A harmonised dataset for Earth system foundation models
Carlos Rodriguez-Pardo, Massimo Tavoni
Why It Matters
What makes this one worth your time
This dataset enables the development of more comprehensive Earth system models by incorporating diverse data types, which can improve understanding and prediction of environmental changes and their socioeconomic impacts.
WorldTensor provides a unified dataset for training Earth system models with both environmental and socioeconomic data.
Summary
The paper introduces WorldTensor, a harmonised global dataset that integrates a wide range of environmental and socioeconomic variables into a standardised spatial and temporal grid, facilitating the development of multimodal Earth system foundation models.
Key contributions
- Creation of a harmonised dataset combining environmental and socioeconomic data.
- Standardisation of data into a 0.25-degree spatial grid and annual temporal framework.
- Distribution of data in NetCDF format with standardised metadata.
Notable insights
- The integration of heterogeneous data sources into a common grid and temporal framework is a significant methodological challenge addressed by WorldTensor.
- The use of NetCDF files with standardised metadata facilitates interoperability and reproducibility in machine learning workflows.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2607.03298v1 Announce Type: cross Abstract: Foundation models for Earth systems have so far been trained primarily on physical climate and weather data, with limited representation of the human systems that both drive and respond to environmental change. The lack of a unified global training resource that combines climate, land, ocean, cryosphere, infrastructure, hazards, and socioeconomic data on a common grid hinders progress toward truly multimodal Earth system foundation models. We present WorldTensor, a harmonised global dataset that aligns hundreds of environmental and socioeconomic variables to a standardised 0.25$^\circ$ spatial grid and annual temporal framework. WorldTensor integrates reanalysis products, remote sensing, emissions inventories, land use reconstructions, hydrological observations, infrastructure and hazard datasets, and socioeconomic indicators within a single representation designed for machine learning workflows. To build the dataset, we regridded inputs across heterogeneous native resolutions and projections, rasterised point and vector datasets into spatially meaningful gridded fields, and reconciled temporal coverages ranging from daily observations to sparse multiyear socioeconomic snapshots. All outputs are distributed as NetCDF files with standardised coordinates, variable metadata, and a common CF metadata convention. WorldTensor provides a reproducible resource for training and evaluating foundation models that learn coupled dynamics across environmental and human systems at planetary scale.