Tropical cyclone-blackout-heatwave compound hazard resilience in a changing climate

Hurricanes (tropical cyclones, TCs) are major initiators of large-scale power system failures in the United States and have left millions without electricity for days in recent events (e.g., Maria 2017, Irma 2017, Harvey 2017, Florence 2018, Sandy 2012). With climate change, TC intensity is projected to increase, potentially worsening outage severity. Concurrently, heat extremes are projected to become more frequent and intense; heatwaves are the leading cause of weather-related mortality, and lack of air conditioning substantially raises fatality risk during heat events. TC-induced blackouts can therefore amplify heatwave impacts. Prior work has suggested TC–heat compound events could affect many more people under warming but often did not incorporate changes in TC climatology or power system reliability. This study investigates how the joint risk of TC-blackout-heatwave compound hazards evolves under climate change. Focusing on Harris County, Texas—a dense, subtropical, TC-prone area—we couple downscaled TC projections, heat index projections, and a physics-based power outage and recovery model to quantify residents’ risk of prolonged heatwaves while without power and to assess resilience and mitigation strategies.

Evidence indicates hurricane intensity will increase with climate change, potentially shifting seasonality as well. Heat extremes beyond human tolerance are projected to become more frequent and severe, and heatwaves are a dominant cause of weather-related mortality, with air conditioning critical for reducing vulnerability. Early research (ref. 27) linked TCs with heatwave impacts and projected larger affected populations under warming but neglected changes in TC climatology and the role of power system reliability and resilience, potentially underestimating risk. Prior empirical analyses have shown power outage distributions are dominated by a small fraction of local disruptions, suggesting inherent network vulnerability. Studies of evacuation, resilience metrics, and infrastructure hardening (including undergrounding) inform adaptation strategies; however, an integrated modeling framework coupling evolving TC and heat hazards with explicit power system failure and recovery has been lacking.

The study integrates climate hazard projections with a physics-based power system model and network analysis.

• TC projection: Use large datasets of synthetic TCs generated by a deterministic-statistical TC model for the Houston area. Historical climate (1981–2000) is based on NCEP reanalysis; future climate (2081–2100) uses six CMIP5 GCMs under RCP8.5 (CCSM4, HadGEM2-ES, IPSL-CM5A-LR, MIROC-5, GFDL-CM3, MRI-CGCM3). Bias-correct storm frequency (relative to NCEP-calibrated 1.5 storms/year) and landfall intensity via CDF quantile mapping; combine GCMs into a single future projection weighted by historical performance. Stochastically resample to generate 10,000 twenty-year simulations for historical and 10,000 for future climates. For each storm, estimate spatial-temporal wind fields using the Holland profile with surface friction and background wind; convert 1-min mean to 3-s gusts.
• Heatwave projection: Compute daily heat index (HI) from near-surface maximum temperature, specific humidity, and surface pressure using NCEP and the six GCMs. Bias-correct future HI by adding historical NCEP–GCM differences (monthly averages interpolated to daily). Account for TC–heatwave interdependence statistically by adding the composite post-TC meteorological anomalies (from observations) to variables used in HI, which reduces HI in the ~5 days after TC passage. Define heatwave as HI > 40.6 °C. Consider compound events when heatwaves start before or on landfall day (conservative).
• Power system outage and recovery modeling: Employ a physics-based model for Harris County’s grid (~1.7 million residents; 551 transmission lines, 23 power plants, 394 substations; ~40,000 low-voltage distribution branches). Apply probabilistic fragility functions to five component types (transmission substations/lines, distribution nodes/lines, local distribution circuits). Update network topology with failures and perform DC power flow on subgrids; shed load to reach steady state. Recovery prioritizes transmission substations/lines and critical facilities, then local distribution networks, with resources allocated per operational strategies. Validate model against Hurricanes Ike (2008) and Harvey (2017) at county and census-tract levels.
• Risk estimation: For each synthetic 20-year simulation, use an agent-based accounting to record each resident’s maximum duration of post-TC blackout and of TC-blackout-heatwave compound hazard. Compute distributions of resident-level exposure durations for historical vs. future climates.
• Network analysis and enhancement strategies: Derive the generalized scaling law between probabilities of local disruptions and their global outage impact across events. Analyze spatial correlation of outage risk with distribution network harmonic mean length at census-tract level. Test three enhancement strategies with varying enhancement rates (fraction of network length protected): random undergrounding of transmission, random undergrounding of distribution, and greedy undergrounding of distribution (iteratively protect root sectors of longest overhead branches near root nodes), with undergrounded segments placed in protective anti-water pipelines.

