# Assessing Microbial Risks from Land Applied Class B Biosolids Through Groundwater Pathway Under Wet Weather Events

Assessing Microbial Risks from Land Applied Class B Biosolids Through Groundwater Pathway Under Wet Weather Events

Abstract

Introduction

The exposure model is the most essential and complicated part of QMRA. Both model structure and parameter selections are crucial for the reliability of the results. To assess the biosolids-associated risks from groundwater pathway, the situation under rainfalls need to be accounted. The rainfall events may increase the resulting risks from biosolids by producing a large amount of water infiltrated in the soil and advancing the microbial transport significantly. The subsurface fate and transport model must take several crucial elements into account, including occurrence values of pathogens in biosolids, the potential routes of infection, the probability of human exposure to the source of the pathogen, as well as the amount that humans would ingest, and the virulence of the infectious agent (U.S. EPA 2000).

The objective of this research is to assess the microbial risk for human exposure to biosolids-associated pathogens through contaminated ground water by integrating available knowledge for modeling transport and fate of microorganisms. Field monitoring studies are used to validate the model prediction. The model is also applied to a site-specific scenario and the risks across pathogens are compared and indicator and pathogen relationships are identified.

Methods

A comprehensive model has been developed for microorganisms released from biosolids-soil mixture on the field surface, infiltrated to the groundwater and finally transport to the nearby drinking water well (Figure 1). A fully explicit infiltration model was developed to predict infiltration depth resulting from specific rainfall events using the Green-Ampt infiltration method (Ravi and Williams 1998; Zhang 2009; Teng, Kumar et al. 2011). Two groundwater transport scenarios were considered, depending on whether or not the infiltrating wetting front from the rainfall event saturates through to the ground water table, creating a fully saturated connection (Figure 1). Scenario 1 (non-saturating rainfall event): if the depth to the water table, h, is larger than wetting front depth, Z, pathogen attenuation includes three processes: vertical transport through saturated soil above the wetting front, vertical transport in unsaturated soil below the wetting front but above the water table, and horizontal transport in saturated soil through groundwater flow to the downstream well. Scenario 2 (saturating rainfall event): if the depth to the water table, h, is smaller than the wetting front depth, Z, during a rainfall event, pathogens transport vertically in saturated soil above the wetting front and then join the saturated horizontal groundwater flow to the downstream well without any attenuation through unsaturated soil. Scenario 2 presents a greater risk of pathogen transport. Based on boundary information calculated from infiltration models, an analytical solution to the advection-dispersion equation was used to model microbial transport and fate through each layer of porous media (either saturated soil or unsaturated soil). The transport models give the final number or concentration of microorganisms in the exposure media of groundwater well water, which was combined with dose-response relationships and used to predict the resulting risks to human health by risk models. The flowchart in Figure 2 shows how the three major components work to complete the risk assessment.

Figure 1 Groundwater exposure model

Figure 2 Flow chart for model development

(1) Infiltration models

A joint Green-Ampt model has been developed to predict the depth of wetting front based on the rainfall information using a analytical solution, which makes it feasible to be encoded in our final product of a spreadsheet-based tool (Galada, Gurian et al. 2011). The joint Green-ampt model predicts the depth of wetting front under the wet weather events (Teng, Kumar et al. 2011). Before the rainfall event, soil above the groundwater table is normally not fully saturated and termed as unsaturated soil. But during and after the rainfall event, the soil above the wetting front is saturated by the rainfall water and assumed to be saturated soil; the soil between the wetting front and the groundwater table remains unsaturated. The thickness of unsaturated soil and saturated soil provides boundary information for the following microbial transport and fate model.

(2) Microbial transport and fate models

The transport model through porous media was developed using advection-dispersion equation incorporating with decay and adsorption to soil. The exposure model was populated with microorganism-specific parameters, such as occurrence data, decay rates in water, soil-water partitioning constants and dose-response data, and provides a time-dependent, microbial concentration profile as a function of distance. Microbial straining is also considered. There are two types of porous media considered for microbial transport and fate modeling, including saturated soil and unsaturated soil.

The governing equation for microbial transport through saturated soil, both for vertical wetting zone transport and horizontal groundwater flow, is the one-dimensional advection-dispersion model with an instantaneous source and including effects of adsorption to soil and first-order inactivation of pathogens (Bedient, et al., 1997) (Equation 1).

