Temporal public perceptions and experiences during water service disruptions: the case of Jackson, Mississippi

Failures in urban water systems are becoming a common occurrence in the US due to disasters, aging infrastructure, and financial constraints, among other concerns. For example, Jackson, Mississippi has experienced reoccurring water outages, burdening community members as they must seek alternative water sources. Prior research has primarily focused on technical, institutional, and public health aspects of water crises, with limited attention to community perspectives. Understanding such social aspects can shed light on, for example, community priorities, levels of trust, mental health concerns, and communication gaps. Here, we document the temporal experiences and perceptions of community members during the Jackson Water Crises. To do so, we qualitatively analyzed news media data and employed topic modeling techniques on social media data from two years, capturing two service disruption events. Our results reveal the cascading impacts of water outages on end-users, including financial, social, and technical issues. For example, over time, results indicate that trust in the government and water providers in Jackson eroded, which may hinder public support. Our results show that social media can be a useful tool for utilities to understand public perceptions in real-time. Recommendations proposed here can inform future responses to water crises in Jackson and other communities, ensuring end-users’ perspectives are incorporated.


Introduction
Access to drinking water is critical to ensure the health, well-being, and security of society (Krajewski et al 2019, Safarpour et al 2022).While access to drinking water is often taken for granted in the United States, this is not the reality for all regions (US Water Alliance 2023).Certain areas, particularly those with vulnerable populations, face reoccurring water crises (i.e. times when drinking water services are unreliable; Sivakumar 2011, Krajewski et al 2019) that have far-reaching implications (Ezell and Chase 2022, Bisgin et al 2023, Kilpatrick et al 2023).For example, after changing drinking water sources, the city of Flint, Michigan experienced high levels of lead in their water system (Chavez et al 2017).This contamination resulted in health issues and political distrust (Nowling and Seeger 2020, Sobeck et al 2020, Ezell and Chase 2022).
Other notable examples include the reoccurring water service outages in Jackson, Mississippi-henceforth referred to as the Jackson Water Crises (JWCs)-where residents in Mississippi's capital city experienced extended periods without access to clean water (Grigg 2023).The JWCs resulted in an array of challenges for residents, such as financial and logistical burdens to acquire bottled drinking water (Nwanaji-Enwerem and Casey 2022) and increased absence rates in public schools (Kim et al 2023).In the long-term, such water crises, especially prolonged events like the JWCs, can lead to a lack of trust in the government, population migration, increased healthcare needs, and reduced willingness to pay water bills (Morckel and Rybarczyk 2018, Morckel and Terzano 2019, Sobeck et al 2020, Johnson 2022).These events display the need to improve the provision of water services going forward, which can be done by documenting lessons learned from previous crises.
Literature on water crises has predominantly focused on technical aspects (e.g.Faust et al 2017, Tiedmann et al 2023, Timmerman 2023), institutional dimensions (e.g.Laugé et al 2015, Chavez et al 2017, Brown et al 2019, Hacker et al 2019, Nowling and Seeger 2020, Tzanakakis et al 2020, Millington and Scheba 2021, Tiedmann et al 2022), and the resulting health impacts (e.g.Ruckart et al 2019, Sobeck et al 2020).For example, Tiedmann et al (2023) analyzed the cascading water-sector challenges resulting from Winter Storm Uri through semi-structured interviews with Texas utilities.The authors evaluated water system resiliency and provided institutional and technical recommendations for utilities and policymakers.Their findings revealed a need to improve communication outage information to end-users, suggesting online maps as an educational tool.From an institutional perspective, Hacker et al (2019) emphasized the importance of improvisation, institutional confidence, and enhanced coordination with various stakeholders when managing infrastructure services during emergencies.In another study focused on institutional failures during the Flint water crisis, Chavez et al (2017) identified both the government and the news media as two main institutions that had failed in terms of accountability and transparency when working with residents.
Although research typically focuses on technical aspects of water crises, there is a body of work focused on end users' perspectives (Faust et al 2013, Getchell and Sellnow 2016, Odimayomi et al 2021, Otts et al 2022).Understanding public perceptions and lived experiences helps to identify community priorities and concerns, gauge trust, uncover inequities, and reveal gaps in communication (Pahl-Wostl 2009, Getchell and Sellnow 2016, Odimayomi et al 2021, EPA 2023).These insights can inform how decision-makers respond to crises, ultimately leading to increased community engagement.Researchers have acknowledged the importance of public feedback about water systems-such community engagement empowers community members to advocate for improved response (Whelton and Cooney 2004, Meyer et al 2018, Fawkes et al 2022).Several studies have analyzed public perceptions of water systems (Krajewski et al 2019, Araya et al 2020, 2021, Odimayomi et al 