Evaluating the potential efficacy of mercury total maximum daily loads on aqueous methylmercury levels in four coastal watersheds

Of the approximately 780 U.S. EPA approved mercury total maximum daily loads (TMDLs), most specify a reduction in total mercury (Hg(T)) loads to reduce methylmercury levels in fish tissue, assuming a 1:1 correspondence. However, mercury methylation is more complex, and therefore, proposed load reductions may not be adequate. Using multiple regression with microlevel and macrolevel variables, the potential efficacy of mercury TMDLs on decreasing aqueous methylmercury levels was investigated in four coastal watersheds: Mugu Lagoon (CA), San Francisco Bay Estuary, Long Island Sound, and south Florida. Hg(T) and methylmercury levels were positively correlated in all watersheds except in Long Island Sound, where spatial differences explained over 40% of the variability in methylmercury levels. A mercury TMDL would be least effective in Long Island Sound due to spatial heterogeneity but most effective in south Florida, where the ratio between aqueous Hg(T) and methylmercury levels was close to 1 and the 95% confidence interval was narrow, indicating a probable reduction in aqueous methylmercury levels if Hg(T) loads were reduced.     

Exposure of the Irish population to PBDEs in food: consideration of parameter uncertainty and variability for risk assessment

Polybrominated diphenyl ethers (PBDEs) are brominated flame retardants used to retard the ignition and/or spread of fire. PBDEs are used in various consumer products, such as textiles, mattresses and TV screens. This study presents a chemical risk assessment for the Irish population based on exposure to PBDEs from food. Special regard is given to the influence of parameter uncertainty and variability on the margins of safety. To quantitatively model uncertainty and variability in concentration data and variability in consumer behavior, a hierarchical probabilistic model was constructed. This model was evaluated using a two-dimensional Monte Carlo simulation (2D-MCS) approach. By considering uncertainty and variability in concentration data, margins of safety (MOS) were derived that are lower by a factor of ～2 compared to MOS based on dose estimates that only consider variability. The lowest MOS is 7.5?×?10⁴ for BDE-99, with impaired spermatogenesis as toxic endpoint. Assuming an MOS of 10⁴ as acceptable, we conclude that there is no significant risk for human health through intake of contaminated food. To investigate whether additional measurements could improve the quality of dose estimates, the statistic “uncertainty-to-variability (UVR)” was developed. By applying the UVR to our dose estimates, we show that, in our case, the datasets contain little uncertainty and additional measurements would not significantly improve the quality of dose estimates.     

Assessment of uncertainty in a probabilistic model of consumer exposure to pesticide residues in food

The assessment of consumer exposure to pesticides is an important part of pesticide regulation. Probabilistic modelling allows analysis of uncertainty and variability in risk assessments. The output of any assessment will be influenced by the characteristics and uncertainty of the inputs, model structure and assumptions. While the use of probabilistic models is well established in the United States, in Europe problems of low acceptance, sparse data and lack of guidelines are slowing the development. The analyses in the current paper focused on the dietary pathway and the exposure of UK toddlers. Three single food, single pesticide case studies were used to parameterize a simple probabilistic model built in Crystal Ball™. Data on dietary consumption patterns were extracted from National Diet and Nutrition Surveys, and levels of pesticide active ingredients in foods were collected from Pesticide Residues Committee monitoring. The effect of uncertainty on the exposure estimate was analysed using scenarios, reflecting different assumptions related to sources of uncertainty. The most influential uncertainty issue was the distribution type used to represent input variables. Other sources that most affected model output were non-detects, unit-to-unit variability and processing. Specifying correlation between variables was found to have little effect on exposure estimates. The findings have important implications for how probabilistic modelling should be conducted, communicated and used by policy and decision makers as part of consumer risk assessment of pesticides.     