(1)

where Dx is coefficient of hydrodynamic dispersion (cm2/h) (= αvx, where α is dispersivity), vx is the average seepage velocity (cm/h), λ is the first order inactivation rate (/h), and R is the retardation factor. The retardation factor is defined as 1+(ρb/n)Kd, where ρb is the bulk dry mass density (g/cm3), n is porosity, and Kd is the equilibrium distribution coefficient (cm3/g).

The transport model through unsaturated soil considered mass transfer across the liquid-solid and liquid-air interfaces and is used for the unsaturated barrier between the water table depth and the wetting front depth, if applicable (Faulkner, Lyon et al. 2002) (Equation 2).

(2) where Cl, Cs, and Ca are the concentration in the liquid, liquid-solid interface, and air-liquid interface, respectively; ρ is the bulk density of the solid matrix; θ is the volumetric moisture content; λl, λs, and λa are the inactivation rate coefficients of pathogen suspended in the liquid, pathogen sorbed at the liquid-solid interface, and pathogen sorbed at the air-liquid interface, respectively.

The straining mechanism was included in the model to capture pathogen removal due to physical straining and is particularly relevant for larger microbes, such as bacteria and protozoa. The straining removal is determined by the coefficient (kstr) and the distance for straining (hstr). Equation 3 was used to predict the concentration of strained microorganisms (Tufenkji 2007).

(3)

where C is the effluent concentration and C0 is the influent concentration, kstr is the straining coefficient, which could be estimated using a correlation based on the microorganism and grain size ratio (Bradford, Simunek et al. 2003) , v is the interstitial microbe velocity, and L is transport distance. The fraction of strained microorganisms is calculated as (1-C/C0).

(3) Risk models

Risks of infection depend on exposed dose, pathogen type, and pathogen-specific dose-response models. Risk models are used to calculate risks of infection based on the predicted environmental concentration from microbial transport and fate models (Equation 4). Exponential models are used to calculate risk of infection per application period from a particular pathogen (Haas, Rose et al. 1999) (Equation 5). Since the microbial transport with groundwater flow produces variable microbial concentrations in well water, daily risk calculations were made for the exposure based on daily dose specific to each day. Risk of infection per application period was calculated as Equation 6. It is crucial to note that the final risks are conditional on the occurrence probability of rainfall events, which can be calculated from the rainfall return periods.

(4)

(5)

(6)

where Concdaily,i is environmental concentrations of microorganisms on the ith day (number/L), Exp is exposure rate (L/day), Dosedaily,i is the daily exposed dose on the ith day (number), Riskdaily,i is daily risk of infection on the ith day, r is the fraction of the ingested microorganisms that survive to initiate infections or host-microorganisms interaction probability, Riskapp is risk of infection per application period, and Dapp is duration of application period (days).

SMART Biosolids model

An environmental dispersion, exposure, and risk model, named the Spreadsheet Microbial Assessment of Risk: Tool for Biosolids (“SMART Biosolids”) was developed by encoding the solutions to the models described above (Galada, Gurian et al. 2011). SMART Biosolids model has a spreadsheet interface for entering inputs and accessing outputs, and add-in macros for repeated iterative computation. The inputs required from a user include site-specific parameters (such as parameters associated with application events, climate and soil characteristics), as well as those for uncertainty analysis (such as the iteration numbers, and option to restore nominal parameter values after an uncertainty analysis). Representative data and references for pathogen-specific fate and transport parameters from diverse sources are stored in the spreadsheets. The outputs include estimates and uncertainties of expected concentrations of microbes in groundwater resulting from biosolids applications, and expected probability of infection by these microbes for nearby residents.

The SMART Biosolids model was applied to a site-specific condition to demonstrate the types of results that the model can generate. A field study representing a typical application event was used for selection of model input parameters. The controlled site with lysimeters and a portable rainfall simulator was monitored to evaluate the leaching and ponding of viral contaminants following land application of biosolids (Wong, Harrigan et al. 2010). Mesophilic anaerobic digested (MAD) biosolids was applied on sandy-loam soil. Portable rainfall simulators applied water on a semi-continuous basis to minimize surface ponding. Based on the total amount of water applied per day, the inputs of rainfall rate and duration were approximated for the model. The inputs describing the site characteristics and application events are shown in Table S1.

Figure 3 Peak concentrations of organisms in groundwater resulting from biosolids application. Setback distances were assumed to be 3-ft water table depth, and 100-ft from application field to nearby well.

Figure 4 Risks of infection per application period for nearby residents by adenoviruses from land-applied biosolids. Error bars show uncertainties with 90% percentiles. Ingestion of groundwater assumes ingestion of 2L water per day, and 3 days per application period.

