Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Mar 28.
Published in final edited form as: Water Resour Res. 2021 Feb 22;57(3):1–11. doi: 10.1029/2020wr028533

Rainfall Washoff of Spores From Concrete and Asphalt Surfaces

Anne M Mikelonis 1, M W Calfee 1, Sang Don Lee 1, Abderrahmane Touati 2, Katherine Ratliff 1
PMCID: PMC8959000  NIHMSID: NIHMS1768700  PMID: 35350225

Abstract

After a biological terrorist attack, understanding the migration of agents such as Bacillus anthracis is critical due to their deadly nature. This is important in urban settings with higher likelihood of human exposure and a large fraction of impervious materials contributing to pollutant washoff. The study goals were to understand the removal of spores from urban surfaces under different rainfall conditions, to compare washoff of two B. anthracis surrogate spores, and to compare two empirical fits for the first flush of spores from small areas. Concrete and asphalt were inoculated with either Bacillus atrophaeus or Bacillus thuringiensis kurstaki spores and exposed to simulated rainfall. The study assessed goodness-of-fit for the Storm Water Management Model (SWMM)’s exponential washoff function compared to an alternative two-stage exponential function. The highest average washoff of spores was 15% for an hour-long experiment. Spore washoff was not significantly different for the two spore types, but there were significant differences in washoff from asphalt versus concrete with more occurring from asphalt. Average kinetic energy of the storm event impacted washoff from asphalt, but not concrete. The two-stage function had a better goodness-of-fit than the SWMM exponential function. As such, emergency responders should be aware that the spread of contamination is impacted by the droplet characteristics of the storm event and the surface material type in the contaminated area; modelers should be aware that different data-fitting approaches may be more appropriate for first-flush calculations of small washoff areas than those used for continuous long-term simulation of large subcatchments.

1. Introduction

The fate and transport of pollutants in urban settings is dictated by numerous complex environmental processes. Urban microclimates, short pollutant travel times, and intricate flow routing around buildings make the identification of migration pathways and the quantification of contamination on surfaces and in storm water runoff challenging. Following a wide-area pathogenic biological release, such as of Bacillus anthracis (the causative agent of anthrax), consequence management activities including sampling, mitigation, decontamination, and waste management may last for many months (D’Amelio et al., 2015; Sinclair et al., 2008). Response strategies and resource management could benefit from a better understanding of the impact of rainfall on pollutant transport, but reference studies describing washoff do not exist for biothreat agents (Mikelonis et al., 2018).

Washoff is the process of removing constituents from surfaces during a period of water runoff. It is a function of many factors including rainfall intensity and/or runoff volumes or rates (Rossman & Huber, 2016; Vaze & Chiew, 2003). For the case of particles, washoff has been mechanistically evaluated through the lens of sediment transport theory, where erosion is related to a critical shear stress that can initiate movement of the particle (Vanoni, 1975). However, several aspects of urban washoff are incongruent with the bulk of work developed for sediment transport. Urban overland flow is very shallow compared to flow in rivers, impervious materials represent a large fraction of surfaces, and rainfall energy in addition to the movement of overland flow can cause particle detachment and motion.

Rainfall simulators have been used for decades as the primary means to study pollutant washoff both in the laboratory and in field settings (Egodawatta et al., 2007; Sartor & Boyd, 1972; Sartor et al., 1974; Vaze & Chiew, 2003). Comprehensive reviews of rainfall simulator design are available (Bowyer-Bower & Burt, 1989; Hall, 1970). The early devices were developed for agricultural plot studies by the United States Soil Conservation Services that directed water upwards approximately 10-feet. The devices achieved intensity values that mimicked naturally occurring rain events, but the droplet sizes and velocities were not realistic of natural rainstorms and potentially underestimated erosivity due to unrepresentative kinetic energies of the simulated storms. Since rainfall intensity is an inherent bulk parameter, rainfall droplet size and velocity must be further characterized to better understand the mechanisms of particle detachment and washoff. Storm energy relationships and rain droplet size frequency distributions have been studied in detail since the 1940s for natural rainfall (Laws, 1941; Laws & Parsons, 1943). In general, raindrops have median drop size values of less than 4 mm and corresponding terminal velocities of less than 9 m/s (Van Dijk et al., 2002).

