Research gaps and priorities for quantitative microbial risk assessment (QMRA).

The coronavirus disease 2019 pandemic highlighted the need for more rapid and routine application of modeling approaches such as quantitative microbial risk assessment (QMRA) for protecting public health. QMRA is a transdisciplinary science dedicated to understanding, predicting, and mitigating infectious disease risks. To better equip QMRA researchers to inform policy and public health management, an Advances in Research for QMRA workshop was held to synthesize a path forward for QMRA research. We summarize insights from 41 QMRA researchers and experts to clarify the role of QMRA in risk analysis by (1) identifying key research needs, (2) highlighting emerging applications of QMRA; and (3) describing data needs and key scientific efforts to improve the science of QMRA. Key identified research priorities included using molecular tools in QMRA, advancing dose-response methodology, addressing needed exposure assessments, harmonizing environmental monitoring for QMRA, unifying a divide between disease transmission and QMRA models, calibrating and/or validating QMRA models, modeling co-exposures and mixtures, and standardizing practices for incorporating variability and uncertainty throughout the source-to-outcome continuum. Cross-cutting needs identified were to: develop a community of research and practice, integrate QMRA with other scientific approaches, increase QMRA translation and impacts, build communication strategies, and encourage sustainable funding mechanisms. Ultimately, a vision for advancing the science of QMRA is outlined for informing national to global health assessments, controls, and policies.

Impacts of Municipal Water-Rainwater Source Transitions on Microbial and Chemical Water Quality Dynamics at the Tap.

When rainwater harvesting is utilized as an alternative water resource in buildings, a combination of municipal water and rainwater is typically required to meet water demands. Altering source water chemistry can disrupt pipe scale and biofilm and negatively impact water quality at the distribution level. Still, it is unknown if similar reactions occur within building plumbing following a transition in source water quality. The goal of this study was to investigate changes in water chemistry and microbiology at a green building following a transition between municipal water and rainwater. We monitored water chemistry (metals, alkalinity, and disinfectant byproducts) and microbiology (total cell counts, plate counts, and opportunistic pathogen gene markers) throughout two source water transitions. Several constituents including alkalinity and disinfectant byproducts served as indicators of municipal water remaining in the system since the rainwater source does not contain these constituents. In the treated rainwater, microbial proliferation and Legionella spp. gene copy numbers were often three logs higher than those in municipal water. Because of differences in source water chemistry, rainwater and municipal water uniquely interacted with building plumbing and generated distinctively different drinking water chemical and microbial quality profiles.

Evaluation of the effectiveness of conservation practices under implementation site uncertainty.

Agricultural nonpoint source pollution is the leading source of water quality degradation in United States, which has led to the development of programs that aim to mitigate this pollution. One common approach to mitigating nonpoint source pollution is the use of best management practices (BMPs). However, it can be challenging to evaluate the effectiveness of implemented BMPs due to polices that limit data sharing. In this study, the uncertainty introduced by data sharing limitations is quantified through the use of a watershed model. Results indicated that BMP implementation improved the overall water quality in the region (up to ∼15% pollution reduction) and that increasing the area of BMP implementation resulted in higher pollution reduction. However, the model outputs also indicated that uncertainty caused by data sharing limitations resulted in variabilities ranging from -160% to 140%. This shows the importance of data sharing among agencies to better guide current and future conservation programs.

Food Footprint as a Measure of Sustainability for Grazing Dairy Farms.

Livestock productions require significant resources allocation in the form of land, water, energy, air, and capital. Meanwhile, owing to increase in the global demand for livestock products, it is wise to consider sustainable livestock practices. In the past few decades, footprints have emerged as indicators for sustainability assessment. In this study, we are introducing a new footprint measure to assess sustainability of a grazing dairy farm while considering carbon, water, energy, and economic impacts of milk production. To achieve this goal, a representative farm was developed based on grazing dairy practices surveys in the State of Michigan, USA. This information was incorporated into the Integrated Farm System Model (IFSM) to estimate the farm carbon, water, energy, and economic impacts and associated footprints for ten different regions in Michigan. A multi-criterion decision-making method called VIKOR was used to determine the overall impacts of the representative farms. This new measure is called the food footprint. Using this new indicator, the most sustainable milk production level (8618 kg/cow/year) was identified that is 19.4% higher than the average milk production (7215 kg/cow/year) in the area of interest. In addition, the most sustainable pasture composition was identified as 90% tall fescue with 10% white clover. The methodology introduced here can be adopted in other regions to improve sustainability by reducing water, energy, and environmental impacts of grazing dairy farms, while maximizing the farm profit and productions.

