Hydrogeology Journal

, Volume 25, Issue 4, pp 969–979

Drinking-water treatment, climate change, and childhood gastrointestinal illness projections for northern Wisconsin (USA) communities drinking untreated groundwater

  • Christopher K. Uejio
  • Megan Christenson
  • Colleen Moran
  • Mark Gorelick
Paper

Abstract

This study examined the relative importance of climate change and drinking-water treatment for gastrointestinal illness incidence in children (age <5 years) from period 2046–2065 compared to 1991–2010. The northern Wisconsin (USA) study focused on municipalities distributing untreated groundwater. A time-series analysis first quantified the observed (1991–2010) precipitation and gastrointestinal illness associations after controlling for seasonality and temporal trends. Precipitation likely transported pathogens into drinking-water sources or into leaking water-distribution networks. Building on observed relationships, the second analysis projected how climate change and drinking-water treatment installation may alter gastrointestinal illness incidence. Future precipitation values were modeled by 13 global climate models and three greenhouse-gas emissions levels. The second analysis was rerun using three pathways: (1) only climate change, (2) climate change and the same slow pace of treatment installation observed over 1991–2010, and (3) climate change and the rapid rate of installation observed over 2011–2016. The results illustrate the risks that climate change presents to small rural groundwater municipalities without drinking water treatment. Climate-change-related seasonal precipitation changes will marginally increase the gastrointestinal illness incidence rate (mean: ∼1.5%, range: −3.6–4.3%). A slow pace of treatment installation somewhat decreased precipitation-associated gastrointestinal illness incidence (mean: ∼3.0%, range: 0.2–7.8%) in spite of climate change. The rapid treatment installation rate largely decreases the gastrointestinal illness incidence (mean: ∼82.0%, range: 82.0–83.0%).

Keywords

USA Climate change Health Rainfall/runoff Municipal groundwater 

Traitement de l’eau potable, changement climatique, et projections des maladies gastro-intestinales chez l’enfant dans les collectivités du nord du Wisconsin (Etats-Unis d’Amérique) buvant de l’eau souterraine non traitée

Résumé

Cette étude a examiné l’importance relative du changement climatique et du traitement de l’eau potable pour l’incidence de la maladie gastro-intestinale chez les enfants (<5 ans) de la période 2046–2065 par rapport à 1991–2010. L’étude du nord du Wisconsin (Etats-Unis d’Amérique) a porté sur les municipalités qui distribuent de l’eau souterraine non traitée. Une analyse de séries chronologiques a d’abord quantifié les précipitations observées (1991–2010) et les associations avec des maladies gastro-intestinales après un contrôle de la saisonnalité et les tendances temporelles. Les précipitations ont probablement transporté des pathogènes vers des sources d’eau potable ou vers des réseaux de distribution d’eau comportant des fuites. S’appuyant sur les relations observées, la deuxième analyse a montré comment le changement climatique et des installations de traitement de l’eau potable pourraient modifier l’incidence des maladies gastro-intestinales. Les précipitations futures ont été modélisées à l’aide de 13 modèles climatiques globaux et trois niveaux d’émissions de gaz à effet de serre. La deuxième analyse a été relancée en utilisant trois voies : (1) le changement climatique seulement, (2) le changement climatique et la même allure lente de déploiement des installations de traitement observée sur la période 1991–2010, et (3) le changement climatique et un rythme rapide de déploiement d’installations de traitement observé sur la période 2011–2016. Les résultats illustrent les risques que présentent le changement climatique pour les petites municipalités rurales alimentées par les eaux souterraines sans traitement d’eau potable. Les changements de précipitations saisonniers liés au changement climatique augmenteront légèrement le taux d’incidence de la maladie gastro-intestinale (moyenne: ∼1.5%, gamme: −3.6–4.3%). Un lent déploiement des installations de traitement conduit à une légère diminution de l’incidence de la maladie gastro-intestinale associée aux précipitations (moyenne: ∼ 3.0%, gamme: 0.2–7.8%) malgré le changement climatique. Un taux rapide d’installations de traitement de l’eau diminue largement l’incidence de la maladie gastro-intestinale (moyenne: ∼82.0%, gamme: 82.0–83.0%).

Tratamiento de agua potable, cambio climático y enfermedades gastrointestinales infantiles en las comunidades del norte de Wisconsin (EE.UU.) que beben agua subterránea no tratada