Rainfall drives washoff processes, but globally rainfall quantities and patterns vary substantially by climatic region. In order to study washoff caused by realistic rainfall events applicable to many areas, a rainfall simulator must have a large operating range. For perspective, according to the United States National Oceanic and Atmospheric Administration’s (NOAA’s) Atlas 14 program data set, 1-h storm events for most locations in the continental United States (CONUS) are less than 50 mm/h for a 2-year return frequency. The largest rain event that most places in the CONUS will experience within a 100-year period is approximately 100 mm/h. (This excludes coastal regions in the southeast and south-central United States/great plains which may experience 1-h storm events over 150 mm/h.) Very high rainfall values are also possible over very short periods of time (NOAA, 2017). Due to this wide variability, predicting washoff amounts through modeling is attractive to aid decision making, but historically storm water quality modeling has been implemented less frequently than water quantity modeling (Chen & Adams, 2006; Obropta & Kardos, 2007). Representative washoff coefficients, relative to other model parameters, have been shown to be in the “topfive” important parameters for generating accurate predictions of water quality along with decay, continuous loss, suspension, and erosion coefficients (Mannina & Viviani, 2010). Despite the recognized relative importance of washoff in simulations and decision making, field and laboratory studies to improve parameterizations are laborious, so opportunities to assess these coefficients through empirical data fitting are rare and limited in scope for routine contaminants and nonexistent for biological agents of concern, such as B. anthracis spores (Butcher, 2003; Egodawatta et al., 2007; Grottker, 1987; Millar, 1999; Sartor et al., 1974). This paper aims to fill this gap by better understanding the rainfall removal of spores from urban surfaces under different rainfall intensities/kinetic energies, to compare washoff of two commonly used B. anthracis surrogate spore types, and to compare two empirical fitting strategies for the first flush of spores from small areas such as parking lots, sidewalks, driveways, and short stretches of roadways.

2. Materials and Methods

2.1. Preparation of Impervious Surfaces

The material coupons used in this research were 35.56-cm (length) × 35.56-cm (width) × 2.54-cm (height) and prepared in-house. Concrete coupons were produced in custom fabricated molds using QUIKRETE sand/topping mix (Product # 1103, The QUIKRETE Companies, Atlanta, GA) according to manufacturer specifications. The coupons were allowed to cure for at least 5 days at ambient indoor conditions covered by plastic. Concrete coupons were used to represent sidewalks and driveways. Asphalt coupons (used to represent roadways and parking lots) were prepared using pills from the North Carolina Department of Transportation Asphalt Laboratory of the Materials and Tests Unit (Raleigh, NC). Multiple cylindrical pills were heated to approximately 138°C for 3–5 h, homogenized (Mud Monster Model #G06160, Goldblatt Industries, Saddle Brook, NJ and 9-Amp 1/2-in Keyed Corded Drill Model # DW130V, DEWALT, Baltimore, MD), and proportioned into custom fabricated coupon molds by a mass that corresponded to the same average density as the cylindrical pills. The asphalt was then pressed using a hydraulic Arbor press (Model LP-500, Devin Mfg. Inc., Arcade, NY) and 1,814 kg of applied pressure.

Meter dosed inhalers (e.g., devices used for asthma medication) loaded with either dry spores of Bacillus atrophaeus (Bg) (ATCC 9372; formerly Bacillus subtilis var. niger and subsequently “Bacillus globigii”) or Bacillus thuringiensis kurstaki (Btk) (ATCC #33679) from a United States Army Laboratory, Dugway Proving Ground were used in these experiments. Both Bg and Btk are biosafety level 1 microorganisms and commonly used surrogates for B. anthracis (Greenberg et al., 2010; Tufts et al., 2014). Spores were loaded onto the coupons as described in Calfee et al. (2013) using a stainless-steel pyramid (Figure 1a). For each experimental trial, three test coupons (concrete/asphalt) and three positive control coupons (stainless steel) were inoculated at target levels of 107 Colony Forming Units (CFUs)/coupon, which is in line with the expected maximum equilibrium concentrations for a surface contamination event. After inoculation with spores, the test coupons were mounted in a custom fabricated collection device that had a 2% slope (Figures 1b–1d). The positive control coupons were sampled for spores using a standard wetted wipe protocol (Brown et al., 2007) to determine spore loading following inoculation. The entire surface of the stainless-steel coupon was first wiped horizontally, the wipe folded to prevent loss of spores, and then wiped vertically, folded again, and wiped along a diagonal path. Test coupons were not wipe sampled, rather spores were collected and analyzed in the runoff water after each simulated rain event.

Figure 1.

Figure 1.

Open in a new tab

Preparation of impervious surfaces. (a) Stainless-steel inoculation pyramid, (b) placement of coupon in holder, (c) sealing coupon in place, and (d) front view of coupon holder.

2.2. Rainfall Simulations

The rainfall simulator used in this work had the ability to simulate 15–115 mm/h rain events (Figure 2). To simulate this range of rainfall rates, the simulator consisted of a manifold with five exchangeable misting nozzles mounted at the top of a 8 m polyvinyl chloride frame with an approximately 1.5 × 1.5 m footprint. The sides of the simulator were enclosed by heavy plastic to prevent water spraying outside the containment area. The simulator was built with direct plumbing to a deionized water source that was used for the simulated rainwater in this work. The nozzles used for the experiments were from Spraying Systems Co. (Glendale Heights, IL). It is important to note that many misting nozzles produce unrealistic droplet size distributions. The nozzles used in this work mimicked droplet size and velocity experienced during natural rainfall events as described by the Gunn–Kinzer curve (Gunn & Kinzer, 1949). More extensive details about the rainfall simulator including parts and methods for characterizing droplet size, velocity, kinetic energy, and spray pattern using a Parsivel2 laser disdrometer (OTT HydroMet, Loveland, CO) are detailed in Mikelonis and Calfee (2018). The test matrix for this research included a total of 15 different impervious material/ spore type/rainfall conditions (Table 1) and Section 3 compares the nozzle performance during the tests included in this study. The goal of this set of tests was to include experiments conducted under a range of rainfall intensities for both spore strains and surface materials combinations and to characterize the rainfall for each experiment by kinetic energy, droplet size, and velocity.

