Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Nov 1.
Published in final edited form as: J Water Resour Plan Manag. 2022 Nov;149(1):10.1061/jwrmd5.wreng-5631. doi: 10.1061/jwrmd5.wreng-5631

Performance and Resilience Analysis of a New York Drinking Water System to Localized and System-Wide Emergencies

Lucinda-Joi Chu-Ketterer 1, Regan Murray 2, Patrick Hassett 3, Josh Kogan 4, Katherine Klise 5, Terranna Haxton 6
PMCID: PMC9907455  NIHMSID: NIHMS1853979  PMID: 36776983

Abstract

Drinking water utilities are vulnerable to both human-caused and natural disasters that can impact the system infrastructure and the delivery of potable water to consumers. Analyzing system performance and resilience can help utilities identify areas of high risk or concern, understand the impacts on consumers, and evaluate response actions during disasters. In this case study, the Water Network Tool for Resilience (WNTR) was used to investigate the performance and resilience of a drinking water system in New York during increased demands due to firefighting, pipe damage, and loss of the source water emergencies. This case study introduced a new combined performance index (CPI) resilience metric, which served to quantify system resilience as a ratio of system performance during an emergency to normal operations. The results revealed that this drinking water system was able to maintain service to most of the consumers during these emergencies due to high redundancy within the system, and conservation efforts extended water service for an additional 20 h. The analysis in this paper can be used by other drinking water utilities to understand their vulnerabilities and evaluate resilience-improving actions in similar disaster scenarios.

Introduction

A drinking water distribution system (WDS) is integral to keeping a community healthy and safe by transporting potable water to residential and commercial consumers. Moreover, a WDS supplies water for critical functions within a community, such as fire protection, health care facilities, and the electrical sector. The system of pipes, tanks, valves, pumps, sensors, supervisory control and data acquisition (SCADA) system, and other components that make up a WDS are vulnerable to a range of threats, from the inevitable aging of materials (Minaei et al. 2019) to natural disasters (Klise et al. 2017; Laucelli and Giustolisi 2015), intentional attacks, such as a cyberattack (Marchese et al. 2020), and climate change (Brazil et al. 2017; Swain et al. 2017). Understanding the potential impacts of these threats and increasing a WDS’s resilience to them is essential to maintaining critical service to consumers.

A WDS can experience localized or system-wide emergencies. For example, fires or pipe breaks affect a localized area within the system, whereas source water and treatment plant issues can affect the entire system. Natural disasters are unique threats to WDSs as they can cause both localized and system-wide impacts. For instance, earthquakes can cause fires and damage to pipes (Hecht et al. 2014), intense cold weather or wildfires can cause pipes to burst (Miller 2021; Hughes 2018), and hurricanes can cause flooding that disrupts connectionsto thewater source (Bacon and Madhani 2017). While emergencies range in severity, any disruptions to a WDS can greatly affect a community’s access to clean water for a period of hours or days. For this reason, understanding the impact of emergencies and the resilience of a WDS is of high importance.

Many definitions of resilience for WDSs are provided within the literature. Bruneau et al. (2003) defined the term using the four Rs of resilience: (1) robustness, (2) redundancy, (3) resourcefulness, and (4) rapidity. Others defined it as the ability of a network to withstand short-term disturbances, overcome disruptions, and rebuild after damage (Hwang et al. 2013). In addition to recovering from damage states, Stevenson et al. (2015) included adapting to past emergencies to learn, plan, and develop mitigation strategies for future emergencies in their definition. To summarize, the resilience of a network describes its ability to continue operations during, and to recover from, a damaged state and can always be improved. While general guidance on preparedness and resilience is available to drinking water utilities (AWWA 2010; NIAC 2010; EPA 2016, 2018a, b, 2019), a quantitative site-specific analysis could provide more detailed information on the impacts on the system to help prioritize capital investments to improve resilience. To help address this need, the EPA and their collaborator, Sandia National Laboratories, developed the Water Network Tool for Resilience (WNTR) (EPA 2020; Klise et al. 2020, 2017). WNTR is an open-source Python package based upon EPANET (Rossman et al. 2020) that simulates emergency scenarios of interest to evaluate the impact on system performance.

Modeling is a useful tool to estimate the impacts of an emergency or disaster scenario on a network and evaluate potential operational actions to improve resilience. Previously studied approaches have included investigations of firefighting, pipe breaks, and source water disruption impacts on a WDS. Firefighting analysis helps identify how well a system can handle increased water demand due to fire and understand how the rest of the system will be impacted (Diao et al. 2016). Similarly, pipe break analysis can help identify which pipes within the system would cause the greatest impact if broken and help prioritize pipe replacement. Corrosion, aging infrastructure, changes in pressure, and environmental factors might cause pipes to leak, break, or burst (Srirangarajan et al. 2013). If a pipe is leaking, water can still flow through the pipe but with a lower flow (Lambert et al. 2013; Gong and Zhou 2017), but if it bursts, water can no longer flow through the pipe (Bicik et al. 2009; Cook et al. 2016; Turner et al. 2012). Diao et al. (2016) modeled pipe breaks by completely closing the pipe, thus mimicking the status of a pipe burst. The same approach was adopted for this case study. However, assuming a single pipe closure does not take into consideration the location of isolation valves since it assumes that these valves are located and accessible within each of the pipes. Walski (1993) first introduced the concept of segments to simulate the impacts of pipe closures since segment analysis incorporates a more realistic approach for valve placement. Recent research explored the optimal isolation valve placement within a system (Giustolisi 2020; Hwang and Lansey 2021; Liu et al. 2017). Additional stress faced by a WDS isa disruptionto their sourcewaterand/or treatment plant. Changes towater quality, limited quantity of sourcewater, and intake damage can impact a WDS’s ability to satisfy customer water demand. These could lead to water treatment plant closures or a reduction in operations, which could cause demand deficits within the network. Khatavkar and Mays (2020) researched the impacts of controlling flow by changing pump operations due to limited water in the system. Water conservation guidelines and regulations are another effective way to address reductions in water availability in a system (DECNY 1989, 2022). By reducing consumer water demands, the limited supply of water in the system can continue providing essential water to consumers.

