ent 374 (2007) 91–101www.elsevier.com/locate/scitotenv

Science of the Total Environm

Ecological risk assessment of lead contamination at rifle and pistolranges using techniques to account for site characteristics

Joseph R. Bennett a,b,⁎, Claire A. Kaufman a, Iris Koch a, Jim Sova a, Ken J. Reimer a,⁎

a Environmental Sciences Group, Royal Military College of Canada, PO Box 17000 Stn Forces, Kingston, Ontario, Canada K7K 7B4b Centre for Applied Conservation Research, Department of Forest Sciences, University of British Columbia,

2424 Main Mall, Vancouver, British Columbia, Canada V6T 1Z4

Received 28 September 2006; received in revised form 1 December 2006; accepted 14 December 2006Available online 24 January 2007

Abstract

Spent ammunition at outdoor rifle and pistol (RP) firing ranges creates a characteristic pattern of contamination, whereby smallareas surrounding backstop berms exhibit extremely high soil lead (Pb) concentrations. We characterized sources, pathways anduncertainty in contaminant accumulation for receptors at two RP ranges in southeastern Ontario. Based on these results, weperformed risk calculations using kriging to estimate risk across “worst-case” species foraging ranges. Range-normalized hazardquotients (RNHQ) were then calculated to estimate risk across each receptor's foraging range. Monte Carlo simulation was used toprovide +2 standard deviation (SD) risk estimates, in order to account for uncertainty in risk parameters. The American robin wasfound to be most at risk (RNHQ=4.10; +2SD=9.24), followed by the short-tailed shrew (RNHQ=0.113; +2SD=0.243) and theeastern cottontail (RNHQ=0.109; +2SD=0.703). Elevated risk for the American robin and short-tailed shrew was due to theirsmall foraging ranges and habit of eating earthworms, which bioaccumulate Pb. Elevated risk for the eastern cottontail was due tovegetation accumulating Pb to levels that were considerably higher than conventional bioaccumulation models would indicate. Theresults of this study clearly emphasize the importance of specific characteristics of RP ranges, such as contamination patterns, dustaccumulation on plant tissues, and contaminant bioaccessibility. These characteristics should be accounted for, not only inperforming risk assessments, but also in choosing remedial options and in routine management practices.© 2006 Elsevier B.V. All rights reserved.

Keywords: Risk assessment; Firing range; Lead; Bioaccumulation; Bioaccessibility; Foraging range

⁎ Corresponding authors. Bennett is to be contacted at Centre forApplied Conservation Research, Department of Forest Sciences,University of British Columbia, 2424 Main Mall, Vancouver, BritishColumbia, Canada V6T 1Z4. Tel.: +1 604 822 1256; fax: +1 604 8229102. Reimer, Tel.: +1 613 530 6000x6161; fax: +1 613 542 9489.

E-mail addresses: jrb5@interchange.ubc.ca (J.R. Bennett),ken.reimer@rmc.ca (K.J. Reimer).

0048-9697/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.scitotenv.2006.12.040

1. Introduction

More than 12000 outdoor small arms firing rangesexist in the United States alone (ITRC, 2003; USEPA,2001). Spent ammunition on these ranges can be a sourceof heavy metal contaminants, of which lead (Pb) hasbeen recognized as the most significant (ITRC, 2003).The potential human health risks associated with bothindoor and outdoor firing ranges have been recognizedfor some time (Tripathi et al., 1991). However, the

92 J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

ecological implications of contamination at small-armsranges have remained relatively unexplored.

Analyses of ecological exposure to Pb at trap andskeet (i.e., shotgun) ranges have been conducted, andhave found potential risk to receptors. Vyas et al. (2000)found elevated Pb concentrations in passerine birdsfrequenting a trap and skeet (TS) range, while Stansleyand Roscoe (1996) found elevated Pb concentrations aswell as evidence of lead toxicity in frogs andmice at a TSrange. Risk calculations performed by Peddicord andLaKind (2000) indicated that food ingestion was themost significant route of exposure for passerine birds atTS ranges, while ingestion of spent shot was the mostsignificant potential route of exposure for mammals.Kendall et al. (1996) analyzed the risk of lead shotgunpellets from hunting and TS range use to regionalpopulations of upland game birds and raptors, andconcluded that while insufficient evidence existed forpopulation-wide effects, an exploration of possiblealternatives to lead shot was merited.

Although the chemical composition of ammunitionused at TS ranges and rifle and pistol (RP) ranges isbroadly similar, the patterns of contamination producedare different (USEPA, 2003). Shotgun shot fired at TSranges impacts in a diffuse pattern with relatively lowvelocity, while bullets fired at RP ranges impact at ahigher velocity in a more discrete area, leading to bullet“pockets” which are denuded of vegetation and exhibithigh metal concentrations. Thus, factors such as projec-tile fragmentation and dust created by impacts are likelyto be more important at RP ranges.