Figure 2.

Figure 2.

Open in a new tab

Rainfall simulator used for experiments.

Table 1.

Test Matrix and Nozzle Performance Characteristics

Test ID Material Spore Nozzle Replicates Average intensity (mm/h) Average kinetic energy (J/m2 h) Average droplet diameter (mm)
A Asphalt Bg HH-14WSQ 3 16 81 0.63
B Asphalt Btk HH-14WSQ 3 15 73 0.63
C Concrete Bg HH-14WSQ 4 16 727 0.63
D Concrete Btk HH-14WSQ 3 16 77 0.62
E Asphalt Bg GG-2.8W 3 29 41 0.6
F Asphalt Btk GG-2.8W 3 28 39 0.60
G Concrete Bg GG-2.8W 3 25 46 0.60
H Concrete Btk GG-2.8W 3 25 44 0.61
I Asphalt Btk HH-30WSQ 3 81 585 0.73
J Concrete Bg HH-1 3 81 767 0.83
K Concrete Btk HH-1 3 91 797 0.82
L Asphalt Bg HH-50WSQ 3 60 746 0.76
M Asphalt Btk HH-50WSQ 2 62 894 0.82
N Concrete Bg HH-50WSQ 3 91 2,156 0.85
O Concrete Btk HH-50WSQ 3 90 2,306 0.98
Open in a new tab

2.3. Runoff Sample Collection

For each test, the rainfall simulator was started, flow checks were performed, and then it was turned off and the coupon holder (containing the inoculated coupon) was placed under the simulator. Sterile 50-mL conical tubes were used to collect runoff water from the outlet of the coupon holder at set timepoints after the commencement of runoff (Figure 1d). The samples were then delivered to an onsite microbiology laboratory with chain of custody for analysis.

2.4. Microbiological Analysis

All samples were manipulated using aseptic techniques, cultured in tryptic soy agar, and plated in triplicate using a spiral plater (Autoplate® Spiral Plating System AP5000, Advanced Instruments LLC, Norwood, MA) followed by incubation at 35°C ± 2°C for 16–19 h. CFUs were enumerated using a QCount® colony counter (Model #530, Advanced Instruments LLC). Samples were reanalyzed by filter-plating larger sample volumes with a 0.45-μm microfunnel (Item # 4804, PALL, Port Washington, NY) when spiral-plated samples yielded fewer than 30 CFU on a single plate. All collection bins, coupon holders, and the aerosol deposition apparatuses were sterilized for 4 h at 200 ppm vaporous hydrogen peroxide prior to experiments. Stainless-steel and concrete coupons were autoclaved on a 121°C gravity cycle. New nonsterilized asphalt coupons were used for each experiment. For quality control of each experiment, sterility swab samples were collected and analyzed for spores from the collection bins, coupon holders, and coupons to ensure cross contamination did not occur between trials.

2.5. Data Analysis

Since the late 1960s, urban storm water quantity and quality have been studied extensively using the United States Environmental Protection Agency’s (U.S. EPA’s) Storm Water Management Model (SWMM) (W. Huber & Roesner, 2013). SWMM includes options to simulate the washoff of pollutants between and during rain events. To this extent, SWMM provides the option of modeling washoff of pollutants from catchments during storm events in three manners: event mean concentrations (EMCs), exponential functions, and rating curves. Since EMCs provide constant runoff concentration throughout the entire simulation and are typically selected based on averaging field-collected local data or estimated for particular contaminants using data collected during the U.S. EPA’s Nationwide Urban Runoff Program (USEPA, 1983), it is not ideal for an application to a rare event, such as a bioterrorist attack. The exponential functions describe a first-order removal process that includes pollutant mass loadings (often referred to as “buildup”) so that washoff ceases when there is no more contaminant remaining and were used in the data analysis portion of this work. Mathematically, the exponential function used for modeling washoff in SWMM is

W(t)=P0e−c1Q(t)c2t (1)

where W(t) is the quantity of constituent remaining on the material surface, P0 is the initial quantity on the surface, c1 and c2 are empirically fitted coefficients, Q(t) is the runoff rate, and t is time. Rating curves and exponential functions differ in only one aspect, the rating curves do not include a starting concentration/ buildup (P0).

The data in this study were also fit to an alternative mathematical function to what is available in SWMM:

f(t)=Ae−k1t+(1−A)e−k2t (2)

where f(t) is the fraction of constituent remaining on the material surface, A, k1, and k2 are empirically fitted coefficients, and t is time. Conceptually, this represents a first-order mass balance approach of two well-mixed reservoirs in parallel with zero new input mass over the course of the washoff. Two reservoirs were integrated into the Alternative Fit to represent the cases of loosely bound spores (fast removal potentially via raindrop impaction) and spores more tightly bound to the surface (slower removal due to shear flow). Physically, A represents a capacity factor or the maximum number of spores on the surface. The k1 coefficient represents how fast spores are initially removed from the coupon and k2 represents the sustained removal of spores from the coupons. This model fit assumes that runoff is relatively constant over a small surface such as a parking lot.