Assessing the risks of pesticide residues to consumers: recent and future developments

Assessing exposure of consumers to pesticide residues is an area of regulatory science that has rapidly developed over the last decade. From simplistic, deterministic models calculating lifetime exposure for adults only, assessment procedures have diversified so that more realistic estimates of long term exposures for adults, schoolchildren, toddlers and infants and short term exposures for adults and toddlers (who generally bound the more extreme consumer patterns) are now carried out. The final assessment of risk still remains a simplistic numeric comparison against hazard assessment based on a wide range of toxicity studies incorporating the appropriate safety or uncertainty factors. As development of risk assessments continues, the use of probabilistic models is becoming an invaluable information tool for quantitative risk management and aiding assessment of cumulative exposure. This paper examines the recent developments in risk assessment and consumer perception of the risks of pesticide residues, and speculates where the future developments in these areas may lie.     

Evaluating variability and uncertainty separately in microbial quantitative risk assessment using two R packages

Quantitative risk assessment has emerged as a valuable tool to enhance the scientific basis of regulatory decisions in the food safety domain. This article introduces the use of two new computing resources (R packages) specifically developed to help risk assessors in their projects. The first package, “fitdistrplus”, gathers tools for choosing and fitting a parametric univariate distribution to a given dataset. The data may be continuous or discrete. Continuous data may be right-, left- or interval-censored as is frequently obtained with analytical methods, with the possibility of various censoring thresholds within the dataset. Bootstrap procedures then allow the assessor to evaluate and model the uncertainty around the parameters and to transfer this information into a quantitative risk assessment model. The second package, “mc2d”, helps to build and study two dimensional (or second-order) Monte-Carlo simulations in which the estimation of variability and uncertainty in the risk estimates is separated. This package easily allows the transfer of separated variability and uncertainty along a chain of conditional mathematical and probabilistic models. The usefulness of these packages is illustrated through a risk assessment of hemolytic and uremic syndrome in children linked to the presence of Escherichia coli O157:H7 in ground beef. These R packages are freely available at the Comprehensive R Archive Network (cran.r-project.org).     

Organophosphorus insecticides in agricultural and residential runoff: field observations and implications for total maximum daily load development

Development of total maximum daily loads (TMDLs) for nonpoint source pollutants requires mass flux estimates for targeted compounds from contributing sources. We measured organophosphorus insecticide concentrations in surface runoff from agricultural and residential land-use sites in a southern Californian watershed over the course of runoff-producing irrigation and rainfall events. Event mean concentrations (EMCs) for chlorpyrifos, diazinon, and malathion exhibited considerable variability among irrigation and storm runoff events at agricultural sites; residential storm runoff EMCs for these compounds were considerably less variable. Event loads and EMCs were higher for runoff events following reported insecticide applications. Organophosphorus insecticide EMCs were not consistently correlated with hydrologic characteristics of runoff events. Our results indicate that on an area basis, loads from residential land may exceed those from sites planted in row crops for a given rainfall depth, suggesting that residential land use warrants explicit consideration in TMDL development and implementation. No consistent first flush effect was discernible for organophosphorus insecticides in storm or irrigation runoff. A relative potency factor approach is introduced to permit evaluation of organophosphorus insecticides on a common toxicological basis and allow development of TMDLs and pollutant control strategies for these compounds as a class.     

Bayesian analysis of a reduced-form air quality model

Numerical air quality models are being used for assessing emission control strategies for improving ambient pollution levels across the globe. This paper applies probabilistic modeling to evaluate the effectiveness of emission reduction scenarios aimed at lowering ground-level ozone concentrations. A Bayesian hierarchical model is used to combine air quality model output and monitoring data in order to characterize the impact of emissions reductions while accounting for different degrees of uncertainty in the modeled emissions inputs. The probabilistic model predictions are weighted based on population density in order to better quantify the societal benefits/disbenefits of four hypothetical emission reduction scenarios in which domain-wide NO(x) emissions from various sectors are reduced individually and then simultaneously. Cross validation analysis shows the statistical model performs well compared to observed ozone levels. Accounting for the variability and uncertainty in the emissions and atmospheric systems being modeled is shown to impact how emission reduction scenarios would be ranked, compared to standard methodology.     

