pith. sign in

arxiv: 2605.09036 · v1 · submitted 2026-05-09 · 💻 cs.LG

PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation

Zesheng Liu , Doyup Kwon , Ning Lin , Maryam Rahnemoonfar This is my paper

Pith reviewed 2026-05-12 01:47 UTC · model grok-4.3

classification 💻 cs.LG
keywords storm surge emulationgraph transformerspeak-aware learningcross-attentioncoastal hazardsclimate forcingtime-series forecastingextreme event modeling
0
0 comments X

The pith

PACT uses a peak-aware graph transformer to outperform baselines in predicting extreme storm surges from atmospheric data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops PACT as an efficient emulator for station-level storm-surge heights driven by atmospheric forcing fields. It encodes spatial patches with GraphSAGE, aggregates via learned station queries in cross-attention, models time with a Transformer encoder, and decodes lead-specific forecasts. A peak-aware auxiliary head plus tail-focused loss and slope regularizer target better capture of extremes and coherent trajectories. This matters for coastal risk work because full hydrodynamic models cannot run the thousands of climate scenarios needed under changing forcings. Tests on US Northeast tide gauges show lower RMSE and MAE than a strong graph baseline, with gains in peak height and tail behavior for both reanalysis and most CMIP6 runs.

Core claim

PACT represents each forcing patch as a graph, encodes spatial structure with GraphSAGE, and uses a learned station query to aggregate node information through cross-attention rather than uniform pooling. A Transformer encoder models temporal dependence across the forcing history, and a horizon-query decoder generates lead-specific forecasts from a shared temporal memory. To better capture extreme events, it couples a lightweight auxiliary peak-aware head with a tailored training objective that includes a tail-focused loss on peak-dominated samples and a horizon-wise slope regularizer to encourage coherent multi-step evolution.

What carries the argument

The peak-aware cross-attention graph transformer architecture together with its auxiliary peak head, tail-focused loss, and horizon-wise slope regularizer.

If this is right

  • Improved peak fidelity directly lowers underestimation of coastal flood heights during extremes.
  • The 3.5-second inference time per winter season enables large climate-scenario ensembles that hydrodynamic models cannot handle.
  • Within-family transfer across CMIP6 forcings supports use in future-climate surge projections, while the reanalysis-to-GCM drop flags a remaining domain gap.
  • The horizon-wise regularizer supports stable multi-day lead forecasts without separate per-horizon models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same peak-aware losses could be tested on other coastal variables such as wave height or total water level to check generality.
  • The persistent reanalysis-GCM transfer gap suggests adding explicit domain-adaptation layers or physics-informed constraints as a next step.
  • Because the station query is learned, the architecture might extend to ungauged locations by interpolating query embeddings from nearby observed sites.
  • Efficiency gains open the door to coupling the emulator inside real-time early-warning systems that update every few hours during a storm.

Load-bearing premise

The peak-aware auxiliary head, tail-focused loss, and horizon-wise slope regularizer will reliably improve extreme-event capture and multi-step coherence without introducing overfitting or sensitivity to the choice of peak-dominated samples.

What would settle it

On held-out tide-gauge records or additional CMIP6 forcings, compute peak-specific RMSE and check whether PACT no longer shows lower values than the spatio-temporal graph baseline while maintaining similar overall RMSE.

Figures

Figures reproduced from arXiv: 2605.09036 by Doyup Kwon, Maryam Rahnemoonfar, Ning Lin, Zesheng Liu.