Figure 4 shows risks of infection from pathogens with uncertainties. Adenovirus is the only pathogen presenting risk by the groundwater pathway. However, the large uncertainties indicate that additional information on the fate and transport of adenovirus in groundwater is needed.

Model validation

The observed results from the field study by Wong et al. (2010) are used for model validation. Since the leachate samples were drawn from the bottom of the lysimeter (240 cm below soil surface), the vertical transport model of SMART Biosolids is used for validation by using 7.87 ft (240 cm) as water table depth. Lysimeter leachate was collected and analyzed for anionic tracer (chloride), microbial tracer (salmonella phage P-22), adenovirus, and somatic phage. Neither adenovirus nor somatic phage was recovered from the leachate samples. P-22 bacteriophage was found in the leachate of three lysimeters (removal rates ranged from 1.8 to 3.2 log10/m). Since the phage P-22 is not one of the available pathogens of concern in the SMART Biosolids model and its transport behavior is similar to coliphage (Tye, Chan et al. 1974), results for coliphage from the SMART Biosolids model are compared to the observed results for phage P-22 in the field study. The outputs of microbial concentration files are examined and compared to the observed results. Three model outputs will be compared with the observed data described here. Each of these outputs will be used to estimate different model parameters (see Table 1): anionic tracer concentrations will be used to estimate hydraulic parameters, P-22 concentrations in the lysimeters will be used to estimate sorption to solids and decay rate, and surface concentrations of phage will be used to estimate attenuation rates. The Solver function in Microsoft Excel will be used to estimate model parameters by minimizing the sum of the squared deviations between the modeled and observed values.

Table 1 Model validation: Comparison of model outputs and field observational data

Microorganisms Approaches Results

Chloride Adjust hydraulic parameters, including saturated water content, depth of saturated soil to fit the breakthrough time;

Then adjust release fraction to fit the breakthrough curves.

Chloride is modeled by using 1 as retardation factor, and 0 as die-off rate. Best fit at saturated water content at 0.69, depth of saturated soil of 90 cm, and saturated release fraction of 0.13.

P-22 Using fitted water content and other hydraulic parameters for chloride;

Adjust retardation factor, release fraction and decay rate to fit breakthrough curves for P-22 Best fit at retardation of 5, release fraction of 0.06, and decay of 0.008 log/hr

P-22 and somatic phage Compare values of decay rate from several models and pick the best decay model Exponential is the best model

(1) Anionic tracer concentration. Wong et al. (2010) observed that the peak of chloride breakthrough occurred around 0.3 pore volumes in each lysimeter, and the peak of P-22 breakthrough varied between lysimeters (<0.1, 0.3 and 0.7 pore volumes). The early time to peak breakthrough of anionic and microbial tracers indicates preferential flow paths. The hydraulic parameters, including saturated water content, depth of saturated soil were adjusted to fit the breakthrough time; then the release fraction was adjusted to fit the breakthrough curves for chloride. Chloride is modeled by using 1 as retardation factor, and 0 as die-off rate. Best fit is at saturated water content of 0.69, depth of saturated soil of 90 cm, and saturated release fraction of 0.13.

(2) P-22 concentrations. Wong et al. (2010) concluded that the difference between breakthrough curves and recoveries between chloride and P-22 was due to sorption. By using the same hydraulic parameters fitted for chloride, the retardation factor, release fraction and decay rate were adjusted to fit the breakthrough curves for P-22. Best fit is at retardation of 5, the release fraction of 0.06, and decay of 0.008 log/hr.

The percentage of adenovirus desorbed from soil (with 8% organic matter) was 2% to 4%, which decreased to 1% from soil with less organic matter (2 %). The recoveries of coliphage from column transport studies were 3.3% to 7.4 %. The fittest release fraction of 6% for P-22, which has the similar transport behavior to coliphage, is consistent with the measured recovery values.

The measured decay of P-22 in the surface water sample was fitted to a first order decay model and the fitted decay rate is 0.0186 log/hr. The fitted decay of 0.008 log/hr shows the persistence of organisms in the groundwater comparing to surface water, considering the effects of UV and temperature .

