Resilience Analysis and Emergency Response Evaluation for Drinking Water Systems

Safe drinking water is provided to consumers through a utility’s drinking water distribution system (WDS). As an integral part of their communities, water utilities must work to minimize any disruptions to their WDS to ensure continuous, safe water service for their consumers. A WDS’s ability to provide water service can be affected by natural disasters, contaminant spills, cyberattacks, and climate change—for these reasons, water utilities should clearly understand how disasters can affect consumers and vital system components in different parts of their service areas.

Water Service in an Emergency

One measure of WDS resilience is the ability to maintain or return water service to consumers in the event of an emergency. Focusing on this area, utilities can identify appropriate response actions that will minimize the number of consumers affected and the disaster-related effects they might face.

Modeling tools can simulate disasters, quantify system resilience, and evaluate emergency response actions. Simulation results can identify components that are crucial for maintaining WDS operations, and with this information utility staff can prioritize infrastructure repair, develop effective emergency response plans, and ensure backup materials and options are in place.

Emergency response plans play a critical role in minimizing WDS disruption during disasters. Developing effective response actions can reduce long-term damage to a system, maintain water service to consumers for longer, and increase a WDS’s overall system resilience to disasters.

Water utilities can also strive for equitable resilience, which considers both system resilience against disasters as well as socioeconomic vulnerability and uneven access to safe water in the community. Paired with typical resilience analysis, water utilities can gain valuable insight about their systems by also considering equity when they identify and prioritize areas for infrastructure repair, new infrastructure, or additional resources.

Hydraulic Modeling

Hydraulic modeling is useful for understanding water flow, demand, and pressure within a WDS during normal operations as well as under various disaster scenarios. Hydraulic modeling can also assess how new infrastructure could increase service to consumers before and after implementation. In addition, these tools can provide quantitative information to evaluate response actions, which can lead to improved responses that lessen system disruptions and build system resilience. Comparisons between different strategies for mitigating service disruptions can be simulated to determine their effectiveness and further improvements.

The US Environmental Protection Agency’s hydraulic modeling software, EPANET, now in its version 2.2, uses WDS components such as pipes, pumps, valves, tanks, and reservoirs to simulate system hydraulics and water quality and their responses over time to changing conditions (Rossman et al. 2020). The Water Network Tool for Resilience (WNTR) extends the capabilities of EPANET and allows for more complex disaster modeling (EPA 2020, Klise et al. 2020).

Damage to a WDS can take many forms, and WNTR can simulate and analyze both the failure process—i.e., how a system can be damaged, and a utility’s potential response activities. For example, WNTR’s time-specific operational controls can be used to stop, modify, and restart a simulation at any point and apply node-specific parameters to allow for more complex simulations, like pipe breaks during an earthquake; WNTR’s built-in resilience metrics quantify the effects.

System Resilience and Equitable Resilience

Simulation results can help with future infrastructure planning, resource allocation, and more effective emergency response development. Resilience metrics summarize a system’s ability to overcome disruptions or damage, and they can be used to compare emergency response actions when planning for potential disasters. Identifying effective responses before an actual emergency helps minimize water service interruptions and prevent unintended consequences. Examples of emergency response actions include water conservation, backup generators, and building redundancy within the system.

Social, economic, and environmental factors such as consumer demographics, income, and toxic substance exposure can help identify areas of concern within a WDS and highlight those needing additional resources and attention to improve equitable resilience. For instance, underserved neighborhood districts could experience greater effects from a disaster caused by pipe age or poor system design. While high-disparity areas might not directly correlate with high-impact areas caused by a disaster, identifying high-disparity areas can help address injustices in resource allocation or help prioritize areas for infrastructure repair.

The following case study focuses on water service availability (WSA), a resilience metric that is calculated as the ratio of the amount of water received to the amount of water requested (Ostfeld et al. 2002). WSA varies between 0 and 1, where 0 indicates consumers at that location receive no water, and a value of 1 indicates consumers at that location had their demands entirely met. While resilience metrics like WSA can somewhat quantify a WDS’s ability to respond to a disaster, solely focusing on system resilience does not consider other factors that can affect a community.

