Using the Pollution Load Index to Evaluate Rooftop Harvested Rainwater Metal(loid) Contamination in Environmental Justice Communities

Abstract

Water scarcity poses a significant public health crisis exacerbated by climate change-induced disruptions to freshwater sources. Rainwater harvesting offers a sustainable solution by harnessing rooftop runoff for domestic use. This study analyzed 577 rooftop-harvested rainwater (RHRW) samples from four Arizona, USA environmental justice communities and 162 control samples from five National Atmospheric Deposition Program wet-only deposition collection sites across Arizona. The samples were tested for metal(loid)s, and the pollution load index (PLI) and Nemerow Integrated Pollution Index (NIPI) were used to assess contamination. The PLI was calculated for 11 known contaminants (As, Pb, Cd, Mn, Al, Cr, Cu, Zn, Ni, Ba, and Be), with the highest contamination factor observed for Ni (1340). PLI levels ranged from 0.118–65.8, with the active mining community Globe-Miami showing the highest range (0.244–65.8). The PLI was significantly greater during the monsoon season than during the winter season for all the communities (p < 0.05). Compared with urban communities (0.118–13.1), active mining communities (0.169–65.8) had higher PLI values. pH was positively correlated with PLI in Tucson (β = ln 0.27). In non-urban/rural mining communities, locations closer to potential contamination sources had higher PLI values (β = ln −0.33 to −0.38). However, in Tucson, the proximity relationship was less defined because of multiple potential contamination sources in urban areas. This study highlights the importance of using indices like PLI and NIPI to assess water quality; PLI reflects cumulative contamination burden, while NIPI contextualizes this burden within potential water uses. The strong positive correlation observed between PLI and NIPI across all use categories supports the validity of both indices and affirms utility. Together, they provide a nuanced understanding of pollution dynamics in RHRW and strengthen the case for public health interventions and ensuring the safety and sustainability of RHRW.

Graphical Abstract:

1. Introduction

Extended periods of drought, a rapidly increasing population, and a growing demand for water has led to regional water scarcity (Nizam et al., 2021), defined as a lack of sufficient potable water to meet daily demands. Climate change reduces both surface and groundwater availability, threatening freshwater access globally. Altered precipitation and increased evapotranspiration decrease surface water, whereas over-extraction depletes groundwater, exacerbating its scarcity (Condon et al., 2020; He et al., 2022; Swain et al., 2022).

Harvesting rainwater is used worldwide as an inexpensive source of water and drought mitigation strategy (Al-Houri et al., 2014; Hasan & Irfanullah, 2022; Mukaromah, 2020). Rooftop harvested rainwater (RHRW) systems collect rainwater from rooftops, directing it through a series of gutters and pipes into storage tanks/cisterns for later use (Al-Houri et al., 2014; Anchan & Shiva Prasad, 2021). This collected rainwater is then used for irrigation, household tasks, and perhaps drinking, contributing significantly to sustainable water management, especially in areas with inconsistent access to clean water sources. Therefore, RHRW helps reduce dependence on centralized water systems, improve resilience to climate-induced water-related challenges, and promote sustainable water management practices by reducing reliance on groundwater and surface water sources (Al-Houri et al., 2014; Anchan & Shiva Prasad, 2021). Additionally, by intercepting and storing rainwater, these systems reduce the volume and velocity of stormwater runoff, thereby mitigating erosion and flooding in urban and rural environments (Mukaromah, 2020).

Environmental justice (EJ) concerns are particularly pertinent when considering RHRW systems. Marginalized or limited-income communities already face disproportionate exposure to environmental hazards and lack access to clean water sources (Clinton, 1994; Gochfeld & Burger, 2011; Schaider et al., 2019). The placement of RHRW systems in areas affected by environmental injustice, e.g., non-urban communities near industrial sites or contaminated land, raises additional concerns about potential contamination of harvested rainwater. These communities may already face increased risk of waterborne illnesses and exacerbating health disparities (Banzhaf et al., 2019; Bullard, 2001; National Academies of Sciences et al., 2017). Additionally, placing RHRW systems in urban communities with multiple contamination sources has similar EJ concerns. These communities often experience a convergence of environmental hazards, including industrial pollution, transportation emissions, and deteriorating infrastructure (F. Li et al., 2022; Moreno-Rodríguez et al., 2015; Pandion et al., 2022), which can compromise water quality and public health. Without careful consideration, there is a risk of exacerbating existing disparities in access to clean water and perpetuating environmental injustices.

Previous studies have examined the impact of roof material (Kim et al., 2016; Lee et al., 2012; Meera & Mansoor Ahammed, 2018; Mendez et al., 2011; Ojo, 2019; Ward et al., 2010; Zdeb et al., 2020), cistern material (Wu et al., 2016), maintenance practices (Dao et al., 2021; Lee et al., 2015), and points-of-use (de Kwaadsteniet et al., 2013; Despins et al., 2009) on harvested rainwater quality. Previous research has also concentrated on microbiological contamination (de Kwaadsteniet et al., 2013; Despins et al., 2009; Gikas & Tsihrintzis, 2012; Hamilton et al., 2019; Moses et al., 2023a, 2023b), physicochemical parameters such as total nitrogen, turbidity, etc. (Despins et al., 2009; Gikas & Tsihrintzis, 2012; Nizam et al., 2021; Radaideh et al., 2008), and the chemical contamination by focusing on individual contaminants (de Kwaadsteniet et al., 2013; Palawat et al., 2023a; Villagómez-Márquez et al., 2023; Zhu et al., 2004). However, an examination of contaminants on an individual basis overlooks the complex reality that contaminants often coexist in the environment. Therefore, there is a need for a unified metric to comprehensively understand the extent of water contamination. By integrating multiple types of contaminant data into a single measure, researchers and policymakers can better assess overall water quality and prioritize mitigation efforts accordingly.

Commonly employed soil and sediment pollution metrics include contamination indices such as the pollution load index (PLI) (Tomlinson et al., 1980), contamination factor (CF) (Pekey et al., 2004), geoaccumulation index (Muller, 1969), and Nemerow Integrated Pollution Index (NIPI) (Su et al., 2022) and the enrichment factor (Sinex & Helz, 1981). These metrics typically involve the analysis of pollutant concentrations relative to background levels or reference standards to determine contamination levels and have been used to study soil and sediments (Jorfi et al., 2017; Mandour et al., 2021; Nasir et al., 2023; Ogbeibu et al., 2014; Rahmanian & Safari, 2022; Safari, 2016; Sukri et al., 2018; Waida et al., 2022; Zeider et al., 2023), marine organisms (Angulo, 1996), and surface water (Ahmad, 2013; Ephsy & Raja, 2023; Waida et al., 2022).

