Downscaling daily wind speed with Bayesian deep learning for climate monitoring

Firas Gerges; Michel C. Boufadel; Elie Bou-Zeid; Hani Nassif; Jason T.L. Wang

doi:https://doi.org/10.1007/s41060-023-00397-6

Downscaling daily wind speed with Bayesian deep learning for climate monitoring

Firas Gerges, Michel C. Boufadel, Elie Bou-Zeid, Hani Nassif, Jason T.L. Wang

New Jersey Institute of Technology, Princeton University

Research output: Contribution to journal › Article › peer-review

Abstract

Wind dynamics are extremely complex and have critical impacts on the level of damage from natural hazards, such as storms and wildfires. In the wake of climate change, wind dynamics are becoming more complex, making the prediction of future wind characteristics a more challenging task. Nevertheless, having long-term projections of some wind characteristics, such as daily wind speed, is crucial for effective monitoring of climate change, and for efficient disaster risk management. Furthermore, accurate projections of wind speed result in optimized generation of wind-based electric power. General circulation models (GCMs) provide long-term simulations (often till year 2100 or more) of multiple climate variables. However, simulations from a GCM are at a grid with coarse spatial resolution, rendering them ineffective to resolve and analyze climate change at the local regional level. Spatial downscaling techniques are often used to map such global large-scale simulations to a local small-scale region. In this paper, we present a novel deep learning framework for spatial downscaling, specifically for forecasting the daily average wind speed at a local station level using GCM simulations. Our framework, named wind convolutional neural networks with transformers, or WCT for short, consists of multi-head convolutional neural networks, followed by stacked transformers, and an uncertainty quantification component based on Bayesian inference. Experimental results show the suitability of WCT when applied on four wind stations in New Jersey and Pennsylvania, USA. Moreover, we use the trained WCT on future GCM simulations to produce local-scale daily wind speed projections up to the year 2100.

Original language	English (US)
Journal	International Journal of Data Science and Analytics
DOIs	https://doi.org/10.1007/s41060-023-00397-6
State	Accepted/In press - 2023

ASJC Scopus subject areas

Information Systems
Modeling and Simulation
Computer Science Applications
Computational Theory and Mathematics
Applied Mathematics

Keywords

Climate change
Convolutional neural networks
Deep learning
Transformer
Wind speed

Access to Document

https://doi.org/10.1007/s41060-023-00397-6

Cite this

@article{374ffd5603cf4754ae975e61f7cb6506,

title = "Downscaling daily wind speed with Bayesian deep learning for climate monitoring",

abstract = "Wind dynamics are extremely complex and have critical impacts on the level of damage from natural hazards, such as storms and wildfires. In the wake of climate change, wind dynamics are becoming more complex, making the prediction of future wind characteristics a more challenging task. Nevertheless, having long-term projections of some wind characteristics, such as daily wind speed, is crucial for effective monitoring of climate change, and for efficient disaster risk management. Furthermore, accurate projections of wind speed result in optimized generation of wind-based electric power. General circulation models (GCMs) provide long-term simulations (often till year 2100 or more) of multiple climate variables. However, simulations from a GCM are at a grid with coarse spatial resolution, rendering them ineffective to resolve and analyze climate change at the local regional level. Spatial downscaling techniques are often used to map such global large-scale simulations to a local small-scale region. In this paper, we present a novel deep learning framework for spatial downscaling, specifically for forecasting the daily average wind speed at a local station level using GCM simulations. Our framework, named wind convolutional neural networks with transformers, or WCT for short, consists of multi-head convolutional neural networks, followed by stacked transformers, and an uncertainty quantification component based on Bayesian inference. Experimental results show the suitability of WCT when applied on four wind stations in New Jersey and Pennsylvania, USA. Moreover, we use the trained WCT on future GCM simulations to produce local-scale daily wind speed projections up to the year 2100.",

keywords = "Climate change, Convolutional neural networks, Deep learning, Transformer, Wind speed",

author = "Firas Gerges and Boufadel, {Michel C.} and Elie Bou-Zeid and Hani Nassif and Wang, {Jason T.L.}",

note = "Funding Information: Funding for this study was provided by the Bridge Resource Program (BRP) from the New Jersey Department of Transportation. Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive licence to Springer Nature Switzerland AG.",

year = "2023",

doi = "https://doi.org/10.1007/s41060-023-00397-6",

language = "English (US)",

journal = "International Journal of Data Science and Analytics",

issn = "2364-415X",

publisher = "Springer Nature Switzerland AG",

}

TY - JOUR

T1 - Downscaling daily wind speed with Bayesian deep learning for climate monitoring

AU - Gerges, Firas

AU - Boufadel, Michel C.