2021).For example, Faust et al (2013) examined the general public's perceptions, including their views on the necessity of new, capital-intensive water infrastructure and whether they supported or opposed such initiatives.Their study emphasized the inclusion of community members as valid stakeholders involved in decision-making.Similarly, Araya et al (2021) integrated community perceptions into water sector decisions after population displacement, highlighting the importance of aligning local authorities' choices with community-supported alternatives.Researchers have also used social media as a way to capture public perceptions in the water sector (Getchell and Sellnow 2016, Moors 2019, Ezell et al 2021, Dwivedi et al 2023).For example, Dwivedi et al (2023) analyzed public opinion about water quality issues, using Twitter to understand how water services affect end-users' quality of life.
Most studies concentrate on either normal operational contexts (e.g.Faust et al 2013, Araya et al 2020) or discrete disaster scenarios (e.g.Sobeck et al 2020, Tiedmann et al 2023), but there is a scarcity of research examining public perceptions and experiences longitudinally.One study by Yang and Faust (2019) examined evolving public perceptions of water systems in shrinking cities from 2013 to 2016 to gauge changes in trust and decision-making participation.Such longitudinal studies reveal pubic response as crises evolve, which can inform water utilities' decisions.Learning from the 2018 California Camp Fire, Odimayomi et al (2021) found that long-term water issues present unique challenges, including changes in water use patterns, the use of alternative water sources, and the installation of in-home water filtration systems.Their study also highlighted the importance of effective communication with residents during extended boil water advisories to foster trust, reduce ambiguities, and enhance collaborative problem-solving.Notably, most existing work uses surveys conducted at one point in time (e.g.Odimayomi et al 2021), limiting our understanding of how public perceptions change as a crisis unfolds.
Here, we study the temporal public perceptions and experiences during the JWCs-reoccurring events that allow for the exploration of how public perceptions and experiences change during prolonged water crises.To do so, we qualitatively analyzed news media data and employed topic modeling techniques on Twitter data.After data collection, Twitter was renamed to X.Such media sources have been commonly used to analyze public perspectives, especially in crisis situations (Moors 2019, Bisgin et al 2023, Grigg 2023).We use both news and Twitter data because they provide different insights.News media includes information for a wide range of ages and backgrounds, while social media, such as Twitter, provides insight into public discourse in real time (Benites-Lazaro et al 2018, Spearing et al 2021, Wahid et al 2022).We collected data over a two-year period that captures two distinct service disruption events, referred to as the first JWC (2021) and the second JWC (2022).These events are described further in the following section.
In this study, our research objectives were to (1) understand public perceptions and experiences during the JWCs, and (2) analyze how these perceptions and experiences changed temporally during the prolonged water crises.Studying such sociotechnical aspects of water crises allows us to make practical and policy recommendations rooted in the public's perspective.This can lead to more comprehensive crisis management, policy development, and proactive measures to mitigate future water crises (US Water Alliance 2023).By studying and learning from past failures, communities and governments can enhance their preparedness and response strategies, improving resilience and well-being.Furthermore, understanding public perceptions provides insight into the consequences of water crises.Taking both perceptions and experiences into consideration provides a comprehensive understanding of the situation, leading to solutions that consider not only how people were impacted but also how they felt about the crises.

The case of Jackson, Mississippi
Jackson, Mississippi, is home to approximately 150 000 residents, 82.5% of whom are Black or African American.The city has a poverty rate of 26.1%, well above the national average of 11.6% (US Census Bureau 2023).Over the years, Jackson has grappled with multiple water crises, dating back to the 1980s (Hernandez 2022, Kim et al 2023).See figure 1 for a timeline of water sector challenges in the city.In February 2021, there was a widespread water outage sparked by a winter storm that caused plant equipment and pipes to freeze, and a reservoir to fail.The 2021 JWC lasted approximately one month, leaving more than 40 000 residents without access to clean water.Just over a year later, the city faced another crisis in late August 2022 due to the Pearl River flooding (Johnson 2022).This flooding caused water treatment issues, leading to water service disruptions for consumers and the enactment of boil water notices throughout the city (Meng 2022, Nwanaji-Enwerem andCasey 2022).Due to the disruption, residents faced many difficulties, such as issues finding alternative water sources and communication challenges regarding water advisories (Sobeck et al 2020, Waldrop 2022).Aging water infrastructure, lack of maintenance, and a dwindling tax base made Jackson's water system vulnerable to such disasters (Grigg 2023).Notably, these challenges are not technical alone, but are deeply intertwined with historical, social, and cultural factors (e.g.historical issues with water billing; Grigg 2023).