(3) Virus Decay rate. Wong et al. (2010) observed somatic phage reached background levels at about day ten. They concluded that microbial pollution from runoff following significant rainfall events is possible when biosolids remain on the soil surface. Die-off rates of P-22 and somatic phage in the surface water were measured, and fit to a first order decay model. The observed die-off rate data can be fit to several other mathematical models, such as exponential (Doyle 2001; Kingsley, Hollinian et al. 2007), biphasic exponential (Doyle 2001), exponentially damped (Doyle 2001), Juneja & Marks (Juneja and Marks 2001; Juneja and Marks 2003), general logistic (Juneja and Marks 2001; Juneja and Marks 2003), Gompertz (Juneja and Marks 2001; Juneja and Marks 2003). Several decay curves of P-22 in surface water are shown. The exponential model fits the data best.

Name General mathematical form Modeled parameters General comments

exponential

k decay constant total number of parameters = 3

N0 initial population

continuous biphasic exponential

k1 1st-phase decay constant total number of parameters = 5

biphasic breakpoint assumed to be t=72 hr, considered model parameter for BIC comparisons

k2 2nd-phase decay constant

N0 initial population

discontinuous biphasic exponential

k1 1st-phase decay constant total number of parameters = 6

biphasic breakpoint assumed to be t=72 hr, considered model parameter for BIC comparisons

k2 2nd-phase decay constant

N01 1st-phase initial population

N02 2nd-phase initial population

Haas

k decay constant general solution to

total number of parameters = 4

a power scale factor

N0 initial population

exponentially damped

k decay constant general solution to

total number of parameters = 4

s damping constant

N0 initial population

Juneja & Marks (1)

k decay constant total number of parameters = 4

m avg. “hits” to cell required for die-off

N0 initial population

Juneja & Marks (2)

a generic constant total number of parameters = 4

b generic constant

N0 initial population

general logistic

k decay constant total number of parameters = 3

N0 initial population

Gompertz

k decay constant total number of parameters = 4

s general constant

N0 initial population

Model application

Two field sites were recently monitored in Imlay County, MI to describe the wet-weather-driven fate of biosolids-associated contaminants (Kumar, Simmons et al. 2010). Wet-weather sampling was conducted for 80 days from June to September (2009) following biosolids application. The field characteristics and application information are used as input for the SMART Biosolids model (Table S2). Risks from different pathogens are compared and the indicator and pathogen relationships are examined.

Figure X Risk of infection per application period from groundwater with a 2-ft water table and setback distance of 150 ft to well.

Figure X Risk of infection per application period from groundwater with varied water table

Figure X Risk of infection per application period from groundwater with varied distance to well

Adenovirus has the highest nominal risk estimate (3.78×10-3), exceeding the next highest risk of 3.5×10-5 due to Cryptosporidium by almost 2 orders of magnitude. While these results are based on many assumptions, including the use of idealized, homogeneous transport models, when interpreted cautiously the results can help prioritize among different risks and identify future research needs. In this case the model uncertainties are large and noteworthy. Further research could be directed towards studying the occurrence and transport of adenoviruses, which have both a high nominal risk estimate and very substantial uncertainty. Similar plots can be prepared to contrast risks from different pathogens across the remaining four exposure pathways (Galada, Gurian et al. 2011).

The results indicate that sandy loam soil was an effective filter for removing organisms for groundwater protection. The depth of water table needs to be considered more seriously than setback distance to well.

Discussion

Conclusion

REFERENCE

SUPPORTING DOCUMENT: TABLES AND FIGURES

Table S1 Inputs for SMART Biosolids model

Parameters Values Units Description

/Remarks

Rainfall intensity 0.25-0.33 cm/h Water application intensity and duration were assumed to be the averaged values: 0.29 cm/h and 90 hours

Rainfall duration 72-108 h

Soil texture class Sandy loam -

Area of application site 20 Acre Assume a square plot

Slope of the plot 4.0 % Natural surface slope between 2 and 6%; assumed 4% (assumed 4%)

Application method Splash-plate spray applicator None

Distance to well 100

feet Minimum setback distances requirement by Virginia State (Evanylo 2009)

Water table depth 3 feet Water table setbacks requirement by Maine State is 3 ft (Eisenberg 2006).

Hydraulic Gradient 0.04 - Assuming equal to overland surface slope

Pathogens and indicators in the biosolids

Potassium chloride 1.87×104 ppm 18.65 g/L (one mole in 4L water=(74.6 g)/(4L)), with assuming 1g/mL density of biosolids.

P-22 3.00×109 PFU/g biosolids 3×1011 PFU/100mL in biosolids with assuming 1g/mL density. Used for coliphage concentration.

Adenovirus 4.20×106 PFU/g biosolids 4.20×108 PFU/100mL in biosolids with assuming 1g/mL density.

