Olof Bergstedt

Fault tree analysis for integrated and probabilistic risk analysis of drinking water systems.

Drinking water systems are vulnerable and subject to a wide range of risks. To avoid sub-optimisation of risk-reduction options, risk analyses need to include the entire drinking water system, from source to tap. Such an integrated approach demands tools that are able to model interactions between different events. Fault tree analysis is a risk estimation tool with the ability to model interactions between events. Using fault tree analysis on an integrated level, a probabilistic risk analysis of a large drinking water system in Sweden was carried out. The primary aims of the study were: (1) to develop a method for integrated and probabilistic risk analysis of entire drinking water systems; and (2) to evaluate the applicability of Customer Minutes Lost (CML) as a measure of risk. The analysis included situations where no water is delivered to the consumer (quantity failure) and situations where water is delivered but does not comply with water quality standards (quality failure). Hard data as well as expert judgements were used to estimate probabilities of events and uncertainties in the estimates. The calculations were performed using Monte Carlo simulations. CML is shown to be a useful measure of risks associated with drinking water systems. The method presented provides information on risk levels, probabilities of failure, failure rates and downtimes of the system. This information is available for the entire system as well as its different sub-systems. Furthermore, the method enables comparison of the results with performance targets and acceptable levels of risk. The method thus facilitates integrated risk analysis and consequently helps decision-makers to minimise sub-optimisation of risk-reduction options.

A theoretical approach to assess microbial risks due to failures in drinking water systems.

A failure in treatment or in the distribution network of a surface water-works could have serious consequences due to the variable raw water quality in combination with an extended distribution. The aim of this study was to examine the theoretical impact of incidents in the drinking water system on the annual risk of infection in a population served by a large water treatment plant in Sweden. Reported incidents in the system were examined and a microbial risk assessment that included three pathogens, Cryptosporidium parvum, rotavirus and Campylobacter jejuni, was performed. The main risk incidents in water treatment were associated with sub-optimal particle removal or disinfection malfunction. Incidents in the distribution network included cross-connections and microbial pollution of reservoirs and local networks. The majority of the annual infections were likely to be due to pathogens passing treatment during normal operation and not due to failures, thus adding to the endemic rate. Among the model organisms, rotavirus caused the largest number of infections. Decentralised water treatment with membranes was also considered in which failures upstream fine-pored membranes would have little impact as long as the membranes were kept intact.

Outbreak of norovirus in Västra Götaland associated with recreational activities at two lakes during August 2004

A large community outbreak of norovirus (NV) gastrointestinal infection occurred in Västra Götaland County, Sweden in August 2004, following attendance at recreational lakes. A frequency age-matched case control study was undertaken of persons who had attended these lakes to identify risk factors. 163 cases and 329 controls were included. Analysis indicates that having water in the mouth while swimming (OR = 4.7; 95% CI 1.1–20.2), attendance at the main swimming area at Delsjön Lake (OR = 25.5; 95% CI 2.5–263.8), taking water home from a fresh water spring near Delsjön lake (OR = 17.3; 95% CI 2.7–110.7) and swimming less than 20 m from shore (OR = 13.4; 95% CI 2.0–90.2) were significant risk factors. The probable vehicle was local contamination of the lake water (especially at the main swimming area). The source of contamination could not be determined

Decay of Bacteroidales Genetic Markers in Relation to Traditional Fecal Indicators for Water Quality Modeling of Drinking Water Sources

The implementation of microbial fecal source tracking (MST) methods in drinking water management is limited by the lack of knowledge on the transport and decay of host-specific genetic markers in water sources. To address these limitations, the decay and transport of human (BacH) and ruminant (BacR) fecal Bacteroidales 16S rRNA genetic markers in a drinking water source (Lake Rådasjön in Sweden) were simulated using a microbiological model coupled to a three-dimensional hydrodynamic model. The microbiological model was calibrated using data from outdoor microcosm trials performed in March, August, and November 2010 to determine the decay of BacH and BacR markers in relation to traditional fecal indicators. The microcosm trials indicated that the persistence of BacH and BacR in the microcosms was not significantly different from the persistence of traditional fecal indicators. The modeling of BacH and BacR transport within the lake illustrated that the highest levels of genetic markers at the raw water intakes were associated with human fecal sources (on-site sewers and emergency sewer overflow). This novel modeling approach improves the interpretation of MST data, especially when fecal pollution from the same host group is released into the water source from different sites in the catchment.

Variability analysis of pathogen and indicator loads from urban sewer systems along a river.

The pathogen loads within surface waters originating from urban wastewater sources needs to be assessed to support drinking water risk estimations and optimal selection of risk reduction measures. Locally reported discharges from sewer systems (>100,000 persons connected) were used to simulate the potential microbial loads into the Göta älv river, Sweden. Using Monte Carlo simulations, the median and 95% percentile (i.e. worst case) of total microbial load from wastewater treatment plants, sewer network overflows and emergency discharges were assessed and presented for dry and wet weather conditions. Wastewater treatment plants with secondary treatment represented a major source of E. coli, norovirus, Giardia and Cryptosporidium. During wet weather, comparably high microbial loads were found for sewer overflows due to heavy rains. Substantial loads were also associated with an incident of the emergency discharge of untreated wastewater. Simulated river water concentrations of faecal indicators (E. coli, sulfite reducing clostridia, somatic coliphages) and pathogens (norovirus, Giardia, Cryptosporidium) were confirmed by river sampling data, suggesting that urban wastewater is the major microbial source for this river.

