Journal article 123 views 14 downloads
Data-driven decadal climate forecasting using Wasserstein time-series generative adversarial networks
,
Marco Lavazza Seranto,
Petr Hajek,
Mohammad Abedin
Annals of Operations Research
Swansea University Author: Mohammad Abedin
-
PDF | Version of Record
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
Download (1.96MB)
DOI (Published version): 10.1007/s10479-023-05722-7
Abstract
Recent trends in global climate modeling, coupled with the availability of more fine-scaledatasets, have opened up opportunities for deep learning-based climate prediction to improvethe accuracy of predictions over traditional physics-based models. For this, however, largeensembles of data are neede...
| Published in: | Annals of Operations Research |
|---|---|
| ISSN: | 0254-5330 1572-9338 |
| Published: |
Springer Science and Business Media LLC
2023
|
| Online Access: |
Check full text
|
| URI: | https://cronfa.swan.ac.uk/Record/cronfa65186 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract: |
Recent trends in global climate modeling, coupled with the availability of more fine-scaledatasets, have opened up opportunities for deep learning-based climate prediction to improvethe accuracy of predictions over traditional physics-based models. For this, however, largeensembles of data are needed. Generative models have recently proven to be a suitablesolution to this problem. For a sound generative model for time-series forecasting, it isessential that temporal dynamics are preserved in that the generated data obey the originaldata distributions over time. Existing forecasting methods aided by generative models arenot adequate for capturing such temporal relationships. Recently, generative models havebeen proposed that generate realistic time-series data by exploiting the combinations ofunsupervised and supervised learning. However, these models suffer from instable learningand mode collapse problems. To overcome these issues, here we propose Wasserstein TimeSeries Generative Adversarial Network (WTGAN), a new forecasting model that effectivelyimitates the dynamics of the original data by generating realistic synthetic time-series data. Tovalidate the proposed forecasting model, we evaluate it by backtesting the challenging decadalclimate forecasting problem. We show that the proposed forecasting model outperformsstate-of-the- art generative models. Another advantage of the proposed model is that onceWTGAN is tuned, generating time-series data is very fast, whereas standard simulators. |
|---|---|
| Keywords: |
Forecasting, climate, deep learning, time series |
| College: |
School of Management |
| Funders: |
Swansea University |