A major drawback for the application of traditional contamination indices to RHRW such as NIPI is that they typically compare contamination levels to enforceable standards, such as the maximum contaminant levels set by the Safe Drinking Water Act, whereas no such enforceable standards exist specifically for RHRW (Tsanov et al., 2023). Nevertheless, the PLI serves as a valuable tool for determining the degree of pollution in environmental matrices by comparing pollutant concentrations in samples to background levels (Krishnakumar et al., 2021; Li et al., 2020). The PLI allows for the identification of significant contaminants and their individual contributions to overall pollution, offering a comprehensive snapshot of contamination severity (Sarkar, 2022).

For these reasons, we employed the PLI in this study to assess metal(loid) contamination in rooftop-harvested rainwater across four EJ communities in Arizona. Specifically, our objectives were to: (1) characterize the contamination factor of metal(loids) in harvested rainwater; (2) evaluate the overall water quality using the PLI and NIPI; (3) identify key factors influencing PLI values, such as land use, seasonality, and proximity to pollution sources. This study addresses the following research question: How does metal(loid) contamination, as measured by the Pollution Load Index, vary across rooftop-harvested rainwater systems in EJ communities, and what factors most significantly influence contamination levels?

2. Methods

2.1. Site Descriptions

Project Harvest (PH) (www.projectharvest.arizona.edu) is a co-created community science initiative that ran between 2016 and 2021 (sample collection time frame 2017–2020). Four Arizona EJ communities were involved (S1 Figure). The participating Arizona communities; Hayden/Winkelman, Globe/Miami, Dewey-Humboldt, and Tucson (59–75% of participants were from the southern metropolitan area) were based on long-standing community-academic partnerships, established trust, and existing relationships developed through participatory research efforts. For additional site descriptions and community engagement details, see: Davis et al., 2020; Moses et al., 2022, 2023a, 2023b; Palawat et al., 2023a; Villagómez-Márquez et al., 2023, Ramírez-Andreotta et al., 2023.

Each community has at least one principal point source of contamination, with some having multiple point and nonpoint sources of contamination identified from the US EPA toxic release inventory (United States Environmental Protection Agency (U.S. EPA), 2024). Dewey-Humboldt (DH) is home to the legacy Iron King Mine and the Humboldt Smelter Superfund site. In 2022, DH had a population of 4,363 and can be described as a urban cluster (Data USA, 2022). Globe/Miami (GM) and Hayden/Winkelman (HW) are active mining communities and due to the town/city proximity, they were grouped in PH. GM is home to the Freeport McMoRan Copper and Gold Mine alongside 1,236 active and legacy mines and mine-adjacent activities (Horton & San Juan, 2023). HW is home to the U.S. EPA alternative superfund site called the ASARCO Hayden Plant site, which consists of a smelter, concentrator, former smelter (Kennecott), and mine tailing facilities (U.S. EPA, n.d.). From October 13, 2019 to July 6, 2020, activities at the HW smelter were temporarily suspended due to a workers strike and the smelting operations were indefinitely suspended in summer 2020 (impacts of the smelting strike on RHRW pollution load were discussed in Palawat et al. (2023a). Hayden, Winkelman, and Miami are rural with <2,500 people and Globe is an urban cluster population of 7,249 (U.S. Census Bureau, 2020a). DH, GM, and HW are together cataloged as non-urban, referred to colloquially herein as, non-urban/rural.

Lastly, Tucson (TU) is an urban community divided into six wards, with a population exceeding 500,000 (City of Tucson, 2023; U.S. Census Bureau, 2020b). The city of south Tucson is a “Pueblo within a city” with a population of just over 5,652, where 84.1% are Hispanic (The City of South Tucson, n.d.). South Tucson is an EJ community due to its proximity to the Tucson International Airport Area Superfund site (Domínguez, 2022). In this study, 59–75% samples were collected from the city of South Tucson (Davis et al., 2020; Palawat et al., 2023a). However, samples from both Tucson and South Tucson are collectively referred to as TU in this study. Like with many urban communities, TU have multiple potential point and nonpoint sources of contamination. However, for statistical modeling purposes, the Davis-Monthan US Air Force Base was selected as the single point source of contamination for this study.

Background samples were obtained from the National Atmospheric Deposition Program (NADP) wet-only deposition rainwater samples from five sites around the AZ (National Atmospheric Deposition Program, 2023; Palawat et al., 2023). The NADP water samples were collected via a lid seal to protect the samples from evaporation and dry deposition contamination (Tanabe, 2019). NADP samples were obtained from the Chiricahua National Monument, Grand Canyon National Park, Organ Pipe Cactus National Monument, Oliver Knoll and Petrified Forest National Park (National Atmospheric Deposition Program, 2023). All five NADP sites are in remote areas away from human activities and the impacts of industrialization and can therefore provide background control levels to help understand the effects of human activities on contamination. NADP samples were collected from all five sites between 2018 and 2021.

2.2. Sample collection and analysis

The primary goal of our sampling strategy was to bracket major rain events in the U.S. Southwest. Historical data from the National Weather Service informed the sampling windows, which were designed around the binary rainfall regime of the North American monsoon and winter wet seasons (National Weather Service, 2017). For 2.5 years, community scientists collected and submitted samples four times a year during the first and last winter rains (December–February) and the first and last monsoon rains (June–September).

In total, 577 community samples (307 in TU, 93 in HW, 124 in GM and 53 in DH) and one hundred and sixty-two (162) NADP samples were collected over the course of the PH. After daily instrument calibration, the RHRW samples were analyzed for pH and electrical conductivity (EC) (Fisher XL-20, Accumet electrode). The samples were further analyzed via inductively coupled plasma–mass spectrometry (ICP–MS, Agilent 7700) at the Arizona Laboratory for Emerging Contaminants (Tucson, Arizona, 85721) for nineteen dissolved metal(loid)s (Be, Al, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Mo, Ag, Cd, Sn, Sb, Ba, Pb). The quality assurance/quality control (QA/QC) protocol employed in this study was adapted from U.S. EPA Method 200.8 and included the use of field and laboratory blanks, certified reference materials and replicates (U.S. EPA, 1996). In general, the calibration standards for the ICP-MS were prepared from multi-element stock solutions (SPEX Certiprep, Metuchen, NJ), the QC sample solutions were from an independent source, e.g., NIST SRM 1643e Trace Elements in Water, and reagent blanks, dilution duplicates, and spiked duplicates were run with every 20–25 samples analyzed as further QA/QC. For additional sample collection, analysis, and QA/QC details, see Palawat et al. (2023b).

