Climate change risk index and municipal bond disclosures of United States drinking water utilities

Drinking water utilities across the United States (US) are exposed to current and future climate change impacts that could affect their ability to provide adequate quantity and quality of water to consumers. The types of climate hazards and magnitude of anticipated changes vary among utilities, due to regional physical climatic conditions. Utilities also vary in their ability to prepare for and recover from negative effects of climate hazards, given differences in existing infrastructure systems and financial resources. Understanding how drinking water utilities are exposed and vulnerable to multiple aspects of climate change can identify gaps in climate resilience planning¹. Integrated multi-component quantitative measures of drinking water utility climate exposure and vulnerability are needed to compare different climate risks facing different utilities and to inform public funding and investment strategies.

The extent to which climate change poses risks depends on the interaction of climate hazards (the climate change-induced physical events and changes that may cause loss of life, injury, or other health impacts) with the exposure (the people and assets in harm’s way of those hazards) and vulnerability (the propensity of those people and assets to be adversely affected) of human and natural systems^2,3. Most current publicly-available climate service tools do not consider exposure and vulnerability in measures of climate risk, resulting in ineffective assessments of how prepared a system is to manage climate hazards⁴.

Climate change presents physical risks to drinking water utilities; hazards such as extreme heat, extreme precipitation and the associated flooding, sea level rise, wildfires, and drought can affect water resources, infrastructure, and utility operations^{5,6,7,8,9,10,11,12}. Increases in drought alone are projected to disrupt sustainable water supply for utilities around the world¹³. Physical climate risks are starting to disrupt US utility operations and finances, including higher operating expenses, loss of revenue from service interruptions, and higher capital expenditures^14,15. Some utilities, including many associated with the Water Utility Climate Alliance, have conducted robust climate impact assessments on their individual systems and offer industry guidance based on their findings^16,17,18,19, but there are many utilities still unaware of how climate change might pose risks to their systems²⁰. The risk facing utilities varies across the country, given differences in climate hazards, exposure, and vulnerability, as well as differences in adaptive capacity, or the ability to respond to changes^21,22. A national screening assessment tool will allow for comparisons of climate risk and provide insights to utilities that have not conducted extensive system-specific assessments. Current drinking water utility performance frameworks use multidimensional approaches to measure vulnerability^23,24,25, characterize system capacity²⁶, and identify affordability and quality outcomes²⁷. There are also numerous individual utility performance indicators that assess specific operational, financial, environmental, and social dimensions of these systems²⁸. While future climate risk is a motivation for developing some of these multidimensional frameworks, no publicly-accessible, generalized index-based assessment currently captures how susceptible US drinking water utilities are to climate change²⁹.

Here we build on existing definitions and frameworks of climate exposure and vulnerability and apply them to the drinking water sector from the perspective of the utility system. Operationally, a utility’s climate exposure includes the utility’s workforce and assets that are susceptible to negative effects from climate hazards. A utility’s climate vulnerability is defined as its propensity to be adversely affected by the climate hazards (i.e., the utility characteristics that make it vulnerable to the climate hazards). Water system vulnerability has social, physical, and financial dimensions²⁵. This work uses a composite index approach to develop measures of utility climate exposure and vulnerability. Combining multiple indicators into a single index^30,31 has been used in other climate change risk and vulnerability metrics, but this approach has not previously been applied for drinking water utilities^32,33.

An effective measure of drinking water utility climate risk that includes hazards, exposure, and vulnerability can be used to identify gaps between projected climate risk and climate risk disclosure (i.e., which utilities are not adequately planning for their projected climate risk). Drinking water utility municipal bonds offer unique insights into current utility financial and operational weaknesses, as well as status of climate risk disclosures^{34,35,36,37,38}. While there are mixed findings about whether physical climate risk has resulted in lower bond ratings, some municipalities have missed bond payments and others have been placed on credit watch in the wake of experiencing climate-related disasters^39,40. Rating agencies consider a utility’s ability to manage impacts of natural hazards in creditworthiness^25,41 and research has found that extreme weather has led to higher cost of capital for water utilities³⁷. To understand how financial markets are assessing utility climate risk, we analyze drinking water utility bond official statements and compare these to the developed measures of projected climate risk.

The objective of this research is to identify drinking water utilities facing climate change risks that might require targeted assistance in climate resilience and adaptation planning. To achieve this, we (1) compile drinking water utility financial and operational data to develop climate exposure and vulnerability indexes, (2) analyze trends in climate risk awareness within drinking water municipal bond official statements, and (3) develop a drinking water utility climate risk index to compare risk across utility geographic location, source water supplies, and population sizes served. Although the primary focus is on climate risks to drinking water utilities, we also consider the perspectives of other decision-makers, including utility customers, bond purchasers, and government agencies. The produced dataset of climate risk index values for 1455 municipal drinking water utilities from across the contiguous US offers a model for how all drinking water utilities should conceptualize climate risk and helps identify utilities in need of climate adaptation and resilience planning assistance.

Drinking water climate exposure and vulnerability indexes

It is challenging to compare climate vulnerabilities among utilities and validate comparisons as there are few databases with publicly accessible information²⁵. Existing drinking water vulnerability and performance indices, which do not address climate risks, have been developed using regional datasets or case study approaches^25,26,27. Our goal was to develop open access drinking water utility vulnerability and exposure indexes for as many systems as possible in the United States, given data availability limitations.