Resumen

Este estudio examinó la importancia relativa del cambio climático y el tratamiento del agua potable para la incidencia de enfermedades gastrointestinales en niños (edad <5 años) del período 2046–2065 en comparación con 1991–2010. El estudio en Wisconsin del norte (EE UU), se centró en los municipios que distribuían agua subterránea no tratada. Un análisis de series de tiempo cuantificó primero las asociaciones de precipitación y enfermedad gastrointestinal observadas (1991–2010) después de controlar la estacionalidad y las tendencias temporales. Es probable que la precipitación transportara patógenos a fuentes de agua potable o a redes de distribución de agua con filtraciones. Sobre la base de las relaciones observadas, el segundo análisis proyecta cómo el cambio climático y la instalación de tratamiento de agua potable puede alterar la incidencia de enfermedades gastrointestinales. Los futuros valores de precipitación fueron modelados por 13 modelos climáticos globales y tres niveles de emisiones de gases de efecto invernadero. El segundo análisis se reanudó utilizando tres trayectorias: (1) sólo el cambio climático, (2) el cambio climático y el mismo ritmo lento de instalación de tratamiento observado durante 1991–2010, y (3) el cambio climático y la rápida instalación observada durante el período 2011–2016. Los resultados ilustran los riesgos que el cambio climático presenta a las pequeñas municipalidades rurales con agua potable de aguas subterráneas sin tratamiento. Los cambios de precipitación estacional relacionados con el cambio climático aumentarán marginalmente la tasa de incidencia de la enfermedad gastrointestinal (media: ∼ 1.5%, rango: −3.6–4.3%). Un ritmo lento de la instalación de tratamiento disminuyó ligeramente la incidencia de la enfermedad gastrointestinal asociada a la precipitación (media: ∼ 3.0%, rango: 0.2–7.8%) a pesar del cambio climático. La tasa de instalación rápida del tratamiento disminuye en gran medida la incidencia de la enfermedad gastrointestinal (media: ∼ 82.0%, rango: 82.0–83.0%).

对(美国)威斯康星州北部社区饮用未处理的地下水进行的饮用水处理、气候变化及童年胃肠疾病的预测

摘要

本研究论述了2046–2065年与1991–2010年相比,气候变化和童年(小于5岁)期胃肠疾病发生率饮用水处理的相对重要性。在(美国)威斯康星州北部进行的研究集中在提供未处理地下水的社区。时间序数分析首次量化了观测到(1991–2010年)的降水和季节性及时间趋势控制后胃肠疾病的关联性。降水很可能传输病原体到饮用水源中或渗漏的供水网络中。在观测的关系基础上,二阶分析预测了气候变化和饮用水处理装置是怎样可能改变胃肠疾病发生率的。采用13个全球气候模型和三个温室气体排放标准模拟了未来的降水值。采用三个途径再进行二阶分析:(1)单单气候变化;(2)气候变化及1991–2010年观测到的处理装置相同的缓慢步伐;(3)气候变化及2011–2016年观测到的快速的装置比率。结果描述了气候变化对饮用水未处理的、采用地下水的小的乡村社区的风险。气候变化相关的季节性降水变化将最低限度地增加胃肠疾病的发病率(平均值:∼1.5%,范围:–3.6–4.3%)。 尽管有气候变化,但处理装置的缓慢步伐多少降低了降水相关的胃肠疾病发病率(平均值:∼3.0%,范围:0.2–7.8%)。快速的处理装置率大大降低了胃肠疾病的发病率(平均值:∼82.0%,范围:82.0–83.0%)

Projeções de tratamento de água para consumo, mudança climática e doenças gastrointestinais em comunidades do Norte de Winsconsin (EUA) consumindo água subterrânea não tratada

Resumo

Esse estudo investigou a importância relativa das mudanças climáticas e do tratamento da água para abastecimento na incidência de doenças gastrointestinais em crianças (idade < 5 anos) para o período 2046–2065 comparado a 1991–2010. O estudo no norte de Winsconsin (EUA) focou em municípios que distribuem água subterrânea não tratada. Uma análise de séries temporais quantificou as associações entre precipitação e doenças gastrointestinais observadas (1991–2010) após controle da sazonalidade e tendências temporais. A precipitação provavelmente transportou patógenos nas fontes de água para abastecimento ou nas redes de distribuição de água com vazamentos. Construída a partir de relações observadas, a segunda análise previu como as mudanças climáticas e a instalação do tratamento da água para abastecimento podem alterar a incidência de doenças gastrointestinais. Valores futuros de precipitação foram modelados por 13 modelos climáticos globais e três níveis de emissões de gases de efeito estufa. A segunda análise foi refeita utilizando três caminhos: (1) apenas mudanças climáticas, (2) mudanças climáticas e o mesmo passo lento na instalação do tratamento observado de 1991 a 2010, e (3) mudanças climáticas e a taxa rápida de instalação observadas de 2011 a 2016. Os resultados ilustram os riscos que as mudanças climáticas apresentam para municípios rurais que utilizam águas subterrâneas sem tratamento de água para abastecimento. Mudanças na precipitação sazonal relacionada às mudanças climáticas aumentarão marginalmente a taxa de incidência de doenças gastrointestinais (média: ∼1.5%, alcance: –3.6–4.3%). Um passo lento na instalação do tratamento de alguma forma diminui a incidência de doenças gastrointestinais associadas a precipitação (média: ∼3.0%, alcance: 0.2–7.8%) apesar das mudanças climáticas. Uma taxa rápida de instalação do tratamento diminui profundamente a incidência de doenças gastrointestinais (média: ∼82.0%, alcance: 82.0–83.0%).

Introduction

Contaminated drinking water is responsible for a notable disease burden in middle and high-income countries. In the US, pathogens in public drinking-water systems cause approximately 4.3–16.4 million gastrointestinal illnesses (GI) per year (Colford et al. 2006; Messner et al. 2006; Reynolds et al. 2008). More specifically, community and non-community groundwater systems respectively contribute to 1.3–5.4 million and 1.1 million GI/year (Colford et al. 2006; Reynolds et al. 2008). There is more uncertainty surrounding groundwater versus surface-water disease burden estimates. Waterborne disease contracted from drinking water sources can result from a variety of mechanisms. For example, a proportion of Cryptosporidium oocysts can survive treatment and present health problems for populations with pre-existing conditions (Yoder and Beach 2010). Water-service-delivery outages, low water pressure, or insufficient chlorine disinfectant residuals may also increase GI risk in finished water distribution systems (Ercumen et al. 2014; Lambertini et al. 2012).