Methods
Various methods can be employed to assess public perceptions and experiences regarding water services, such as qualitative analysis of interviews (Spearing et al 2020), media analysis (Moors 2019), and text mining techniques, including sentiment analysis and topic modeling (Wahid et al 2022, Bisgin et al 2023, Dwivedi et al 2023).These methods enable researchers to understand the social context, a necessary first step in offering policy recommendations to address underlying drivers of water crises.Here, we utilized qualitative analysis and topic modeling to analyze news media and Twitter data throughout the two-year study period.Figure 2 outlines our data collection and analysis.

Data collection
Publicly available news articles related to the JWCs were gathered using Nexis Uni (2023), an online news database.We reviewed news articles to ensure their relevance to our study, alignment with our research objectives, and to remove duplicate content, resulting in a final sample of 44 articles for qualitative analysis (see figure 2 for more details).The news data encompassed two distinct time periods: February to April 2021, which corresponds to the first JWC (2021) triggered by a winter storm, and August to October 2022, during the second JWC (2022) which was sparked by the Pearl River flooding (see figure 1 for more details).
Twitter data were collected using the full-archive premium search application programming interface with the assistance of the gtdownloader library (Acosta-Sequeda and Derrible 2023).To capture public perceptions, we collected tweets related to the JWCs from February 2021 (the onset of the first crisis) until February 2023 (the end point of our study) using 'water' , 'water crisis' , and 'boil water' within 'Jackson' keywords as search terms.Notably, the search query for collecting news media and social media differed due to the nature of the content.News articles may not always contain direct public views, so we incorporated keywords like 'Public' and 'Resident' to identify articles including resident quotes and perspectives.Following the same process as with news media, we analyzed tweets for two distinct time periods: February-April 2021 and August-October 2022.

Data analysis 2.2.1. Qualitative analysis of news media
A hybrid inductive-deductive qualitative analysis was used to identify themes in the news articles (Spearing et al 2021).Qualitative analysis is a valuable method for assessing non-numeric data, allowing researchers to explore social problems (Saldaña 2021, Spearing et al 2021).The coding process was performed using NVivo software.First, each excerpt, representing statements from individuals discussing water-sector challenges was coded as either a public perception or a public experience during the crisis (a deductive approach based on our research questions).Next, we used an inductive approach to identify specific themes (i.e.child codes), such as health concerns and the use of bottled water.Table 1 shows a simplified coding dictionary (a complete coding dictionary can be found in supplementary information; SI).To validate our analysis, we performed an intercoder reliability check resulting in a satisfactory level of agreement with a kappa coefficient of approximately 0.8 (Mezzich et al 1981).After developing the coding dictionary, we created frequency tables for the whole dataset, as well as for the first and second JWC separately.Separating our results based on the crises allows us to understand how public perceptions and experiences changed over time.

Topic modeling of social media data
Topic modeling was used to understand emergent trends in Twitter data.Topic modeling is a machine learning technique based on natural language processing.This technique facilitates the identification of underlying themes and topics within large textual datasets by organizing and categorizing the information (Chuang et al 2012).The Latent Dirichlet Allocation (LDA) algorithm, one of the most used algorithms, iteratively assigns words to topics and topics to documents to find the most likely topic assignments for each word and document.LDA assumes that documents contain a random mixture of latent topics, and each topic is represented as a probability distribution over a set of words (Chuang et al 2012, Benites-Lazaro et al 2018, Dwivedi et al 2023).Before topic modeling, we pre-processed the Twitter data (e.g.removing hyperlinks and Twitter usernames, converting all words to lowercase, eliminating punctuation, and adding stop words).Next, we applied the LDA algorithm to Twitter data (1) over a two-year period, (2) during the first JWC, and (3) during the second JWC; which allowed for temporal comparison.
Determining the optimal number of topics depends on the aim of the study and dataset size (Benites-Lazaro et al 2018, Wahid et al 2022).We examined various methods to classify the topics, settling on nine topics.There are several approaches to assign titles to emerging topics, such as manually checking the data set, using visualization charts, or employing measuring scores (Chuang et al 2012, Sievert and Shirley 2014, Wahid et al 2022).In this study, we used the LDAvis library developed by Sievert and Shirley (2014) to visualize the topics generated by the LDA model.LDAvis provides two panels for visualization-one offers a global view of topics while the other presents a horizontal bar chart showing individual keywords for selected topics.The bar chart displays the overall term frequency and the estimated term frequency within the selected topic, facilitating topic interpretation.