Nine indicators were drawn from three different public databases (Table 1 and SI Fig. S1) to capture dimensions of climate vulnerability and exposure of drinking water utilities. These include five utility system characteristics chosen to capture dimensions of vulnerability and four chosen to capture dimensions of exposure. The choices of indicators and utility system characteristics were informed by data availability, and data limitations restrict the number of utilities assessed in the analysis. To enable comparison of the climate risk index with information from municipal bonds, the analysis was further limited to municipal water utilities that have issued a municipal bond in the past 10 years. See Methods for additional details about the data sources.

Distributions of the vulnerability indicators (Fig. 1, left) reveal the relative numbers of utilities experiencing each aspect of vulnerability. The wide distribution for infrastructure age indicates that some utilities are older and others are younger, while the heavily right skewed distribution for water quality health indicates most utilities are currently in compliance with regulations. There is a minimal correlation between physical system health and financial health (R² = 0.25), indicating that, for this subset of utilities, a utility’s finances do not guarantee a reliable physical infrastructure system. The five indicators were combined into an overall vulnerability index (see “Methods” section for more details). The map in Fig. 1 identifies the utilities with combined vulnerability indexes at the 90th percentile or higher. Around 70% of the most vulnerable utilities are located in the Upper Midwest, Ohio Valley, and Northeast. This is due to older infrastructure, slightly lower operating ratios, and in the Upper Midwest and Northeast, higher drinking water non-compliance scores. For these most vulnerable utilities, vulnerability is mainly concentrated in one dimension (i.e., poor physical system health or poor financial health), with around half of the utilities (48.5%) showing high physical vulnerability, 28.3% showing high financial vulnerability, and 26.1% high in both dimensions of vulnerability. Many of these utilities, including Toledo, OH; Troy, NY; and Saginaw, MI, have been identified previously as vulnerable utilities by other composite indices using different indicators^25,27.

Most utilities have overall lower exposure to climate change compared to a few highly exposed systems and the distributions of the four exposure indicators (Fig. 2, left) are all very narrow, indicating limited variations between utilities. The four exposure indicators were combined into an overall exposure index (see “Methods” section for more details). The map in Fig. 2 shows that the utilities with relatively higher exposure (90th percentile exposure index) are distributed across the country; every region has some exposed utilities. Many of the most exposed utilities have high per capita employment rates. Texas has a large portion of the most exposed utilities with high per capita service connections or service areas. Some cities, including Hillsborough, CA and Ferndale, MI, have high annual costs of service and thus larger exposure of their finances if disruptions to the system were to occur.

Combined climate risk scores

Drinking water utility climate risk is calculated from the product of the individual exposure and vulnerability indices, as well as mid-century climate change projection hazard index values (see Methods section)¹. The distribution of the combined risk index (Fig. S8) is right skewed; a few utilities have comparatively higher climate risk. The hazard index values are larger on average than the other indices, meaning this component has a greater influence on the combined risk index. This is an accurate representation of how climate risk will manifest in drinking water utilities; most will experience changes in climate hazards, while few have notably higher exposure or vulnerability. This is an important property of this risk model because the changes in climate hazards are external stressors that might be unknown to the system, compared to other system financial or infrastructure characteristics that the utility has more ability to affect.

Drinking water utility climate risk is geographically distributed across the US (Fig. 3). The regions with the highest average risk are the Northeast and West (Fig. S9) but the South, Upper Midwest, and Ohio Valley also have large numbers of utilities classified as higher risk (Fig. S10). While Western US regions are likely to see comparatively larger changes in climate hazards, smaller magnitude changes in hazards in the Northeast, South, Upper Midwest, and Ohio Valley coupled with existing vulnerabilities and exposure, means these utilities are also at risk. These utilities could experience more difficulty recovering from climate change hazards, given the lack of current adaptive capacity. Surface water systems have a higher average climate risk score than groundwater systems (Fig. S14) and a slightly larger portion of utilities that use surface water (27%, 264 out of 991 systems) are classified as having high climate risk, compared to those that use groundwater (21%, 93 out of 442 systems), see SI Fig. S15. Utilities serving more than 1 million customers have the highest average climate risk scores, likely due to their proximity to the coasts, though overall there are minimal differences in average climate risk score by population size served (Fig. S16). The utilities with high or moderate climate risk scores serve ~67 million customers.

One indicator from each of the vulnerability, exposure, and hazard indices was omitted (infrastructure age, service costs, and sea level rise projections) to assess how sensitive the climate risk index is to the included indicators. Infrastructure age (represented through a proxy of median housing age) was included in the index to capture the health of infrastructure assets. There is regional variation in housing age across the country so removing this indicator from the index had an effect on regional averages (Fig. S11). The Northeast, Upper Midwest, and Ohio Valley had comparatively lower risk values when infrastructure age was not considered, though the modified index was strongly correlated with the original index (R² = 0.80). The service cost of delivering a set amount of water per household was included in the index to capture utility financial exposure. Service costs also vary widely across the country, and there were numerous utilities included in the sample with missing cost data (regional medians were used instead). Omitting service costs had a minimal overall effect on the risk index (correlation between modified index values and original was significant, R² = 0.88). Higher service costs in California meant the West region had a lower average risk index value when this indicator was omitted (Fig. S12). Sea level rise was included in the original hazard index because it can affect groundwater systems and buried infrastructure. Omitting sea level rise had a minimal effect on regional averages of the risk index, as so many utilities are unaffected (R² = 0.96), see Fig. S13.