Water treatment and a well-functioning distribution system improve water quality and reduce GI illness, physician visits, and outbreaks (Craun and Calderon 2006; Kuusi et al. 2003; Teschke et al. 2010; Tulchinsky et al. 2000; Beaudeau et al. 2010). A well-functioning distribution system maintains physical, hydraulic, and water quality integrity from the treatment facility to the point of use (National Research Council 2006). In the US, there are different regulations for municipal surface water, groundwater and private wells. Public surface-water systems must meet minimum treatment levels under the Surface Water Treatment Rules. Existing public groundwater regulations (Groundwater Rule) mandate source protection and water quality monitoring but do not require treatment. Federal regulations do not apply to private wells. Certain US states require private well-water testing when buying or selling property.

The drinking water source and its associated regulations have important implications on waterborne pathogens and associated human illnesses. Across the US, approximately 27% of public groundwater supply wells contain human enteric viruses (US Environmental Protection Agency 2006). Households consuming water from wells in areas with a higher density of septic systems are more likely to contract GI (Borchardt et al. 2003). A randomized community intervention of disinfection in an area served by municipal groundwater attributed 6–22% of GI to drinking-water viral pathogens (Borchardt et al. 2012).

In general, precipitation has been shown to increase GI pathogen concentrations in both surface and groundwater (Corsi et al. 2014; Dwight et al. 2002; Harper et al. 2011; Abbaszadegan et al. 2003; Bradbury et al. 2013). Moreover, several studies have demonstrated a positive association between precipitation and GI incidence. A study in northern Wisconsin found precipitation increased GI cases in communities served by untreated municipal groundwater (Uejio et al. 2014), while another study in Milwaukee found a similar association between increased precipitation and GI in an area served primarily by municipal treated surface water (Drayna et al. 2010). Similarly, in Massachusetts, precipitation was associated with GI in areas where sewage/stormwater runoff discharged into a surface-water drinking source (Jagai et al. 2015). This systematic relationship implies that hydrologic events are transporting pathogens into drinking water sources or finished water distribution networks (Naumova et al. 2005).

Climate change may intensify the hydrologic cycle and projections suggest annual Wisconsin precipitation may increase 4–8% by midcentury (Trenberth 1999; Vavrus and Behnke 2014); furthermore, physical theory and climate models suggest extreme precipitation events (>5.1 cm) will become stronger and more frequent (Trenberth 1999; Vavrus and Behnke 2014). Large precipitation events may promote microbial leaching from septic systems through the soil and into groundwater (Nicosia et al. 2001; Shadford et al. 1997). Similarly, there is some evidence that extreme precipitation events disproportionately (non-linearly) increase the risk of contracting GI (Curriero et al. 2001; Harper et al. 2011; Naumova et al. 2005).

Given the observed associations between precipitation and GI, increased illness may be an anticipated consequence of climate change, especially among younger children who are more susceptible to and more likely to suffer complications from waterborne pathogens. On the other hand, installing water treatment is an important barrier against waterborne pathogens and may be an adaptation strategy, especially in areas regulated under the less stringent Groundwater Rule. The study quantifies how a range of projected precipitation changes and drinking-water treatment installation upgrades may alter childhood GI incidence rates. Predictive statistical models use 13 high resolution (downscaled) precipitation projections, three greenhouse gas emissions scenarios, and two treatment installation rates. The study compared actual childhood GI incidence rates in the period 1991–2010 with predicted rates for 2046–2065. The study considered three plausible future pathways: (1) only climate change, (2) climate change and a slow rate of treatment installation (similar to the rate of installation observed for 1991–2010) or (3) climate change and a rapid rate of installation (the rate observed for 2011–2016).

Materials and methods

Study area

The study area includes five northern Wisconsin municipalities distributing minimally treated groundwater served by the Marshfield Clinic (Fig. 1). The Marshfield Clinic provided healthcare to almost all (∼95%) of the 4,097 residents in the five municipalities including 307 children (DeStefano et al. 1996; US Census Bureau 2016). The study focused on young children (age < 5 years) with a home address in a study municipality. The study area is characterized by agriculturally productive Antigo soils which include Almena, Auburndale, Brit, Onamia, and Spencer silt loams and peat soils (Voigtlander 2008). The study municipalities minimally treated their drinking water through fluoridation and adjusting the water’s pH to control corrosion. Thus, conventional drinking-water treatments such as flocculation, coagulation, settling, filtration, or disinfection were not applied to the drinking water.
Fig. 1

Map of the study area’s five municipalities accessing minimally treated groundwater. The Marshfield Clinic’s Healthcare facilities are located in the study municipalities

Data

Health outcomes and population

The study reanalyzed a time series of GI healthcare visits from young children (age < 5 years, N = 291) over the summer and fall (June–November) from period 1991–2010. In the study area, only summer and fall GI cases were systematically related to precipitation (Uejio et al. 2014). This may be due to the changing composition of pathogens in the winter/spring (e.g. norovirus, rotavirus) versus the summer/fall (e.g. enterovirus, cryptosporidium). In the United States, healthcare providers assign one or more International Classification of Diseases (ICD, 9th revision) billing codes to each healthcare encounter based on a patient’s symptoms, diagnoses, or procedures. A billing code is not necessarily a medical diagnosis. Billing codes are primarily collected to reimburse healthcare providers but are also widely used in public health research.

