Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring

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

doi:10.1007/978-3-031-26419-1_34

Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring

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

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Scopus citations

Abstract

General Circulation Models (GCMs) are established numerical models for simulating multiple climate variables, decades into the future. GCMs produce such simulations at coarse resolution (100 to 600 km), making them inappropriate to monitor climate change at the local regional level. Downscaling approaches are usually adopted to infer the statistical relationship between the coarse simulations of GCMs and local observations and use the relationship to evaluate the simulations at a finer scale. In this paper, we propose a novel deep learning framework for forecasting daily precipitation values via downscaling. Our framework, named Precipitation CNN or PCNN, employs multi-head convolutional neural networks (CNNs) followed by Bahdanau attention blocks and an uncertainty quantification component with Bayesian inference. We apply PCNN to downscale the daily precipitation above the New Jersey portion of the Hackensack-Passaic watershed. Experiments show that PCNN is suitable for this task, reproducing the daily variability of precipitation. Moreover, we produce local-scale precipitation projections for multiple periods into the future (up to year 2100).

Original language	English (US)
Title of host publication	Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings
Editors	Massih-Reza Amini, Stéphane Canu, Asja Fischer, Tias Guns, Petra Kralj Novak, Grigorios Tsoumakas
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	565-580
Number of pages	16
ISBN (Print)	9783031264184
DOIs	https://doi.org/10.1007/978-3-031-26419-1_34
State	Published - 2023
Externally published	Yes
Event	22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022 - Grenoble, France Duration: Sep 19 2022 → Sep 23 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13717 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022
Country/Territory	France
City	Grenoble
Period	9/19/22 → 9/23/22

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
General Computer Science

Keywords

Climate change
Convolutional neural networks
Machine learning
Statistical downscaling

Access to Document

10.1007/978-3-031-26419-1_34

Cite this

Gerges, F., Boufadel, M. C., Bou-Zeid, E., Darekar, A., Nassif, H., & Wang, J. T. L. (2023). Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring. In M.-R. Amini, S. Canu, A. Fischer, T. Guns, P. Kralj Novak, & G. Tsoumakas (Eds.), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings (pp. 565-580). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13717 LNAI). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-26419-1_34

Gerges, Firas ; Boufadel, Michel C. ; Bou-Zeid, Elie et al. / Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring. Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings. editor / Massih-Reza Amini ; Stéphane Canu ; Asja Fischer ; Tias Guns ; Petra Kralj Novak ; Grigorios Tsoumakas. Springer Science and Business Media Deutschland GmbH, 2023. pp. 565-580 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{d2705706242544a28703db96a473be2c,

title = "Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring",

abstract = "General Circulation Models (GCMs) are established numerical models for simulating multiple climate variables, decades into the future. GCMs produce such simulations at coarse resolution (100 to 600 km), making them inappropriate to monitor climate change at the local regional level. Downscaling approaches are usually adopted to infer the statistical relationship between the coarse simulations of GCMs and local observations and use the relationship to evaluate the simulations at a finer scale. In this paper, we propose a novel deep learning framework for forecasting daily precipitation values via downscaling. Our framework, named Precipitation CNN or PCNN, employs multi-head convolutional neural networks (CNNs) followed by Bahdanau attention blocks and an uncertainty quantification component with Bayesian inference. We apply PCNN to downscale the daily precipitation above the New Jersey portion of the Hackensack-Passaic watershed. Experiments show that PCNN is suitable for this task, reproducing the daily variability of precipitation. Moreover, we produce local-scale precipitation projections for multiple periods into the future (up to year 2100).",

keywords = "Climate change, Convolutional neural networks, Machine learning, Statistical downscaling",

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

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022 ; Conference date: 19-09-2022 Through 23-09-2022",

year = "2023",

doi = "10.1007/978-3-031-26419-1_34",

language = "English (US)",

isbn = "9783031264184",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "565--580",

editor = "Massih-Reza Amini and St{\'e}phane Canu and Asja Fischer and Tias Guns and {Kralj Novak}, Petra and Grigorios Tsoumakas",

booktitle = "Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings",

address = "Germany",

}

Gerges, F, Boufadel, MC, Bou-Zeid, E, Darekar, A, Nassif, H & Wang, JTL 2023, Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring. in M-R Amini, S Canu, A Fischer, T Guns, P Kralj Novak & G Tsoumakas (eds), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13717 LNAI, Springer Science and Business Media Deutschland GmbH, pp. 565-580, 22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022, Grenoble, France, 9/19/22. https://doi.org/10.1007/978-3-031-26419-1_34

Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring. / Gerges, Firas; Boufadel, Michel C.; Bou-Zeid, Elie et al.
Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings. ed. / Massih-Reza Amini; Stéphane Canu; Asja Fischer; Tias Guns; Petra Kralj Novak; Grigorios Tsoumakas. Springer Science and Business Media Deutschland GmbH, 2023. p. 565-580 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13717 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring

AU - Gerges, Firas

AU - Boufadel, Michel C.

AU - Bou-Zeid, Elie

AU - Darekar, Ankit

AU - Nassif, Hani

AU - Wang, Jason T.L.

PY - 2023

Y1 - 2023

N2 - General Circulation Models (GCMs) are established numerical models for simulating multiple climate variables, decades into the future. GCMs produce such simulations at coarse resolution (100 to 600 km), making them inappropriate to monitor climate change at the local regional level. Downscaling approaches are usually adopted to infer the statistical relationship between the coarse simulations of GCMs and local observations and use the relationship to evaluate the simulations at a finer scale. In this paper, we propose a novel deep learning framework for forecasting daily precipitation values via downscaling. Our framework, named Precipitation CNN or PCNN, employs multi-head convolutional neural networks (CNNs) followed by Bahdanau attention blocks and an uncertainty quantification component with Bayesian inference. We apply PCNN to downscale the daily precipitation above the New Jersey portion of the Hackensack-Passaic watershed. Experiments show that PCNN is suitable for this task, reproducing the daily variability of precipitation. Moreover, we produce local-scale precipitation projections for multiple periods into the future (up to year 2100).

AB - General Circulation Models (GCMs) are established numerical models for simulating multiple climate variables, decades into the future. GCMs produce such simulations at coarse resolution (100 to 600 km), making them inappropriate to monitor climate change at the local regional level. Downscaling approaches are usually adopted to infer the statistical relationship between the coarse simulations of GCMs and local observations and use the relationship to evaluate the simulations at a finer scale. In this paper, we propose a novel deep learning framework for forecasting daily precipitation values via downscaling. Our framework, named Precipitation CNN or PCNN, employs multi-head convolutional neural networks (CNNs) followed by Bahdanau attention blocks and an uncertainty quantification component with Bayesian inference. We apply PCNN to downscale the daily precipitation above the New Jersey portion of the Hackensack-Passaic watershed. Experiments show that PCNN is suitable for this task, reproducing the daily variability of precipitation. Moreover, we produce local-scale precipitation projections for multiple periods into the future (up to year 2100).

KW - Climate change

KW - Convolutional neural networks

KW - Machine learning

KW - Statistical downscaling

UR - http://www.scopus.com/inward/record.url?scp=85151046332&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85151046332&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-26419-1_34

DO - 10.1007/978-3-031-26419-1_34

M3 - Conference contribution

AN - SCOPUS:85151046332

SN - 9783031264184

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 565

EP - 580

BT - Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings

A2 - Amini, Massih-Reza

A2 - Canu, Stéphane

A2 - Fischer, Asja

A2 - Guns, Tias

A2 - Kralj Novak, Petra

A2 - Tsoumakas, Grigorios

PB - Springer Science and Business Media Deutschland GmbH

T2 - 22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022

Y2 - 19 September 2022 through 23 September 2022

ER -

Gerges F, Boufadel MC, Bou-Zeid E, Darekar A, Nassif H, Wang JTL. Bayesian Multi-head Convolutional Neural Networks with Bahdanau Attention for Forecasting Daily Precipitation in Climate Change Monitoring. In Amini MR, Canu S, Fischer A, Guns T, Kralj Novak P, Tsoumakas G, editors, Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2022, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 565-580. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-26419-1_34