Systems Approach to Climate, Water, and Diarrhea in Hubli-Dharwad

Oct 26, 2016 - It is informed using water quality and diarrhea data from Hubli-Dharwad, India—a city with an intermittent piped water supply exhibit...
0 downloads 12 Views 748KB Size
Article pubs.acs.org/est

Systems Approach to Climate, Water, and Diarrhea in HubliDharwad, India Jonathan Mellor,*,† Emily Kumpel,‡ Ayse Ercumen,§ and Julie Zimmerman∥ †

Department of Civil and Environmental Engineering, University of Connecticut, Storrs, Connecticut 06269, United States Aquaya Institute, Nairobi, Kenya § Division of Epidemiology, University of California at Berkeley, Berkeley, California 94720, United States ∥ Department of Chemical and Environmental Engineering, Yale University, New Haven, Connecticut 06511, United States ‡

S Supporting Information *

ABSTRACT: Anthropogenic climate change will likely increase diarrhea rates for communities with inadequate water, sanitation, or hygiene facilities including those with intermittent water supplies. Current approaches to study these impacts typically focus on the effect of temperature on all-cause diarrhea while excluding precipitation and diarrhea etiology while not providing actionable adaptation strategies. We develop a partially mechanistic, systems approach to estimate future diarrhea prevalence and design adaptation strategies. The model incorporates downscaled global climate models, water quality data, quantitative microbial risk assessment, and pathogen prevalence in an agent-based modeling framework incorporating precipitation and diarrhea etiology. It is informed using water quality and diarrhea data from Hubli-Dharwad, Indiaa city with an intermittent piped water supply exhibiting seasonal water quality variability vulnerable to climate change. We predict all-cause diarrhea prevalence to increase by 4.9% (Range: 1.5−9.0%) by 2011−2030, 11.9% (Range: 7.1−18.2%) by 2046−2065, and 18.2% (Range: 9.1−26.2%) by 2080−2099. Rainfall is an important modifying factor. Rotavirus prevalence is estimated to decline by 10.5% with Cryptosporidium and E. coli prevalence increasing by 9.9% and 6.3%, respectively, by 2080−2099 in this setting. These results suggest that ceramic water filters would be recommended as a climate adaptation strategy over chlorination. This work highlights the vulnerability of intermittent water supplies to climate change and the urgent need for improvements.



INTRODUCTION Many have postulated that anthropogenic climate change is likely to inordinately burden the 1.8 billion people worldwide without consistent access to microbially safe water1 and the 2.5 billion without improved sanitation2 through increased prevalence of diarrheal diseases.3 This is because temperature, precipitation, infrastructure, and etiology can impact diarrhea rates. Currently, diarrheal diseases lead to an estimated 842 000 deaths annually worldwide.4 Global studies have estimated that climate change will increase diarrhea risk by 22−29% by the end of the century,5 which could undermine recent progress in reducing the global burden of diarrheal diseases. This predicted increase is predicated by the overall positive association between increasing ambient temperature and allcause diarrhea prevalence. Overall, diarrhea incidence has been shown to increase by 3−11% per degree Celsius temperature increase in Fiji,6 Bangladesh,7 and Peru.8,9 A recent meta-analysis confirmed these trends.10 However, the effect size and direction (positive or negative association with increasing temperature) can vary with season,8,11 socioeconomic status7 and presence of improved water, sanitation, and hygiene (WASH) infrastructure.12 © XXXX American Chemical Society

Previous research has shown that precipitation is also a relevant climatic variable affecting diarrheal diseases. A study in Botswana measured a bimodal cyclical pattern in diarrhea incidence with peaks in both wet and dry seasons,11 while other studies in Fiji6 and Bangladesh7 measured increases in diarrhea incidence both above and below threshold rainfall amounts. A study in rural Ecuador found that heavy rainfall following a wet period decreased diarrhea risk while heavy rainfall following a dry period increased risk.13 Such studies are consistent with a recent systematic review that found increases in diarrheal disease rates following heavy rainfall and flooding events in 71% and 76%, respectively, of reviewed studies.14 Adding to the complexity, diarrhea is a symptom of infections that can be caused by a number of different enteric pathogens.15 The relative importance of these pathogens varies regionally16 and seasonally,17 which is likely to have important implications for disease burdens under future climate scenarios. For example, Received: Revised: Accepted: Published: A

May 2, 2016 October 6, 2016 October 26, 2016 October 26, 2016 DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology rotavirus tends to be positively associated with cooler and drier weather in tropical climates18 while cryptosporidiosis is more prevalent in warmer and wetter weather.19 Temperature, precipitation and etiology can impact diarrhea prevalence. However, prior studies that have combined climate change models with the epidemiological evidence to estimate the climate change impacts on diarrhea rates globally have only incorporated the effects of temperature on all-cause diarrhea.5,20,21 These studies generalize regional epidemiological results globally, do not incorporate precipitation effects, and do not consider diarrhea etiology. These epidemiological results also do not rely on a mechanistic framework and therefore have limited ability to inform adaptation strategies to improve resiliency to climate events. An alternate approach is to treat the climate, water, pathogen, and infrastructure components as a complex system and to integrate the relevant human, engineered, and natural components that may influence system behavior.22 Agent-based models (ABMs) are useful tools for incorporating many components to study complex systems23 and have been used to study the WASH system in South Africa24 and ceramic water filters (CWFs) to prevent diarrhea.25 At least 300 million people around the world receive piped water that is supplied only intermittently, for limited durations, throughout Asia, Africa, Latin America, and eastern Europe.26 These systems provide water intermittently and remain pressurized only during supply cycles. When the water is off, the pipes return to atmospheric pressure. Although the exact mechanisms are uncertain, contamination can enter the water pipes during these low pressure periods.26,27 Intermittent piped water supplies have been shown to exhibit substantial seasonal microbiological water quality variability28 that could indicate their vulnerability to the changes in temperature and precipitation caused by climate change. Modeling how such supplies are impacted by climate change can help to develop adaptation strategies to mitigate any negative impacts of climate on biological drinking water quality and resulting diarrhea prevalence. An understanding of how such supplies will be impacted by climate change is urgent given the rapid population growth expected in low and middle income countries over the next several decades29 and the vulnerability of those regions to climate change.30 One possible interim strategy is household point-of-use (POU) water treatment; such technologies could help communities adapt to climate change by improving water quality until they receive safe and reliable drinking water supplies. There are many different POU technologies available to communities including flocculent/disinfectant sachets, ceramic siphons, and sodium dichloroisocyanurate tablets31 as well as biosand systems, Lifestraw filters and solar water disinfection.32 Two common POU technologies that have shown promise at improving household microbial water quality33 and at reducing diarrhea prevalence across many regions34,35 are ceramic water filters (CWFs) and the Safe Water System (SWS). The CWF is a porous ceramic filter that uses both physical filtration and coated silver nanoparticles to inactivate pathogens. Users of the SWS dose sodium hypochlorite to disinfect water in specially designed sanitary water storage containers. CWFs or SWSs could help communities adapt to climate change to reduce both current and future risk of diarrheal diseases until they receive safe and reliable drinking water supplies. However, CWFs are likely not effective at removing viruses,36 while the levels of chlorine used in the SWS are ineffective at inactivating protozoa.37

In this study, we employ an ABM approach to (1) develop a partially mechanistic systems-based method to estimate the risk of diarrheal diseases due to consumption of water from intermittent water supplies under future climate scenarios (2) study the relative importance of temperature, rainfall, and diarrhea etiology on diarrhea prevalence, and (3) investigate the ability of POU household water treatment devices to mitigate increased diarrhea risk. These three goals were achieved by combining global climate model (GCM) results with empirical weather, water quality, and diarrhea prevalence data through a quantitative microbial risk assessment (QMRA) model using an ABM framework. The empirical evidence was based on previously collected longitudinal microbial water quality,28 child diarrhea prevalence,38 and weather data collected in Hubli-Dharwad, India.