Gaps in water utility climate risk understanding

The developed index identifies which utilities face the highest risks from climate hazards but does not account for any actions the utility might have taken to address climate change and reduce their risk. Twenty-three percent (378 out of 1611 utilities) of all the analyzed municipal bond official statements mentioned climate change, though a higher portion of revenue bonds (33.0%) compared to general obligation bonds (19.7%) mentioned climate change. A similar fraction (30%) of all municipal offering statements mention climate change⁴². This could indicate a recognition by municipalities that repayment based on income from service rates is more susceptible to climate hazards than repayment based on taxes, or there could be a confounding size effect as larger utilities might be more likely to issue revenue bonds than general obligation bonds. Prior work also found that many drinking water utility managers do not consider their systems to be at risk²⁰. Looking at trends in population served and geographical region (Figs. S17 and S18), utilities serving larger population sizes and located in the West or Northwest disclosed climate risk more often than other utilities. Climate change mentions are becoming more common, with 30% of the utilities discussing climate change within their bond statements in 2023 and 2024 (Fig. S19), compared with only 9.1% of utilities in statements from before 2020.

Utilities that mention climate change in their municipal bond official statements have similar distributions of bond ratings (Fig. S20) and climate risk (Fig. S21) when compared with those that do not disclose such climate risk (i.e., there are no discernible effects of mentioning climate risk within an official statement on having a better bond rating). Similarly, there is no observed relationship between the drinking water bond ratings and climate risk disclosures (R² = 0.12) and a slightly negative relationship between bond ratings and climate risk index values (R² = −0.39). This aligns with other research findings that physical climate risks are largely not yet factored into municipal bond markets or ratings^43,44.

A comparison of calculated climate risk index values and climate change awareness identified high risk utilities that might not be aware of their risks. This category of utilities are good candidates for policy interventions that educate systems about all potential links between climate hazards and utility functions²⁰. A larger proportion of utilities classified as having high climate risk mentioned climate change in bond statements, compared to those with lower overall climate risk (Fig. S22). This could indicate current guidance and recommendations are successfully being communicated to the highest risk utilities. Similarly, a slightly higher percentage of utilities exposed to multiple types of climate hazards (27.3%) mentioned climate change in their bond statements, compared to only 22.9% of utilities that are not projected to exceed multiple climate thresholds (Fig. S23).

Of the 146 drinking water utilities within the 90th percentile of climate risk scores, a majority (65%, 95 utilities) did not mention climate change in their official bond statements. Most of these utilities are located in the South (the majority of which are located in Texas), Ohio Valley, or Midwest. Of the high-risk utilities that did mention climate change, most are located in the West (mainly California). About half of the most at-risk utilities in the Northeast mentioned climate change in their bond statements and half didn’t; there is no discernable difference in the characteristics of either group of utilities.

Along with overall mentions of climate change, the frequency with which the drinking water utilities mentioned specific hazards (drought, flooding, wildfires, extreme heat, and sea level rise) within their bond statements was considered. Of the identified climate hazards, flooding risk was mentioned most frequently, followed by drought (Table S8). Some of the hazards were mentioned more frequently alongside mentions of climate change; most of the bonds that mentioned extreme heat and sea level rise also mentioned climate change (94.9% and 88.6%, respectfully) while only around half of the bonds that mentioned drought or flooding also mentioned climate change (Table S9). This could indicate that utilities perceive extreme heat and sea level risk as climate-related while drought and flooding are less climate-related. The utilities that mentioned drought and wildfires are projected to see larger changes in these hazards, indicating an alignment of risk perceptions and actual risk (Fig. S24). The opposite was observed for flooding and extreme heat (i.e., the systems that mentioned these hazards have lower projected changes in these hazards), indicating drinking water utilities might need more targeted guidance about whether and how extreme precipitation and extreme heat can affect their system’s reliability. There might be institutional understanding gaps about which hazards climate change will exacerbate and which of these pose the greatest risks to system reliability. Historical flood and drought may bring these more to mind than slower changes associated with increasing extreme heat.

The overall climate risk index captures the system’s risk while their bond official statement captures their risk awareness; combining both reveals which utilities are potential  candidates for climate adaptation or resiliency planning assistance (Fig. 4). Utilities classified as Low Priority are those that have lower climate risk index values and mentioned climate change within their bond statements, indicating that they might be taking proactive actions. Those classified as High Priority have higher climate risk index values and did not mention climate change in their bond statements. The portion of analyzed utilities classified as High Priority systems (36%) reflects an important misalignment between risk awareness and risk disclosure. Many of the Low Priority utilities are suburban and in close proximity to others with the same classification (for instance, the suburbs around Nashville, TN and Chicago, IL), compared to the higher priority utilities. Many of the High Priority utilities serve smaller population sizes; the average population size served by a High Priority system is around 52,000, compared to 137,000 people served by a Low Priority system. A larger fraction of smaller systems is classified as High Priority systems (Fig. S25). This is notable, as smaller utilities have lower adaptive capacity^21,22 and the analysis does not include even smaller utilities (those serving fewer than 10,000 customers). The utilities omitted from this analysis likely experience similar climate hazards to those included in the analysis that are geographically close to them, but more information is needed about their exposure and vulnerability. Omitting these utilities biases our findings toward larger systems, which may have more resources and are more often in urban areas. This bias may mask additional risk associated with small systems, which may have fewer resources and be found disproportionately in rural areas.

Validating the results of this work is challenging, as utilities will likely not experience the full effects of climate change hazards for multiple decades. However, recent extreme weather events that have disrupted drinking water utility operations and caused financial losses can provide context to the identified classifications. For instance, in 2022, a flooding event coupled with existing system vulnerabilities left 180,000 residents in Jackson, MI without drinking water⁴⁵. This work classifies Jackson, MI as a High Priority system.