To analyze the collected data, cumulative CFUs for each coupon’s timepoints were calculated by integrating under the curve of the average CFU in each runoff sample using the trapezoidal rule. The number of spores remaining on the coupon at each timepoint was calculated by subtracting the cumulative CFU from the average value of the stainless-steel positive control coupon associated with the test. Finally, to control for differences in starting concentrations (i.e., positive control values) between tests, the calculated number of spores remaining on the coupon at each timepoint was divided by the average positive control value to produce the average fraction of spores remaining on the coupon. The open source statistical software R version 3.6.1 was used for the data analysis. Curve fitting was performed using nonlinear regression with least squares criterion to estimate parameters from each experiment using the nls2 R package’s random search functionality to obtain c1 and c2 for SWMM’s exponential washoff function (Equation 1) (Grothendieck, 2013). The measured runoff rate from each replicate coupon was averaged at each timepoint and used as Q(t). The nls function in R was used for the Alternative Fit (Equation 2). The Akaike information criterion (AIC), a statistical estimator of model prediction error, was used for relative comparison of the SWMM and Alternative Fits (with a lower value indicating better relative fit). Due to a plateau in spore washoff after approximately 15 min of exposure to simulated rain, the data set failed the assumptions of normality and homoscedasticity required for traditional analysis of variance techniques. Nonparametric analysis and significance testing was performed using the nparLD R software package (Noguchi et al., 2012). A value of 0.05 was used to assess significance. The washoff data and R code for curve fitting code and significance testing are in the supporting information of this paper (Data Sets S1 and S2).

3. Results

3.1. Rainfall Simulator Performance

An analysis of rainfall simulator performance was conducted and presented here because different nozzles were used for some of the material–spore combinations. Average rainfall kinetic energy varied for each test from 39 to 2,306 J/(m2 h) (Figure 3a and Table 1). The kinetic energy was primarily controlled by varying rainfall intensity from 16 to 91 mm/h (Figure 3b). The characterization data indicated that droplet size played less of a role because it was held fairly constant for all experiments at an average value of 0.7 mm (Figure 3c); as expected, when droplet size varied between tests the size variation scaled linearly with changes in kinetic energy (Figure 3d). It is important to note that researchers have documented the need to apply correction factors to Parsivel2 data to correct for measurement bias of certain droplet distributions (Angulo-Martínez & Barros, 2015; Raupach & Berne, 2015). Developing correction factors was outside the scope of this work, so the kinetic energy values should be interpreted as relative measurements rather than absolute values.

Figure 3.

Figure 3.

Open in a new tab

Rainfall simulator characterization results. (a) Kinetic energy by nozzle and impervious material, (b) intensity by nozzle and impervious material, (c) droplet diameter by nozzle and impervious material, and (d) relationship between average kinetic energy by droplet size and intensity. Kinetic energy and droplet diameter values obtained using a Parsivel2 disdrometer. Intensity values were obtained by measuring volume accumulated in a bin over a set time interval. Error bars represent standard deviation from multiple tests using the same conditions. Nozzle HH-14WSQ was used in Tests A–D, GG-2.8W in Tests E–H, HH-30WSQ in Test I, HH-1 in Tests J and K, and HH-50WSQ in Tests L–O.

The nozzles used for these experiments performed similarly for the asphalt and concrete test conditions in terms of average rainfall kinetic energy, intensity, and droplet diameter with one exception (Figure 3). The HH-50WSQ nozzle used during concrete experiments produced a larger average intensity and droplet size than when used during the asphalt test. The reason for this is unknown but may be due to slight corrosion of the nozzle altering rainfall characteristics. The removal data collected for these tests do not indicate this difference had an influence on the test results (i.e., concrete spore removal [Tests N & O] was lower than asphalt [Tests L & M] when using HH-50WSQ). The data reveal two clusters of nozzles performance conditions. The GG-2.8W and HH-14WSQ nozzles were used to produce lower intensities/droplet sizes and as a result lower kinetic energy values with an average of 160 J/(m2 h) across test conditions. Conversely, the HH-1, HH-30WSQ, and HH-50WSQ produced higher kinetic energy values with an average of 983 J/(m2 h) across test conditions.