METHODS Study Setting. This study is based on data collected during a prior study in Hubli-Dharwad located in Karnataka, India, that measured water quality and diarrhea prevalence in children concurrently in a large cohort of households over the course of approximately one year.28,38 The twin cities of Hubli-Dharwad have a combined population of approximately 900 000 residents. The climate in the study area is characterized by stark seasonal variation. The hot and dry season lasts from February to June followed by a monsoon season with cooler temperatures and abundant rainfall. October to February has very little rainfall and more moderate temperatures. The municipal water supply in Hubli-Dharwad relies on reservoir and lake water that is treated at two treatment plants using standard water treatment methods, which then supply water to the municipal piped water distribution system.28 Despite this initial treatment, the intermittency of the supply for the majority of wards (administrative boundaries) leaves pipes unpressurized for long periods of time and results in overall low water pressure during supply hours, which can adversely affect microbial water quality. During the period of data collection in the prior study, water was supplied once every 1 to 8 days, with a median frequency of 5 days in most parts of the city. A subset of the city’s users were upgraded to receive a continuous water supply that operates 24 h a day, 7 days a week. The purpose of the prior study was to compare drinking water quality and diarrhea prevalence between the continuous and intermittent water delivery systems.28,38 Our present study focuses on the intermittently operated piped water system which exhibited substantial microbial contamination as well as seasonal variability in water quality. Indicator bacteria concentrations were significantly higher in the rainy season (ANOVA, p < 0.01), which also coincided with warmer ambient temperatures. Indicator bacteria concentrations were likewise higher within 24 h after a rainfall28 possibly indicating that runoff may be washing human/animal feces into the water supply. Given these weather-dependent trends, it is likely that the piped water quality in this setting will be affected by future climate change as the region warms and precipitation patterns change.3 Indicator Bacteria and Diarrhea Prevalence Data Collection. During the prior study, diarrheal prevalence data were collected longitudinally from 1,951 households with an intermittent water supply in Hubli-Dharwad. Caregivers reported the seven-day prevalence of symptoms (i.e., occurrence of diarrhea in the 7 days preceding the interview) for children under the age of 5 at four quarterly visits between November B

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Figure 1. Overall modeling framework and data sources.

2010 and February 2012.38 These prevalence data were used to validate the model presented here. A total of 875 water samples were collected between November 2010 and November 2011 from households with an intermittent water supply in Hubli-Dharwad as part of the prior studies.28 The sampling technique is described in that study.28 The 100 mL samples were taken from both taps and household storage containers and were analyzed for Escherichia coli (E. coli) and total coliform bacteria using Colilert Quanti-tray 2000 (IDEXX Laboratories Inc., Westbrook, ME, U.S.A.) (methods in Kumpel and Nelson 2013). Total coliform bacteria were used in our model because of their relatively high correlation with enteric pathogens.39 However, E. coli has been shown to be a better indicator of diarrhea risk.40 Therefore, the model was developed to use either indicator. Multiple imputation was used to fill in incomplete water quality data as is described in the Supporting Information, SI. Modeling Approach. The ABM had 1000 representative households with 2 children in each household, where the quality of the water consumed by the children was determined by the ambient weather conditions. Their propensity to have diarrhea was then modeled as a function of the quality of consumed water. The model was run at a daily time step for 1825 days (5 years). These population numbers and run time were chosen to be computationally manageable, while providing statistically significant results. All simulations were conducted using Netlogo, a modeling package for ABMs.41 The overall modeling framework is shown in Figure 1. Stochastic Weather Simulator. This modeling framework requires that daily weather be simulated for both baseline and future climates. To study how global climate change impacts daily weather, the results of global climate models (GCMs) need to be downscaled. This can be done using LARS-WG, which is a stochastic weather simulator that produces synthetic daily temperature and rainfall data for baseline and future climates

using historic weather data for a specific geographical location.42 LARS-WG incorporates 15 GCM simulations for three Special Report on Emissions Scenarios (SRES)43 from the Intergovernmental Panel on Climate Change (IPCC) Assessment Report 4 to stochastically generate daily weather for each predicted future climate. These scenarios are the A2 (Separated World), A1B (Rich World), and B1 (Sustainable World).43 LARS-WG can output daily weather for the three SRES scenarios for three time periods: 2011−2030, 2046−2065, and 2080− 2099. The generator was first trained with the available 43 years of observed daily minimum and maximum temperature and rainfall data from the Indian Metrological Department for Gadag, India, a city approximately 60 km east of Hubli-Dharwad where data were available. Once trained, daily minimum, maximum temperature and precipitation values were then simulated for 1000 years for each modeled scenario. The choice of 1000 years was arbitrarily made to ensure a diverse set of yearly climates. One of the 1000 years worth of synthetic data was randomly selected for each year the model was run. Those weather values constituted each simulated day’s weather from January 1 to December 31 when the next year’s worth of synthetic data were randomly selected from the 1000 years of simulated data. Water Quality Matching Routine. The goal of this study is to understand how weather and climate can impact microbiological water quality and resulting diarrhea prevalence. It therefore necessary to correlate ambient weather conditions to indicator bacteria concentrations. On each day in the model, the child “agents” consume water whose quality is determined as a function of the simulated day’s weather. To do this, a correlational routine matches each model day’s simulated weather with field water quality measurements taken on days with the most similar weather. The routine first searches for water quality measurements taken within a stochastically (continuous linear distribution) varied 2−22 Julian day range when the actual rainfall totals were within the same range as the C