Estimation of uncertainty and variability in bacterial growth using Bayesian inference. Application to Listeria monocytogenes

The usefulness of risk assessment is limited by its ability or inability to model and evaluate risk uncertainty and variability separately. A key factor of variability and uncertainty in microbial risk assessment could be growth variability between strains and growth model parameter uncertainty. In this paper, we propose a Bayesian procedure for growth parameter estimation which makes it possible to separate these two components by means of hyperparameters. This model incorporates in a single step the logistic equation with delay as a primary growth model and the cardinal temperature equation as a secondary growth model. The estimation of Listeria monocytogenes growth parameters in milk using literature data is proposed as a detailed application. While this model should be applied on genuine data, it is highlighted that the proposed approach may be convenient for estimating the variability and uncertainty of growth parameters separately, using a complete predictive microbiology model.     

农药残留膳食暴露评估模型研究进展

农药残留膳食暴露风险评估为化学品风险管理、食品安全与人类健康领域的重点关注问题，有效的定量风险评估模型是开展农药残留膳食暴露评价的基础工具。本文结合国内外农药残留膳食暴露风险评估研究进展，旨在归纳农药残留膳食暴露风险评估模型，包括确定性评估模型、概率性评估模型和累积性评估模型；根据模型演变、适用情形与评估需求，比对分析模型间优缺点，探讨模型不确定度、评估软件、数据库应用等关键因子。同时立足我国农药残留膳食风险评估现状，展望模型相关参数整合的紧迫性与重要性，为完善农药残留膳食暴露风险评估体系建设提供科学参考。     

Optimization of stormwater filtration at the urban/watershed interface

Environmental pollution from cities is a major ecological problem attributed to contaminated runoff from nonpoint sources. The U.S. Environmental Protection Agency's guidance on implementation of total maximum daily loads (TMDL) does not adequately cover methods to improve waters impaired by nonpoint sources. To comply with TMDLs, cities may install filters in curb inlets, or use other Best Management Practices (BMPs). We tested 10 different filters and found their effectiveness in retaining pollutants ranged from 0 to >90%, depending on combinations of pollutant types (metals, pathogens, and total suspended sediments (TSS)) and filter materials. Hence, the decision to deploy filters into curb inlets must consider land use patterns associated with specific categories of pollutants generated within cities. We developed a geographic information system (GIS)-enabled model for estimating and mitigating emissions of pollutants from urban regions into watersheds. The model uses land use categories and pollutant loadings to optimize strategic placement of filters to accommodate TMDLs. For example, in a city where the landuse pattern generates 4 x 10(6) kg of TSS, 55 kg of Cd, and 2 x 10(3) kg of Zn per year into 498 curb inlets that discharge into a sensitive watershed, the optimized placement of 137, 92, and 148 filters can achieve TMDL endpoints for each pollutant, respectively. We show further that 158 strategically placed filters effectively meet the requirements simultaneously for all three pollutants, a result at least 5 times more effective than random placement of filters.     

Microbial growth modelling with artificial neural networks

There is a growing interest in modelling microbial growth as an alternative to time-consuming, traditional, microbiological enumeration techniques. Several statistical models have been reported to describe the growth of different microorganisms, but there are accuracy problems. An alternate technique 'artificial neural networks' (ANN) for modelling microbial growth is explained and evaluated. Published data were used to build separate general regression neural network (GRNN) structures for modelling growth of Aeromonas hydrophila, Shigella flexneri, and Brochothrix thermosphacta. Both GRNN and published statistical model predictions were compared against the experimental data using six statistical indices. For training data sets, the GRNN predictions were far superior than the statistical model predictions, whereas the GRNN predictions were similar or slightly worse than statistical model predictions for test data sets for all the three data sets. GRNN predictions can be considered good, considering its performance for unseen data. Graphical plots, mean relative percentage residual, mean absolute relative residual, and root mean squared residual were identified as suitable indices for comparing competing models. ANN can now become a vehicle whereby predictive microbiology can be applied in food product development and food safety risk assessment.     

