MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting

Accurately subseasonal daily sea surface temperature prediction (SSTP) is significant for forecasting and mitigating extreme climate events related to sea surface temperature (SST). However, this scale's forecasting lies at the transitional zone between short-term forecasting and long-ter...

Full description

Saved in:
Bibliographic Details
Main Authors: Qi He, Zhenfeng Lan, Wei Song, Wenbo Zhang, Yanling Du, Wei Zhao
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10919018/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850276242769575936
author Qi He
Zhenfeng Lan
Wei Song
Wenbo Zhang
Yanling Du
Wei Zhao
author_facet Qi He
Zhenfeng Lan
Wei Song
Wenbo Zhang
Yanling Du
Wei Zhao
author_sort Qi He
collection DOAJ
description Accurately subseasonal daily sea surface temperature prediction (SSTP) is significant for forecasting and mitigating extreme climate events related to sea surface temperature (SST). However, this scale's forecasting lies at the transitional zone between short-term forecasting and long-term climate prediction, requiring simultaneous consideration of small-scale variations crucial for the former and large-scale variations fundamental to the latter. Thus, achieving precise subseasonal daily SSTPs is challenging. In this study, we introduce a novel multiscale periodic transformer (MSPT) to predict subseasonal daily SST, which can account for temporal variations at various scales. Initially, MSPT integrates fast Fourier transform and multilayer perceptron to extract all potential periodic scales and adaptively identify critical ones. Each periodic scale features an independent branch composed of patch embedding and Transformer encoder, dedicated to specifically learning temporal variations at that scale. Only the outputs of critical branches are weighted and aggregated to obtain effective multiperiodic scale characteristics. This approach effectively decouples complex temporal patterns, enabling the model to capture reliable dependencies that are beneficial for improving subseasonal forecasting. Furthermore, by introducing additional multivariate attention, our improved Transformer encoder can capture the inherent multivariate correlations of SST dynamics, perfecting the representation of temporal variations at specific periodic scales. Extensive subseasonal forecasting experiments conducted at four locations in the South China Sea demonstrate that MSPT achieves state-of-the-art performance in 10–30 d subseasonal daily SSTPs, validating the effectiveness of multiscale periodic information in improving subseasonal forecasting.
format Article
id doaj-art-7d8aaff3e2f74fe7bb17c703a278d7b5
institution OA Journals
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-7d8aaff3e2f74fe7bb17c703a278d7b52025-08-20T01:50:22ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01188399841510.1109/JSTARS.2025.354952410919018MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature ForecastingQi He0https://orcid.org/0009-0009-5509-0983Zhenfeng Lan1https://orcid.org/0009-0000-2028-415XWei Song2https://orcid.org/0000-0002-0604-5563Wenbo Zhang3https://orcid.org/0000-0003-1375-0081Yanling Du4https://orcid.org/0000-0003-1613-4022Wei Zhao5https://orcid.org/0000-0002-7906-1054College of Information Technology, Shanghai Ocean University, Shanghai, ChinaCollege of Information Technology, Shanghai Ocean University, Shanghai, ChinaCollege of Information Technology, Shanghai Ocean University, Shanghai, ChinaCollege of Information Technology, Shanghai Ocean University, Shanghai, ChinaCollege of Information Technology, Shanghai Ocean University, Shanghai, ChinaCollege of Information Technology, Shanghai Ocean University, Shanghai, ChinaAccurately subseasonal daily sea surface temperature prediction (SSTP) is significant for forecasting and mitigating extreme climate events related to sea surface temperature (SST). However, this scale's forecasting lies at the transitional zone between short-term forecasting and long-term climate prediction, requiring simultaneous consideration of small-scale variations crucial for the former and large-scale variations fundamental to the latter. Thus, achieving precise subseasonal daily SSTPs is challenging. In this study, we introduce a novel multiscale periodic transformer (MSPT) to predict subseasonal daily SST, which can account for temporal variations at various scales. Initially, MSPT integrates fast Fourier transform and multilayer perceptron to extract all potential periodic scales and adaptively identify critical ones. Each periodic scale features an independent branch composed of patch embedding and Transformer encoder, dedicated to specifically learning temporal variations at that scale. Only the outputs of critical branches are weighted and aggregated to obtain effective multiperiodic scale characteristics. This approach effectively decouples complex temporal patterns, enabling the model to capture reliable dependencies that are beneficial for improving subseasonal forecasting. Furthermore, by introducing additional multivariate attention, our improved Transformer encoder can capture the inherent multivariate correlations of SST dynamics, perfecting the representation of temporal variations at specific periodic scales. Extensive subseasonal forecasting experiments conducted at four locations in the South China Sea demonstrate that MSPT achieves state-of-the-art performance in 10–30 d subseasonal daily SSTPs, validating the effectiveness of multiscale periodic information in improving subseasonal forecasting.https://ieeexplore.ieee.org/document/10919018/Multiperiodic scale modelingsea surface temperature (SST)South China sea (SCS)subseasonal forecastingtemporal transformer
spellingShingle Qi He
Zhenfeng Lan
Wei Song
Wenbo Zhang
Yanling Du
Wei Zhao
MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Multiperiodic scale modeling
sea surface temperature (SST)
South China sea (SCS)
subseasonal forecasting
temporal transformer
title MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting
title_full MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting
title_fullStr MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting
title_full_unstemmed MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting
title_short MSPT: A Transformer-Based Model Using Multiscale Periodic Information for 10–30 d Subseasonal Daily Sea Surface Temperature Forecasting
title_sort mspt a transformer based model using multiscale periodic information for 10 x2013 30 d subseasonal daily sea surface temperature forecasting
topic Multiperiodic scale modeling
sea surface temperature (SST)
South China sea (SCS)
subseasonal forecasting
temporal transformer
url https://ieeexplore.ieee.org/document/10919018/
work_keys_str_mv AT qihe msptatransformerbasedmodelusingmultiscaleperiodicinformationfor10x201330dsubseasonaldailyseasurfacetemperatureforecasting
AT zhenfenglan msptatransformerbasedmodelusingmultiscaleperiodicinformationfor10x201330dsubseasonaldailyseasurfacetemperatureforecasting
AT weisong msptatransformerbasedmodelusingmultiscaleperiodicinformationfor10x201330dsubseasonaldailyseasurfacetemperatureforecasting
AT wenbozhang msptatransformerbasedmodelusingmultiscaleperiodicinformationfor10x201330dsubseasonaldailyseasurfacetemperatureforecasting
AT yanlingdu msptatransformerbasedmodelusingmultiscaleperiodicinformationfor10x201330dsubseasonaldailyseasurfacetemperatureforecasting
AT weizhao msptatransformerbasedmodelusingmultiscaleperiodicinformationfor10x201330dsubseasonaldailyseasurfacetemperatureforecasting