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

The N50, α, and k parameters are empirically derived constants based on feeding experiments where volunteers are given various doses of pathogens and monitored for signs of infection. To better represent the variety of immune system responses in a heterogeneous population, a bootstrap method was used45,46 to generate 1000 different possible parameter combinations for each of the three pathogens. Data from Ward (1986)47 was used to calculate the rotavirus parameters, Messner (2001) for Cryptosporidium48 and DuPont (1971) for E. coli.49 Each child on each day was then randomly assigned one of these parameter sets for each pathogen and their probability of becoming infected was calculated. The dose (d) was calculated by multiplying the household’s daily water quality by a stochastically (continuous linear distribution) varied ingestion amount of between 1 and 2.4 L per child per day.50 If a child is infected, then the probability of them becoming symptomatic (and hence reporting symptoms during a survey) is 90.6% for rotavirus, 80.4% for E. coli and 91.3% for Cryptosporidium.44 The probabilities for all three pathogens were combined for each child on each day to determine the probability of all-cause diarrhea. Results are presented both in terms of both the pathogen-specific and allcause diarrhea prevalence. The model assumes that children get sick immediately after being exposed to contaminated water. Although some have suggested there is a lag period between weather events and diarrhea impacts,13,19 the data used for this study did not support the need for such lagging. This modeling approach focused on the risk associated with water consumption. While it is well-known that there are many other pathways for pathogens to be ingested,51,52 the use of QMRA to quantify the risk of exposure to drinking water has been advocated for previously22,53 and related prior studies using water pathways have proven useful despite this limitation.24,25,54−57 Modified Pathogen-Specific Infection Risk Due to Temperature and Precipitation. Across a broad range of climates, E. coli, rotavirus, and Cryptosporidium incidence rates are dependent on temperature and/or precipitation, which then need to be incorporated into the model.18,19,58 This was accomplished by modifying the probability of infection for these three pathogens using regional reports of pathogen temperature and precipitation dependence.58−60 The precipitation dependence of rotavirus was calculated based on empirical evidence from South Asia.59 Those data were incorporated using eq 4, which modifies the probability of infection (eq 2). These values are generally close to a value of 1 under most precipitation conditions. R is the total rainfall (in mm) over the preceding 30 days and 53.6 mm is the average monthly rainfall for the Hubli-Dharwad area.

simulated day’s rainfall total. Daily rainfall ranges were 0 mm, 0− 1 mm, 1−10 mm, and >10 mm. If no water quality measurements were taken within the 2−22 Julian day range that had rainfall totals that were within same ranges, then the date range was expanded by 7-day increments until at least one suitable day was found within the expanded range. The subset of days with appropriate rainfall totals were searched to find the day with the minimum temperature differential as defined by eq 1. ∥tfield‐min − tsimulated‐min| + |tfield‐max − tsimulated‐max∥

(1)

Each of the 1000 households was then randomly assigned to have one of the 1 to 10 water quality measurements made on that day. Representative Pathogens. Diarrhea is a symptom of an infection by an enteric pathogen.16 Cryptosporidium parvum, pathogenic E. coli, and rotavirus were identified as being the most common enteric pathogens in India44 and were used in this study. They represent the three main classes of microorganisms that cause diarrhea: protozoa, bacteria, and viruses.16 Together these three pathogens have been shown to cause 86.7% of diarrhea among under five year old hospitalized patients in the south Indian city of Vellore.44 Indicator to Pathogen Ratios. After the indicator bacteria levels are determined by the ambient weather for each household, it is necessary to convert those indicator bacteria concentrations to pathogen concentrations. The ratio between E. coli or total coliform indicator bacteria and enteric pathogens depends upon many variables, including the pathogen of interest, water type, source of pathogen, pathogen detection method, and sample size for pathogen detection.39 To incorporate this uncertainty, the indicator to pathogen ratio for each pathogen of interest was stochastically (continuous linear distribution) varied between upper and lower limits during model runs. These lower and upper limits were set using a calibration procedure. Each of the indicator to pathogen ratio range upper and lower limits were randomly varied until the optimized limits were found. These optimized lower and upper limits were determined using the community all-cause diarrhea rates from the prior study38 as well as regional reports of diarrhea etiology.44 Dose−response relationships were determined using the Quantitative Microbial Risk Assessment (QMRA) routine described below. The calibrated lower and upper limits of the pathogen ratios along with the baseline diarrhea etiology are given in Table S1. An identical calibration routine was conducted using both E. coli and total coliform as the indicator bacteria. The results for total coliform simulations are summarized in the Results section and figures. Analogous results for E. coli are given in the SI. Quantitative Microbial Risk Assessment. QMRA is a wellestablished technique used to calculate the probability that a person will be infected with a diarrhea-causing pathogen if they are exposed.45 In QMRA, the probability of infection is calculated using the dose response relationships. For this study, the eqs 2 and 3 for rotavirus/E. coli and Cryptosporidium, respectively, were used. Pinfection is the probability of infection, d is the ingested dose, N50 is the median infective dose, α is the dimensionless infectivity constant, and k is also an infectivity constant.45 Pinfection

⎡ ⎤−α d 1/ α (2 − 1)⎥ = 1 − ⎢1 + N50 ⎣ ⎦

Pinfection = 1 − e−kd

−0.0003 × R + (1 + 0.003 × 53.6)

(4)

Given the high nonlinearity seen in the temperature dependence of rotavirus, all of the independently reported measurements of the rotavirus Z-score as a function of temperature reported by Jagai et al. (2012)59 were used as a modifying factor. The full plot of prevalence modification values for all temperatures can be found in Figure S1. For each day in the model, one of the prevalence modification values corresponding to the simulated day’s temperature was randomly selected and used to modify the infection probability (eq 2). A similar method was used for the temperature dependence of Cryptosporidium. Previously published monthly measurements of diarrhea positive stool samples from Delhi and Vellore60 as a function of temperature were converted into a prevalence

(2) (3) D

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

A second method varied LARS-WG baseline daily temperatures by ±5 °C to determine the temperature dependence of diarrhea prevalence and compare it to previously published epidemiological studies. A similar method was used to multiply LARS-WG baseline daily precipitation amounts by a factor of 0 to 2 for days when it rained. To test the importance of including rainfall to predict future diarrhea prevalence, the analysis was repeated with the rainfall functionality bypassed. In this case, the correlational water quality routine only considered temperature when searching for the closest possible weather days. The model was run using both E. coli and total coliform as the indicator bacteria. Results for total coliform as the indicator are given in the Results section, because those results produced a better fit with the data (Figure S6). Results for E. coli are given in the SI. Results. The LARS-WG daily mean maximum temperature output for all three SRES emission scenarios are shown in Figure S2. Those plots indicate a general warming trend in HubliDharwad across all three emissions scenarios (B1, A1B, and A2) with warmer temperatures predicted to occur as the 21st century progresses due to anthropogenic greenhouse gas emissions. As expected, warming was greatest in the A2 (Separated World) scenario followed by the A1B (Rich World) and B1 (Sustainable World) scenarios. Rainfall amounts exhibited inconsistent trends between scenarios and time periods in both average daily magnitude and direction (Figure S3). The bootstrapped dose−response parameters (α and N50) for E. coli and rotavirus are given in Figures S4 and S5. The median, minimum, and maximum values for all three pathogens are given in Table S2. The modeled results for total coliform as the indicator bacteria were consistent with the field results on 85% of weeks (Student’s t test, p > 0.05) based on 1000 model runs (Figure S6). The model demonstrated that all-cause diarrhea prevalence was dependent on temperature and rainfall. The all-cause diarrhea dependence as a function of ambient temperature departure from baseline using total coliform bacteria as the indicator is shown in Figure 2. This dependence exhibits a near linear trend (R2 = 0.995 p < 0.001) by about 6.4% per degree Celsius. In contrast, diarrhea prevalence decreased with increased rainfall, but the effect is less pronounced than the temperature dependence (Figure 3).

