A District Sectorization for Water Network Protection from Intentional Contamination Academic research paper on "Civil engineering"

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedía Engineering 70 (2014) 515 - 524

Procedía Engineering

www.elsevier.com/locate/procedia

12th International Conference on Computing and Control for the Water Industry, CCWI2013

A district sectorization for water network protection from

intentional contamination

A. Di Nardoa*, M. Di Natalea, D. Musmarraa, G. F. Santonastasoa, V. Tzatchkovb,

V.H. Alcocer-Yamanakab

aDepartment of Civil Engineering of Second University of Naples, via Roma 29, Aversa (CE) 80014, Italy bUrban Hydraulics Dept., Mexican Institute of Water Technology, Paseo Cuauhnahuac 8532, Jiutepec Morelos 62550, Mexico

Abstract

The introduction of cyanide with a backflow attack into a water system was studied. The recent development of techniques for water network sectorization, aimed to improve the management of water systems, represents also an efficient way to protect networks from intentional contamination. The possibility of closing gate valves by a remote control system to create an i-DMA (isolated District Meter Area) can reduce the risk of contamination and thus the extent of damage of a terroristic attack. The study proposes a novel technique for designing i-DMAs compatible with hydraulic performance and optimized for water network protection.

© 2013 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility ofthe CCWI2013 Committee

Keywords: water network protection, district metering, water contamination, malicious attack, sectorization.

1. Introduction

Water distribution networks are exposed to different potential sources of accidental and intentional contamination (US EPA, 2003). The first one is related essentially to problems of occasional bad source water quality, dysfunction of chlorine stations or pipe breaks; while the second one concerns malicious attacks

* Corresponding author. Tel.: +39-081-5010202; fax: +39-081-5037370. E-mail address: armando.dinardo@unina2.it

1877-7058 © 2013 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility of the CCWI2013 Committee

doi:10.1016/j.proeng.2014.02.057

represented by intentional introduction of a contaminant at the network sources or contaminant injection in a network pipe (Nilsson et al., 2005; Clark et al., 2006) or by a backflow that occurs when a pump system is utilized to overcome the pressure gradient of network pipes (Kroll, 2010). Water contamination by terrorist attacks is a major risk for society and can have serious consequences such as poisoning and infectious diseases; after September 11, 2001 many countries adopted guidelines for water quality monitoring and emergency action plans (HSPDs, 2002; US EPA, 2003, 2009; CER, 2005).

A malicious act may consist of introduction of chemical, biochemical or radioactive contaminants in the water supply network. Sabotage is most effective when the hazardous substance is injected as a concentrated liquid or powder. The most hazardous substances are biotoxins (organic substances that can cause serious poisoning) and biological agents (viruses and/or bacteria).

Studies available in the literature have been focused on the ability of common sensors to detect noticeable changes in water quality when a contaminant is present (USEPA, 2005; Hall et al., 2007), especially by monitoring parameters such as pH, conductivity, Total Organic Carbon (TOC), turbidity and residual chlorine, coupled with interpretive algorithms (McKenna et al., 2007; Umberg, 2008; Kroll and King, 2010). Other studies are focused on the optimal positioning of measurement stations and identification of point source contamination (Rico-Ramirez et al., 2007; Ostfeld et al., 2008; Chang et al., 2011). These techniques are very helpful to develop Early Warning Systems (EWS) but they are not so effective in assessing the impact of actions to be taken to guarantee safety and security of users.

When a water supply network contamination incident is identified, two main actions should be carried out: a) alert users not to use water and b) close the sector of the network in order to limit health risks.

Early warning is crucial for the first action to be successful, while the effectiveness of the second one depends on the possibility of closing pipes to disconnect network sectors. Early warning requires a good distribution of fast warning sensors over the network (Kroll and King 2010); pipe closing can only be done if network sectorization (or partitioning) has been envisaged in the planning phase.

