A harmonised dataset for Earth system foundation models

2026-07-03

AI Summary

This work addresses a critical limitation in current Earth system foundation models—the absence of a unified training dataset that integrates heterogeneous, multi-source data spanning climate, land, ocean, cryosphere, infrastructure, hazards, and socioeconomic domains. To overcome this, the study constructs WorldTensor, a standardized dataset that harmonizes hundreds of environmental and socioeconomic variables into a common 0.25° spatial grid and annual temporal framework. Through techniques including regridding, rasterization of vector and point data, and temporal alignment, the resulting dataset adheres to Climate and Forecast (CF) metadata conventions and is distributed in NetCDF format. WorldTensor enables deep multimodal integration across natural and human systems, establishing a reproducible, high-quality benchmark for training and evaluating planetary-scale coupled models.

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.

Problem

• Earth system foundation models
• harmonised dataset
• multimodal data integration
• socioeconomic data
• global training resource

Innovation

• harmonised dataset
• Earth system foundation models
• multimodal integration
• spatiotemporal standardisation
• machine learning for Earth systems