An integrated method for quantifying and managing extreme weather risks and liabilities for industrial infrastructure and operations

The physical forces and environmental stressors that occur during extreme weather events place facilities at risk for infrastructure failures, loss of operation and production, and highly impactful chemical releases, all of which directly affect a company's bottom line. Hurricane Harvey (2017) resulted in over 100 such failures and chemical releases. There is a pressing need today for risk predictions that incorporate and account for evolving environmental factors such as continuous sea level rise. Such nonstatic (nonstationary) risk management approaches will allow us to more accurately predict storm surge flooding as a function of time and provide more realistic short‐term and long‐term (on the order of decades) predictions to assist in actionable planning. An integrated three‐part approach to assessing the risk of infrastructure damage and chemical releases and the resulting business and legal consequences are presented in this work. This approach consists of (a) temporally variant and spatially localized probabilistic predictions of flooding and forces related to flooding (FloodScore) with unprecedented resolution; (b) detailed impact predictions on facility infrastructure including structural, mechanical, and electrical elements based on the predictions from step (a); and (c) a quantitative means of scoring the environmental/financial risk and consequences of chemicals released as derived from step (b). This integrated approach, which assesses risk of losses in both the near term and out to 50 years in the future, includes the assessment of ecological and human impact levels and provides actionable information for resiliency and risk mitigation planning.

(NOAA) estimated the total losses in 2018 to be $91 billion. 1 The year 2018 is ranked as the fourth highest in total cost, behind years 2017, 2005, and 2012. Additionally, when evaluating data from 1980 to 2018, the annual average number of events is 6.2 (Consumer Price Index [CPI]-adjusted), while the annual average from 2014 to 2018 is 12.6 (CPI-adjusted). While the growing frequency of extreme weather event occurrence is still debatable, experts agree that climate change will only compound the already increasing severity of tropical cyclones, coastal flooding, and wildfires. 1 It is therefore important for corporations and government entities alike to consider the changing, nonstationary conditions when assessing the risk of infrastructure destruction, business interruption, and consequences of chemical releases resulting from extreme weather events. It is this information that can best assist in mitigating these risks.
Weather-related vulnerabilities to multiple sectors, including the energy sector, 2 continue to increase. In response, companies typically adopt one of three approaches to extreme-weather-related risk mitigation. The most common approach builds reactive response plans based on static or stationary data such as historical flood maps and information from past events. 3 While flood maps do provide valuable historical insight, they do not account for other factors such as sea level rise and associated increases in the severity of storm surges that are expected over the next 30 years. A more progressive risk management approach leverages dynamic sets of data that account for both the changing environment and the accelerated pace of data availability. Dynamic data sets include information on rising sea levels, rising ocean surface water temperatures, increasing severity of storms, and changes in land use and habitat. The third and most proactive approach to extreme-weather-related risk management focuses on using nonstationary data in a longer term planning horizon. Companies who subscribe to this approach plan for weather-related occurrences with an agreed-upon risk tolerance and leverage probabilistic predictions to build risk mitigation and resiliency around their facilities. This proactive assessment and mitigation approach can help inform not only the risk management of existing facilities, but also the location of new and future facilities, design specifications, and the degree of resiliency that should be built into the design.
Today's standard of care in developing infrastructure and protecting against the increasing severity of extreme weather events does not explicitly require that nonstationarity be considered. In fact, the development of such new standards and codes may well be a slow process due to its technical and scientific complexity as well as the broad and varied make-up of their stake holders. Furthermore, ambiguity exists on how owners, designers, and engineering procurement and construction contractors should consider the changing environment. On the other hand, recent flooding events and the destruction these events have caused to infrastructure and society is well documented. As a result, new facilities that are designed and built using traditional methods based on stationary weather models may face liability exposure associated with losses caused by extreme weather events. Such claims may allege that the owners, designers, and/or builders knew or should have known of the risks and should have therefore protected against them by exceeding current codes and standards requirements. Such liability exposure constitutes a part of the overall risk exposure.
Companies wishing to more rigorously assess and mitigate extreme-weather-related risk can benefit from an approach wherein detailed weather, engineering, environmental, and health analyses are integrated into a systematic methodology. Sole reliance on information from past occurrences and use of historical data sets can inadvertently result in plans that are ill-equipped to withstand the future environment. Leveraging dynamic data sets instead, which account for future changes, can greatly improve risk management outcomes.
This improvement is especially enhanced when proactive engineering analyses that examine failure modes resulting from extreme weather forces are coupled with assessments of environmental and health risks and consequences of potential chemical releases. To underscore this point, during the 2017 weather event known as Hurricane Harvey, over 100 sites released hazardous pollutants. 4 To properly and rigorously address the risk assessment of infrastructure damage, chemical releases, and business and legal consequences as described above, an integrated three-part approach was developed that consists of (a) time-dependent probabilistic predictions of flooding and forces related to flooding with unprecedented resolution; (b) detailed impact predictions on facility infrastructure including structural, mechanical, and electrical elements resulting from step (a); and (c) a quantitative means of scoring the environmental/financial risk and consequences of chemicals released from these predictions. This integrated and localized approach to determining facility-level asset vulnerabilities, quantifying potential impacts, identifying risk management actions, and implementing risk transfer strategies provides actionable information for resiliency and risk mitigation planning.