Table S2 Model application: Input parameters for SMART Biosolids model

Parameter Value Unit Description

/Remarks

Rainfall intensity 4.83 cm/h Average precipitation is 1.9 inches (=4.83cm) from the figure (average 5.8 mm from the text). Assuming intensity = 4.83cm/h, and duration = 1 h.

Rainfall duration 1 h

Soil texture class Loam -

Area of application site 20 Acre Assume a square plot.

Slope of the plot 4.0 % Natural surface slope between 2 and 6%; assumed 4%.

Application method Splash-plate spray applicator None

Biosolids application rate 1.37 dry tons biosolids

/acre =(5.91 L/m2)(5.2g/0.1L)(4046.86m2/acre)(1kg/1000g)(1/907ton/kg), assuming 100% biosolids deposited on the field surface.

Water Table Depth 2

feet Assumed to be 2 feet; also vary: 0 feet, 1feet, 2 feet, 3feet

Distance to Well 150 Feet Assume the nearby ditch as a well which is 150 ft away from the boundary of the plot with biosolids application (i.e., equal to existing width of buffer for ditch); also vary: 50 ft, 100 ft, and 150 ft.

Hydraulic Gradient 0.04 - Assuming equal to overland surface slope

Figure S4 Decay curves of P-22 and somatic phage in surface water samples from L4, L5, and L6.

Reference

Bradford, S. A., J. Simunek, et al. (2003). "Modeling colloid attachment, straining, and exclusion in saturated porous media." Environmental Science & Technology 37(10): 2242-2250.

Doyle, M. P. (2001). Predictive Modeling and Risk Assessment. Food Microbiology: Fundamentals and Frontiers. Washington, D.C., ASM Press.

Eisenberg, J. N. (2006). "Application of a Dynamic Model to Assess Microbial Health Risks Associated with Beneficial Uses of Biosolids." WERF: 1-124.

Evanylo, G. K. (2009). Agricultural Land Application of Biosolids in Virginia: Regulations, Retreived from http://pubs.ext.vt.edu/452/452-302/452-302.html.

Faulkner, B. R., W. G. Lyon, et al. (2002). Predicting attenuation of viruses during percolation in soils. Cincinnati, OH 45268, National Risk Management Research Laboratory, Office of Research and Development, U.S. Environmental Protection Agency.

Galada, H., P. L. Gurian, et al. (2011). Site Specific Risk Assessment Tools for Land Applied Biosolids, Water Environment Research Foundation. SRSK3R08.

Haas, C. N., J. B. Rose, et al. (1999). Quantitative Microbial Risk Assessment. New York, NY., John Wiley & Sons, Inc.

Juneja, V. K. and H. Marks (2001). Discussion of Non-linear survival curves. Presented at Society of Risk Analysis Workshop, Decision Support Tools for Microbial Risk Assessment, Annual Meeting, Seattle, WA.

Juneja, V. K. and H. M. Marks (2003). "Mathematical description of non-linear survival curves of Listeria monocytogenes as determined in a beef gravy model system at 57.5 to 65 癈." Innovative Food Science & Emerging Technologies 4(3): 307-317.

Kingsley, D. H., D. R. Hollinian, et al. (2007). "Inactivation of a norovirus by high-pressure processing." Applied and Environmental Microbiology 73(2): 581-585.

Kumar, A., F. J. Simmons, et al. (2010). "A case study of a wet weather-friven fate of biosolids-associated contaminants (in preparation)."

Ravi, V. and J. R. Williams (1998). "Estimation of infiltration rate in the vadose zone: Compilation of simple mathematical models." Volume I(EPA/600/R-97/128a).

Teng, J., A. Kumar, et al. (2011). "Determination of Critical Rainfall Events for Quantitative Microbial Risk Assessment of Land-Applied Soil Amendments (Accepted)." Journal of Hydrologic Engineering.

Tufenkji, N. (2007). "Modeling microbial transport in porous media: Traditional approaches and recent developments." Advances in Water Resources 30(6-7): 1455-1469.

Tye, B.-K., R. K. Chan, et al. (1974). "Packaging of an oversize transducing genome by Salmonella phage P22." Journal of Molecular Biology 85(4): 485-490.

Wong, K., T. Harrigan, et al. (2010). "Evaluation of leaching and ponding of viral contaminants following land application of biosolids on sandy-loam soil (in preparation)."

Zhang, H., B. Olson, M.S. Olson, P.L. Gurian, C. Harte, J. Teng. (2009). Extending the Risk Assessment Framework for Pathogens in Biosolids. WEF Residuals and Biosolids Conference. Portland, OR.