modifying factor for the probability of infection (Tables S4 and S5). On each model day, one of these two data sets was randomly selected. Then, values corresponding to the simulated day’s ambient temperature were used to modify the infection probability (eq 3). Since Cryptosporidium was not found to be highly correlated with precipitation in the Ajjampur (2010) study conducted in India,60 Cryptosporidium infection rates were not modified by precipitation. While it is true that Cryptosporidium can be impacted by precipitation,19 the Ajjampur study was the most geographically relevant study for this application. A recent meta-analysis of the climatic drivers of E. coli indicates a positive association between diarrheagenic E. coli incidence and monthly mean ambient temperature.58 These results were incorporated into the present model by modifying the infection probability (eq 2) according to the incident rate ratios (IRRs) found in Table S6. One of the 18 IRRs was randomly selected on each model day. Temperature deviations were calculated by subtracting the modeled monthly mean temperatures from the historic monthly mean temperatures. POU Technologies. CWFs and SWSs were chosen to simulate potential interim adaptation strategies. These POU technologies reduce indicator organism concentrations for domestic water consumers, however they have been shown to decline in microbial removal efficacy over time in many areas.33 This could explain their observed decline in reducing diarrhea.35 Declining compliance and inadequate maintenance can also lead to reduced effectiveness.25 Therefore, in order to realistically depict the effectiveness of CWFs and SWSs in the field, microbial effectiveness data from Mellor et al. (2014)33 were used, which are consistent with prior studies.25,61 These two technologies were chosen due to the availability of the previously collected data, and the fact that they are both widely available and effective technologies.62 In the Mellor et al. (2014) study, CWF log removal efficiencies declined from an average of 2.20 to 1.18 over the course of the year-long study likely due to poor maintenance and gradually deteriorating filters. Similarly, SWS log removal efficiencies went from an average of 2.37 to 1.60 over the same time period. This was incorporated into the present ABM using a linear approximation for the indicator bacteria removal efficiency of CWFs and SWSs: LRV = m × tSWS/CWF‐life + LRV0

(5)

LRV is the log reduction value of the POU technology, m is the linear slope, tSWS/CWF‑life is the length of time (in days) since the SWS or CWF was introduced to a household and LRV0 is a stochastically (normal distribution) varied initial LRV. The values of these parameters are given in Table S3. The model assumes that each household keeps their filter for one year before being replaced and, when replaced, the filter performs like new. User compliance was assumed to be 75% in the model based on prior self-reports from the same prior study.33 By using this empirical data, the model is able simulate realistic POU-related behavior. Validation. Two methods were used to validate the ABM. The first validation method used measured water quality data for each field sampling day during the year of data collection, which was used to calculate diarrheal disease prevalence as described in the previous sections (i.e., indicator to pathogen ratios and QMRA). For this validation routine, the modeled weekly diarrhea rates were then compared to the empirically measured weekly diarrhea rates using Student’s t tests.

Figure 2. Changes in diarrhea prevalence as a function of temperature. E

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

increase (Range: 9.6−23.4%) by the 2046−2065 period, and to a 24.0% increase from baseline (Range: 14.3−34.1%) by 2080− 2099. This is 2.6% (2011−2030), 4.0% (2046−2065), and 5.8% (2080−2099) increase over the full model that includes rainfall as a predictor of water quality. When the two POU adaptation strategies are compared (Figure 5), the results in indicate that both CWFs and SWSs are

Figure 3. Changes in diarrhea prevalence as a function of rainfall. Rainfall multiplier multiples the daily simulated rainfall values by a given amount.

Baseline (measured) weekly diarrhea prevalence was approximately 8.4% of children during the week preceding the interview. The full results of the model for all three SRES scenarios using total coliform as the indicator predict that allcause diarrhea prevalence would increase with time, and there are significant mean differences between the three scenarios [F(2, 1007) = 19.03, p < 0.001]. The B1, A1B, and A2 scenarios, respectively, had the following 5-year child mean diarrhea cases: 24.5, 25.0, and 25.3 (Figure 4 (a)). The average diarrhea prevalence is predicted to increase by 4.9% (Range: 1.5−9.0%) during the 2011−2030 time period rising to a 11.9% increase (Range: 7.1−18.2%) by the 2046−2065 period, and to a 18.2% increase from baseline (Range: 9.1−26.2%) by 2080−2099 (Figure 4(a)). Concurrent with this increase in all-cause diarrhea, diarrhea caused by rotavirus is predicted to decreases by 5.5% by 2046− 2065 and by 10.5% by 2080−2099. Conversely, diarrhea cases attributable to Cryptosporidium are predicted to increase by 5.0% by 2046−2065 and by 9.9% by 2080−2099. Likewise, diarrhea cases attributable to E. coli are expected to increase by 2.8% by 2046−2065 and by 6.3% by 2080−2099. These pathogenspecific changes are relative to a 2011−2030 baseline and are statistically significant. When rainfall is excluded from the water quality correlation routine the model predicts that diarrhea prevalence would be greater than is predicted by the full model (Figure 4(b)). In this case, diarrhea prevalence is predicted to increase by 7.5% (Range: 5.2−12.3%) during the 2011−2030 time period rising to a 15.9%

Figure 5. Diarrhea changes for all climate scenarios and time periods for users of CWFs and SWSs. Data indicate that both CWF and SWS can be effective diarrhea reducing tools, but that CWF are increasingly better toward the end of the century. Darker colored triangles represent SWS and the lighter colored circles are CWF.

predicted to be effective at reducing diarrhea prevalence well below baseline levels in both the near and long-term. The model results also indicate that CWFs are predicted to maintain their effectiveness as an all-cause diarrhea reduction tool in the future as Cryptosporidiosis and E. coli become more prevalent under warmer conditions. The efficacy of CWF in reducing diarrhea started at 62.3% (Range: 61.5−63.2%) during the 2011−2030 and declined marginally to 60.4% (Range: 58.7−61.9%) by the 2046−2065 period, and to 58.8% (Range: 57.0−61.0%) by 2080−2099. SWS efficacy declined, as diarrhea reductions from baseline went from a high of 58.3% (Range: 56.6−59.6%) during the 2011−2030 time period, declining to 54.8% (Range: 51.9−

Figure 4. (a) Results from the full model. All-cause diarrhea prevalence increases with time with increasing differences between scenarios. Each data point indicates a separate GCM model run that was incorporated into LARS-WG. (b) The full model results excluding rainfall. These results indicate that rainfall is an important modifier of diarrhea prevalence in the model. F