| EXTREME WEATHER IMPACTS AND FAILURES
The physical forces and environmental stressors that occur during extreme weather events place facilities at risk for infrastructure fail-

| Electrical substation failure
One of the many dramatically visible effects of Superstorm Sandy was the darkening of the skyline of most of lower Manhattan, which is shown in Figure 1. 5 Flooding at two Con Edison transmission substations at the East River Complex accounted for the outages of 10 electrical distribution networks in lower Manhattan. In addition to the loss of these ten networks, two networks were preemptively deenergized by Con Edison and one network shutdown due to flooding at a substation near the East River in the Seaport area. [6][7][8] The peak water level as measured at the Battery (at the southern tip of lower Manhattan) was 14.06 ft above mean lower water level (MLLW). This peak level was 2 ft higher than the maximum National Weather Service forecast and also exceeded any known historical flood level by several feet. 9 The peak water level observed in the vicinity of the Con Edison East River Complex was 13.8 ft (MLLW), which similarly exceeded the forecasted values, and resulted in several feet of flooding at street level. 10 As Figure 2 shows, the streets became rivers and flood waters overtopped or dislodged the temporary protective measures at the East River Complex that had been installed up to an elevation of 13.6 ft (MLLW). 6,7 Therefore, the damage sustained at the East River Complex was the result of a peak water level only 0.2 ft above the temporary protective measures. Complex. 11 Con Edison reported that the outages at the East River Complex were due to flooding of critical components of the low-voltage protective relay system as well as components of the system that maintains flow of pressurized dielectric oil for insulating feeders. 6,7 Published photographs, such as that shown in Figure 4 10 demonstrate the effects of flood waters on relay equipment. Also during Sandy, failures at the East River Complex produced arcing and explosions that were visible from across the East River in Brooklyn. A sequence of still images from a video of this event is shown in Figure 5. 12 As demonstrated by the Con Edison experience during Sandy, the failure of energized equipment caused catastrophic damage both to equipment that was directly submerged, as well as damage to upstream and downstream equipment that was not directly submerged. The overall damage at the East River Complex was sufficient

| Generator dual tank air vent failure
Another example illustrates how Superstorm Sandy, and the associated flooding, caused an unexpected power outage in a building that had been specifically protected against this scenario. As depicted in Figure 6, emergency generators and a suspended fuel tank were located at the top of a building to eliminate the risk of water damage and maintain electricity during a weather event. In this design, the suspended tank was supplied by pumping fuel from a tank in the basement. Water entered the basement fuel tank through a vent located at the elevation of the "100-year flood" scenario, which was ultimately only slightly below the actual flood water level. When the water that entered the basement fuel tank was pumped to the emergency generators, the generators malfunctioned, causing a power outage. Relocating a vent is among the simpler and easier measures available for preventing flood intrusion. With a proper risk assessment, the risk could have been identified, and the vent moved to the "10, 000-year flood" elevation at minimal cost. Additionally, the notion of a "100-year flood" is only meaningful as an indication of the past. It does not utilize data from the present or use forward-looking probabilities for future flooding events. The present framework is designed to specifically address this inconsistency between the past, present, and future.

