pith. sign in

arxiv: 2605.16332 · v1 · pith:PZSRZDIRnew · submitted 2026-05-05 · 📊 stat.AP

Data-Driven Climate Outage Risk Characterization and Resilience Analysis in Joint Power-Communication Networks

Yoneke Graham , Gelila Webster , Tina Tran , Sohini Roy This is my paper

Pith reviewed 2026-05-20 23:57 UTC · model grok-4.3

classification 📊 stat.AP
keywords climate outagesresilience analysispower-communication networkscascade simulationdata-driven frameworkcoastal risksevere weatherinterdependency
0
0 comments X

The pith

Climate outages in joint power-communication networks cause far larger resilience gaps on coasts, dropping operability to 17.6 percent after cascades.

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

The paper combines analysis of more than 525,000 real-world outage records with simulations of how failures spread through linked power and communication networks. It shows climate-related outages are rising by about 9,100 events each year and hit coastal states with greater severity, while severe weather stands out as the strongest predictor of bad outcomes. When these patterns are turned into four representative failure scenarios and run through cascade models on a standard test grid, coastal extreme weather produces much deeper losses of system function than inland cases. A reader would care because the work demonstrates that simple counts of outages miss the extra damage created when power and communication layers fail together, which matters for deciding where to strengthen infrastructure as climate events increase.

Core claim

Using the EAGLE-I national outage dataset from 2015-2023, the analysis shows climate-related outages increase by roughly 9,100 events per year and impose greater severity on coastal states. An interpretable logistic regression identifies Severe Weather as the dominant predictor of severe outage risk. Guided by this, four geographically representative failure scenarios evaluated with MIIM-based cascade simulation on the IEEE 118-bus system with communication overlay reveal that coastal scenarios produce substantially larger resilience gaps, with the Extreme Coastal Severe Weather scenario reducing post-cascade operability to 17.6 percent. Aggregate outage statistics alone underestimate co

What carries the argument

MIIM-based cascade simulation applied to four failure scenarios on the IEEE 118-bus power system with a communication network overlay, after logistic regression on empirical outage data identifies severe weather as the main driver.

If this is right

  • Coastal scenarios produce substantially larger resilience gaps than the inland case.
  • Aggregate outage statistics alone underestimate coastal risk.
  • Cross-layer cascade propagation amplifies geographic damage in ways revealed only through interdependency-aware simulation.

Where Pith is reading between the lines

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

  • The same data-plus-simulation method could be applied to future climate projections to rank specific coastal regions by resilience priority.
  • Adding other coupled systems such as transportation or water networks might show even wider geographic differences in outage impact.

Load-bearing premise

The four geographically representative failure scenarios and the MIIM-based cascade simulation on the IEEE 118-bus system with a communication network overlay accurately reflect real-world dynamics of climate-induced outages in joint power-communication networks.

What would settle it

Repeating the cascade simulations on actual regional grid data with verified communication dependencies and finding that the coastal versus inland operability gap disappears or reverses under the same weather inputs.

Figures

Figures reproduced from arXiv: 2605.16332 by Gelila Webster, Sohini Roy, Tina Tran, Yoneke Graham.

Figure 1
Figure 1. Figure 1: Annual climate-related outage counts in the EAGLE-I dataset (2015–2023), showing a strong upward trend with a peak in 2021. Raw event types were consolidated into standardized categories. In this paper, climate-related outages are defined as the union of Severe Weather, Winter Storm, and Natural Disaster. The main variables used in this section are event category, year, state, duration, and maximum custome… view at source ↗
Figure 2
Figure 2. Figure 2: Top 10 states by climate-related outage count (2015–2023) Climate-related outages account for about 79.9% of all records in the cleaned dataset, indicating that the outage landscape is dominated by weather- and climate-associated disruptions. Among the climate-related categories, Severe Weather is by far the most prevalent, substantially exceeding both Natural Disaster and Winter Storm in total count. The … view at source ↗
Figure 5
Figure 5. Figure 5: C. Results and Interpretation On the test set of 105,072 records, the model achieves an accuracy of 0.664, precision of 0.363, recall of 0.454, and AUC-ROC of 0.634. The full classification report is summarized in Table II. TABLE II. LOGISTIC REGRESSION CLASSIFICATION REPORT Class Precision Recall F1-score Support Non-Severe 0.80 0.73 0.77 78,804 Severe 0.36 0.45 0.40 26,268 Accuracy - - 0.664 105,072 Macr… view at source ↗
Figure 3
Figure 3. Figure 3: ROC curve for the logistic regression model on the test set [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Test-set confusion matrix for the logistic regression model. As shown in [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Initial failure set breakdown by entity type across all four scenarios. B. System Level Operability [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: System operability before and after cascade for all four scenarios, with resilience gaps indicated. The overall resilience outcomes are summarized in Table IV and visualized in [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Post-cascade failed entity counts by type across all four scenarios. E. Integrated Interpretation Taken together, the results of Sections III through VI form a coherent analytical pipeline. Section III established that climate-related outages are increasing, that Severe Weather is the dominant event category, and that coastal states bear a greater severity burden. Section IV then showed that Severe Weather… view at source ↗
read the original abstract