Limitations
As with any study, there are limitations.First, as a primary source, news articles and social media data are susceptible to inaccuracies and biases.It is important to note that the biases present in the data can be seen as beneficial-our research objective is to identify and code these public perceptions, whatever the biases.Further, social media usage varies across different age groups.For instance, Twitter is more commonly used by younger individuals, but news media represent the perspectives of a broader range of people.Therefore, incorporating both sources allows for a more comprehensive understanding of the subject matter.Another limitation of our study is related to the LDA algorithm, which is not deterministic, and results depend on input parameters and preprocessing of data, like considering specific stop words (Benites-Lazaro et al 2018).To reduce this limitation, we employed multiple runs of the LDA algorithm with different stop words to ensure robustness and consistency in the topic modeling results.The algorithm and libraries used do not give us the tweet count per topic because they treat all data as one document.However, they do provide keyword frequencies, which can be helpful for topic labeling.

Results
In this section, we show results in different timeframes-the whole dataset and during each separate crisis.This allows for insight into temporal changes in public perceptions and experiences.An in-depth explanation of our results is confined to the discussion section.

Qualitative analysis of news media
Table 2 shows the results from the qualitative analysis of news media.It is important to note that the frequency does not indicate importance but instead shows that the code is repeated more often in the data.As shown in table 2, about 25% of the excerpts were about peoples' perceptions of the factors that led to the JWCs and 75% of the excerpts were about their lived experiences during the crises.
People attributed the crises to social (69.1%), financial (17.5%), and technical (13.4%) factors.Residents were aware that technical and financial aspects such as aging infrastructure and tax/billing issues (financial code) contributed to the JWCs, but most discussion was focused on social aspects.The most frequent child codes under the social parent code are negative attitudes towards decision-makers and government agencies and perceptions about race as a factor leading to JWCs.This focus on social aspects was similar across crises, but the relative frequency of excerpts coded to the financial code increased during the second JWC (from 7.4% in the first crisis to 21.4% in the second crisis).This may be showing that the public was becoming more aware of the financial challenges experienced by the water utility and city government over time.Residents experienced technical (40.5%) and financial challenges (9.5%), such as water quality issues and increased financial burdens, but 50% of the references were related to social experiences.Main social experiences during the JWCs included challenges to access alternative sources of water (28.9%),public health concerns and related sicknesses (26.8%), emotional distress (17.6%), and altruism and community bonds (13.4%).Between the two crises, the relative frequency of financial experiences increased, while technical and social experiences remained relatively consistent.Furthermore, the overall frequencies of codes in the second JWC more closely resembled the complete dataset because the second crisis had significantly more articles and, in turn, references in the data.a Relative frequency is the percent of all excerpts coded to each parent code.

Topic modeling of social media data
We identified nine topics for the entire two-year period studied, as well as for the first and second crises separately.We then categorized the topics into public perceptions and lived experiences (consistent with the qualitative analysis).Topics, their definition, and examples are shown in table 3. The topics for each crisis are provided in the SI.As expected, the qualitative analysis of news media primarily categorized codes as lived experiences of residents (74.5%), while the emergent topics from Twitter data were mostly related to public perceptions (seven of nine topics).This difference can be attributed to the inherent nature of the sources.News media predominantly reports the stories of people's lives and their experiences, while social media platforms like Twitter directly reflect people's perceptions and attitudes.
Figure 3 shows how the topics differed between crises.Similar topics emerged in both crises, including boil notices, technical infrastructure issues, racism, and political issues.We see that people compared the JWCs with other events, but the specific event varied between the first (Austin, TX) and the second JWC (Flint, MI).This shows that other crises occurring at the same time can influence public perceptions in other contexts.In the first JWC, there was a focus on water quality and water quantity (i.e.technical considerations), but this shifted to focus on a need for long-term solutions and financial problems in the second JWC.This shift shows that prolonged water crises can increase the frustration of residents, causing them to advocate for more sustainable solutions (e.g.funding and managerial changes).Notably, the number of tweets about the JWCs increased from the first to the second JWC, as shown in figure 3, meaning that people tweeted more about their perceptions and experiences over time.

