GeoAI-Cubes is a spatial and temporal database from different sensors designed to provide pre-processed information from a variety of Earth observation systems or external data fully aligned at the pixel or tile scale.

Geoaicube is an initiative promoted by the Catalonia Space 2030 Strategy. 

The dataset is optimized to efficiently train artificial intelligence models focused on geospatial applications. All Geoaicube data is in Zarr format, georeferenced in the same WGS84 UTM31N projection and covering the same geographical extent and with the labeling of the Land Cover Map of Catalonia, major changes 2023:

• UTM X West: 427000.
• UTM X East: 459000.
• UTM Y North: 4630900.
• UTM Y South: 46113000.

Reuse of information

The use of the data is subject to a Creative Commons Attribution 4.0 International license.    More information

Contains Sentinel Copernicus data modified by the ICGC.

It is requested that all methodologies and results obtained by the different scientific groups be shared with the ICGC via email: icgc@icgc.cat.

Viewer

Full window viewer

Dataset Description

The following cartographic and additional information layers have been converted to ZARR format (more efficient for data ingestion processes into AI models):

• Geodatacube_orto2024.zarr: 25 cm resolution (32.9 GB).
• Geodatacube_S22024_2025.zarr: 10 m resolution (1.19 GB).
• GeoDataCube_access_XARRAYvsZARR_cat.ipynb: Jupyter Notebook explaining access to information in zarr format.
• Field_ROI_geodatacube.gpkg: GPKG of crop 1 for Sentinel-2 time series analysis in the notebook.
• Field2_ROI_geodatacube.gpkg: GPKG of crop 2 for Sentinel-2 time series analysis in the notebook.
• bbox_epsg32631.geojson: File in geojson format of the geographic scope.

The usability of GEOAICUBES, happens in I.A. to have good labels. The proposal is that these come together with the publication of the aforementioned geoaicubes (high and low resolution) with the labeling of the Land Cover Map of Catalonia, major changes 2023. 

Download

The data can be downloaded directly from the address https://datacloud.icgc.cat/datacloud/geodatacubes.

Due to the large volume of files available, it is recommended to download using the FTP client.

Server: ftp.icgc.cat | User: descarregues01 | Password: descarregues01

Examples of the available set of images

Sentinel 2 RGB	Orthoimage RGB	nDSM
Sentinel-2 IRC	Orthoimage RGB IRC

Land cover map	Legend
Graphic representation	Class	Name	Color	Class	Name	Color
	1	Herbaceous crops		22	"Eixample"
	2	Vegetable gardens, nurseries and forced crops		23	Lax urban areas
	3	Vineyards		24	Isolated buildings in rural areas
	4	Olive groves		25	Isolated residential areas
	5	Other woody crops		26	Green areas
	6	Crops in transformation		27	Industrial, commercial and/or service areas
	7	Dense coniferous forests		28	Sports and leisure areas
	8	Dense deciduous, broadleaf forests		29	Mining and/or landfill areas
	9	Dense sclerophyll and laurel forests		30	Zones in transformation
	10	Scrub		31	Road network
	11	Open forests of conifers		32	Urban bare soil
	12	Deciduous, broadleaf open forests		33	Airport areas
	13	Sclerophyll and laurel open forests		34	Railway network
	14	Meadows and grasslands		35	Port areas
	15	Riverside forest		36	Reservoirs
	16	Sòl nu forestal		37	Lakes and lagoons
	17	Burnt areas		38	Watercourses
	18	Rocks and swamps		39	Ponds
	19	Beaches		40	Artificial canals
	20	Wetlands		41	Sea
	21	Urban center		0	Outside the border