An Attention-Based Stochastic Simulator for Multisite Extremes to Evaluate Nonstationary, Cascading Flood Risk

Adam Nayak; Pierre Gentine; Upmanu Lall

arxiv: 2509.14162 · v2 · submitted 2025-09-17 · ⚛️ physics.geo-ph · physics.ao-ph· physics.data-an

An Attention-Based Stochastic Simulator for Multisite Extremes to Evaluate Nonstationary, Cascading Flood Risk

Adam Nayak , Pierre Gentine , Upmanu Lall This is my paper

Pith reviewed 2026-05-18 16:06 UTC · model grok-4.3

classification ⚛️ physics.geo-ph physics.ao-phphysics.data-an

keywords flood risk simulationattention mechanismmultisite extremesnonstationary riskclimate variabilityinsurance portfoliosspatiotemporal coherenceMississippi basin

0 comments

The pith

An attention-based framework simulates multisite flood events that are coherent in space and time and linked to climate variability.

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

The paper introduces a multisite flood simulation framework that combines attention-based analog retrieval with stochastic generation of flood sequences. This approach aims to produce portfolios of floods at over 100 sites in the Mississippi River Basin that respect spatial and temporal dependencies while being conditioned on interannual climate variability. A sympathetic reader would care because current tools do not adequately address flood risks at the interannual to decadal scales relevant to insurance contracts and financial planning. The framework also uses explainable AI and wavelet analysis to link the simulated flood clusters to large-scale climate drivers, making the results physically interpretable.

Core claim

The multisite flood simulation framework produces spatiotemporally coherent flood portfolios conditioned on interannual climate variability, yielding physically interpretable flood clusters for portfolio-scale loss simulation and plausible out-of-sample flood risk catalogs.

What carries the argument

Attention-based analog retrieval paired with stochastic multivariate sequence generation for flood frequency, intensity, and duration at multiple sites.

If this is right

Generates spatiotemporally coherent flood portfolios.
Conditions the simulations on interannual climate variability.
Produces physically interpretable flood clusters linked to climate drivers.
Supplies plausible out-of-sample flood risk catalogs for insurance assessment.

Where Pith is reading between the lines

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

Similar attention-based approaches might be useful for simulating other spatially correlated extremes like droughts.
Combining this with financial models could enable direct simulation of cascading losses in portfolios.
Testing the model on data from other river basins would check its generalizability beyond the Mississippi.

Load-bearing premise

Attention-based analog retrieval combined with stochastic multivariate sequence generation can accurately reproduce nonstationary spatial-temporal flood dependencies across sites without post-hoc tuning.

What would settle it

Observing whether the simulated flood clusters match the spatial patterns and timing of actual historical floods in the Mississippi River Basin over periods not used in training.

read the original abstract

Flood risk is correlated in space and time, challenging insurance systems that rely on diversification across assets. Financial instruments governing flood coverage are typically structured as 1 to 5-year contracts, exposing portfolios to climate-driven risk at interannual-to-decadal scales. Yet existing tools address climate risk either through seasonal forecasts extending only months or multidecadal projections misaligned with fiscal horizons, leaving a critical gap in actionable flood risk simulation. We introduce a multisite flood simulation framework combining attention-based analog retrieval with stochastic generation of multivariate flood frequency, intensity, and duration sequences. Applied to over 100 sites in the Mississippi River Basin, the model produces spatiotemporally coherent flood portfolios conditioned on interannual climate variability. Explainable AI attribution paired with wavelet analysis links simulated clustering to large-scale climate drivers, yielding physically interpretable flood clusters for portfolio-scale loss simulation. The framework provides plausible, out-of-sample flood risk catalogs for interannual-to-decadal insurance risk assessment and financial planning.

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.

Desk Editor's Note private letter to a colleague

This paper combines attention-based analog retrieval with stochastic multivariate generation to target interannual flood risk simulation, but the abstract shows no validation numbers or baselines.