Discussion
Incorporating public perceptions into water infrastructure management enables people to act as social sensors, leading to community-driven solutions and enhancing social sustainability and community capitals (Kaminsky and Javernick-Will 2014).Here, we displayed how this can be done in the context of the JWCs.We discuss the temporal public perceptions, as well as experiences that emerged from media data.

Temporal public perceptions during the JWCs
Our results revealed three main categories of public perceptions: social, financial, and technical aspects that people perceived as leading to the crises.It is important to note that managerial issues, such as improper operations and maintenance, were often mentioned within the broader context of technical, political, and financial themes.For example, the technical incompetence of water managers was discussed in the context of operational and technical issues (e.g.'That's the issue now-poor maintenance for far too long'; Martin et al 2021).Additionally, concerns about state oversight of drinking water services were implied within discussions of political issues.We see that people may not explicitly articulate management issues but rather allude to them indirectly within broader contexts.
Political reasons emerged as a key factor in both news and social media-most excerpts criticized politicians, decision-makers, and government agencies.People mostly expressed negative attitudes towards authorities and their plans for Jackson (e.g.'The state has consistently ignored Jackson's asks for help.We are not a high-priority for the people in power'; Minta 2022b).On social media, people utilized the hashtag #msleg, which is associated with the Mississippi Legislature, to question the state officials' actions.For example, one tweet stated, 'How can a state justify spending $70 million in welfare money on rich people's agendas when hard-working folks in Jackson can't even get water to brush their teeth?It's hard to believe state officials care about the Black folks in Jackson.#msleg.' (BoiJuanda 2022).Notably, @tatereeves, the governor, was frequently mentioned in tweets, followed by several media-related accounts.The political perceptions code frequency from the news media data increased slightly from the first to the second JWC.Additionally, the frequency of keywords related to politicians, like mayor and governor, increased in the second JWC.This underscores the adverse consequences of prolonged or recurring water crises, as it reflects public frustration and the quest to pinpoint the root causes of the crises, including identifying individuals who can offer solutions.
Race emerged as a significant topic during both crises, consistent with previous studies on other events like the Flint water crisis (Ezell and Chase 2022, Bisgin et al 2023, Kilpatrick et al 2023).In our study, race was one of the most frequent codes in the qualitative analysis and a prominent topic in topic modeling during both JWCs.Tweets such as '40 000 residents of Jackson were left without potable water….It's what happens when a state with a sorry racial history fails to help ensure that the citizens of a largely Black city have access [to] one of the most basic human needs: safe, clean water' (Milwaukee County Human Rights Commission 2021) and 'It's shameful.There is no question in my mind that if Jackson was 70% white, there would be a greater investment in water infrastructure' (Hassanein 2022) show the discourse around race and water inequities.Considering previous research and our results, it is crucial to acknowledge race as a significant social factor in the water sector, emphasizing the importance of continuing to study environmental justice (Meng 2022, Bisgin et al 2023).
The poverty rate in Jackson is 26.1%, significantly higher than the national average of 11.6%.While the qualitative analysis reveals that people perceive the community's income level as a contributing factor to the JWCs, its relative frequency is lower compared to political reasons and race.Nonetheless, understanding the implications of poverty is crucial because cities with high poverty rates often face challenges securing the necessary funding for capital improvements (Grigg 2023).Further, poverty can lead to a dwindling tax base, leaving numerous needs unmet (Blake 2022).Dissatisfaction with the tax and billing system is evident in news media (financial perceptions)-e.g.'Even though we are the capital city, even though we do contribute more to the state's coffers than any other city, there has not been reciprocal reinvestment in the city of Jackson.' (Northey 2021).Despite the enactment of a water system customer bill of rights in 2019 (City of Jackson 2019), people expressed concerns about not receiving water bills, inability to pay high water bills, and the lack of reinvestment.Topic modeling revealed different trends related to finances, focusing more on the general funding constraints of the city (e.g.lack of funding for capital projects) and the decision-makers prioritization of grants.Notably, the financial problems topic did not emerge as a topic in the first JWC, while there were various tweets about financial problems in the second.This trend emphasizes the negative financial impacts of prolonged crises.
Public perceptions about technical aspects of water systems also emerged which is not surprising given that the ASCE's (2020) report card ranked Mississippi's water infrastructure a D due to its aging and deteriorating drinking water infrastructure (ASCE 2020).Statements related to technical perceptions highlight concerns about the city's aging water infrastructure and operational issues in Jackson (e.g.'Our infrastructure is failing people.It is broken, and you cannot put new things on top of old broken things'; Blake 2022).Interestingly, the relative frequency of technical perceptions decreased over time in the news data (from 18.5% to 11.4%, respectively).Technical infrastructure issues emerged during the topic modeling of social media data; however, tweets discussing technical issues of water infrastructure were often intertwined with political criticism.For example, 'We asked…and asked…politicians, reps, and aids WHAT about the WATER!? What are we gonna do when the winter hits an area already battling infrastructure problem has frozen pipes??' (UgliWorld 2022).This emphasizes that people do not perceive technical infrastructure issues as a single main factor resulting in JWCs, but instead see that the provision of service is influenced by other financial, political, and external factors.

