Development of seismic vulnerability index for water distribution network (SVI-WDN) in the province of Special Region of Yogyakarta, Indonesia: A conceptual framework Open Access

The Special Region of Yogyakarta is one of the areas located in the fire ring zone, which has a high risk of disasters due to primary hazards such as earthquakes, tsunamis, and volcano eruptions (Muntafi et al. 2020). These hazards cause the most severe impact, with extended casualties emphasizing the failure of infrastructures in the post-disaster period (Eastman et al. 1995). During this period, water supply infrastructures are also most vulnerable and required. According to Li (2007), facilities such as water intake, treatment, and distribution pipe networks, are prone to several disasters. This was consistent (Ratnapradipa et al. 2012; Hlavinek et al. 2008), where earthquakes caused severe effects by rupturing water infrastructures, pipes, spare parts, and materials. Damage to the water supply infrastructure also creates shortages, leading to various diseases such as cholera, typhoid, and diarrhoea (Alexoudi et al. 2007). This requires a vulnerability assessment method to avoid severe damage to water supply infrastructure. The methods that are required for assessment method are classified into four categories, (a) Index-based method, (b) geographical information system (GIS)-based decision support tools, (c) visualization tools, and (d) dynamic computer models (Satta 2014). An index-based approach provides a single unitless aggregated value for vulnerability level expression. This value simplifies complex and interacting parameters, leading to its usefulness and easy understanding for management tools (Balica et al. 2012). Many previous studies also used the index method to estimate the vulnerability level of a system (Hosseini & Mirza-Hessabi 1999; Balica et al. 2012; Adhikary et al. 2018; Sekovski et al. 2020; Baig et al. 2021).

In the development of an index, the estimation of the criteria/sub-criteria is very important, although no suitable method was observed for weight determination (Böhringer & Jochem 2011). This determination method is often obtained through the availability of data, as well as the relationship pattern between parameters and time constraints (Sutadian et al. 2017). In this process, two methods are commonly used, namely statistical and participatory techniques. In Zardari et al. (2015), the procedure of weight identification was unclear through the statistical method, leading to the higher values of insignificant criteria. This method subsequently requires a high number of sample sizes, which are commonly problematic in cases with limited data availability (Hutcheson & Sofroniou 1999). It also mostly assumes linear relationships among selected criteria/sub-criteria, which is often uncommon in related cases. Meanwhile, the participatory methods, e.g., Delphi and AHP techniques, are highly popular for index estimation. According to Franklin & Hart (2007) and Hartwich (1998), Delphi was described as an expensive method with a time-consuming process, compared to other related techniques (Zardari et al. 2015). For AHP, a pairwise comparison sequence is mostly used on the relative values ​​between two criteria/sub-criteria. It also has similar accurate results as Delphi and is described as an easy-to-use process carried out in a relatively short time (Sutadian et al. 2017). Based on these methodological benefits and drawbacks, this study selected the index-based method developed by the participatory approach of AHP. Many reports also developed the vulnerability index for water supply systems (Nojima 2008; Karamouz et al. 2010), which was only implemented in specific regions or countries. This regional restriction led to unguaranteed implementations in other areas. Therefore, this study aims to develop a seismic vulnerability index for water distribution network (SVI-WDN) framework for the Special Region of Yogyakarta. This SVI-WDN framework is a tool aimed achieving a consensus among many aspects, which eventually produce a single value (index) representing the vulnerability status of water distribution systems. Outcomes from the SVI-WDN will be the handling prioritization of pipe segments in the study area. The structure of this report is observed as follows: (A) the importance of SVI and the index development methods, (B) the occurrences of earthquakes in the Special Region of Yogyakarta and their impact on water infrastructure based on a previous incident (the 2006 earthquake), (C) the literature review of the existing seismic vulnerability assessment and outputs on the criteria and sub-criteria mostly relevant in representing the level of SVI, and (D) a detailed description of the AHP application, to develop the seismic vulnerability index.