Figure 1
Figure 1. Figure 1: Overview of the computational setup used in this study. (a) The ADCIRC com￾putational domain over the western North Atlantic, with the fixed local study region highlighted in red. The dashed blue line defines the extents of the computational mesh used for numerical simulations. (b) Enlarged view of the study region showing the four tide-gauge stations used in the analysis: Lewes, Delaware; Boston, Massachu… view at source ↗
Figure 2
Figure 2. Figure 2: Overall architecture of PACT. 3.2 Spatial Graph Encoding of Local Atmospheric Forcing We begin by encoding each station-centered forcing patch as a graph so that the model can learn spatially structured representations of the local atmospheric state be￾fore any station-level aggregation is applied. Specifically, given a forcing graph Gt−k(where k = 0, 6h, 12h) at each time step in the input window, we use … view at source ↗
Figure 3
Figure 3. Figure 3: Diagram of station-query readout. The station query is formed by adding a learn￾able station token to the output of a small station encoder that processes fixed station metadata. The resulting query attends over the node embeddings produced by the GraphSAGE encoder through cross-attention, yielding a station-specific representation of the forcing field. Formally, let Uτ ∈ R N×d denote the matrix of node em… view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative examples of surge level time series predictions at Battery, NY, made by PACT with peak aware loss on NCEP Reanalysis dataset. MSE loss as the base PACT configuration, and PACT trained with the peak-aware loss as the best PACT configuration. To keep the main text focused, the detailed ablation studies on spatial mean cen￾tering of pressure anomalies and on the peak-aware design are presented in … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative examples of surge level time series predictions on year 2080-2081 in the CNRM dataset, with the model trained on different datasets. Having established the within-dataset performance of PACT, we now examine cross￾dataset generalization, which is substantially more demanding because the atmospheric forcing distributions can differ across reanalysis and climate-model products. We focus on the bes… view at source ↗
Figure 6
Figure 6. Figure 6: Event-level peak prediction accuracy on the Battery station for future-year CMIP6 forcings. Each point corresponds to a detected surge event and plots the predicted peak magni￾tude ˆp against the ground-truth peak magnitude p (in meters). Events are identified as clusters of exceedances above the 95th percentile within each year, and predicted/ground-truth events are paired in temporal order with a 48-hour… view at source ↗
Figure 7
Figure 7. Figure 7: Distributional fidelity of storm-surge peak magnitudes on the Battery station for future-year CMIP6 forcings. We report Gaussian kernel density estimates of event peak magni￾tudes (meters) for the ADCIRC-driven ground truth (GT) and model predictions (Baseline-12h, PACT-Base, PACT-Best), using a shared bandwidth for comparability. Panels correspond to different CMIP6 datasets (AWI, CNRM, EC EARTH, MPI, MRI… view at source ↗
Figure 8
Figure 8. Figure 8: Severity-conditioned peak error on the Battery station for future-year CMIP6 forc￾ings. Peak RMSE (meters) is computed as a function of event severity by binning ground-truth peak magnitudes into quantile bins spanning the 1st–99th percentiles and evaluating RMSE within each bin. The resulting curve is smoothed with a short moving-window mean, and bins with fewer than five events are omitted. Panels corres… view at source ↗
read the original abstract

Accurate and efficient storm-surge emulation is essential for coastal hazard assessment, yet high-fidelity hydrodynamic models remain too expensive for large scenario ensembles and rapid evaluation under heterogeneous climate forcings. We present PACT, a peak-aware cross-attention graph transformer for efficient station-level storm-surge prediction from atmospheric forcing fields. PACT represents each forcing patch as a graph, encodes spatial structure with GraphSAGE, and uses a learned station query to aggregate node information through cross-attention rather than uniform pooling. A Transformer encoder models temporal dependence across the forcing history, and a horizon-query decoder generates lead-specific forecasts from a shared temporal memory. To better capture extreme events, we introduce a peak-aware learning strategy that couples a lightweight auxiliary peak-aware head with a tailored training objective, including a tail-focused loss on peak-dominated samples and a horizon-wise slope regularizer to encourage coherent multi-step evolution. Across multiple tide-gauge stations along the US Northeast coast, PACT outperforms a strong spatio-temporal graph neural network baseline in both RMSE and MAE. Diagnostics show improved peak fidelity and tail preservation for reanalysis and most CMIP6 datasets. PACT is also computationally efficient, requiring about 3.5~s to generate a full winter-season surge trajectory for one year after training. Under distribution shift across five CMIP6 forcings, PACT transfers well within the CMIP6 family but degrades markedly when transferring from reanalysis to climate-model forcings, highlighting a persistent reanalysis--GCM gap.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper introduces PACT, a peak-aware cross-attention graph transformer for station-level storm-surge emulation from atmospheric forcings. It encodes spatial structure with GraphSAGE, aggregates via learned station queries in cross-attention, models temporal dependence with a Transformer encoder, and decodes lead-specific forecasts; a peak-aware strategy adds an auxiliary head, tail-focused loss on peak-dominated samples, and horizon-wise slope regularizer. The central empirical claim is that PACT outperforms a strong spatio-temporal GNN baseline in RMSE and MAE across US Northeast tide-gauge stations, with improved peak fidelity and tail preservation on reanalysis and most CMIP6 datasets, while being computationally efficient (~3.5 s per winter-season trajectory) and transferring reasonably within the CMIP6 family but degrading from reanalysis to GCM forcings.

Significance. If the reported gains prove robust and attributable to the proposed components, the work would be significant for enabling large-ensemble coastal hazard assessments under climate change, where high-fidelity hydrodynamic models are too costly. The efficiency and explicit focus on extremes address practical needs in operational and CMIP6-driven risk modeling.

