Abstract
Nine in ten major outages in the US have been caused by hurricanes. Long-term outage risk is a function of climate change-triggered shifts in hurricane frequency and intensity; yet projections of both remain highly uncertain. However, outage risk models do not account for the epistemic uncertainties in physics-based hurricane projections under climate change, largely due to the extreme computational complexity. Instead they use simple probabilistic assumptions to model such uncertainties. Here, we propose a transparent and efficient framework to, for the first time, bridge the physics-based hurricane projections and intricate outage risk models. We find that uncertainty in projections of the frequency of weaker storms explains over 95% of the uncertainty in outage projections; thus, reducing this uncertainty will greatly improve outage risk management. We also show that the expected annual fraction of affected customers exhibits large variances, warranting the adoption of robust resilience investment strategies and climate-informed regulatory frameworks.
Subject terms: Natural hazards, Energy infrastructure, Climate change
Introduction
Energy infrastructure is the backbone of our digital economy as well as societal health and safety, and uninterrupted access to electricity is essential for sustaining modern life2. Hurricanes are the leading cause of major power outages26 and have been responsible for over $900 billion in damage and thousands of fatalities in the US over the last 40 years30,36. In 2012, for example, Hurricane Sandy left over 8.5 million customers in the U.S. without power and caused over $7 billion in economic loss and 225 fatalities35.
In an effort to minimize the impact of hurricane-induced power outages, several research studies have developed forecast models to provide estimates of the extent of power outages prior to a storm landfall7,29,31,33,34,37. Such short-term forecasts allow utilities to improve their preparation, response and recovery efforts. They also allow emergency management and response agencies to better plan population evacuations and locate emergency shelters12,27,32. However, long term resilience investment decisions in electric power systems require the ability to credibly characterize long-term power outage risks, warranting the inclusion of climate change as a force affecting hurricane activity6.
There exists a significant body of literature on the relationship between climate change and hurricane activity, with most studies reporting a decrease in storm frequency due to climate change20,48,49,53. However, a non-trivial number of studies project an increase in frequency of hurricanes under similar future climate scenarios3,10,56. Different choices of climate models, initial parametrization of the models21, and spatial resolution of the models52 are among the reasons that cause such variability in projections of future hurricane activity. There is also much uncertainty about the projected future hurricane intensity under climate change. While there is consensus about an increase in future storm intensity, the extent of the projected increase varies among different studies20 ranging from 2 to 14%14,20,21,54. Importantly, regional predictions of hurricane activity under climate change depend on the pattern of warming and so may be fundamentally limited on these long timescales20,22. Thus, it is not clear that science will be able to reduce these epistemic uncertainties in the foreseeable future. Moreover, potential economic damages are known to increase with storm frequency and intensity19,39, including at the extremes4.
While many studies have investigated the impacts of climate change on electricity demand and supply28,38,41, very few have examined climate-change driven hurricane risk to electric power distribution systems12,47 which are most vulnerable to sever weather and climate events16. Resilient electric power infrastructure planning in hurricane prone regions hinges on the ability to systematically characterize and efficiently propagate the uncertainties in hurricane activity under climate change. However, existing approaches rely on computationally expensive methods, which limits their potential for adoption by local and regional planners as well as regulating agencies. Moreover, the physics-based projections of hurricane activity under climate change are not integrated into the existing hurricane outage risk models under climate change. Instead, simple probabilistic assumptions are made to perturb hurricane activity.
We address these fundamental gaps by proposing a transparent and efficient approach to, for the first time, link the knowledge provided by physics-based projections of hurricane activity under climate change with electric power distribution infrastructure risk models, providing a pathway for better integration of climate physics within intricate engineering risk models. We do this by coupling the latest synthesis of physics-based climate change projections with a validated power outage forecast model to predict the impact of future hurricane activity on coastal power distribution systems and then attribute the uncertainty in power outage risk to epistemic uncertainty in future frequency and intensity of hurricanes. Specifically, we utilize results from a recently-published study of projections of hurricane activity20 to quantify the uncertainty in these projections and fit parameterized distributions to future intensity and frequency of hurricanes. This allows us to efficiently propagate the uncertainty in projections through a validated hurricane-induced outage model13,47 in order to directly link changes in hurricane activity to outage risk. Particularly, we seek to answer two fundamental questions: (a) is long-term power outage risk driven primarily by climate change-triggered shifts in intensity or frequency?, and (b) what is the distribution of the expected annual outages due to climate change triggered shifts in hurricane frequency and intensity?