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

conclusion of this work. The model predicts that diarrhea prevalence attributable to Cryptosporidium and E. coli may increase, while rotavirus prevalence will likely decrease. This trend is consistent with the epidemiological evidence from the Carlton et al. (2016)10 meta-analysis which estimated an incident rate ratio for bacterial diarrhea of 1.07 (95% CI 1.04−1.10) and for viral diarrhea 0.96 (95% CI 0.82−1.11). Moreover, while the temperature and precipitation dependence of different pathogens was well-known previously for a variety of pathogens,10,18,19,63 to the best of our knowledge, this is the first use of a partially mechanistic modeling framework to incorporate this information to predict diarrhea prevalence under future climates in low-income settings. Given the pathogen heterogeneity between different geographical locations seen in recent global studies of enteric infections16,64 and the seasonal nature of many pathogens, this study suggests that a knowledge of regional diarrhea etiology may be important for developing regionally effective adaptation strategies and should be included in future studies. In our scenario, excluding rainfall as a salient environmental predictor of water quality increased the predicted diarrhea risk under future climates. This can be seen by comparing Figure 4(a) with Figure 4(b). In Figure 4(b) we see generally higher diarrhea rates especially for later time periods. While researchers have previously reported on the relevance of precipitation as a modifier of microbial water quality53,65−68 and in both allcause 10,11,13,14 and pathogen-specific diarrhea prevalence,14,18,19,63 its importance under future climate scenarios was unclear and rainfall was frequently excluded from studies of climate change impacts on diarrhea rates. Specifically, water quality studies have reported that factors including the timing of rainfall (i.e., recent rainfall, wet versus dry season or early versus late rainy season) and seasonal total rainfall can be positively or negatively correlated with water quality.65−68 Our results confirm that rainfall may be a critical modifying risk factor for all-cause diarrhea prevalence and should be considered in future studies. The decrease of diarrhea prevalence with increasing precipitation seen in Figure 3 could be due to the fact that the monsoonal climate is associated with cooler temperatures in this region of India (Figures S2 and S3). Since temperature effects are larger than precipitation effects, the cooler weather is apparently decreasing diarrhea prevalence. Considerations of diarrhea etiology are important for determining options for improving water quality, as illustrated by the POU results. These results are relevant for policy makers and implementing nongovernmental organizations. Both CWF and SWS technologies have been advocated for as a means of diarrhea reduction. Our results indicate that the ability to remove Cryptosporidium will be increasingly important as the earth warms, making CWFs more preferable compared to SWS technologies (Figure 5). The importance of Cryptosporidium was corroborated by recent results from the Global Enteric Multicenter Study.16 Moreover, the increasing availability of effective rotavirus vaccines69 may further decrease the fraction of diarrhea attributable to rotavirus. Although CWFs are one widespread technology, other filtration-based technologies including biosand filters are also likely to be effective against cryptosporidium.70 However, although POU devices can be an important interim approach, upgrading piped systems to consistently provide safe water is the preferable option given poor compliance and suboptimal microbiological efficacy that can impact the effectiveness of POU technologies.25,71

57.3%) by the 2046−2065 period, and down to 51.4% (Range: 47.3−56.0%) by 2080−2099. Identical analyses were conducted using E. coli as the indicator bacteria and are summarized in the SI. Results were qualitatively similar albeit with a larger effect size for diarrhea prevalence during later time periods (Figure S9). A notable exception was the all-cause diarrhea prevalence rates as a function of temperature departures below normal, which showed little trend (Figure S8). Discussion. The results indicate that (1) intermittent water supplies are vulnerable to climatic influences and are thus likely to contribute to an increase in diarrhea prevalence as the climate changes (Figure 4(a)), (2) rainfall (Figure 3) and diarrhea etiology are predicted to be highly relevant modifying factors, (3) diarrhea prevalence attributable to Cryptosporidium and E. coli are predicted to increase, while rotavirus prevalence will likely decrease, (4) POU technologies like CWFs and SWSs could provide interim protection against both current and future risk, but CWFs are preferred (Figure 5). Ultimately, on the basis of this study and the prior study,28 intermittent supplies should be upgraded to continuous operation to reduce this waterborne disease risk. Given the rapid urbanization of low and middle-income countries expected over the coming decades, coupled with cities’ reliance on intermittent water supplies, there is an urgent need to study the vulnerability of those supplies to climatic influences. These results indicate that such supplies may lead to a 18.2% increase (Range: 9.1−26.2%) in diarrhea prevalence by the end of the century (Figure 4(a)). This increase is at the lower end of previous global estimates, which predict a 8−11% increase for the period 2010−2039, 15−20% by 2040−2069 and 22−29% by 2070−2099.5 However, most prior studies exclude rainfall. They also generally do not explicitly consider diarrhea etiology. Although, being epidemiological in nature, they do take into account nonwater pathways, which are not explicitly modeled using our approach. Our model predicts diarrhea prevalence will increase by approximately 6.4% per degree Celsius above current average temperatures (Figure 2). This increase is consistent with previous regional epidemiological studies which measured a 7% (95% CI 3−10%) per degree Celsius temperature increase across a broad range of climates.10 This consistency is evidence that this partially mechanistic modeling framework could have predictive power. It is also notable that our model predicts increasingly large differences between SRES scenarios as the century progresses. From Figure 4(a), it is evident that all-cause diarrhea prevalence rates are relatively insensitive to SRES emissions scenarios in the near term and diarrhea rates will increase even if nations take more aggressive action to limit greenhouse gas emissions. However, by the end of the century, there are increased differences in diarrhea rates between the SRES scenarios as the divergent emission scenarios exhibit larger predicted temperature differences. These results also highlight the large range of possible climate change outcomes and the large differences between downscaled GCMs particularly in regards to future precipitation estimates. When using E. coli as an indicator bacteria, we see larger differences between SRES scenarios (Figure S9). This is likely due to the higher sensitivity of all-cause diarrhea rates to higher temperatures when using E. coli as the indicator bacteria (Figure S8). The importance of incorporating diarrhea etiology when predicting all-cause diarrhea prevalence in the future is another G

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

pathogen-specific infection risk due to temperature and precipitation has the potential to duplicate some of the impact that weather has on water quality. This might amplify the impact of weather events on diarrhea rates. However, the model does adequately replicate the epidemiological evidence so this effect is likely small. Finally, the use of only three pathogens to represent all-cause diarrhea may limit the applicability of these results and makes our results somewhat conservative. However, these three pathogens represent the three major pathogen taxa. Different pathogens within each of these taxa tend to be similarly influenced by temperature.10 Therefore, although this simplification may impact the magnitude of climate change’s impact on diarrhea, it is unlikely to modify the overall relationships identified by this analysis. Our modeling framework can be easily modified to incorporate other pathogens. Future field work should focus on studying the environmental, animal, and human transport of regionally important pathogens through communities to understand better how people are exposed during different climatological conditions. These studies should include both hydrologic transport, as well as behavioral adaptation (e.g., hand-washing, water treatment compliance, pathogens associated with surfaces, water source choice, sanitation usage, and so forth). Studies should also be conducted in rural regions and different climates where water scarcity is another important variable. Such empirical measurements can inform pathogen exposure models which can be used to build similar coupled modeling approaches that include human, engineered and environmental components. When empirical evidence is unavailable from a specific site, this modeling approach can also incorporate data from geographically and socioeconomically similar areas as we have done. Despite these caveats, this modeling framework is a first step toward incorporating and understanding the importance of many of the salient risk factors associated with future diarrhea prevalence. These results suggest that climate change poses a risk to vulnerable populations and appropriate strategies ought to be undertaken to mitigate this risk. This work can therefore have implications for the 1.8 billion globally who did not have consistent access to a safe water supply76 and, more specifically, the 300 million people who rely on similar intermittent water supplies.26