major comments (3)
  1. [Abstract / peak-aware learning strategy] Abstract and methods (peak-aware learning strategy): No ablation studies or sensitivity analyses isolate the contribution of the auxiliary peak-aware head, tail-focused loss on peak-dominated samples, and horizon-wise slope regularizer to the claimed peak-fidelity and tail-preservation gains. Without these, it remains unclear whether improvements stem from these elements or from the GraphSAGE + cross-attention architecture and the peak-selection heuristic itself.
  2. [Results / evaluation] Results and evaluation sections: The manuscript provides no quantitative details on baseline specifications, exact RMSE/MAE values with error bars, statistical tests, data splits, or loss formulations, despite the central claim resting on outperformance across reanalysis and CMIP6 datasets. This absence prevents assessment of effect sizes and reliability.
  3. [Transferability / CMIP6 experiments] Transferability analysis: The noted degradation from reanalysis to CMIP6 forcings is reported but not examined for interaction with the peak-aware components; if the tail-focused loss and slope regularizer overfit reanalysis peak statistics, this could explain the distribution-shift sensitivity and undermine generalization claims.
minor comments (2)
  1. [Abstract] The abstract states performance improvements without referencing specific tables or figures containing the supporting metrics, making the summary harder to connect to the evidence.
  2. [Methods] Notation for the tail-focused loss weight and slope regularizer coefficient is introduced but not explicitly tied to the free-parameter list or training protocol.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on our manuscript. We have carefully considered each major comment and provide point-by-point responses below, along with our plans for revisions.

read point-by-point responses

  1. Referee: [Abstract / peak-aware learning strategy] Abstract and methods (peak-aware learning strategy): No ablation studies or sensitivity analyses isolate the contribution of the auxiliary peak-aware head, tail-focused loss on peak-dominated samples, and horizon-wise slope regularizer to the claimed peak-fidelity and tail-preservation gains. Without these, it remains unclear whether improvements stem from these elements or from the GraphSAGE + cross-attention architecture and the peak-selection heuristic itself.

    Authors: We agree that ablation studies are essential to isolate the impact of the peak-aware components. In the revised manuscript, we will include comprehensive ablation experiments that systematically remove the auxiliary peak-aware head, the tail-focused loss, and the horizon-wise slope regularizer, reporting their effects on RMSE, MAE, peak fidelity, and tail preservation metrics. This will help attribute the performance gains more precisely. revision: yes

  2. Referee: [Results / evaluation] Results and evaluation sections: The manuscript provides no quantitative details on baseline specifications, exact RMSE/MAE values with error bars, statistical tests, data splits, or loss formulations, despite the central claim resting on outperformance across reanalysis and CMIP6 datasets. This absence prevents assessment of effect sizes and reliability.

    Authors: We acknowledge the need for greater transparency in the quantitative results. We will revise the results section to include exact RMSE and MAE values with error bars (e.g., standard deviations over multiple seeds or cross-validation folds), detailed specifications of the baseline model, the data split methodology, full mathematical formulations of all loss terms, and results of statistical significance tests (such as paired t-tests) to support the outperformance claims. revision: yes

  3. Referee: [Transferability / CMIP6 experiments] Transferability analysis: The noted degradation from reanalysis to CMIP6 forcings is reported but not examined for interaction with the peak-aware components; if the tail-focused loss and slope regularizer overfit reanalysis peak statistics, this could explain the distribution-shift sensitivity and undermine generalization claims.

    Authors: This is a valid concern regarding potential overfitting of the peak-aware strategy to reanalysis data. To address it, we will extend the transferability analysis in the revision by evaluating ablated models (lacking the peak-aware elements) under the same reanalysis-to-CMIP6 transfer scenarios. We will report whether the degradation is more or less pronounced without these components, thereby clarifying their role in the observed distribution shift. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical architecture and evaluation

full rationale

The paper describes a neural architecture (GraphSAGE + cross-attention + Transformer + auxiliary peak head) trained with a composite loss and evaluated via RMSE/MAE on tide-gauge data. No equations, uniqueness theorems, or first-principles derivations are presented that could reduce to fitted quantities by construction. All performance claims are direct empirical comparisons against a baseline; the auxiliary components are motivated heuristically but their effect is measured on held-out data rather than assumed. No self-citation load-bearing steps appear in the provided text. This is standard self-contained ML experimentation.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The abstract does not enumerate specific hyperparameters or background assumptions in detail; the model depends on standard graph and transformer components plus custom loss terms whose weighting is not specified.

free parameters (2)
  • tail-focused loss weight
    Weighting factor applied to peak-dominated samples in the tailored training objective.
  • slope regularizer coefficient
    Strength of the horizon-wise slope regularizer used to encourage coherent multi-step forecasts.
axioms (1)
  • domain assumption Graph representation of atmospheric forcing patches via GraphSAGE captures the spatial structure relevant to station-level surge.
    Invoked by the choice to encode forcing patches as graphs rather than grids or images.

pith-pipeline@v0.9.0 · 5581 in / 1438 out tokens · 48110 ms · 2026-05-12T01:47:50.524786+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

109 extracted references · 109 canonical work pages

  1. [1]

    Proceedings of the 38th International Conference on Machine Learning , pages =

    Perceiver: General Perception with Iterative Attention , author =. Proceedings of the 38th International Conference on Machine Learning , pages =. 2021 , editor =

    work page 2021

  2. [2]

    and Ying, Rex and Leskovec, Jure , title =

    Hamilton, William L. and Ying, Rex and Leskovec, Jure , title =. NIPS , year =

    work page

  3. [3]

