Research LetterOpen Access

The Human Right to Water: A 20-Year Comparative Analysis of Arsenic in Rural and Carceral Drinking Water Systems in California

Published:CID: 097701 by:2


Access to safe drinking water is considered a universal human right.1 In the United States, exposure to arsenic contamination in drinking water disproportionately impacts small, groundwater-reliant communities and communities of color.2,3 These inequities are driven by a combination of natural, built, and sociopolitical factors.4 The United Nations calls upon states to especially safeguard the right to safe water for groups that may face difficulties exercising this right, such as incarcerated people.1 Limited research exists on water quality in prisons; however, prisons in the Southwestern United States have elevated arsenic concentrations compared with other community water systems (CWSs) in the region.5

Inorganic arsenic is an odorless, colorless carcinogen, common in California’s San Joaquin Valley groundwater.3,6 In 2001, the U.S. Environmental Protection Agency lowered the maximum contaminant level (MCL) for arsenic from 50 micrograms per liter50μg/L to a running annual average of less than 10 micrograms per liter<10μg/L.7 This stronger standard went into effect in 2006. In 2012, California passed its Human Right to Water bill (Assembly Bill 685) mandating safe, affordable, and accessible water for all.

In this article we present a comparative analysis of 20 y of data (2001–2021) on arsenic concentrations in the CWSs serving Kern Valley State Prison (KVSP) and three neighboring rural communities: Allensworth, Delano, and McFarland. Our objective was to better understand trends in water quality, compliance, and treatment following adoption of the revised arsenic MCL and to elucidate differences, if any, between neighboring incarcerated and nonincarcerated populations.


We selected KVSP because of its well-documented history of arsenic contamination.5 The Allensworth Community Services District, City of Delano, and City of McFarland CWSs are located in close proximity to KVSP, rely exclusively on groundwater sources, and serve greater than 500>500 people each. All three communities have median household incomes of less than 60 percent<60% of California’s statewide average (Table 1).

Table 1 Key socioeconomic and drinking water-related characteristics for the four study sites.

AllensworthCity of DelanoCity of McFarlandKern Valley State Prison
Demographic and economic indicators
 Population: ACS estimatea575 (plus or minus 162±162)b52,886 (plus or minus 31±31)b14,823 (plus or minus 34±34)bData unavailable
 Population: SDWIS estimatec52152,65815,1055,300
 Median household incomea,d33,214 dollars$33,214 (plus or minus 14,921 dollars±$14,921)b43,641 dollars$43,641 (plus or minus 4,601 dollars±$4,601)b35,346 dollars$35,346 (plus or minus 3,476 dollars±$3,476)bData unavailable
 Poverty rate (CA mean equals 13.11 percentmean=13.11%) (%)a,e47.6319.3230.21Data unavailable
Governance and water supply
 Type of local governanceCommunity services districtCity councilCity councilNA
 Active public groundwater supply wells (as of 2021, Q3) (n)c21432
SDWA violation data: 2001–2021f
 Arsenic MCL violations (n)c,g11271219
 Arsenic monitoring or treatment technique violations (n)c3080
 Most recent MCL violation: arsenic (as of 2021, Q3)cQ2 of 2020Q4 of 2012Q1 of 2013Q4 of 2012
 Most recent MCL violation: any (as of 2021, Q3)2020 arsenic violation2019 1,2,3-trichloro-propane violationh2019 1,2,3-trichloro-propane violationh2019 total coliform rule violation
Sampling results from served water sources: 2001–2021ilowercase italic n equals 150n=150lowercase italic n equals 1,714n=1,714lowercase italic n equals 136n=136lowercase italic n equals 426n=426
 Mean arsenic level plus or minus standard deviation±SD (microgram per literμg/L)9.27plus or minus 2.88±2.883.41plus or minus 6.71±6.718.43plus or minus 6.68±6.687.51plus or minus 8.19±8.19
 Median arsenic level (IQR) (microgram per literμg/L)8.95 (5.00)0 (4.27)7.95 (4.43)5.00 (5.00)
 Min. and max. arsenic level (microgram per literμg/L)3.7, 230, 562, 760, 50.4
 Samples exceeding 10 micrograms per liter10μg/L arsenicc34% (lowercase italic n equals 51n=51, max: 23 micrograms per liter23μg/L)8% (lowercase italic n equals 141n=141, max: 56 micrograms per liter56μg/L)20% (lowercase italic n equals 27n=27, max: 76 micrograms per liter76μg/L)19% (lowercase italic n equals 81n=81, max: 50 micrograms per liter50μg/L)
 Posttreatment or post-blending samples exceeding 10 micrograms per liter10μg/L arsenicc,j18% (lowercase italic n equals 13n=13, max: 23 micrograms per liter23μg/L)0% (lowercase italic n equals 0n=0 of 1,250)13% (lowercase italic n equals 9n=9, max: 18 micrograms per liter18μg/L)12% (lowercase italic n equals 47n=47, max: 50 micrograms per liter50μg/L)
Arsenic treatment status and funding
 Approximate state funding for arsenic remediationk496,000 dollars$496,000 interim solutions; 390,000 dollars$390,000 planning grant20.5 million dollars$20.5 million loan and 818,000 dollars$818,000 construction grant232,000 dollars$232,000 planning grant, 3.7 million dollars$3.7 million construction grant6.2 million dollars$6.2 million planning and construction funding
 Arsenic treatment status (as of 2021, Q3)cNo, but the water from two wells was blendedYes, wellhead treatment on four wellsYes, wellhead treatment on one wellYes, treatment on blended water

