Risk assessment of urban water network triggered by social behavior due to rising water demand within households

Urban water sustainability research highlights the significant impact of social and spatial interactions on water consumption in an urban water network (UWN) (Kamran et al. 2024; Marcal et al. 2024; Xia et al. 2024). The literature explores this topic from various angles, including the relationship between water use and assessed tax value and building age (Chang et al. 2017); the relationship between water consumption and household characteristics such as size, presence of swimming pools, income level, and the proportion of elderly residents (Jayarathna et al. 2017); the connection between seasonal water consumption patterns and education level, outdoor space size, and household size (House-Peters et al. 2010). Furthermore, Medina-Rivas et al. (2022) highlighted differences in water access based on factors such as intermittent water supply, reliance on external sources for water, neighborhood altitude, and socioeconomic status. Lu et al. (2018) investigated variables, including the availability of water sources, domestic water prices, and the adoption of water-saving devices in relation to overall water usage. However, there is limited understanding regarding the dynamics within the household-UWN complex when considering both cognitive aspects and spatially driven risk assessment related to UWNs.

Chang et al. (2017) illustrated that the interplay between social and physical environments influences domestic water use in UWNs. They found that water use in single-family homes exhibits significant spatial dependance and neighborhood influence. Water demand management policies aim to reduce consumption and optimize urban water resources (Stavenhagen et al. 2018). Furthermore, the interaction between citizens, UWN infrastructure, and domestic technologies is crucial for optimizing water consumption (de Sousa & Dias Fouto 2024). Shahangian et al. (2021) noted that urban areas are susceptible to water scarcity and may not ensure adequate water availability. Viewing the water supply chain reveals the dynamics between physical and human elements, with citizens as pivotal actors in UWN usage. Thus, examining the psycho-social factors behind conservation behaviors in urban households is vital for developing interventions that encourage behavioral changes, which rely on acceptance and voluntary participation (Shahangian et al. 2021). Identifying household attitudes toward water consumption is essential to finding effective strategies for reducing internal water demand (Randolph & Troy 2008; Almulhim & Abubakar 2024). The diversity of actions by different stakeholders necessitates varied approaches in water policy aimed at improving water use. Therefore, understanding the factors that influence individuals' perceptions, behaviors, and intentions toward sustainable water use is critical. This includes implementing preventive measures, educating citizens, and raising social awareness to enhance water utilization efficiency and decision-making in the face of water scarcity challenges (Straus et al. 2016; Kang et al. 2017).

Daily water consumption in residential and indoor sectors rose during and after the COVID-19 quarantine, irrespective of climate or population size (Sabzchi-Dehkharghani et al. 2023). Changes in cleaning methods and the shift to remote work contributed to this increase (Kalbusch et al. 2020; Feizizadeh et al. 2021a). The decline in water usage by nonresidential users was offset by the rise in residential demand, with the most significant reductions seen in industrial activities, followed by the commercial sector (Kalbusch et al. 2020; Feizizadeh et al. 2021a).

Tavares et al. (2023) demonstrated that increased demand during the COVID-19 pandemic was influenced by the spatial location in UWN. Ribas et al. (2024) found that the rise in water consumption was more pronounced in middle- and low-income neighborhoods compared to high-income areas, highlighting the impact of sociodemographic factors. They emphasized the importance of resilience in water resource systems for challenging scenarios, such as the overlap of a pandemic and severe drought. Vizanko et al. (2024) utilized an agent-based model to analyze demand and water age in the UWN during the pandemic, noting a spatial shift in demand and an increase in water age, particularly in residential areas near industrial centers. Additionally, prior studies reported declining urban wastewater quality during the COVID-19 pandemic (Wan et al. 2024).