The recent development of techniques for Water Network Partitioning (WNP) that divides the water network in District Meter Area (DMA) (Wrc/WSA/WCA 1994), directed to improve the management of water systems (Di Nardo and Di Natale, 2011; Di Nardo et al., 2013c), represents also an efficient way to protect networks from biological agents (Grayman et al., 2009; Murray et al., 2010). More recently, in Di Nardo et al. (2013a), the authors proposed a methodology to reduce the risk of intentional contamination of a water supply network using the technique of Water Network Sectorization (WNS). Sectorization is achieved by closing gate valves in the network pipes that link the DMAs (Tzatchkov et al., 2006). In this condition, wherein each district in the system is completely separated (or isolated) from all other districts, the isolated district can be named i-DMA, as proposed in Di Nardo et al. (2012) and in Di Nardo et al. (2013c).

A study for water network protection with WNS carried out by Di Nardo et al. (2013 a) showed that: a) DMA isolation is more effective than only water network partitioning, b) the higher the number of DMAs in a WNP, the better the protection for the users; c) WNP reduces the extent of risk because several introduction points are needed to have a wide negative impact on the network; d) WNP allows to activate easier protection measures because it is possible to disconnect a small part of the network; e) the methodology respects the criteria of dual-use value (Kroll and King 2010).

In this paper different WNPs, obtained with a recent methodology in compliance with hydraulic performance, based on graph partitioning techniques (Di Nardo et al., 2011) were investigated in order to evaluate the effects of different WNPs for water network protection. In this way the study analysed the benefits of defining i-DMAs compatible with hydraulic performance using different weights on pipes and nodes.

The intentional contamination was modelled as proposed in Di Nardo et al. (2013a) by the introduction of cyanide at a reservoir and with a backflow attack into a real-water system defining the most dangerous points for a deliberate contaminant delivery.

The analysis was carried out with different partitioning and sectorization scenarios on a real multiple source water distribution network in Italy. The results show a significant reduction of risk for users.

2. Contamination model

In this section the definition of the characteristics of the contamination incident are first described. A water supply network may be intentionally contaminated in several ways: various contaminants (chemical, biochemical or radioactive) may be introduced into one or more points of the water system (sources, reservoirs, tanks or generic points). In this work, two cases of intentional contamination were studied; specifically a) the introduction of cyanide at a reservoir and b) the backflow attack into a real-water system was modeled in order to define the most dangerous points for a deliberate contaminant delivery. The contaminant chosen is potassium cyanide (Patnaik, 2007), an inorganic compound, highly soluble in water (716 g/l) and extremely toxic (LD50 = 2.86 mg/kg). It is worth highlighting that choosing a specific contaminant is important to define the corresponding lethal concentration but this does not undermine the general value of the proposed approach that can be applied also to other contaminants (i.e. aldicarb, anthrax culture, fluoroacetate, nicotine, ricin, sarin, VX, etc.).

2.1. Backflow attack

The backflow attack scenario was borrowed from Di Nardo et al. (2013a) with some little modifications that increase the effects of contamination. Specifically the backflow is created with a pump system - easy to find on the market - which can introduce a contaminant into the water system by overcoming the pressure gradient of network pipes, and disseminate it into the network affecting the areas surrounding the introduction point. The malicious attack is carried out at a single point and, theoretically, by a single terrorist equipped with a small number of simple devices (a small pump, a backpack to transport the pollutant, etc.) that would allow him to commit the crime unnoticed.

A backflow attack can be easily accomplished mixing cyanide with water in a bathroom tub in any house and pumping the solution into the water network. In this study the bathroom tub was redefined with a higher dimension, as compared to Di Nardo et al. (2013a), more similar to the real ones; for this reason the number of exposed users is significantly higher. The introduction point can be everywhere in the system. The most dangerous for the users introduction points can be identified employing the EPANET2 water quality tool. Different assumptions have been made for evaluating the hazardous effects: a) every network node corresponds to a given number of users; b) a given amount of potassium cyanide (flow rate and concentration) is introduced into the water network for 2 hours (8.00 am and 10.00 am) (a two-hour interval is enough to refill the tub and mix cyanide (Di Nardo et al., 2013a); c) the lethal concentration of potassium cyanide in water for a user whose bodyweight is 70 kg is 200 mg/l (Patnaik, 2007).