read the letter

The main thing here is that the authors built a framework using attention to pull historical analogs and then stochastically generate sequences of flood frequency, intensity, and duration across many sites, conditioned on interannual climate variability. They apply it to over 100 locations in the Mississippi basin and add wavelet analysis plus explainable AI to tie the resulting clusters back to large-scale drivers. This setup aims to produce coherent flood portfolios for 1-5 year insurance horizons where current tools fall short. What is new is the specific pairing of attention retrieval with stochastic multivariate output for nonstationary multisite extremes, plus the post-hoc interpretability step. The paper does a clear job naming the practical gap between seasonal forecasts and long projections, and the focus on portfolio-scale loss simulation and out-of-sample catalogs is directly useful for risk applications. The soft spots are around evidence. The description supplies no quantitative metrics, error bars, baseline comparisons, or out-of-sample test results, so it is hard to judge whether the attention and stochastic pieces actually reproduce the spatial-temporal dependencies or if the clusters are physically interpretable without extra tuning. The central assumption that the framework captures key nonstationary patterns across sites without missing drivers needs checking against data. This work is for hydrologists, climate risk modelers, and people building tools for financial adaptation. Readers interested in machine learning for extremes or stochastic simulation of cascading events could get value from the technical approach, provided the full results support the claims. The paper shows straightforward engagement with the problem and existing methods, so it qualifies as serious thinking on its own terms. I would send it for peer review so referees can examine the implementation details and any empirical support.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces a multisite flood simulation framework that combines attention-based analog retrieval with stochastic generation of multivariate sequences for flood frequency, intensity, and duration. Applied to over 100 sites in the Mississippi River Basin, the approach generates spatiotemporally coherent flood portfolios conditioned on interannual climate variability. Post-hoc explainable AI attribution and wavelet analysis are used to link simulated flood clusters to large-scale climate drivers, with the goal of producing plausible out-of-sample risk catalogs for interannual-to-decadal insurance and financial planning applications.

Significance. If validated, the framework would address a recognized gap in tools for simulating nonstationary flood risk at time scales matching typical insurance contract lengths. The combination of attention mechanisms for capturing spatial dependencies with stochastic sequence generation and wavelet-based interpretability could enable more realistic portfolio-scale loss modeling under climate variability, offering advantages over purely seasonal or multidecadal approaches.

major comments (2)

[Abstract and Results] Abstract and Results: The central claims that the framework produces 'spatiotemporally coherent flood portfolios' and 'plausible, out-of-sample flood risk catalogs' with 'physically interpretable flood clusters' are not accompanied by any quantitative validation metrics, error bars, baseline comparisons (e.g., against historical catalogs or alternative simulators), or out-of-sample test statistics. This absence directly undermines evaluation of the coherence and physical interpretability assertions.
[Methods] Methods: The description of attention-based analog retrieval combined with stochastic multivariate generation does not specify how nonstationarity is explicitly handled, what loss or objective functions are used during training, or whether post-hoc tuning is required to match observed spatial-temporal dependencies. Without these details, it is unclear whether the weakest assumption—that the method accurately reproduces dependencies across sites without missing key drivers—holds.

minor comments (1)

[Abstract] The abstract would benefit from a concise parenthetical definition or citation for 'attention-based analog retrieval' to improve accessibility for readers outside the immediate subfield.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify how to strengthen the presentation of our work. We respond to each major comment below.

read point-by-point responses

Referee: [Abstract and Results] Abstract and Results: The central claims that the framework produces 'spatiotemporally coherent flood portfolios' and 'plausible, out-of-sample flood risk catalogs' with 'physically interpretable flood clusters' are not accompanied by any quantitative validation metrics, error bars, baseline comparisons (e.g., against historical catalogs or alternative simulators), or out-of-sample test statistics. This absence directly undermines evaluation of the coherence and physical interpretability assertions.

Authors: We agree that the abstract and results summary would be strengthened by explicit quantitative support. The manuscript contains visual and qualitative demonstrations of spatiotemporal coherence and physical interpretability, but we acknowledge that numerical metrics, error bars, baseline comparisons, and out-of-sample statistics are not reported alongside the central claims. We will revise the Results section and abstract to add quantitative validation metrics (including correlation-based measures of spatial coherence, error statistics on flood characteristics, and comparisons to a baseline independent-site simulator) together with error bars from ensemble runs and explicit out-of-sample test statistics. revision: yes
Referee: [Methods] Methods: The description of attention-based analog retrieval combined with stochastic multivariate generation does not specify how nonstationarity is explicitly handled, what loss or objective functions are used during training, or whether post-hoc tuning is required to match observed spatial-temporal dependencies. Without these details, it is unclear whether the weakest assumption—that the method accurately reproduces dependencies across sites without missing key drivers—holds.

Authors: We agree that additional methodological detail is warranted. The manuscript conditions the attention-based retrieval on interannual climate indices to capture nonstationarity, but we will expand the Methods section to explicitly describe how this conditioning is implemented, to state the loss and objective functions used for training the attention and stochastic components, and to clarify that no post-hoc tuning is performed. We will also add discussion and supporting analysis addressing the reproduction of cross-site dependencies and the potential for missing drivers, including sensitivity checks on the attention mechanism. revision: yes

Circularity Check

0 steps flagged

No significant circularity; framework is a self-contained generative method

full rationale

The manuscript introduces a multisite flood simulation framework that combines attention-based analog retrieval with stochastic generation of multivariate sequences, conditioned on interannual climate variability. No equations, fitted parameters, or derivation steps are presented that reduce a claimed prediction or result to the model's own inputs by construction. The central claims rest on the architecture's ability to produce spatiotemporally coherent outputs and on post-hoc wavelet attribution for interpretability, rather than on any self-referential fitting or renaming of known quantities. The approach is framed as a new methodological contribution for out-of-sample risk catalogs, with no load-bearing self-citations or uniqueness theorems invoked to close the logical chain. This constitutes a self-contained modeling pipeline without detectable circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on domain assumptions about the effectiveness of attention mechanisms and stochastic models for extremes; no explicit free parameters or invented entities are named.