Understanding the public's lived experiences during the JWCs
The lived experiences of end-users provide valuable insights about the shortcomings of water systems (Araya et al 2020).Understanding how end-users interact with their water systems and how they are impacted by service challenges can shed light on the ways people engage with the political processes that govern water-sector decision-making.This knowledge can help reduce uncertainty about who is affected, how they experience the impacts, and which users have the potential to influence positive change.Also, it is crucial to take into account public lived experiences, such as issues with water quality and quantity, challenges in accessing alternative water sources, and financial problems that may exacerbate existing household vulnerabilities.
Across data sources, challenges to acquire bottled water (e.g.waiting in long lines), including the financial burden to do so, were discussed.For instance, asking for help was an emergent topic in social media data.Many of these tweets referred to bottled water (e.g.'Why are we not talking about Jackson?The city has run out of bottled water to give to residents.It's the largest city in Mississippi, and it's 80% Black.Their water system has failed because of years of neglect.Water is beyond a basic need.We are neglecting our neighbors.';Basile 2022), displaying the need to understand and address bottlenecks in water service delivery.Our study aligns with previous work emphasizing the need to effectively allocate alternative water sources given that drinking bottled water is a key protective action (Lindell et al 2017, Ezell et al 2021).For example, the shifting and closing of water distribution sites after the Flint water crisis resulted in challenges for residents, leading to skepticism about certain water resources and causing people to reject government measures (Ezell et al 2021).Such issues were also present during the JWCs.For example, one excerpt mentioned that '…Some waited more than 2 and a half hours, only to reach the front of the line and be told that the water was all gone.' (Waldrop 2022).Effective management of alternative water sources can alleviate burdens on residents and help improve community resilience during water crises.However, the relative frequency of the challenges to access alternative sources of water code was consistent in both JWCs, highlighting that barriers in bottled water distribution during the first crisis were still widely discussed during the second.
Although the availability and accessibility of water, particularly bottled water, presented social challenges for residents, there were also financial implications.The cost of water was more frequently mentioned during the second crisis, indicating that while it may not have been a primary concern initially, the recurrent need for bottled water may have been a more significant worry for households over time.Some residents expressed concerns about the ongoing costs, saying, 'we're on a budget, we have to go buy water all the time.' (Williams and Reily 2022).In addition to the cost of alternative water sources, our data revealed other consequences like the closure of local businesses.For example a resident stated, 'So my main income is washing laundry.And so, I didn't have water for two days.So, when the water stopped, my income stopped, essentially.But I also deliver food.But the water crisis has affected the restaurants here.So, I have to go out farther to make some type of money and deliver food.' (Sanchez et al 2022).
Among the technical experiences, water quality and quantity problems were prominent, as expected.Both of these codes emerged as topics in the first JWC, along with boil notices as a separate topic in both crisis periods.Numerous tweets emphasize the recurring nature of boil water alerts, which serve as a shock and stressor for the community, as demonstrated by one tweet: 'I'm back in Jackson, Mississippi, which has had 'Boil Water' alerts for so long that there are products like this 'Welcome to Boil Water Alert Mississippi' shirt.' (Baggett 2022).Previous research on water crises has underscored the safety issues arising from prolonged boil water advisories, emphasizing the importance of effective communication and clarity in issuing such protective actions (Lindell et al 2017, Odimayomi et al 2021, Tomko et al 2023).Statements like 'I didn't know there was a boil-water alert' (Cohen 2022a), coded under the boil-water notices theme in this study, reveal issues communicating boil water notices.Our finding, along with previous research, shows the need for communication-focused studies during water crises.Water quantity was linked to people struggling to meet basic needs, such as flushing toilets, cooking, brushing teeth, and showering.The water quality parent code encompasses various aesthetic and health-related factors.Additionally, the water crisis created pressing public health concerns within the community.People expressed worries about potential exposure to contaminated water, as seen in quotes such as 'I spend most of my shower looking away so I'm not accidentally ingesting water' (Williams and Reily 2022), and concerns about their children's health, with statements like 'I have to keep reminding my younger kids, 'Do not drink the water out of that faucet.'…They're young and they don't understand.' (Phillis and Pettus 2022).Even pets were discussed as one resident mentioned, 'I don't even give my dog the drinking water, it's just that bad.' (Cohen 2022b).
As various studies have shown, water crises can lead to not only physical health impacts but also mental health issues (Nowling and Seeger 2020)-the challenges experienced during a water crisis can overwhelm society (Sobeck et al 2020).Our results align with these findings, as emotional distress emerged as one of the codes in news media during the JWCs.Residents expressed their struggles and concerns, stating, 'We are all very tired that we have to work so hard just to get water, something that should be readily available to anyone and everyone at any time' (Williams and Reily 2022).These findings emphasize the indirect results of water access issues and the need to provide physical and mental health support during crises, mainly prolonged events, which is in line with Sobeck et al's (2020) work acknowledging the relationship between mental health and the Flint water crisis.
The JWCs have led to a loss of trust in the government, utility providers, and the water infrastructure, something that was more evident during the second JWC.In the second event, individuals expressed their lack of trust through statements such as, 'But I don't know how many folks will trust that or how long it will take.If you've lived in [a] place this long where you couldn ′ t trust the water, and they told you all of a sudden, it's safe, you just have to wonder if you can believe it.I won't be drinking it today.Would you?' (Blake 2022).This erosion of trust is concerning as it could hinder future planning and implementing mitigation strategies, similar to what occurred during the prolonged water crisis in Flint (Morckel and Terzano 2019, Nowling and Seeger 2020, Ezell et al 2021).Water providers could use public perceptions about water quality and quantity to anticipate a lack of support among customers and mitigate potential opposition (Yang and Faust 2019).The lack of trust in the government can have significant implications, leading to, for example, population migration (which emerged as one of the social experiences child codes), further exacerbating financial challenges in the city (Morckel andTerzano 2019, Nowling andSeeger 2020).This sentiment is reflected in statements such as, 'The other towns are too close to move to.If something isn't done, there won't be any business in Jackson.' (Williams and Reily 2022).
Despite these challenges, altruism and community bonds emerged as key experiences.In crises, communities often organize to help respond.For example, individual volunteerism, and donations of bottled water and money to support one another were discussed on social media.Similarly, quotes from news media revealed the resilient spirit of the community, with expressions like 'People here are resilient, they are coming together with a level of organizing that is rooted in the same communities that fueled the modern civil rights movement,' and 'I'm lucky that some of my neighbors brought me some water.'(Blake 2022, Schrader 2022).This finding demonstrates that during crises, people tend to place trust in each other and strengthen their community bonds.Decision-makers can leverage this sense of solidarity to enhance their responses to crises and effectively address emergency situations.