We find that hurricane-induced power outage risk under climate change is predominantly driven by the uncertainty in future frequency of hurricanes. Precisely, uncertainty in future projections of the frequency of hurricanes explains over 95% of the total uncertainty in future outage projections. We also show that the expected annual fraction of customers affected by power outage exhibit rather large variances. Given the multi-decadal timescales of power infrastructure planning, this wide range of possible future outcomes highlights the urgent need for a more robust approach for setting and enforcing reliability standards toward effective management of the power system under climate change.
Projected outage risk
To characterize long-term power outage risk, we first create a baseline scenario to represent present-day hurricane impacts on power distribution systems. The baseline scenario is generated by simulating 2000 years of storm-induced outages, using historical data and a validated outage forecast model13. To examine impacts under climate change then, our model takes as input three factors that define changes in hurricane activity and whose distributions are provided in a recent synthesis work20: the overall frequency factor, , the frequency factor for intense storms, , and the intensity factor, . Distinguishing between overall vs. intense storm frequency allows for shifts within the distribution toward higher-intensity storms. These factors represent fractional increases or decreases in their respective components in a future climate state. They are set to one under the baseline scenario and take uncertain values according to climate change projections. Their distributions may be directly estimated from the projections of hurricane activity under climate change. Here, we fit parameterized distributions to the data from the latest synthesis of physics-based climate change projections by Knutson et al.20 (see “Methods” for details). We then leverage an efficient methodology for stochastic computations59 to propagate the uncertainty in hurricane activity through the outage forecast model to evaluate future power outage risk and its range of uncertainties. Finally, we leverage global sensitivity analysis44 to investigate whether the uncertainty in long-term power outage risk is driven primarily by the uncertainty in future intensity or frequency (“Methods”). Our proposed framework is computationally efficient, and it offers a simple and transparent way for quantitative attribution of infrastructure impact risk to key parameters governing changes in hurricane activity. These key advantages render the framework ideal for decision-support tools under climate change and provide a path for integrating climate knowledge within infrastructure impact models to better inform planners, regulators, and policymakers.
To simulate the 2,000 years of storm-induced power outages under the baseline scenario, the frequency and intensity of hurricanes are sampled from historical distributions; i.e., the dashed line distributions in Fig. 1a,b. Landfall location is drawn from the historical record aggregated to 50km stretches of the coastline as shown by the circles in Fig. 1c. Hurricane tracks and wind fields (see thick black line with the gray band in Fig. 1c) are also modeled using historical data and then fed to a validated power outage forecast model13,47 (““Methods”) to estimate the outage risk in terms of the fraction of the population that loses power in each simulated storm (colored map in Fig. 1c). The fractions are finally aggregated across four distinct regions: the Gulf, upper Atlantic, lower Atlantic and Florida. Figure 2 depicts the distributions of the yearly fractions of customers affected by power outage in these four US regions for the baseline scenario. Based on these distributions, we calculate baseline statistics critical for decision making: the expected value, , and the 95th percentile, , (see Supplementary Information for the latter) of the yearly fraction of affected population.
To perturb hurricane activity under climate change, we assign probability distributions to the frequency factor, the frequency factor for very severe storms, and the intensity factor, , and , based on the recently-published synthesis of future hurricane projections for a 2K global mean temperature increase by Knutson et al.20. Figure 3a–c, generated directly from the raw data (Knutson, pers. comm.), depict the variability in these factors for the North Atlantic basin. It can be seen that hurricane frequency is mostly projected to decrease, but the frequency of very intense storms, i.e., category 4 and 5, is expected to increase. Considering the longer upper tail of the distributions, we model each factor using shifted log-normal distributions with the same interquartile ranges as those reported in Knutson et al.20.