2.2. EPANET quality model

Cyanide contamination has been simulated by means of EPANET2, a water quality simulation module that allows to model the transport of a dissolved species travelling down the length of a pipe with the same average velocity as the carrier fluid and may react (either growing or decaying) at a given rate. Longitudinal dispersion is usually not an important transport mechanism under most operating conditions (although it can be considered, if needed, as shown by Tzatchkov et al. (2002)). This means there is no intermixing of mass between adjacent parcels of water traveling down a pipe. The EPANET2 water quality simulator uses a Lagrangian time-based approach to track the fate of discrete parcels of water as they move along pipes and mix together at junctions between fixed-length time steps (Liou and Kroon, 1987).

The average velocity u corresponds to the water flow velocity computed with the hydraulic simulation module of EPANET2. The values of u are used as convective transport velocities in the following equation, along the i-th network pipe, adopted to describe transport of dissolved species through each pipe:

In equation (1) Ci represents the concentration of the dissolved species i (i.e. contaminant concentration), x is the position along the pipe and r is the rate of reaction (mass/volume/time) as a function of concentration. In this study, the reaction rate was assumed to be zero assuming that the reaction of cyanide with water is negligible during the time considered. At network nodes instantaneous and complete mixing is assumed, and the following equations are written:

in which Ck,out represents the contaminant concentration leaving node k; Qk,ext and Ck,ext represent the external input contaminant flow and concentration, respectively; QUn and Qin are flow and concentration in the i-th pipe; and Ik is the set of network pipes inflowing to node k.

2.3. Scenarios analised

Different scenarios were analysed to test the effects of a district sectorization for water network protection from intentional contamination. The different steps of the proposed methodology are illustrated in the flow chart of Fig. 1:

a) Hydraulic simulation of the water supply network by EPANET2;

b) Water Network Partitioning design;

c) Malicious attack simulation by the EPANET2 water quality module in different scenarios;

d) Computation of the total Number of Exposed Users (Neu) and the number of exposed users that consumed more than the lethal dose LD50 (Neu50);

e) Minimization of Neu50 changing the number and dimension of DMAs.

Fig. 1. Flow chart of the methodology proposed for WNP design for water network protection.

Specifically, starting from the INPUT data of the water network (with n nodes, m pipes, node water demand distribution Qi and node elevations z, with i=1..n, source heads Hs, with s=1..r reservoirs, and pipe length Lj with j=1..m); pipe flow qj, node pressure heads h and pipe head loss AHj can be calculated by the hydraulic simulation module of EPANET2 (step a).

The next step b) consists in the water network partitioning obtained using a novel methodology proposed by Di Nardo et al. (2013 a), based on a graph partitioning technique that allows to obtain the number and dimensions of the i-DMAs compatible with the level of service requested from users, based on different weights and balancing the flow, the length of pipes, or the power dissipated in the network. The choice of different weights allows to obtain different WNP layouts in terms of dimension, shape and hydraulic characteristics of each DMA. In the case

study the effects of this choice for network protections was investigated. The hydraulic simulation module of EPANET2 was used to compute the weights to be assigned to each pipe as explained below.

The step c) consists in simulation of the malicious attack, as described in the previous section, using the water quality simulation module of EPANET2. Different scenarios can be defined in order to test the effectiveness of this methodology. For each scenario the node where the worst consequences are expected in case of a backflow attack has been identified after a number of water quality simulations by EPANET2 by inserting a pump system with the same cyanide concentration at each network node and finding the corresponding maximum value of the number of exposed users (Neu).

After comparing the results obtained with the scenarios considered (step d) in terms of Neu, it is possible, returning to step b, to change the criteria used to define the DMAs and, consequently, the number and dimension of DMAs (step e). In this way a higher number of smaller districts can be defined and it is possible to reduce the negative effects of the network contamination because by closing a smaller DMA the number of exposed users is decreased.

This methodology can be applied not only to intentional and severe contamination but also to accidental and less severe contamination caused by poor source water quality, accidental backflows, or any other similar problem that can occur in a water distribution system.

3. Case study

In order to demonstrate the proposed methodology and to investigate the positive effects of WNP, the real water distribution network of Parete (Caserta, Italy), already studied by the authors (Di Nardo and Di Natale, 2012), has been used.