All scenarios highlighted in the following case study were conducted using WNTR. Specifically, this example shows how to use hydraulic modeling with specific metrics to assess drinking water resilience to infrastructure damage, pump and tank outages, and loss of access to source water. In addition, the example shows how socioeconomic and environmental census tract and other data can be used to identify and prioritize areas for WDS infrastructure renewal or replacement.

Pennsylvania Case Study

A WDS in the US state of Pennsylvania was used to demonstrate how hydraulic resilience modeling tools can simulate disasters, analyze the effects, and evaluate emergency response actions. The utility produces an average of 70 mgd of water, serving 300,000 consumers, and is considered a large system, with more than 74,000 pipes. Four disaster scenarios were simulated with this model:

• Pipe breaks

• Pump outage

• Tank outage

• Loss of source water

The following sections outline how each disaster scenario was simulated and the relevant results.

Pipe Criticality

Pipe criticality analysis determines the number of points of water consumption (junctions) and consumers affected by damage to a single pipe that causes changes in water flow, reduces system pressure, and in extreme cases, prevents water delivery. For this study, pipe criticality analysis was simulated by closing individual pipes completely for 48 hours during a simulation. This is a simple approach to simulating pipe damage as it only affects downstream flow. Pipe damages can also be simulated as pipe bursts or leaks.

Pipes with diameters greater than 7 inches were included in the analysis, for a total of 30,542 pipes. This analysis assumed that pipe isolation valves were located at the end of each pipe. More realistic valve locations in which multiple pipes would need to be closed for a repair are also available within the tool.

Junctions are considered impacted if they experience pressure below a specified minimum threshold. The minimum pressure that determines whether consumers still receive some water was 5 psi, while the required pressure to ensure consumers receive all requested water was 20 psi. If a junction experienced pressure below 20 psi, it was considered impacted since it would no longer receive all of its expected demand.

Figure 1 shows the number of impacted junctions for different pipe diameters, while Figure 2 shows the number of affected consumers for the same pipe diameters. Both figures demonstrate that most of the pipes analyzed in this study did not affect many junctions or consumers, with only 0.17% of all analyzed pipes affecting more than 500 junctions and only 0.27% of all analyzed pipes affecting more than 5,000 consumers. The greatest number of junctions affected by a single pipe break was 9,692, and the greatest number of consumers affected by a single pipe break was 150,248. These results are unsurprising given the hydraulic redundancy in the system, where alternate hydraulic paths are typically available in most areas of the system.

Figure 1. Pipe Criticality Analysis: Junctions (n) Affected on the Basis of Pipe Diameter.

Figure 2. Pipe Criticality Analysis: Consumers (n) Affected on the Basis of Pipe Diameter.

While the trends in Figures 1 and 2 are very similar, the difference in magnitude suggests that pipes of greater diameter affect junctions associated with more consumers. This type of analysis helps utilities identify the pipes in their system that affect their consumers most if they were out of service for an extended time. With this kind of information, utilities can determine whether additional resources should be available to expediate pipe repair.

Repair or replacement strategies might prioritize returning service quickly to the largest number of consumers, but this could leave certain areas undermaintained. Taking socioeconomic and environmental factors into account can help identify additional vulnerable areas and ensure equitable safe drinking water delivery within the WDS.

In this study, three socioeconomic and environmental census tract data categories, pipe criticality consumer impacts, and historical pipe break count data were used to identify potential priority areas to improve equitable resilience. The socioeconomic and environmental census tract data included the count of women and children, average household income, and blood lead levels; these were layered over the WDS map and associated with each pipe.

All factors were normalized on a scale of 0 to 1, where greater values (closer to 1) are higher priorities than lower values (closer to 0). Average household income was an exception in which a value of zero was considered higher priority.

While this case study provides only a single example of how pipe criticality data can be used in conjunction with socioeconomic and environmental census tract data to assess system and equity resilience, results from any disaster analysis can be used for similar purposes. Utilities can combine data sets to identify and prioritize areas in their system for infrastructure repair, new resources, or other beneficial maintenance.

Two approaches were used to identify each pipe’s priority:

• Equal weight: all factors were considered equally important and multiplied by the same weight value.

• Proportional weight: factors were ranked from highest normalized value to lowest and multiplied by different weight values.

As an example of the latter approach, if the blood lead levels associated with a given pipe had a higher normalized value than the average household income, the normalized blood lead level value would be multiplied by a higher weight value than the normalized average household income. Once multiplied, the values for each factor were summed to determine the pipe’s priority.