Note: Socioeconomic data and results from our drinking water analyses are presented for four neighboring community water systems in California’s southern San Joaquin Valley. Analyses are based on lowercase italic n equals 3,984n=3,984 water quality monitoring samples taken across these four systems from 2001 to 2021, of which lowercase italic n equals 2,426n=2,426 samples were from served water sources. Samples below the detection limit were included as reported in California’s Drinking Water Watch.9 ACS, American Community Survey; CA, California; IQR, interquartile range; max, maximum; MCL, maximum contaminant level; min, minimum; NA, not applicable; Q, quarter; SD, standard deviation; SDWA, Safe Drinking Water Act; SDWIS, Safe Drinking Water Information System.

aInformation from 5-y ACS 2019 estimates (most recent year available) for these census-designated geographies.

bMargin of error bounds reflect a 90% confidence interval around the estimate.

cInformation calculated based on data from California’s Drinking Water Watch,9 which is partly sourced from the U.S. Environmental Protection Agency’s SDWIS.

dMedian household income in the past 12 months, in 2019 inflation-adjusted U.S. dollars.

ePercentage of the population whose income in the past 12 months was below the federal poverty level.

fSDWA violation data.

gThe MCL is the highest level of a contaminant allowed in drinking water. MCL violations are assigned by the Division of Drinking Water when a public water system exceeds the MCL.

hCalifornia’s state MCL for 1,2,3-trichloropropane is 0.005 micrograms per liter0.005μg/L.

iServed water sources refers to water sources served to community water system customers, which excludes raw water samples from treated and blended sources.

jWater quality samples taken after arsenic treatment (i.e., in Delano, McFarland, and Kern Valley State Prison), or after blending the water from two wells without treatment (i.e., in Allensworth).

kInformation from the California Department of Corrections and Rehabilitation (2013)8 and California State Water Resources Control Board (CSWRCB) Department of Financial Assistance (DFA).9 DFA confirmed funding for Allensworth, Delano, and McFarland from 2014 to 2021 (B. Chase, Supervising Water Resource Control Engineer, CSWRCB, personal communication). “Interim solutions” refers to grant funding for bottled water deliveries and school water filling stations.

