OpenContrails: Human-Labeled Contrail Masks on GOES-16 ABI Imagery (Google Research)

Provider
Google Research (Ng et al.)
License
CC-BY-4.0 (source)
Stored at
Google Cloud Storage, gs://goes_contrails_dataset/20250305 (us)
Access
Free and public. No account with the data provider is needed.
Category
Climate & Weather
Source
https://arxiv.org/abs/2304.02122

About this dataset

OpenContrails is a benchmark for detecting contrails, the line-shaped ice clouds that aircraft leave behind, in geostationary satellite imagery. Human labelers drew per-pixel contrail masks on GOES-16 Advanced Baseline Imager (ABI) infrared scenes, and each labeled frame comes with a sequence of neighboring ABI frames so that models can use temporal context, as the detection model in the paper does. Automated contrail detection is needed to build and test contrail avoidance, a low-cost way to reduce the climate warming caused by aviation. Measured 2026-11-17: the 20250305 release (an update of the 20230419 release, which sits beside it in the bucket) holds tfrecords/ with training shards (train.tfrecords-000NN-of-00100, about 2.5 to 3 GB each), a 4.5 MB dataset instruction PDF, a README pointing to an example Colab for parsing the TFRecords, and the GOES-16 acknowledgement (GOES data carry no use restrictions). Load the records with TensorFlow tf.data. Labels and model outputs are released under CC BY 4.0; cite Ng et al., Contrail Detection on GOES-16 ABI With the OpenContrails Dataset, IEEE Transactions on Geoscience and Remote Sensing 62 (2024).

How to access this dataset in Python

import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('goes_contrails_dataset/20250305/', detail=False)[:10]

Lists the files with anonymous access. Runs on any machine, with no cloud account.

Data structure

Files

Example file: 20250305/tfrecords/train.tfrecords-00000-of-00100

Path pattern: {YYYYMMDD}/{seg1}/{filename}

Analyze this dataset without downloading it

On Scigantic, OpenContrails: Human-Labeled Contrail Masks on GOES-16 ABI Imagery (Google Research) 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 Climate & Weather datasets