Numerous techniques exist in the literature for assessing vulnerabilities in water resource systems using Geographic Information System (GIS) based frameworks and statistical methods. Examples include seismicity analysis in UWN (Zohra et al. 2012), evaluation of aquifers to anthropogenic pollution (Jesiya & Gopinath 2019; Sadeghfam et al. 2021; Casadiegos-Agudelo et al. 2024) and subsidence (Bagheri et al. 2024), and analysis of water scarcity (Nguyen et al. 2020; Leveque et al. 2024). While frameworks for quantifying vulnerability are often subjective, some techniques, such as the fuzzy catastrophe scheme (FCS), mitigate this subjectivity by employing fuzzy membership and catastrophe functions (Sadeghfam et al. 2020a, 2021). Some researchers have explored the relationship between resilience and vulnerability in network infrastructure (Soldi et al. 2015; Mortula et al. 2020). Resilience assesses the overall strength of the infrastructure, whereas vulnerability focuses on specific components and their susceptibility to risks and potential consequences (Soldi et al. 2015). Vairavamoorthy et al. (2007) developed the a software tool, which integrates with GIS to create maps illustrating the risk of pollutant intrusions from sewers, drains, and ditches.

Spatial patterns significantly influence household water consumption (Medina-Rivas et al. 2022). Galán et al. (2009) developed an agent-based model in a GIS environment integrating social, urban dynamics, and water consumption to simulate various demand scenarios. This model highlighted the impact of urban dynamics on population movements, housing types, and territorial changes while also considering socio-geographical aspects such as technology and cognitive factors. Feizizadeh et al. (2021b) used regression analysis to explore the spatial relationship between water consumption and different urban locations, calculating consumption index proportions to assess sustainability. Their findings revealed higher water consumption rates in degraded areas and slums. Chang et al. (2017) applied ordinary least squares and spatial error regression models to analyze the effect of spatial correlation on water usage patterns, concluding that older inner-city areas have lower water consumption than newer suburban regions across all four studied cities.

This section deals with cognitive determinants, in which the following notation was defined as follows: attitude toward water-saving (⁠⁠); subjective norms (SNs) (⁠⁠); perceived behavioral control (PBC) (⁠⁠); attitude toward COVID-19 (⁠⁠); perceived risk (⁠⁠); perceived modifying approaches (PMAs) (⁠⁠); intention to save urban water resources (⁠⁠); sustainable urban water use behavior (⁠⁠).

Attitude refers to an individual's positive or negative evaluation of engaging in a specific behavior (Ajzen 1991), and a positive attitude is associated with a positive intention (Hussein et al. 2017). The evidence provided by Abadi (2019), Wan et al. (2018), and Ru et al. (2018) show that attitude is one of the determinants of intention to protect water and energy resources and intention to use urban green areas. Among various factors influencing behavioral intention, attitude holds the greatest influence, followed by SNs, PBC, personal normative beliefs, and moral intensity (Jafarkarimi et al. 2016). A survey of 1,196 households across the UK found that attitude plays a substantial role in determining water-saving intentions (Russell & Knoeri 2020). Environmental concerns and perceived risk indirectly impact the intention to conserve water through their influence on attitude, SNs, and PBC. Interestingly, high levels of informational advertising may enhance the effect of attitude on the intention to save water but weaken the effect of PBC (Si et al. 2022).

H1₍₎: Attitude toward water-saving behavior has a positive effect on behavioral intention.

SNs represent how people who are respected and considered approachable by citizens encourage them to conserve water resources. This pressure is, indeed, created by these people and encourages citizens to perform frugal behaviors. According to Ajzen (1991), individuals are likelier to engage in a particular behavior when they receive positive and inspiring messages from others, such as friends, family, and colleagues. This notion of SNs influencing behavioral intention is supported by the research conducted by Abdelradi et al. (2017). People's intention to use urban green areas is influenced by SNs (Wan et al. 2018). Furthermore, Wan et al. (2017) discovered that two interactive factors, namely experiential attitude and subjective norm, as well as instrumental attitude and subjective norm, impact recycling intention. Additionally, SNs play a role in increasing the likelihood of recycling among individuals with a positive experiential attitude and motivating those with limited knowledge about the benefits of recycling to engage in recycling behaviors. Moreover, Perren & Yang (2015) demonstrated that SNs contribute to individuals' intention to conserve water.

H2: SNs have a positive effect on behavioral intention.