2.3. Statistical analysis

The method limits of detection (MLOD) were obtained by multiplying the ICP–MS instrument detection limits by the laboratory preparation factors (Palawat et al., 2023b). Values below the detection limit were substituted with the method limit of detection divided by the square root of 2. The contamination factor was used to determine the extent of contamination compared with that of the background samples (W. Ahmad et al., 2021). There was no significant difference in NADP concentrations between sites or years. Therefore, the mean NADP concentration for each analyte was used as the background concentration.

2.3.1. Contamination Factor and Pollution Load Index

The contamination factor of each sample was obtained via (1):

$$\mathrm{Contamination}\mathrm{factor}=\frac{C_n}{C_b}$$ (1)

where ${\mathrm{C}}_{\mathrm{n}}$ is the concentration of an element in the RHRW sample and ${\mathrm{C}}_{\mathrm{b}}$ is the mean NADP concentration for each analyte. CF < 1 represents low contamination, 1 ≤ CF < 3 represents moderate contamination, 3 ≤ CF < 6 represents considerable contamination, and CF ≥ 6 represents very high contamination (W. Ahmad et al., 2021; Jannat et al., 2023). For nutrients such as K and Mg, the contamination factor (CF) indicates nutrient enrichment, whereas for pollutants such as Pb and Cd, the CF represents anthropogenic influences on pollutant concentrations. The PLI was then calculated for As, Pb, Cd, Mn, Al, Cr, Cu, Zn, Ni, Ba and Be via (4).

$$\text{PLI}=\sqrt[n]{C{F}_1\times C{F}_2\times C{F}_3\dots \dots \dots \times C{F}_n}$$ (4)

The PLI calculation uses the geometric mean of the contamination factors rather than the arithmetic mean. The geometric mean is appropriate because it accounts for the multiplicative nature of contamination data and reduces the influence of extreme values, providing a more balanced representation of overall pollution. Contamination factors often vary by several orders of magnitude; thus, using the arithmetic mean could disproportionately weight a few very high or very low values, distorting the true overall pollution level. The geometric mean ensures that each metal contributes proportionally to the index, offering a more stable and representative measure of cumulative pollution.

2.3.2. Nemerow Integrated Pollution Index (NIPI)

The NIPI is another widely applied method for comprehensive water quality assessment. It integrates both the average and maximum values of individual pollutant indices, thereby emphasizing the influence of the most critical contaminant while also considering the broader contribution of all measured pollutants (Dey et al., 2021; Kong et al., 2019; Su et al., 2022). This approach allows for a more robust evaluation by combining multiple factors into a single, weighted index. The NIPI is particularly effective in identifying cases where a single pollutant dominates the overall pollution profile, ensuring that such extremes are not masked by averaging. The NIPI uses the following formula:

First, the single factor pollution index is calculated using;

$$\text{P}=\frac{C_i}{S_i}$$

where $C_i$ is the measured concentration of a given element in the RHRW samples from the study area, and $S_i$ is the corresponding standard/recommendation/maximum level used as a baseline. Given that the RHRW intended uses can vary across households, this study categorized water use into the following primary uses and their associated baseline : (1) Human-Drinking, U.S. Environmental Protection Agency’s (US EPA) primary and secondary maximum contaminant levels (MCL) under the Safe Drinking Water Act (US EPA, 2015), (2) Irrigation, U.S. Department of Agriculture’s “recommended maximum irrigation concentration for continuous use on all soils” (Pick, 2011), and (3) Livestock-Drinking, USDA’s Livestock and Poultry Drinking Water Recommended Upper Limit (Pick, 2011). The NIPI was then calculated for each designated use based on all 11 identified pollutants where data were available using the formula;

$$\text{NIPI}=\sqrt{\frac{P_{avg}^2+P_{max}^2}{2}}$$

Where ${\mathrm{P}}_{\mathrm{avg}}$= the average value of the single-factor index

${\mathrm{P}}_{\max }$= the maximum value of the single-factor index

2.3.3. Air Quality Data

Air quality data was obtained from the U.S. EPA AirData Air Quality Monitor Program (U.S. EPA, 2014), covering the period spanning December 2017 to February 2020. PM₁₀ concentrations (μg m⁻³) were collected as 24-hour block averages. Data collection was organized by community and season. PM₁₀ data was available for Tucson, Globe/Miami, and Hayden-Winkelmann. Dewey-Humboldt had no air quality monitors within a 50-mile radius. In Globe/Miami and Hayden-Winkelmann, only one air quality monitor was present in each community, whereas Tucson had multiple air quality monitors. However, the South Tucson air quality monitor was selected. The mean and standard deviation of PM₁₀ concentrations were calculated for both the monsoon and winter seasons.

2.3.4. Statistical Modeling

Linear mixed models were built to assess the factors affecting metal(loid) concentrations in harvested rainwater samples (Bates et al., 2015; Hajduk, 2019; Quinn & Keough, 2002; Zuur et al., 2009). Owing to community-level variability, individual communities were modeled separately. The data were analyzed via a method similar to that used by Palawat et al. (2023a), with the addition of a backward stepwise model selection process. To better meet the assumptions of normality, all the dependent variables were natural log transformed. For each community, site was a random variable, and the fixed variables included season, proximity to the point source, and pH. The distance to the point source was normalized in each community for ease of analysis and comparison. The data were normalized by subtracting the average distance in each community from each site’s distance and then dividing that by the standard deviation of the community.

Additionally, different sublocations and geopolitical boundaries were often included as predictor variables in the model when present: GM (Miami/Claypool, Globe, Icehouse Canyon/Six Shooter Canyon), HW (Hayden, Winkelman) and Ward in TU (Wards 1–6). If the models did not meet the assumptions (linearity, independence, and homogeneity of residuals), they were not tabulated or summarized. Additionally, models with low R² values (<10%) are reported with a note stating that the models inadequately explained the variance. The same modeling method was used to assess the factors influencing the PLI according to community and overall factors. Figures were created with back-transformed data for ease of interpretation, whereas tables retain the natural log-transformed data directly from the model outputs. All the statistical analyses were conducted in RStudio (R Core Team, 2023) via the following packages: “tidyverse”, “ggplot2”(Wickham, 2016), “easystats” (Daniel et al., 2020), “lme4” (Bates et al., 2015), “lmertest” (Kuznetsova et al., 2017), “car”, “MASS”(Ripley et al., 2024), “envStats”(Millard, 2013), “dplyr” (Wickham et al., 2020) and “performance” (Lüdecke et al., 2021).