• Validation: The outage-recovery model reproduces observed outages for Ike and Harvey; estimated 5-day post-event outages: Harvey ~1% (0–1%), Ike ~63% (60–66%).
• Post-TC heatwaves (RCP8.5, 2081–2100 vs. 1981–2000): Probability of a post-TC heatwave lasting >5 days increases from 2.7% (1.8–4.2%) to 20.2% (12.0–31.5%). Probability of >12-day events becomes 7.5% (4.8–11.9%) from near-zero. Relative risk (future/historical) rises with duration: ~5× (3–9) for 1-day, 7× (4–12) for 5-day, and 22× (14–33) for 13-day heatwaves.
• TC-induced blackout risk: Over a 20-year period, residents affected by any TC blackout increase from 50% (historical) to 73% (future). Share of residents facing at least one >5-day post-TC outage rises from 14% to 44%. Worst-case share not affected declines from 12.8% to 2.7%. For 90% worst recovery cases, full repair within ~15 days historically vs. potentially >25 days in future severe TCs.
• Compound hazard risk: Expected share of residents experiencing at least one >5-day TC-blackout-heatwave event over 20 years increases 23×, from 0.8% to 18.2%. This exceeds the product of independent increases for outages and heatwaves, indicating increased co-occurrence of strong TCs and prolonged heatwaves under climate change.
• Spatial heterogeneity: All census tracts’ power outage risk at least doubles; compound hazard risk increases more strongly. Urban core areas (center/lower county) have lower risks than rural/outer areas, linked to substation density and distribution network topology.
• Scaling law: The largest 20% of local disruptions account for 72% (71–75%) of total service interruptions, indicating disproportionate non-local impact from a minority of local failures; results are robust across events.
• Resilience enhancement: Greedy undergrounding of distribution networks is far more effective than random strategies. Undergrounding the first 5% of distribution lines near root nodes reduces expected residents facing >5-day compound hazard from 18.2% to 11.3% (a 6.9% absolute reduction), about ~15× the benefit of random distribution undergrounding (~0.5%) and ~>60× that of random transmission undergrounding (~0.1%).

Findings indicate that climate change can dramatically increase TC-induced blackout risk and, even more, the compounded risk of enduring heatwaves without power. The time-scale dependence of relative risk (peaking around 13-day heatwave durations) challenges current recovery benchmarks (~5+ days) and suggests potential "resonance" with climate-changed hazards, necessitating faster, more robust recovery. Additional time frames relevant to decision-making show rising compound risk: current (2000–2019) 2.2%, near future (2020–2039) 5.1%, mid-century (2040–2059) 6.7%, reaching 18.2% by late-century under RCP8.5. Even holding TC frequency at historical levels, late-century compound risk remains elevated (11.2%), while under RCP2.6 with unchanged TC activity it rises modestly to ~1.0%, implying strong mitigation could largely avoid the risk escalation. Network analyses reveal that a small fraction of local failures drives most outages and that longer local distribution network lengths (common in low-density areas) correlate with higher outage risk, highlighting how urban form influences vulnerability. Greedy undergrounding of strategic distribution segments near root nodes offers substantial, cost-efficient risk reduction, consistent with the dominance of distribution-level failures. Despite potential capacity upgrades and backup resources (e.g., generators, solar), outage resilience under TCs remains constrained by physical damage and protracted repair; even 50% capacity increases do not significantly enhance TC resilience in the model. Incorporating worker heat stress suggests that extreme heat can further slow recovery (raising expected affected share from 18.2% to 23.3%), underscoring compounding operational challenges. The methodology and insights likely generalize to subtropical U.S. coastal megacities and support integrated climate-adaptation planning.

By coupling downscaled TC projections, bias-corrected heat index projections, and a physics-based power outage and recovery model, this study quantifies how TC-blackout-heatwave compound hazard risk escalates under climate change. For Harris County, the expected share of residents experiencing at least one >5-day heatwave without power in a 20-year period increases from 0.8% historically to 18.2% by late century under RCP8.5. The co-occurrence of strong TCs and prolonged heatwaves intensifies beyond what independent scaling would suggest. Targeted resilience measures—especially greedy undergrounding of strategic distribution network segments—can substantially and cost-effectively mitigate risk (e.g., 5% undergrounding reduces expected affected residents to 11.3%). The framework provides a basis for unified reliability- and resilience-oriented grid design against both daily operations and extreme events. Future research should extend the compound hazard modeling to include flood-induced outages, detailed operational logistics, socioeconomic dynamics, and coupled economic disruption, enabling refined cost–benefit analyses of adaptation strategies and supporting resilient, sustainable development in TC-prone regions.

Key limitations and uncertainties include: (1) Hazard scope focuses primarily on wind and wind-driven treefall; flooding and debris impacts are not explicitly modeled and can impede early recovery, potentially elevating risk in flood-prone areas. (2) The heatwave–outage compounding is conservatively defined (heatwaves starting before/on landfall) and uses statistical corrections for TC–HI interdependence; complex physical interactions may differ. (3) Distribution network topology and protective devices are simplified (star-like/MST approximations; limited modeling of protective actions); while validated at daily time scales for Ike and Harvey, finer operational details are not captured. (4) Worker heat stress, grid aging, and population–infrastructure scaling can degrade recovery and increase risk; only worker heat constraints were explored in sensitivity (raising late-century compound risk from 18.2% to 23.3%). (5) Assumptions about future grid operation and capacity upgrades (even +50% capacity) show limited impact on TC resilience in this framework; real-world investments and technology changes could alter outcomes. (6) Uncertainty in future TC frequency/intensity and heat extremes remains; although sensitivity analyses suggest qualitatively robust increases, precise magnitudes are uncertain.