OlmoEarth Pretraining Dataset: Multimodal Satellite Image Time Series

Provider
Allen Institute for AI (Ai2), OlmoEarth team
License
ODbL-1.0 (source)
Size
272 files
Format
CSV (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/allenai/olmoearth_pretrain_dataset)
Access
Free and public. No account with the data provider is needed.
Category
Earth Observation & Geospatial
Source
https://github.com/allenai/olmoearth_pretrain/blob/main/docs/Pretraining-Dataset.md

About OlmoEarth Pretraining Dataset: Multimodal Satellite Image Time Series

Pretraining corpus for the OlmoEarth remote sensing foundation models from Ai2. It holds 285,288 samples, each a 2560 x 2560 m UTM grid cell with a 360-day time range, sampled from about 120 OpenStreetMap feature categories. Each sample carries up to three satellite image time series as monthly (30-day) mosaics, Sentinel-2 L2A (12 bands), Sentinel-1 IW GRD (VV, VH) and Landsat 8/9 OLI-TIRS Collection 2 Level-1 (11 bands), plus auxiliary rasters: USDA Cropland Data Layer, ERA5 monthly reanalysis, rasterised OpenStreetMap, SRTM elevation, ESA WorldCereal 2021, ESA WorldCover 2021 and WRI/Meta canopy height maps, resampled to power-of-two multiples of 10 m. The listing (measured 2026-11-14) shows one folder per modality (10_sentinel2_l2a_monthly, 10_landsat_monthly, 10_era5_10_monthly and so on) with a per-modality CSV index, lon_lats.json and an h5py_data folder; the repository uses about 16.1 TB. Read the rasters with rasterio and the h5py_data files with h5py, following the olmoearth_pretrain GitHub documentation. Released by Ai2 under ODbL, with each source's attribution listed on the card.

How to load OlmoEarth Pretraining Dataset: Multimodal Satellite Image Time Series in Python

import pandas as pd
df = pd.read_csv("hf://datasets/allenai/olmoearth_pretrain_dataset/10_cdl.csv")
df.head()

Data structure

File types in sampled files: CSV

Files

Example file: 10_cdl.csv

Analyze this dataset without downloading it

On Scigantic, OlmoEarth Pretraining Dataset: Multimodal Satellite Image Time Series opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Earth Observation & Geospatial datasets