Hydraulic metrics can be used to evaluate WDS performance during normal and damaged states. In this study, metrics are categorized as performance metrics and resilience metrics. Performance metrics are used to describe the performance of a WDS during a simulation. Water utilities use a similar approach through key performance indicators (KPI) to assess their performance across the following five fields: (1) customers, (2) operations, (3) environment, (4) human capital, and (5) corporate governance (Minelli 2021). While not explicitly referred to as such, water service availability (WSA) could be considered a KPI-related metric. WSA quantifies a system’s ability to deliver consumer demand and is calculated as the ratio of delivered demand to expected (requested) demand (Ostfeld et al. 2002).

Resilience metrics are used to quantify the resilience of a WDS during a damage state analysis. The modified resilience index (MRI) was proposed in 2008 (Jayaram and Srinivasan 2008) to quantify the system’s total head and is calculated as the ratio of surplus head to required head (Jeong et al. 2017). While MRI is very similar to the widely used Todini index (Todini 2000; Baños et al. 2011; Tsakiris and Spiliotis 2012; Atkinson et al. 2014; Zhan et al. 2020), it is more accurate when applied to systems with multiple sources. MRI has been used in research compared to other resilience metrics (Baños et al. 2011; Paez et al. 2018).

The purpose of this case study is to demonstrate the application of WNTR to simulate localized and system-wide emergencies and calculate the associated impacts. With Poughkeepsie, New York, drinking water utility (POK) as the case study, the emergencies analyzed in this paper include increased water demand due to firefighting, large diameter pipe closures, and the loss of the source water in which analysis parameters were chosen based on POK’s needs. Performance metrics are used to quantify system impact due to localized emergencies, and resilience metrics are used to quantify system resilience against system-wide emergencies. This case study also introduces a new resilience metric, combined performance index (CPI).

Case Study

Poughkeepsie, New York, is located approximately 113 km (70 mi) north of New York City along the Hudson River. Raw water taken from the Hudson River is filtered, disinfected, and distributed to approximately 80,000 individuals within the City of Poughkeepsie, Town of Poughkeepsie, the Dutchess County Water Authority, and the Town of Hyde Park (Dutchess County Water and Wastewater Authority 2019). The water treatment facility is owned by both the City and Town of Poughkeepsie. It has a maximum production capacity of 1.02 m3/s [23.3 million gallons per day (MGD)] but produces approximately 0.54 m3/s (12.3 MGD) per year. Pough-keepsie was identified as a candidate for drinking water resilience research because the water treatment facility is located in the estuarine tidal floodplain of the Hudson River, and threats of sea level rise had already prompted Poughkeepsie to work with EPA to assess extreme weather risks. Furthermore, water quality vulnerabilities exist, such as: (1) nearly 322 km (200 mi) of the total 507 km (315 mi) of the Hudson River has been placed on EPA’s National Priorities List of the country’s most contaminated hazardous waste sites (EPA 2021), and Poughkeepsie sources its water from this contaminated region; (2) the Hudson River experiences a moving seasonal salt front, which can adversely affect water treatment processes; and (3) the Hudson River acts as a corridor for rail- and barge-transported crude oil, presenting spill-related risks to the water supply. With Poughkeepsie’s source water being vulnerable, it is a priority to understand its drinking water system’s resilience to any disasters that disrupt the treatment facility operations. For this reason, POK wanted to understand the impact of source water disruptions on their system. POK was also interested in identifying critical pipes to inform capital and operational investments.

The POK water distribution model provided by the administrator of the Poughkeepsie Water Treatment Facility was used to simulate the various emergency scenarios. The POK model (Fig. 1) is composed of 2 reservoirs, 3 tanks, 6,622 residential, commercial, and industrial demand nodes, 10 flow control and pressure-reducing valves, 16 pumps (grouped into 7 pump stations), and 7,198 pipes across the City of Poughkeepsie and the Town of Poughkeepsie. The city and town are connected to the same treatment plant but have different personnel to maintain their respective infrastructure. Due to this, geographic information system (GIS) data from Dutchess County GIS Data Inventory (Dutchess County GIS Metadata Record 2018) was used to identify nodes, and associated emergency impacts, in each region.

Fig. 1.

Fig. 1.

Open in a new tab

Depiction of the Poughkeepsie, New York, drinking water utility (POK) distribution system. The square represents the reservoir, and triangles represent tanks. The shaded area represents the city, and the unshaded area represents the town.

Methods

The POK case study simulated three localized emergencies: (1) fire-fighting criticality, (2) pipe criticality, and (3) segment criticality, and one system-wide emergency: loss of water source. Normal operations were simulated for 400 h, during which the POK model maintained an average pressure of 56 mH2O (80 psi). To understand the potential range of impacts during localized scenarios, tank levels were used as a surrogate for system storage to determine when the criticality analysis should be simulated. To account for fill patterns of the tanks, low and high tank storage times were identified after 100 h once the tank levels stabilized. Thus, a comparison of system impacts between when the tanks were nearly drained and when the tanks were mostly full could be made. The combined low tank storage was calculated as 44,165 m3 (11.7 × 106 gal:) and occurred 120 h into the simulation, while the combined high tank storage was 51,548 m3 (13.6 × 106 gal:) and occurred 106 h into the simulation.

WNTR

WNTR was chosen for its benefits and flexibility over existing hydraulic modeling tools. WNTR extends the capabilities of EPANET to simulate disruptive incidents, quantify the impacts, and evaluate recovery efforts. The flexibility of the Python environment also allows the user to easily customize analysis, such as a stochastic analysis for a range of probabilistic scenarios of interest to a drinking water utility. In this case study, the pressure-dependent demand (PDD) model was used since it more accurately reflects the supplied demand during system disruptions.

CriticalityMaps, a supplemental WNTR package, was used for the localized emergency scenarios within the POK model. The CriticalityMaps package helps facilitate the analysis of increased water demand (for firefighting), pipe bursts, and segment (i.e., multiple connected pipes) bursts within a system (Hassett 2019). In WNTR, users can generate valve layers to represent isolation valve placement throughout the WDS to group pipes and nodes into segments.

