Historical climate data, unlocked

Recovering centuries of weather observations with AI.

Everything we know about climate change and variability depends on observations. oldWeather.ai uses modern AI to transform handwritten weather logs, ship records, and archival documents into accessible digital climate data.

See an example Visit the rainfall project
An illustration of AIs repairing a weather station

Why this matters

Weather observations are one of our most valuable climate records.

People have been observing the weather for centuries, and the billions of observations they have made are an invaluable scientific resource.

Hidden data

Hundreds of millions of historical observations are still unavailable to science, locked away in national archives, personal journals, and local records.

Better climate insight

Unlocking these records can reveal historical climate patterns, improve long-term trend analysis, and strengthen future climate predictions.

AI at scale

oldWeather.ai aims to digitize and analyze handwritten logs cheaply and rapidly with machine learning, preserving the past while making it usable for science.

Featured example

50 million daily rainfall observations, recovered from paper archives.

See an example oldWeather.ai project: Auto Daily Rainfall — 50 million observations of daily rainfall from England and Wales, transformed from paper archive into an accessible digital record by AI vision-language models.