Investigations at RP ranges (e.g. Astrup et al., 1999;Rooney et al., 1999; Chen and Daroub, 2002; Cao et al.,2003) have found that contamination at these sites isdistinguished by the following characteristics: 1) con-tamination composed largely of particulate (and generallyinsoluble) forms of Pb, at concentrations ranging fromnear background to N90000mg kg−1 at the bullet pocketsof backstop berms; 2) heterogeneous contamination pat-terns, which are difficult to measure accurately due to theparticulate nature of the contaminant; 3) highly variablesize of contaminant particles, ranging from relatively inertand immobile shotgun slugs to potentially soluble andmobile microscopic fragments; and 4) relatively smallaffected areas, which typically represent only a portion ofthe foraging ranges of receptor organisms. RP ranges areoften located near areas of secondary succession, whichmay be home to a large number of potential ecologicalreceptors. There can also be considerable dust producedby impacts at bullet pockets, a phenomenon that has beenlittle recognized as an ecologically significant contami-nant transport mechanism at RP ranges.

Because of the unique characteristics of contaminantpatterns at RP ranges, conventional techniques used inecological risk assessment (ERA) (e.g. USEPA, 1991;Suter et al., 2000) must be modified. In particular,contaminant intake models need to reflect a small, veryhighly contaminated area with potentially significantaerial fallout. We therefore undertook a thoroughcharacterization of pathways for four common receptorsat RP ranges in eastern North America. We sought toanswer the following questions: 1) What are the relevantexposure pathways; 2) How bioavailable is the Pbcontamination at small-arms ranges; and 3) Doescontamination at the small but highly impacted areasat RP ranges represent an ecological risk to mammalianand avian receptors? The methodology presented maybe used, with appropriate adjustments, as a template forecological risk assessments at small-arms ranges ineastern North America and elsewhere.

2. Materials and methods

2.1. Study sites

Three RP ranges were considered in this study, alllocated in southeastern Ontario. The ranges were built tothe same specifications, using material from a commonsource, and are used for training law enforcement officers.Between 20000 and 29000 rounds of a variety of calibersare fired per year at each range. Personnel dischargeweapons from several firing lines along the 100 m lengthof the ranges, into 8–10 bullet pockets along∼4.5 m highbackstop berms. The local bedrock is limestone, with soildepth ranging from b50 cm to N1.5 m; soil pH at all threeranges is circum-neutral (ESG, 2005). The immediatesurroundings of the firing ranges consist of fallow fieldsand immature forest and shrubs.

Firing ranges were designated A, B and C. Riskcalculations were performed for Ranges A and B, whichwere the subject of an environmental assessment in-volving extensive soil sampling. Ranges A and B arelocated side by side at the same site (Fig. 1), while RangeC, which is located∼30 km distant, was used for analysisof contaminant accumulation in receptor organisms. Atthe time of sampling, Ranges A, B and C had beenaccumulating bullets for 3, 4 and 5 years, respectively.

2.2. Receptor organisms

Risk calculations were performed for four species,which were chosen to represent receptors across a rangeof trophic levels and life history strategies, and in theinterest of broader applicability to RP ranges in eastern

Fig. 1. Layout of Firing Ranges A and B, showing outline of berm andsample locations.

93J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

North America. Exposures and hazard quotients (HQ)were calculated for individual organisms; the selectedreceptors could potentially be used as analogues forspecies where individual-level risk (as opposed topopulation-level risk) was deemed important. Thefollowing species were chosen:

2.2.1. Eastern cottontail (Sylvilagus floridanus)The eastern cottontail is a common resident in farm-

lands, fields and hedge rows in eastern and central NorthAmerica (Chapman et al., 1980). These areas are anal-ogous to those found at our study sites. During recon-naissance work at Range A, an eastern cottontail wasobserved foraging on the berm face. The diet of theeastern cottontail includes diverse herbaceous and woodyplants, with the former being preferentially consumedduring the summer months and the latter being consumedduring the winter (Chapman et al., 1980). This receptorcould be exposed to Pb assimilated into plant tissues andfrom contaminated dust adhering to plant leaves.