US drinking water utilities are at risk from climate change, regardless of location, size, and source water, and utilities with higher climate risk scores serve around 67 million customers. While the climate risk index was developed using a sample of only 1455 municipal drinking water utilities, the simplicity of the model and accessibility of input data allows any US drinking water utility to calculate its own climate risk and compare it to other utilities. Drinking water utility managers should gather their own system’s data for the indicators from Table 1, assess climate projections for their system¹, calculate their system’s climate risk index using Eqs. 1 and 2, and produce these assessments for the public. The developed index is intended to be useful in comparing climate risks facing utilities across the country, but the most accurate picture of actual risk requires in-depth climate impact assessments by individual systems.

A primary limitation of this work is that, due to data availability, it was developed for municipally-owned drinking water utilities serving more than 10,000 customers that had issued bonds recently. Many smaller systems likely have high climate risks and limited resources to access hazard projections necessary to plan for those risks. Future work should expand the climate risk index to include all drinking water utilities to enable prioritization of planning and financial assistance decisions. This would require public reporting of infrastructure health and financial data for all utilities. Additional studies could also further explore how to overcome institutional knowledge gaps about which climate hazards pose risks to system reliability and how these risks should be disclosed to the public. Although all contiguous US drinking water utilities are projected to experience changes in climate hazards by mid-century¹, this work identifies a subset of High Priority utilities that are potentially misaligned in their technical risk and risk disclosure. These utilities, with high physical risk but no mention of this risk in their municipal bonds, should be prioritized for assistance in climate resilience and adaptation planning. For instance, this prioritization could be directly tied to funding instruments such as Drinking Water State Revolving Funds, which allow states to determine whether and how funding assistance is awarded. While many utilities in the Western US are seemingly aware of their climate risk, utilities in the South, Ohio Valley, Midwest, and to a lesser extent, Northeast, may not be as aware, based on information disclosed in their bond official statements.

While there are limitations to using municipal bond statements to assess climate risk disclosures (see “Limitations” section), the low fraction of utilities mentioning climate issues in bond disclosures (23%) suggests that there is likely a considerable amount of undisclosed financial liability associated with at-risk utilities. The municipal bonds analyzed in this study represent a total of $39.3 billion dollars of debt, most of which will mature within the next 20 years (Fig. S26). Of that total amount, $9.2 billion (23%) has been issued by utilities that we project to have high climate risk and potentially limited risk awareness. This large amount of capital is a financial liability for the issuers and owners of the municipal debt, as well as the community members dependent on the long-term reliability of these drinking water systems. This section offers practical suggestions for how drinking water utility managers, bond purchasers, and governmental agencies can utilize the climate risk index and the findings in this work to increase their climate preparedness.

Ensuring High Priority utilities get the assistance they need in preparing for climate change requires governmental support. Federal and state regulatory bodies should use these results to target climate adaptation planning resources and available funding. The 2021 Bipartisan Infrastructure Law⁴⁶ recommends states incorporate climate resilience criteria into Drinking Water State Revolving Fund programs⁴⁷; the climate risk index could serve as one metric included in that prioritization. Government agencies must play a role in requiring climate adaptation and providing publicly-accessible, high-quality climate risk information. Utilities vary in awareness of how climate change might pose risks to the reliability of their systems²⁰ and in their access to private-sector climate information services⁴, leading to information asymmetry. Dependence on the federal government as a backstop may lead to less local investment in resiliency measures, creating a moral hazard. Thus, it is critical that public sector agencies work to improve risk awareness of drinking water utilities and require adaptation planning.

At the state-level, the presented drinking water utility prioritization classifications reveal which states have the most at-risk utilities (Fig. 5). Texas, the state with the largest number of High Priority drinking water utilities (74) does not have a state-led climate adaptation plan⁴⁸ and has a very low rate of climate change mentions within bond statements (5%). Requiring climate assessments for drinking water utilities at the state-level in Texas could improve climate risk awareness for those utilities. This finding highlights the importance of state-level policies to target gaps in climate planning and adaptation. The prioritization is also useful in assessing which utilities might require less attention. For instance, many utilities in Colorado have high climate hazard index values, but their low exposure and vulnerability index values combined with an awareness of and investment in climate resilience has resulted in extensive climate adaptation planning^49,50.

Bond purchasers and ratings agencies are interested in the financial liability of the debt issued by the most at-risk utilities. While physical climate risks do not yet appear to be factored into bond ratings, the Low Priority utilities have a higher average bond value ($57.6 million) and longer time to maturity (16.4 years) compared to the High Priority utilities ($17.5 million value, 11.7 years). Municipal issuers with relatively lower-value and shorter time-to-maturity bonds could have limited access to credit, which further diminishes their ability to invest in resilience and adaptation measures. Utilities omitted from this analysis because they have not issued a bond in the past 10 years also have a limited ability to invest in large capital projects, highlighting another layer of financial inequalities. Climate risk is more relevant to longer-term municipal bonds, as these bonds will mature closer to when the climate hazards are projected to be experienced. This complements other research findings that long-term municipal bonds are more affected by climate risk compared to short-term bonds^38,44,51.