3. Results

3.1. Detection and descriptive statistics of individual metal(loid)s

Descriptive statistics of analyte concentrations across Project Harvest communities and NADP control sites are presented in S1 Table. Most analytes were detected in over 75% of the samples, with the exception of Ag (56.0%), Be (53.7%), and Se (48.9%). Wide concentration ranges were observed for Al, Ba, Cr, Cu, Mn, Ni, and Zn across all communities, including the control samples. Overall, analyte concentrations in the control samples were generally lower than those measured in the four PH communities.

3.2. Contamination factor, PLI and NIPI

To determine the contamination factor across all four communities, the concentrations of the nineteen analytes were compared to those of the background samples (S2 Table). The contamination factors in the four communities showed different levels of anthropogenic influence on various metal(loid)s. GM had the highest geometric mean contamination factors for Al (1.45), Be (0.741), Co (3.99), Fe (1.03), Mn (1.75), and Ni (3.37). DH had the highest contamination factor for Cr (0.991). HW had the highest contamination levels for Ag (0.380), As (5.65), Cd (6.58), Cu (14.8), Pb (2.33), Sb (1.44), and Se (0.176). The TU had the highest contamination factors for Mo (9.25), Sn (1.03), V (1.45), and Zn (9.78). A contamination factor greater than one indicates that the concentration of an analyte is higher than the background level in uncontaminated environments. Therefore, GM had the greatest anthropogenic influence on Al, Co, Fe, Mn, and Ni; HW had the greatest anthropogenic influence on As, Cd, Cu, Pb, and Sb; and TU had the greatest anthropogenic influence on Mo, Sn, V, and Zn.

The PLI and NIPI serve as cumulative metrics that capture the overall pollution burden by integrating multiple contaminants. Based on the combined PLI geometric mean, pollution severity across the communities ranked from highest to lowest as follows: HW (2.49), GM (2.43), TU (1.65), DH (1.18) (Table 1A; Figure S2). However, NIPI classifications were highly sensitive to the applied categorical baselines. When the U.S. EPA primary and secondary MCL were used as baselines, most samples across all communities were classified as having no pollution: 79.2% in DH, 64.2% in GM, 75% in HW, and 78.2% in TU (Table S3, Figure S3). In contrast, when applying the irrigation and drinking-livestock levels, a broader NIPI distribution and higher pollution was observed (Table 1B). Despite these classification differences, there was a statistically significant and positive Spearman correlation between PLI and NIPI across all use categories (S4 Figure). The strongest correlation was observed for drinking water use (ρ = 0.83), while the lowest was for irrigation water use (ρ = 0.65).

Table 1A.

Pollution load index winter and monsoon summary by community.

Community	Mean (sd)	Median (min, max)	Geomean (geometric standard deviation)
Dewey-Humboldt (DH)
Monsoon (N=19)	2.46 (2.04)	1.65 [0.511, 7.88]	1.84 (2.18)
Winter ((N=34))	1.23 (1.14)	0.893 [0.148, 5.99]	0.925 (2.11)
Dewey-Humboldt Combined(N=53)	1.67 (1.62)	1.06 [0.148, 7.88]	1.18 (2.27)
Globe/Miami
Winter (N=79)	3.20 (5.06)	1.54 [0.244, 24.8]	1.82 (2.51)
Monsoon (N=45)	6.71 (10.6)	3.69 [0.369, 65.8]	4.02 (2.52)
Globe/Miami Combined (N=124)	4.48 (7.67)	1.97 [0.244, 65.8]	2.43 (2.70)
Hayden/Winkelman
Winter (N=67)	2.48 (1.77)	2.30 [0.169, 10.2]	1.93 (2.19)
Monsoon (N=26)	8.04 (10.6)	4.23 [1.13, 47.8]	4.83 (2.62)
Hayden/Winkelman Combined (N=93)	4.03 (6.26)	2.45 [0.169, 47.8]	2.49 (2.53)
Tucson
Winter (N=220)	1.85 (1.59)	1.40 [0.118, 13.1]	1.45 (1.98)
Monsoon (N=87)	2.76 (1.89)	2.33 [0.341, 11.3]	2.29 (1.83)
Tucson Combined (N=307)	2.11 (1.73)	1.56 [0.118, 13.1]	1.65 (1.99)
All Communities
Winter (N= 400)	2.17 (2.72)	1.46 [0.118, 24.8]	1.53 (2.18)
Monsoon (N=177)	4.51 (7.15)	2.71 [0.341, 65.8]	2.88 (2.31)
All Communities Combined (N=577)	2.89 (4.68)	1.68 [0.118, 65.8]	1.86 (2.33)

Table 1B.

Nemerow Integrated Pollution Index in the PH Communities by Different Water Uses.

	Human-Drinking		Irrigation		Livestock-Drinking
Community	Mean (sd)	Median (Min-Max)	Mean (sd)	Median (Min-Max)	Mean (sd)	Median (Min-Max)
Dewey-Humboldt	0.45 (0.79)	0.13 (0.05-4.46)	0.6 (0.29)	0.71 (0.02-1.3)	0.16 (0.25)	0.08 (0.02-1.12)
Globe/Miami	1.29 (2.92)	0.34 (0.05-18.56)	0.91 (1.44)	0.71 (0.03-9.74)	0.27 (0.52)	0.08 (0.01-3.9)
Hayden/Winkelman	0.61 (1.07)	0.25 (0.02-8.59)	0.88 (1.25)	0.71 (0.05-11.24)	0.54 (1.1)	0.2 (0.01-8.55)
Tucson	0.59 (1.9)	0.19 (0.02-24.9)	0.67 (0.56	0.71 (0.03-4.72)	0.16 (0.26)	0.08 (0.01-2.5)

3.3. Factors influencing the pollution load index

3.3.1. Seasonal influences on the PLI

The estimated PLI was analyzed for seasonal variation via linear and linear mixed models, as PH RHRW samples were collected at both the beginning and end of the winter and during monsoon rain events (S5 Table). Overall, the rooftop-harvested rainwater collected in the monsoon season had a greater PLI than did the samples collected in the winter (Figure 1a). This trend was consistent across all four communities when analyzed individually (Figure 1b); however, GM and HW presented more evident differences in the estimated PLI, with a peak during the monsoon season.