Firefighting Criticality

Firefighting criticality analysis assessed the impact of firefighting demands on the system and used performance metrics to quantify the results. The analysis applied a firefighting demand at a given node and identified which nodes in the system experienced reductions in pressure as a result. In this analysis, firefighting demands were applied to all nodes connected to pipes with diameters between 0.15 m (6 in) and 0.20 m (8 in), as requested by POK, for a total of 2,871 nodes (simulations). A firefighting demand of 0.16 m3/s (2,500 GPM) was applied at each node for a duration of two hours (AWWA 2008) and simulated at both low and high tank storage times (120–122 and 106–108 h, respectively). This resulted in two firefighting criticality scenarios: (1) FCHS: 2-h firefighting demand during high tank storage; and (2) FCLS: 2-h firefighting demand during low tank storage.

Pipe Criticality

Pipe criticality analysis identified critical pipes within the distribution system and used performance metrics to quantify the results. To determine the criticality of every single pipe, pipes were closed individually during the simulation to identify impacted nodes and the resulting impacted population for a total of 7,198 pipes (simulations). Since POK can typically repair broken pipes within 48 to 96 h, the pipes were closed for 48 or 96 h beginning at either hour 120 (low storage) or 106 (high storage) h in the simulation. This resulted in four pipe criticality scenarios: (1) PC48LS: 48-h pipe burst during low tank storage, (2) PC48HS: 48-h pipe burst during high tank storage, (3) PC96LS: 96-h pipe burst during low tank storage, and (4) PC96HS: 96-h pipe burst during high tank storage. All scenarios were simulated for an additional 24 h after the bursts, for a total duration of 178 h (48-h burst) or 240 h (96-h burst).

Segment Criticality

Segment criticality helped identify critical segments within the distribution system and used performance metrics to quantify the results. To determine the criticality of a segment, the segment was closed during the simulation and the number of impacted nodes was used to calculate the population impacted. Since POK did not provide the locations of their isolation valves, valve placement was determined using WNTR’s valve generator. All 7,198 pipes were included in this analysis with N-2 strategically placed isolation valves for a total of 2,120 segments (simulations). N-2 denoted that for every node in the system connected to three or more pipes, two of those pipes did not contain an isolation valve. For nodes connected to one or two pipes, those pipes did not contain an isolation valve. The same 48-h and 96-h burst simulations in pipe criticality were used for segment criticality for a total offour scenarios: (1) SC48LS: 48-h segment burst during low tank storage, (2) SC48HS: 48-h segment burst during high tank storage, (3) SC96LS: 96-h segment burst during low tank storage, and (4) SC96HS: 96-h segment burst during high storage.

Source Water Loss Analysis

POK was interested in the loss of source water analysis due to system vulnerabilities, including freezing of their water intake, saltwater intrusion incidents from drought conditions, and contamination in the Hudson River. To simulate this in WNTR, the pipe connected to the reservoir was closed at hour 106 (high storage) for different durations depending on the source water loss scenario. Two recovery and three conservation scenarios were simulated for a total of five scenarios. Recovery scenarios assumed access to the source water was lost for a specified duration, while conservation scenarios assumed water conservation efforts were implemented during the loss of source water analysis. Recovery times were chosen based on when the city (133 h) and town (144 h) lost pressure to evaluate system resilience given different recovery efforts. Recovery scenarios were simulated by reopening the closed pipe connected to the reservoir (i.e., source). Conservation efforts are a common mitigation technique to help alleviate devastating impacts on consumers during an emergency scenario (e.g., loss of source water). The five source water loss scenarios were (1) SWLR1: access to the source water was restored after the city was unable to meet consumer demand due to low pressures in the system 133 h into the simulation, (2) SWLR2: access to the source water was restored after both the city and town were unable to meet consumer demand due to low pressures in the system 144 h into the simulation, (3) SWLC1: loss of source water with zero conservation efforts implemented, (4) SWLC2: loss of source water with moderate conservation effort implemented as a 25% reduction of water usage, and (5) SWLC3: loss of source water with maximum conservation effort implemented as a 40% reduction of water usage and stopped service to the three highest consumers. All scenarios had a simulation duration of 400 h. The longer simulation duration was used to observe the system returning to normal operations during the recovery scenarios, which was not necessary for the criticality analyses as the system returned to initial conditions within one or two timesteps.

Simulations evaluated the effect of different conservation approaches on the POK model’s response during a loss of source water scenario. SWLC1 simulated the loss of the source water without changes to node demand. SWLC2 simulated the loss of the source water with a 25% reduction in node demand based on 2015 California drought regulations (CalEPA 2015; Plumer 2015). SWLC3 simulated the loss of the sourcewater with a 40% reduction in node demand and stopped service to the three largest consumers. These parameters were chosen based on feedback from POK as well as 2019 drought regulations in Lincoln, Nebraska, which reduced individual water consumption by 50%, and commercial and industrial water consumption by 20% (Matteson 2019). While the Nebraska regulation reported a 50% reduction in consumption, hydraulic convergence errors occurred for any conservation effort simulated above 40%.

Analysis Metrics

The number of impacted nodes, impacted population, and performance metrics were used to quantify the impacts of the localized emergencies while resilience metrics were used to quantify the resilience of the POK model during a system-wide emergency. For the firefighting criticality, a list of impacted nodes was generated for each firefighting node. Similarly, during pipe and segment criticality, a list of impacted nodes was generated for each pipe, or segment, burst. The impacts from an individual criticality simulation ranged from zero to all nodes and consumers impacted within the system. A node was considered impacted if the pressure was below the minimum pressure at any point during the emergency assuming it was not below minimum pressure during normal operations. For the POK case study, the minimum pressure was defined as 14.06 mH2O (20 psi), as regulated in the New York codes, rules, and regulations for public drinking water systems (NYCRR 1981), and the required pressure was defined as 17.58 mH2O (25 psi). Minimum pressure is the threshold for nodes to receive a portion of the requested water demand, while required pressure is the threshold for nodes to receive the full requested water demand. Population impacts were calculated per node using WNTR’s population method with an assumed 6.13 × 10−6 m3/s (140 gal. per day) water consumption per capita. Using this assumption, the POK model population was calculated as 80,122, with 23,736 in the city and 56,386 in the town, and was consistent with the consumers served by POK. WNTR’s population metric can be used when census tract data is not available or incomplete.