The total amount of bonds held by High Priority drinking water utilities (Table S10) can offer insights into the total value of risk, though the amounts presented in this work only reflect the most recent bond issuances, not the total amount of issued debt. Six states (Michigan, Illinois, California, Massachusetts, Virginia, and Texas) have High Priority municipal drinking water utilities with total recent bond debt around or above $500 million dollars and an average bond maturity of over 10 years (Fig. 6); the purchasers of these municipal bonds might not be aware of the risks facing these drinking water utilities. Ratings agencies should consider adjusting ratings to reflect climate risk, even for short-term issuances. To more accurately quantify the value of risk facing utilities, the Environmental Protection Agency could require utility managers to estimate capital improvements needed to address future climate hazards within the next Drinking Water Infrastructure Needs Survey and Assessment⁵².

The developed drinking water utility climate risk index is useful for multiple stakeholders, including utility customers, bond purchasers, and regulatory agencies. It provides drinking water utilities with a better understanding of the projected future risk they face and allows them to compare their risk with other utilities serving different population sizes, using different source waters, and located in different regions. Utilities can use this information for planning and decision-making. This climate risk index also allows for more accurate climate risk disclosures, especially in the municipal bond market. Further, the proposed classifications identify which utilities should be prioritized when government agencies provide climate resilience planning and assistance; these utilities are candidates for immediate federal or state support for climate adaptation.

A dataset of mid-century climate change projections and hazard index values for 42,786 drinking water utilities across the contiguous US was used as the core dataset for this analysis¹. This dataset was created using a multi-hazard approach, accounting for changes in extreme heat, extreme precipitation, freeze-thaw cycles, sea level rise, wildfires, and water supply stress, as well as gradual changes in average temperature and precipitation (captured through the proxy of utility energy demand), and drew on climate projections from the U.S. Government’s Climate Mapping for Resilience and Adaptation tool⁵³ and the Climate Risk and Resilience Portal (ClimRR) dataset⁵⁴. The mid-century, Relative Concentration Pathway 4.5 climate emission scenario was used and both sets of projections were derived from Localized Constructed Analogs (LOCA) downscaled climate model ensembles⁵⁵. Existing research analyzing multi-hazard climate risks to drinking water systems outside the US combines hazards with information about utility system infrastructure⁵⁶. In this work, we link service area boundaries for all US community water systems⁵⁷ and system information from the Safe Drinking Water Information System database (SDWIS) for all utilities using Public Water System Identification Numbers. Public water systems are required to annually report water quality violation and enforcement information, as well as facility and infrastructure information to SDWIS⁵⁸. The SDWIS database was further used to compile information about populations served, the number of service connections, and the length and duration of water quality violations (captured through the Drinking Water Non-Compliance Score). Added context about the utility’s financial, environmental, and demographic conditions were acquired from the Municipal Drinking Water Database (MDWD)⁵⁹. This dataset contains demographic information, details about the municipal government structure, and financial measures as reported in the Government Finance Database⁶⁰. The MDWD was used for data about municipal employees, the age of infrastructure, and the prevalence of poverty rates within the utility’s customer base.

While the MDWD includes financial metrics about the overall municipality linked to the drinking water utility, it does not include direct measures of a utility’s financial health or service rates. The Electronic Municipal Market Access (EMMA) database was used to extract water utility financial information from municipal bond official statements for each municipal drinking water system. Many municipal drinking water utilities use municipal bonds to finance capital improvement projects^35,36, and bond statements can offer information about current utility financial and operational weaknesses. Official bond statements are disclosure documents published by municipal issuers that list information about the bond offering, including operating expenses and revenue (including service rates), other financial information, and bond ratings⁶¹. Issued municipal bonds are given ratings by external ratings agencies using holistic and case-based analysis, including information from official statements, that reflects a rating agency’s opinions on the bond’s default risk⁶². Municipal bond official statements have previously been used to collect information about drinking water utility customers and water rates⁶³. We conducted searches for each municipal drinking water system within EMMA and extracted the most recent bond official statement issued by that utility or its associated municipality over the past 10 years. In addition to providing information about utility operating ratios, bond rating, and service costs, the bonds were also used to provide insights into climate risk awareness. This is discussed in the Using Municipal Bond Climate Risk Disclosures to Identify Gaps in Water Utility Climate Risk Preparedness section.

Evaluated drinking water utility systems

We evaluate 1455 of the more than 50,000 community drinking water utilities in the contiguous US. The main data limitation was the use of data from the MDWD, which includes only drinking water utilities serving more than 10,000 people (2219 utilities). Of those utilities, only 1609 were linked to municipalities that issued bonds in the last 10 years. Of the analyzed 1609 utilities, some were missing data (reported as NA) for one or more of the nine exposure and vulnerability indicators needed to create the indices. Some of these NA values were able to be replaced with data from the MDWD or median values, so the final dataset contains 1455 utilities. The indicator with the largest portion of missing data was utility service costs; these were extracted from bond statements and not every utility included service rates. These missing values were replaced with median regional service costs. Omitting this indicator was assessed in the sensitivity analysis. See SI Fig. S1 for details of the dataset restrictions and Table S1 for more details about how the treatment of missing values. Drinking water utilities specifically mentioned in this article were contacted and none provided comments.

The analyzed sample of utilities serves a combined 113.7 million people, or 33% of the total US population. See SI Figs. S2–S4 for details about population served, source water supply, location, and projected climate hazard index values for the utilities included in the sample. In addition to omitting smaller drinking water utilities in the US, this sample has a smaller proportion of utilities served by groundwater and a much larger proportion served by surface water. In terms of regional representation, the sample overrepresents utilities located in the Ohio Valley and Upper Midwest and underrepresents utilities in the Northern Rockies and Plains, Southeast, and Southwest; regions were designated using the National Oceanic and Atmospheric Administration (NOAA) Climate Regions (Fig. S5)⁶⁴. The states with the largest number of utilities included in the sample are Texas, Illinois, Massachusetts, and California (Fig. S6). See SI Table S3 for more details comparing the sample with all drinking water utilities. Despite the limitations imposed by the availability of data in national datasets, the underlying model structure was designed so that individual utilities can use their own system’s data to determine a climate risk score for comparison.