GI cases recorded one of the following codes in the first or second billing position: specified gastrointestinal infections such as cholera, salmonellosis (including typhoid/paratyphoid), shigellosis, amebiasis, protozoal intestinal disease, and other bacterial and viral pathogens (code 001-009.9). The case definition also included broader codes for unspecified gastroenteritis (code 558.9) or diarrhea (code 787.91) since GI is infrequently attributed to a specific pathogen (Gangarosa et al. 1992; Hoxie et al. 1997). The case definition excluded patients with laboratory confirmed localized Salmonella infections (003.2), staphylococcal, and Clostridia food poisoning (005–005.3). The case definition also excluded patients who filled or were prescribed antibiotics within 7 days of the GI visit. Finally, cases with repeated GI visits up to 30 days after the initial visit were excluded to increase the likelihood that the visits were separate illnesses. The US Census Bureau’s (2016) year 2000 census, which was in the middle of the study period, provided observed counts of children.

Observed and projected climate information

The Parameter-elevation Regressions on Independent Slopes Model (PRISM) provided historical daily precipitation at a 4-km2 resolution. PRISM incorporates ancillary geographic information (e.g. elevation) to statistically interpolate conditions in between weather observations (Daly et al. 2008). PRISM is well correlated with the study area’s daily precipitation from weather stations (Pearson’s Correlation Coefficient: 0.87; NOAA 2016). The present study used PRISM instead of weather stations to align with the geographic resolution of the climate change projections.

The climate projection uncertainty can be attributed to natural climate variability, and societal and scientific processes (Hawkins and Sutton 2011). Natural climate variability refers to normal cycles of drier and wetter than average conditions. The 20-year study period minimized but did not eliminate climate variability uncertainty. The Special Report on Emissions Scenarios (SRES) provides societal uncertainty information (Nakicenovic and Swart 2000).

The present study used three SRES pathways (A1B, A2, B1) to illustrate future scenarios of changes in population, economic growth and distribution, technology, and policy in the year 2050 (Table 1). Population growth follows the same trajectory in the A1B and B1 scenarios. The A1B scenario contains rapid economic and energy efficiency growth and a balance of fossil fuel and alternative energy sources. In the A2 scenario, fragmented international relationships lead to regional economic and slower technological growth but larger human population levels compared to the other two scenarios. In the B1 scenario, there is a large shift toward service occupations, alternative energy sources, global economic solutions, and environmental sustainability. For the years 2046–2065 (represented by the SRES projection to the year 2050), the greatest greenhouse gas emissions levels are produced by the A2 scenario, followed by A1B, and B1.
Table 1

The key characteristics between the three emissions scenarios (A1B, A2, B1) for population, economic growth, technology, and greenhouse gas emissions in 2050 (Nakicenovic and Swart 2000)

 

Scenario in 2050

A1B

A2

B1

Population (billion)

8.7

11.3

8.7

World GDP (1012 1990 US $/year)

181

82

136

Per capita income ratio: developed countries and economies in transition to developing countries

2.8

6.6

3.6

Primary energy (1018 J/year)

1,347

971

813

Share of coal in primary energy (%)

14

30

21

Share of zero carbon in primary energy (%)

36

18

30

Carbon dioxide emissions, fossil fuels (gigatons of carbon/year)

16.0

16.5

11.7

Carbon dioxide emissions, land use (gigatons of carbon/year)

0.4

0.9

−0.4

The A1B scenario corresponds to rapid economic and energy efficiency growth. The B1 scenario produces the lowest greenhouse gas emissions. Under the A2 scenario, there is gradual economic and technological growth and larger human population levels

Scientific uncertainty is inherent in modeling precipitation and climate change. Climate models may represent precipitation, climate change feedbacks, and sensitivities to greenhouse gas emissions in different ways. The Wisconsin Initiative on Climate Change Impacts (WICCI) provided high resolution or downscaled (∼10-km resolution) Coupled Model Intercomparison Project Phase 3 projections from the period 2046–2065. Using precipitation values from a large suite of models provides information on scientific uncertainty. This study used all available WICCI models which included 13 models for the A1B and B1 scenarios and nine from the A2 scenario (Table 2).
Table 2

Wisconsin Initiative on Climate Change Impacts Global Climate Models that provided precipitation projections for 2046–2065. The table lists the model’s country of origin, horizontal resolution, and number of vertical levels, before statistical downscaling to ∼10 km

Model name

Country

Horizontal resolution [km]

Vertical resolution [levels]

CGCM3.1 (T47)

Canada

3.75

L31

CGCM3.1 (T63)

Canada

2.8

L31

CNRM_CM3

France

2.8

L45

CSIRO_mk3.0

Australia

2.8

L18

CSIRO_mk3.5

Australia

2.8

L18

GFDL_CM2.0

United States

2.5

L24

GISS_AOM

United States

3.5

L12

IAP_FGOALS

China

2.8

L26

MIROC3.2 (medium resolution)

Japan

3.5

L20

MIROC3.2 (high resolution)

Japan

1.2

L56

ECHO_G

Germany/Korea

3.75

L19

MPI_ECHAM5

Germany

2.8

L31

MRI_CGCM2.3.2a

Japan

3.5

L30

The WICCI projections are statistically downscaled and debiased using quantile mapping to remove systematic errors and provide high spatial resolution information (Wood et al. 2004). Projected daily precipitation values were created by randomly sampling the downscaled probability density function. For each emissions scenario and global climate model, there are three model runs initialized with slightly different conditions. The analysis used the average of the three daily projection runs.

