GeoCryoAI

Bradley A. Gay, PhD | NASA Postdoctoral Program Fellow | JPL, California Institute of Technology

GeoCryoAI is a hybridized ensemble learning framework composed of stacked convolutional layers and memory-encoded recurrent neural networks. This multimodal deep learning architecture simultaneously ingests and analyzes in situ measurements, airborne remote sensing observations, and process-based modeling outputs exhibiting disparate spatiotemporal sampling and data densities. If these resources prove helpful and are incorporated, repurposed, and/or modules are extracted and reused, please cite this repository, the companion dataset and source code in the ORNL DAAC repository, and the JGR-MLC manuscript.

Gay, B. A., Pastick, N. J., Watts, J. D., et al., 2025. Decoding the Spatiotemporal Complexities of the Permafrost Carbon Feedback with Multimodal Ensemble Learning. Journal of Geophysical Research, Machine Learning and Computation. Under Review.

Gay, B.A., et al. 2025. GeoCryoAI | Decoding the Spatiotemporal Complexities of the Permafrost Carbon Feedback with Multimodal Ensemble Learning. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/2371

Gay, B., Pastick, N., Watts, J., Armstrong, A., Miner, K., & Miller, C. (2024). geocryoai (Version 1.0.0) [Computer software]. https://www.github.com/bradleygay/geocryoai

Relevant Manuscripts and Datasets

Gay, B. A., Pastick, N. J., Watts, J. D., et al., 2025. Decoding the Spatiotemporal Complexities of the Permafrost Carbon Feedback with Multimodal Ensemble Learning. Journal of Geophysical Research, Machine Learning and Computation. In Production.

Gay, B.A., N.J. Pastick, J.D. Watts, A.H. Armstrong, K. Miner, and C.E. Miller. 2025. GeoCryoAI Permafrost, Thaw Depth and Carbon Flux in Alaska, 1969-2022. Preprint. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/2371

Gay, B. A., Pastick, N. J., Watts, J. D., et al., 2024. Forecasting Permafrost Carbon Dynamics in Alaska with Earth Observation Data and Artificial Intelligence, ESS Open Archive. https://essopenarchive.org/users/524229/articles/1225858-forecasting-permafrost-carbon-dynamics-in-alaska-with-earth-observation-data-and-artificial-intelligence

Gay, B. A., Zfle, A. E., Armstrong, A. H., et al. Investigating Permafrost Carbon Dynamics in Alaska with Artificial Intelligence, December 26, 2023. ESS Open Archive. https://doi.org/10.22541/essoar.170355056.64772303/v1

Gay, B. A., Zfle, A. E., Armstrong, A. H., et al. Investigating High-Latitude Permafrost Carbon Dynamics with Artificial Intelligence and Earth System Data Assimilation, December 26, 2023. ESS Open Archive. https://doi.org/10.22541/essoar.170355053.35677457/v1

Gay, B.A., Pastick, N.J., Zfle, A.E., Armstrong, A.H., Miner, K.R., Qu, J.J., 2023. Investigating permafrost carbon dynamics in Alaska with artificial intelligence. Environmental Research Letters 18. https://doi.org/10.1088/1748-9326/ad0607

Gay, B. A., (2023). Investigating High-Latitude Permafrost Carbon Dynamics with Artificial Intelligence and Earth System Data Assimilation. (Order No. 30488695, George Mason University). ProQuest Dissertations and Theses, 281. Retrieved from https://www.proquest.com/dissertations-theses/investigating-high-latitude-permafrost-carbon/docview/2826111475/se-2

Large Data Files

This repository contains large data files that have been chunked for storage. To reconstruct the original files:

1. Clone the repository

2. Install required packages:

pip install h5py pandas numpy

3. Run the reconstruction script:

python chunk_reassembly.py

The script will reconstruct:

• ensembletensor.h5 (from h5chunks)
• finalfcfch4altmonthly1kmds.parquet (from parquet_chunks)