Looking at the municipal bond data, this sample of 1455 utilities has bonds with an average maturity date of 2037, an average issuance of $25.9 million dollars, and an average rating of Aa2/AA/AA (Table S4). Thirty-six percent of the bonds were issued directly by a drinking water utility, while the rest were issued by a municipality that oversees the water utility as a business-type activity. Most of the analyzed bonds were issued in the past 5 years and are general obligation bonds (Fig. S7), in which the issuer guarantees repayment using government taxation power. Bonds issued directly by the water utility are more likely to be revenue bonds that guarantee repayment based on income from service rates.

Climate vulnerability index development

Indicators included in the vulnerability index (Fig. 7) were chosen to capture financial and physical utility system characteristics that affect its propensity to be adversely affected by climate change. Financial dimensions of vulnerability capture the ability of a utility to pay for treatment and delivery of clean water and maintain its fiscal health into the future²⁶, as well as the socioeconomics of the utility customer base, which may affect its ability to fund and maintain system improvements through rate increases. Physical utility system characteristics, or the utility’s infrastructure health, are those that make it potentially vulnerable to the climate hazards identified in the combined hazard index^2,3,25 and those that affect the reliability of the utility’s infrastructure assets. The intersection of physical and financial aspects affects the ability of the utility to invest in repairs or prevent infrastructure failure.

Three factors are included as indicators of utility system financial health: utility operating ratio, municipal bond rating, and percentage of served population below the poverty line. A utility operating ratio, or the ratio of operating revenues to operating expenses, is used in assessments of drinking water financial capability and capacity because they capture a utility’s cost recovery. Ratios greater than or equal to one indicate the utility can cover expenses through revenue sources^26,65.

Bond ratings are also included as an indicator as they measure perceived risks of a utility’s creditworthiness, or the reliability a system can repay debt. Most utility infrastructure financing relies on the bond market and a change in rating can signal management or financial issues facing a utility^35,63,65. The nomenclature of ratings and results vary slightly across rating agencies but take the form of very highly rated bonds with the lowest risk (e.g., AAA) to bonds with the highest risk of default (e.g., C), see Table S2. Lower municipal bond ratings can affect borrowing costs, limiting capital investment programs⁶³. While bond ratings are not commonly included in drinking water vulnerability assessment due to data limitations, they have been used in other analyses of utility financial health⁶³.

Poverty rates are an important inclusion in social vulnerability indices⁶⁶ and previous assessments of drinking water-specific vulnerability identified the influence of a utility’s social vulnerability, captured through poverty prevalence, on the utility’s financial vulnerability^25,26. Utilities with high populations served below the poverty line have challenges raising rates to increase revenue without causing affordability issues or increasing numbers of delinquent accounts⁶⁷.

Physical dimensions of water utility vulnerability capture the ability of a utility’s infrastructure system to maintain service given changes in severity, frequency, or duration of climate hazards. The health of a utility’s physical system depends on asset conditions; infrastructure in poorer conditions is more likely to be damaged when exposed to hazards. Age of the infrastructure system is used to capture infrastructure condition, as older infrastructure is closer to its end-of-life and is more likely to be compromised by climate hazards⁶⁸. Exact ages for different infrastructure components are not commonly accessible, so median housing age in the utility service area is used as a proxy for infrastructure health. Comparing median housing age data to pipe age data⁶⁹ for a subset of the data shows a correlation (R² = 0.63), with the housing age about 4 years older than the pipe age, on average. A similar validation was used and found by Hughes²⁵.

Another aspect of a utility’s physical system health is how well it meets drinking water quality standards. Complying with water quality regulation violations is an important performance outcome for utilities²⁶, and violations can reveal underlying limitations in a system’s ability to properly treat drinking water. From a utility management perspective, violations of the Safe Drinking Water Act reflect existing operational vulnerabilities. Violations can be health-based, revealing exceedances of maximum contaminant levels for contaminants or failure to meet treatment technique requirements, or non-health-based, revealing failures in reporting or public notification). While the health-based violations reveal the largest vulnerabilities in a system’s ability to produce safe water, non-health-violations are also important inclusions in drinking water vulnerability assessments²⁶ because they may indicate administrative capacity limitations. The Drinking Water Non-Compliance Score is included as an indicator because it accounts for the severity and duration of both health and non-health water quality violations⁷⁰. This indicator is used to capture the challenges a system has complying with water quality regulations; higher scores indicate more challenges faced by the utility⁷¹. The two physical system health indicators, Drinking Water Non-Compliance Score and infrastructure age, were averaged to create one physical system indicator, that was then combined with equal weighting with the financial health indicator for the vulnerability index.

Climate exposure index development

Climate exposure is defined from the perspective of the water utility; it is the people and assets of the water utility that could be negatively affected by climate hazards^2,3. A utility is considered more exposed to climate change if it has more assets on a per capita basis (i.e., a utility has more exposed assets when delivering water to the same number of people). The indicators included in the exposure index (Fig. 8) capture the different facets of a utility’s assets: infrastructure, workforce, and finances.