axioms (2)

domain assumption Attention-based analog retrieval can identify relevant historical climate states to condition flood simulations on interannual variability.
Invoked as the core mechanism for producing coherent, climate-conditioned flood portfolios.
domain assumption Stochastic generation of multivariate sequences can produce realistic joint distributions of flood frequency, intensity, and duration across sites.
Required for the claim of spatiotemporally coherent portfolios.

pith-pipeline@v0.9.0 · 5712 in / 1539 out tokens · 76341 ms · 2026-05-18T16:06:07.556023+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

58 extracted references · 58 canonical work pages

[1]

Zscheischler, J. et al. A typology of compound weather and climate events. Nature Reviews Earth & Environment 1 , 333–347 (2020)

work page 2020
[2]

Zscheischler, J. et al. Future climate risk from compound events. Nat. Clim. Chang. 8 , 469–477 (2018)

work page 2018
[3]

Tellman, B. et al. Satellite imaging reveals increased proportion of population exposed to ﬂoods. Nature 596 , 80–86 (2021)

work page 2021
[4]

B., Das Bhowmik, R

Kumar, K. B., Das Bhowmik, R. & Mujumdar, P. P. Revising ﬂood return periods by accounting for the co‐occurrence between ﬂoods and their potential drivers. Int. J. Climatol. e8783 (2025)

work page 2025
[5]

& Lall, U

Amonkar, Y., Doss-Gollin, J. & Lall, U. Compound Climate Risk: Diagnosing Clustered Regional Flooding at Inter-Annual and Longer Time Scales. Hydrology 10 , 67 (2023)

work page 2023
[6]

& Lall, U

Hwang, J. & Lall, U. Increasing dam failure risk in the USA due to compound rainfall clusters as climate changes. npj Nat. Hazards 1 , 1–9 (2024)

work page 2024
[7]

& Lall, U

Nayak, A., Gentine, P. & Lall, U. Financial losses associated with US ﬂoods occur with surprisingly frequent, low return period precipitation. Under Review (2025)

work page 2025
[8]

& Lall, U

Jain, S. & Lall, U. Floods in a changing climate: Does the past represent the future? Water Resources Research 37 , (2001)

work page 2001
[9]

Michel-Kerjan, E. O. Catastrophe Economics: The National Flood Insurance Program. J. Econ. Perspect. 24 , 165–186 (2010)

work page 2010
[10]

& Cooke, R

Kousky, C. & Cooke, R. Explaining the failure to insure catastrophic risks. Geneva Pap. Risk Insur. Issues Pract. 37 , 206–227 (2012)

work page 2012
[11]

& Lall, U

Nayak, A., Zhang, M., Gentine, P. & Lall, U. Catastrophic ‘Hyperclustering’ and Recurrent 19 Losses: Diagnosing U.S. Flood Insurance Insolvency Triggers. (2025)

work page 2025
[12]

& Moradkhani, H

Sohrabi, M., Moftakhari, H. & Moradkhani, H. Analyzing compound ﬂooding drivers across the US Gulf Coast states. Geophys. Res. Lett. 52 , e2025GL114769 (2025)

work page 2025
[13]

& Lall, U

Nayak, A., Gentine, P. & Lall, U. A nonstationary stochastic simulator for clustered regional hydroclimatic extremes to Characterize compound ﬂood risk. J. Hydrol. X 25 , 100189 (2024)

work page 2024
[14]

& Lall, U

Bonnafous, L. & Lall, U. Space-time clustering of climate extremes amplify global climate impacts, leading to fat-tailed risk. Nat. Hazards Earth Syst. Sci. 21 , 2277–2284 (2021)

work page 2021
[15]

Slater, L. J. et al. Nonstationary weather and water extremes: a review of methods for their detection, attribution, and management. Hydrol. Earth Syst. Sci. 25 , 3897–3935 (2021)

work page 2021
[16]

Merz, B. et al. Understanding heavy tails of ﬂood peak distributions. Water Resour. Res. 58 , e2021WR030506 (2022)

work page 2022
[17]

J., III, Aiyyer, A

Hwang, J., Schreck, C. J., III, Aiyyer, A. & Sankarasubramanian, A. Understanding the organizing scales of winter ﬂood hydroclimatology and the associated drivers over the coterminous United States. J. Hydrol. X 27 , 100200 (2025)

work page 2025
[18]