Figure 1.

Effect plots visualizing the estimated pollution load index by season in (a) all communities combined (b) in individual communities.

3.3.2. Influence of Proximity on the PLI

The proximity to an identified point source of pollution was significant only in the GM and HW models. In both communities, sampling locations further from the pollution point source had lower PLI values than did sampling at closer locations (Figure 2A, B). Although proximity was not significant in the PLI models for DH and TU (S5 Table), some individual analytes were significantly influenced by proximity in TU (S5 Table). The concentrations of Be, Cu, and Pb significantly decreased farther from Davis than they did at the Air Force Base (S4 Table). However, the concentration of Zn increased with increasing distance from the air force base (S4 Table). Given these results, it is important to note that the reliability of these models was limited on the basis of model diagnostics.

Figure 2.

Effect plots of normalized proximity to potential point sources of contamination. (A) Pollution load index (PLI) trend in Globe/Miami, Arizona, USA in relation to the Freeport McMoran active mine, (B) PLI trend in Hayden/Winkleman, Arizona, USA in relation to the ASARCO smelter and (C) an interaction between proximity to the Davis-Monthan Air force base and (C1) winter and (C2) monsoon season influencing Zn concentration in Tucson, Arizona, USA.

Additionally, the interaction between season and proximity significantly affected the zinc concentration in Tucson, and the monsoon season had a significantly greater zinc concentration further from the AFB than did the winter. Nonetheless, a positive association existed for zinc in Tucson, further from the AFB, both in the winter and during the monsoon season (Figure 2, C1–2). Moreover, location had a significant effect on the concentrations of As and Cd in HW and there were significantly greater concentrations of As and Cd in Hayden than in Winkelman (Figure 3A). Ward location significantly affected the concentrations of Mo and Sb in TU. Mo was highest in wards one, five and six, whereas Sb was highest in wards five and six (Figure 3B).

Figure 3.

Effect plots of (A) location on As and Cd concentration in Hayden/Winkelmann, Arizona, USA, (B) Mo and Sb concentration in in Tucson, Arizona, USA

3.3.3. pH Influence on the PLI

A two-sample t- test comparing mean pH levels in winter to those in the monsoon season revealed significantly greater pH levels during the monsoon season (6.13 ± 0.73) than during the winter season (5.80 ± 0.69) across all communities (p value = 2.97e-06) (S6 Table). Specifically, at TU, the mean pH was 5.86 ± 0.69 in the winter, whereas it was 6.18 ± 0.72 in the monsoon season (p value = 0.001835). In DH, the mean pH was 5.91± 0.730 in the winter, whereas it was 6.47 ± 0.504 in the monsoon season (p value = 0.003359). In GM, the mean pH was 5.57 ± 0.66 in the winter, whereas it was 5.80 ± 0.82 in the monsoon season (p value = 0.07495). Finally, in HW, the mean pH was 55.81 ± 0.62 in the winter, whereas it was 6.25 ± 0.51 in the monsoon season (p value = 0.007966).

In the TU PLI model, pH was a significant factor influencing the PLI, where an increase in pH was correlated with higher PLI values (Figure 4, Table 2). This positive relationship extended to most analyzed elements (Al, As, Ba, Cd, Co, Cr, Cu, Mn, Mo, Ni, Pb, Sb, V, and Zn) in TU, aligning with the general trend where relatively high pH levels are associated with increased metal(loid) concentrations (S5 Figure, S4 Table). However, an exception to this pattern was noted for Be, where a negative correlation with pH was observed, where higher concentrations were associated with lower pH (S5 Figure).

Figure 4.

Effect plots of pH on pollution load index (PLI) in Tucson, Arizona, USA (urban community).

Table 2.

24-Hour block average PM₁₀ (0-10 μm) concentrations (μg/m³) across the four communities from 2017-2020 in Winter (December - February) and Monsoon (June - September).

Community	Season	2017 - 2018	2018 - 2019	2019 - 2020
GM	Winter	23± 9	12± 6	13± 5
	Monsoon	25±17	21±10	-2
HW	Winter	30± 23	17± 6	9± 6
	Monsoon	35± 23	26±10	-
TU	Winter	35± 14	23± 18	24± 8
	Monsoon	24± 12	21± 8	-
DH	Winter	---1	---	--
	Monsoon	---	---	---

^---1No air quality monitors were located in Dewey-Humboldt

^-2Monsoon samples for 2020 were not collected due to the COVID-19 pandemic.

In contrast, in the other locations studied, pH did not significantly impact the overall PLI model. Nevertheless, when models examined individual analytes, pH levels were significantly associated with specific analytes in these locations. In DH, Ba was positively correlated with pH. Similarly, in the GM, As and Ba both demonstrated positive associations with the PLI, whereas in the HW, As alone was positively correlated with pH (S4 Table).

3.3.4. Influence of land use on the PLI

The models assessing the impact of land use emphasize that the active mining communities (GM and HW) have the highest contamination, and the most negative proximity slope compared with the urban community (Figure 5, Table 2). However, if a site is far enough from point sources (approximately greater than 1.5 km), the PLI of the active mining community falls below the TU PLI (Figure 5).

Figure 5.

Pollution Load Index (PLI) patterns in mining communities with a prominent, single point source of pollution and an urban community with multiple pollution sources.

Air quality data was obtained to contextualize the PLI seasonal differences and its relationship to PM₁₀ concentrations in the communities (Table 2). In GM and HW, PM₁₀ concentrations were consistently higher during the monsoon season compared to the winter season. However, this trend was reversed in Tucson and the winter PM₁₀ concentration was higher than the monsoon concentration in both 2018 and 2019.

4. Discussion

4.1. Comparing the PLI to the NIPI

The NIPI is a widely applied integrated pollution assessment tool that synthesizes multiple contaminant concentrations into a single metric relative to established environmental quality standards or designated uses (Su et al., 2022). NIPI incorporates both average and maximum pollutant concentrations to reflect overall and worst-case pollution conditions (Kong et al., 2019). However, the use of only the average and maximum values may obscure the broader distribution of contaminant concentrations, potentially overlooking important patterns such as multimodality or frequent exceedances below the maximum threshold. Additionally, the NIPI can overweigh extreme values, disproportionately influencing the final score and potentially exaggerating risk in the presence of a single high concentration outlier. As a result, in the current study, even though the drinking water standards are lower and more restrictive, the NIPI sometimes produces higher index values for irrigation or livestock water uses, simply due to one or two extreme concentrations. This outcome can misrepresent the relative risk of exposure across different uses.