Time series analysis

The time series analysis associated weekly precipitation against GI cases during summer and fall (June to November) over 1991–2010. One strength of the procedure is that it compares a population against itself to “cancel out” risk factors that gradually change over time. Time series analysis was conducted with generalized additive models (GAM) in R using the gamm4 package (Wood 2006). A negative binomial distribution analyzed the overdispersed (variance > mean) GI counts.

Equation (1) outlines the GAM structure where μ represents the expected weekly GI counts. The equation’s intercept is β0 and monthly indicator variables (μmonth) controlled for seasonality. The beta coefficient (β1) measures the strength of the association between weekly precipitation (X1) and GI counts. Yearly indicator variables (μyear) and a cubic regression spline (s, up to 10 degrees of freedom) of the study week (X2) controlled for secular trends. Autocorrelation and partial autocorrelation functions of the GAM residuals verified that temporal autocorrelation had been adequately controlled.
$$ \log \left(\mu \right)={\beta}_0+{\mu}_{\mathrm{month}}+{\mu}_{\mathrm{year}}+{\beta}_1\cdotp {X}_1+s\left({X}_2\right) $$
(1)

There may be a temporal lag between precipitation and GI healthcare visitations. This period may be related to hydrologic transport, the pathogen’s incubation period, and healthcare seeking behavior. Pathogen transport, survival, and groundwater hydrogeology studies have not been conducted in the study area’s unconsolidated sand and gravel aquifer. Studies from comparable aquifers suggest groundwater flow velocities are on the order of meters per day but may be over 100 m/day in riverbank filtration areas (Pang 2009). Precipitation may further increase the rate of virus transport by mobilizing viruses and colloids and promoting soil transport (Nicosia et al. 2001; Shadford et al. 1997); thus, precipitation may plausibly transport pathogens into groundwater sources over a wide range of periods (days to weeks).

A complimentary study commonly elucidated adenoviruses, enteroviruses, and norovirus genogroup I in the study area’s drinking water systems (Borchardt et al. 2012). The norovirus incubation period is short (∼1.7 days) compared to adenoviruses and enteroviruses (2–10 days; Lee et al. 2013). After experiencing symptoms, there may be a delay before a patient is seen by a medical provider. A Milwaukee (Wisconsin) study suggested this period was 2 days but it may be slightly longer in areas located far from a healthcare provider (Gorelick et al. 2011); thus, there may be a few-days to multi-week lag between precipitation events and subsequent GI healthcare visitations.

Model-building selected the best-fitting concurrent or preceding weeks’ precipitation metric. A screening procedure separately associated each precipitation lag against cases while controlling for seasonality and secular trends. The best-fitting (lowest deviance) statistical model provided information for the future disease projections. The time series results are reported as the change in the relative risk of reporting GI per centimeter of precipitation. For a relative risk of one, there is no relationship between precipitation and GI. Similarly, relative risks significantly greater than one are positively associated with GI.

Projected climatic change and drinking-water treatment installation

The primary goal of projections in this study is to develop a plausible range of future childhood GI rates. The methodology was developed for projecting the benefits of air pollution interventions but has since been applied to other health outcomes (National Research Council 2002; Peng et al. 2011). Figure 2 outlines the data sources and steps in projecting the future GI disease burden. The time series quantified the observed relationship between precipitation and childhood GI over 1991–2010. This statistical model forms the basis for estimating a range of future GI over 2046–2065. Separate estimates are created for the three SRES pathways and three drinking-water treatment installation rates (nine total estimates). Each estimate contains a distribution of GI rates corresponding to precipitation changes projected by the high resolution climate projections.
Fig. 2

Flow diagram outlining the relationship between the study data sources and analysis. A time series analysis first quantified the observed relationship between precipitation and childhood gastrointestinal illness over 1991–2010. Next, this statistical model projects future gastrointestinal illness based on a range of drinking-water treatment and climate-change scenarios

Equation (2) estimates precipitation-associated GI counts in 2046–2065 (F) from GI counts in weeks without precipitation (Yb), time series precipitation relative risk (B1), and projected population at risk (P). Precipitation associated GI comprises a portion of the total GI disease burden. For each SRES scenario, the study created a distribution of projected GI by substituting each climate model’s weekly projected summer and fall precipitation values (X1) over 2046–2065.
$$ F={Y}_{\mathrm{b}}\cdotp \left({e}^{B_1\cdotp {X}_1^{\prime }-1}\right)\cdotp P $$
(2)

This study considered three future GI disease burden pathways. The first pathway only considered future precipitation’s (X1) impact on GI incidence. The size of the population at-risk (P) is the same as the observational period (P1991–2010). The study presumes conventional treatment and chlorination will produce a 4-log virus removal or inactivation. The second and third pathways considered both future precipitation and very different rates of installing municipal drinking water treatment. Water treatment was modeled by decreasing the future population at-risk by the same number of children that gained access to treated water (P = P1991–2010Pnew treatment). Thus, the projections presumed that the observed relationship between precipitation and GI cases (B1) are the same in the future.