Each step of the model outlined in Figure 1 has an associated uncertainty that was accounted for in our coupled model. The downscaling of the climate models is the first source of uncertainty especially for precipitation.42 This variability of temperature and precipitation is captured through the different GCM runs shown in Figures S2 and S3, which is why we ran the ABM for each of the GCMs and IPCC scenarios and graphed the results of each model run (e.g., Figures 4 and 5). The next significant area of uncertainty has to do with the POU compliance and effectiveness. It is well known that compliance rates vary and that even those using POU technologies frequently consume untreated water.72 We have accounted for this uncertainty using empirical evidence of POU technology effectiveness and compliance data.33 This uncertainty is represented through the error bars in Figure 5 and range of numerical results. The remaining areas of uncertainty are represented by the ranges of the numerical results and the error bars in the associated figures (e.g., Figures 2−5). Indicator bacteria to pathogen ratios and QMRA are both areas of uncertainty in the model. We accounted for the indicator bacteria to pathogen ratios by stochastically varying our three fitting parameters as shown in Table S1. QMRA uncertainty was incorporated by the bootstrapping technique we employed (e.g., Table S2 and Figures S4 and S5). The relationship between weather and pathogen-specific diarrhea incidence is an active area of research and can be modified by a number of environmental, social and economic influences.22 We accounted for this heterogeneity by integrating the published raw data into the model negating the need for making assumptions about those relationships (e.g., Tables S4−S6 and Figure S1). Despite the promise of this modeling framework, there are a number of important limitations. These results are likely only applicable to intermittent piped water supplies in urban or periurban settings in low-income tropical countries. The temperature and precipitation dependence of water quality would likely be different in a rural location or for those relying on water from point sources such as boreholes, springs, or rainwater. Furthermore, water accessibility is also an important determinant of health. If people are forced to travel further to collect adequate water supplies in the face of increased climate variability, then their health is likely to suffer.73 This scenario is more likely in rural locations where water sources can be located several kilometers away even during rainy periods.74 This modeling framework also does not take into account economic growth, which is likely to reduce enteric infections through improved access to WASH infrastructure particularly in regions like India that are experiencing rapid economic advancement. Our results therefore represent a worst-case scenario. These results are also limited to investigating the risk of ingesting pathogens through water, while it is well-known that enteric pathogens are transported via a variety of environmental pathways including hands, foods, and flies among others.51 However, this is a common assumption for related modeling frameworks.24,25,54−57 We also assumed that the CWF technology would be replaced each year. Less frequent replacement would likely have decreased efficacy. Next, the causal relationship between the indicator bacteria used in this study and diarrhea infection is an active area of research.40,75 To investigate this, the model was run for both E. coli and total coliform bacteria as indicators of microbial risk. Although there were some quantitative differences in the results between the two approaches, the qualitative results were consistent and do not modify the overall conclusions or recommendations of this study. The modification of our



ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.6b02092. Supporting figures, tables, and the results using E. coli (PDF)



AUTHOR INFORMATION

Corresponding Author

*Phone: +1 (860) 486-0548. Fax: +1 (860) 486-2298. E-mail: [email protected] (J.M.). Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We would like extend a special thank you to our collaborators at the Center for Multidisciplinary Research (CMDR) in HubliDharwad, India. In particular Dr. Nayanatara Nayak, Dr. Narayan Bilava, and Mr. Madhu Reddy were all integral in the data H

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

diarrhea in Dhaka, Bangladesh. Journal of Clinical Microbiology 1999, 37, 3458−3464. (18) Levy, K.; Hubbard, A. E.; Eisenberg, J. N. Seasonality of rotavirus disease in the tropics: a systematic review and meta-analysis. International Journal of Epidemiology 2009, 38, 1487−1496. (19) Jagai, J. S.; Castronovo, D. A.; Monchak, J.; Naumova, E. N. Seasonality of cryptosporidiosis: A meta-analysis approach. Environ. Res. 2009, 109, 465−478. (20) Campbell-Lendrum, D.; Woodruff, R.; Prüss-Ü stün, A.; Corvalán, C. Quantifying the health impact at national and local levels. WHO 2007, 85, 235. (21) Hodges, M.; Belle, J. H.; Carlton, E. J.; Liang, S.; Li, H.; Luo, W.; Freeman, M. C.; Liu, Y.; Gao, Y.; Hess, J. J.; et al. Delays in reducing waterborne and water-related infectious diseases in China under climate change. Nat. Clim. Change 2014, 4, 1109−1115. (22) Mellor, J. E.; Levy, K.; Zimmerman, J.; Elliott, M.; Bartram, J.; Carlton, E.; Clasen, T.; Dillingham, R.; Eisenberg, J.; Guerrant, R.; et al. Planning for climate change: The need for mechanistic systems-based approaches to study climate change impacts on diarrheal diseases. Sci. Total Environ. 2016, 548, 82−90. (23) An, G. Dynamic knowledge representation using agent-based modeling: ontology instantiation and verification of conceptual models. Methods Mol. Biol. 2009, 500, 445−468. (24) Mellor, J. E.; Smith, J. A.; Learmonth, G. P.; Netshandama, V. O.; Dillingham, R. A. Modeling the complexities of water, hygiene, and health in Limpopo Province, South Africa. Environ. Sci. Technol. 2012, 46, 13512−13520. (25) Mellor, J.; Abebe, L.; Ehdaie, B.; Dillingham, R.; Smith, J. Modeling the sustainability of a ceramic water filter intervention. Water Res. 2014, 49, 286−299. (26) Kumpel, E.; Nelson, K. L. Intermittent water supply: prevalence, practice, and microbial water quality. Environ. Sci. Technol. 2016, 50, 542−553. (27) Kumpel, E.; Nelson, K. L. Mechanisms affecting water quality in an intermittent piped water supply. Environ. Sci. Technol. 2014, 48, 2766−2775. (28) Kumpel, E.; Nelson, K. L. Comparing microbial water quality in an intermittent and continuous piped water supply. Water Res. 2013, 47, 5176−5188. (29) United Nations Department of Economic and Social Affairs. World Population to 2300; 2004; Vol. 236. (30) Christenson, E.; Elliott, M.; Banerjee, O.; Hamrick, L.; Bartram, J. Climate-Related Hazards: A Method for Global Assessment of Urban and Rural Population Exposure to Cyclones, Droughts, and Floods. Int. J. Environ. Res. Public Health 2014, 11, 2169−2192. (31) Mohamed, H.; Clasen, T.; Njee, R. M.; Malebo, H. M.; Mbuligwe, S.; Brown, J. Microbiological effectiveness of household water treatment technologies under field use conditions in rural Tanzania. Trop. Med. Int. Health 2016, 21, 33−40. (32) Clasen, T.; Alexander, K.; Sinclair, D.; Boisson, S.; Peletz, R.; Chang, H.; Majorin, F.; Cairncross, S. Interventions to improve water quality for preventing diarrhoea. Cochrane Database of Systematic Reviews 2015, DOI: 10.1002/14651858.CD004794.pub3. (33) Mellor, J. E.; Kallman, E.; Oyanedel-Craver, V.; Smith, J. A. Comparison of Three Household Water Treatment Technologies in San Mateo Ixtatán, Guatemala. J. Environ. Eng. 2015, 141, 04014085. (34) Arnold, B. F.; Colford, J. M. Treating water with chlorine at pointof-use to improve water quality and reduce child diarrhea in developing countries: a systematic review and meta-analysis. Am. J. Trop. Med. Hyg. 2007, 76, 354−364. (35) Hunter, P. Household water treatment in developing countries: comparing different intervention types using meta-regression. Environ. Sci. Technol. 2009, 43, 8991−8997. (36) Bielefeldt, A. R.; Kowalski, K.; Schilling, C.; Schreier, S.; Kohler, A.; Scott Summers, R. Removal of virus to protozoan sized particles in point-of-use ceramic water filters. Water Res. 2010, 44, 1482−1488. (37) Fayer, R. Effect of sodium hypochlorite exposure on infectivity of Cryptosporidium parvum oocysts for neonatal BALB/c mice. Appl. Environ. Microbiol. 1995, 61, 844−846.

collection for the previous studies used in this research. We would also like to thank Dr. Kara Nelson from UC Berkeley for all of her valuable help and advice in preparing this manuscript. This work was funded by Yale University’s Climate and Energy Institute.