Parete is a town with 10,800 inhabitants in a densely populated area in the southern part of the province of Caserta (Italy), whose water supply network is sketched in Fig. 2. Its main hydraulic characteristics for the EPANET2 model are reported in Di Nardo and Di Natale (2012). The network is supplied from two sources and water consumption is exclusively for residential use with a prevalence of buildings built in 1970's and 1980's, with 3 to 4 floors.

After the hydraulic simulation (step a), the WNP of Parete (step b) was defined using the novel methodology presented in Di Nardo et al. (2011). The tables and figures reported in this paper refer to four DMAs but simulations were carried out also for three DMAs confirming the behavior already observed by Di Nardo et al. (2013a) that a WNP with a higher number of DMAs protects better the users. The methodology, based on graph partitioning techniques, allows to obtain different WNPs, considering weights for both edges and vertices (s,, co,).

In this study case the weights chosen were: dissipated power in each pipe yqjAHj (Di Nardo and Di Natale, 2011; Greco et al., 2012) for the edges (s,), length and volume of water in pipes linked to each node (®,), and node water demand. Because the aim of this study was to analyze protection actions corresponding to different WNP and WNS layouts, three different partitioning layouts were compared each of them obtained using the methodology proposed in Di Nardo et al. (2013a), but with different weights, as follows:

• Swxpi with weights sj= yqjAHj and c, = j Lj j2 where mt is the subset of pipes j incident to node i;

• SWNP2 with weights sj= yqjAHj and ®,= £ ™L j Vj j 2 where Vj is the volume of each pipe j incident to node i;

• SWNP3 with weights sj= yqjAHj and m,= Qi (water demand at i-th node);

Then, the results for Neu, Neu50, and the length of contaminated pipes Lep for the following scenarios were compared:

• Snnd) Network with No Districts (NND);

• SWNPi) Water Network Partitioning with four DMAs;

• SWNSi-1..4) Water Network Sectorization with isolation of a single DMA (i-DMA1 or i-DMA2 or i-DMA3 or i-DMA4) at 8.00 am (one hour from the beginning of the attack) or at 10.00 am (three hours from the beginning

of the attack).

Each WNP was defined following step e) of the flow chart in Fig. 1, returning to step b) and changing the weight of nodes, according to methodology proposed by Di Nardo et al. (2011), in order to reduce the number of exposed users Neu50.

As reported in Di Nardo et al. (2013a), the choice of closing the contaminated DMA an hour (at 8.00 am) or three hours (at 10.00 am) after the beginning of the malicious attack is a reasonable hypothesis for a water network that is not equipped with an early warning system (that allows shorter detection times). In this case, after a couple of hours, it can be realistically assumed that the authorities are alerted and they ordered to close the network district.

In the Table 1, the three different SWNPi and SNND were compared using performance indices introduced in Di Nardo and Di Natale (2010) and Di Nardo et al. (2013b). The performance indices of the WNP layouts (SWNPi) show a slight alteration compared to original network layout (SNND) confirming the effectiveness of the methodology of the chosen water network partitioning technique. An excellent result was obtained with weights on the nodes equal to pipe lengths corresponding to SWNPj with an alteration of the resilience of less than 1%. Tables 2 to 4 compare the results of the simulation for the different scenarios with scenario SNND.

All simulations were carried out assuming a pump system in the corresponding DMA that introduced potassium cyanide into the node that generates the worst damage for users. These scenarios are illustrated in Figures 2 and 3: a yellow triangle indicates the insertion point in each district but specific information about nodes, cyanide concentration and other simulation details are not reported in order to protect the Parete water supply system.

Table 1. Performance indices for each WNP.

snnd swnp1 swnp2 swnp3

hme. a, [m] 31.05 30.93 30.64 29.99

hml, . [m] 21.36 21.22 22.10 12.21

Ird - 0.77 6.27 7.90

In the Table 2, a comparison of simulation results for SWNP1 is reported. The values for WNP1 show an insignificant reduction (-1.68%) of Neu50 for this simple water network partitioning without any isolations. For valve closure an hour after the attack the best result is obtained for the WNS scenario with isolation of DMA2, with a reduction of Neu and Neu50 of up to 86.47% and 85.21% respectively, although the result for the isolation of DMA3 is poor with a reduction of only 12.80%.