In the study area, there was no relationship between precipitation and GI in municipalities with treated drinking water (Uejio et al. 2014). The second pathway applied the amount of water treatment installation observed over 1991–2010 to the future period. This relatively slow rate decreased the population accessing untreated water by 4.3%. Only one municipality serving 13 children (Pnew treatment) started chlorinating their water supply. The third pathway considered the rapid rate of installation over 2011–2016. The population accessing untreated water decreased by 82.5%, as two larger municipalities serving 132 children (Pnew treatment) initiated conventional treatment. The projections are reported as an annual incidence per 10,000 children by dividing precipitation-associated GI counts (F) by the observed population (P1991–2010) and multiplying by 10,000. The projected rates are compared against the observational period.

Results

Observational period

This section summarizes the demographic and epidemiological characteristics of cases in the study area. Male children comprised a larger proportion of cases (62.5%) than female (37.5%) children. There was a higher proportion of cases in the youngest age categories [30% (<1 year), 25% (1 to < 2 years)] than the slightly older groups—17% (2 to < 3 years), 17% (3 to < 4 years), and 11% (4 to < 5 years). The most common types of patient health insurance were Medicaid (52.5%), commercial insurance (33.3%), and other (14.2%); thus, lower income households suffer from a disproportionately higher GI disease burden.

The summer and fall GI incidence rate was 397 cases per 10,000 children. Almost all patients first received care during clinic visits (83.3%) compared to the emergency room (10.8%) and inpatient hospital care (4.2%). Correspondingly, most patients (83.3%) only required medical attention on the first day of his/her healthcare visit. Of the remaining patients, the median inpatient stay was 6 days (range: 1–22 days).

Table 3 reports the time series models that associated precipitation against childhood GI cases at five different temporal lags (0–4 weeks prior). The table reports the relative risk adjusted for seasonality and secular trends, 95% confidence intervals, p-values, and goodness of fit (deviance). The concurrent week’s precipitation was the best fitting time series model. Each centimeter of weekly precipitation linearly increased GI by 11% (adjusted risk ratio: 1.11, 95% confidence interval: 1.03–1.19, p < 0.01). In the fall, GI cases seasonally increased, but not significantly, compared to the summer months. Temperature was not significantly associated with GI incidence. The precipitation time series models for 2 and 4 weeks prior suggest an inverse precipitation and GI relationship but the associations are marginally statistically significant (p > = 0.05). In the study area, the current week’s precipitation tends to be negatively associated with precipitation 4 weeks prior (p = 0.16).—for this reason, the marginally significant 2 and 4 week precipitation lags may be capturing temporally structured weather changes.
Table 3

Time series results associating precipitation over different temporal lags against childhood gastrointestinal illness cases

Model

Relative risk

95% confidence interval

p values

Deviance explained

Concurrent precipitation (cm)

1.11

1.03, 1.19

< 0.01

18.0

Previous week’s precipitation (cm)

1.00

0.99, 1.01

0.91

16.1

Precipitation 2 weeks prior (cm)

0.99

0.98, 1.00

0.06

17.4

Precipitation 3 weeks prior (cm)

1.00

0.99, 1.01

0.70

16.1

Precipitation 4 weeks prior (cm)

0.99

0.98, 1.00

0.05

16.6

The reported relative risk is adjusted for seasonality and long-term trends. The table also reports the upper and lower 95% relative risk confidence intervals. The concurrent week’s precipitation exhibits the strongest and most consistent association with cases

Projected disease burden

During 1991–2010, the summer season (311.9 mm/year) was typically wetter than the fall (232.1 mm/year). Table 4 summarizes the 2046–2065 summer precipitation projection changes compared to the 1991–2010 long-term average. For the three greenhouse gas scenarios, the table reports the mean and range of projected changes and the proportion of models projecting wetter conditions. Most projections suggest that summer and fall conditions will become wetter; thus, there is more scientific uncertainty about the magnitude than the direction of precipitation change. Summer is projected to receive a greater increase in precipitation (mean: 31–37 mm, range: −35–93 mm) compared to the fall (mean: 2–11 mm, range: −54–34 mm). Notably, the precipitation projections are similar in spite of different greenhouse gas emissions levels. The scientific uncertainty from climate models is larger than the societal uncertainty contained in the three scenarios.
Table 4

The projected change in average seasonal precipitation from global climate models and three scenarios in 2046–2065 compared to 1991–2010

Scenario

Summer

Fall

Mean change (range) [mm]

No. of wetter models/total models

Mean change (range) [mm]

No. of wetter models/total models

A1B

37 (−35–93)

11/13

6 (−40–27)

10/13

A2

31 (−18–55)

8/9

11 (−7–34)

8/9

B1

34 (−27–82)

10/13

2 (−54–34)

9/13

Summer was defined as June, July, August and fall as September, October, November. The A1B scenario corresponds to rapid economic and energy efficiency growth. The B1 scenario produces the lowest greenhouse gas emissions. Under the A2 scenario, there is gradual economic and technological growth and larger human population levels.

