers, v1.4
Official Google Earth Engine catalog documentation for the 1984–2021 asset, bands, resolution, data range, citation, and access method.
Open sourceHigh-resolution mapping of global surface water and its long-term changes
The peer-reviewed Nature publication describing the original Global Surface Water mapping method and long-term global analysis.
Open sourceGlobal Surface Water Data Users Guide
Official user documentation referenced by the JRC for dataset descriptions and usage details.
Open source