Given the probability distributions of the frequency factor, the frequency factor for intense storms, and the intensity factor, we obtain the probability distribution of expected yearly fraction of affected customers through building a Polynomial Chaos surrogate11,59 (“Methods”). Figure 4a shows the median, the interquartile range, and the 5th and 95th percentiles for the change in the expected yearly fraction of affected customers in the four considered regions. It is found that (see Supplementary Information) there is about 60–65% chance that the expected fraction of customers without power will decrease under climate change, in all regions. Importantly, the changes in the expected yearly fraction of affected customers for all regions exhibit rather large variances, with 90% confidence intervals that span a wide range from large (more than 30%) decreases to large (about 40%) increases. This large variability in the expected yearly fractions of impacted customers makes power system resilience planning and management especially challenging. Can we understand what causes such a large uncertainty?
To answer this question, we next investigate whether this long-term power outage risk is driven primarily by the uncertainty in future intensity or frequency. This is done through a global sensitivity analysis using Sobol’ sensitively indices45. The Sobol’ indices, illustrated in Fig. 4b, represent portions of variance of expected yearly impacted customers attributed to the uncertainty in frequency factor, the frequency factor for very severe storms, and the intensity factor. It can be seen that, in all four regions, more than 95% of variability in the expected yearly fraction of customers that experience power outage is caused by the uncertainty in frequency of non-intense tropical storms. Additionally, in all regions, more than 97% of observed variability in Fig. 4a is due to the uncertainty in future frequency of storms, i.e., both non-intense and intense storms. On the other hand, uncertainty in future intensity of storms barely contributes to the variability in the expected fraction of impacted costumers.
The non-significant contribution of uncertainty in the intensity of storms to power outage risk can be explained by the small variance of the change in future storm intensity. In fact, it is evident in Fig. 3a that the range of projected change in hurricane intensity in North Atlantic basin is relatively small. On the other hand, that same figure shows that the variability in projected climate change-triggered changes in the frequency of non-intense and intense storms is large. Yet, the variability in projected frequency of intense storms contributes to less than 3% of variability in power outage risk in all regions. This can be explained by the small frequency of intense storms for the baseline scenario with an average of 0.12 per year compared to 4.19 for non-intense storms. An extreme scenario of a 100% increase in the frequency of future storms, for instance, translates to average frequencies of 0.24 and 8.38 storms per year for intense and non-intense storms, respectively. Consequently, it is mainly the variability in future frequency of non-intense hurricanes which drives the uncertainty in the expected faction of affected customers. This suggests that more precise projections of the frequency of non-intense hurricanes are crucial to arrive at significantly improved predictions of hurricane-induced power outage risk under climate change.
Discussion
Hurricanes are among the most costly and devastating natural hazards, wreaking havoc on the built environment each year. Climate change is projected to shift hurricane activity. However, there is little consensus on the degree of shifts in hurricane activity among different climate models and it is not clear whether science will be able to reduce this epistemic uncertainty anytime soon. These uncertainties may render the current long-term infrastructure resilience investments sub-optimal under climate change. Moreover, existing approaches for characterizing uncertainty in projected risk are computationally expensive “black boxes” that are not practical for use in long-term planning. In addition, there is often a disconnect between the physics-based climate projections and infrastructure risk models. We propose a novel and yet computationally cheap framework to propagate the uncertainties in climate projections to infrastructure risk models and to characterize the primary driver of uncertainty in hurricane-induced power outage risk. Specifically, we propagate the uncertainty in projection of future hurricane activity defined by a recently-published synthesis study20 to power outage risk under climate change using a validated state-of-the-art power outage model. While our uncertainty-informed analysis reveals a modest chance of decrease in the expected yearly fraction of affected customers under climate change, we also find a very wide range of outcomes, from more than 30% decrease to about 40% increase, which translates to significant risk for long-term power grid planning. Specifically, this variability needs to be accounted for in power distribution systems contingency planning in terms of both emergency operations and systems hardening. It is also of critical importance for utility regulatory commissions that tend to set reliability standards based on historical system performances35.