3.2. Spore Washoff

The spore washoff data followed a general trend of exponential decay with an initial large flush of spores occurring within the first 15 min of rainfall followed by a slower removal plateau for the remainder of the hour (Figure 4). Washoff did not significantly vary by spore type (p-value = 0.33), but it was significantly different for asphalt versus concrete (p-value = 0.034). For the asphalt coupons, increased removal between tests was directly related to greater average kinetic energy values of the simulated rain. The spore removal from the concrete coupons was within the variation of tests regardless of changes in rainfall intensity/kinetic energy. Figure S1 graphs coupon runoff rate over the course of the experiments according to spore and material type. In these experiments, spore washoff did not correlate with runoff rate (i.e., higher washoff did not necessarily occur with higher runoff rate). The mathematical functions described by Equation 2 (SWMM Fit) and Equation 3 (Alternative Fit) were used to fit the experimental data to assess the best approach for estimating spore removal. An example of one test is provided to demonstrate how the SWMM Fit approximates the data as compared to the Alternative Fit (Figure 5a). The Alternative Fit consistently produced smaller standard errors than the SWMM Fit and had an average AIC of −100 (σ = 26) whereas the SWMM Fit AIC averaged −41 (σ = 15). The SWMM Fit both underpredicted and overpredicted the measured data, depending on the test (Figure 5b). Graphs for all tests with both the SWMM Fit and Alternative Fit are in the supporting information (Figure S2).

Figure 4.

Figure 4.

Open in a new tab

Washoff results by material and spore type. The measured rainfall intensity, kinetic energy, and average droplet size for each test are summarized in Table 1. Error bars represent standard error of replicate coupons.

Figure 5.

Figure 5.

Open in a new tab

(a) Average washoff curve for Bg spores from asphalt coupons (Test L) with the SWMM Fit and Alternative Fits. (b) Performance of fitting for all tests. Bg, Bacillus atrophaeus; SWMM, Storm Water Management Model.

The coefficients and AIC values generated for all the experiments are summarized numerically in Tables S1 and S2 in the supporting information. For the SWMM Fit, c1 was found to have a median value of 0.01 (σ = 0.006) and c2 was found to have a median value of 0.57 (σ = 0.04) across all rainfall tests. The median values did not substantially vary by material type (concrete: c1 = 0.01, c2 = 0.57; asphalt c1 = 0.01, c2 = 0.55). For the Alternative Fit, the coefficient “A” represents the removal capacity factor. The median value of A was 0.02 (σ = 0.03) for all rainfall tests. The k1 coefficient represents how fast spores are initially removed from the coupon (its median value was 19 [σ = 72] across all rainfall tests) and k2 represents the sustained removal of spores from the coupons (its median value was 0.02 [σ = 0.03] across all rainfall tests). The median values for k1 of the Alternative Fit varied substantially by material type with asphalt having a larger initial removal than concrete (concrete: A = 0.02, k1 = 8, and k2 = 0.01; asphalt: A = 0.02, k1 = 31, and k2 = 0.06).

4. Discussion

Previous research found that the total amount of solids washed off streets was dependent on rainfall intensity and slope of the surface roughness (Muthusamy et al., 2018; Sartor & Boyd, 1972). Slope was held constant during these experiments, but there was not an observed relationship between rainfall intensity or runoff rate and spore removal for either material type. This may have been due to the low slope and small size of the coupons failing to generate a change in shear velocity with higher rainfall amounts. Instead, there was a correlation between kinetic energy and increased spore removal from asphalt coupons and no clear pattern for the concrete coupons. Others have also observed soil erosion differences related to kinetic energy and droplet size (Bubenzer & Jones, 1971).

In this study, more spores were removed from the asphalt than the concrete coupons. This is consistent with data patterns in the SWMM 4 User Manual for street solids removed from concrete and asphalt roadways at the same rainfall rate (W. C. Huber et al., 1988; Sartor & Boyd, 1972), but inconsistent with a more recent study where concrete consistently experienced more washoff of particles than asphalt (Zhao et al., 2018). Several material properties may have contributed to significantly less washoff of spores from concrete than asphalt in this study. First, it is hypothesized that due to concrete’s higher porosity compared to asphalt (on average 13 times more water by mass for the coupons used in this study) and the small size of the spores (approximately 1 μm equivalent spherical diameter) that rainfall may have initially transported spores into the concrete coupon’s pores and crevasses leading to lower washoff amounts. The surface charge of asphalt and concrete may also be different leading to different adhesion forces between the spores and the materials and contributing to different removals. Once spores and the material surfaces are covered in natural organic matter these surface charge differences may be mitigated (Mikelonis et al., 2020). Finally, a third hypothesis is that pH changes due to hydroxyl groups on concrete could contribute to inactivation of spores at very basic pH values on the surface/in the pores and lead to reduced washoff from concrete versus asphalt. Studies addressing these mechanistic hypotheses were not found and warrant further investigation.

Values of washoff coefficients in the literature vary substantially. For the SWMM Fit, the general consensus is that c1 varies 3–5 orders of magnitude and that c2 is typically between 1.1 and 2.6 in sediment transport research (Vanoni, 1975). In one study, an urban catchment was modeled in SWMM and the washoff coefficients were used as calibration parameters for total suspended solids removal and found to average 0.18 for c1 and 2.35 for c2 (Di Modugno et al., 2015). During a road salt washoff study, c1 values varied between 0.18 and 0.25 and c2 values ranged from 2.2 to 2.6 (Trenouth et al., 2015). Another paper cited values from different studies ranging from 0.13 to 500 for c1 and 0.9 to 2.35 for c2, with the higher end of the range being for total phosphorus (Chow et al., 2012). Considering so much variability in the literature and the lack of variables such as traffic and wind in this controlled laboratory study, the median values from this work of 0.01 for c1 and 0.57 for c2 are consistent with what others have found for more traditional urban pollutants. The spores used are approximately 1 μm in diameter, so the lower value for c2 is consistent with the soluble solids range of 0–1 rather than settleable solids (Mannina & Viviani, 2010).