PBC is a natural involuntary characteristic that refers to an individual's confidence or ability to engage in a specific behavior, and it plays a crucial role in determining behavioral intention (Song et al. 2012). Several studies have demonstrated that perceptions of control predict individuals' intention to conserve water (Perren & Yang 2015; Shahangian et al. 2021). It was also proven that PBC positively affects people's intention to use urban green areas (Wan et al. 2018). Ru et al. (2018) gained insight into energy conservation; for example, they found that PBC is the most determining factor for people's intention to save energy. One aspect of PBC is financial power, which relates to how individuals allocate funds from their revenue sources toward environmental projects. In this context, Wangui (2016) discovered that individuals with higher income sources are more likely to support environmental projects.

H3a/3b: PBC has a positive effect on (a) behavioral intention and (b) water-saving behavior.

The findings of SEM analysis conducted by Nguyen et al. (2022) demonstrate that the perceived risk of COVID-19 significantly impacts attitudes and PBC. In their study, Liu et al. (2022) collected 470 questionnaires from Chinese university students. They found that students' perception of positive COVID-19 information positively influenced their travel intention through PBC, travel attitudes, and travel motivations. Notably, PBC acted as a mediating variable, explaining a substantial portion of the effect of positive COVID-19 information perception on travel motivation and intention. Moreover, Lucarelli et al. (2020) discovered that individuals who possess a higher awareness of the interconnections between COVID-19 and climate change exhibit greater intentions to engage in pro-environmental behaviors. This heightened awareness, particularly regarding the environment's negative impacts caused by motor vehicles (including road safety burdens) and climate change issues, significantly influences the intention to use bicycles. Additionally, Irawan et al. (2022) found that positive affect related to COVID-19 significantly influenced SNs and PBC.

H4: Attitude toward COVID-19 has a positive effect on behavioral intention.

Perceived risk to health and well-being might influence health-promoting behavior (Anthonj et al. 2022). Exposure to the media, ability to gather information, trust in the government, and trust in the news media are important factors in shaping the public's risk perception of the virus, which in turn affects their intention to participate in social and economic activities (Choi et al. 2018). Choi et al. (2013) conducted a study on risk perception toward street food and found that perceived risks have a negative impact on consumers' attitudes and perceived benefits of street food. Furthermore, risk perception negatively influences behavioral intention. Meanwhile, efficacy belief and perceived risk can affect behavioral intention (Wibhisono & Salamah 2022). Environmental risk perception significantly influences environmental concerns, which subsequently impact behavioral intention. These constructs act as mediators of sustainable consumption behavior. The findings indicate that sustainable consumption behavior is associated with environmental concerns, which are influenced by increased environmental risk perception (Saari et al. 2021). Perceived risk indirectly affects water conservation intentions through attitude, SNs, and PBC. The influence of perceived risk on health-promoting behavior has been explored in previous research (Anthonj et al. 2022).

H5: Perceived risk has a positive effect on behavioral intention.

Evidence shows that pushing motives contribute to forecasting the behavioral intention to experience and is enhanced by the quality of the experience and perceived value (Dean & Suhartanto 2019). In this vein, punished activities are terminated more quickly than reinforced activities (Lamb et al. 1980). As induced from the literature review, reinforcement and punishment significantly affect motivation, values, decision-making, dealing with conflicts, and human performance (Asadullah et al. 2019); compliance might occur as a result of anticipated reward that is more likely to induce behavior change (Greitemeyer & Kazemi 2008). Furthermore, the inclination to adhere to information technology is significantly impacted by the perceived justice of the punishment, which is adversely affected by the actual punishment. When considering the perceived fairness of the punishment, the impact of satisfaction on the intention to comply diminishes, and the perceived usefulness loses its significance (Xue et al. 2011).

H6: PMAs have a positive effect on behavioral intention.

In a study conducted by Cooper (2017), the theory of planned behavior was employed to examine the relationship between urban water restrictions and compliance. Using structural equation modeling, the findings revealed that intentions to comply with these restrictions had a noteworthy and positive influence on reported compliance behavior. Abadi (2019) indicates that intention is a remarkable driver in shaping water-saving behavior in reviving endangered ecosystems.

H7: Behavioral intention has a positive effect on water-saving behavior.