We exploit the simplicity of our framework to demonstrate the uncertainty in long-term service disruptions in the power distribution systems is primarily driven by the uncertainty in projection of frequency of non-intense hurricanes. Our proposed framework can be used by electric utility planners and policymakers to make uncertainty-informed resilient investment decisions. It can also provide insights to climate scientists on the practical implications of uncertainty in climate change projections for power systems management. In particular, the results suggest that efforts that can reduce the uncertainty in predictions of future frequency of hurricanes of all intensities more accurately on the basin scale, or ideally the regional scale, will substantially improve the predictions of hurricane-induced power outage risks under climate change.
We finally note that the study presented here can be extended to account for potential changes in distribution of landfall locations for a more comprehensive analysis of power outage risk under climate change24. Additionally, the statistical relationships in the outage model used here can be coupled with a simulation model to include grid-hardening measures such as undergrounding power lines. By formulating outage models based on alternative grid designs or topologies, our framework would allow for considering technological changes, such as an increased adoption of renewable energy or distributed generation, in evaluating long term power outage risks.
Methods
Baseline scenario: historical data and the simulation framework
The historical data and simulation framework used to generate the baseline scenario in this study are based on the work by Staid et al.47; below we provide a summary, and refer the reader to the original text for details.
Storm seeding
We rely on the historical record of hurricanes making landfall in the continental United States to determine the baseline behavior of hurricanes. Based on storms occurring between 1851 and 2018, we fit a Poisson distribution to annual frequency of storms making landfall and build empirical distributions for the location of landfall and the storm intensity using the HURDAT2 database25,50. For each replicated year, a realization is drawn from the distribution of annual frequency to determine the number of storms occurring. For each simulated storm within a given year, sample realizations are drawn from the distribution of historical landfall locations and distribution of maximum windspeed at landfall. These samples determine where the storm makes landfall and how strong it is when it does.
Storm track model
For each simulated storm, we create a track using a Random Forest model trained on historical storm movement in the same region. For example, for landfall locations in the Gulf of Mexico we rely on historical storms along the Gulf Coast to train the Random Forest model. This ensures that the storm movement follows realistic weather patterns. The track is created concurrently with simulated storm decay. For each six-hour time step following landfall, the maximum wind speed decays according to the model proposed by Kaplan and Demaria18. This proceeds until the maximum wind speed falls below the threshold for tropical storm classification (1-min sustained winds of 34 knots).
Wind field model
With the storm track and forward moving speed fully specified, we then employ a wind field model to calculate the expected distribution of winds for the areas exposed to the storm. To characterize the wind field in this study, we adopt a model proposed by Willoughby et al.58 and use the stormwindmodel R package1 to calculate the storm radius, the maximum 3-s gust wind, and the duration of winds above 20 m/s for the centroid of each census tract within the storm’s radius.
Power outage model
Finally, we use wind parameters calculated via the wind field model as input to a statistical power outage prediction model developed by Guikema et al.13, The model has been trained and validated on historical hurricane-induced power outage events to represent the relationship between wind characteristics and the fraction of customers expected to lose power from a given storm.
Climate-informed projection
To study the impact of climate change on , we introduce three multiplicative factors: the frequency factor, , the frequency factor for very severe storms, , and the intensity factor, . Specifically, to account for the changes in the frequency of storms, we multiply the mean of the Poisson distribution, that is fitted to the historical data and is used to sample the number of storms in each replicated year, by the frequency factor. Since several studies have suggested that the change in frequency of very intense storms shows a different pattern20, we assign a specific multiplicative factor, , to the mean of Poisson distribution for frequency of very intense storms. Lastly, to account for the changes in intensity of storms, we multiply the randomly sampled maximum wind speed for each hurricane by the intensity factor, . Note that , , and are equal to one under baseline scenario and their values under climate change are uncertain. Their uncertainty are evident in Fig. 3, which is reprinted from a synthesis study by knutson et al.20. Considering the heavy tails of distributions for North Atlantic basin, we assign shifted log-normal distributions with the same interquartile ranges depicted in Fig. 3 to represent the variability in , and . We then sample from these log-normal distributions to draw realizations for the frequency factor, the frequency factor for very severe storms, and the intensity factor for each simulation year under climate change. Note also that we do not consider the impact of climate change on storm sizes in our analysis. We do this given that the storm size is not expected to change significantly with climate change23, in line with recent theoretical advances in our understanding of storm size5.