The SWMM Fit struggled balancing the rapid spore washoff with the gradual plateau and therefore failed to replicate the shape of the observed spore washoff curve. Due to the structure of the coefficients in the exponent, it was computationally challenging to fit using standard nonlinear curve fitting approaches in R and a random search feature had to be implemented. This led to some individual tests having very poor fits. In the SWMM Fit, Q(t) is intentionally raised by c2 within the exponent for flexibility in modeling a diverse set of physical scenarios. The presence of this complicated exponential structure allows the model to simulate peaks of pollution during the middle of a storm, not just at the start (Rossman & Huber, 2016). While this mathematical flexibility is useful for a wider range of scenarios, we observed for simple pollutant decay it can present challenges for fitting data. On the other hand, since the Alternative Fit employed a two-stage exponential to describe removal it was able to capture the temporal variability of the rapid first flush of spores and the slower plateaued removal of spores. This reinforces the work of several others that have proposed “capacity factors” to better model washoff (Egodawatta et al., 2007; Muthusamy et al., 2018). The results were summarized according to the median coefficient values because the Alternative Fit experienced two instances where k1 was substantially higher than for the rest of the tests. Others have also observed convergence to optimum values that diverge substantially from the mean or have no clear physical significance (Al Ali et al., 2016). Additionally, the coefficient structure is computationally easier to fit using statistical software packages than the nested exponentials in the SWMM Fit and likely contributes to its higher AIC value.

The spore removal in these experiments only represent the “first flush” of a pollutant or the instance where concentration is higher at the start of the storm event and then decreases rapidly. Within natural catchments, contaminants may form secondary peaks during the middle of rainfall events. In these instances, the SWMM Fit’s coefficient structure has the capacity to accommodate a decrease followed by an increase while the Alternative Fit does not. The SWMM Fit’s value should not be discounted within a stormwater model based on these experimental results Without both empirically fitted coefficients, it is only possible to model decreasing concentrations of constituent as the remaining buildup P0 also decreases over time. These data merely indicate that for localized areas that experience washoff that decreases rapidly the SWMM Fit may not be accurate due to its complicated exponential structure causing issues for curve fitting routines used to generate coefficients. To better understand implications to overall mass transport, both the Alternative and the SWMM Fits would need to be tested and compared within a calibrated stormwater model for a range of rainfall events and compared to field data. Our research program plans to test this in conjunction with future field campaigns. As such, the Alternative Fit is proposed only for use with modeling the first flush of pollution. Additionally, the Alternative Fit, as used here, is only dependent on time, not runoff rate or rainfall intensity. This relationship was observed for the spores on small coupons. However, in larger, more complex natural catchments, it is expected that additional parameters will need to be incorporated into the coefficients of the Alternative Fit as factors such as flow length and area will vary and rainfall characteristics will have a greater impact. At this point, it is recommended the Alternative Fit be considered for simple reaches only such as parking lots or small portions of roads.

5. Conclusions

Spore washoff from concrete or asphalt did not vary significantly by spore type. The median washoff coefficients for the SWMM Fit were within expected ranges from sediment transport. Values of 0.1 for c1 and 0.57 for c2 represent the best fit from this study’s data to use in storm water models for impervious land uses in SWMM. The rainfall results highlight the important role of kinetic energy in spore washoff from asphalt surfaces. As such, emergency responders should be aware that the spread of contamination is impacted by the droplet characteristics of the storm event (not just the intensity) and the distribution of surface materials in the contaminated area. For these constant rainfall intensity experiments, the highest average washoff of spores from replicate coupons was 15%. Future rainfall simulator work is planned to better understand the relationship of spore washoff to variable intensity rain events and drying cycles. The Alternative Fit captured the shape of the washoff data better and had a lower AIC than the SWMM Fit, indicating it would be beneficial to build capabilities into SWMM to model spore removal using a two-stage exponential decay function. Further research into how best to incorporate runoff rate or rainfall intensity into the Alternative Fit in larger scale field experiments would also be beneficial. With this capability, future studies should focus on a sensitivity analysis assessing the impact of using the Alternative versus SWMM Fit within several calibrated storm water models.