Nakamura, J., Lall, U., Kushnir, Y., Robertson, A. W. & Seager, R. Dynamical structure of extreme ﬂoods in the U.s. midwest and the United Kingdom. J. Hydrometeorol. 14 , 485–504 (2013)

work page 2013
[19]

Flood Hydroclimatology

Hirschboeck, K. Flood Hydroclimatology. in Flood Geomorphology vol. 27 (John Wiley & Sons, 1988)

work page 1988
[20]

Tsai, W.-M. et al. Co‐occurring atmospheric features and their contributions to precipitation extremes. J. Geophys. Res. 130 , (2025)

work page 2025
[21]

Bowers, C., Seraﬁn, K. A. & Baker, J. W. Temporal compounding increases economic 20 impacts of atmospheric rivers in California. Sci Adv 10 , eadi7905 (2024)

work page 2024
[22]

& O’Reilly, C

Patterson, M. & O’Reilly, C. H. Climate models struggle to simulate observed North Paciﬁc jet trends, even accounting for tropical paciﬁc sea surface temperature trends. Geophys. Res. Lett. 52 , (2025)

work page 2025
[23]

Seneviratne, S. et al. Changes in climate extremes and their impacts on the natural physical environment. Preprint at https://doi.org/10.7916/D8-6NBT-S431 (2012)

work page doi:10.7916/d8-6nbt-s431 2012
[24]

Allan, R. P. & Soden, B. J. Atmospheric warming and the ampliﬁcation of precipitation extremes. Science 321 , 1481–1484 (2008)

work page 2008
[25]

& Portmann, R

Sun, Y., Solomon, S., Dai, A. & Portmann, R. W. How often does it rain? J. Clim. 19 , 916–934 (2006)

work page 2006
[26]

F., Smith, R

Wehner, M. F., Smith, R. L., Bala, G. & Duﬀy, P. The eﬀect of horizontal resolution on simulation of very extreme US precipitation events in a global atmosphere model. Clim. Dyn. 34 , 241–247 (2010)

work page 2010
[27]

Lima, C. H. R., Lall, U., Troy, T. J. & Devineni, N. A climate informed model for nonstationary ﬂood risk prediction: Application to Negro River at Manaus, Amazonia. J. Hydrol. 522 , 594–602 (2015)

work page 2015
[28]

& Lall, U

Kwon, H.-H., Brown, C., Xu, K. & Lall, U. Seasonal and annual maximum streamﬂow forecasting using climate information: application to the Three Gorges Dam in the Yangtze River basin, China / Prévision d’écoulements saisonnier et maximum annuel à l'aide d'informations climatiques: application au Barrage des Trois Gorges dans le bassin du Fleuve Yangtze, Ch...

work page 2009
[29]

& Lall, U

Kwon, H.-H., Brown, C. & Lall, U. Climate informed ﬂood frequency analysis and prediction in Montana using hierarchical Bayesian modeling. Geophys. Res. Lett. 35 , (2008). 21

work page 2008
[30]

& Lall, U

Sankarasubramanian, A. & Lall, U. Flood quantiles in a changing climate: Seasonal forecasts and causal relations. Water Resour. Res. 39 , (2003)

work page 2003
[31]

Liu, Y., Wright, D. B. & Lorenz, D. J. A nonstationary stochastic rainfall generator conditioned on global climate models for design ﬂood analyses in the Mississippi and other large river basins. Water Resour. Res. 60 , e2023WR036826 (2024)

work page 2024
[32]

& Peleg, N

Cache, T., Bevacqua, E., Zscheischler, J., Müller-Thomy, H. & Peleg, N. Simulating realistic design storms: A joint return period approach. Water Resour. Res. 61 , e2024WR039739 (2025)

work page 2025
[33]

& Lall, U

Najibi, N., Devineni, N. & Lall, U. Compound continental risk of multiple extreme ﬂoods in the United States. Geophys. Res. Lett. 50 , (2023)

work page 2023
[34]

Yu, X., Xu, Y.-P., Guo, Y., Chen, S. & Gu, H. Synchronization frequency analysis and stochastic simulation of multi-site ﬂood ﬂows based on the complicated vine copula structure. Hydrol. Earth Syst. Sci. 29 , 179–214 (2025)

work page 2025
[35]

Maduwantha, P. et al. Generating boundary conditions for compound ﬂood modeling in a probabilistic framework. EGUsphere 1–29 (2025) doi:10.5194/egusphere-2025-1557

work page doi:10.5194/egusphere-2025-1557 2025
[36]

Nearing, G. S. et al. Technical note: Data assimilation and autoregression for using near-real-time streamﬂow observations in long short-term memory networks. Hydrol. Earth Syst. Sci. 26 , 5493–5513 (2022)

work page 2022
[37]

Kratzert, F. et al. Toward improved predictions in ungauged basins: Exploiting the power of machine learning. Water Resour. Res. 55 , 11344–11354 (2019)

work page 2019
[38]