The infrastructure assets the utility owns include the distribution system, pump stations, treatment plants, and any storage facilities. While utilities have extensive information on their facilities, this information is not gathered in any public centralized database for all US utilities. Limited information, including the number of treatment plants and the number of service connections are available from SDWIS. The number of service connections per capita and the service area per capita, considered together, are used as a proxy for total exposed assets (i.e., how much more infrastructure the system maintains on a per capita basis). As the size of the distribution system affects economies of scale for the utility^72,73, the service area per capita is used to capture the relative length and complexity of the utility’s infrastructure. The service connections per capita captures the relative exposure of the utility’s service lines²⁷. The Climate Vulnerability Index, a measure that characterizes vulnerability and exposure generally, uses a similar per capita for lane miles to capture transportation infrastructure exposure⁷⁴. There are limitations to using this proxy measure for total exposed assets, namely that the exact location and size of utility infrastructure assets affect exposure as well (i.e., a large treatment plant in a flood zone is more exposed than a smaller treatment plant outside a flood zone). This index is intended to provide insights for the entire country, but individual utilities should use more detailed information about their facilities to properly assess their individual exposure. Furthermore, while source water supplies can affect a utility’s exposure to different climate hazards^75,76, this index does not distinguish between surface water and groundwater systems. Differences in source water were considered in analyzing the trends in climate risk index values.

Also included in this exposure index are utility personnel; these personnel are susceptible to negative consequences of extreme climate hazards like heat waves and flooding events. Drinking water utilities have identified their employees as exposed to these climate hazards¹⁸. The number of employees per capita was used because a utility has increased workforce exposure if it has more employees per capita. As each utility’s total personnel is not publicly available, the number of municipal employees per capita was used as a proxy. Comparing the number of municipal full-time employees to drinking water utility full time employees for a subset of Wisconsin utilities⁶⁹, there was a strong correlation (R² = 0.85) between the values, with the overall municipality having about 20 times as many employees as the utility, on average.

Finally, the utility’s financial exposure can be captured in the cost of delivering a set amount of water per household. Utilities generate revenue through service rate charges and the affordability of these rates has consequences for how much revenue the utility earns⁶⁵. Comparing household service charges reveals how financially exposed a utility is, were a climate hazard to interrupt service or reduce a customer’s ability to pay their water bills. The household annual cost of service was calculated using water rates reported in utility bond official statements, assuming 2.65 people per household each consuming 50 gallons per day⁷⁷.

Not included in this combined exposure index is the exposure of the customers the system serves. These customers are exposed to the effects of climate hazards on their drinking water utility, and they may experience negative effects of those climate hazards, depending upon the actions the utility takes to mitigate those effects. However, the scope of this work focuses on exposure of the water utility itself.

Combined climate risk index development

Composite indices for drinking water climate exposure and climate vulnerability were combined with the previously developed drinking water climate hazard index database¹. The composite index approach captures multidimensional facets and assumes each indicator component has a distinct influence on the combined index⁷⁸. The different indicators (({X}_{i})) were normalized (scale 0 to 1) using a min-max normalization^79,80. This approach is minimizes the effects of extreme values^32,79, and has been used in other drinking water and climate indices^81,82, see Eq. 1.

The combined exposure and vulnerability indexes were then aggregated and divided by their number of components³⁰, see Eq. 2. The indicators were all weighted equally^31,83, which is the best choice when there is a lack of theoretical justification for a different weighting scheme^31,83. Other drinking water and climate risk indices also used this weighting scheme^25,32. The final combined climate risk index was calculated using the product of the exposure and vulnerability indices, along with the hazard index¹, see Eq. 2. This formulation follows IPCC frameworks for modeling climate risk^2,3 and has been used in other approaches that developed separate hazard, exposure, and vulnerability indices,  assigned equal weights, and then multiplied them together for a composite index of climate risk³².

Sensitivity analysis

Composite indices are useful for modeling multidimensional system properties that are not directly measurable, but decisions about their construction, including which indicators to include and how they are combined, affect whether the results are meaningful^83,84,85.

The min-max method used to normalize the indictors reduces the effects of outliers because all indicators have identical ranges^32,79. The linear aggregation method used to combine the indicators assumes compensation, or that one indicator’s lower can be compensated by another’s higher value^31,86. An important consideration when constructing composite indices is whether the individual indicators are related, as the assumption is each indicator has a distinct effect. Different aspects of utility vulnerability are expected to be related²⁶, and socially vulnerable communities experience increased risks of drinking water quality violations⁸⁷. Indicators were chosen to capture different facets of vulnerability and exposure, and indicator collinearity was tested to assess whether the individual indicators are independent of one another⁷⁸. Low collinearity was observed between the indicators (SI Tables S5 and S6), suggesting the selected indicators represent different aspects of exposure and vulnerability, and that they can be used together.

The climate risk index was developed from the hazard, exposure, and vulnerability indices, including a total of 16 indicators. Correlation analyses were performed to examine the relationships between each indicator component and the composite index (SI Table S7). The overall vulnerability index and exposure index are moderately correlated with the risk index (R² = 0.75 and R² = 0.62, respectively) while the hazard index is not correlated (R² = 0.20). Infrastructure age (captured by the proxy of median housing age) is the individual indicator with the strongest correlation with the risk index (R² = 0.66). The exposure, vulnerability, and hazard indices have low multicollinearity between themselves, indicating each is contributing a unique aspect to the overall climate risk index.