The performance metrics, WSA, and average system pressure were used to understand the behavior of the POK model during the three criticality analyses. WSA is calculated using Eq. (1)

WSA=DiDexp (1)

where Di = delivered demand; and Dexp = expected demand at node i. WSA can be between 0 and 1, where 0 indicates the system is unable to deliver any water to the consumer and 1 indicates the system delivers all of the requested water demand to the consumer. A WSA of 0 occurs when the pressure at a node is below the set minimum pressure, and a WSA of 1 occurs when the pressure at a node exceeds the required pressure. For this case study, WSA was calculated using the water_service_availability function in WNTR.

To quantify the impacts on the POK model during the system-wide emergency, two resilience metrics were used. The first metric, MRI, was calculated per junction using Eq. (2):

MRI=(Pi+Elevi)−(Preq+Elevi)Preq+Elevi (2)

where Pi = pressure at node i; Preq is required pressure; and Elevi = elevation of node i. The head for each node is calculated as the sum of pressure and node elevation, while the required head is calculated as the sum of required pressure and node elevation. In this study, the required pressure was constant, so MRI was directly related to the node pressure. MRI values range from less than −1 to greater than 1. A negative value indicates that the node head falls below the required head, while a positive value indicates that the node has a surplus head. An MRI value of 0 indicates the head is equal to the required head. For this case study, MRI was calculated using the modified_resilience_index function in WNTR.

While MRI has been used in academic and research settings, drinking water utility staff might be unfamiliar with the metric and associated terminology. For this reason, a new resilience metric, CPI, was developed with the intention to resemble similar water utility KPIs. CPI builds upon WSA while taking into consideration junction pressure. While junction demand is calculated using the specified pressure thresholds, it does not capture large changes in pressure when the lowest experienced pressure does not drop below the required pressure. Although consumers can still receive all expected water, large pressure changes can potentially cause damage to the pipes and system. Potential damage includes burst pipes from pressure surges and water hammers (Ghidaoui et al. 2005). CPI quantifies a system’s ability to meet standard operational node pressure and WSA conditions during an emergency. It is calculated using Eq. (3)

CPI=∑i=1NPiPbiN+∑i=1NDiDbiN2 (3)

where Pi = pressure at node i; Pbi = pressure at node i during normal conditions; Di = delivered demand at node i; Dbi = delivered demand during normal conditions at node i; and N = total number of nodes. CPI can be 0 or greater, where 0 indicates the system is unable to meet any water demand and all nodes have lost pressure. A CPI of 1 indicates the system is operating at normal operations, while a value greater than 1 indicates the system pressure or flow exceeds normal operations. While performance and resilience metrics can be used to describe a WDS’s response to an emergency, one metric cannot fully describe a system’s response or resilience. For this reason, multiple metrics were used to report and summarize the results of this case study.

Results and Discussion

Performance and resilience metrics were used to assess the POK model’s response to localized and system-wide emergencies. These analysis metrics revealed that the POK model was able to continue servicing most of the population during localized emergencies, and recovery and conservation efforts allowed the system to maintain water service to consumers for longer during system-wide emergencies.

Firefighting Criticality

For the firefighting criticality analysis, the firefighting node with the greatest impact, in both scenarios, affected 1,273 consumers, which was only 1.59% of the total population. Of the 2,871 firefighting simulations analyzed, the percentage of firefighting nodes that impacted at least one consumer was 72% and 48% during FCLS and FCHS, respectively. Most of the 2,871 simulations (36% in FCHS and 70% in FCLS) impacted between 11 to 500 consumers. When tanks were at low storage levels (FCLS), 675 additional nodes impacted the system, of which 92% were in the city with the remaining 8% in the town. Of these additional nodes, the one with the greatest impact affected 456 consumers (0.57% of the total population). The 10 firefighting nodes with the greatest impacts in FCHS affected an average of 1,082 consumers (1.35% of the total population), while the 10 firefighting nodes with the most impacts in FCLS affected an average of 1,103 consumers (1.38% of the total population). Table 1 shows the numerical breakdown of population impacts for both scenarios. Fig. 2 shows the spatial distribution of firefighting nodes and their impacts throughout the system with the areas of most difference between FCHS and FCLS highlighted. Only firefighting nodes with at least one impacted consumer are included in the figure. Firefighting nodes that impacted greater than 500 consumers were near large corporate, medical, and college campuses. Most of the nodes with impacts in FCLS were in the city.

Table 1.

Summary of the number of firefighting node simulations per population impacted intervals for each firefighting criticality scenario

Population impacted (count) Scenario
FCHS FCLS
0 1,482 807
1–10 329 22
11–500 1,021 2,003
501–1,000 32 30
>1,000 7 9
Open in a new tab

Fig. 2.

Fig. 2.

Open in a new tab

Population impacted by pressures less than 14.06 mH2O (20 psi) during (a) FCHS compared to during (b) FCLS. The size of the circle increases with population impact where the smallest circles indicate nodes that impacted between 1 and 10 consumers, and the largest circles indicate nodes that impacted over 1,000 consumers. Nonvisible nodes indicate nodes not included in the analysis [i.e., nodes connected to pipes with diameters less than 0.1524 m (6 in) or greater than 0.2032 m (8 in)] or nodes that resulted in zero population impacts. The square represents the reservoir, and triangles represent tanks.

Pipe Criticality

Most of the pipe criticality simulations (79%) impacted zero consumers across all four pipe criticality scenarios. Only 0.40% of the pipes in both PC48HS and PC48LS, 0.44% in PC96LS, and 0.41% in PC96HS impacted over 5,000 consumers. Furthermore, only three pipes in each scenario affected all 80,122 consumers in the system and were all connected to the reservoir. Of the 7,198 pipes analyzed, 21% of them impacted at least one consumer across all four scenarios. Only 20% of analyzed pipes impacted between 1 and 500 consumers.