Climate-driven power outages pose a growing threat to U.S. grid reliability, yet empirical outage studies and interdependency-based resilience analyses are rarely integrated. This paper presents a data-driven framework that integrates empirical outage characterization with cascade failure simulation in joint power-communication networks. Using the EAGLE-I national outage dataset (2015-2023, above 525,000 records), we characterize the climate-outage landscape through descriptive analysis and hypothesis testing, finding that climate-related outages increase by roughly 9,100 events per year and impose a significantly greater severity burden on coastal states. An interpretable logistic regression model then identifies the main predictors of severe outage risk, with Severe Weather emerging as the dominant factor. Guided by these findings, we construct four geographically representative failure scenarios and evaluate them using MIIM-based cascade simulation on the IEEE 118-bus system with a communication network overlay. Coastal scenarios produce substantially larger resilience gaps than the inland case, with the Extreme Coastal Severe Weather scenario reducing post-cascade operability to 17.6 percentage. The results show that aggregate outage statistics alone underestimate coastal risk, as cross-layer cascade propagation amplifies geographic damage in ways revealed only through interdependency-aware simulation.

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

2 major / 2 minor

Summary. The manuscript presents a data-driven framework that integrates empirical characterization of climate-related power outages from the EAGLE-I dataset (2015-2023, >525,000 records) with MIIM-based cascade simulations on the IEEE 118-bus power system augmented by a communication network overlay. It reports that climate-related outages increase by roughly 9,100 events per year with greater severity in coastal states, identifies Severe Weather as the dominant predictor via logistic regression, constructs four geographically representative failure scenarios, and concludes that extreme coastal severe weather scenarios reduce post-cascade operability to 17.6% while cross-layer cascades amplify geographic damage beyond aggregate statistics.

Significance. If the central simulation results hold, the work has moderate significance for applied statistics in infrastructure resilience by showing how empirical predictors can inform interdependency modeling and reveal that aggregate outage counts underestimate coastal risks. The use of a large public dataset and a standard benchmark system is a positive for reproducibility, though the geographic mapping step remains a key untested link.

major comments (2)
  1. [Cascade simulation and scenario construction] The headline claim that 'coastal scenarios produce substantially larger resilience gaps' with the Extreme Coastal Severe Weather scenario reducing post-cascade operability to 17.6% rests on four 'geographically representative' failure scenarios. However, the IEEE 118-bus system is a synthetic, geography-agnostic test case with no embedded coordinates, coastal topology, or region-specific vulnerabilities. It is therefore unclear how the logistic regression predictors (e.g., Severe Weather) are mapped onto specific buses/lines to produce topology-driven coastal-inland distinctions rather than arbitrary failure-set choices. This mapping step is load-bearing for the amplification claim and requires explicit validation or sensitivity checks.
  2. [Empirical analysis and logistic regression] The abstract and results sections report quantitative claims (9,100 annual increase, 17.6% operability, logistic regression coefficients) without accompanying details on data cleaning, model validation, cross-validation, error bars, or sensitivity to the choice of the four scenarios. These omissions directly affect the reliability of the central resilience-gap conclusions.
minor comments (2)
  1. [Simulation setup] Clarify the exact definition and implementation of the communication network overlay on the IEEE 118-bus system, including how interdependencies are modeled (e.g., which nodes are coupled and under what failure propagation rules).
  2. [Abstract and results] The phrase '17.6 percentage' should be corrected to '17.6 percent' or '17.6%' for standard statistical reporting.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review of our manuscript. We address each major comment point by point below, with clear indications of planned revisions to improve clarity and transparency.