We analyzed publicly available drinking water quality monitoring and violation data for 2001–2021, downloaded from California’s Drinking Water Watch, which is partly sourced from the U.S. Environmental Protection Agency’s Safe Drinking Water Information System.9 For our case analysis, we compared MCL and monitoring violation occurrence and frequency by CWS, and we calculated and compared running average arsenic concentrations for each water source in each CWS (Figure 1A–D). We further analyzed disaggregated sampling points and the 2001–2021 averages for served water (i.e., water served to CWS customers) (Figure 1E–H; Table 1). To assess arsenic remediation effectiveness, we disaggregated publicly reported monitoring data to calculate the number and percentage of posttreatment and post-blending samples exceeding 10 micrograms per liter10μg/L (Table 1). We communicated with regional water engineers to confirm our understanding of publicly available water source labeling and arsenic remediation information. Data and R scripts used for our analyses are available in our supporting information files (

Figures 1A to 1D are ribbon and line graphs titled Allensworth, Delano, Kern Valley State Prison, and McFarland under Loess-Smoothed Curves, plotting micrograms per liter (parts per billion), ranging from 0 to 40 in increments of 10 (y-axis) across M C L violations and Monitoring violations from the year 2001 to 2021 in increments of 2 (x-axis) for Pretreatment or blending, Blended, Posttreatment, Posttreatment and blending, Raw (Untreated), and Maximum Contaminant Level. Figures 1E to 1H are dot graphs titled Allensworth, Delano, Kern Valley State Prison, and McFarland under Individual Sampling Records for Served Water Sources, plotting micrograms per liter (parts per billion), ranging from 0 to 60 in increments of 20 (y-axis) across M C L violations and Monitoring violations from the year 2001 to 2021 in increments of 2 (x-axis) for Pretreatment or blending, Blended, Posttreatment, Posttreatment and blending, Raw (Untreated), and Maximum Contaminant Level.

Figure 1. Public water supply arsenic data for the four study sites (Allensworth, Delano, McFarland, Kern Valley State Prison), January 2001–May 2021. All panels depict reported arsenic concentrations. (A–D) depict loess-smoothed water quality data for each water sampling point within each CWS, with shading for 95% confidence intervals around the moving averages. Pre- and post-arsenic-remediation (i.e., blending or treatment) water quality data are included for comparison and assessment of average arsenic remediation effectiveness. Both MCL violation data and monitoring and reporting violation data for arsenic are shown below each panel for year(s) with violations. For comparison with (A–D), (E–H) depict arsenic concentration sampling data for served water sources only (i.e., water sources served to CWS customers, excluding raw water samples from treated sources). Although running annual averages are used to assess MCL compliance, scatter plot figures reveal served samples that exceeded the MCL, enabling pre- and post-arsenic-treatment comparison. The asterisk in (F) indicates that pretreatment water quality monitoring samples taken around that date were likely sampled during the treatment plant commissioning process and, if so, would not have been served to customers (personal communication from a Water Resource Control Engineer at the California State Water Resources Control Board); however, we include these monitoring results here because engineers at the CSWRCB could not definitively confirm that these samples were not served to customers. In all panels, samples below the detection limit were included as reported in California’s Drinking Water Watch9 and dashed lines indicate the 2006 legal limit (i.e., MCL) change to 10 micrograms per liter10μg/L. Note: CSWRCB, California State Water Resources Control Board; CWS, community water system; loess, locally estimated scatterplot smoothing; max, maximum; MCL, maximum contaminant level.


Over the time period analyzed, all four systems served water in exceedance of the revised arsenic MCL in multiple years, and all received violations for exceedances (Figure 1). Mean arsenic levels in served water sources from 2001 to 2021 ranged from 3.4 plus or minus 6.7 micrograms per liter3.4μg/L [standard deviation (SD) = 6.7 μg/L] in Delano, to 9.3 plus or minus 2.9 micrograms per liter9.3μg/L (SD = 2.9 μg/L) in Allensworth (lowercase italic n equals 2,426n=2,426 samples from served water sources across four systems) (Table 1). All four systems also received MCL violations for other contaminants (e.g., nitrate, total coliforms, 1,2,3-trichloropropane) over this 20-y period. Disaggregated sampling results for served water sources (Figure 1E–H) reveal that, following arsenic remediation efforts, multiple samples remained greater than 10 micrograms per liter>10μg/L in every system except Delano (Table 1). Uniquely among the three CWSs with arsenic treatment in place, KVSP had several posttreatment water samples with arsenic levels greater than 20 micrograms per liter>20μg/L (Figure 1G). From 2019 to 2021, Allensworth and McFarland also exhibited short periods during which arsenic remediation efforts were not optimized (Figure 1E,H).