Taking the three pipes connected to the reservoir out of consideration, the 10 pipes with the greatest impacts affected an average of 56,693 consumers (71% of the total population) in PC48LS, an average of 49,318 consumers (62% of the total population) in PC48HS, an average 59,814 consumers (75% of the total population) in PC96LS, and an average of 53,185 consumers (66% of the total population) in PC96HS. Table 2 shows the numerical break-down of population impacts for all four scenarios. All four pipe criticality scenarios produced similar results, suggesting that the POK model was reliable and redundant for pipe bursts up to 96 h. In particular, most pipes impacted less than 0.06% of the consumers. Redundant systems have more pathways to transport water to consumers, leading to fewer disruptions to consumers during localized emergencies. However, a few pipes affected the whole system, and these should be identified so that mitigation options to reduce the impacts on consumers could be explored.

Table 2.

Summary of the number of pipe burst simulations per population impacted intervals for each pipe criticality scenario

Population impacted (count) Scenario
PC48LS PC48HS PC96LS PC96HS
0 5,668 5,686 5,657 5,657
1–500 1,459 1,442 1,467 1,469
501–2,500 42 40 38 41
2,501–5,000 0 1 4 1
>5,000 29 29 32 30
Open in a new tab

Fig. 3 shows the spatial distribution of pipes and their impacts throughout the system, with the areas of most difference between PC48HS and the other scenarios highlighted. Only pipes with at least one impacted consumer are included in the figure. Pipes that impacted more than 5,000 consumers were either transmission lines, pipes connected to tanks and pumps, or pipes in regions of the system with few alternate routes. While WSA and average system pressure can provide an overall insight into system impacts, depending on the size of the system, localized impacts can be lost when averaging over the entire system. For this reason, Fig. 4 provides WSA and node pressure values for both the entire system and just the impacted nodes for two pipe burst simulations during PC96LS. To evaluate the range of impacts, Pipe A, which impacted 300 consumers, and Pipe B, which impacted 3,000 consumers, were selected. Pipe A had an immediate and consistent reduction in WSA and node pressure during the 96-h simulation, while Pipe B had more fluctuations. For both pipes, the reductions in WSA and node pressure were only observed when focusing on the specific nodes that Pipe A and B impacted in comparison to the whole system. The WSA and pressure fluctuations in Fig. 4(b) indicated that some of the 3,000 impacted consumers experienced lower WSA and pressures than others. In contrast, the consistent change in WSA and pressure in Fig. 4(a) suggested that all 300 impacted consumers experienced similar impacts. Thus, system average WSA and pressure might be poor indicators for identifying unmet demand and low pressures due to localized emergencies among many consumers.

Fig. 3.

Fig. 3.

Open in a new tab

Population impacted by pressures less than 14.06 mH2O (20 psi) during PC48HS (a) compared to the population impacts from (b) PC96HS, (c) PC96LS, and (d) PC48LS. Line thickness increases with population impact where the thinnest lines indicate pipes that had zero population impacts and the thickest lines indicate pipes that impacted over 5,000 consumers. The square represents the reservoir, and triangles represent tanks.

Fig. 4.

Fig. 4.

Open in a new tab

Average WSA and system pressure for (a) Pipe A; and (b) Pipe B during PC96LS. The shaded area from hours 120–216 indicates the 96-h pipe burst.

Segment Criticality

The segment with the greatest impact, in all four scenarios, affected all 80,122 consumers in the system. These two segments were connected to the reservoir. Of the 2,120 segments analyzed, the percentage of segments that impacted at least one consumer was 65% for SC48LS, SC48HS, and SC96LS, and 66% for SC96HS. Most of the 2,120 simulations (62% in SC48LS and 63% in SC48HS, SC96LS, and SC96HS) affected between 1 and 500 consumers.

Taking the two segments connected to the reservoir out of consideration, the 10 segments with the greatest impacts affected an average of 53,647 consumers (67% of the total population) in SC48LS, an average of 52,393 consumers (65% of the total population) in SC48HS, an average of 55,579 consumers (69% of the total population) in SC96LS, and an average of 53,272 consumers (66% of the total population) in SC96HS. Table 3 shows the numerical breakdown of population impacts for all four scenarios. Segments that impacted more than 5,000 consumers either included transmission lines or were located near tanks, pumps, or regions of the system with few alternate routes. As the locations for the segments with the largest impact were similar to pipe locations in pipe criticality, a separate figure for segment criticality was not included in this paper.

Table 3.

Summary of the number of segment burst simulations per population impacted intervals for each segment criticality scenario

Population impacted (count) Scenario
SC48LS SC48HS SC96LS SC96HS
0 738 737 733 722
1–500 1,320 1,325 1,325 1,337
501–2,500 31 28 30 30
2,501–5,000 1 2 2 2
>5,000 30 28 30 29
Open in a new tab

The numerical breakdown of population impacts for segment criticality followed a similar trend to pipe criticality, in which the majority of segments impacted fewer than 500 consumers. Since multiple pipes make up a single segment, there were 2,120 segments for all 7,198 pipes. Thus, the number of pipes that impacted zero consumers was almost sevenfold the number of segments. Despite this, 97% of the segments impacted fewer than 500 consumers, further suggesting high redundancy within the system.

Additionally, the number of segments impacting 500 or fewer consumers included most of the pipes that were smaller than 0.3048 m (12 in) in diameter, which had very little impact on the system when broken. Furthermore, the amount of storage in the system and the duration of the pipe burst appeared to have little effect on the number of impacted consumers.

Source Water Loss Analysis

Fig. 5 shows the results from the source water loss scenarios. The trends for each region (i.e., town, city, and whole network) were similar across all source water loss scenarios. For this reason, the system average for WSA and pressure are shown. SWLR1 restored access to the source water 133 h into the simulation, 13 h after most of the city lost WSA and pressure (27 h after the initial shutdown). SWLR2 restored access to the source water 144 h into the simulation, 3 h after the whole system lost WSA and pressure (38 h after initial shutdown). During SWLR1, WSA was restored within 33 h, and pressure stabilized within 77 h. During SWLR2, WSA was restored within 66 h, and pressure stabilized within 144 h [Fig. 5(b)]. POK was also interested in understanding the effectiveness of conservation efforts during a source water loss scenario. The WSA and average pressure results from all three conservation scenarios are shown in Fig. 5(c). SWLC1 maintained WSA for 35 h after source water loss, while WSA was maintained for 42 h and 58 h in SWLC2 and SWLC3, respectively. Maximum conservation efforts extended WSA by 20 h for a total of 55 h before a complete shutdown. A second reduction in WSA and pressure during conservation is observed in Fig. 5, representing the differences between the town and city losing access to water.