Recommendations
Here, we propose several practical and research recommendations that can be used to better respond to future water sector crises.
• Better utilize social media for crisis response in the water sector: Our study displayed how social media can be used to better understand community priorities and experiences.Such information should be collected by utilities and used in emergency situations.For instance, in emergency situations, social media platforms can serve as channels for the public to seek assistance, allowing government agencies and utility providers to promptly address concerns and provide necessary support.Notably, this may be challenging as utilities are often under resourced, especially when responding to disasters.We argue that there is a need to provide communication-focused training and support to utilities to aid in such efforts.• Consider the effects of cascading disasters in emergency planning: Both JWCs occurred due to an initial stress (e.g.winter storm, flooding), but given infrastructure and managerial issues, water outages occurred.
Researchers have studied such phenomena as 'cascading events' (National Academies of Sciences and Medicine 2022).Our study showed that during such cascading events, public perceptions are dynamic and constantly changing.Such trends need to be incorporated into emergency response plans (ERPs) at utilities.Planning should be designed to anticipate scenarios where multiple crises intersect.water systems, a predominately Black community.Residents strongly perceived that race plays a significant role in these crises, as indicated by its prevalence in news media discussions (40.3% frequency) and its emergence as a prominent theme in our topic modeling analysis.Our findings align with previous research that explored water service racial disparities (Horney et al 2020).Beyond allocating resources and ensuring access to clean water, it is crucial to foster community engagement, particularly in underserved areas.This can be achieved by establishing policies that mandate the active involvement of residents in decision-making processes related to water infrastructure (Fletcher et al 2022).By integrating equity as a core component of water operations and considering community input in both routine and crisis scenarios, water utilities can make significant strides in improving service levels for marginalized communities.This approach extends to timely repairs and maintenance in disadvantaged neighborhoods, ensuring that essential services are not compromised based on socio-economic factors (Osman and Faust 2021).• Enhance transparency and communication during water-sector crises: The lack of trust in the government was a major issue during the JWCs.Previous studies also mentioned that water-sector disasters, mainly longlasting failures, often result in residents' distrust in the government, and their utility providers (Morckel and Terzano 2019, Sobeck et al 2020, Odimayomi et al 2021).Such issues may be addressed by enhancing transparency and communication with the public, including timely and accurate information about the status of water outages (Yang and Faust 2019).As suggested in previous literature, water safety education events may be useful to build trust (Odimayomi et al 2021).These initiatives help residents comprehend their roles and responsibilities in ensuring water safety while enhancing collaborative problem-solving skills.