The sensitivity of the climate risk index to the included indicators can be analyzed by assessing how results change when certain indicators are omitted. Indicators from each of the vulnerability, exposure, and hazard indices were omitted (Infrastructure Age, Service Costs, and Sea Level Rise) to analyze how the climate risk index changed. Infrastructure age was chosen because while median housing age has been used as a proxy for infrastructure age in other work²⁵, a utility’s pipes might be considerably older or younger than the housing stock and age itself is only an indirect measure of asset health. Service costs were chosen because this indicator had the most missing data (median service costs within a NOAA Region were used instead). Sea level rise was chosen for the hazard index because of its difficulty to predict⁸⁸ and because it only affects coastal utilities (an uneven regional distribution).

Using municipal bond climate risk disclosures to identify gaps in water utility climate risk preparedness

The developed water utility climate risk index captures the determinants of risk (hazard, exposure, and vulnerability) but does not include any utility actions aimed at addressing climate change. The actions a system takes in response to climate change are also important drivers of overall risk⁸⁹. While research has found a lack of climate adaptation planning occurring at drinking water utilities²⁰, climate risk disclosures from bond official statements offer important insights about which utilities have acknowledged potential risks of climate change. Issued municipal bonds are given ratings by external ratings agencies using holistic and case-based analysis, including information from official statements, that reflects a rating agency’s opinions on the bond’s default risk⁶². Rating agencies have begun to consider physical climate risks to municipal debt issuers when assigning bond ratings^90,91, making bond official statements an important source of information for climate risk disclosures. S&P Global rates water and sewer utilities as having strong identification of operation risks if “climate risk assessment is incorporated into planning and operations as a potential risk to the system.”⁹⁰ Utilities may score as vulnerable if they have not explicitly addressed climate risk either in plans or operations. There are mixed conclusions about whether physical climate risk is actually linked to lower bond ratings. Some research has found links between increased exposure to certain climate hazards (like sea level rise) and lower municipal bond ratings and higher borrowing costs^38,44,51. Research focused on municipal water bonds found that specific hazards like droughts and floods were factored into bond markets³⁷. Other work has found that the full scope of physical climate hazard risk has not yet been priced into the municipal market⁴³. Regardless, professional organizations recommend that water utilities provide information to bond rating agencies about climate risk and adaptation planning⁹², making bond official statements a useful source of information about their perceptions of climate risk.

In addition to using municipal bond official statements to extract drinking water utility financial information (described previously), these statements were also searched for mentions of climate change and resilience, as well as mentions of specific climate hazards, including drought, flood, wildfire, heat, and sea level rise. Given that rating agencies explicitly consider physical climate risks when assessing their ratings^90,91, these mentions were assumed to capture utility awareness of and possible responses to climate change. While the lack of mentions was assumed to indicate less awareness or concern from the utility about climate risk, utilities might have other reasons for not explicitly using the term climate change, which are discussed further in the Limitations section. The climate risk awareness identified in the bond official statements was also compared to and combined with the calculated climate risk index scores for each utility to reveal which utilities might be uninformed about their future climate risk and which might need targeted climate adaptation assistance.

Limitations

A primary limitation is the drinking water utility sample analyzed in this work; focusing on municipal utilities that serve more than 10,000 customers and have issued a municipal bond in the past 10 years limits the applicability of the complete model to utilities serving smaller population sizes. The larger the system, the greater its adaptive capacity, or ability to respond to changes^21,22. While utilities serving fewer than 10,000 customers might have lower exposure (i.e., fewer assets and smaller systems), these systems likely have much higher existing vulnerabilities (i.e., more susceptible to disruptions) because of lower adaptive capacity. Smaller utilities have more drinking water quality violations and less capacity to comply with regulations⁹³. Omitting smaller systems from this analysis likely underestimates the number of systems nationwide that have higher climate risks and that should be prioritized for adaptation assistance. Additional analyses will be needed to prioritize assistance among these smaller systems.

Many smaller systems are groundwater systems, and an additional limitation of this work is that the index does not distinguish between surface water and groundwater systems. Source water supply type can affect exposure to different climate hazards, but there is limited work quantifying how climate hazards affect the operations and infrastructure of groundwater systems differently from surface water systems. Future work could model how certain climate hazards damage infrastructure and disrupt operations differently. Segmenting the sample to drinking water utilities that have recently issued bonds also introduces biases, as these omitted utilities might have decided against bond issuance because of an underlying financial or operational vulnerability. A larger portion of the utilities that have not issued bonds (1.5%) are in the 98^th percentile for WQ non-compliance scores than the utilities that have issued bonds (0.5%).

There are also limitations associated with using mentions of climate change within municipal bonds to identify risk awareness and preparedness. “Climate change” as a term has a range of associated political framings and misunderstandings⁹⁴, which influence use of the term. Political affiliations and personal experiences of extreme weather affect perceptions of climate risk^95,96,97 and there are regional differences in public understanding of whether global climate change is happening and whether it could harm people in the US^98,99. Given these differences, some municipalities and utilities might be hesitant to explicitly use the term in their bond official statements even while rating agencies are requiring a consideration of these physical climate risks when assigning ratings^90,91. Lastly, many ratings agencies are focused on short-term creditworthiness and utilities might have disclosed longer-term climate risks or concerns directly to ratings agencies and not explicitly mentioned them in official statements. Searching the bond statements for explicit mentions of certain climate-related hazards (drought, flooding, heat, sea level rise) alongside explicit mentions of climate change revealed only a small fraction of the bonds that mentioned climate change didn’t mention one of the hazards (3.2%) and around a third (27.3%) of those that mentioned a hazard didn’t mention of climate change.