Supplementary Material

Supporting Information S1
Dataset S1

Key Points:

  • Washoff of Bacillus atrophaeus and Bacillus thuringiensis kurstaki spores was not significantly different

  • Greater spore washoff occurred from asphalt than concrete coupons

  • Washoff results were fit to EPA’s SWMM exponential washoff function and compared to an alternative two-stage exponential removal function

Acknowledgments

The authors would like to thank the laboratory staff at Jacobs Technology Inc., including Jason Colon and Chris Fuller for executing and overseeing rainfall simulation experiments and Lee Brush, Brian Sechrest, Eric Morris, Jonathan Sawyer, and Steve Terll for their data collection and rainfall simulator construction and maintenance contributions. Denise Aslett, Kathleen Butler, Ahmed Abdel-Hady, Brian Ford, and Mariela Monge of Jacobs Technology Inc. are acknowledged for biological enumeration. Katherine Ratliff developed the python script used throughout this work to analyze Parsivel2 data. Thank you to Joan Bursey and Ramona Sherman for providing the EPA quality assurance review and Michael Tryby and Jeff Szabo for providing EPA technical reviews. The U.S. Environmental Protection Agency (EPA) through its Office of Research and Development directed the research described herein under EP-C-15–008 with Jacobs Technology Inc. It has been reviewed by the Agency but does not necessarily reflect the Agency’s views. No official endorsement should be inferred. EPA does not endorse the purchase or sale of any commercial products or services.

Footnotes

Data Availability Statement

The data set for this paper is available at the following doi.org/10.23719/1518925.