    2025 , eprint=

    GRIT-LP: Graph Transformer with Long-Range Skip Connection and Partitioned Spatial Graphs for Accurate Ice Layer Thickness Prediction , author=. 2025 , eprint=

    work page 2025

  4. [4]

    Perceiver

    Andrew Jaegle and Sebastian Borgeaud and Jean-Baptiste Alayrac and Carl Doersch and Catalin Ionescu and David Ding and Skanda Koppula and Daniel Zoran and Andrew Brock and Evan Shelhamer and Olivier J Henaff and Matthew Botvinick and Andrew Zisserman and Oriol Vinyals and Joao Carreira , booktitle=. Perceiver. 2022 , url=

    work page 2022

  5. [5]

    and Kaiser,

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N. and Kaiser,. Attention is all you need , year =. Proceedings of the 31st International Conference on Neural Information Processing Systems , pages =

    work page

  6. [6]

    Are Sixteen Heads Really Better than One? , url =

    Michel, Paul and Levy, Omer and Neubig, Graham , booktitle =. Are Sixteen Heads Really Better than One? , url =

    work page

  7. [7]

    Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned

    Voita, Elena and Talbot, David and Moiseev, Fedor and Sennrich, Rico and Titov, Ivan. Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019

    work page 2019

  8. [8]

    International Conference on Learning Representations , year=

    Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer , author=. International Conference on Learning Representations , year=

    work page

  9. [9]

    British Machine Vision Conference , year=

    Revisiting single-gated Mixtures of Experts , author=. British Machine Vision Conference , year=

    work page

  10. [10]

    , title =

    Barron, Jonathan T. , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    work page

  11. [11]

    Casey Dietrich and Wei Du and Philip B

    Antonia Sebastian and Jennifer Proft and J. Casey Dietrich and Wei Du and Philip B. Bedient and Clint N. Dawson , keywords =. Characterizing hurricane storm surge behavior in Galveston Bay using the SWAN+ADCIRC model , journal =. 2014 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2014.03.002 , url =

    work page doi:10.1016/j.coastaleng.2014.03.002 2014

  12. [12]

    National Weather Digest , volume=

    The role of the SLOSH model in National Weather Service storm surge forecasting , author=. National Weather Digest , volume=

    work page

  13. [13]

    Earth's Future , volume =

    Begmohammadi, Amirhosein and Lin, Ning and Xi, Dazhi and Blackshaw, Christine , title =. Earth's Future , volume =. doi:https://doi.org/10.1029/2025EF006063 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2025EF006063 , note =

    work page doi:10.1029/2025ef006063

  14. [14]

    Butler and M

    T. Butler and M. U. Altaf and C. Dawson and I. Hoteit and X. Luo and T. Mayo. Data Assimilation within the Advanced Circulation (ADCIRC) Modeling Framework for Hurricane Storm Surge Forecasting. Monthly Weather Review. 2012. doi:10.1175/MWR-D-11-00118.1

    work page doi:10.1175/mwr-d-11-00118.1 2012

  15. [15]

    A framework for flexible peak storm surge prediction , journal =

    Benjamin Pachev and Prateek Arora and Carlos del-Castillo-Negrete and Eirik Valseth and Clint Dawson , keywords =. A framework for flexible peak storm surge prediction , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2023.104406 , url =

    work page doi:10.1016/j.coastaleng.2023.104406 2023

  16. [16]

    and Scheffner, Norman , year =

    Luettich, Jr, Richard and Westerink, J. and Scheffner, Norman , year =. ADCIRC: An Advanced Three-Dimensional Circulation Model for Shelves, Coasts, and Estuaries. Report 1. Theory and Methodology of ADCIRC-2DDI and ADCIRC-3DL , journal =

    work page

  17. [17]

    High-fidelity hurricane surge forecasting using emulation and sequential experiments , volume =

    Plumlee, Matthew and Asher, Taylor and Chang, Won and Bilskie, Matthew , year =. High-fidelity hurricane surge forecasting using emulation and sequential experiments , volume =. The Annals of Applied Statistics , doi =

    work page

  18. [18]

    and Hagen, S

    Bilskie, Matthew V. and Hagen, S. C. and Alizad, K. and Medeiros, S. C. and Passeri, D. L. and Needham, H. F. and Cox, A. , title =. Earth's Future , volume =. doi:https://doi.org/10.1002/2015EF000347 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2015EF000347 , abstract =

    work page doi:10.1002/2015ef000347

  19. [19]

    Journal of Marine Science and Engineering , VOLUME =

    Wright, Lynn Donelson and Thom, Bruce Graham , TITLE =. Journal of Marine Science and Engineering , VOLUME =. 2023 , NUMBER =

    work page 2023

  20. [20]