2.2.2. American robin (Turdus migratorius)The American robin is a common three season

resident of diverse habitats in the study area (Saueret al., 2006), and has been observed foraging at Ranges Aand B (personal observation). Its diet varies widely

throughout the year, fluctuating from sugar and lipid-richfruits in the fall and winter to invertebrates during thebreeding season (Wheelwright, 1986). Because of itshabit of foraging on the ground as well as its dietarypreferences, the American robin is potentially vulnerableto Pb exposure at firing ranges (Peddicord and LaKind,2000). Earthworms may comprise a large portion of therobin's diet in the spring and are known to bioaccumu-late Pb (e.g. Ireland, 1979; Reid and Watson, 2005).Thus, the American robin represents an avian omnivorethat is potentially exposed to Pb through consumption ofearthworms in contaminated areas at RP ranges.

2.2.3. Red-tailed hawk (Buteo jamaicensis)The red-tailed hawk is the most common and wide-

spread raptor in North America (Sauer et al., 2006), andis particularly common in patchy areas of mixed forestand grasslands, where prey visibility is high (Baker andBrooks, 1981). The red-tailed hawk consumes varyingproportions of birds and mammals, depending on avail-ability (Fitch et al., 1946). It was chosen as a receptororganism that represents a top predator, which wouldlikely consume species that are themselves exposed tocontaminants.

2.2.4. Short-tailed shrew (Blarina brevicauda)The short-tailed shrew is a small, carnivorous mam-

mal, which lives and forages in shallow burrows. Earth-worms often comprise a large portion of its diet, puttingthe shrew at risk of bioaccumulation of heavy metals(Reinecke et al., 2000). Its subterranean habit means theamount of incidentally consumed soil in the shrew's dietis particularly high (cf. Talmage and Walton, 1993). Thelife history characteristics of the short-tailed shrewtherefore make it potentially vulnerable to contaminationat RP ranges.

2.3. Soil sampling and analysis

Soils at Ranges A and B were thoroughly sampled inorder to determine contaminant patterns. Soil sampleswere collected at several depths, using standard tech-niques in a stratified pattern (Fig. 1). Due to knownissues of sample heterogeneity at RP ranges (USEPA,2003), sample preparation followed a protocol designedto minimize analytical uncertainty. Samples werehomogenized in the laboratory, and soil was sieved tob2 mm, the fraction which is most frequently used in RPrange studies (e.g. Astrup et al., 1999; Rooney et al.,1999; Chen and Daroub, 2002; Cao et al., 2003;Hardison et al., 2004), and was assumed to potentiallyform a component of incidentally consumed soil. While

94 J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

larger fractions may represent an ongoing source ofmetals to the biologically relevant fraction, they are notimmediately bioavailable to the chosen receptors, and ifanalyzed along with the finer fraction, could inflateconcentrations beyond that which is bioavailable. Fromthe b2 mm fraction, 30 g of soil was homogenized for1 min using a mortar and pestle, and then a 0.5 gsubsample was re-homogenized using a mortar andpestle. This subsample was heated with 2 mL of nitricacid and 6 mL of hydrochloric acid overnight so that thevolume was reduced to 1–2 mL. The solution was thenmade up to 25 mL with distilled deionized water, filteredthrough a Whatman® No. 40 filter paper, and analyzedby a Varian Vista AX simultaneous inductively coupledplasma atomic emission spectrometer (ICP-AES), usinga detection limit of 10 mg kg−1. Analytical accuracy ofsoil results, as well as all other results presented, wasverified through the analysis of standard referencematerials. Analytical precision of all results was verifiedusing 10% field duplicates and an additional 10%analytical duplicates.

No a priori assumption was made as to theimportance of bullet constituents in contributing toecological risk, therefore concentrations of As, Cu, Pb,Sb and Zn were measured. However, risk calculationsproduced negligible results for all metals but Pb; onlyresults for Pb are presented. Contaminant profilesindicated that concentrations attenuated quickly withsoil depth. Risk calculations were thus completed forsurface (0–5 cm depth) concentrations, in the interest ofboth conservatism in estimating risk and biologicalrelevance. Surface Pb concentrations patterns aresummarized in Table 1.

2.3.1. Contaminant pathway analysisContaminant pathwayswere constructed based on site-

specific data and literature information, and include onlysignificant pathways. Based on ERA guidance (USEPA,1991; Suter et al., 2000), air was not considered asignificant pathway; and based on soil pH and contam-

Table 1Summary of surface (0–5 cm) soil Pb concentrations at RP Rangesstudied

Range Number ofsamples

Maximum[Pb] mg kg−1

Minimum[Pb] mg kg−1

Mean [Pb] mgkg−1 (SE)

A 80 26700 b10 1910 (569)B 73 16400 b10 1260 (389)C 23 27600 12 6170 (2040)

Risk calculations were performed for Ranges A and B; Range C wasused for analysis of contaminant accumulation in receptor organisms.Number of samples includes ∼10% duplicates.

inant profiles (ESG, 2005), water was also considered aninsignificant pathway. Food and soil ingestion weretherefore considered the significant pathways.

