Google Satellite Embedding V1 - small areas (2017-2024)

This dataset provides the Google Satellite Embedding subset extracted through an internal pipeline for the entire UK. The underlying source data is the annual AlphaEarth Foundations Satellite Embedding dataset, which is produced by Google DeepMind.

IMPORTANT - This is a beta version of the data product. We have released it as a preview to allow users exploration and experimentation while we perform final quality assurance and additional checks. If you encounter any issue or have questions, please get in touch with us at imago@liverpool.ac.uk We will shortly be releasing a more stable version. Please check https://imago.ac.uk for news and updates.

Data and Resources

Additional Info

Field Value
Author Imago Team
Maintainer Imago RSE Team
Version 1.0.0
Last Updated June 26, 2026, 15:08 (UTC)
Created April 10, 2026, 12:47 (UTC)
Size
constraints None
content This dataset provides Google Satellite Embedding values (64-dimensional, L2-normalised vectors, with values ranging between −1 and +1) with 64 columns (from band1_mean to band64_mean) for the entire UK. The source data are Google Satellite Embedding V1, processed at 10-metre spatial resolution, for each year from 2017 to 2024 (eight total). Data extraction was performed using a geospatial pipeline to export tiled embeddings. This data has been aggregated to small area geographies across the United Kingdom: Lower layer Super Output Areas (LSOAs) for England and Wales, Data Zones for Scotland, and Small Areas for Northern Ireland.
crs EPSG:27700, OSGB36/British National Grid
data_quality Embedding values are derived from the Google Satellite Embedding dataset V1, which provides global coverage; this product focuses on the UK. The original data was extracted as GeoTIFFs and reprojected to EPSG:27700. The extracted GeoTIFFs were rescaled by a factor of 32767 to optimise storage. Further details on the methodology will be provided in the upcoming technical report. Users may consult the Google Embedding introduction (https://medium.com/google-earth/ai-powered-pixels-introducing-googles-satellite-embedding-dataset-31744c1f4650) for more information on applications and limitations of the source imagery. **IMPORTANT** - This version is released as beta. This means that, although we have applied our quality standard in its production, we are still assessing its overall quality and identifying minor issues that need resolution before a more stable version can be released. As soon as we are ready, we will issue a new release. Please check https://imago.ac.uk for news and updates.
data_source
file_id
lineage The underlying methods and source information used to construct the dataset are documented in the upcoming technical report. The Google Satellite Embedding dataset V1 was accessed via Google Earth Engine API (https://developers.google.com/earth-engine/datasets/catalog/GOOGLE_SATELLITE_EMBEDDING_V1_ANNUAL), covering a period from 1 January 2017 to 31 December 2024. Data was processed at 10-metre spatial resolution for British National Grid tiles, reprojected from multiple UTM zones to EPSG:27700. All processing was conducted by the Imago Team.
source Imago: Data Service for Imagery
spatial_coverage United Kingdom
spatial_resolution Small area (LSOA / Data Zone / Small Area)
temporal_coverage 01-01-2017:31-12-2024
temporal_resolution annual