Table 2. Simulation results for SWNP1

Scenario DMA1 DMA2 DMA3 DMA4 TOT

Neu50 Lep N,u Neu50 Lep Neu Neu50 Lep N,u Neu50 Lep Neu Neu50 Lep

- - [km] - - [km] - - [km] - - [km] - - [km]

Snnd 2382 2382 7.32 1102 1102 3.69 2277 1653 2.52 1514 1514 4.85 7275 6651 21.49

swnp1 2382 2382 7.32 1102 1102 3.69 2277 1541 2.52 1514 1514 4.85 7275 6539 21.49

swns1-1 2269 2269 6.92 0 0 0.00 0 0 0.00 0 0 0.00 2269 2269 6.92

8.00 am swns1-2 0 0 0.00 984 984 3.03 0 0 0.00 0 0 0.00 984 984 3.03

swns1-3 0 0 0.00 0 0 0.00 5800 5800 7.29 0 0 0.00 5800 5800 7.29

swns1-4 0 0 0.00 141 141 0.48 0 0 0.00 1514 1472 4.85 1655 1613 6.14

swns1-1 2382 2269 7.32 0 0 0.00 0 0 0.00 0 0 0.00 2382 2269 7.32

10.00 am swns1-2 1645 710 5.56 1102 1102 3.69 0 0 0.00 0 0 0.00 2747 1812 9.33

swns1-3 852 0 0.85 0 0 0.00 5800 5800 7.29 0 0 0.00 6652 5800 8.17

swns1-4 2382 1176 7.32 1102 1061 3.69 1827 0 1.98 1514 1514 4.85 6825 3751 20.50

In Table 3, a comparison of simulation results for SWNP2 is reported. For the case of WNP with no DMA isolation of (SWNP2) Neu is the same as that for SNND, while an insignificant reduction (5%) can be observed for

Neuso- In this scenario no additional security measure is taken and the effects of the contamination incident on the exposed users depend exclusively on the reduction of loop level and hydraulic section of the water system due to WNP. In contrast, WNS provides a significant reduction in almost all scenarios both with closures after an hour and after three hours, as reported in Table 3. The values of Neu and Neu50 for SWNS2_i are significantly lower (4,125 and 1,606 respectively, corresponding to a reduction of 43.30% and 75.85%). In SWNS2-2 and SWNS2-3 the reduction of the contamination impact is also lower with a reduction of 60.96% and 57.30% for SWNS2-2, and 76.85% and 75.84% for SWNS2_3. The results for SWNS2-4 are not so good, with a reduction of 21.64% for Neu and 14.28% for NeuS0. Obviously with a closure after three hours the simulation results show lower protection as reported in last four lines of Table 3.

Table 3. Simulation results for SWNP2 Scenario DMA1 DMA2 DMA3 DMA4 TOT

Neu Neu50 Lep [km] Neu Neu50 Lep [km] Neu Neu50 LLep [km] Neu Neu50 Lep [km] Neu Neu50 LLep [km]