Polynomial chaos surrogate and sensitivity analysis
To facilitate the uncertainty quantification, we approximate as series expansions in orthonormal Hermite polynomials (Polynomial Chaos) of , , and . We use regression to approximate the coefficients of the series expansion and select the polynomial order and number of training samples required for surrogate construction through convergence analysis. Interested readers are referred to11,17,55,59 for more details and theoretical background on Polynomial Chaos surrogates and their computational efficiency as uncertainty quantification engines.
We use one of the most widely used global sensitivity analysis approaches, namely Sobol’ method43,45. In this approach, the variance of system’s response is decomposed to summation of contributions from different inputs to the model and their interaction. The sensitivity of a particular response (say the expected fraction of costumers without power) to different inputs (say the frequency factors) then manifests itself in their associated Sobol’ indices. More specifically, a square-integrable function is expressed as a sum of elementary basis functions15:
1 |
where d is the dimensionality of the input, , and is a constant. It was shown by Sobol46 that the above decomposition is unique if the basis functions in the expansion are orthogonal. Accordingly, a functional decomposition of the variance is available9:
2 |
where , and so on for higher order interactions with the operation of mathematical expectation. Finally, the variance-based sensitivity indices, the so-called Sobol’ indices, read
3 |
It is readily understood that the Sobol’ indices also satisfy
4 |
The Sobol’ indices, therefore, specify the contribution to the total variance, , from given inputs or input combinations. Monte Carlo sampling based approaches can also be used to evaluate the Sobol’ indices42. However, these approaches can be computationally costly due to the small convergence rate of Monte Carlo sampling particularly in the case of this study when the sought statistic is that of a expected value which itself is based on upwards of tens of hundreds of “simulated years”. Alternatively, Sobol’ indices can be calculated with minimal computational cost when is approximated as series expansion of orthonoromal polynomials. For example, in Eq. (3), can be readily calculated as sum of squares of the coefficients associated with the subset of the polynomial basis functions which are functions of only. Similarly, can be calculated as summation of squared coefficients of all polynomial basis functions with order equal to or greater than one.
Supplementary information
Acknowledgements
The authors would like to acknowledge Purdue Climate Change Research Center (PCCRC), the Center for Environment (C4E) as well as the NSF Grants CRISP-1832688, CMMI-1826161 and CMMI-1826155. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.
Author contributions
D.C., R.N, A.L., and M.T. conceived the idea. A.S., D.C., R.N, A.L., and M.T. supervised the study. N.A. conducted the uncertainty modeling and propagation as well as the sensitivity analysis and led the write-up of the manuscript. B.R. led the data collection and processing as well as hurricane impact simulation runs. All authors reviewed and edited the manuscript.
Data availability
The simulation code and data are available upon request.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Roshanak Nateghi and Mazdak Tootkaboni.
Contributor Information
Roshanak Nateghi, Email: rnateghi@purdue.edu.
Mazdak Tootkaboni, Email: mtootkaboni@umassd.edu.
Supplementary information
is available for this paper at 10.1038/s41598-020-72207-z.
References
- 1.Anderson, B., Schumacher, A., Guikema, S., Quiring, S., & Ferreri, J.. Stormwindmodel: Model Tropical Cyclone Wind Speeds, (2018). R package version 0.1.1. https://CRAN.R-project.org/package=stormwindmodel.