    2019 , doi =

    Storm surge return levels induced by mid-to-late-twenty-first-century extratropical cyclones in the Northeastern United States , journal =. 2019 , doi =

    work page 2019

  21. [21]

    and Curchitser, Enrique N

    Ramos-Valle, Alexandra N. and Curchitser, Enrique N. and Bruyère, Cindy L. and McOwen, Sean , title =. Journal of Geophysical Research: Atmospheres , volume =. doi:https://doi.org/10.1029/2020JD033266 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2020JD033266 , note =

    work page doi:10.1029/2020jd033266

  22. [22]

    and Lin, Ning and Oppenheimer, Michael and Lai, Ching-Yao , title =

    Lockwood, Joseph W. and Lin, Ning and Oppenheimer, Michael and Lai, Ching-Yao , title =. Journal of Geophysical Research: Atmospheres , volume =. doi:https://doi.org/10.1029/2022JD037617 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2022JD037617 , note =

    work page doi:10.1029/2022jd037617

  23. [23]

    Tseng and C.D

    C.M. Tseng and C.D. Jan and J.S. Wang and C.M. Wang , keywords =. Application of artificial neural networks in typhoon surge forecasting , journal =. 2007 , issn =. doi:https://doi.org/10.1016/j.oceaneng.2006.09.005 , url =

    work page doi:10.1016/j.oceaneng.2006.09.005 2007

  24. [24]

    Scientific Reports , year =

    Timothy Tiggeloven and Anaïs Couasnon and Chiem van Straaten and Sanne Muis and Philip Ward , title =. Scientific Reports , year =

    work page

  25. [25]

    Data-Driven Modeling of Global Storm Surges , volume =

    Tadesse, Michael Getachew and Wahl, Thomas and Cid, Alba , year =. Data-Driven Modeling of Global Storm Surges , volume =. Frontiers in Marine Science , doi =

    work page

  26. [26]

    Prediction of storm surge and inundation using climatological datasets for the Indian coast using soft computing techniques , volume =

    Sahoo, Bishnupriya , year =. Prediction of storm surge and inundation using climatological datasets for the Indian coast using soft computing techniques , volume =. Soft Computing , doi =

    work page

  27. [27]

    Reza Hashemi and Malcolm L

    M. Reza Hashemi and Malcolm L. Spaulding and Alex Shaw and Hamed Farhadi and Matt Lewis , year =. An efficient artificial intelligence model for prediction of tropical storm surge , volume =

    work page

  28. [28]

    A real time storm surge forecasting system using ADCIRC , journal =

    Fleming, Jason and Fulcher, Crystal and Luettich, Jr, Richard and Estrade, Brett and Allen, Gabrielle and Winer, Harley , year =. A real time storm surge forecasting system using ADCIRC , journal =

    work page

  29. [29]

    Report 1

    ADCIRC: An Advanced Three-Dimensional Circulation Model for Shelves, Coasts, and Estuaries. Report 1. Theory and Methodology of ADCIRC-2DDI and ADCIRC-3DL. , author=

    work page

  30. [30]

    Bulletin of the American Meteorological Society , year = 1996, month = mar, volume =

    The NCEP/NCAR 40-Year Reanalysis Project. Bulletin of the American Meteorological Society , year = 1996, month = mar, volume =

    work page 1996

  31. [31]

    and Bony, S

    Eyring, V. and Bony, S. and Meehl, G. A. and Senior, C. A. and Stevens, B. and Stouffer, R. J. and Taylor, K. E. , TITLE =. Geoscientific Model Development , VOLUME =. 2016 , NUMBER =

    work page 2016

  32. [32]

    Egbert and Svetlana Y

    Gary D. Egbert and Svetlana Y. Erofeeva. Efficient Inverse Modeling of Barotropic Ocean Tides. Journal of Atmospheric and Oceanic Technology. 2002. doi:10.1175/1520-0426(2002)019<0183:EIMOBO>2.0.CO;2

    work page doi:10.1175/1520-0426(2002)019 2002

  33. [33]

    and Emanuel, K

    Lin, N. and Emanuel, K. A. and Smith, J. A. and Vanmarcke, E. , title =. Journal of Geophysical Research: Atmospheres , volume =. doi:https://doi.org/10.1029/2009JD013630 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2009JD013630 , abstract =

    work page doi:10.1029/2009jd013630

  34. [34]

    Using surrogate modeling to predict storm surge on evolving landscapes under climate change

    Mohammad Ahmadi Gharehtoragh and David Johnson. Using surrogate modeling to predict storm surge on evolving landscapes under climate change. npj Natural Hazards. 2024

    work page 2024

  35. [35]

    Developing a novel hybrid deep learning model for extratropical storm surge forecasting: A case in the Bohai Sea , journal =

    Zhicheng Zhu and Chengqing Ruan and Qinrong Liu and Zhifeng Wang and Jinsheng Qi , keywords =. Developing a novel hybrid deep learning model for extratropical storm surge forecasting: A case in the Bohai Sea , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.csr.2025.105621 , url =

    work page doi:10.1016/j.csr.2025.105621 2026

  36. [36]

    and Simionato, Claudia G

    Dinapoli, Matías G. and Simionato, Claudia G. , title =. Journal of Geophysical Research: Oceans , volume =. doi:https://doi.org/10.1029/2024JC021685 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2024JC021685 , note =

    work page doi:10.1029/2024jc021685

  37. [37]

