The increase of volume, complexity and demand for geospatial data, such as climate models and satellite data, requires the adaptation of storage, reproducibility and traceability of the high-volume, large-scale data. Traditional formats such as NetCDF or TIFF, although reliable and efficient, do not come with the flexibility and cloud-native adaptability of Zarr files.
This project presents a library, PCoRaS, designed to create multidimensional data cubes in Zarr format from different sources of geospatial and raster data. The library takes one or multiple raster-based datasets as input and generates an N-dimensional data cube according to user-defined variables, dimensions and shared metadata attributes. By allowing users to define which variables should be extracted and how dimensions should be structured, the library supports the transformation of heterogeneous raster inputs into organized analysis-ready data cubes. A particular feature of the library is its support for dual temporal dimensions, enabling the aggregation of variable data over specific temporal frames, such as months, while preserving a broader temporal dimension representing one or multiple years. This functionality is especially relevant for climate and environmental datasets where seasonal, monthly, and long-term temporal structures must coexist within the same data cube.
This project works on top of Raster2STAC. The package extracts the metadata from raster data, generating json files compatible with STAC, creates STAC collections with items and assets, and automatically uploads the data to a S3 bucket. The combination of Zarr data cube generation, STAC metadata and S3 publication contributes to a FAIR and cloud-native geospatial infrastructure, allowing climate data and satellite data to comply with the requirements of scalable access, interoperability and reproducibility. Overall, this work contributes a reusable and scalable workflow that supports more interoperable, reproducible and cloud-ready geospatial data processing