Development and evaluation of a comprehensive drought index.

Droughts are known as the world's costliest natural disasters impacting a variety of sectors. Despite their wide range of impacts, no universal drought definition has been defined. The goal of this study is to define a universal drought index that considers drought impacts on meteorological, agricultural, hydrological, and stream health categories. Additionally, predictive drought models are developed to capture both categorical (meteorological, hydrological, and agricultural) and overall impacts of drought. In order to achieve these goals, thirteen commonly used drought indices were aggregated to develop a universal drought index named MASH. The thirteen drought indices consist of four drought indices from each meteorological, hydrological, and agricultural categories, and one from the stream health category. Cluster analysis was performed to find the three closest indices in each category. Then the closest drought indices were averaged in each category to create the categorical drought score. Finally, the categorical drought scores were simply averaged to develop the MASH drought index. In order to develop predictive drought models for each category and MASH, the ReliefF algorithm was used to rank 90 variables and select the best variable set. Using the best variable set, the adaptive neuro-fuzzy inference system (ANFIS) was used to develop drought predictive models and their accuracy was examined using the 10-fold cross validation technique. The models' predictabilities ranged from R2 = 0.75 for MASH to R2 = 0.98 for the hydrological drought model. The results of this study can help managers to better position resources to cope with drought by reducing drought impacts on different sectors.

Evaluating the significance of wetland restoration scenarios on phosphorus removal.

Freshwater resources are vital for human and natural systems. However, anthropogenic activities, such as agricultural practices, have led to the degradation of the quality of these limited resources through pollutant loading. Agricultural Best Management Practices (BMPs), such as wetlands, are recommended as a valuable solution for pollutant removal. However, evaluation of their long-term impacts is difficult and requires modeling since performing in-situ monitoring is expensive and not feasible at the watershed scale. In this study, the impact of natural wetland implementation on total phosphorus reduction was evaluated both at the subwatershed and watershed levels. The study area is the Saginaw River Watershed, which is largest watershed in Michigan. The phosphorus reduction performances of four different wetland sizes (2, 4, 6, and 8 ha) were evaluated within this study area by implementing one wetland at a time in areas identified to have the highest potential for wetland restoration. The subwatershed level phosphorus loads were obtained from a calibrated Soil and Water Assessment Tool (SWAT) model. These loads were then incorporated into a wetland model (System for Urban Stormwater Treatment and Analysis IntegratioN-SUSTAIN) to evaluate phosphorus reduction at the subwatershed level and then the SWAT model was again used to route phosphorus transport to the watershed outlet. Statistical analyses were performed to evaluate the spatial impact of wetland size and placement on phosphorus reduction. Overall, the performance of 2 ha wetlands in total phosphorus reduction was significantly lower than the larger sizes at both the subwatershed and watershed levels. Regarding wetland implementation sites, wetlands located in headwaters and downstream had significantly higher phosphorus reduction than the ones located in the middle of the watershed. More specifically, wetlands implemented at distances ranging from 200 to 250 km and 50-100 km from the outlet had the highest impact on phosphorus reduction at the subwatershed and watershed levels, respectively. A multi criteria decision making (MCDM) method named VIKOR was successfully executed to identify the most suitable wetland size and location for each subwatershed considering the phosphorus reduction and economic cost associated with wetland implementation. The methods introduced in this study can be easily applied to other watersheds for selection and placement of wetlands while considering environmental benefits and economic costs.