This study introduces a novel methodology for assessing the risk posed by UWN to not meeting demand (NMD) through the utilization of both technical and cognitive data. As far as the authors are aware, there is a scarcity of studies that explore the interplay between the social and physical environments of UWNs. This integration was accomplished using FCS, which has shown effectiveness in various other contexts. Additionally, the study identifies several other innovative elements, which are detailed below:

• (1) To examine the impacts of the COVID-19 pandemic on water consumption in the UWN from a spatial perspective;

• (2) To determine the divers that trigger the behavior of water resource conservation;

• (3) To probe into the tendencies of residents in different urban areas toward conserving urban water resources.

In the context of the risk of NMD, intrinsic vulnerability relates to the inherent attributes of a UWN that hinder its ability to meet demands, irrespective of external factors or human activities. In the study, four data layers are incorporated and outlined as follows: (i) The distance from reservoirs (D), where the reservoirs supply the required water, heads into the UWN. There is clear support that increasing D increases head losses and the possibility of NMD increases. (ii) and (iii) The network density (N) and the consumers' density (C) can increase water demand in networks and the possibility of NMD. (iv) Pressure in the network (P) directly reveals the state of the water head in the network, and lower values indicate the possibility of NMD.

In this module, cognitive and perception variables were conceptually defined, and operational definition was rendered. Further explanations about the cognitive model and the interactions were provided in Section 1.1 and Figure 1.

The questionnaire validity and the indicator's reliability were evaluated through the partial least square (PLS) method, which confirms the relevancy of the questionnaire items and the generalizability of the results. The evaluation included assessing Cronbach's reliability indices and applying composite reliability (CR) to compare the results of two reliability indices. Notably, both Cronbach's alpha and CR measurements should exceed a cutoff value of 0.70. Additionally, convergent validity was employed using the average variance extracted (AVE) index, which should exceed a cutoff value of 0.50 (Abadi et al. 2019). Discriminant validity is confirmed when inter-construct correlations are lower than the square root of each construct's AVE (Wilson & Lankton 2009). Cross-loading analysis (CLA) was conducted to ensure that an item does not load on more than one latent variable. The analyses by Cronbach's alpha, CR, AVE, and CLA demonstrate that the questionnaire is a valid measurement tool (see Table 2). Notably, the second subscript shows the indicator defined for the related constructions.

Table 2

Measures of cross-loading

Construct/indicator .	.	.	.	.	.	.	.	.
	0.753
	0.758
	0.673
		0.704
		0.679
		0.678
		0.701
		0.787
			0.563
			0.730
			0.589
			0.620
				0.696
				0.626
				0.921
				0.622
					0.604
					0.605
					0.542
					0.733
						0.221
						0.259
						0.268
						0.295
						0.448
						0.254
						0.289
						0.306
							0.663
							0.843
							0.810
								0.282
								0.375
								0.298
								0.443
								0.444
								0.559
								0.456
								0.445
								0.595
								0.354
								0.664
								0.336
								0.278
								0.406
								0.533

Construct/indicator .	.	.	.	.	.	.	.	.
	0.753
	0.758
	0.673
		0.704
		0.679
		0.678
		0.701
		0.787
			0.563
			0.730
			0.589
			0.620
				0.696
				0.626
				0.921
				0.622
					0.604
					0.605
					0.542
					0.733
						0.221
						0.259
						0.268
						0.295
						0.448
						0.254
						0.289
						0.306
							0.663
							0.843
							0.810
								0.282
								0.375
								0.298
								0.443
								0.444
								0.559
								0.456
								0.445
								0.595
								0.354
								0.664
								0.336
								0.278
								0.406
								0.533

View Large

To improve the reliability of the results, the indicators with a Cronbach's alpha that reach the permissible limit were removed from Table 2. The cross-loading values for path analysis are displayed in Table 3. Notably, a path analysis refers to exploring potential sequential relationships between the considered attitudes, behaviors, and water consumption. In summary, our findings indicate that the indicator variables load well on the latent variables, exceeding the threshold of 0.70, indicating strong validity.