& Silva, A

Kazadi, A., Doss‐Gollin, J., Sebastian, A. & Silva, A. FloodGNN-GRU: a spatio-temporal graph neural network for ﬂood prediction. Environ Data Sci 3 , e21 (2024)

work page 2024
[39]

Sun, Y. Q. et al. Can AI weather models predict out-of-distribution gray swan tropical 22 cyclones? Proc. Natl. Acad. Sci. U. S. A. 122 , e2420914122 (2025)

work page 2025
[40]

Vaswani, A. et al. Attention is all you need. arXiv [cs.CL] (2017)

work page 2017
[41]

& Compo, G

Torrence, C. & Compo, G. P. A Practical Guide to Wavelet Analysis. Bull. Am. Meteorol. Soc. 79 , 61–78 (1998)

work page 1998
[42]

& Yan, Q

Sundararajan, M., Taly, A. & Yan, Q. Axiomatic attribution for deep networks. arXiv [cs.LG] (2017)

work page 2017
[43]

& Boulet, B

Zhang, H., Nie, P., Sun, L. & Boulet, B. Nearest neighbor multivariate time series forecasting. IEEE Trans. Neural Netw. Learn. Syst. 36 , 12606–12618 (2025)

work page 2025
[44]

Cowpertwait, P. S. P. Further developments of the neyman‐scott clustered point process for modeling rainfall. Water Resour. Res. 27 , 1431–1438 (1991)

work page 1991
[45]

& Eagleson, P

Entekhabi, D., Rodriguez-Iturbe, I. & Eagleson, P. S. Probabilistic representation of the temporal rainfall process by a modiﬁed Neyman‐Scott Rectangular Pulses Model: Parameter estimation and validation. Water Resour. Res. 25 , 295–302 (1989)

work page 1989
[46]

Kavvas, M. L. & Delleur, J. W. The stochastic and chronologic structure of rainfall sequences--Application to Indiana. docs.lib.purdue.edu (1975)

work page 1975
[47]

T., Lall, U., Zagona, E

Rajagopalan, B., Erkyihun, S. T., Lall, U., Zagona, E. & Nowak, K. A nonlinear dynamical systems‐based modeling approach for stochastic simulation of streamﬂow and understanding predictability. Water Resour. Res. 55 , 6268–6284 (2019)

work page 2019
[48]

& Khalil, A

Kwon, H.-H., Lall, U. & Khalil, A. F. Stochastic simulation model for nonstationary time series using an autoregressive wavelet decomposition: Applications to rainfall and temperature. Water Resources Research 43 , (2007)

work page 2007
[49]

C., Rajagopalan, B

Nowak, K. C., Rajagopalan, B. & Zagona, E. Wavelet Auto-Regressive Method (WARM) for multi-site streamﬂow simulation of data with non-stationary spectra. J. Hydrol. 410 , 1–12 23 (2011)

work page 2011
[50]

T., Rajagopalan, B., Zagona, E., Lall, U

Erkyihun, S. T., Rajagopalan, B., Zagona, E., Lall, U. & Nowak, K. Wavelet‐based time series bootstrap model for multidecadal streamﬂow simulation using climate indicators. Water Resour. Res. 52 , 4061–4077 (2016)

work page 2016
[51]

R., Stedinger, J

Olsen, J. R., Stedinger, J. R., Matalas, N. C. & Stakhiv, E. Z. Climate Variability and Flood Frequency Estimation for the Upper Mississippi and Lower Missouri Rivers. J. Am. Water Resour. Assoc. 35 , 1509–1523 (1999)

work page 1999
[52]

& Foley, J

Twine, T., Kucharik, C. & Foley, J. Eﬀects of El Niño-Southern Oscillation on the climate, water balance, and streamﬂow of the Mississippi River basin. Journal of Climate 18 , 4840–4861 (2005)

work page 2005
[53]

& Dee, S

Muñoz, S. & Dee, S. El Niño increases the risk of lower Mississippi River ﬂooding. Sci. Rep. 7 , (2017)

work page 2017
[54]

Transformers are Graph Neural Networks

K Joshi, C. Transformers are Graph Neural Networks. arXiv [cs.LG] (2025)

work page 2025
[55]

& Kaheil, Y

Lall, U., Devineni, N. & Kaheil, Y. An Empirical, Nonparametric Simulator for Multivariate Random Variables with Diﬀering Marginal Densities and Nonlinear Dependence with Hydroclimatic Applications. Risk Anal. 36 , 57–73 (2016)

work page 2016
[56]

& Sharma, A

Lall, U. & Sharma, A. A Nearest Neighbor Bootstrap For Resampling Hydrologic Time Series. Water Resour. Res. 32 , 679–693 (1996)

work page 1996
[57]