Optimization of bioenergy crop selection and placement based on a stream health indicator using an evolutionary algorithm.

The emission of greenhouse gases continues to amplify the impacts of global climate change. This has led to the increased focus on using renewable energy sources, such as biofuels, due to their lower impact on the environment. However, the production of biofuels can still have negative impacts on water resources. This study introduces a new strategy to optimize bioenergy landscapes while improving stream health for the region. To accomplish this, several hydrological models including the Soil and Water Assessment Tool, Hydrologic Integrity Tool, and Adaptive Neruro Fuzzy Inference System, were linked to develop stream health predictor models. These models are capable of estimating stream health scores based on the Index of Biological Integrity. The coupling of the aforementioned models was used to guide a genetic algorithm to design watershed-scale bioenergy landscapes. Thirteen bioenergy managements were considered based on the high probability of adaptation by farmers in the study area. Results from two thousand runs identified an optimum bioenergy crops placement that maximized the stream health for the Flint River Watershed in Michigan. The final overall stream health score was 50.93, which was improved from the current stream health score of 48.19. This was shown to be a significant improvement at the 1% significant level. For this final bioenergy landscape the most often used management was miscanthus (27.07%), followed by corn-soybean-rye (19.00%), corn stover-soybean (18.09%), and corn-soybean (16.43%). The technique introduced in this study can be successfully modified for use in different regions and can be used by stakeholders and decision makers to develop bioenergy landscapes that maximize stream health in the area of interest.

Application of risk-based multiple criteria decision analysis for selection of the best agricultural scenario for effective watershed management.

Effective watershed management requires the evaluation of agricultural best management practice (BMP) scenarios which carefully consider the relevant environmental, economic, and social criteria involved. In the Multiple Criteria Decision-Making (MCDM) process, scenarios are first evaluated and then ranked to determine the most desirable outcome for the particular watershed. The main challenge of this process is the accurate identification of the best solution for the watershed in question, despite the various risk attitudes presented by the associated decision-makers (DMs). This paper introduces a novel approach for implementation of the MCDM process based on a comparative neutral risk/risk-based decision analysis, which results in the selection of the most desirable scenario for use in the entire watershed. At the sub-basin level, each scenario includes multiple BMPs with scores that have been calculated using the criteria derived from two cases of neutral risk and risk-based decision-making. The simple additive weighting (SAW) operator is applied for use in neutral risk decision-making, while the ordered weighted averaging (OWA) and induced OWA (IOWA) operators are effective for risk-based decision-making. At the watershed level, the BMP scores of the sub-basins are aggregated to calculate each scenarios' combined goodness measurements; the most desirable scenario for the entire watershed is then selected based on the combined goodness measurements. Our final results illustrate the type of operator and risk attitudes needed to satisfy the relevant criteria within the number of sub-basins, and how they ultimately affect the final ranking of the given scenarios. The methodology proposed here has been successfully applied to the Honeyoey Creek-Pine Creek watershed in Michigan, USA to evaluate various BMP scenarios and determine the best solution for both the stakeholders and the overall stream health.

Application of analytical hierarchy process for effective selection of agricultural best management practices.

In this study an analytical hierarchy process (AHP) was used for ranking best management practices (BMPs) in the Saginaw River Watershed based on environmental, economic and social factors. Three spatial targeting methods were used for placement of BMPs on critical source areas (CSAs). The environment factors include sediment, total nitrogen, and total phosphorus reductions at the subbasin level and the watershed outlet. Economic factors were based on total BMP cost, including installation, maintenance, and opportunity costs. Social factors were divided into three favorability rankings (most favorable, moderately favorable, and least favorable) based on area allocated to each BMP. Equal weights (1/3) were considered for the three main factors while calculating the BMP rank by AHP. In this study three scenarios were compared. A comprehensive approach in which environmental, economic, and social aspects are simultaneously considered (Scenario 1) versus more traditional approaches in which both environmental and economic aspects were considered (Scenario 2) or only environmental aspects (sediment, TN, and TP) were considered (Scenario 3). In Scenario 1, only stripcropping (moderately favorable) was selected on all CSAs at the subbasin level, whereas stripcropping (49-69% of CSAs) and residue management (most favorable, 31-51% of CSAs) were selected by AHP based on the watershed outlet and three spatial targeting methods. In Scenario 2, native grass was eliminated by moderately preferable BMPs (stripcropping) both at the subbasin and watershed outlet levels due the lower BMP implementations cost compared to native grass. Finally, in Scenario 3, at subbasin level, the least socially preferable BMP (native grass) was selected in 100% of CSAs due to greater pollution reduction capacity compared to other BMPs. At watershed level, nearly 50% the CSAs selected stripcropping, and the remaining 50% of CSAs selected native grass and residue management equally.