Table 3

The results of path analysis

.	Cronbach's alpha .	CR .	AVE .	.	.	.	.	.	.	.	.
	0.771	0.772	0.531	0.728
	0.838	0.836	0.506	0.389	0.711
	0.72	0.721	0.395	0.279	0.424	0.629
	0.817	0.813	0.528	0.203	0.316	0.284	0.726
	0.72	0.717	0.391	0.315	0.392	0.444	0.48	0.625
	0.7	0.682	0.266	0.444	0.382	0.502	0.323	0.683	0.495
	0.816	0.818	0.602	0.39	0.369	0.385	0.35	0.532	0.516	0.775
	0.794	0.777	0.199	0.423	0.5	0.562	0.367	0.585	0.603	0.449	0.447

.	Cronbach's alpha .	CR .	AVE .	.	.	.	.	.	.	.	.
	0.771	0.772	0.531	0.728
	0.838	0.836	0.506	0.389	0.711
	0.72	0.721	0.395	0.279	0.424	0.629
	0.817	0.813	0.528	0.203	0.316	0.284	0.726
	0.72	0.717	0.391	0.315	0.392	0.444	0.48	0.625
	0.7	0.682	0.266	0.444	0.382	0.502	0.323	0.683	0.495
	0.816	0.818	0.602	0.39	0.369	0.385	0.35	0.532	0.516	0.775
	0.794	0.777	0.199	0.423	0.5	0.562	0.367	0.585	0.603	0.449	0.447

Note. PMAs, perceived motivating approaches; AVE, average variance extracted; CR, composite reliability. Discriminant analysis (inter-correlation of latent variables).

View Large

To determine the mean differences between the expected and observed correlations, we also employed the standardized root mean square (SRMS), a crucial indicator of model fit. The SRMS quantifies the difference between the implied correlation matrix of the model and the observed correlation. Henseler et al. (2014) suggested SRMS as a suitable metric for PLS–SEM (Partial Least Squares – Structural Equation Modeling) to prevent model misspecification. Hu & Bentler (1998) noted that SRMS values below 0.08 or 0.10 are optimal. SRMS in the present study is 0.65, which shows that the model adequately fits the data. The blind-folding method was used to estimate the component's cross-validity.

The evidence from PLS–SEM showed that attitude toward water-saving behavior has a positive effect on behavioral intention (⁠ = 0.16). Therefore, H1 was confirmed. Other results pertain to the positive effect of attitude toward COVID-19 (⁠ = 0.10) and perceived risk (⁠ = 0.25) on behavioral intention, therefore, resulting in affirmed H4 and H5. We gained insight into the effect of PMAs on behavioral intention; as this variable increases, it is more likely to form the behavioral intention to perform the water-saving behavior (⁠ = 0.17), providing the confirmation of H6. A positive and substantial effect of behavioral intention on water-saving behavior is also indicated by the greatest value of the beta coefficient linked with the path of Intention → Behavior (⁠ = 0.45), supporting the validity of H7. R² and Cohen's d demonstrate that the conceptual model effectively matches the data, as seen in Table 4. The regression model's findings using tolerance confirmed a strong correlation between the independent variables, as seen in Table 4.

Table 4

Results of hypothesis testing, path coefficients, t-value, and p-value

Hypotheses .	Paths .	Path coefficients .	Standard error .	t-value .	p-value .	Decision .
H1		0.16	0.09	2.54	0.01	✓
H2		0.08	0.06	1.90	0.06	×
H3		0.08	0.06	1.87	0.06	×
H4		0.10	0.09	2.27	0.02	✓
H5		0.25	0.12	2.82	0.005	✓
H6		0.17	0.10	3.32	0.001	✓
H7		0.45	0.16	10.01	0.001	✓
= 0.388, Adjusted = 0.377, = 0.61
= 0.203, Adjusted = 0.200, = 0.25

Hypotheses .	Paths .	Path coefficients .	Standard error .	t-value .	p-value .	Decision .
H1		0.16	0.09	2.54	0.01	✓
H2		0.08	0.06	1.90	0.06	×
H3		0.08	0.06	1.87	0.06	×
H4		0.10	0.09	2.27	0.02	✓
H5		0.25	0.12	2.82	0.005	✓
H6		0.17	0.10	3.32	0.001	✓
H7		0.45	0.16	10.01	0.001	✓
= 0.388, Adjusted = 0.377, = 0.61
= 0.203, Adjusted = 0.200, = 0.25

Note. Paths and standardized regression weights. was measured by formulae (⁠⁠).