Homeowners Flock to Last-Resort Insurance Policies

Eaglesham, J. Homeowners Flock to Last-Resort Insurance Policies. The Wall Street Journal (2023)

work page 2023
[58]

NFIP Debt

FEMA. NFIP Debt. https://www.fema.gov/case-study/nﬁp-debt (2024)

work page 2024

[1] [1]

Zscheischler, J. et al. A typology of compound weather and climate events. Nature Reviews Earth & Environment 1 , 333–347 (2020)

work page 2020

[2] [2]

Zscheischler, J. et al. Future climate risk from compound events. Nat. Clim. Chang. 8 , 469–477 (2018)

work page 2018

[3] [3]

Tellman, B. et al. Satellite imaging reveals increased proportion of population exposed to ﬂoods. Nature 596 , 80–86 (2021)

work page 2021

[4] [4]

B., Das Bhowmik, R

Kumar, K. B., Das Bhowmik, R. & Mujumdar, P. P. Revising ﬂood return periods by accounting for the co‐occurrence between ﬂoods and their potential drivers. Int. J. Climatol. e8783 (2025)

work page 2025

[5] [5]

& Lall, U

Amonkar, Y., Doss-Gollin, J. & Lall, U. Compound Climate Risk: Diagnosing Clustered Regional Flooding at Inter-Annual and Longer Time Scales. Hydrology 10 , 67 (2023)

work page 2023

[6] [6]

& Lall, U

Hwang, J. & Lall, U. Increasing dam failure risk in the USA due to compound rainfall clusters as climate changes. npj Nat. Hazards 1 , 1–9 (2024)

work page 2024

[7] [7]

& Lall, U

Nayak, A., Gentine, P. & Lall, U. Financial losses associated with US ﬂoods occur with surprisingly frequent, low return period precipitation. Under Review (2025)

work page 2025

[8] [8]

& Lall, U

Jain, S. & Lall, U. Floods in a changing climate: Does the past represent the future? Water Resources Research 37 , (2001)

work page 2001

[9] [9]

Michel-Kerjan, E. O. Catastrophe Economics: The National Flood Insurance Program. J. Econ. Perspect. 24 , 165–186 (2010)

work page 2010

[10] [10]

& Cooke, R

Kousky, C. & Cooke, R. Explaining the failure to insure catastrophic risks. Geneva Pap. Risk Insur. Issues Pract. 37 , 206–227 (2012)

work page 2012

[11] [11]

& Lall, U

Nayak, A., Zhang, M., Gentine, P. & Lall, U. Catastrophic ‘Hyperclustering’ and Recurrent 19 Losses: Diagnosing U.S. Flood Insurance Insolvency Triggers. (2025)

work page 2025

[12] [12]

& Moradkhani, H

Sohrabi, M., Moftakhari, H. & Moradkhani, H. Analyzing compound ﬂooding drivers across the US Gulf Coast states. Geophys. Res. Lett. 52 , e2025GL114769 (2025)

work page 2025

[13] [13]

& Lall, U

Nayak, A., Gentine, P. & Lall, U. A nonstationary stochastic simulator for clustered regional hydroclimatic extremes to Characterize compound ﬂood risk. J. Hydrol. X 25 , 100189 (2024)

work page 2024

[14] [14]

& Lall, U

Bonnafous, L. & Lall, U. Space-time clustering of climate extremes amplify global climate impacts, leading to fat-tailed risk. Nat. Hazards Earth Syst. Sci. 21 , 2277–2284 (2021)

work page 2021

[15] [15]

Slater, L. J. et al. Nonstationary weather and water extremes: a review of methods for their detection, attribution, and management. Hydrol. Earth Syst. Sci. 25 , 3897–3935 (2021)

work page 2021

[16] [16]

Merz, B. et al. Understanding heavy tails of ﬂood peak distributions. Water Resour. Res. 58 , e2021WR030506 (2022)

work page 2022

[17] [17]

J., III, Aiyyer, A

Hwang, J., Schreck, C. J., III, Aiyyer, A. & Sankarasubramanian, A. Understanding the organizing scales of winter ﬂood hydroclimatology and the associated drivers over the coterminous United States. J. Hydrol. X 27 , 100200 (2025)

work page 2025

[18] [18]

Nakamura, J., Lall, U., Kushnir, Y., Robertson, A. W. & Seager, R. Dynamical structure of extreme ﬂoods in the U.s. midwest and the United Kingdom. J. Hydrometeorol. 14 , 485–504 (2013)

work page 2013

[19] [19]

Flood Hydroclimatology

Hirschboeck, K. Flood Hydroclimatology. in Flood Geomorphology vol. 27 (John Wiley & Sons, 1988)

work page 1988

[20] [20]

Tsai, W.-M. et al. Co‐occurring atmospheric features and their contributions to precipitation extremes. J. Geophys. Res. 130 , (2025)

work page 2025

[21] [21]