Table 5 reports the projected GI rates for nine future scenarios (three greenhouse gas emissions by three drinking-water treatment installation scenarios). The table reports the mean and range of combined summer and fall GI incidence from the time series analysis and climate models described in Table 4. The first pathway only considered future summer and fall precipitation’s (Δx) impact on GI incidence. The methodology presumes hydrologic events will continue to transport pathogens into untreated municipal groundwater in the future. Precipitation-attributable GI incidence over 2046–2065 is compared to the observed 1991–2010 rate. Since not all GI is coincident with precipitation, the total GI incidence rate (397/10,000 children) is greater than the precipitation attributable rate (230/10,000 children). Without additional drinking-water treatment installation, increased precipitation from climate change may marginally increase the GI rate 1.4–1.7% (range: –3.6–4.3%) to ∼234/10,000 children in 2046–2065. Interestingly, the projected GI rates are similar regardless of the greenhouse gas emissions scenario. By comparison, there is notably more GI variability from the different climate model’s precipitation projections; thus, GI projections are more sensitive to scientific uncertainty from the climate models than the rate of greenhouse gas emissions.
Table 5

The mean projected annual GI incidence associated with precipitation for three greenhouse gas emissions scenarios and pathways

Scenario

Climate change (range) [cases/10,000]

Climate change + adaptation (1991–2010) (range) [cases/10,000]

Climate change + adaptation (2011–2016) (range) [cases/10,000]

A1B

234 (222, 239)

224 (212, 229)

41 (39, 42)

A2

234 (227, 237)

223 (218, 227)

41 (40, 42)

B1

233 (222, 240)

223 (220, 239)

41 (39, 42)

The table also reports the range of GI incidences in parentheses. The three pathways were: (1) only climate change, (2) climate change and a slow pace of drinking-water treatment installation over 1991–2010, (3) climate change and the rapid rate of installation over 2011–2016. The A1B scenario corresponds to rapid economic and energy efficiency growth. The B1 scenario produces the lowest greenhouse gas emissions. Under the A2 scenario, there is gradual economic and technological growth and larger human population levels

Projected disease burden with drinking-water treatment installation

The second and third pathways considered both climate change and initiating drinking-water chlorination (Table 4). Both pathways modeled adaptation by lowering the population at-risk by the number of children gaining access to treated water. The second pathway’s GI incidence rate proportionately decreased (mean: 2.7–3.0%, range: 0.2–7.8%) based on the 1991–2010 treatment adoption. Treating drinking water more than compensated for the increased risk climate change presented to non-treated areas. Based on the more aggressive treatment installation over 2011–2016, the precipitation-associated GI incidence (mean: 82%, range: 82.0–83.0%) correspondingly decreased.

Discussion

A complimentary study investigated the relationship between precipitation and childhood GI across different drinking-water systems in the study area (Uejio et al. 2014). The study investigated three types of systems: treated municipal groundwater, untreated municipal groundwater, and private wells. Interestingly, precipitation was only systematically related to GI in untreated municipal drinking-water areas. Since precipitation was unrelated to GI in the other water systems, this suggests that drinking water is the primary source of precipitation associated GI risk. Nonetheless, the study design cannot rule out the contribution of other precipitation associated pathways—for example, precipitation events often increase indicator bacteria concentrations and the risk of contracting GI while swimming in recreational water bodies (e.g. Patz et al. 2008; Nevers and Whitman 2011); thus, the drinking-water treatment pathways may somewhat overstate the reduction in GI incidence.

This study applies an established methodology to quantify the benefits of additional drinking-water treatment installation compared to the risks presented by climate change. This procedure relies upon local studies showing that residents contract GI from untreated drinking water and have increased risk during precipitation events (Borchardt et al. 2012; Uejio et al. 2014). Precipitation may increase GI incidence on the order of 1.4–1.7% (range: −3.6–4.3%) without additional drinking-water treatment; thus, climate change is a secondary public health concern compared to the untreated drinking water disease burden, which may reasonably bound the climate change risks for the remaining 56 Wisconsin municipal water systems without disinfection which serve 65,000 people. Installing and maintaining drinking water treatment will address existing health disparities and make communities more resilient to climate change.

This section discusses the modeling process assumptions and major sources of uncertainty. The time series model based on observed relationships (1991–2010) projected future (2046–2065) GI risk. The relatively small GI sample size (N = 291) lowered the precision of the time series effect estimates. The statistical model presumes existing GI and precipitation relationships will apply to the future; however, changing pathogen pollution sources (e.g. aging water distribution pipes) or host susceptibility (e.g. vaccinations against waterborne pathogens) may alter future relationships. The climate models agree that summer and fall will become wetter but the magnitude of changes is uncertain. The primary source of precipitation uncertainty comes from the climate models instead of the greenhouse gas emissions rates. The climate models use different statistical approaches to reproduce convective summer season precipitation.

The study also considered the uncertainty related to a range of water treatment installation rates. The scenarios considered observed rates of treatment installation over two periods—over 1991–2010, the relatively slow pathway decreased the population without water treatment by 4.3%; whereas in contrast, the rapid rate of installation over 2011–2016 decreased the population at risk by 82.5%. Across Wisconsin, municipalities distributing groundwater initiated treatment at a leisurely pace (one municipality every 2–5 years) more comparable to the slow pathway. The rapid installation rate is not inevitable and is significantly faster than the background rate. Achieving this adaptation rate requires substantial political will, funding, and technical expertise (Howard et al. 2010); political will may be limited as evidenced by a state law stopping agencies from requiring drinking water treatment (Wisconsin Act 19 2011).