snnd 721 721 2.96 1232 1232 4.60 3144 3144 9.10 2178 1554 3.25 7275 6651 21.49

Swnp2 721 721 2.96 1232 1232 4.60 3144 3144 9.10 2178 1215 3.25 7275 6312 21.49

swns2-1 721 664 2.96 1232 865 4.60 1993 0 7.11 179 77 0.23 4125 1606 16.03

8.00 am swns2-2 0 0 0.00 0 0 0.00 2840 2840 8.74 0 0 0.00 2840 2840 8.74

swns2-3 0 0 0.00 0 0 0.00 922 845 0.81 762 762 0.79 1684 1607 1.65

swns2-4 0 0 0.00 0 0 0.00 0 0 0.00 5701 5701 8.02 5701 5701 8.02

swns2-1 721 721 2.96 1232 1191 4.60 3144 1849 9.10 1728 179 2.25 6825 3940 20.5

10.00 am swns2-2 0 0 0.00 0 0 0.00 2840 2840 8.74 0 0 0.00 2840 2840 8.74

swns2-3 0 0 0.00 0 0 0.00 1432 1147 1.55 762 762 0.79 2194 1909 2.39

swns2-4 0 0 0.00 0 0 0.00 978 0 0.95 5701 5522 8.02 6679 5522 9.25

In Table 4 a comparison of the simulation results for SWNP3 is reported. The values for WNP show a slight reduction (19.92% for Neu and 12.40% for Neu50) while the WNS scenario shows the best results in terms of general risk reduction: the lowest value of Neu50 is 45.80% for SWNS3-2 which is a very good result compared to the rest of the scenarios that have a reduction of 14.28% for SWNS2-4 and 12.80% for SWNS2-4.

This result of SWNP3 arises from balancing of the four DMAs with weights equal to the water demand Qt that entails the main cause of exposure risk for users. This WNP, as reported in Tab. 1, has a resilience deviation index equal to 7.90% with a slight alteration of hmax, equal to 29.99 m, but a more significant alteration of hmin, equal to 12.21 m, compared to SNND, SWNP1 and SWNP2.

Table 4. Simulation results for Swnp3 Scenario DMA1 DMA2 DMA3 DMA4 TOT

Neu Neu50 LLep [km] Neu Neu50 Lep [km] Neu Neu50 Lep [km] Neu Neu50 Lep [km] Neu Neu50 LLep [km]

snnd 2303 1679 2.70 0 0 0.00 2668 2668 8.24 2304 2304 8.52 7275 6651 21.49

swnp3 3022 3022 3.25 2804 2804 3.41 0 0 0.00 0 0 0.00 5826 5826 7.47

swnp3-1 2277 2195 2.52 0 0 0.00 0 0 0.00 0 0 0.00 2277 2195 2.52

8.00 am swnp3-2 3022 801 3.25 2804 2804 3.41 0 0 0.00 0 0 0.00 5826 3605 7.47

swnp3-3 0 0 0.00 0 0 0.00 2178 1820 6.30 0 0 0.00 2178 1820 6.30

swnp3-4 2139 0 2.42 0 0 0.00 1682 1601 3.80 1255 1255 3.08 5076 2856 10.32

swnp3-1 2507 2507 2.82 0 0 0.00 0 0 0.00 0 0 0.00 2507 2507 2.82

10.00 am swnp3-2 3022 3022 3.25 2804 2804 3.41 0 0 0.00 0 0 0.00 5826 5826 7.47

swnp3-3 0 0 0.00 0 0 0.00 2566 2127 8.10 0 0 0.00 2566 2127 8.1

swnp3-4 2139 0 2.42 0 0 0.00 2668 1831 8.24 1255 888 3.10 6062 2719 15.32

The analysis of the simulation results with a DMA isolation three hours after the beginning of the attack shows, practically in all scenarios, that protection with WNS is insignificant with a large number of exposed users.

Fig.2 shows the three SWNP of the Parete network, in which each DMA is indicated with the green lines, compared with SNND; it is possible observe that the dimension and, consequently, the shape of each DMA is different.

Then in Fig. 2a the exposed nodes are indicated with a different colour (blue and red) and symbol (full and empty circle), as explained in the legend of the figure; the diameter of each circle is proportional to the number of exposed users, the higher the diameter, the larger is Neu or Neu50 corresponding to the each individual node. Fig. 2 shows clearly that SWNP3 is the best partitioning to protect network from contamination; indeed both Neu and Neu50, equal to 5826, are lower than for the other three scenarios, although the risk mitigation is low without sectorization. Fig. 2d shows a good result of SWNP3 also in terms of Lep with a significant reduction of the total length of contaminated pipes, equal to 7.47 Km.

Fig. 2. Effect of contamination in different scenarios SWNPi compared with Snnd.

Fig. 3 reports the WNS effects of SWNP3 with isolation at 8.00 am, in which the yellow triangle indicates the insertion point in each district (more information about nodes, cyanide concentration and other simulation details are not provided in order to protect the Parete water supply system). As seen from the figure, cyanide contamination is significantly reduced the DMA isolation one hour after the terrorist attack.