Bowers, C., Seraﬁn, K. A. & Baker, J. W. Temporal compounding increases economic 20 impacts of atmospheric rivers in California. Sci Adv 10 , eadi7905 (2024)

work page 2024

[22] [22]

& O’Reilly, C

Patterson, M. & O’Reilly, C. H. Climate models struggle to simulate observed North Paciﬁc jet trends, even accounting for tropical paciﬁc sea surface temperature trends. Geophys. Res. Lett. 52 , (2025)

work page 2025

[23] [23]

Seneviratne, S. et al. Changes in climate extremes and their impacts on the natural physical environment. Preprint at https://doi.org/10.7916/D8-6NBT-S431 (2012)

work page doi:10.7916/d8-6nbt-s431 2012

[24] [24]

Allan, R. P. & Soden, B. J. Atmospheric warming and the ampliﬁcation of precipitation extremes. Science 321 , 1481–1484 (2008)

work page 2008

[25] [25]

& Portmann, R

Sun, Y., Solomon, S., Dai, A. & Portmann, R. W. How often does it rain? J. Clim. 19 , 916–934 (2006)

work page 2006

[26] [26]

F., Smith, R

Wehner, M. F., Smith, R. L., Bala, G. & Duﬀy, P. The eﬀect of horizontal resolution on simulation of very extreme US precipitation events in a global atmosphere model. Clim. Dyn. 34 , 241–247 (2010)

work page 2010

[27] [27]

Lima, C. H. R., Lall, U., Troy, T. J. & Devineni, N. A climate informed model for nonstationary ﬂood risk prediction: Application to Negro River at Manaus, Amazonia. J. Hydrol. 522 , 594–602 (2015)

work page 2015

[28] [28]

& Lall, U

Kwon, H.-H., Brown, C., Xu, K. & Lall, U. Seasonal and annual maximum streamﬂow forecasting using climate information: application to the Three Gorges Dam in the Yangtze River basin, China / Prévision d’écoulements saisonnier et maximum annuel à l'aide d'informations climatiques: application au Barrage des Trois Gorges dans le bassin du Fleuve Yangtze, Ch...

work page 2009

[29] [29]

& Lall, U

Kwon, H.-H., Brown, C. & Lall, U. Climate informed ﬂood frequency analysis and prediction in Montana using hierarchical Bayesian modeling. Geophys. Res. Lett. 35 , (2008). 21

work page 2008

[30] [30]

& Lall, U

Sankarasubramanian, A. & Lall, U. Flood quantiles in a changing climate: Seasonal forecasts and causal relations. Water Resour. Res. 39 , (2003)

work page 2003

[31] [31]

Liu, Y., Wright, D. B. & Lorenz, D. J. A nonstationary stochastic rainfall generator conditioned on global climate models for design ﬂood analyses in the Mississippi and other large river basins. Water Resour. Res. 60 , e2023WR036826 (2024)

work page 2024

[32] [32]

& Peleg, N

Cache, T., Bevacqua, E., Zscheischler, J., Müller-Thomy, H. & Peleg, N. Simulating realistic design storms: A joint return period approach. Water Resour. Res. 61 , e2024WR039739 (2025)

work page 2025

[33] [33]

& Lall, U

Najibi, N., Devineni, N. & Lall, U. Compound continental risk of multiple extreme ﬂoods in the United States. Geophys. Res. Lett. 50 , (2023)

work page 2023

[34] [34]

Yu, X., Xu, Y.-P., Guo, Y., Chen, S. & Gu, H. Synchronization frequency analysis and stochastic simulation of multi-site ﬂood ﬂows based on the complicated vine copula structure. Hydrol. Earth Syst. Sci. 29 , 179–214 (2025)

work page 2025

[35] [35]

Maduwantha, P. et al. Generating boundary conditions for compound ﬂood modeling in a probabilistic framework. EGUsphere 1–29 (2025) doi:10.5194/egusphere-2025-1557

work page doi:10.5194/egusphere-2025-1557 2025

[36] [36]

Nearing, G. S. et al. Technical note: Data assimilation and autoregression for using near-real-time streamﬂow observations in long short-term memory networks. Hydrol. Earth Syst. Sci. 26 , 5493–5513 (2022)

work page 2022

[37] [37]

Kratzert, F. et al. Toward improved predictions in ungauged basins: Exploiting the power of machine learning. Water Resour. Res. 55 , 11344–11354 (2019)

work page 2019

[38] [38]

& Silva, A

Kazadi, A., Doss‐Gollin, J., Sebastian, A. & Silva, A. FloodGNN-GRU: a spatio-temporal graph neural network for ﬂood prediction. Environ Data Sci 3 , e21 (2024)

work page 2024

[39] [39]

