A Bayesian Deep Learning Approach to Near-Term Climate Prediction
Since model bias and associated initialization shock are serious shortcomings that reduce prediction skills in state-of-the-art decadal climate prediction efforts, we pursue a complementary machine-learning-based approach to climate prediction. The example problem setting we consider consists of pre...
Published in: | Journal of Advances in Modeling Earth Systems |
---|---|
Main Authors: | , , , , , |
Language: | unknown |
Published: |
2023
|
Subjects: | |
Online Access: | http://www.osti.gov/servlets/purl/1906948 https://www.osti.gov/biblio/1906948 https://doi.org/10.1029/2022ms003058 |
id |
ftosti:oai:osti.gov:1906948 |
---|---|
record_format |
openpolar |
spelling |
ftosti:oai:osti.gov:1906948 2023-07-30T04:05:28+02:00 A Bayesian Deep Learning Approach to Near-Term Climate Prediction Luo, Xihaier Nadiga, Balasubramanya T. Park, Ji Hwan Ren, Yihui Xu, Wei Yoo, Shinjae 2023-01-05 application/pdf http://www.osti.gov/servlets/purl/1906948 https://www.osti.gov/biblio/1906948 https://doi.org/10.1029/2022ms003058 unknown http://www.osti.gov/servlets/purl/1906948 https://www.osti.gov/biblio/1906948 https://doi.org/10.1029/2022ms003058 doi:10.1029/2022ms003058 97 MATHEMATICS AND COMPUTING 54 ENVIRONMENTAL SCIENCES 2023 ftosti https://doi.org/10.1029/2022ms003058 2023-07-11T10:17:15Z Since model bias and associated initialization shock are serious shortcomings that reduce prediction skills in state-of-the-art decadal climate prediction efforts, we pursue a complementary machine-learning-based approach to climate prediction. The example problem setting we consider consists of predicting natural variability of the North Atlantic sea surface temperature on the interannual timescale in the pre-industrial control simulation of the Community Earth System Model. While previous works have considered the use of recurrent networks such as convolutional LSTMs and reservoir computing networks in this and other similar problem settings, we currently focus on the use of feedforward convolutional networks. In particular, we find that a feedforward convolutional network with a Densenet architecture is able to outperform a convolutional LSTM in terms of predictive skill. Next, we go on to consider a probabilistic formulation of the same network based on Stein variational gradient descent and find that in addition to providing useful measures of predictive uncertainty, the probabilistic (Bayesian) version improves on its deterministic counterpart in terms of predictive skill. Finally, we characterize the reliability of the ensemble of machine learning models obtained in the probabilistic setting by using analysis tools developed in the context of ensemble numerical weather prediction. Other/Unknown Material North Atlantic SciTec Connect (Office of Scientific and Technical Information - OSTI, U.S. Department of Energy) Journal of Advances in Modeling Earth Systems 14 10 |
institution |
Open Polar |
collection |
SciTec Connect (Office of Scientific and Technical Information - OSTI, U.S. Department of Energy) |
op_collection_id |
ftosti |
language |
unknown |
topic |
97 MATHEMATICS AND COMPUTING 54 ENVIRONMENTAL SCIENCES |
spellingShingle |
97 MATHEMATICS AND COMPUTING 54 ENVIRONMENTAL SCIENCES Luo, Xihaier Nadiga, Balasubramanya T. Park, Ji Hwan Ren, Yihui Xu, Wei Yoo, Shinjae A Bayesian Deep Learning Approach to Near-Term Climate Prediction |
topic_facet |
97 MATHEMATICS AND COMPUTING 54 ENVIRONMENTAL SCIENCES |
description |
Since model bias and associated initialization shock are serious shortcomings that reduce prediction skills in state-of-the-art decadal climate prediction efforts, we pursue a complementary machine-learning-based approach to climate prediction. The example problem setting we consider consists of predicting natural variability of the North Atlantic sea surface temperature on the interannual timescale in the pre-industrial control simulation of the Community Earth System Model. While previous works have considered the use of recurrent networks such as convolutional LSTMs and reservoir computing networks in this and other similar problem settings, we currently focus on the use of feedforward convolutional networks. In particular, we find that a feedforward convolutional network with a Densenet architecture is able to outperform a convolutional LSTM in terms of predictive skill. Next, we go on to consider a probabilistic formulation of the same network based on Stein variational gradient descent and find that in addition to providing useful measures of predictive uncertainty, the probabilistic (Bayesian) version improves on its deterministic counterpart in terms of predictive skill. Finally, we characterize the reliability of the ensemble of machine learning models obtained in the probabilistic setting by using analysis tools developed in the context of ensemble numerical weather prediction. |
author |
Luo, Xihaier Nadiga, Balasubramanya T. Park, Ji Hwan Ren, Yihui Xu, Wei Yoo, Shinjae |
author_facet |
Luo, Xihaier Nadiga, Balasubramanya T. Park, Ji Hwan Ren, Yihui Xu, Wei Yoo, Shinjae |
author_sort |
Luo, Xihaier |
title |
A Bayesian Deep Learning Approach to Near-Term Climate Prediction |
title_short |
A Bayesian Deep Learning Approach to Near-Term Climate Prediction |
title_full |
A Bayesian Deep Learning Approach to Near-Term Climate Prediction |
title_fullStr |
A Bayesian Deep Learning Approach to Near-Term Climate Prediction |
title_full_unstemmed |
A Bayesian Deep Learning Approach to Near-Term Climate Prediction |
title_sort |
bayesian deep learning approach to near-term climate prediction |
publishDate |
2023 |
url |
http://www.osti.gov/servlets/purl/1906948 https://www.osti.gov/biblio/1906948 https://doi.org/10.1029/2022ms003058 |
genre |
North Atlantic |
genre_facet |
North Atlantic |
op_relation |
http://www.osti.gov/servlets/purl/1906948 https://www.osti.gov/biblio/1906948 https://doi.org/10.1029/2022ms003058 doi:10.1029/2022ms003058 |
op_doi |
https://doi.org/10.1029/2022ms003058 |
container_title |
Journal of Advances in Modeling Earth Systems |
container_volume |
14 |
container_issue |
10 |
_version_ |
1772817402972602368 |