Geostatistical modeling of the spatial distribution of soil dioxins in the vicinity of an incinerator. 1. Theory and application to Midland, Michigan

Deposition of pollutants around point sources of contamination, such as incinerators, can display complex spatial patterns depending on prevailing weather conditions, the local topography, and the characteristics of the source. Deterministic dispersion models often fail to capture the complexity observed in the field, resulting in uncertain predictions that might hamper subsequent decision-making, such as delineation of areas targeted for additional sampling or remediation. This paper describes a geostatistical simulation-based methodology that combines the detailed process-based modeling of atmospheric deposition from an incinerator with the probabilistic modeling of residual variability of field samples. The approach is used to delineate areas with high levels of dioxin TEQ(DF)-WHO98 (toxic equivalents) around an incinerator, accounting for 53 field data and the output of the EPA Industrial Source Complex (ISC3) dispersion model. The dispersion model explains 43.7% of the variance in the soil TEQ data, whereas the regression residuals are spatially correlated with a range of 776 m. One hundred realizations of soil TEQ values are simulated on a grid with a 50 m spacing. The benefit of stochastic simulation over spatial interpolation is 2-fold: (1) maps of simulated point TEQ values can easily be aggregated to the geography that is the most relevant for decision making (e.g., census block, ZIP codes); and (2) the uncertainty at the larger scale is simply modeled by the empirical distribution of block-averaged simulated values. Incorporating the output of the atmospheric deposition model as a spatial trend yields a more realistic prediction of the spatial distribution of TEQ values than log-normal kriging using only the field data, in particular, in sparsely sampled areas away from the incinerator. The geostatistical model provided guidance for the study design (census block-based population sampling) of the University of Michigan Dioxin Exposure Study (UMDES), focused on quantifying exposure pathways to dioxins from industrial sources, relative to background exposures.     

Cost effectiveness of regulation-compliant filtration to control sediment and metal pollution in urban runoff

The implementation of Total Maximum Daily Load (TMDL) to control urban runoff presents major structural and managerial challenges for cities. We developed a decision support system (DSS) for TMDL compliance at the city level to solve for a phased, least-cost strategy toward meeting four TMDLs using stormwater filtration. Based on a case-study city, we modeled wet weather flows and associated discharge of Total Suspended Sediment (TSS), cadmium, copper, and zinc to receiving waters by coupling U.S. EPA's Storm Water Management Model (SWMM v. 5.0) with the geographic dataset of the urban drainage network. We linked a mixed integer linear programming algorithm to the watershed model for deriving cost-effective selection and placement of curb inlet filters to meet mass- and concentration-based TMDL requirements. The least cost solution for meeting the city's TMDL waste load allocations for TSS (73.9% reduction), Cd (50.6% reduction), Cu (30.0% reduction), and Zn (55.7% reduction) would require 1071 filter inserts at a cost of $1.7 million. In contrast, random placement of 1071 filters or uniform placement of 1266 filters is effective only for TSS and would cost $4.0 million and $4.8 million, respectively. Our results demonstrate the increases in cost-effectiveness of using an optimization-based DSS for urban watershed management.     