Small drinking-water systems often have aging infrastructure and limited finances, water system expertise, and human resources. Water treatment technologies vary based on their treatment efficacy, raw water quality, and installation and operation costs. Prefabricated package filtration systems which contain coagulation, flocculation, settling and filtration systems may be an attractive small water system option. Package systems require less maintenance than conventional systems and are relatively more affordable. Relatively cheaper filters (e.g. slow sand filters, diatomaceaous earth) effectively reduce virus concentrations but require high quality source water and have moderate installation and operation costs (Wang et al. 2006). Small municipalities may also consider purchasing treated water from a nearby municipality, which can be the most cost effective option.

The costs of installing conventional or prefabricated drinking-water treatment systems, upgrading infrastructure, or drilling new supply wells can be prohibitive for small rural municipalities. Two government programs, the US Environmental Protection Agency’s Drinking Water State Revolving Fund (DWSRF) and US Department of Agriculture’s Water and Waste Disposal Loan and Grant Program, have special provisions for small drinking-water systems (1996 Safe Drinking Water Act Amendments P.L. 104–82, Consolidated Farm and Rural Development Act of 1972 P.L. 92–419). Both programs provide below-market-rate-interest loans to install or upgrade drinking water treatment, storage, or transmission. Since 2009, municipalities with small populations, lower incomes, and high unemployment rates can use the DWSRF to pay down principal infrastructure costs.

The DWSRF provided financial assistance to two municipalities in the study ($4.6 million USD) to install drinking water treatment and drill a new supply well. In Wisconsin, there are more projects that apply for financial assistance than are awarded funds; however, if the other study municipalities chose to apply for assistance, they would become high priority DWSRF projects. The financial assistance priority score is based on acute health effects from microbial organisms, population size, median household income, and county unemployment rate (Wisconsin Department of Natural Resources and The Department of Administration 2015).

A community may consider future population sizes and per capita costs when making long-term infrastructure investments. Demographic projections suggest people will continue to move away from rural municipalities to larger urban enclaves. The Wisconsin population and household projections estimates the study area population may decrease 20% from the year 2010–2040 (Egan-Robertson 2014). Declining population will further lower the total number of GI cases across all three pathways but not alter the incidence rate. The continued development of innovative and cost-effective filtration and/or treatment technologies may benefit these municipalities (US Environmental Protection Agency 2015).

This study is the first to project GI using precipitation instead of temperature. Worldwide, diarrhoeal disease risk increases 3–11% per degree (°C) temperature change (Hales et al. 2014). Ambient temperatures will increase the replication rate of bacterial pathogens (McMichael et al. 2004; Kolstad and Johansson 2011; Ramakrishnan 2011; Semenza et al. 2012). Most climate change studies focus on temperature and GI associations instead of hydrology for at least two reasons. First, there is very high confidence that temperatures will become hotter worldwide, while the direction of precipitation change varies regionally (Stocker et al. 2013). Second, there are fewer international studies of local water and foodborne GI pathways and risk factors (Hales et al. 2014). In this study’s area, there is strong agreement among precipitation models and complementary knowledge of drinking water GI pathways.

Conclusions

This study quantified how additional drinking water treatment and climate change may alter 2046–2065 GI incidence in young children (age < 5 years) compared to 1991–2010. Building on observed relationships, the analysis considered future pathways. Climate change alone was considered in the first pathway, while the other pathways examined both climate change and “slow” and “aggressive” rates of drinking-water treatment installation. More seasonal precipitation will marginally increase the GI incidence rate (mean: ∼1.5%, range: −3.6–4.3%). Water treatment increases the resilience of municipal groundwater systems to climate change. Aggressive rates of treatment installation can drastically reduce GI attributed to precipitation (∼82.0%). Following the aggressive pathway will require political will, expanded funding, and technical assistance. The results highlight the continued need for innovative and cost-effective drinking-water technologies for small systems.

Acknowledgements

This work was partially supported by the Centers for Disease Control and Prevention (grant 1U01EH000428-01) and National PERISHIP Dissertation Fellowship funded by The National Science Foundation, University of Colorado Natural Hazards Center, Swiss Re, and the Public Entity Risk Institute. Stephen Vavrus, Kevin Braun, and Ruben Behnke kindly shared the Wisconsin Initiative on Climate Change Impacts climate projections. We thank Mark A. Borchardt, Joan B. Rose, and anonymous reviewers whose comments significantly improved the article.

Copyright information

© Springer-Verlag Berlin Heidelberg 2017

Authors and Affiliations

  • Christopher K. Uejio
    • 1
  • Megan Christenson
    • 2
  • Colleen Moran
    • 2
  • Mark Gorelick
    • 3
  1. 1.Department of GeographyFlorida State UniversityTallahasseeUSA
  2. 2.Wisconsin Department of Health ServicesMadisonUSA
  3. 3.Children’s Hospital of Wisconsin-Milwaukee CampusMilwaukeeUSA

Personalised recommendations