Additionally, another notable limitation of NIPI is its dependence on predefined benchmarks corresponding to specific water uses (Su et al., 2022). PH community scientists RHRW use is multifaceted and includes landscaping (33%), trees (21%), gardening (39%), swimming pool (3%), other (3%), car washing (1%) and <1% for drinking, dish washing and bathing (Villagómez-Márquez et al. 2023b). In situations such as this where water uses can be multifaceted and regulatory standards may be incomplete or non-existent, applying NPI can be challenging and may oversimplify complex exposure scenarios (Zhu et al., 2004).

The PLI addresses some of these limitations by providing a more balanced and robust assessment by incorporating the geometric mean of concentration ratios relative to those of control or reference sites assumed to be minimally impacted (Hakanson, 1980; Tomlinson et al., 1980). This method reduces the influence of single outliers and allows for a more holistic understanding of cumulative pollution burden across multiple contaminants. PLI is particularly valuable in environmental justice contexts, where chronic, low-level exposures across several pollutants may pose significant health risks that NIPI might underemphasize or obscure. This comparative approach makes PLI more adaptable for contexts where designated use benchmarks are lacking or where water use patterns do not fit conventional categories.

Importantly, our study found a strong positive correlation between PLI and NIPI across designated uses (Spearman’s rho ranging from 0.60 to 0.73), which lends empirical support to the validity of PLI as a robust pollution indicator. This correlation suggests that PLI and NPI broadly reflect similar pollution gradients despite methodological statistical differences. However, the reliance of NIPI on designated use benchmarks may introduce bias or limit its applicability, whereas PLI’s reference-based approach can be more inclusive but requires careful selection of appropriate control sites to avoid under- or overestimating pollution levels. Therefore, while NIPI remains a valuable and widely recognized tool for pollution assessments, especially when water use designations are clear/regulatory criteria exist, PLI provides a practical alternative that better accommodates complex, multifaceted water use scenarios as highlighted here.

4.2. Contamination Factor Patterns in Project Harvest Communities

In general, the contamination profile of each individual community is shaped by its industrial characteristics. In GM, a community with active copper mining, there was considerable contamination of Cd, Co and Ni and a very high contamination factor of Cu and Zn. This finding is consistent with previous studies that reported a high contamination factor of heavy metals around copper mines (Liu et al., 2020; Zhou et al., 2007). Additionally, HW (an active copper smelter and a designated alternative Superfund site) had moderate contamination of Co, Mn, Ni, Pb, Sb and V; considerable contamination of As and Zn; and very high contamination of Cd, Cu, and Mo. However as highlighted by Palawat et al. (2023a), mining and smelting activities were suspended between October 13^th 2019 to July 6^th 2020 in HW. This may have altered the contamination pattern observed in the rainwater samples harvested in winter 2019 and summer 2020. Currently, smelting activities in HW are indefinitely suspended. In DH (a legacy mining community), there was a low contamination factor (< 1) for most contaminants, including Cd, Mn and Pb, and moderate contamination with As, Co, and Cu. The highest contamination factor was recorded for Zn. Finally, in TU (the urban community), there was moderate contamination of As, Cd, Co, Ni, Pb, Sb, Sn, and V; a considerable contamination factor (1 ≤ CF < 3) of Cu; and a very high contamination factor (CF ≥ 6) of Mo and Zn.

4.3. Factors Affecting PLI in Harvested Rainwater

4.3.1. Season

Season significantly influenced the PLI in all four communities. During the North American monsoon season, local rainfall increases significantly, leading to increased precipitation and runoff (Becker, 2021; Crimmins, 2006; Moreno-Rodríguez et al., 2015). PM₁₀ concentrations in the GM and HW were higher during the monsoon season compared to the winter, a trend attributed to the frequent dust storms and occasional haboobs that characterize the monsoon season (Eagar et al., 2017; Sorooshian et al., 2013; White et al., 2023). The elevated temperatures during the monsoon drive thermal and barometric gradients, which contribute to increased atmospheric deposition, amplified by dust storms and the accumulation of dust on rooftops (Eagar et al., 2017). Additionally, wind erosion is more pronounced during this season, leading to saltation and the suspension of fine particulate matter, which can settle on rooftops and infiltrate cisterns ( (Pandion et al., 2022).

Rainfall during the monsoon also plays a dual role: it cleanses aerosols and gases from the atmosphere but simultaneously washes accumulated debris from rooftop surfaces into rainwater harvesting systems (Ogolla & Olal, 2023). This combination of enhanced atmospheric deposition, intensified wind erosion, and rain-induced surface wash-off processes exacerbates contamination in harvested rainwater during the monsoon season.

4.3.2. pH

pH was a significant factor for the PLI in the urban community and pH was highly correlated with season across all communities. Between March and June in Arizona, fine soil, sulfate and organics dominate the particulate matter (PM) mass due to hot and dry meteorological conditions (Moreno-Rodríguez et al., 2015; Palawat et al., 2023). On the other hand, antecedent precipitation suppresses dust flux in winter, leading to less suspended particulate matter (PM) in the winter. During the monsoon season, gustier winds lead to greater summer dust flux (Reheis & Urban, 2011). The particles that can be transported long distances are generally <2.5 μm (PM_2.5) and contain a high proportion of fine soil, organic carbon, sulfates and nitrates. The highest PM2.5 levels are measured between April and May because of dry conditions and high wind speeds (Sorooshian et al., 2013). Therefore, the pH of rain during the monsoon season is greater (6.09±1.13) because of the presence of crustal acid-neutralizing components, such as calcium carbonate, magnesium, potassium and sodium ions, in dust and fine PM (Sorooshian et al., 2013). This value is greater than the average pH of natural rain, which is 5.6 (Baird & Cann, 2005). In urban areas, high traffic activity, road construction and industrialization may also contribute to dust deposition on rooftops (Igbinosa & Aighewi, 2017). Therefore, the relatively high pH of rain during the monsoon season and the washing of dust deposited on rooftops into RHRW cisterns lead to an increase in the pH of harvested rainwater compared with that in the winter season.