Between the two approaches, the equal weight approach is useful if a utility wants to consider multiple factors and is not concerned about any of them in particular. However, if a utility is mainly concerned about prioritizing its most critical factor, the proportional weight approach can help identify these areas better. While individual data sets provide utilities with enough information to prioritize areas, taking multiple factors into consideration can help utility leaders make more comprehensive decisions.

Table 1 illustrates how pipe priority can differ when focusing on individual factors or by using the equal weight or the proportional weight approaches. For instance, pipe A is the highest in priority when looking at only consumer impacts (from pipe criticality analysis), but it is the lowest when using the proportional weight approach. This demonstrates how considering multiple factors when planning for maintenance and repairs within the WDS can lead to different prioritization.

Table 1.

Priority Rankings for Three Pipes in the Water Distribution System

Pipe	Single Factor	All Factors
	Consumer impacts	Equal weight	Proportional weight
A	1	2	3
B	3	3	1
C	2	1	2

A ranking of 1 is the highest priority, 3 is the lowest priority.

Parts A and B of Figure 3 show areas of high priority when applying the equal weight and proportional weight approaches, respectively. Using the equal weight approach and comparing overall pipe priorities, Figure 3, part A, shows two higher-priority areas, while the majority of the system is considered lower in priority. However, the proportional weight approach in Figure 3, part B, highlights additional areas within the WDS that could be prioritized depending on the other factors of interest. With either approach, the value of a pipe’s overall priority is informative only when compared with another pipe.

Figure 3. Priority Areas Based on Socioeconomic and Environmental Factors for Equal Weight (A) and Proportional Weight (B) Approaches^a.

^aHigher value corresponds to higher priority

Breakdown analysis can be conducted for individual pipes, pressure zones, or census tracts to understand which factors identify an area as a high priority. For our case study, breakdown analysis was completed at the pipe level, although it is reported on the census tract level.

Figure 4 shows the distribution of all five factors in priority for the equal weight (Figure 4, part A) and the proportional weight (Figure 4, part B) approaches for each census tract in a single pressure zone of the system. Within this pressure zone, the pipe break consumer-impacts factor was zero and thus had no effect on a pipe’s priority, regardless of the approach. Conversely, blood lead levels and women and children counts were typically strong indicators for identifying high-priority areas when using either approach, and these factors were particularly influential when using the proportional weight approach.

Figure 4. Socioeconomic and Environmental Factors Breakdown by Census Tract Using Equal Weight (A) and Proportional Weight (B) Approaches^a.

^aHigher value corresponds to higher priority

If a utility wants to identify high-priority areas while considering multiple factors but is mainly concerned about one, breakdown analysis paired with the proportional weight approach can be particularly insightful. Since the equal weight approach considers all factors the same, it can be difficult to identify the leading factor for a given census tract. For example, average household income appears to contribute equally for census tracts 4 and 5 (Figure 4, part A). Conversely, the proportional weight approach (Figure 4, part B) shows that average household income is the leading factor in census tract 4, while blood lead level is the leading factor in census tract 5. In the event of limited resources or time-sensitive situations, understanding the most influential factors in determining priority can be beneficial.

Pump Outages

Pumps are a vital component within a WDS, and changes in pumping can be affected by power outages, damage, or routine maintenance. Short- and long-term effects of a pump outage can include lack of supplied water and reduced water pressure. If depressurization occurs, system water quality can be compromised, and long-term damage can include an increased rate of pipe breaks.

For this study, pump controls were changed to simulate a power outage at a single pump station (made up of six high-service pumps) starting at hour 24 for 72 hours. An emergency response was simulated in which a backup pump capable of pumping 13 mgd was also brought online during the outage. Average systemwide WSA was used to evaluate the system during the pump outage and pump outage with backup pump scenarios (Figure 5, part A).

Figure 5. Systemwide Water Service Availability for Pump Outage (A), Tank Outage (B), and Loss of Water Source (C) Analyses^a.

WSA—water service availability

^aHigher value corresponds to more demand being met

All three scenarios were simulations with and without emergency response actions.