Modeling the hydrological significance of wetland restoration scenarios.

Wetlands provide multiple socio-economic benefits, among them mitigating flood through short- and long-term water storage functions and assisting with reduction of downstream flood peaks. However, their effectiveness in controlling floods is dictated by wetland size and distribution within a watershed. Due to the complexity of wetland hydrological processes at the watershed scale, the Soil and Water Assessment Tool (SWAT) was used to study the impact of wetland restoration on streamflow rates and peaks in the Shiawassee River watershed of Michigan. Wetland restoration scenarios were developed based on combinations of wetland area (50, 100, 250, and 500 ha) and wetland depth (15, 30, 61, and 91 cm). Increasing wetland area, rather than depth, had a greater impact on long-term average daily streamflow. Wetland implementation resulted in negligible reductions in daily peak flow rates and frequency of peak flow events at the watershed outlet. In developing high impact areas for wetland restoration, similar locations were identified for reduction of subbasin and watershed outlet streamflow. However, the best combinations of area/depth differed depending on the goal of the restoration plan.

Evaluating the capabilities of watershed-scale models in estimating sediment yield at field-scale.

Many watershed model interfaces have been developed in recent years for predicting field-scale sediment loads. They share the goal of providing data for decisions aimed at improving watershed health and the effectiveness of water quality conservation efforts. The objectives of this study were to: 1) compare three watershed-scale models (Soil and Water Assessment Tool (SWAT), Field_SWAT, and the High Impact Targeting (HIT) model) against calibrated field-scale model (RUSLE2) in estimating sediment yield from 41 randomly selected agricultural fields within the River Raisin watershed; 2) evaluate the statistical significance among models; 3) assess the watershed models' capabilities in identifying areas of concern at the field level; 4) evaluate the reliability of the watershed-scale models for field-scale analysis. The SWAT model produced the most similar estimates to RUSLE2 by providing the closest median and the lowest absolute error in sediment yield predictions, while the HIT model estimates were the worst. Concerning statistically significant differences between models, SWAT was the only model found to be not significantly different from the calibrated RUSLE2 at α = 0.05. Meanwhile, all models were incapable of identifying priorities areas similar to the RUSLE2 model. Overall, SWAT provided the most correct estimates (51%) within the uncertainty bounds of RUSLE2 and is the most reliable among the studied models, while HIT is the least reliable. The results of this study suggest caution should be exercised when using watershed-scale models for field level decision-making, while field specific data is of paramount importance.

Linking biological integrity and watershed models to assess the impacts of historical land use and climate changes on stream health.

Land use change and other human disturbances have significant impacts on physicochemical and biological conditions of stream systems. Meanwhile, linking these disturbances with hydrology and water quality conditions is challenged due to the lack of high-resolution datasets and the selection of modeling techniques that can adequately deal with the complex and nonlinear relationships of natural systems. This study addresses the above concerns by employing a watershed model to obtain stream flow and water quality data and fill a critical gap in data collection. The data were then used to estimate fish index of biological integrity (IBI) within the Saginaw Bay basin in Michigan. Three methods were used in connecting hydrology and water quality variables to fish measures including stepwise linear regression, partial least squares regression, and fuzzy logic. The IBI predictive model developed using fuzzy logic showed the best performance with the R (2) = 0.48. The variables that identified as most correlated to IBI were average annual flow, average annual organic phosphorus, average seasonal nitrite, average seasonal nitrate, and stream gradient. Next, the predictions were extended to pre-settlement (mid-1800s) land use and climate conditions. Results showed overall significantly higher IBI scores under the pre-settlement land use scenario for the entire watershed. However, at the fish sampling locations, there was no significant difference in IBI. Results also showed that including historical climate data have strong influences on stream flow and water quality measures that interactively affect stream health; therefore, should be considered in developing baseline ecological conditions.