    Catalano and Anthony J

    Arielle J. Catalano and Anthony J. Broccoli. Synoptic Characteristics of Surge-Producing Extratropical Cyclones along the Northeast Coast of the United States. Journal of Applied Meteorology and Climatology. 2018. doi:10.1175/JAMC-D-17-0123.1

    work page doi:10.1175/jamc-d-17-0123.1 2018

  38. [38]

    Journal of Geophysical Research: Oceans , volume =

    Marsooli, Reza and Lin, Ning , title =. Journal of Geophysical Research: Oceans , volume =. doi:https://doi.org/10.1029/2017JC013434 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2017JC013434 , year =

    work page doi:10.1029/2017jc013434

  39. [39]

    Scientific Reports , volume=

    Machine learning-based assessment of storm surge in the New York metropolitan area , author=. Scientific Reports , volume=. 2022 , publisher=

    work page 2022

  40. [40]

    Kyprioti and Christopher Irwin and Alexandros A

    Aikaterini P. Kyprioti and Christopher Irwin and Alexandros A. Taflanidis and Norberto C. Nadal-Caraballo and Madison C. Yawn and Luke A. Aucoin , keywords =. Spatio-temporal storm surge emulation using Gaussian Process techniques , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2022.104231 , url =

    work page doi:10.1016/j.coastaleng.2022.104231 2023

  41. [41]

    , title=

    Adeli, Ehsan and Sun, Luning and Wang, Jianxun and Taflanidis, Alexandros A. , title=. Neural Computing and Applications , year=. doi:10.1007/s00521-023-08719-2 , url=

    work page doi:10.1007/s00521-023-08719-2

  42. [42]

    Irish and Michelle T

    Jun-Whan Lee and Jennifer L. Irish and Michelle T. Bensi and Douglas C. Marcy , keywords =. Rapid prediction of peak storm surge from tropical cyclone track time series using machine learning , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2021.104024 , url =

    work page doi:10.1016/j.coastaleng.2021.104024 2021

  43. [43]

    Tausía and S

    J. Tausía and S. Delaux and P. Camus and A. Rueda and F. Méndez and K.R. Bryan and J. Pérez and C.G.R. Costa and R. Zyngfogel and A. Cofiño , keywords =. Rapid response data-driven reconstructions for storm surge around New Zealand , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.apor.2023.103496 , url =

    work page doi:10.1016/j.apor.2023.103496 2023

  44. [44]

    URL https://www.sciencedirect.com/science/article/abs/ pii/S0895611123000241

    Wenhong Xie and Guangjun Xu and Hongchun Zhang and Changming Dong , keywords =. Developing a deep learning-based storm surge forecasting model , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.ocemod.2023.102179 , url =

    work page doi:10.1016/j.ocemod.2023.102179 2023

  45. [45]

    Soft Computing , year=

    Wang, Tiantian and Liu, Tiezhong and Lu, Yunmeng , title=. Soft Computing , year=. doi:10.1007/s00500-022-07508-8 , url=

    work page doi:10.1007/s00500-022-07508-8

  46. [46]

    Artificial neural network-based storm surge forecast model: Practical application to Sakai Minato, Japan , journal =

    Sooyoul Kim and Shunqi Pan and Hajime Mase , keywords =. Artificial neural network-based storm surge forecast model: Practical application to Sakai Minato, Japan , journal =. 2019 , issn =. doi:https://doi.org/10.1016/j.apor.2019.101871 , url =

    work page doi:10.1016/j.apor.2019.101871 2019

  47. [47]

    Water , VOLUME =

    Chao, Wei-Ting and Young, Chih-Chieh and Hsu, Tai-Wen and Liu, Wen-Cheng and Liu, Chian-Yi , TITLE =. Water , VOLUME =. 2020 , NUMBER =

    work page 2020

  48. [48]

    Journal of Marine Science and Engineering , VOLUME =

    Wei, Zhangping and Nguyen, Hai Cong , TITLE =. Journal of Marine Science and Engineering , VOLUME =. 2022 , NUMBER =

    work page 2022

  49. [49]

    Storm surge modeling in the AI era: Using LSTM-based machine learning for enhancing forecasting accuracy , journal =

    Stefanos Giaremis and Noujoud Nader and Clint Dawson and Carola Kaiser and Efstratios Nikidis and Hartmut Kaiser , keywords =. Storm surge modeling in the AI era: Using LSTM-based machine learning for enhancing forecasting accuracy , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2024.104532 , url =

    work page doi:10.1016/j.coastaleng.2024.104532 2024

  50. [50]

    Bias correction of operational storm surge forecasts using Neural Networks , journal =

