Skip to content

Dataset · CMU-OSPO

Daily and annual climate data for 210 US cities

CMU-OSPO/protostar-climate-train-to-2018

Historical temperature and precipitation records for US cities up to 2018 for climate risk forecasting.

What's inside

A tabular dataset of 10M to 100M rows containing daily and annual temperature and precipitation metrics across 210 US cities through December 31, 2018. Sourced from Carnegie Mellon University's open archives (KiltHub) and enriched with an annual cooling degree days column, the dataset is distributed under a CC0 license as the training split for the Protostar Challenge.

What you can do with it

  • Training time-series forecasting models for weather and climate trends
  • Modeling climate risks and estimating urban power grid loads
  • Benchmarking tabular and gradient-boosted algorithms on long historical series
Tasks
Time series forecasting
Size
10M–100M rows
Downloads this month
1.4K
Likes
4

How to load it

from datasets import load_dataset

ds = load_dataset("CMU-OSPO/protostar-climate-train-to-2018")