Although all four CWSs were in compliance with the arsenic MCL as of the third quarter of 2021, sample levels and daily concentrations fluctuate and can periodically exceed legal limits with no violations recorded (Figure 1). Because compliance with the arsenic MCL is determined using a running annual average by water system sampling point, in communities with arsenic levels near the MCL, the number of MCL violations (Table 1) likely underestimates the risks of chronic exposures to arsenic, and underreports potential violations of the human right to safe water. Disaggregated sampling results (Figure 1E–H) more accurately reflect the health risks from arsenic levels than do the running annual averages used to assess legal compliance. Metrics such as the 95th percentile of sample results, or the number of samples exceeding half of the MCL, are useful additional tools for tracking such exposure risks.3,10

In low-income rural settings, persistent and known water-related injustices can reach across carceral boundaries. Unlike the other CWSs in our study, KVSP was built following notification of the new arsenic MCL in 2001. After 2006, KVSP was out of compliance for 7 y. Despite a 6 million dollar$6-million state investment for arsenic remediation (Table 1),8 violations of the human right to water persisted at KVSP with respect to day-to-day water safety and access to alternatives (Figure 1C,G). Bottled water is sold at KVSP, but because incarcerated people can be paid at most 56 dollars per month$56/month per California regulations (15 California Code of Regulations, Section 3041.2) it is not a viable safe water alternative. By comparison, Allensworth residents still lack a long-term solution for their arsenic exposure, although the state subsidizes bottled water access. Because federal laws situate the responsibility for CWS financing primarily at a local level, small CWSs serving low-income communities often lack the funding needed to adequately mitigate exposures to water contaminants such as arsenic.4 Repeated individual sampling results for arsenic greater than 10 micrograms per liter>10μg/L in Allensworth, KVSP, and McFarland (Figure 1E,G,H) reveal the limitations of current arsenic remediation efforts.

Our analysis is bounded by reported CWS data and limited by a lack of water quality testing data at the point of use; thus, we cannot properly estimate individual exposures. Generalization of our findings to other carceral and rural communities may be limited by the particulars of KVSP and the other study communities.


Overall, our findings illustrate that a) structural challenges to the realization of the human right to safe water occur pre- and post-arsenic treatment and unfold in distinct ways for incarcerated and nonincarcerated rural communities, b) human right to water violations can persist even following state investments for remediation, and c) annually averaged water quality data used to track and publicly report violations of the Safe Drinking Water Act provide only a partial guide to whether the human right to water is being realized. Publicly available, disaggregated monitoring data enables a more nuanced comparison of water quality and of progress toward the human right to water.


A.C. and D.P. contributed to the study concept and design. J.R., E.H., Z.Z., S.K., X.Z., and C.D. contributed to data extraction and organization. J.R., E.H., Z.Z., S.K., X.Z., C.D., Z.H., and A.C. conducted initial data analysis. J.R. conducted primary and final analysis. J.R., E.H., and A.C. conducted outreach and ground-truthing. J.R., I.R., J.V., D.P., and A.C. drafted the manuscript. J.R., I.R., and A.C. reviewed and revised the manuscript. J.R., I.R., D.P., and A.C. offered supervision.

We thank M. Hagan, B. Chase, B. Potter, C. Fischer, and L. Pham for their assistance and guidance. This work was partially supported by the National Science Foundation Graduate Research Fellowship Program under grant DGE 1752814 to J.R. Additional funding was provided through the University of California’s Undergraduate Research Apprenticeship Program and the Berkeley Fellowship.

URLs for all data sources as well as the R scripts used for data processing and analysis are included in the supporting information files (


J.R. serves as an uncompensated board member at Community Water Center, an NGO that works to achieve safe and affordable drinking water access for Californians. All other authors declare they have no actual or potential competing financial interests.