- 2.Anderson K. When the lights go out. Nat. Energy. 2020;5(3):189–190. [Google Scholar]
- 3.Camargo SJ. Global and regional aspects of tropical cyclone activity in the CMIP5 models. J. Clim. 2013;26(24):9880–9902. [Google Scholar]
- 4.Chavas D, Yonekura E, Karamperidou C, Cavanaugh N, Serafin K. US hurricanes and economic damage: Extreme value perspective. Nat. Hazards Rev. 2013;14(4):237–246. [Google Scholar]
- 5.Chavas DR, Reed KA. Dynamical aquaplanet experiments with uniform thermal forcing: System dynamics and implications for tropical cyclone genesis and size. J. Atmos. Sci. 2019;76(8):2257–2274. [Google Scholar]
- 6.Contento A, Xu H, Gardoni P. Climate Adaptation Engineering. New York: Elsevier; 2019. Risk analysis for hurricanes accounting for the effects of climate change; pp. 39–72. [Google Scholar]
- 7.Darestani YM, Shafieezadeh A, DesRoches R. Effects of adjacent spans and correlated failure events on system-level hurricane reliability of power distribution lines. IEEE Trans. Power Deliv. 2017;33(5):2305–2314. [Google Scholar]
- 8.Di Lorenzo, P. usmap: US Maps Including Alaska and Hawaii (2019). R package version 0.5.0. https://CRAN.R-project.org/package=usmap.
- 9.Efron B, Stein C. The jackknife estimate of variance. Ann. Stat. 1981;20:586–596. [Google Scholar]
- 10.Emanuel KA. Downscaling CMIP5 climate models shows increased tropical cyclone activity over the 21st century. Proc. Nat. Acad. Sci. 2013;110(30):12219–12224. doi: 10.1073/pnas.1301293110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ghanem RG, Spanos PD. Stochastic Finite Elements: A Spectral Approach. North Chelmsford: Courier Corporation; 2003. [Google Scholar]
- 12.Guikema, S., & Nateghi, R. Modeling power outage risk from natural hazards. In Oxford Research Encyclopedia of Natural Hazard Science. (2018).
- 13.Guikema SD, Nateghi R, Quiring SM, Staid A, Reilly AC, Gao M. Predicting hurricane power outages to support storm response planning. IEEE Access. 2014;2:1364–1373. [Google Scholar]
- 14.Hill KA, Lackmann GM. The impact of future climate change on TC intensity and structure: A downscaling approach. J. Clim. 2011;24(17):4644–4661. [Google Scholar]
- 15.Hoeffding W. Breakthroughs in Statistics. Berlin: Springer; 1992. A class of statistics with asymptotically normal distribution; pp. 308–334. [Google Scholar]
- 16.Ji C, Wei Y, Mei H, Calzada J, Carey M, Church S, Hayes T, Nugent B, Stella G, Wallace M, et al. Large-scale data analysis of power grid resilience across multiple US service regions. Nat. Energy. 2016;1(5):1–8. [Google Scholar]
- 17.Jones BA, Doostan A, Born GH. Nonlinear propagation of orbit uncertainty using non-intrusive polynomial chaos. J. Guid. Control Dyn. 2013;36(2):430–444. [Google Scholar]
- 18.Kaplan J, DeMaria M. A simple empirical model for predicting the decay of tropical cyclone winds after landfall. J. Appl. Meteorol. 1995;34(11):2499–2512. [Google Scholar]
- 19.Klotzbach PJ, Bell MM, Bowen SG, Gibney EJ, Knapp KR, Schreck I, Carl J. Surface pressure a more skillful predictor of normalized hurricane damage than maximum sustained wind. Bull. Am. Meteorol. Soc. 2020;20:20. [Google Scholar]
- 20.Knutson T, Camargo SJ, Chan JCL, Emanuel K, Ho C-H, Kossin J, Mohapatra M, Satoh M, Sugi M, Walsh K, et al. Tropical cyclones and climate change assessment: Part II. Projected response to anthropogenic warming. Bull. Am. Meteorol. Soc. 2019;20:19. [Google Scholar]
- 21.Knutson TR, Tuleya RE. Impact of CO-induced warming on simulated hurricane intensity and precipitation: Sensitivity to the choice of climate model and convective parameterization. J. Clim. 2004;17(18):3477–3495. [Google Scholar]