Watershed scale PFAS fate and transport model for source identification and management implications.

Developing strategic plans for the remediation and mitigation of pre- and polyfluoroalkyl substances (PFAS) in soil, groundwater, and surface water requires an understanding of the fate and transport of these chemicals on a regional scale. To fill this knowledge gap, we developed a distributed hydrogeochemical model and applied it to a large-scale watershed with various point and non-point sources of a long-chain, highly persistent PFAS compound known as perfluorooctane sulfonic acid (PFOS). The results showed that the developed model could reproduce the spatiotemporal concentration of PFOS across a large and diverse watershed. Herein, our first objective was to quantify the PFOS transport from the unsaturated zone to the groundwater and surface water via leaching, surface runoff, lateral flow, and sediment transport. The second objective was to identify factors influencing PFOS release from confirmed and suspected PFAS sites and urban and agricultural areas. The modeling results show that surface runoff played a significant role in PFOS transport, with urban areas and industrial sites being major contributors. In addition, sediment transport was found to be a notable pathway for PFOS release, particularly from sites with biosolids application. Further analysis revealed the relative importance of topography, soil water retention, and water-solid adsorption factors in determining PFOS transport dynamics at the watershed scale for better source identification and targeted management.

Large-scale assessment of PFAS compounds in drinking water sources using machine learning.

The monitoring of Per and Polyfluoroalkyl substances (PFAS) in drinking water sources has significantly increased due to their recognition as a major public health concern. This information has been utilized to assess the importance of potential explanatory variables in determining the presence and concentration of PFAS in different regions. Nevertheless, the significance of these variables and the reliability of the methods in regions beyond where they were initially tested is still uncertain. Hence, our research pursues two main objectives: 1) to evaluate the validity of the aforementioned variables and methods for several PFAS species in a different area and 2) to build on existing modeling work; a new PFAS predictive model is introduced which is more reliable in determining the presence and concentration of PFAS at a regional level. To achieve these goals, we reconstructed four state-of-the-art models using a statewide dataset available for Michigan. These models involve spatial regression techniques, classification and regression random forest algorithms, and boosted regression trees. They also include numerous explanatory variables, such as features of local soil and hydrology and the number of nearby contamination sources. Then, we use a Bayesian selection approach to find the most relevant among these variables. Finally, we employ the most relevant covariates to assess PFAS occurrence and estimate their concentration using a novel combination of machine learning algorithms and conditional autoregressive (CAR) modeling. In the first case, PFAS occurrence was assessed with an accuracy comparable to the reconstructed models (>90%) while using significantly fewer variables. In the second case, by maintaining low data requirements, the estimated concentrations of most PFAS compounds were more closely aligned with available observations compared to previous methods, with correlation coefficients ρ > 0.90 and R2 > 0.77.

Evaluation of targeting methods for implementation of best management practices in the Saginaw River Watershed.