Fig. 5.

Fig. 5.

Open in a new tab

Average WSA and system pressure for (a) the whole network, city, and town with no conservation efforts; (b) recovery efforts; and (c) conservation efforts during source water loss scenarios. The vertical dashed line indicates the start of the water source loss at 106 h.

Similarly, MRI and CPI values decreased after the loss of the source water, which indicated when the city and town individually lost pressure. Fig. 6 provides MRI and CPI values for the recovery and conservation source water loss scenarios. SWLR2 had more negative MRI values compared to SWLR1, suggesting that on average the nodes during SWLR2 were unable to meet the required node head. Increasing levels of conservation efforts resulted in fewer negative MRI values, indicating improved resilience during water source loss with conservation efforts. SWLR1 recovered faster than SWLR2 since SWLR2 had a CPI of 0 for a longer time period. Similarly, increasing conservation efforts reduced the durations of time periods with a CPI value of 0, indicating the POK model was able to maintain WSA and pressure on consumers longer when conservation efforts were implemented. Fig. 6 shows that MRI and CPI had similar trends with only slight differences for a few timesteps, but with the exception that CPI did not fall below zero. The similarity between the two metrics suggests that CPI could be an adequate surrogate resilience metric for MRI while using values that are easier to interpret.

Fig. 6.

Fig. 6.

Open in a new tab

MRI and CPI for (a) recovery; and (b) conservation efforts during the source water loss scenarios. The vertical dashed line indicates the start of the water source loss at 106 h.

Conclusions and Future Research

This case study used WNTR and an associated supplemental package to explore the impacts of localized emergencies related to firefighting, pipe, and segment criticality and resilience against a system-wide source water loss emergency for the drinking water system in Poughkeepsie, New York. Identifying vulnerable locations in a system can help with disaster planning. Additionally, the case study introduced a new resilience metric, CPI, that related system resilience to WSA and system pressure. Even though CPI seems intuitive to understand, it could have some limitations. While CPI is an arithmetic average of the system WSA and pressure and assigns equal weight to both variables, using weighted WSA and pressure values might provide a more dynamic summary of the system’s response to an emergency. This could also allow for the metric to be adaptable between localized and system-wide scenarios. Furthermore, the results of this study are deterministic and do not account for data uncertainties since only a single variation of each scenario was simulated. In particular, the duration of pipe bursts analyzed in this case study did not seem to stress the system, so the effects of longer durations could be explored in the future. Despite this, the results from the localized and system-wide emergencies were simulated based on POK’s interests and were consistent with their past experiences with similar emergencies. Working directly with POK allowed for more tailored scenarios, realistic modeling parameters, and ongoing feedback, which produced useful results to share with Poughkeepsies’ joint water board for capital and operational investment planning.

While this case study looked at the impacts of disaster scenarios on the POK model, similar methods can be used by researchers, consultants, and water utilities to gain valuable information about their distribution systems and resilience to a range of disasters or disruptions. Such findings can help mitigate the most critical deficits. Future analysis could analyze the impacts of back-to-back disasters on a system or how a combination of different mitigation responses could more quickly restore a system to its initial conditions. The methods for firefighting, pipe, and segment criticality, and the source water loss analysis are valuable for assessing the performance and resilience of water distribution systems.

Acknowledgments

The authors would like to acknowledge Randy Alstadt, Joint Water Board Administrator at Poughkeepsie’s Water Treatment Facility, for his input and support during this analysis.

Footnotes

Disclaimer

The EPA through its Office of Research and Development funded and managed the research described herein under Interagency Agreement (IA # DW08992524701) with the Department of Energy’s Oak Ridge Associated Universities (ORAU) and Interagency Agreement (IA #DW08992513801) with the Department of Energy’s Sandia National Laboratories. It has been subjected to review by the Office of Research and Development and approved for publication. Any mention of trade names, manufacturers, or products does not imply an endorsement by the United States Government or the US Environmental Protection Agency. EPA and its employees do not endorse any commercial products, services, or enterprises. 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 US 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 US Department of Energy or the United States Government.

Data Availability Statement

Some or all data, models, or code used during the study were provided by a third party (water distribution network model). Direct requests for these materials may be made to the provider as indicated in the Acknowledgements.

