This repository holds code relevant to Sweeney et al., 2023.
Data: https://zenodo.org/records/8286589
Notebooks contain the code to train CNN's and make predictions based on observational data from 1980-2022. This code is not created to run out of the box, but instead needs data to be downloaded and paths to data must be respecified in the code provided.
Training of CNNs used in Sweeney et al., 2023 was done on 1 Tesla GPU from Casper provided by NCAR’s Computational and Information Systems Laboratory. With these specifications the Arctic notebook takes ~20 mins to run, and the global notebook takes ~45 mins.