| PREDICTION OF FLOODING FROM EXTREME WEATHER EVENTS
As presented in Section 1, the first step in assessing infrastructure risk and potential damage is to develop temporally variant and spatially localized predictions of flooding related to an extreme weather event.
In this section, the development of Jupiter's methodology and modeling workstream is presented to illustrate the real-time (operational) prediction of events.

| Methodology
By definition, predicting compound extreme events requires modeling multiple hazards that can contribute to a peril. In coastal zones, the canonical compound event is a storm surge combined with heavy rain.
Often antecedent conditions, such as nearly-to-completely saturated soils or a high water table, can contribute to the peril as well. While

| Use of fragility curves for complex facilities
For assessing the risk of damage to critical infrastructure and equipment during flooding events, it is necessary to use multiple fragility curves both in series for cascading events and in parallel for events impacting independent systems that are driven by the flood. For example, a piece of electrical equipment or a tank containing a hazardous substance may be behind a levee, in which case fragility curves must be used to estimate the probability of levee failure or overtopping, followed by fragility curves to estimate the probability of damage states to the electrical equipment or rupture of the tank, given that water has breached the levee. The process for creating fragility curves for equipment in water events can be challenging, namely because of the numerous relevant intensity measures and because they must be created for each piece of equipment. Fragility curves are typically created through observations after events occur, such as for electrical substation equipment performance during earthquakes, 18 prior experience with such equipment, or laboratory testing of the equipment at numerous intensities.
In case of the latter, the probability of a damage state at each intensity is observed, and then a curve is fit through the data (ie, a lognormal cumulative distribution function using the method of moments) for each damage state. For larger infrastructure, such as a protective levee where laboratory test data does not exist and is impractical to perform, the fragility curves may be determined through numerical simulation. The following section outlines the use of fragility curves.

| Levee/flood wall breach
Flood protection systems, such as the levees and floodwalls in New Orleans, can provide the illusion of safety, while risk assessments reveal that there is a substantial risk. Figure 10   determined for a given scenario with the likelihood of a specific outcome and is shown in Figure 11. Incorporation of nonstationary data into the vulnerability analysis allows us to determine and assess the evolution of the risk profile at a given site.

| CONCLUSIONS
In the United States, recent extreme weather events have resulted in profound social and economic impacts. These events highlight the importance for corporations and government entities alike to consider a changing, nonstationary set of conditions and to assess, quantify, and mitigate the associated risk of infrastructure destruction, business interruption, and consequences of chemical releases resulting from extreme weather events. Weather-related vulnerabilities to multiple sectors, including the energy sector, 2 continue to increase.
The standard of care in developing infrastructure and protecting against the increasing severity of extreme weather events and more specifically how to address the nonstationary aspect of these events is a subject that is receiving great attention and serious consideration by the public sector, developers and writers of codes and standards, insurance underwriters, EPC companies, and owners. However, the development of new rules can be a slow process due to the technical and scientific complexity of this issue, as well as the broad and varied make-up of the stake holders. In light of the well documented recent flooding events and the destruction, these have caused to infrastructure and society, new facilities that are being designed and built using the traditional methods based on stationary assumptions may face liability caused by extreme weather events. Such claims may allege that the owners, designers, and/or builders knew or should have known of the risks and should have therefore protected against them more than is currently required by current codes and standards. This extreme weather event liability exposure now constitutes part of the overall risk.
The clear path forward for companies wishing to more rigorously assess, quantify, and mitigate extreme weather-related risk is to implement an approach that integrates detailed weather, engineering, environmental, and health analyses into a systematic methodology.
In this article, we presented an integrated three-part approach to assessing risk of infrastructure damage and chemical releases and the resulting business and legal consequences. This includes a nonstationary methodology to quantify the extent of flooding and its related forces, the resulting impact on infrastructure, and a method to quantify environmental and financial risk associated with these predictions. This localized approach to determining facility-level asset vulnerabilities, quantifying potential impacts, identifying risk management actions, and implementing risk transfer strategies provides actionable information for resiliency and risk mitigation planning and management.