AU - Bou-Zeid, Elie

AU - Nassif, Hani

AU - Wang, Jason T.L.

N1 - Funding Information: Funding for this study was provided by the Bridge Resource Program (BRP) from the New Jersey Department of Transportation. Publisher Copyright: © 2023, The Author(s), under exclusive licence to Springer Nature Switzerland AG.

PY - 2023

Y1 - 2023

N2 - Wind dynamics are extremely complex and have critical impacts on the level of damage from natural hazards, such as storms and wildfires. In the wake of climate change, wind dynamics are becoming more complex, making the prediction of future wind characteristics a more challenging task. Nevertheless, having long-term projections of some wind characteristics, such as daily wind speed, is crucial for effective monitoring of climate change, and for efficient disaster risk management. Furthermore, accurate projections of wind speed result in optimized generation of wind-based electric power. General circulation models (GCMs) provide long-term simulations (often till year 2100 or more) of multiple climate variables. However, simulations from a GCM are at a grid with coarse spatial resolution, rendering them ineffective to resolve and analyze climate change at the local regional level. Spatial downscaling techniques are often used to map such global large-scale simulations to a local small-scale region. In this paper, we present a novel deep learning framework for spatial downscaling, specifically for forecasting the daily average wind speed at a local station level using GCM simulations. Our framework, named wind convolutional neural networks with transformers, or WCT for short, consists of multi-head convolutional neural networks, followed by stacked transformers, and an uncertainty quantification component based on Bayesian inference. Experimental results show the suitability of WCT when applied on four wind stations in New Jersey and Pennsylvania, USA. Moreover, we use the trained WCT on future GCM simulations to produce local-scale daily wind speed projections up to the year 2100.

AB - Wind dynamics are extremely complex and have critical impacts on the level of damage from natural hazards, such as storms and wildfires. In the wake of climate change, wind dynamics are becoming more complex, making the prediction of future wind characteristics a more challenging task. Nevertheless, having long-term projections of some wind characteristics, such as daily wind speed, is crucial for effective monitoring of climate change, and for efficient disaster risk management. Furthermore, accurate projections of wind speed result in optimized generation of wind-based electric power. General circulation models (GCMs) provide long-term simulations (often till year 2100 or more) of multiple climate variables. However, simulations from a GCM are at a grid with coarse spatial resolution, rendering them ineffective to resolve and analyze climate change at the local regional level. Spatial downscaling techniques are often used to map such global large-scale simulations to a local small-scale region. In this paper, we present a novel deep learning framework for spatial downscaling, specifically for forecasting the daily average wind speed at a local station level using GCM simulations. Our framework, named wind convolutional neural networks with transformers, or WCT for short, consists of multi-head convolutional neural networks, followed by stacked transformers, and an uncertainty quantification component based on Bayesian inference. Experimental results show the suitability of WCT when applied on four wind stations in New Jersey and Pennsylvania, USA. Moreover, we use the trained WCT on future GCM simulations to produce local-scale daily wind speed projections up to the year 2100.

KW - Climate change

KW - Convolutional neural networks

KW - Deep learning

KW - Transformer

KW - Wind speed

UR - http://www.scopus.com/inward/record.url?scp=85160822445&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85160822445&partnerID=8YFLogxK

U2 - https://doi.org/10.1007/s41060-023-00397-6

DO - https://doi.org/10.1007/s41060-023-00397-6

M3 - Article

SN - 2364-415X

JO - International Journal of Data Science and Analytics

JF - International Journal of Data Science and Analytics

ER -

Downscaling daily wind speed with Bayesian deep learning for climate monitoring

Abstract

ASJC Scopus subject areas

Keywords

Access to Document

Other files and links

Fingerprint

Cite this