Higher pH was correlated with higher concentrations of Al, As, Ba, Cd, Co, Cr, Cu, Mn, Mo, Ni, Pb, Sb, V, and Zn (S3 Figure). This finding aligns with studies that show associations between pH and metal concentrations in water, e.g. in groundwater, a relatively high pH has been correlated with high arsenic concentrations (Katsoyiannis & Katsoyiannis, 2007; Spayd, 2023). On the other hand, Chapman et al. (2008) reported a negative association between pH and Al, Cu, Pb, and rainwater. However, their pH ranged from 5.1–9.1, whereas the pH ranged from 3.50–8.49 in the present study (S5 Table). At lower pH values, Fe(III) oxides exhibit pH-dependent charge characteristics, being positive at low pH values and negative at high pH values (Tombácz, 2009). This adsorption can significantly affect metal(loid) mobility and concentration in solution. Therefore, the relationships between pH and metal concentration and speciation are influenced by the composition of the water, presence of organic matter, redox conditions, and biological activity (Saalidong et al., 2022).

Interestingly, the Be concentration was negatively correlated with pH. Within a pH range of 5–8, Be tends to precipitate as insoluble hydroxides or form hydrated complexes (International Agency for Research on Cancer, 1993). Under dilute acidic conditions with a pH below 5, positive ions can be generated. Above a pH of 8, it forms negative ions referred to as beryllates (U.S. Environmental Protection Agency, 1998). In this study, 90% (n= 475) had a pH between 5 and 8, and it is logical to conclude that Be precipitated into insoluble hydroxide, which was filtered during the filtration process (International Agency for Research on Cancer, 1993).

4.4. Urban Community with Multiple Sources

In Tucson, contamination levels persistently remain elevated even at considerable distances from identified point sources. This suggests that pollution in urban settings is not solely attributed to a singular source but is instead widespread, likely stemming from a combination of point and nonpoint sources. Between 2017 and 2019 alone, 27 facilities released emissions into the air, soil, and water (U.S. EPA, 2024b). These included multiple metal recycling facilities, cement manufacturing plants, an international airport, jet manufacturing facilities, a U.S. Department of Defense Air Force base, fabricated metal facilities, plastics and rubber industries, a hazardous waste facility, nonmetallic mineral product industries, and petroleum bulk terminals (U.S. EPA, 2024b). The diversity and distribution of these sources pose a challenge in isolating a single contamination site for study, unlike in rural communities, where contamination often stems from a single identifiable source (Deshmukh et al., 2020; Martínez-Bravo & Martínez-del-Río, 2019).

Consequently, for the sake of comparability, the Tucson Air Force Base was chosen after a thorough evaluation of the US EPA toxicity release inventory (TRI) “form R” reports (Palawat et al., 2023). However, selecting a single point source for this study may not comprehensively capture the myriad point sources of pollution present in Tucson, as is often the case in urban environments. In addition to point sources such as industrial facilities, nonpoint sources such as vehicular emissions, contamination from road construction, commercial and industrial emissions, and urban landscaping practices significantly contribute to the degradation of environmental quality (Deshmukh et al., 2020; Dietrich et al., 2021; Paton & Haacke, 2021; Pitt et al., 2005). These nonpoint sources compound environmental challenges by collectively exacerbating contamination, making contamination a ubiquitous issue regardless of proximity to specific industrial facilities. Therefore, utilizing proximity to a single source as a metric for understanding or predicting pollution loads in urban cities is inadequate.

The ubiquity of environmental pollution in Tucson indicates shared environmental challenges. However, there remains an inequitable distribution of environmental burdens across Tucson. Wards one, five, and six, predominantly situated in the city of South Tucson, presented elevated concentrations of Mo and Sn. The clustering of higher concentrations of Mo and Sn in wards one, five, and six signifies disproportionate exposure to environmental hazards among residents in these areas. Historically marginalized and disadvantaged populations residing in South Tucson often bear the brunt of environmental pollution due to discriminatory land-use policies, industrial siting decisions, and limited access to resources and political power (Gee & Payne-Sturges, 2004; Maantay, 2001; Maantay et al., 2010). This localized increased exposure makes a strong case for EJ within this community, particularly regarding their racial and socioeconomic composition (Palawat et al., in prep.). As such, wards one, five, and six represent important EJ communities within Tucson, where addressing environmental inequalities is paramount.

4.5. Non-urban/Rural Mining Communities

Place serves as a critical lens for understanding environmental injustices. Geographic location not only dictates proximity to pollution sources but also intersects with socioeconomic factors, shaping vulnerability to environmental injustices (Banzhaf et al., 2019; Chakraborty et al., 2011). In the present study, the highest PLIs were observed in GM and HW, the communities with active mining and smelting activities. Rainwater samples collected close to active copper, coal, and lead mines have been reported to contain relatively high metal(loid) concentrations (Mahato et al., 2016; Palawat et al., 2023a; Wright et al., 2024).

Further supporting this interpretation, Palawat et al. (2023a) conducted multifactor analysis, correlation analysis, and statistical modeling across the same four communities and observed significant associations between RHRW As/Pb concentrations and proximity to pollution point sources, particularly in HW and GM. Additionally, a recent Pb source apportionment study by Alqattan et al. (2025) confirmed that mining was the dominant source of lead in GM, further reinforcing the link between active extractive activities and elevated metal levels in rooftop-harvested rainwater.

In this study, proximity to a point source had a negative relationship with the pollution load index in rural active mining communities. Therefore, individuals living close to a mine are exposed to greater pollution loads than are those living further away (Chakraborty et al., 2011; Maantay et al., 2010; Palawat et al., 2023a; Peng et al., 2023). While there is overall increased exposure for all community members, individuals residing in close proximity to point sources may experience greater hazard exposure and health risks, including elevated rates of respiratory illnesses, cancer, birth defects, and other health disparities linked to environmental exposures (Brender et al., 2011; U.S. EPA, 2017). This study emphasized the significant impact of proximity to resource extraction on community health outcomes in communities with active mining and/or smelting activities.

It is important to study the differential pollution loads in rural mining communities because of structural vulnerabilities and infrastructural inadequacies that are prominent in rural settings. The public health challenges in the rural U.S are twofold. First, compared with those in urban communities, health outcomes are worse, especially for communities with active resource extraction and underfunded health departments (Leider et al., 2020).

