Clouds project provides an unsupervised machine learning algorithm for automated clustering and pattern discovery in cloud imagery, and a dataset resulting from the algorithm that is applied to 22 years (2000-2021) of Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA's Aqua and Terra satellites to contribute to the democratization of climate research.
Our project offers the AI-driven Cloud Classification Atlas AICCA, an unsupervised deep learning-based novel cloud classification dataset. About AICCA, please visit our AICCA page.
Our unsupervised deep learning algorithm, rotation-invariant cloud clustering RICC clusters 22 years of ocean satellite images from the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA's Aqua and Terra instruments—198 million patches, each roughly 100 km x 100 km (128 x 128 pixels)—into 42 AI-generated cloud classes. AICCA translates 801 TB of satellite images into 54.2 GB of class labels and cloud top and optical properties, a reduction by a factor of 15,000.
Clouds team now temporarily limits AICCA data upon users on requests. Please contatct tkurihana@uchicago.edu
You need to register Globus, a high-speed data transfer service, to download AICCA - patch and grid-cell dataset from following link:
Version.1 (Complete 2000 -- 2021)
*Nov 3rd: Clouds team is now temporarily limited AICCA data for intenal only.
*Nov 3rd: Clouds team is now temporarily limited AICCA data for intenal only. This CSV format version was supposed to a pre-stage for grid-cell dataset. But we are aware that the dataformat is easily used with Pandas and decided to publickly open this data.
*Nov 3rd: Clouds team is now temporarily limited AICCA data for intenal only. Version.1 (Complete 2000 -- 2021)
A novel cloud classification dataset produced by applying modern unsupervised deep learning methods to identify robust and meaningful clusters of cloud patterns. AICCA delivers in a compact form (tens of gigabytes of class labels, with high spatial and temporal resolution) information currently accessible only as hundreds of terabytes of multi-spectral images. AICCA enables data-driven diagnosis of patterns of cloud organization, provide insight into their evolution on timescales of hours to decade
If you use AICCA dataset for your work, please cite the paper and dataset:
@article{kurihana2022aicca,
AUTHOR = {Kurihana, Takuya and Moyer, Elisabeth J. and Foster, Ian T.},
TITLE = {AICCA: AI-Driven Cloud Classification Atlas},
JOURNAL = {Remote Sensing},
VOLUME = {14},
YEAR = {2022},
NUMBER = {22},
ARTICLE-NUMBER = {5690},
URL = {https://www.mdpi.com/2072-4292/14/22/5690},
ISSN = {2072-4292},
DOI = {10.3390/rs14225690}
}
A rotation-invariant cloud clusering (RICC) is a data-driven unsupervised learning apporoach that leverages rotaion-invariant autoencoder and hierarchical agglomerative clustering to automate the clustering of cloud patterns and textures without any assumptions concerning artitifical cloud categories.
If you find RICC is applicable and useful, please cite this paper:
@article{kurihana2021cloud,
title={Data-Driven Cloud Clustering via a Rotationally Invariant Autoencoder},
author={Kurihana, Takuya and Moyer, Elisabeth and Willett, Rebecca and Gilton, Davis and Foster, Ian},
journal={IEEE Transactions on Geoscience and Remote Sensing},
year={2022},
volume={60},
pages={1-25},
doi={10.1109/TGRS.2021.3098008}}
This work is supported by the AI for Science program of the Center for Data and Computing at the University of Chicago and
Center for Robust Decisionmaking on Climate and Energy Policy (RDCEP).
We thank Argonne Leadership Computing Facility and University of Chicago’s Research Computing Center for access to computing resources.