- 22.Knutson TR, McBride JL, Chan J, Emanuel K, Holland G, Landsea C, Held I, Kossin JP, Srivastava AK, Sugi M. Tropical cyclones and climate change. Nat. Geosci. 2010;3(3):157–163. [Google Scholar]
- 23.Knutson TR, Sirutis JJ, Zhao M, Tuleya RE, Bender M, Vecchi GA, Villarini G, Chavas D. Global projections of intense tropical cyclone activity for the late twenty-first century from dynamical downscaling of cmip5/rcp4. 5 scenarios. J. Clim. 2015;28(18):7203–7224. [Google Scholar]
- 24.Kossin JP, Emanuel KA, Vecchi GA. The poleward migration of the location of tropical cyclone maximum intensity. Nature. 2014;509(7500):349–352. doi: 10.1038/nature13278. [DOI] [PubMed] [Google Scholar]
- 25.Landsea CW, Franklin JL. Atlantic hurricane database uncertainty and presentation of a new database format. Mon. Weather Rev. 2013;141(10):3576–3592. [Google Scholar]
- 26.Lin N. Tropical cyclones and heatwaves. Nat. Clim. Change. 2019;9(8):579–580. [Google Scholar]
- 27.MacKenzie CA, Barker K. Empirical data and regression analysis for estimation of infrastructure resilience with application to electric power outages. J. Infrastruct. Syst. 2013;19(1):25–35. [Google Scholar]
- 28.Maia-Silva D, Kumar R, Nateghi R. The critical role of humidity in modeling summer electricity demand across the United States. Nat. Commun. 2020;11(1):1–8. doi: 10.1038/s41467-020-15393-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Mensah AF, Dueñas-Osorio L. Efficient resilience assessment framework for electric power systems affected by hurricane events. J. Struct. Eng. 2016;142(8):C4015013. [Google Scholar]
- 30.Mukherjee S, Nateghi R, Hastak M. A multi-hazard approach to assess severe weather-induced major power outage risks in the US. Reliabil. Eng. Syst. Saf. 2018;175:283–305. [Google Scholar]
- 31.Nateghi R. Multi-dimensional infrastructure resilience modeling: An application to hurricane-prone electric power distribution systems. IEEE Access. 2018;6:13478–13489. [Google Scholar]
- 32.Nateghi R, Guikema SD, Quiring SM. Comparison and validation of statistical methods for predicting power outage durations in the event of hurricanes. Risk Anal. Int. J. 2011;31(12):1897–1906. doi: 10.1111/j.1539-6924.2011.01618.x. [DOI] [PubMed] [Google Scholar]
- 33.Nateghi R, Guikema S, Quiring SM. Power outage estimation for tropical cyclones: Improved accuracy with simpler models. Risk Anal. 2014;34(6):1069–1078. doi: 10.1111/risa.12131. [DOI] [PubMed] [Google Scholar]
- 34.Nateghi R, Guikema SD, Quiring SM. Forecasting hurricane-induced power outage durations. Nat. Hazards. 2014;74(3):1795–1811. [Google Scholar]
- 35.Nateghi R, Guikema SD, Wu Y, Bayan Bruss C. Critical assessment of the foundations of power transmission and distribution reliability metrics and standards. Risk Anal. 2016;36(1):4–15. doi: 10.1111/risa.12401. [DOI] [PubMed] [Google Scholar]
- 36.NECI. US billion-dollar weather and climate disasters. NOAA National Centers for Environmental Information (NECI) updates this data periodically. https://www.ncdc.noaa.gov/billions. (2019).
- 37.Ouyang M, Duenas-Osorio L. Multi-dimensional hurricane resilience assessment of electric power systems. Struct. Saf. 2014;48:15–24. [Google Scholar]
- 38.Perera ATD, Nik VM, Chen D, Scartezzini J-L, Hong T. Quantifying the impacts of climate change and extreme climate events on energy systems. Nat. Energy. 2020;5(2):150–159. [Google Scholar]
- 39.Pielke RA, Jr, Gratz J, Landsea CW, Collins D, Saunders MA, Musulin R. Normalized hurricane damage in the United States: 1900–2005. Nat. Hazards Rev. 2008;9(1):29–42. [Google Scholar]
- 40.R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna (2020). https://www.R-project.org/.