read point-by-point responses

  1. Referee: [Cascade simulation and scenario construction] The headline claim that 'coastal scenarios produce substantially larger resilience gaps' with the Extreme Coastal Severe Weather scenario reducing post-cascade operability to 17.6% rests on four 'geographically representative' failure scenarios. However, the IEEE 118-bus system is a synthetic, geography-agnostic test case with no embedded coordinates, coastal topology, or region-specific vulnerabilities. It is therefore unclear how the logistic regression predictors (e.g., Severe Weather) are mapped onto specific buses/lines to produce topology-driven coastal-inland distinctions rather than arbitrary failure-set choices. This mapping step is load-bearing for the amplification claim and requires explicit validation or sensitivity checks.

    Authors: We agree that the synthetic IEEE 118-bus system lacks inherent geographic features, making the mapping from empirical predictors to specific failure sets a critical step that requires fuller documentation. The four scenarios were constructed by using the logistic regression results (with Severe Weather as the dominant predictor and coastal states showing higher severity in the EAGLE-I data) to inform differentiated failure initiation rates and sets: coastal scenarios applied elevated probabilities and larger initial failure clusters to a designated subset of buses and lines chosen to reflect higher connectivity and load, while the inland scenario used baseline rates. This approach is intended to illustrate potential amplification under interdependencies rather than claim literal geographic fidelity. We will revise the manuscript to add an explicit 'Scenario Construction' subsection detailing the mapping rationale, the exact criteria and sizes of the failure sets, the probability assignments derived from regression coefficients, and sensitivity analyses across alternative mappings to demonstrate robustness of the 17.6% operability result. revision: yes

  2. Referee: [Empirical analysis and logistic regression] The abstract and results sections report quantitative claims (9,100 annual increase, 17.6% operability, logistic regression coefficients) without accompanying details on data cleaning, model validation, cross-validation, error bars, or sensitivity to the choice of the four scenarios. These omissions directly affect the reliability of the central resilience-gap conclusions.

    Authors: We acknowledge that greater methodological detail is needed to support the quantitative claims. The manuscript will be revised to incorporate a dedicated methods appendix or expanded section covering: data cleaning steps for the EAGLE-I records (filtering for climate-related causes, handling of missing data and duplicates, and temporal aggregation); the full logistic regression specification including all predictors, coefficient estimates with standard errors, and significance levels; validation metrics such as cross-validation procedure, AUC-ROC, and goodness-of-fit tests; confidence intervals or standard errors for key statistics including the reported annual increase of approximately 9,100 events; and sensitivity checks on the four scenarios to show how outcomes vary with parameter choices. These additions will directly address concerns about reliability. revision: yes

Circularity Check

0 steps flagged

No significant circularity; results derive from external data and standard benchmark simulation

full rationale

The paper integrates descriptive analysis and logistic regression on the external EAGLE-I dataset (2015-2023) to identify predictors such as Severe Weather, then constructs four failure scenarios guided by those empirical findings before applying MIIM-based cascade simulation on the independent IEEE 118-bus test system with a communication overlay. The headline quantitative claim (17.6% post-cascade operability under Extreme Coastal Severe Weather) is an output of the simulation step rather than a quantity defined by or fitted to the regression parameters themselves. No self-citations, uniqueness theorems, or ansatzes from prior author work are invoked to force the geographic resilience-gap result by construction, and the benchmark system plus public dataset supply independent content against which the simulation can be evaluated. The derivation chain therefore remains self-contained.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on the representativeness of the chosen test system and simulation method plus the accuracy of the external dataset for climate-outage labeling, without introducing new postulated entities.

free parameters (1)
  • Logistic regression coefficients for outage severity predictors
    The interpretable logistic regression identifies Severe Weather as dominant, implying fitted coefficients that depend on the specific dataset partitioning and variable selection.
axioms (2)
  • domain assumption The EAGLE-I dataset provides unbiased and complete records of climate-related outages across the U.S. from 2015-2023.
    Invoked in the descriptive analysis and hypothesis testing of the 525,000 records to establish the 9,100 annual increase and coastal severity burden.
  • domain assumption The IEEE 118-bus system augmented with a communication overlay and the MIIM cascade model sufficiently approximate real joint power-communication network behavior under climate stress.
    Invoked when constructing the four geographically representative failure scenarios and reporting the 17.6% operability result.