References

  1. Al Ali S, Bonhomme C, & Chebbo G. (2016). Evaluation of the performance and the predictive capacity of build-up and wash-off models on different temporal scales. Water, 8(8), 312. [Google Scholar]
  2. Angulo-Martínez M, & Barros A. (2015). Measurement uncertainty in rainfall kinetic energy and intensity relationships for soil erosion studies: An evaluation using PARSIVEL disdrometers in the Southern Appalachian Mountains. Geomorphology, 228, 28–40. [Google Scholar]
  3. Bowyer-Bower T, & Burt T. (1989). Rainfall simulators for investigating soil response to rainfall. Soil Technology, 2(1), 1–16. [Google Scholar]
  4. Brown GS, Betty RG, Brockmann JE, Lucero DA, Souza CA, Walsh KS, et al. (2007). Evaluation of a wipe surface sample method for collection of Bacillus spores from nonporous surfaces. Applied and Environmental Microbiology, 73(3), 706–710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bubenzer D, & Jones GBA Jr. (1971). Drop size and impact velocity effects on the detachment of soils under simulated rainfall. Transactions of the ASAE, 14(4), 625–628. [Google Scholar]
  6. Butcher JB (2003). Buildup, washoff, and event mean concentrations. Journal of the American Water Resources Association, 39, 1521–1528. [Google Scholar]
  7. Calfee MW, Lee SD, & Ryan SP (2013). A rapid and repeatable method to deposit bioaerosols on material surfaces. Journal of Microbiological Methods, 92(3), 375–380. [DOI] [PubMed] [Google Scholar]
  8. Chen J, & Adams BJ (2006). Analytical urban storm water quality models based on pollutant buildup and washoff processes. Journal of Environmental Engineering, 132(10), 1314–1330. [Google Scholar]
  9. Chow M, Yusop Z, & Toriman ME (2012). Modelling runoff quantity and quality in tropical urban catchments using Storm Water Management Model. International journal of Environmental Science and Technology, 9(4), 737–748. [Google Scholar]
  10. D’Amelio E, Gentile B, Lista F, & D’Amelio R. (2015). Historical evolution of human anthrax from occupational disease to potentially global threat as bioweapon. Environment International, 85, 133–146. [DOI] [PubMed] [Google Scholar]
  11. Di Modugno M, Gioia A, Gorgoglione A, Iacobellis V, La Forgia G, Piccinni AF, & Ranieri E. (2015). Build-up/wash-off monitoring and assessment for sustainable management of first flush in an urban area. Sustainability, 7(5), 5050–5070. [Google Scholar]
  12. Egodawatta P, Thomas E, & Goonetilleke A. (2007). Mathematical interpretation of pollutant wash-off from urban road surfaces using simulated rainfall. Water Research, 41(13), 3025–3031. [DOI] [PubMed] [Google Scholar]
  13. Greenberg DL, Busch JD, Keim P, & Wagner DM (2010). Identifying experimental surrogates for Bacillus anthracis spores: A review. Investigative Genetics, 1(1), 1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Grothendieck G. (2013). Non-linear regression with brute force. R package version 0.2. cran.r-project.org. https://cran.r-project.org/web/packages/nls2/nls2.pdf
  15. Grottker M. (1987). Runoff quality from a street with medium traffic loading. The Science of the Total Environment, 59, 457–466. [Google Scholar]
  16. Gunn R, & Kinzer GD (1949). The terminal velocity of fall for water droplets in stagnant air. Journal of Meteorology, 6(4), 243–248. [Google Scholar]
  17. Hall M. (1970). A critique of methods of simulating rainfall. Water Resources Research, 6(4), 1104–1114. [Google Scholar]
  18. Huber W, & Roesner L. (2013). The history and evolution of the EPA SWMM. Paper presented at Fifty Years of Watershed Modeling Past, Present and Future, ECI Symposium Series, Boulder, CO, ECI Symposium Series September 24–26, 2012. http://dc.engconfintl.org/watershed/29 [Google Scholar]
  19. Huber WC, Dickinson RE, & Barnwell TO Jr. (1988). Storm Water Management Model; version 4 (EPA/600/3–88/001a). Athens, GA: Environmental Protection Agency. [Google Scholar]
  20. Laws JO (1941). Measurements of the fall-velocity of water-drops and raindrops. Eos, Transactions American Geophysical Union, 22(3), 709–721. [Google Scholar]
  21. Laws JO, & Parsons DA (1943). The relation of raindrop-size to intensity. Eos, Transactions American Geophysical Union, 24(2), 452–460. [Google Scholar]
  22. Mannina G, & Viviani G. (2010). An urban drainage stormwater quality model: Model development and uncertainty quantification. Journal of Hydrology, 381(3–4), 248–265. [Google Scholar]
  23. Mikelonis A, Boe T, Calfee W, & Lee SD (2018). Urban fate and transport modeling of contaminants: The unique needs of emergency response and the potential for adapting existing models. Journal of Water Management Modeling, CHIJWMM 2018:C447. [Google Scholar]
  24. Mikelonis A, & Calfee MW (2018). Development of laboratory equipment and protocols for the assessment of rain, water wash-down, and channelized flow for removal of spores on urban surfaces (EPA/.600/R-18/322). Washington, DC: U.S. Environmental Protection Agency. [Google Scholar]
  25. Mikelonis A, Ratliff K, & Youn S. (2020). Laboratory results and mathematical modeling of spore surface interactions in stormwater runoff. Journal of Contaminant Hydrology, 235, 103707. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Millar RG (1999). Analytical determination of pollutant wash-off parameters. Journal of Environmental Engineering, 125(10), 989–992. [Google Scholar]
  27. Muthusamy M, Tait S, Schellart A, Beg MNA, Carvalho RF, & de Lima JL (2018). Improving understanding of the underlying physical process of sediment wash-off from urban road surfaces. Journal of Hydrology, 557, 426–433. [Google Scholar]
  28. NOAA. (2017). Precipitation frequency data server. Silver Spring, MD: U.S. Department of Commerce. [Google Scholar]
  29. Noguchi K, Gel YR, Brunner E, & Konietschke F. (2012). nparLD: An R software package for the nonparametric analysis of longitudinal data in factorial experiments. Journal of Statistical Software, 50(12), 1–23. 10.18637/jss.v050.i1225317082 [DOI] [Google Scholar]
  30. Obropta CC, & Kardos JS (2007). Review of urban stormwater quality models. Deterministic, Stochastic, and Hybrid Approaches1, 43(3), 1508–1523. [Google Scholar]
  31. Raupach TH, & Berne A. (2015). Correction of raindrop size distributions measured by Parsivel disdrometers, using a two-dimensional video disdrometer as a reference. Atmospheric Measurement Techniques, 8(1), 343. [Google Scholar]
  32. Rossman L, & Huber W. (2016). Storm Water Management Model reference manual volume III—Water quality (EPA/600/R-16/093). Washington, DC: US EPA Office of Research and Development. [Google Scholar]
  33. Sartor JD, & Boyd GB (1972). Water pollution aspects of street surface contaminants (EPA-R2–72-081). Washington, DC: Office of Research and Monitoring, U.S. Environmental Protection Agency. [Google Scholar]
  34. Sartor JD, Boyd GB, & Agardy FJ (1974). Water pollution aspects of street surface contaminants. Journal (Water Pollution Control Federation), 46, 458–467. [PubMed] [Google Scholar]
  35. Sinclair R, Boone SA, Greenberg D, Keim P, & Gerba CP (2008). Persistence of category A select agents in the environment. Applied and Environmental Microbiology, 74(3), 555–563. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Trenouth WR, Gharabaghi B, & Perera N. (2015). Road salt application planning tool for winter de-icing operations. Journal of Hydrology, 524, 401–410. [Google Scholar]
  37. Tufts JA, Calfee MW, Lee SD, & Ryan SP (2014). Bacillus thuringiensis as a surrogate for Bacillus anthracis in aerosol research. World Journal of Microbiology and Biotechnology, 30(5), 1453–1461. [DOI] [PubMed] [Google Scholar]
  38. USEPA. (1983). In Division WP (Ed.), Results of the Nationwide Urban Runoff Program (Accession Number PB84–185552). Washington, DC: NTIS. [Google Scholar]
  39. Van Dijk A, Bruijnzeel L, & Rosewell C. (2002). Rainfall intensity–kinetic energy relationships: A critical literature appraisal. Journal of Hydrology, 261(1–4), 1–23. [Google Scholar]
  40. Vanoni VA (1975). Sedimentation engineering. Manuals and reports on engineering practice (No. 54). American Society of Civil Engineers. [Google Scholar]
  41. Vaze J, & Chiew FH (2003). Study of pollutant washoff from small impervious experimental plots. Water Resources Research, 39(6), 1160. 10.1029/2002WR001786 [DOI] [Google Scholar]
  42. Zhao H, Jiang Q, Xie W, Li X, & Yin C. (2018). Role of urban surface roughness in road-deposited sediment build-up and wash-off. Journal of Hydrology, 560, 75–85. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information S1
NIHMS1768700-supplement-Supporting_Information_S1.pdf (409.5KB, pdf)
Dataset S1
NIHMS1768700-supplement-Dataset_S1.csv (85.1KB, csv)

RESOURCES