References

  1. Atkinson S, Farmani R, Fayyaz A, and Butler D. 2014. “Reliability indicators for water distribution system design: Comparison.” J. Water Resour. Plann. Manage. 140 (2): 160–168. 10.1061/(ASCE)WR.1943-5452.0000304. [DOI] [Google Scholar]
  2. AWWA (American Water Works Association). 2008. Distribution system requirements for the fire protection. AWWA M31. Denver: AWWA. [Google Scholar]
  3. AWWA (American Water Works Association). 2010. Risk and resilience management of water and wastewater systems. AWWA J100–10 (R13). Denver: AWWA. [Google Scholar]
  4. Bacon J, and Madhani A. 2017. “Drinking water issues, fire intensify Harvey crisis.” Accessed June 23, 2021. https://www.usatoday.com/story/news/nation/2017/08/31/weary-firefighters-begin-block-by-block-hunt-harvey-survivors/619603001/.
  5. Baños R, Reca J, Martínez J, Gil C, and Márquez A. 2011.“Resilience indexes for water distribution network design: A performance analysis under demand uncertainty.” Water Resour. Manage. 25 (10): 2351–2366. 10.1007/s11269-011-9812-3. [DOI] [Google Scholar]
  6. Bicik J, Kapelan Z, and Savic DA´. 2009. “Operational perspective of the impact of failures in water distribution systems.” In Proc., World Environmental and Water Resources Congress 2009, 1–10. Reston, VA: ASCE. [Google Scholar]
  7. Brazil L, Grigg N, Kummerow C, and Waskom R. 2017. “Water and climate: Charting the path to a sustainable future.” J. Water Resour. Plann. Manage. 143 (10): 01817002. 10.1061/(ASCE)WR.1943-5452.0000832. [DOI] [Google Scholar]
  8. Bruneau M, Chang S, Eguchi R, Lee G, O’Rourke T, Reinhorn A, Shinozuka M, Tierney K, Wallace W, and Von Winterfeldt D. 2003. “A framework to quantitatively assess and enhance the seismic resilience of communities.” Earthquake Spectra 19 (4): 733–752. 10.1193/1.1623497. [DOI] [Google Scholar]
  9. California Environmental Protection Agency. 2015. “2015 Resolution–NO. 2015–0032.” Accessed July 9, 2021. https://www.waterboards.ca.gov/board_decisions/adopted_orders/resolutions/res15.shtml.
  10. Cook D, Husband P, and Boxall J. 2016. “Operational management of trunk main discolouration risk.” Urban Water J. 13 (4): 382–395. 10.1080/1573062X.2014.993994. [DOI] [Google Scholar]
  11. DECNY (Department of Environmental Conservation New York). 1989. “Water conservation requirements.” Accessed February 9, 2022. https://www.dec.ny.gov/lands/86945.html.
  12. DECNY (Department of Environmental Conservation New York). 2022. “Saving water makes good sense.” Accessed February 9, 2022. https://www.dec.ny.gov/lands/5009.html.
  13. Diao KG, Sweetapple C, Farmani R, Fu G, Ward S, and Butler D. 2016. “Global resilience of water distribution systems.” Water Res. 106 (4): 383–393. 10.1016/j.watres.2016.10.011. [DOI] [PubMed] [Google Scholar]
  14. Dutchess County GIS Metadata Record. 2018. “Municipal boundary file for Dutchess County.” Accessed February 19, 2021. https://gis.ny.gov/gisdata/inventories/details.cfm?DSID=1156.
  15. Dutchess County Water and Wastewater Authority. 2019. “Drinking water quality report.” Accessed July 1, 2020. https://dutchessny.gov/Departments/Water-Wastewater/Docs/2019_Poughkeepsie_AWQR.pdf.
  16. EPA. 2016. “Flood resilience: A basic guide for water and wastewater utilities.” Accessed May 27, 2021. https://www.epa.gov/waterutilityresponse/flood-resilience-basic-guide-water-and-wastewater-utilities.
  17. EPA. 2018a. “Drought response and recovery guide for water utilities.” Accessed May 27, 2021. https://www.epa.gov/waterutilityresponse/drought-response-and-recovery-guide-water-utilities.
  18. EPA. 2018b. “Earthquake resilience guide.” Accessed May 27, 2021. https://www.epa.gov/waterutilityresponse/earthquake-resilience-guide.
  19. EPA. 2019. “Power resilience guide for water and wastewater utilities.” Accessed May 27, 2021. https://www.epa.gov/communitywaterresilience/power-resilience-guide-water-and-wastewater-utilities.
  20. EPA. 2020. “Water network tool for resilience (WNTR).” Accessed November 11, 2020. https://github.com/USEPA/WNTR.
  21. EPA. 2021. “National priorities list (NLP) sites—By State.” Accessed June 15, 2021. https://www.epa.gov/superfund/national-priorities-list-npl-sites-state#NY.
  22. Ghidaoui M, Zhao M, McInnis D, and Axworthy D. 2005. “A review of water hammer theory and practice.” ASME Appl. Mech. Rev. 58 (1): 49–76. 10.1115/1.1828050. [DOI] [Google Scholar]
  23. Giustolisi O. 2020. “Water distribution network reliability assessment and isolation valve system.” J. Water Resour. Plann. Manage. 146 (1): 04019064. 10.1061/(ASCE)WR.1943-5452.0001128. [DOI] [Google Scholar]
  24. Gong C, and Zhou W. 2017. “First-order reliability method-based system reliability analyses of corroding pipelines considering multiple defects and failure modes.” Struct. Infrastruct. Eng. 13 (11): 1451–1461. 10.1080/15732479.2017.1285330. [DOI] [Google Scholar]
  25. Hassett P. 2019. “Criticalitymaps.” Accessed June 1, 2019. https://criticalitymaps.readthedocs.io/en/latest/criticality.html.
  26. Hecht P, Stanton S, and Smith D. 2014. “Earthquake in Napa, Calif., leaves widespread damage.” Accessed June 23, 2021. https://www.govtech.com/em/disaster/earthquake-napa-calif-widespread-damage.html.
  27. Hughes R. 2018. “California wildfires: The day Paradise burned down.” Accessed June 6, 2021. https://www.bbc.com/news/world-us-canada-46198498.
  28. Hwang H, Forrester A, and Lansey K. 2013. “Resilien ce of regional water supply systems.” In Proc., World Environmental and Water Resources Congress 2013: Showcasing the Future, 946–954. Reston, VA: ASCE. [Google Scholar]
  29. Hwang H, and Lansey K. 2021. “Isolation valve impact on failure severity and risk analysis.” J. Water Resour. Plann. Manage. 147 (3): 04020110. 10.1061/(ASCE)WR.1943-5452.0001320. [DOI] [Google Scholar]
  30. Jayaram N, and Srinivasan K. 2008. “Performance-based optimal design and rehabilitation of water distribution networks using life cycle costing.” Water Resour. Res. 44 (1): 516. 10.1029/2006WR005316. [DOI] [Google Scholar]
  31. Jeong G, Wicaksono A, and Jang D. 2017. “Revisiting the resilience index for water distribution networks.” J. Water Resour. Plann. Manage. 143 (8): 04017035. 10.1061/(ASCE)WR.1943-5452.0000792. [DOI] [Google Scholar]