Increasing concerns regarding water quality in the Great Lakes region are mainly due to changes in urban and agricultural landscapes. Both point and non-point sources contribute pollution to Great Lakes surface waters. Best management practices (BMPs) are a common tool used to reduce both point and non-point source pollution and improve water quality. Meanwhile, identification of critical source areas of pollution and placement of BMPs plays an important role in pollution reduction. The goal of this study is to evaluate the performance of different targeting methods in 1) identifying priority areas (high, medium, and low) based on various factors such as pollutant concentration, load, and yield, 2) comparing pollutant (sediment, total nitrogen-TN, and total phosphorus-TP) reduction in priority areas defined by all targeting methods, 3) determine the BMP relative sensitivity index among all targeting methods. Ten BMPs were implemented in the Saginaw River Watershed using the Soil and Water Assessment Tool (SWAT) model following identification of priority areas. Each targeting method selected distinct high priority areas based on the methodology of implementation. The concentration based targeting method was most effective at reduction of TN and TP, likely because it selected the greatest area of high priority for BMP implementation. The subbasin load targeting method was most effective at reducing sediment because it tended to select large, highly agricultural subbasins for BMP implementation. When implementing BMPs, native grass and terraces were generally the most effective, while conservation tillage and residue management had limited effectiveness. The BMP relative sensitivity index revealed that most combinations of targeting methods and priority areas resulted in a proportional decrease in pollutant load from the subbasin level and watershed outlet. However, the concentration and yield methods were more effective at subbasin reduction, while the stream load method was more effective at reducing pollutants at the watershed outlet. The results of this study indicate that emphasis should be placed on selection of the proper targeting method and BMP to meet the needs and goals of a BMP implementation project because different targeting methods produce varying results.

Connecting microbial, nutrient, physiochemical, and land use variables for the evaluation of water quality within mixed use watersheds.

As non-point sources of pollution begin to overtake point sources in watersheds, source identification and complicating variables such as rainfall are growing in importance. Microbial source tracking (MST) allows for identification of fecal contamination sources in watersheds; when combined with data on land use and co-occuring variables (e.g., nutrients, sediment runoff) MST can provide a basis for understanding how to effectively remediate water quality. To determine spatial and temporal trends in microbial contamination and correlations between MST and nutrients, water samples (n = 136) were collected between April 2017 and May of 2018 during eight sampling events from 17 sites in 5 mixed-use watersheds. These samples were analyzed for three MST markers (human - B. theta; bovine - CowM2; porcine - Pig2Bac) along with E. coli, nutrients (nitrogen and phosphorus species), and physiochemical paramaters. These water quality variables were then paired with data on land use, streamflow, precipitation and management practices (e.g., tile drainage, septic tank density, tillage practices) to determine if any significant relationships existed between the observed microbial contamination and these variables. The porcine marker was the only marker that was highly correlated (p value <0.05) with nitrogen and phosphorus species in multiple clustering schemes. Significant relationships were also identified between MST markers and variables that demonstrated temporal trends driven by precipitation and spatial trends driven by septic tanks and management practices (tillage and drainage) when spatial clustering was employed.

Modeling the persistence of viruses in untreated groundwater.

Several factors can affect virus behavior and persistence in water sources. Historically linear models have been used to describe persistence over time; however, these models do not consider all of the factors that can affect inactivation kinetics or the observed patterns of decay. Meanwhile, applying the appropriate persistence model is critical for ensuring that decision makers are minimizing human health risk in the event of contamination and exposure to contaminated groundwater. Therefore, to address uncertainty in predictions of decay or virus concentrations over time, this study fit seventeen different linear and nonlinear mathematical models to persistence data from a previously conducted sampling study on drinking water wells throughout the United States. The models were fit using Maximum Likelihood Estimation and the best fitting models were determined by the Bayesian Information Criterion. The purpose of the study was to identify the best model for estimating decay of viruses in groundwater and to determine if model uncertainty contributes to erroneous predictions of viral contamination when only conventional models are considered. For the datasets analyzed in this study, the Juneja and Marks models and the exponential damped model were more representative of the persistence of viruses in groundwater than the traditionally used linear models. The results from this study were then evaluated with classification trees in order to identify more relevant modeling methodology for future research. The classification trees aid in narrowing the scope of appropriate persistence models based on characteristics of the experimental conditions and water sampled.

Ecohydrological modeling for large-scale environmental impact assessment.