    Paulina Tedesco and Jean Rabault and Martin Lilleeng Sætra and Nils Melsom Kristensen and Ole Johan Aarnes and Øyvind Breivik and Cecilie Mauritzen and Øyvind Sætra , keywords =. Bias correction of operational storm surge forecasts using Neural Networks , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.ocemod.2024.102334 , url =

    work page doi:10.1016/j.ocemod.2024.102334 2024

  51. [51]

    Journal of Geophysical Research: Machine Learning and Computation , volume =

    Nagaraj, Meghana and Rodriguez, Alejandra and Wahl, Thomas , title =. Journal of Geophysical Research: Machine Learning and Computation , volume =. doi:https://doi.org/10.1029/2025JH000650 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2025JH000650 , note =

    work page doi:10.1029/2025jh000650

  52. [52]

    Advancing storm surge forecasting from scarce observation data: A causal-inference based Spatio-Temporal Graph Neural Network approach , journal =

    Wenjun Jiang and Jize Zhang and Yuerong Li and Dongqin Zhang and Gang Hu and Huanxiang Gao and Zhongdong Duan , keywords =. Advancing storm surge forecasting from scarce observation data: A causal-inference based Spatio-Temporal Graph Neural Network approach , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2024.104512 , url =

    work page doi:10.1016/j.coastaleng.2024.104512 2024

  53. [53]

    2026 , eprint=

    Flo: A data-driven limited-area storm surge model , author=. 2026 , eprint=

    work page 2026

  54. [54]

    Journal of Advances in Modeling Earth Systems , volume=

    Simulations for CMIP6 with the AWI climate model AWI-CM-1-1 , author=. Journal of Advances in Modeling Earth Systems , volume=. 2020 , publisher=

    work page 2020

  55. [55]

    Geoscientific Model Development , volume=

    The Beijing climate center climate system model (BCC-CSM): The main progress from CMIP5 to CMIP6 , author=. Geoscientific Model Development , volume=. 2019 , publisher=

    work page 2019

  56. [56]

    0.3) , author=

    The Canadian earth system model version 5 (CanESM5. 0.3) , author=. Geoscientific Model Development , volume=. 2019 , publisher=

    work page 2019

  57. [57]

    Earth System Dynamics , volume=

    Simulation of observed climate changes in 1850--2014 with climate model INM-CM5 , author=. Earth System Dynamics , volume=. 2018 , publisher=

    work page 2014

  58. [58]

    Geoscientific Model Development , volume=

    Replicability of the EC-Earth3 Earth system model under a change in computing environment , author=. Geoscientific Model Development , volume=. 2020 , publisher=

    work page 2020

  59. [59]

    Journal of Advances in Modeling Earth Systems , volume=

    Implementation of the CMIP6 forcing data in the IPSL-CM6A-LR model , author=. Journal of Advances in Modeling Earth Systems , volume=. 2020 , publisher=

    work page 2020

  60. [60]

    Geoscientific Model Development , volume=

    Description and basic evaluation of simulated mean state, internal variability, and climate sensitivity in MIROC6 , author=. Geoscientific Model Development , volume=. 2019 , publisher=

    work page 2019

  61. [61]

    2) for the high-resolution model intercomparison project (HighResMIP) , author=

    Max planck institute earth system model (MPI-ESM1. 2) for the high-resolution model intercomparison project (HighResMIP) , author=. Geoscientific Model Development , volume=. 2019 , publisher=

    work page 2019

  62. [62]

    0: Description and basic evaluation of the physical component , author=

    The Meteorological Research Institute Earth System Model version 2.0, MRI-ESM2. 0: Description and basic evaluation of the physical component , author=. Journal of the Meteorological Society of Japan. Ser. II , volume=. 2019 , publisher=

    work page 2019

  63. [63]

    and Taflandis, A

    Toofani Movaghar, P. and Taflandis, A. A. and Nadal-Caraballo, N. C. and Zhang, J. , title =. Journal of Geophysical Research: Machine Learning and Computation , volume =. doi:https://doi.org/10.1029/2025JH001032 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2025JH001032 , note =

    work page doi:10.1029/2025jh001032

  64. [64]

    A review of tropical cyclone-generated storm surges: Global data sources, observations, and impacts

    Needham, Hal F. and Keim, Barry D. and Sathiaraj, David , title =. Reviews of Geophysics , volume =. doi:https://doi.org/10.1002/2014RG000477 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2014RG000477 , abstract =

    work page doi:10.1002/2014rg000477

  65. [65]

    La Houille Blanche , volume =

    Xavier Bertin , title =. La Houille Blanche , volume =. 2016 , publisher =. doi:10.1051/lhb/2016020 , URL =

    work page doi:10.1051/lhb/2016020 2016

  66. [66]

    Journal of Geophysical Research: Oceans , volume =

    Qian, Xiaojuan and Hwang, Sooncheol and Son, Sangyoung , title =. Journal of Geophysical Research: Oceans , volume =. doi:https://doi.org/10.1029/2023JC020400 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2023JC020400 , note =

    work page doi:10.1029/2023jc020400

  67. [67]