Risk assessment for Escherichia coli O157:H7 in marketed unpasteurized milk in selected East African countries

We carried out a study to assess the risk associated with the presence of Shiga toxigenic Escherichia coli (STEC) in informally marketed unpasteurized milk in urban East Africa. Data for the risk models were obtained from on-going and recently completed studies in Kenya and Uganda. Inputs for the model were complemented with data from published literature in similar populations. A fault-tree scenario pathway and modular process risk model approach were used for exposure assessment. Hazard characterization was based on a socioeconomic study with dose-responses derived from the literature. We used a probabilistic approach with Monte Carlo simulation and inputs from farm and household surveys. The qualitative analysis suggested a low to moderate risk of infection from consuming milk and that the widespread consumer practice of boiling milk before consumption was an important risk mitigator. Quantitative analysis revealed that two to three symptomatic STEC infections could be expected for every 10,000 unpasteurized milk portions consumed, with a possible range of 0 to 22 symptomatic cases. Sensitivity analyses to assess the uncertainty and variability associated with the model revealed that the factor with the greatest influence on disease incidence was the prevalence of STEC in dairy cattle. Risk assessment is a potentially useful method for managing food safety in informal markets.     

Quantitative microbial risk assessment exemplified by Staphylococcus aureus in unripened cheese made from raw milk

This paper discusses some of the developments and problems in the field of quantitative microbial risk assessment, especially exposure assessment and probabilistic risk assessment models. To illustrate some of the topics, an initial risk assessment was presented, in which predictive microbiology and survey data were combined with probabilistic modelling to simulate the level of Staphylococcus aureus in unripened cheese made from raw milk at the time of consumption. Due to limited data and absence of dose-response models, a complete risk assessment was not possible. Instead, the final level of bacteria was used as a proxy for the potential enterotoxin level, and thus the potential for causing illness. The assessment endpoint selected for evaluation was the probability that a cheese contained at least 6 log cfu S. aureus g(-1) at the time of consumption; the probability of an unsatisfactory cheese, P(uc). The initial level of S. aureus, followed by storage temperature had the largest influence on P(uc) at the two pH-values investigated. P(uc) decreased with decreasing pH and was up to a factor of 30 lower in low pH cheeses due to a slower growth rate. Of the model assumptions examined, i.e. the proportion of enterotoxigenic strains, the level of S. aureus in non-detect cheeses, the temperature limit for toxin production, and the magnitude and variability of the threshold for an unsatisfactory cheese, it was the latter that had the greatest impact on P(uc). The uncertainty introduced by this assumption was in most cases less than a factor of 36, the same order of magnitude as the maximum variability due to pH. Several data gaps were identified and suggestions were made to improve the initial risk assessment, which is valid only to the extent that the limited data reflected the true conditions and that the assumptions made were valid. Despite the limitations, a quantitative approach was useful to gain insights and to evaluate several factors that influence the potential risk and to make some inferences with relevance to risk management. For instance, the possible effect of using starter cultures in the cheese making process to improve the safety of these products.     

A risk assessment approach applied to the growth of Erwinia carotovora in vegetable juice for variable temperature conditions

Risk assessment for food spoilage relies on probabilistic models of microbial growth to predict the likelihood that microbial populations will exceed predefined spoilage levels. To assist in the design and management of industrial food quality systems, predictive microbiological models have to incorporate major risk factors such as the variability in the microbial strain, environment and initial contamination levels. In addition, the application of results measured under laboratory conditions to the less controlled environment of an industrial process usually also involves uncertainty. Extra information regarding this uncertainty must be factored into industrial microbial risk assessment. In this paper, based on our previous analysis of the growth of Erwinia carotovora we show how different factors contribute to the risk of microbial spoilage of vegetable juice and we demonstrate an effective way of including these factors into risk assessment models. The association of risk components with different unavoidable and manageable factors is also valuable for the development of optimal strategies for reducing microbial risk.     