The comparison between Fig. 2 and Fig. 3 shows clearly the positive effect of DMA isolation both in terms of number of exposed users and length of contaminated pipes.

Fig. 3. The effects of contamination in different scenarios SWNS3

4. Conclusions

Hydraulic and water quality simulations showed that network protection can be achieved by DMA partitioning and sectorization; the latter may significantly decrease contaminant propagation and protect a greater part of the users from cyanide ingestion. The study confirms some insights of a previous study carried out by Di Nardo et al. (2013a): a) DMA isolation is more effective than water network partitioning alone, and b) WNP reduces the risk because several points of contaminant introduction are needed to produce a wide negative impact on the network. Simulation results show that WNP obtained by using node water demands as weights represents the best layout for water network protection, but it also corresponds to the maximum alteration of the performance indices although compatible with the level of service for users. Then, the simulation results show that water network partitioning without additional sectorization measures does not improve significantly the protection of the water system. On the contrary, the isolation of the attacked DMAs is always much more effective in reducing the number of exposed users, being the efficacy dependant on how fast the district isolation is. An isolation achieved three hours after the beginning of the terroristic attack turned out to be practically useless for user protection. Further studies are needed to improve network protection and to test the proposed procedure on networks of large dimensions.

Finally, it is worth highlighting that the methodology used to design the WNP showed excellent results as reported in the simulation with an alteration of the resilience of the original network of less than 1%.

References

CER (Centre for European Reform), 2005. The EU and counter-terrorism, Centre for European Reform, London.

Chang, N.B., Pongsanone, N.P., Ernest, A., 2011. Comparisons between a rule-based expert system and optimization models for sensor

deployment in a small drinking water network, Expert Systems with Applications, 38; pp. 10685-10695.

Clark, R.M., Chandrasekaran, L., Buchberger, S.B., 2006. Modeling the propagation of waterborne disease in water distribution systems: results from a case study, 8th WSDA Symposium, Cincinnati.

Di Nardo, A., Di Natale, M., 2012. A design support methodology for district metering of water supply networks, Eds: Kevin E. Lansey, Christopher Y. Choi, Avi Ostfeld, Ian L. Pepper In: Water Distribution Systems Analysis 2010 Proceedings of of the 12th International Conference, 12-15 September 2010, Tucson, Arizona, ASCE, pp. 870-887.

Di Nardo, A., Di Natale, M., 2011. A heuristic design support methodology based on graph theory for district metering of water supply networks, Engineering Optimization, 43(2), pp. 193-211.

Di Nardo, A., Di Natale, M., Guida, M., Musmarra, D., 2013a. Water Network Protection from Intentional Contamination by Sectorization. Water Resources Management, 27(6), pp. 1837-1850.

Di Nardo, A., Di Natale, M., Santonastaso, G.F., Tzatchkov, V.G., Alcocer-Yamanaka, V.H., 2012. Water Network Sectorization based on genetic algorithm and minimum dissipated power paths. In: Proceeedings of WaterLossEurope201. London:International Water Association, Ferrara, 23-35 Maggio 2012.

Di Nardo, A., Di Natale, M., Santonastaso, G.F., Venticinque, S., 2011. Graph partitioning for automatic sectorization of a water distribution system, Eds: Savic, D.A., Kapelan, Z., Butler, D., Proceedings of 11th International Conference on Computing and Control for Water Industry (CCWI). Urban Water Management: Challenges and Opportunities, 3-7 September 2011, Centre for Water Systems, University of Exeter, Exeter (UK), 3, pp. 841-846.

Di Nardo, A., Di Natale, M., Santonastaso, G.F., Tzatchkov, V.G., V.H. Alcocer- Yamanaka, 2013b. Performance indices for Water Network Partitioning. 5th Annual IWA International Conference on Benchmarking and Performance Assessment, April 9-12, Medelhn, Colombia, 1-11.

Di Nardo, A., Di Natale, M., Santonastaso, G., Tzatchkov, V., and Alcocer-Yamanaka, V. 2013c. Water Network Sectorization Based on Graph Theory and Energy Performance Indices J. Water Resour. Plann. Manage., DOI 10.1061/(ASCE)WR.1943-5452.0000364 (Feb. 28, 2013).