Conclusion
In this study, we conducted an in-depth analysis of public perceptions and experiences during the JWCs, exploring temporal changes.To do so, we employed a qualitative analysis of news articles and topic modeling on social media posts related to the JWCs.Our results display the complex interplay of social, technical, and financial factors that influence water provision.Emergent public perceptions during the JWCs include negative attitudes towards decision-makers, perceptions that race was a contributing factor to the crises, and eroding trust in the water provider.Our results emphasize water-sector inequities present in urban water systems, revealing the need for integrated, long-term solutions that are based on community perspectives and prioritize vulnerable communities.Temporally, our results showed that experiences and perceptions are dynamic and change with prolonged events.For example, financial burdens placed on households were discussed more in the second crisis.Our study underscores the importance of understanding public perceptions and experiences during prolonged water crises.Neglecting these perspectives can lead to significant knowledge gaps regarding service levels, infrastructure challenges, and user satisfaction.Results reveal the importance of studying the mental health implications of water crises in developed countries, with a focus on people's experiences due to water accessibility issues (e.g. increase in stress, anxiety, or depression).The public perceptions and experiences revealed here can inform ERPs in Jackson, as well as other communities facing water crises.For example, access to bottled water was a challenge during the JWCs, especially the need to provide long-term resources.Plans for equitable long-term bottled water distribution should be included in ERPs instead of solely focusing on immediate response.While we focused on public opinions expressed through news media and social media, future research should directly engage community members through focus groups, surveys, and interviews.Findings from this work can be used to design interview protocols and surveys to further understand community members' experiences.

Figure 1 .
Figure 1.Timeline showing key dates for Jackson's water system.

Figure 2 .
Figure 2. Data collection and analysis process.

Figure 3 .
Figure 3.Comparison between emergent topics revealed through social media in each JWC.

Table 1 .
(Hassanein 2022)g dictionary for news media.This is not going to be the last time something like this happens in the United States, because we have an increasingly unpredictable climate, and we have aging infrastructure-and we have widening wealth gaps'(Hassanein 2022) ''Our infrastructure is failing people.It is broken, and you cannot put new things on top of old broken things.' (Blake 2022) Experiences Lived experiences of people during the JWCs.'You gotta get in and start boiling water for everything that you're gonna be using in service.Not only do we have to boil water just to wash dishes, for the bar, for glasses, but there's the $200 or $300 a day in ice purchases, canned sodas, bottled water, things of that nature.' (Judin and Pittman 2022) Social Social issues experienced because of the JWCs, such as people's health, emotions, desire to leave, and community bonds.'So, it's a real inconvenience.And it is a public safety issue, particularly when you're talking about our children, and you're talking about elders, in the community.It's a really bad public safety issue, for not just my family, but for all of our families, in Jackson, Mississippi.' (Blackwell et al 2022) Financial Statements about financial impacts experienced because of the JWCs.'We're on a budget, we have to go buy water all the time.' (Williams and Reily 2022) Technical Statements about technical issues people experienced with their water system.'Monday had no water, yesterday water was usable to flush toilets, and they said we could boil it, but I didn't use it.'(Baker-Jordan 2022)

Table 2 .
Relative frequency of references from news media.

Table 3 .
Emergent topics from social media data based on topic analysis.

•
Explore mental health problems related to water outages in developed countries: We recommend further investigation into the relationship between water access, water insecurity, and mental health in developed countries.Prior work typically focuses on developing countries (e.g.Wutich et al 2020), but our research suggests the need to expand this work in developed contexts.Future studies should delve deeper into this area, contributing to a better understanding of the mental health implications of water-related challenges in developed settings.•Use an equity and social justice lens in research and practice: Our results revealed inequities within Jackson's