Based on Figure 1, the damage incurred by the pipeline network included the activities of installation, replacement, as well as repair of pipes and bridges. In Yogyakarta, 10,952 damage points were observed, costing around Rp. 5 billion. The size of the damaged pipe also had a diameter of 50 to 250 mm. Other damaged areas were shallow wells and much industrial spring water broncaptering. These were the two primary water sources for 70–95% of inhabitants (BAPPENAS and Worldbank 2006). The service area belonging to the Water Utilities of Yogyakarta City was also highly affected with the highest damaged pipe.

Using the AHP method, the major steps in weight determination are as follows:

• a.

  Establish a hierarchy.

• b.

  Calculate the sub-criteria weight and SVI-WDN.

• c.

  Visualize the index into the GIS map.

There were three criteria found in the existing literature, namely physical, environmental and operational criteria. The physical criteria included the externals properties of the piping network. Meanwhile, the environmental aspects mainly dealt with the external factors influencing WDN's performance, with the operational criteria focusing on the activities required to continuously initiate the water supply system, toward providing the necessary service. Another review was conducted to ensure that the selected criteria and sub-criteria represented the study area's physical, environmental, and operational characteristics by excluding contradictions.

An extensive literature review including journal articles, regulations, standards, technical guidances and reports was conducted to identify the initial criteria, sub-criteria, and value range of SVI-WDN. In this case, sub-criteria score was classified into various scales. However, some studies did not classify a criterion due to the assessment of vulnerability using only the sub-criteria.

According to Nojima (2008), Adhikary et al. (2018), and Zohra et al. (2012) seismic vulnerability was examined for lifeline infrastructures. Other reports also explained multi-disaster vulnerability assessment in different disaster-prone countries, including Japan, Canada, Algeria, and Italy. The proposed experimental methods included statistical estimation, a hierarchal fuzzy expert system, multi-criteria decision-making, and hydraulic modelling. Table 1 presents the criteria and sub-criteria from these literature reviews; seven studies analysed the vulnerability index for global water infrastructure.

Based on Table 1, subsequent selections were conducted by considering all the criteria and sub-criteria relevant to the area and condition of the Yogyakarta water supply system. Several sub-criteria were also filtered based on two factors, namely relevancy and data availability. Moreover, the excluded sub-criteria from Table 1 or another source included (1) liquefaction, (2) Repair rate, (3) intersection pipe fault (Zohra et al. 2012; Adhikary et al. 2018), (4) number of repairs, (5) peak ground velocity (PGV), (6) road loading, (7) water table level, (8) daily traffic, (9) cost of repair, (10) damage to surroundings (11) loss of protection, and (12) type of service area (Yan & Vairavamoorthy 2003; Fares & Zayed 2008). Liquefaction was inrelevant due to following: liquefaction was a relatively minor event in Yogyakarta, and the system managed by water utility companies was mostly installed in the city or sub-district capital, which was from the incidence of landslides. Data road loading and daily traffic were also not considered because the pipe burial location was at the roadside or sidewalk. Based on data unavailability, the other irrelevant sub-criteria were subsequently excluded. The values for pressure, leakage record, water quality, and discontinuity were obtained from other literature sources related to Indonesian WDN. According to the structural analysis and connectivity, the vulnerability index was calculated to identify the susceptible parts (Pinto et al. 2010). This indicated the requirement of vulnerability, to represent a score used for index calculations in a formula (Eastman et al. 1995). The utilized value ranges in this development included vulnerability (Zohra et al. 2012) and several regulations related to local and international pipe standards. In this experiment, various scores were assigned to the range value of the sub-criteria, where the lowest and highest estimations had low and great effects (not vulnerable and highly vulnerable), respectively. After filtration through the two initially grouped factors, the sub-criteria were re-grouped into environmental, physical, and operational standards. Table 2 shows the selected criteria and sub-criteria.

The criteria, sub-criteria and their related scores were examined through questionnaire analysis with the experts, which contained government officials, academics, as well as water utilities and supply system specialists. This questionnaire contained agreement or disagreement questions, with the criteria and sub-criteria derived from literature reviews. The experts were also permitted to suggest the new criteria/sub-criteria and modify their value ranges. To determine the final criteria, sub-criteria, and value range for SVI-WDN, two screening procedures were established. Firstly, only the criteria and sub-criteria. selected by more than 50% of experts were used in the SVI-WDN calculation. Secondly, analyses were conducted by examining the experts' inputs in the response column. In this case, the similar inputs suggested by more than two experts were included in the list of criteria/sub-criteria/range values.