Sun, Y. Q. et al. Can AI weather models predict out-of-distribution gray swan tropical 22 cyclones? Proc. Natl. Acad. Sci. U. S. A. 122 , e2420914122 (2025)

work page 2025

[40] [40]

Vaswani, A. et al. Attention is all you need. arXiv [cs.CL] (2017)

work page 2017

[41] [41]

& Compo, G

Torrence, C. & Compo, G. P. A Practical Guide to Wavelet Analysis. Bull. Am. Meteorol. Soc. 79 , 61–78 (1998)

work page 1998

[42] [42]

& Yan, Q

Sundararajan, M., Taly, A. & Yan, Q. Axiomatic attribution for deep networks. arXiv [cs.LG] (2017)

work page 2017

[43] [43]

& Boulet, B

Zhang, H., Nie, P., Sun, L. & Boulet, B. Nearest neighbor multivariate time series forecasting. IEEE Trans. Neural Netw. Learn. Syst. 36 , 12606–12618 (2025)

work page 2025

[44] [44]

Cowpertwait, P. S. P. Further developments of the neyman‐scott clustered point process for modeling rainfall. Water Resour. Res. 27 , 1431–1438 (1991)

work page 1991

[45] [45]

& Eagleson, P

Entekhabi, D., Rodriguez-Iturbe, I. & Eagleson, P. S. Probabilistic representation of the temporal rainfall process by a modiﬁed Neyman‐Scott Rectangular Pulses Model: Parameter estimation and validation. Water Resour. Res. 25 , 295–302 (1989)

work page 1989

[46] [46]

Kavvas, M. L. & Delleur, J. W. The stochastic and chronologic structure of rainfall sequences--Application to Indiana. docs.lib.purdue.edu (1975)

work page 1975

[47] [47]

T., Lall, U., Zagona, E

Rajagopalan, B., Erkyihun, S. T., Lall, U., Zagona, E. & Nowak, K. A nonlinear dynamical systems‐based modeling approach for stochastic simulation of streamﬂow and understanding predictability. Water Resour. Res. 55 , 6268–6284 (2019)

work page 2019

[48] [48]

& Khalil, A

Kwon, H.-H., Lall, U. & Khalil, A. F. Stochastic simulation model for nonstationary time series using an autoregressive wavelet decomposition: Applications to rainfall and temperature. Water Resources Research 43 , (2007)

work page 2007

[49] [49]

C., Rajagopalan, B

Nowak, K. C., Rajagopalan, B. & Zagona, E. Wavelet Auto-Regressive Method (WARM) for multi-site streamﬂow simulation of data with non-stationary spectra. J. Hydrol. 410 , 1–12 23 (2011)

work page 2011

[50] [50]

T., Rajagopalan, B., Zagona, E., Lall, U

Erkyihun, S. T., Rajagopalan, B., Zagona, E., Lall, U. & Nowak, K. Wavelet‐based time series bootstrap model for multidecadal streamﬂow simulation using climate indicators. Water Resour. Res. 52 , 4061–4077 (2016)

work page 2016

[51] [51]

R., Stedinger, J

Olsen, J. R., Stedinger, J. R., Matalas, N. C. & Stakhiv, E. Z. Climate Variability and Flood Frequency Estimation for the Upper Mississippi and Lower Missouri Rivers. J. Am. Water Resour. Assoc. 35 , 1509–1523 (1999)

work page 1999

[52] [52]

& Foley, J

Twine, T., Kucharik, C. & Foley, J. Eﬀects of El Niño-Southern Oscillation on the climate, water balance, and streamﬂow of the Mississippi River basin. Journal of Climate 18 , 4840–4861 (2005)

work page 2005

[53] [53]

& Dee, S

Muñoz, S. & Dee, S. El Niño increases the risk of lower Mississippi River ﬂooding. Sci. Rep. 7 , (2017)

work page 2017

[54] [54]

Transformers are Graph Neural Networks

K Joshi, C. Transformers are Graph Neural Networks. arXiv [cs.LG] (2025)

work page 2025

[55] [55]

& Kaheil, Y

Lall, U., Devineni, N. & Kaheil, Y. An Empirical, Nonparametric Simulator for Multivariate Random Variables with Diﬀering Marginal Densities and Nonlinear Dependence with Hydroclimatic Applications. Risk Anal. 36 , 57–73 (2016)

work page 2016

[56] [56]

& Sharma, A

Lall, U. & Sharma, A. A Nearest Neighbor Bootstrap For Resampling Hydrologic Time Series. Water Resour. Res. 32 , 679–693 (1996)

work page 1996

[57] [57]

Homeowners Flock to Last-Resort Insurance Policies

Eaglesham, J. Homeowners Flock to Last-Resort Insurance Policies. The Wall Street Journal (2023)

work page 2023

[58] [58]

NFIP Debt

FEMA. NFIP Debt. https://www.fema.gov/case-study/nﬁp-debt (2024)

work page 2024