Precipitation effects on microbial pollution in a river: lag structures and seasonal effect modification.

Background The river Göta Älv is a source of freshwater for 0.7 million swedes. The river is subject to contamination from sewer systems discharge and runoff from agricultural lands. Climate models projects an increase in precipitation and heavy rainfall in this region. This study aimed to determine how daily rainfall causes variation in indicators of pathogen loads, to increase knowledge of variations in river water quality and discuss implications for risk management. Methods Data covering 7 years of daily monitoring of river water turbidity and concentrations of E. coli, Clostridium and coliforms were obtained, and their short-term variations in relation with precipitation were analyzed with time series regression and non-linear distributed lag models. We studied how precipitation effects varied with season and compared different weather stations for predictive ability. Results Generally, the lowest raw water quality occurs 2 days after rainfall, with poor raw water quality continuing for several more days. A rainfall event of >15 mm/24-h (local 95 percentile) was associated with a three-fold higher concentration of E. coli and 30% higher turbidity levels (lag 2). Rainfall was associated with exponential increases in concentrations of indicator bacteria while the effect on turbidity attenuated with very heavy rainfall. Clear associations were also observed between consecutive days of wet weather and decreased water quality. The precipitation effect on increased levels of indicator bacteria was significant in all seasons. Conclusions Rainfall elevates microbial risks year-round in this river and freshwater source and acts as the main driver of varying water quality. Heavy rainfall appears to be a better predictor of fecal pollution than water turbidity. An increase of wet weather and extreme events with climate change will lower river water quality even more, indicating greater challenges for drinking water producers, and suggesting better control of sources of pollution.

Epidemiology and estimated costs of a large waterborne outbreak of norovirus infection in Sweden.

A large outbreak of norovirus (NoV) gastroenteritis caused by contaminated municipal drinking water occurred in Lilla Edet, Sweden, 2008. Epidemiological investigations performed using a questionnaire survey showed an association between consumption of municipal drinking water and illness (odds ratio 4·73, 95% confidence interval 3·53-6·32), and a strong correlation between the risk of being sick and the number of glasses of municipal water consumed. Diverse NoV strains were detected in stool samples from patients, NoV genotype I strains predominating. Although NoVs were not detected in water samples, coliphages were identified as a marker of viral contamination. About 2400 (18·5%) of the 13,000 inhabitants in Lilla Edet became ill. Costs associated with the outbreak were collected via a questionnaire survey given to organizations and municipalities involved in or affected by the outbreak. Total costs including sick leave, were estimated to be ∼8,700,000 Swedish kronor (∼€0·87 million).

Cost-effectiveness analysis of risk-reduction measures to reach water safety targets

Identifying the most suitable risk-reduction measures in drinking water systems requires a thorough analysis of possible alternatives. In addition to the effects on the risk level, also the economic aspects of the risk-reduction alternatives are commonly considered important. Drinking water supplies are complex systems and to avoid sub-optimisation of risk-reduction measures, the entire system from source to tap needs to be considered. There is a lack of methods for quantification of water supply risk reduction in an economic context for entire drinking water systems. The aim of this paper is to present a novel approach for risk assessment in combination with economic analysis to evaluate risk-reduction measures based on a source-to-tap approach. The approach combines a probabilistic and dynamic fault tree method with cost-effectiveness analysis (CEA). The developed approach comprises the following main parts: (1) quantification of risk reduction of alternatives using a probabilistic fault tree model of the entire system; (2) combination of the modelling results with CEA; and (3) evaluation of the alternatives with respect to the risk reduction, the probability of not reaching water safety targets and the cost-effectiveness. The fault tree method and CEA enable comparison of risk-reduction measures in the same quantitative unit and consider costs and uncertainties. The approach provides a structured and thorough analysis of risk-reduction measures that facilitates transparency and long-term planning of drinking water systems in order to avoid sub-optimisation of available resources for risk reduction.

Replacement predictions for drinking water networks through historical data.

Lifetime distribution functions and current network age data can be combined to provide an assessment of the future replacement needs for drinking water distribution networks. Reliable lifetime predictions are limited by a lack of understanding of deterioration processes for different pipe materials under varied conditions. An alternative approach is the use of real historical data for replacement over an extended time series. In this paper, future replacement needs are predicted through historical data representing more than one hundred years of drinking water pipe replacement in Gothenburg, Sweden. The verified data fits well with commonly used lifetime distribution curves. Predictions for the future are discussed in the context of path dependence theory.

Estimation of pathogen concentrations in a drinking water source using hydrodynamic modelling and microbial source tracking.

The faecal contamination of drinking water sources can lead to waterborne disease outbreaks. To estimate a potential risk for waterborne infections caused by faecal contamination of drinking water sources, knowledge of the pathogen concentrations in raw water is required. We suggest a novel approach to estimate pathogen concentrations in a drinking water source by using microbial source tracking data and fate and transport modelling. First, the pathogen (norovirus, Cryptosporidium, Escherichia coli O157/H7) concentrations in faecal contamination sources around the drinking water source Lake Rådasjön in Sweden were estimated for endemic and epidemic conditions using measured concentrations of faecal indicators (E. coli and Bacteroidales genetic markers). Afterwards, the fate and transport of pathogens within the lake were simulated using a three-dimensional coupled hydrodynamic and microbiological model. This approach provided information on the contribution from different contamination sources to the pathogen concentrations at the water intake of a drinking water treatment plant. This approach addresses the limitations of monitoring and provides data for quantitative microbial risk assessment (QMRA) and risk management in the context of faecal contamination of surface drinking water sources.