Assessing TMDL effectiveness using flow-adjusted concentrations: a case study of the Neuse River,North Carolina

Integrated control of both point and nonpoint source water pollution using Total Maximum Daily Load (TMDL) assignments will be a major regulatory focus over the next decade. We propose the use of "flow-adjusted" pollutant concentrations to evaluate the effectiveness of management actions taken to meet approved TMDLs. Pollutant concentrations are usually highly correlated with streamflow, and flow is strongly weather-dependent. Thus, pollutant loads, which are calculated as pollutant concentration multiplied by streamflow, have a large weather-dependent variance component. This natural variation can be removed by calculating flow-adjusted concentrations. While such values are not a direct measure of pollutant load, they make it easier to discern changes in streamwater quality. Additionally, they are likely to be a better predictor of pollutant concentrations in the receiving waterbody. We demonstrate the use of this technique using long-term nutrient data from the Neuse River in North Carolina. The Neuse River Estuary has suffered many eutrophication symptoms, and a program to reduce nutrient loading has been in place for several years. We show that, in addition to revealing recent reductions in nutrient inputs, annual flow-adjusted riverine nutrient concentrations show a more pronounced relationship with estuarine nutrient concentrations than do annual nutrient loads. Thus, we suggest that the calculation of flow-adjusted concentrations is a useful technique to aid in assessment of TMDL implementation.     

Biological variability and exposure assessment

Predictive models are now commonly used for exposure assessment, with growth parameters defined for each microbial species. In this study, we tried to take into account microbial growth variability among strains of a single species. Bacillus cereus in pasteurized milk was chosen to illustrate the influence of the biological variability on the outcome of exposure assessment. Each parameter of the exposure assessment (growth parameters, shelf-life conditions) was characterized by a probability distribution describing variability and/or uncertainty. The impact of the intra-species variability on the result of the exposure assessment was then quantified and discussed. Two simple domestic shelf life conditions were tested. The results confirm that the biological variability has a great impact on the accuracy of the result and should not be systematically neglected.     

Spatial variability and uncertainty in ecological risk assessment: a case study on the potential risk of cadmium for the little owl in a Dutch river flood plain

This paper outlines a procedure that quantifies the impact of different sources of spatial variability and uncertainty on ecological risk estimates. The procedure is illustrated in a case study that estimates the risks of cadmium for a little owl (Athene noctua vidalli) living in a Dutch river flood plain along the river Rhine. A geographical information system (GIS) was used to quantify spatial variability in contaminant concentrations and habitats. It was combined with an exposure and effect model that uses Monte Carlo simulation to quantify parameter uncertainty. Spatial model uncertainty was assessed by the application of two different spatial interpolation methods (classification and kriging) and foraging ranges. The results of the case study show that parameter uncertainty is the main type of uncertainty influencing the risk estimate, and to a lesser extent spatial variability, while spatial model uncertainty was of minor importance. Compared to the deterministically calculated hazard index for the little owl (0.9), inclusion of spatial variability resulted in a median hazard index that can vary between 0.8 and 1.4. It is concluded that a single estimator for a whole flood plain may over- or underestimate risks for specific parts within the flood plain. Further research that expands the procedure presented in this paper is necessary to improve the incorporation of spatial factors in ecological risk assessment.     

Opportunities for Reducing Total Maximum Daily Load (TMDL) Compliance Costs: Lessons from the Chesapeake Bay

The Chesapeake Bay Total Maximum Daily Load (TMDL) program is an unprecedented opportunity to restore the Chesapeake Bay, yet program costs threaten to undermine its complete implementation. Analyses of Bay TMDL program design and implementation were used to relate program cost-effectiveness to choices in (1) compliance definitions, (2) geographic load allocations, and (3) approaches to engaging unregulated sources. A key finding was that many design choices require choosing an acceptable level of risk of achieving water quality outcomes, and a lack of data can lead to precautionary choices, which increase compliance costs. Furthermore, although some choices managed costs, others decisions may have reduced the potential for cost savings from water quality trading and payment programs. In particular, the choice by some states to distribute the portion of load reductions that improve water quality in the Bay mainstem to many small basins is likely to diminish the potential for market development or reduce funding for the most cost-effective nutrient and sediment reduction practices. Strategies for reducing costs of future TMDLs include considering diminishing marginal returns early in the TMDL design to balance costs and risks in regulatory goal setting and to design rules and incentives that promote innovation and cost-effective compliance strategies.