Grayman, W.M., Murray, R., Savic, D.A. 2009. Effects of redesign of water systems for security and water quality actors. In: Starrett, S. (Ed.), Proc. of the World Environmental and Water Resources Congress. 49 Kansas City, MO.

Greco, R., Di Nardo, A., Santonastaso, G.F., 2012. Resilience and entropy as indices of robustness of water distribution networks. Journal of Hydroinformatics, 14(3), pp. 761-771.

Hall, J., Zaffiro, A.D., Marx, R.B., Kefauver, P.C., Krishnan, E.R., Haught, R.C., Herrmann, J.G., 2007. On-Line Water Quality Parameters as Indicators of Distribution System Contamination. Jour. AWWA, 99(1), pp. 66-67.

HSPDs, 2002. Public Health Security and Bioterrorism Preparedness and Response, Act of 2002 (Bioterrorism Act).

Kroll, D., 2010. Protecting world water supplies against backflow attacks, a Water and Wastewater International, 25 (2), 4 p.

Kroll, D., King, K. 2010. Methods for evaluating water distribution network early warning systems, Journal of American Water Works Association, 102 (1), pp. 79-89.

Liou, C.P., Kroon, J.R., 1987. Modeling the propagation of waterborne substances in distribution networks, J. AWWA, 79(11), pp. 54-58.

McKenna, S.A., Hart, D.B., Klise, K.A., Cruz, V.A., Wilson, M.P., 2007. Event Detection From Water Quality Time Series. Proc. ASCE World Envir. & Water Resources Congress, Tampa, Fla.

Murray, R., Grayman, W.M., Savic, D.A., Farmani, R., 2010. Effects of DMA redesign on water distribution system performance. In: Boxall, J. & Maksimovic, C. (Eds.), Proceedings of the Tenth International Conference on Computing and Control in the Water Industry 2009. London, CRC Press.

Nilsson, K.A., Buchberger, S.G., Clark, R.M., 2005. Simulating Exposures to Deliberate Intrusions into Water Distribution Systems, Journal of Water Resources Planning and Management, 131(3), pp. 228-236.

Ostfeld, A., 2008. The Battle of water Sensor Networks (BWSN): A Design Challenge for Engineers and Algorithms. J. Water Resour. Plng and Mgmt., ASCE, 134 (6); pp. 556-568.

Patnaik, P., 2007. A comprehensive guide to the hazardous properties of chemical substances. Third Edition, J. Wiley & Sons, New Jersey. DOI: 10.1002/9780470134955.ch16.

Rico-Ramirez, V., Frausto-Hernandez, S., Diwekar, U.M., Hernandez-Castro, S., 2007. Water networks security: A two-stage mixed-integer stochastic program for sensor placement under uncertainty, Computers and Chemical Engineering, Elsevier, 31; pp. 565-573.

Tzatchkov, V.G., Alcocer-Yamanaka, V.H., Bourguett-Ortiz, V.H., 2006. Graph theory based algorithms for water distribution network sectorization projects. In Proc. of the 8th Annual Water Distribution Systems Analysis Symposium WDSA, Cincinnati, Ohio, USA.

Tzatchkov, V.G., Aldama, A.A., Arregurn, F.I., 2002. Advection-dispersion-reaction modeling in water distribution networks, Journal of Water Resources Planning and Management, ASCE, 128(5), pp. 334-342.

Umberg, K., 2008. Evaluation of Water Quality Event Detection Systems Deployed at the First Water Security Initiative Pilot Utility. Proc. 2008 AWWA Water Security Congress, Cincinnati.

US EPA, 2003. Response Protocol Toolbox: Planinng for and Responding to Drinking Water Contamination Threats and Incidents, online version: http://www.epa.gov/safewater/watersecurity/pubs/guide_response_overview.pdf

US EPA, 2009. National primary drinking water regulations. US. Environmental protection agency. Washington DC, USA, EPA 816-F-09-004.

Wrc/WSA/WCA Engineering and Operations Committee, 1994. Managing leakage: UK water industry managing leakage. Rep. A-J. London: WRc/WSA/WCA.