    Rezuanul and Lee, Chia-Ying and Mandli, Kyle T

    Islam, Md. Rezuanul and Lee, Chia-Ying and Mandli, Kyle T. and Takagi, Hiroshi , title=. Scientific Reports , year=. doi:10.1038/s41598-021-95825-7 , url=

    work page doi:10.1038/s41598-021-95825-7

  68. [68]

    Dietrich and M

    J.C. Dietrich and M. Zijlema and J.J. Westerink and L.H. Holthuijsen and C. Dawson and R.A. Luettich and R.E. Jensen and J.M. Smith and G.S. Stelling and G.W. Stone , keywords =. Modeling hurricane waves and storm surge using integrally-coupled, scalable computations , journal =. 2011 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2010.08.001 , url =

    work page doi:10.1016/j.coastaleng.2010.08.001 2011

  69. [69]

    Dietrich, J. C. and Tanaka, S. and Westerink, J. J. and Dawson, C. N. and Luettich, R. A. and Zijlema, M. and Holthuijsen, L. H. and Smith, J. M. and Westerink, L. G. and Westerink, H. J. , title=. Journal of Scientific Computing , year=. doi:10.1007/s10915-011-9555-6 , url=

    work page doi:10.1007/s10915-011-9555-6

  70. [70]

    Dullaart, J. C. M. and Muis, S. and de Moel, H. and Ward, P. J. and Eilander, D. and Aerts, J. C. J. H. , TITLE =. Natural Hazards and Earth System Sciences , VOLUME =. 2023 , NUMBER =

    work page 2023

  71. [71]

    Bernier, N. B. and Thompson, K. R. , title =. Journal of Geophysical Research: Oceans , volume =. doi:https://doi.org/10.1029/2005JC003168 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2005JC003168 , abstract =

    work page doi:10.1029/2005jc003168

  72. [72]

    and Aerts, Jeroen C

    Muis, Sanne and Verlaan, Martin and Winsemius, Hessel C. and Aerts, Jeroen C. J. H. and Ward, Philip J. , title=. Nature Communications , year=. doi:10.1038/ncomms11969 , url=

    work page doi:10.1038/ncomms11969

  73. [73]

    O'Neill, B. C. and Tebaldi, C. and van Vuuren, D. P. and Eyring, V. and Friedlingstein, P. and Hurtt, G. and Knutti, R. and Kriegler, E. and Lamarque, J.-F. and Lowe, J. and Meehl, G. A. and Moss, R. and Riahi, K. and Sanderson, B. M. , TITLE =. Geoscientific Model Development , VOLUME =. 2016 , NUMBER =

    work page 2016

  74. [74]

    Muis, Sanne and Aerts, Jeroen C. J. H. and Á. Antolínez, José A. and Dullaart, Job C. and Duong, Trang Minh and Erikson, Li and Haarsma, Rein J. and Apecechea, Maialen Irazoqui and Mengel, Matthias and Le Bars, Dewi and O’Neill, Andrea and Ranasinghe, Roshanka and Roberts, Malcolm J. and Verlaan, Martin and Ward, Philip J. and Yan, Kun , title =. Earth's ...

    work page doi:10.1029/2023ef003479

  75. [75]

    Journal of Marine Science and Engineering , VOLUME =

    Qin, Yue and Su, Changyu and Chu, Dongdong and Zhang, Jicai and Song, Jinbao , TITLE =. Journal of Marine Science and Engineering , VOLUME =. 2023 , NUMBER =

    work page 2023

  76. [76]

    Kyprioti and Alexandros A

    Aikaterini P. Kyprioti and Alexandros A. Taflanidis and Norberto C. Nadal-Caraballo and Madison Campbell , keywords =. Storm hazard analysis over extended geospatial grids utilizing surrogate models , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.coastaleng.2021.103855 , url =

    work page doi:10.1016/j.coastaleng.2021.103855 2021

  77. [77]

    2024 , eprint=

    AIFS -- ECMWF's data-driven forecasting system , author=. 2024 , eprint=

    work page 2024

  78. [78]

    Adam and Gao, Xiang and Hodges, Kevin , title =

    Chinta, Sandeep and Schlosser, C. Adam and Gao, Xiang and Hodges, Kevin , title =. Earth's Future , volume =. doi:https://doi.org/10.1029/2024EF005289 , url =. https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2024EF005289 , note =

    work page doi:10.1029/2024ef005289

  79. [79]

    2026 , eprint=

    StormNet: Improving storm surge predictions with a GNN-based spatio-temporal offset forecasting model , author=. 2026 , eprint=

    work page 2026

  80. [80]

    Monthly Weather Review , volume=

    Composite analysis of North Atlantic extratropical cyclones in NCEP--NCAR reanalysis data , author=. Monthly Weather Review , volume=

    work page

Showing first 80 references.