Second, air quality monitoring data show that urban areas have more intensive monitoring than do rural areas (U.S. EPA, 2023; Zeider et al., 2021, 2023). In the current study, only one air quality monitor was located in GM, an active mining community and HW, home to an alternative superfund site. DH had no air monitor within a 50-mile radius, even though it is home to the legacy Iron King Mine and the Humboldt Smelter Superfund site. However, in TU, there were nine PM₁₀ monitors in the Tucson metropolitan area. These structural vulnerabilities and infrastructural inadequacies, which lead to information disparities in rural areas, may exacerbate the impact of resource extraction activities on community health and well-being and compound people’s susceptibility to environmental exposures.

Additionally, DH with legacy mining activity had a PLI>1, indicating the lasting effect of legacy contamination (Cooke et al., 2024; Nascimento et al., 2023). However, the DH PLI was much lower than those of GM and HW (active mining communities) and TU (urban community). The low PLI observed in DH may be attributed to the extensive clean-up activities organized by the USEPA and Arizona Department of Environmental Quality (ADEQ) that spanned 2008–2023, involving residential surface soil clean up and ongoing environmental monitoring (ADEQ, 2023). Consequently, the concentration of metal(loid)s in rainwater collected from areas with legacy contamination that have undergone cleanup is significantly lower than that in rainwater collected from areas with active mining and urban communities, emphasizing the environmental impact of remedial management and active contamination sources on RHRW quality.

4.6. Limitations

An acknowledged limitation of this study was the inconsistent sampling frequency and the limited demographic representation of project participants, as described by Davis et al. (2020), Moses et al. (2022), Moses et al. (2023a), and Palawat et al. (2023a). Additionally, although PM₁₀ concentration data from the AirData Air Quality Monitoring Program was used to explore the relationship between air quality and PLI, the extent of correlation is limited; PM₁₀ measurements are based solely on particle size (≤10 microns) rather than chemical composition and thus do not provide metal(loid) concentration data.

The PLI itself also has important limitations. First, by reducing the complexity of environmental pollution to a single numerical value, the PLI may oversimplify the diverse and potentially synergistic impacts of individual contaminants. This aggregation can mask the specific toxicological or environmental effects of certain pollutants that would otherwise warrant separate attention. Second, the PLI assumes equal weighting for all contaminants, which may not accurately reflect their varying degrees of environmental or health risk. Furthermore, the PLI employed in this study does not capture microbial or organic contaminants, which can also affect water quality and public health outcomes. Another challenge in applying the PLI to RHRW is the absence of enforceable water quality standards for harvested rainwater; therefore, background concentrations from the National Atmospheric Deposition Program (NADP) were used as reference values. Our focus, however, was to highlight the degree of contamination irrespective of potential water use, particularly given that current standards and guidelines do not account for exposure to contaminant mixtures.

Finally, it is important to emphasize that this study was designed to apply and evaluate the use of the PLI as a tool for comprehensive water quality assessment. While other analytical approaches such as principal component analysis (PCA) or multiple factor analysis (MFA) could provide additional insights into contamination patterns and potential sources, they were beyond the scope of this study. This work does not present a human health risk assessment or a source apportionment analysis; rather, it offers a focused examination of metal(loid) contamination in rooftop-harvested rainwater systems using the PLI framework.

Despite its limitations, the PLI remains a valuable tool in the broader context of environmental management and policy development. PLI provides a convenient way to assess overall pollution levels and can serve as a useful tool for initial screening and comparison purposes. Using the PLI to assess water quality provides a quick method for evaluating overall water quality conditions. Therefore, the PLI provides a straightforward and systematic approach for evaluating the quality of harvested rainwater by aggregating multiple pollutants into a single index. This simplification enables quick assessments of the potential risks associated with rooftop rainwater harvesting, aiding in decision-making for water use and treatment strategies.

5. Conclusion

The goal of this study was to investigate metal(loid) contamination in RHRW systems across four EJ communities in Arizona. Central to our investigation was the use of the PLI and NIPI to distill complex contaminant data into composite indices, offering a comprehensive snapshot of RHRW quality. The PLI, which compares contaminant levels to reference background samples, proved to be a flexible and context-sensitive indicator, well-suited for characterizing cumulative pollution in multifaceted and unregulated water use scenarios such as rainwater harvesting. In contrast, NIPI, which benchmarks contaminant concentrations against established regulatory standards/recommendations, provided a designated-use-oriented perspective on pollution severity.

In this study, PLI offered broader applicability given the lack of regulatory benchmarks for most RHRW uses. However, NIPI’s inclusion strengthened the study by highlighting how pollution classifications can shift significantly depending on the intended water use and applied benchmarks/thresholds. This dual-index approach enriches our understanding by presenting complementary perspectives: PLI reflects cumulative contamination burden, while NIPI contextualizes this burden within potential water uses. Importantly, the strong positive correlation observed between PLI and NIPI across all use categories supports the validity of both indices and affirms their utility. Together, they provide a nuanced understanding of pollution dynamics in RHRW and strengthen the case for public health interventions.

Land use is a significant driver of pollution load, and a disproportionate burden of environmental pollution is faced by non-urban/rural communities near resource extraction activities. In these communities, location influences environmental quality, with proximity to point sources of pollution correlated with increased contamination risk. On the other hand, urban areas are characterized by multiple layers of pollution from point and nonpoint sources, leading to an increased PLI in RHRW. Moreover, atmospheric and rooftop dust accumulation influenced harvested rainwater quality, particularly through interactions with dust-borne calcium carbonate and subsequent increases in pH. Our study identified pH as a significant driver of pollution load, where rainwater with higher pH values presented elevated PLI values. Additionally, the pollution load varied with season, with a greater PLI observed during the monsoon season than during the winter months, which was attributed to greater dust and rain events.

As climate change intensifies drought conditions in arid regions, the role of RHRW as a supplementary water source will grow. By integrating both PLI and NIPI, this study provides a robust and multifaceted assessment framework that can inform sustainable rainwater harvesting practices, guide regulatory development, and foster transparency in water quality communication. Ultimately, our findings highlight the urgent need for sustainable interventions, equitable access to safe water, and policy frameworks that reflect the diverse uses of harvested rainwater. Prioritizing environmental justice in these efforts is essential for addressing systemic disparities, protecting public health, and ensuring a resilient and equitable water future for all.

Highlights.

1. Pollution load index (PLI) used to determine rooftop harvested rainwater quality

2. PLI primary drivers are industrial activity, pH, and seasonal changes

3. Proximity to pollution point source affect PLI in rural communities

4. Urban communities have multiple, diverse pollution sources impacting PLI

5. Environmental justice issues exacerbate rooftop harvested rainwater PLI challenges

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request. The corresponding author is working to make the data available in a public repository.