The system’s average WSA decreased by about 20% within an hour of the pump outage, and it dropped more than 80% from the original level at 56 hours into the pump outage (Figure 5, part A). For the emergency response scenario, the single backup pump was able to slightly improve the overall WSA during the outage, but the system still experienced an 80% decline in WSA 56 hours into the outage. This analysis identified that additional backup pump capacity or other emergency response actions are necessary to prevent major effects on the system from this type of pump outage scenario.

Tank Outages

Changes to water storage tanks can greatly affect water service. A tank might not be able to supply water for a variety of reasons, reducing the available water storage capacity and associated system pressure the tank provides. For instance, a pipe leading in or out of the tank could break or leak, water in the tank could become contaminated, or the tank might need to be closed for routine maintenance. For every event, it is important for water utilities to understand how long they can provide water to consumers without their tanks in service.

Similar to pump outages, emergency response actions can minimize how consumers are affected during tank outages. For this case study, a tank outage was simulated by closing the pipes leading into and out of a 133-million-gallon water storage tank at hour 24 of the simulation and then reopening them after 72 hours. An emergency response was also simulated in which valves to other pressure zones were opened to allow additional water to reach the area normally served by the tank while it was nonoperational. The average systemwide WSA was used to track the system’s response to each scenario (Figure 5, part B).

As the model results show, the emergency response of opening additional connections to other pressure zones was generally effective for the simulated tank outage scenario (Figure 5, part B). Without the additional open connections, WSA decreased by more than 40% at 70 hours into the tank outage. Conversely, the system was able to retain 30% more WSA during the period of highest stress by opening the additional connections. Similarly, additional emergency response actions for the tank outage could be simulated and evaluated.

Loss of Source Water

Changes to a utility’s source water or its ability to withdraw water from existing sources can leave it unable to pump water into the WDS. Access to a source water could be disrupted because of issues with the intake, pump failure, power outages, or water contamination. In the event that access to source water is compromised, a utility has only the water already stored in the system to continue service to all users.

For this case study, a loss of access to the source water was simulated by closing all pipes connected to the reservoir at hour 24 of the simulation for 72 hours, and the effects to the average systemwide WSA were recorded. An emergency response of mandatory 25% water conservation was also simulated, and the WSA results were compared.

Figure 5, part C, shows the WSA for both source-water-loss scenarios. The emergency response of having consumers reduce their water usage by 25% did not result in less impact on WSA during the first 24 hours after source water loss, but it elevated the WSA generally throughout the entire outage period. About 25 hours after the loss of source water, the WSA was 20% higher compared with the case if water conservation efforts had not been in place. However, demand management eventually failed, around 70 hours after the loss of source water (simulation hour 94), as the WSA decreased close to 10% in both scenarios. This decrease suggested that emergency curtailing of consumer demands was effective only for a limited period and that more aggressive demand management efforts (>25%) and other water sources are needed if the original source of supply is inaccessible for longer periods.

Resilience Modeling Analysis Builds Robust Systems

This case study has shown how WDS modeling tools and analysis can be used to assess a drinking water utility’s resilience to various disasters. WNTR was used to simulate how disasters, such as pipe breaks, pump and tank outages, or loss of source water, could affect a WDS. The model was also used to quantify system resilience during simulated disasters. Utilities should note that socioeconomic and environmental data layers can be used to identify high-priority areas within their systems where greater investment could improve equitable resilience.

While this case study reported on the specific resilience analysis results of a Pennsylvania drinking water system, the methods described can be applied to any WDS model to provide insight into a specific system’s resilience. Simulating disasters and their corresponding emergency responses can help utilities understand potential impacts on consumers and evaluate the effectiveness of specific actions.

Resilience modeling tools such as WNTR can assist utilities with these complex disaster scenarios by identifying critical components within their system. Water service availability was used in this case study, but additional metrics include tank capacity and water age. Incorporating this kind of analysis can also help in designing more robust systems in terms of materials and layouts, and it can also be used to plan for available backup and replacement equipment. By employing these tools, drinking water utilities can investigate options to build equitable resilience into their systems.

Key Takeaways.

Modeling tools can simulate possible disasters to water distribution systems, including pipe breaks, power outages, and changes to water source availability.

Resilience analysis can help water utilities compare emergency response strategies for disasters that could affect their water distribution systems.

Socioeconomic and environmental factors are useful to identify priority areas within a water distribution system to improve equitable resilience.