  32. Khatavkar P, and Mays L. 2020. “Real-time operation of water-supply canal systems under limited electrical power and/or water availability.” J. Water Resour. Plann. Manage. 146 (4): 04020012. 10.1061/(ASCE)WR.1943-5452.0001191. [DOI] [Google Scholar]
  33. Klise K, Bynum M, Moriarty D, and Murray R. 2017. “A software framework for assessing the resilience of drinking water systems to disasters with an example earthquake case study.” Environ. Model. Software 95 (4): 420–431. 10.1016/j.envsoft.2017.06.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Klise K, Hart D, Moriarty D, Bynum M, Murray R, Burkhardt J, and Haxton T. 2020. “Water network took for resilience (WNTR) user manual: Version 0.2.3.” https://wntr.readthedocs.io/en/latest. [Google Scholar]
  35. Lambert A, Fantozzi M, and Thornton J. 2013. “Practical approaches to modeling leakage and pressure management in distribution systems–progress since 2005.” In Proc., 12th Int. Conf. on Computing and Control for the Water Industry—CCWI2013. Perugia, Italy: Elsevier. [Google Scholar]
  36. Laucelli D, and Giustolisi O. 2015. “Vulnerability assessment of water distribution networks under seismic actions.” J. Water Resour. Plann. Manage. 141 (6): 04014082. 10.1061/(ASCE)WR.1943-5452.0000478. [DOI] [Google Scholar]
  37. Liu HX, Walski T, Fu G, and Zhang C. 2017. “Failure impact analysis of isolation valves in water distribution network.” J. Water Resour. Plann. Manage. 143 (7): 04017019. 10.1061/(ASCE)WR.1943-5452.0000766. [DOI] [Google Scholar]
  38. Marchese D., Jin A Fox-Lent C, and Linkov I. 2020. “Resilience for smart water systems.” J. Water Resour. Plann. Manage. 146 (1): 02519002. 10.1061/(ASCE)WR.1943-5452.0001130. [DOI] [Google Scholar]
  39. Matteson C. 2019. “Water conservation tips for drought, flood and other disasters.” Accessed February 22, 2021. https://drought.unl.edu/Publications/News.aspx?id=336. [Google Scholar]
  40. Miller R. 2021. “Millions of gallons have leaked from burst water pipes in just one Texas city: ‘That is an incredible amount of water.” Accessed June 23, 2021. https://www.usatoday.com/story/news/nation/2021/02/18/texas-winter-weather-frozen-pipes-burst-food-damage-expected/4489631001/. [Google Scholar]
  41. Minaei A, Haghighi A, and Ghafouri H. 2019. “Computer-aided decision-making model for multiphase upgrading of aged water distribution mains.” J. Water Resour. Plann. Manage. 145 (5): 04019008. 10.1061/(ASCE)WR.1943-5452.0001070. [DOI] [Google Scholar]
  42. Minelli A. 2021. “Performance benchmarking in the water utilities sector.” Accessed February 24, 2022. https://www.performancemagazine.org/benchmarking-water-utilities-sector/. [Google Scholar]
  43. New York Codes, Rules, and Regulations. 1981. “Title: Section 5–1.27—Adequacy of distribution system.” Accessed March 22, 2022. https://regs.health.ny.gov/content/section-5-127-adequacy-distribution-system.
  44. NIAC (National Infrastructure Advisory Council). 2010. “A framework for establishing critical infrastructure resilience goals.” Accessed February 8, 2022. https://www.dhs.gov/xlibrary/assets/niac/niac-a-framework-for-establishing-critical-infrastructure-resilience-goals-2010-10-19.pdf. [Google Scholar]
  45. Ostfeld A, Kopgan D, and Shamir U. 2002. “Reliability simulation of water distribution systems—Single and multiquality.” Urban Water 4 (1): 53–61. 10.1016/S1462-0758(01)00055-3. [DOI] [Google Scholar]
  46. Paez D, Filion Y, and Suribabu C. 2018. “Correlation analysis of reliability surrogate measures in real size water distribution networks.” In Proc., WDSA/CCWI Joint Conf. Ontario, Canada: Queen’s Univ. and the Royal Military College in Kingston. [Google Scholar]
  47. Plumer B. 2015. “A guide to California’s drought and water crisis.” Accessed February 22, 2021. https://www.vox.com/2015/4/29/18094190/california-drought-water. [Google Scholar]
  48. Rossman L, Woo H, Tryby M, Shang F, Janke R, and Haxton T. 2020. “EPANET 2.2 user manual.” Accessed March 22, 2022. https://cfpub.epa.gov/si/si_public_record_report.cfm?Lab=CESER&dirEntryId=348882. [Google Scholar]
  49. Srirangarajan S, Allen M, Preis A, Iqbal M, Lim HB, and Whittle AJ. 2013. “Wavelet-based burst event detection and localization in water distribution systems.” J. Signal Process. Syst. 72 (1): 1–16. 10.1007/s11265-012-0690-6. [DOI] [Google Scholar]
  50. Stevenson JR, Vargo J, Ivory V, Bowie C, and Wilkinson S. 2015.“Resilience benchmarking & monitoring review.” Accessed February 9, 2022. http://resiliencechallenge.nz/w-pcontent/uploads/2018/05/1ResilienceBenchmarkMonitoring-Deliverable1.pdf. [Google Scholar]
  51. Swain ED, Gómez-Fragoso J, and Torres-Gonzalez S. 2017. “Projecting impacts of climate change on water availability using artificial neural network techniques.” J. Water Resour. Plann. Manage. 143 (12): 04017068. 10.1061/(ASCE)WR.1943-5452.0000844. [DOI] [Google Scholar]
  52. Todini E. 2000. “Looped water distribution networks design using a resilience index based heuristic approach.” Urban Water 2 (2): 115–122. 10.1016/S1462-0758(00)00049-2. [DOI] [Google Scholar]
  53. Tsakiris G, and Spiliotis M. 2012. “Applying resilience indices for assessing the reliability of water distribution systems.” Water Utility J. 3: 19–27. [Google Scholar]
  54. Turner J, Qiao J, Lawley M, Richard J-P, and Abraham D. 2012. “Mitigating shortage and distribution costs in damaged water networks.” Socio-Econ. Plann. Sci. 46 (4): 315–326. 10.1016/j.seps.2012.02.001. [DOI] [Google Scholar]
  55. Walski T. 1993. “Water distribution valve topology for reliability analysis.” Reliab. Eng. Syst. Saf. 42 (1): 21–27. 10.1016/0951-8320(93)90051-Y. [DOI] [Google Scholar]
  56. Zhan X, Meng F, Liu S, and Fu G. 2020. “Comparing performance indicators for assessing and building resilient water distribution systems.” J. Water Resour. Plann. Manage. 146 (12): 06020012. 10.1061/(ASCE)WR.1943-5452.0001303. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Some or all data, models, or code used during the study were provided by a third party (water distribution network model). Direct requests for these materials may be made to the provider as indicated in the Acknowledgements.

RESOURCES