Ecohydrological models are frequently used to assess the biological integrity of unsampled streams. These models vary in complexity and scale, and their utility depends on their final application. Tradeoffs are usually made in model scale, where large-scale models are useful for determining broad impacts of human activities on biological conditions, and regional-scale (e.g. watershed or ecoregion) models provide stakeholders greater detail at the individual stream reach level. Given these tradeoffs, the objective of this study was to develop large-scale stream health models with reach level accuracy similar to regional-scale models thereby allowing for impacts assessments and improved decision-making capabilities. To accomplish this, four measures of biological integrity (Ephemeroptera, Plecoptera, and Trichoptera taxa (EPT), Family Index of Biotic Integrity (FIBI), Hilsenhoff Biotic Index (HBI), and fish Index of Biotic Integrity (IBI)) were modeled based on four thermal classes (cold, cold-transitional, cool, and warm) of streams that broadly dictate the distribution of aquatic biota in Michigan. The Soil and Water Assessment Tool (SWAT) was used to simulate streamflow and water quality in seven watersheds and the Hydrologic Index Tool was used to calculate 171 ecologically relevant flow regime variables. Unique variables were selected for each thermal class using a Bayesian variable selection method. The variables were then used in development of adaptive neuro-fuzzy inference systems (ANFIS) models of EPT, FIBI, HBI, and IBI. ANFIS model accuracy improved when accounting for stream thermal class rather than developing a global model.

Ecohydrological model parameter selection for stream health evaluation.

Variable selection is a critical step in development of empirical stream health prediction models. This study develops a framework for selecting important in-stream variables to predict four measures of biological integrity: total number of Ephemeroptera, Plecoptera, and Trichoptera (EPT) taxa, family index of biotic integrity (FIBI), Hilsenhoff biotic integrity (HBI), and fish index of biotic integrity (IBI). Over 200 flow regime and water quality variables were calculated using the Hydrologic Index Tool (HIT) and Soil and Water Assessment Tool (SWAT). Streams of the River Raisin watershed in Michigan were grouped using the Strahler stream classification system (orders 1-3 and orders 4-6), k-means clustering technique (two clusters: C1 and C2), and all streams (one grouping). For each grouping, variable selection was performed using Bayesian variable selection, principal component analysis, and Spearman's rank correlation. Following selection of best variable sets, models were developed to predict the measures of biological integrity using adaptive-neuro fuzzy inference systems (ANFIS), a technique well-suited to complex, nonlinear ecological problems. Multiple unique variable sets were identified, all which differed by selection method and stream grouping. Final best models were mostly built using the Bayesian variable selection method. The most effective stream grouping method varied by health measure, although k-means clustering and grouping by stream order were always superior to models built without grouping. Commonly selected variables were related to streamflow magnitude, rate of change, and seasonal nitrate concentration. Each best model was effective in simulating stream health observations, with EPT taxa validation R2 ranging from 0.67 to 0.92, FIBI ranging from 0.49 to 0.85, HBI from 0.56 to 0.75, and fish IBI at 0.99 for all best models. The comprehensive variable selection and modeling process proposed here is a robust method that extends our understanding of watershed scale stream health beyond sparse monitoring points.

Modeling Escherichia coli removal in constructed wetlands under pulse loading.

Manure-borne pathogens are a threat to water quality and have resulted in disease outbreaks globally. Land application of livestock manure to croplands may result in pathogen transport through surface runoff and tile drains, eventually entering water bodies such as rivers and wetlands. The goal of this study was to develop a robust model for estimating the pathogen removal in surface flow wetlands under pulse loading conditions. A new modeling approach was used to describe Escherichia coli removal in pulse-loaded constructed wetlands using adaptive neuro-fuzzy inference systems (ANFIS). Several ANFIS models were developed and validated using experimental data under pulse loading over two seasons (winter and summer). In addition to ANFIS, a mechanistic fecal coliform removal model was validated using the same sets of experimental data. The results showed that the ANFIS model significantly improved the ability to describe the dynamics of E. coli removal under pulse loading. The mechanistic model performed poorly as demonstrated by lower coefficient of determination and higher root mean squared error compared to the ANFIS models. The E. coli concentrations corresponding to the inflection points on the tracer study were keys to improving the predictability of the E. coli removal model.