REFERENCES

(1) Onda, K.; LoBuglio, J.; Bartram, J. Global access to safe water: accounting for water quality and the resulting impact on MDG progress. Int. J. Environ. Res. Public Health 2012, 9, 880−894. (2) WHO, UNICEF, Progress on drinking water and sanitation 2015 update. 2015. (3) Barros, V.; Field, C. Climate Change 2014: Impacts, Adaptation, and Vulnerability: Working Group II Contribution to the IPCC 5th Assessment Report of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, U.K., 2014; Vol. 5. (4) Prüss-Ustün, A.; Bartram, J.; Clasen, T.; Colford, J. M.; Cumming, O.; Curtis, V.; Bonjour, S.; Dangour, A. D.; De France, J.; Fewtrell, L.; et al. Burden of disease from inadequate water, sanitation and hygiene in low- and middle-income settings: a retrospective analysis of data from 145 countries. Trop. Med. Int. Health 2014, 19, 894−905. (5) Kolstad, E. W.; Johansson, K. A. Uncertainties associated with quantifying climate change impacts on human health: a case study for diarrhea. Environ. Health Perspect. 2010, 119, 299. (6) Singh, R. B.; Hales, S.; de Wet, N.; Raj, R.; Hearnden, M.; Weinstein, P. The influence of climate variation and change on diarrheal disease in the Pacific Islands. Environ. Health Perspect. 2001, 109, 155. (7) Hashizume, M.; Armstrong, B.; Hajat, S.; Wagatsuma, Y.; Faruque, A. S.; Hayashi, T.; Sack, D. A. Association between climate variability and hospital visits for non-cholera diarrhoea in Bangladesh: effects and vulnerable groups. International Journal of Epidemiology 2007, 36, 1030− 1037. (8) Checkley, W.; Epstein, L. D.; Gilman, R. H.; Figueroa, D.; Cama, R. I.; Patz, J. A.; Black, R. E. Effects of El Niño and ambient temperature on hospital admissions for diarrhoeal diseases in Peruvian children. Lancet 2000, 355, 442−450. (9) Lama, J. R.; Seas, C. R.; León-Barúa, R.; Gotuzzo, E.; Sack, R. B. Environmental temperature, cholera, and acute diarrhoea in adults in Lima, Peru. Journal of Health, Population and Nutrition 2004, 399−403. (10) Carlton, E. J.; Woster, A. P.; DeWitt, P.; Goldstein, R. S.; Levy, K. A systematic review and meta-analysis of ambient temperature and diarrhoeal diseases. International Journal of Epidemiology 2016, 45, 117− 130. (11) Alexander, K. A.; Carzolio, M.; Goodin, D.; Vance, E. Climate change is likely to worsen the public health threat of diarrheal disease in Botswana. Int. J. Environ. Res. Public Health 2013, 10, 1202−1230. (12) Bhavnani, D.; Goldstick, J. E.; Cevallos, W.; Trueba, G.; Eisenberg, J. N. Impact of rainfall on diarrheal disease risk associated with unimproved water and sanitation. Am. J. Trop. Med. Hyg. 2014, 90, 705−711. (13) Carlton, E. J.; Eisenberg, J. N.; Goldstick, J.; Cevallos, W.; Trostle, J.; Levy, K. Heavy rainfall events and diarrhea incidence: the role of social and environmental factors. Am. J. Epidemiol. 2014, 179, 344−352. (14) Levy, K.; Woster, A. P.; Goldstein, R. S.; Carlton, E. J. Untangling the Impacts of Climate Change on Waterborne Diseases: a Systematic Review of Relationships between Diarrheal Diseases and Temperature, Rainfall, Flooding, and Drought. Environ. Sci. Technol. 2016, 50, 4905− 4922. (15) Walker, C. L. F.; Sack, D.; Black, R. E. Etiology of diarrhea in older children, adolescents and adults: a systematic review. PLoS Neglected Trop. Dis. 2010, 4, e768. (16) Kotloff, K. L.; Nataro, J. P.; Blackwelder, W. C.; Nasrin, D.; Farag, T. H.; Panchalingam, S.; Wu, Y.; Sow, S. O.; Sur, D.; Breiman, R. F.; et al. Burden and aetiology of diarrhoeal disease in infants and young children in developing countries (the Global Enteric Multicenter Study, GEMS): a prospective, case-control study. Lancet 2013, 382, 209−222. (17) Albert, M. J.; Faruque, A.; Faruque, S.; Sack, R.; Mahalanabis, D. Case-control study of enteropathogens associated with childhood I