pith-pipeline@v0.9.0 · 5747 in / 1630 out tokens · 46226 ms · 2026-05-20T23:57:33.547762+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

9 extracted references · 9 canonical work pages

  1. [1]

    Analyzing at-scale distribution grid response to extreme temperatures,

    S. Hanif, M. Mukherjee, S. Poudel, M. G. Yu, R. A. Jinsiwale, T. D. Hardy, and H. M. Reeve, “Analyzing at-scale distribution grid response to extreme temperatures,” Applied Energy, vol. 337, p. 120886, May 2023

    work page 2023

  2. [2]

    Revealing growing and disparate vulnerability in the U.S. power system: A spatiotemporal analysis of nationwide outages from 2014 to 2023,

    B. Li, J. Ma, O. A. Omitaomu, and A. Mostafavi, “Revealing growing and disparate vulnerability in the U.S. power system: A spatiotemporal analysis of nationwide outages from 2014 to 2023,” International Journal of Disaster Risk Reduction, vol. 133, p. 105980, 2026

    work page 2014

  3. [3]

    A New Model to Analyze Power and Communication System Intra -and-Inter Dependencies,

    S. Roy, H. Chandrasekaran, A. Pal and A. Sen, "A New Model to Analyze Power and Communication System Intra -and-Inter Dependencies," 2020 IEEE Conference on Technologies for Sustainability (SusTech), Santa Ana, CA, USA, 2020, pp. 1-8

    work page 2020

  4. [4]

    Resilience assessment and planning in power distribution systems: Past and future considerations,

    S. Paul, A. Poudyal, S. Poudel, A. Dubey, and Z. Wang, “Resilience assessment and planning in power distribution systems: Past and future considerations,” Renewable and Sustainable Energy Reviews, vol. 188, p. 113859, 2023

    work page 2023

  5. [5]

    Assessment of Line Outage Prediction Using Ensemble Learning and Gaussian Processes During Extreme Meteorological Events,

    A. Unlu and M. Peña, “Assessment of Line Outage Prediction Using Ensemble Learning and Gaussian Processes During Extreme Meteorological Events,” Wind, vol. 4, no. 4, pp. 342–362, 2024

    work page 2024

  6. [6]

    Deep Learning - Based Weather -Related Power Outage Prediction with Socio - Economic and Power Infrastructure Data,

    X. Wang, N. Fatehi, C. Wang, and M. H. Nazari, “Deep Learning - Based Weather -Related Power Outage Prediction with Socio - Economic and Power Infrastructure Data,” in Proc. IEEE Power & Energy Society General Meeting (PESGM), 2024, pp. 1–5

    work page 2024

  7. [7]

    Machine Learning Model Development to Predict Power Outage Duration (POD): A Case Study for Electric Utilities,

    B. Ghasemkhani, R. A. Kut, R. Yilmaz, D. Birant, Y. A. Arıkök, T. E. Güzelyol, and T. Kut, “Machine Learning Model Development to Predict Power Outage Duration (POD): A Case Study for Electric Utilities,” Sensors, vol. 24, no. 13, p. 4313, 2024

    work page 2024

  8. [8]

    Resilience of Interdependent Communication and Power Distribution Networks against Cascading Failures,

    W. K. Chai, V. Kyritsis, K. V. Katsaros, and G. Pavlou, “Resilience of Interdependent Communication and Power Distribution Networks against Cascading Failures,” in Proc. IFIP Networking Conference , 2016, pp. 37–45

    work page 2016

  9. [9]

    Identification and Mitigation of False Data Injection using Multi State Implicative Interdependency Model (MSIIM) for Smart Grid,

    S. Roy and A. Sen, "Identification and Mitigation of False Data Injection using Multi State Implicative Interdependency Model (MSIIM) for Smart Grid," 2021 IEEE International Conference on Communications Workshops (ICC Workshops) , Montreal, QC, Canada, 2021, pp. 1-6

    work page 2021