View Large

The paper takes on board the risk of UWNs to NMD using social and spatially driven GIS datasets. The paper takes a step toward combining social and technical datasets in a UWN problem, but there is room to elaborate on a modeling strategy. The developed methodology suffers from subjectivity with expert judgment and uncertainty. The FCS is employed to decrease the subjectivity with the rate and weight values of data layers. However, the lack of uncertainty assessment is one of the main drawbacks, which can be investigated in two forms of uncertainty with data and models in future studies. Recently, the Bayesian model averaging (Gharekhani et al. 2022) and generalized likelihood uncertainty estimation (Sadeghfam et al. 2021) have been applied in the aquifer vulnerability studies, which can be appropriate techniques in UWN studies.

The study incorporates four data layers, which comprise the distance from reservoirs (D), the network density (N), the consumers' density (C), and the pressure in the network (P). The selection of these data layers heavily depends on data availability. Additional data layers can also be incorporated, such as pipe material, the exact time of repair or replacement of each pipe, and water loss reports. Also, some beneficial data layers can be prepared using mathematical tools such as the EPANET solver by simulation of predefined consumption scenarios. The inadequate available data prevented the study from using such models, but it can be a research direction for future studies.

As mentioned in the earlier section, behavioral intention is positively affected by attitude toward water-saving behavior. Attitude is a seminal variable in psychological science used to study water-saving behavior, especially in studies that offer immediate implications. This variable indicates online evaluations and views about episodes under study or memory-based judgments. Attitude, in essence, represents the self-declarations of desirability and undesirability of objectives. In urban water resources, it would be important to take residents' views and know how they think of water use and water conservation schemes. It lets researchers and water authorities know what factors drive or restrain the adoption of water offerings or policies. This finding complies with studies such as Abadi (2019), Wan et al. (2018), and Ru et al. (2018). Attitude toward COVID-19 and perceived risk are other variables that account for behavioral intention. With increasing water consumption during COVID-19, people are likely to establish conservation attitudes and intentions, especially if they become aware of water restriction risks in urban areas when the media periodically report the restrictions of urban water sources. For example, Lucarelli et al. (2020) state that, due to COVID-19, people have a stronger intention to shape conservation behaviors and use conservation technologies. The literature shows that perceived risk triggers develop and affect behavioral intention (Wibhisono & Salamah 2022).

This study investigates the NMD risk in a UWN, focusing on the social behavior of residents, denoted by hazard, and the intrinsic characteristics of UWN, denoted by vulnerability. The hazard was assessed through 356 questionnaires, and vulnerability was determined using FCS to reduce inherent subjectivity. The classified risk in five bands shows that the central, southwest, and southeast regions are in Band 5, aligning with traditional city areas. The findings highlight that attitudes significantly influence the model, affecting the intention to conserve UWN. Thus, urban water managers and policymakers should consider this cognitive factor, as they often rely on their perspectives, which may differ from consumer interests. While boards may reach a consensus on policy content, citizens usually have diverse views and lifestyles that lead to differing acceptance of these policies. Therefore, water and wastewater officials must incorporate citizen feedback into UWN management, particularly in post-COVID-19, which has shaped public perceptions.

The study also reveals that a heightened sense of risk correlates with a stronger intention to protect urban water sources, especially when media coverage highlights water quality issues. Collective awareness plays a vital role in fostering protective behaviors. Evidence suggests that citizens are inclined to take action to preserve urban water resources, as indicated by a high path coefficient. Consequently, we recommend organizing educational sessions on optimal water usage at the household level, preferably in mosques and cultural centers, with participation from city officials.

One limitation of this research is its reliance on quantitative data, which may contain systematic and nonsystematic errors, thus requiring caution in generalizing the findings. Additionally, the uncertainty in data and GIS processing was not examined. For instance, pressure data from measurement accuracy and temporal variations contribute to data uncertainty, while spatial interpolation and its associated parameters relate to GIS processing uncertainty. Future studies could address these shortcomings.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.