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology (38) Ercumen, A.; Arnold, B. F.; Kumpel, E.; Burt, Z.; Ray, I.; Nelson, K.; Colford, J. M., Jr Upgrading a piped water supply from intermittent to continuous delivery and association with waterborne illness: a matched cohort study in urban India. PLoS Med. 2015, 12, e1001892. (39) Wu, J.; Long, S.; Das, D.; Dorner, S. Are microbial indicators and pathogens correlated? A statistical analysis of 40 years of research. J. Water Health 2011, 9, 265−278. (40) Gruber, J. S.; Ercumen, A.; Colford, J. M., Jr Coliform bacteria as indicators of diarrheal risk in household drinking water: systematic review and meta-analysis. PLoS One 2014, 9, e107429. (41) Tisue, S.; Wilensky, U. International Conference on Complex Systems; 2004; pp 16−21. (42) Semenov, M. A.; Stratonovitch, P. Use of multi-model ensembles from global climate models for assessment of climate change impacts. Climate Research 2010, 41, 1. (43) Nakicenovic, N.; Swart, R. Special report on emissions scenarios. Nakicenovic, N., Swart, R., Eds.; Cambridge University Press: Cambridge, U.K., July 2000, 2000, 1; p 612. (44) Ajjampur, S.; Rajendran, P.; Ramani, S.; Banerjee, I.; Monica, B.; Sankaran, P.; Rosario, V.; Arumugam, R.; Sarkar, R.; Ward, H.; et al. Closing the diarrhoea diagnostic gap in Indian children by the application of molecular techniques. J. Med. Microbiol. 2008, 57, 1364−1368. (45) Haas, C. N.; Rose, J. B.; Gerba, C. P. Quantitative Microbial Risk Assessment; John Wiley & Sons: New York, 1999. (46) CAMRA, Quantitative Microbial Risk Assessment (QMRA) Wiki. 2014; http://qmrawiki.msu.edu/, Accessed: April 20, 2014. (47) Ward, R. L.; Bernstein, D. I.; Young, E. C.; Sherwood, J. R.; Knowlton, D. R.; Schiff, G. M. Human rotavirus studies in volunteers: determination of infectious dose and serological response to infection. J. Infect. Dis. 1986, 154, 871−880. (48) Messner, M. J.; Chappell, C. L.; Okhuysen, P. C. Risk assessment for cryptosporidium: a hierarchical bayesian analysis of human dose response data. Water Res. 2001, 35, 3934−3940. (49) DuPont, H. L.; Formal, S. B.; Hornick, R. B.; Snyder, M. J.; Libonati, J. P.; Sheahan, D. G.; LaBrec, E. H.; Kalas, J. P. Pathogenesis of Escherichia coli diarrhea. N. Engl. J. Med. 1971, 285, 1−9. (50) Howard, G.; Bartram, J. Domestic Water Quantity, Service Level, And Health; World Health Organization: Geneva, 2003. (51) Eisenberg, J.; Scott, J.; Porco, T. Integrating disease control strategies: balancing water sanitation and hygiene interventions to reduce diarrheal disease burden. Am. J. Public Health 2007, 97, 846. (52) Pickering, A. J.; Julian, T. R.; Marks, S. J.; Mattioli, M. C.; Boehm, A. B.; Schwab, K. J.; Davis, J. Fecal contamination and diarrheal pathogens on surfaces and in soils among Tanzanian households with and without improved sanitation. Environ. Sci. Technol. 2012, 46, 5736− 5743. (53) Sterk, A.; Schijven, J.; de Nijs, T.; de Roda Husman, A. M. Direct and Indirect Effects of Climate Change on the Risk of Infection by Water-Transmitted Pathogens. Environ. Sci. Technol. 2013, 47, 12648− 12660. (54) Enger, K.; Nelson, K.; Rose, J.; Eisenberg, J. The joint effects of efficacy and compliance: a study of household water treatment effectiveness against childhood diarrhea. Water Res. 2013, 47, 1181− 1190. (55) Enger, K.; Nelson, K.; Clasen, T.; Rose, J.; Eisenberg, J. Linking Quantitative Microbial Risk Assessment and Epidemiological Data: Informing Safe Drinking Water Trials in Developing Countries. Environ. Sci. Technol. 2012, 46, 5160−5167. (56) Brown, J.; Clasen, T. High adherence is necessary to realize health gains from water quality interventions. PLoS One 2012, 7, e36735. (57) Hunter, P. R.; Zmirou-Navier, D.; Hartemann, P. Estimating the impact on health of poor reliability of drinking water interventions in developing countries. Sci. Total Environ. 2009, 407, 2621−2624. (58) Philipsborn, R.; Ahmed, S. M.; Brosi, B. J.; Levy, K. Climatic drivers of diarrheagenic Escherichia coli: A systematic review and metaanalysis. J. Infect. Dis. 2016, 214, 6. (59) Jagai, J. S.; Sarkar, R.; Castronovo, D.; Kattula, D.; McEntee, J.; Ward, H.; Kang, G.; Naumova, E. N. Seasonality of rotavirus in South

Asia: a meta-analysis approach assessing associations with temperature, precipitation, and vegetation index. PLoS One 2012, 7, e38168. (60) Ajjampur, S. S. R.; Liakath, F. B.; Kannan, A.; Rajendran, P.; Sarkar, R.; Moses, P. D.; Simon, A.; Agarwal, I.; Mathew, A.; O’Connor, R.; et al. Multisite study of cryptosporidiosis in children with diarrhea in India. Journal of clinical microbiology 2010, 48, 2075−2081. (61) Brown, J.; Sobsey, M. Microbiological effectiveness of locally produced ceramic filters for drinking water treatment in Cambodia. J. Water Health 2010, 8, 1−10. (62) Rosa, G.; Clasen, T. Estimating the scope of household water treatment in low-and medium-income countries. Am. J. Trop. Med. Hyg. 2010, 82, 289−300. (63) Ahmed, S. M.; Lopman, B. A.; Levy, K. A Systematic Review and Meta-Analysis of the Global Seasonality of Norovirus. PLoS One 2013, 8, e75922. (64) The MAL-ED Network Investigators. The MAL-ED Study: A multinational and multidisciplinary approach to understand the relationship between enteric pathogens, malnutrition, gut physiology, physical growth, cognitive development, and immune responses in infants and children up to 2 years of age in resource-poor environmentsClinical Infectious Diseases 2014, 59, S193−S206. (65) Odagiri, M.; Schriewer, A.; Daniels, M. E.; Wuertz, S.; Smith, W. A.; Clasen, T.; Schmidt, W.-P.; Jin, Y.; Torondel, B.; Misra, P. R.; et al. Human fecal and pathogen exposure pathways in rural Indian villages and the effect of increased latrine coverage. Water Res. 2016, 100, 232− 244. (66) Daniels, M. E.; Smith, W. A.; Schmidt, W.-P.; Clasen, T.; Jenkins, M. W. Modeling Cryptosporidium and Giardia in ground and surface water sources in rural India: associations with latrines, livestock, damaged wells, and rainfall patterns. Environ. Sci. Technol. 2016, 50, 7498−7507. (67) Kostyla, C.; Bain, R.; Cronk, R.; Bartram, J. Seasonal variation of fecal contamination in drinking water sources in developing countries: A systematic review. Sci. Total Environ. 2015, 514, 333−343. (68) Levy, K.; Hubbard, A. E.; Nelson, K. L.; Eisenberg, J. N. Drivers of water quality variability in northern coastal Ecuador. Environ. Sci. Technol. 2009, 43, 1788−1797. (69) Madhi, S. A.; Cunliffe, N. A.; Steele, D.; Witte, D.; Kirsten, M.; Louw, C.; Ngwira, B.; Victor, J. C.; Gillard, P. H.; Cheuvart, B. B.; et al. Effect of Human Rotavirus Vaccine on Severe Diarrhea in African Infants. N. Engl. J. Med. 2010, 362, 289−298. (70) Sobsey, M.; Stauber, C.; Casanova, L.; Brown, J.; Elliott, M. Point of use household drinking water filtration: a practical, effective solution for providing sustained access to safe drinking water in the developing world. Environ. Sci. Technol. 2008, 42, 4261−4267. (71) Clasen, T. Household water treatment and safe storage to prevent diarrheal disease in developing countries. Current environmental health reports 2015, 2, 69−74. (72) Boisson, S.; Kiyombo, M.; Sthreshley, L.; Tumba, S.; Makambo, J.; Clasen, T. Field assessment of a novel household-based water filtration device: a randomised, placebo-controlled trial in the Democratic Republic of Congo. PLoS One 2010, 5, e12613. (73) Pickering, A. J.; Davis, J. Freshwater availability and water fetching distance affect child health in sub-Saharan Africa. Environ. Sci. Technol. 2012, 46, 2391−2397. (74) Mellor, J. E.; Watkins, D.; Mihelcic, J. Rural water usage in East Africa: Does collection effort really impact basic access? Waterlines 2012, 31, 215−225. (75) Levy, K.; Nelson, K. L.; Hubbard, A.; Eisenberg, J. N. Rethinking indicators of microbial drinking water quality for health studies in tropical developing countries: case study in northern coastal Ecuador. Am. J. Trop. Med. Hyg. 2012, 86, 499. (76) UNICEF, WHO. Global water supply and sanitation assessment: 2000 report; 2003.

J

DOI: 10.1021/acs.est.6b02092 Environ. Sci. Technol. XXXX, XXX, XXX−XXX