Furthermore, variation was observed in the values of the sub-criteria, as shown in Table 2. Since the scales are different, the criteria need to be normalized before multiplication by their weights. This contained a score of 1 to 7, with the vulnerability level description displayed in Table 3.

To calculate the sub-criteria weight, a pairwise comparison matrix was initially developed by combining several sub-criteria and various assigned scales to rate the stakeholders' perceptions (Supriadi et al. 2018). In this study, the 1–9 point scale was used to express the level of relative importance among sub-criteria. This scale was selected due to being easy to use, simple, and straightforward (Zhang et al. 2019). Table 4 presents the 1–9 scale, which exhibited the level of importance. For example, scales 1 and 9 showed that sub-criteria i was equally important and extremely more important compared to j, respectively. Conversely, a scale of 1/9 indicated that i was extremely less important to j.

According to Saaty (1990) the pairwise comparison matrix only considered the sub-criteria grouped similar criteria. This proved that the matrix was differentiated into three groups, namely physical, environmental and operational criteria.

With the dimension of N × N for each criterion, the pairwise comparison matrix was created, where N is the number of compared sub-criteria In this process, the participants were only instructed to fill the upper diagonal part of the matrix, since the lower sector had the positive reciprocal value of the upper triangle. For example, when the sub-criteria i compared to j has a scale of 9, is subsequently observed for the comparison of j-i. Table 5 shows the detailed pairwise comparison matrix.

Table 5

View Large

To determine the order of priority, this matrix underwent a normalization process. The AHP method also needs to be equipped with the calculation of the consistency index and ratio (CI and CR), to determine the CL (consistency level) of user input. When the CR ≤ 0.1, the user's answers were optimally consistent with the resulting solution (Saaty, 1990).

The SVI was implemented in the study area by identifying the condition of the pipe segment based on the sub-criteria. The score was then assigned to this segment before calculating the normalized value seismic vulnerability level was calculated.

This study presents the conceptual framework for developing a SVI-WDN in the Special Region of Yogyakarta, Indonesia. The SVI-WDN aims to assist water authorities to prioritize the pipe segment handling and maintenance considering many physical, environmental and operational aspects. The index-based method was subject to local conditions, so far studies of the vulnerability index for water distribution network were generated outside Indonesia. This is the first study that presents a vulnerability index for water distribution network in Indonesia. Indonesia particularly the Special Region of Yogyakarta has the unique characteristics of water distribution network and area, therefore it is crucial to develop a specific index for the Special Region of Yogyakarta. There were three criteria and 15 sub-criteria for representing conditions of water distribution networks and their environment which were derived from diverse literature. Each sub-criteria have values range and corresponding scores to describe the vulnerability level. Experts for this study consist of government officials, academics and a water supply specialist. The Analytical Hierarchy Process (AHP) was implemented to determine weight for selected criteria and sub-criteria. The vulnerability index was an aggregation of the normalized score that was multiplied by the weight of sub-criteria.

This study offers new insight into vulnerability index development for water distribution networks in the Special Region of Yogyakarta, Indonesia, which has not been researched in the past. However, this framework was designed to be limited for a certain region and only to the earthquake hazard. Therefore, future work will focus on incorporating a robustness analysis to test the applicability of the criteria and their weight to be implemented to other regions in Indonesia. Furthermore, the framework will be extended to include multi-disaster as Indonesia is a country which experiences many fatal disasters.

We would like to thank the Program of Thesis Recognition Grant (Rekognisi Tugas Akhir) from the Directorate of Research, University of Gadjah Mada, Indonesia for fully funding for this research.

This work was supported by the Program of Thesis Recognition Grant.

All authors contributed to the process of literature collection and review as well as design a conceptual framework. All author has read and approved the final manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.