- 41.Reyna JL, Chester MV. Energy efficiency to reduce residential electricity and natural gas use under climate change. Nat. Commun. 2017;8(1):1–12. doi: 10.1038/ncomms14916. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Saltelli A. Making best use of model evaluations to compute sensitivity indices. Comput. Phys. Commun. 2002;145(2):280–297. [Google Scholar]
- 43.Saltelli A, Tarantola S, Campolongo F, Ratto M. Sensitivity Analysis in Practice: A Guide to Assessing Scientific Models. Oxford: Wiley Online Library; 2004. [Google Scholar]
- 44.Saltelli A, Ratto M, Andres T, Campolongo F, Cariboni J, Gatelli D, Saisana M, Tarantola S. Global Sensitivity Analysis: The Primer. New York: Wiley; 2008. [Google Scholar]
- 45.Sobol IM. Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates. Math. Comput. Simul. 2001;55(1–3):271–280. [Google Scholar]
- 46.Sobol IM, et al. Sensitivity estimates for nonlinear mathematical models. Math. Modell. Comput. Exp. 1993;1(4):407–414. [Google Scholar]
- 47.Staid A, Guikema SD, Nateghi R, Quiring SM, Gao MZ. Simulation of tropical cyclone impacts to the US power system under climate change scenarios. Clim. Change. 2014;127(3–4):535–546. [Google Scholar]
- 48.Sugi M, Murakami H, Yoshimura J. On the mechanism of tropical cyclone frequency changes due to global warming. J. Meteorol. Soc. Jpn. Ser. II. 2012;90:397–408. [Google Scholar]
- 49.Sugi M, Yoshida K, Murakami H. More tropical cyclones in a cooler climate? Geophys. Res. Lett. 2015;42(16):6780–6784. [Google Scholar]
- 50.Trice, T., & Landsea, C. HURDAT: Hurricane Re-Analysis Project. R package version 0.2.3. https://CRAN.R-project.org/package=HURDAT. (2019)
- 51.U.S. Census Bureau. TIGER/Line Shapefiles (Machinereadable Data Files). https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-line-file.html. Accessed 11 Jan 2019 (2019).
- 52.Vecchi GA, Delworth TL, Murakami H, Underwood SD, Wittenberg AT, Zeng F, Zhang W, Baldwin JW, Bhatia KT, Cooke W, et al. Tropical cyclone sensitivities to co 2 doubling: Roles of atmospheric resolution, synoptic variability and background climate changes. Clim. Dyn. 2019;53(9–10):5999–6033. [Google Scholar]
- 53.Walsh KJE, Camargo SJ, Vecchi GA, Daloz AS, Elsner J, Emanuel K, Horn M, Lim Y-K, Roberts M, Patricola C, et al. Hurricanes and climate: The US CLIVAR working group on hurricanes. Bull. Am. Meteorol. Soc. 2015;96(6):997–1017. [Google Scholar]
- 54.Walsh KJE, McBride JL, Klotzbach PJ, Balachandran S, Camargo SJ, Holland G, Knutson TR, Kossin JP, Lee T, Sobel A. Tropical cyclones and climate change. Wiley Interdiscip. Rev. Clim. Change. 2016;7(1):65–89. [Google Scholar]
- 55.Wan X, Em Karniadakis G. Multi-element generalized polynomial chaos for arbitrary probability measures. SIAM J. Sci. Comput. 2006;28(3):901–928. [Google Scholar]
- 56.Wehner M, Reed KA, Stone D, Collins WD, Bacmeister J. Resolution dependence of future tropical cyclone projections of CAM5. 1 in the US CLIVAR hurricane working group idealized configurations. J. Clim. 2015;28(10):3905–3925. [Google Scholar]
- 57.Wickham, H. ggplot2: Elegant Graphics for Data Analysis. Springer, New York (2016). (ISBN 978-3-319-24277-4). https://ggplot2.tidyverse.org.
- 58.Willoughby HE, Darling RWR, Rahn ME. Parametric representation of the primary hurricane vortex. Part II: A new family of sectionally continuous profiles. Mon. Weather Rev. 2006;134(4):1102–1120. [Google Scholar]
- 59.Xiu D. Numerical Methods for Stochastic Computations: A Spectral Method Approach. Princeton: Princeton University Press; 2010. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The simulation code and data are available upon request.