In order to characterize the contribution of the rele-vant pathways to receptor exposure, contaminants wereanalyzed in both grasses and earthworms. No earth-worms were found in areas with N10000 mg kg−1 soil;thus, a model could not be constructed that would spanthe range of soil concentrations needed. The Oak RidgeNational Laboratory (ORNL) generalized model foraccumulation of Pb in earthworms (Sample et al., 1998a)was used, since it incorporates data from soil [Pb]N20000 mg kg−1. Logistical constraints also dictatedthat ORNL generalized models (Sample et al., 1998b)were used in analysis of contaminants in prey items ofthe red-tailed hawk.

A total of 22 vegetation samples (including twoduplicates) were collected at Range C. Samples werecollected within a 50 cm radius of corresponding soilsample locations. Only the above-ground, non-woodyportion of the plant was collected, in order to simulatethe non-winter feeding habits of the eastern cottontail.Vegetation samples were comprised of ∼85% grasses(Poaceae), as well as a mixture of herbaceous and woodyspecies. Vegetation appeared healthy in all areas,including those adjacent to bullet pockets.

Published soil intake rates for herbivores that arecommonly used in risk assessments (e.g. USEPA, 1993)do not account for sites such as RP ranges, where largequantities of potentially contaminated dust are producedwhen bullets impact backstop berms. Herbivores for-aging at RP ranges can be exposed to this dust throughconsuming plants in the fallout zone. We thereforeanalyzed vegetation samples both as collected, and thenrinsed four times with deionized water, so the contribu-tion made by dust to total plant metal concentrationscould be determined. Plant samples were analyzed byICP-AES, using a detection limit of 5 mg kg−1. Resultsfor unwashed plant samples reflected both plant andincidental soil/dust consumption.

2.3.2. BioaccessibilitySoil and unwashed plant samples collected at Range

C were subjected to in vitro bioaccessibility analysis inorder to estimate the bioavailability of contaminants toreceptors. Without the incorporation of bioaccessibilityinformation, the conservative assumption is made thatreceptors will absorb 100% of the contaminant theyingest. Incorporation of bioaccessibility analyses intoERA provides more realistic estimates of contaminantuptake, and thus better estimates of receptor risk (Oomenet al., 2003; Peijnenburg and Jager, 2003). A growing

95J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

number of risk assessors are incorporating bioaccessi-bility analyses into ERAs, and reference manuals andvalidated analytical methods have been published in aneffort to standardize these tests (e.g. USEPA, 2004).

Soil samples were analyzed using two different in vitrobioaccessibility models. The first model simulated mam-malian gastrointestinal conditions, in order to mimic di-gestion in the short-tailed shrew. Sample preparationand analytical methods followed Kelley et al. (2002) andUSEPA (2004). The second model simulated gastrointes-tinal conditions of the American robin and involvedmodifications of an avian model (Levengood andSkowron, 2001) to reflect the robin's gastrointestinaltract (ESG, 2005).

Unwashed plant samples were analyzed using an invitro bioaccessibility model to simulate gastrointestinalconditions of the eastern cottontail. The USEPA (2004)method was modified in order to reflect the dietarycharacteristics of the eastern cottontail: a correction to∼67% moisture content in vegetation was used (ESG,2005). Bioaccessibility extracts were analyzed by ICP-AES, using a detection limit of 5 mg kg−1. No directinference to in vivo bioaccessibility values was possible,as USEPA (2004) models do not span the Pb levels foundin our study; however, our method provides a reasonableestimate of bioavailability. A linear regression of natural-log (ln)-transformed bioaccessible [Pb] versus soil [Pb]was used to construct a model to infer bioavailablevegetation [Pb] for risk calculations. Polynomial regres-sions did not explain significantly greater variation thanthe linear model.

2.3.3. Risk calculationsDaily intake of a contaminant through food ingestion

was calculated as follows, using receptor-specific pa-rameters from standard sources (e.g. USEPA, 1993):

EDIf ¼Xm

i−1

ðCi � IRi � RAFiÞBW

ð1Þ

where EDIf is the estimated daily intake of contaminantthrough food ingestion (mg kg−1 d), Ci is the con-centration of contaminant in food type i (mg kg−1), IRi isthe ingestion rate (kg d) of food type i, RAFi is therelative absorption factor (bioaccessibility) for food typei at concentration Ci (unitless), and BW is the averagebody weight of the organism (kg).

Ingestion of soil, dust and grit may occur when soiland dust particles become adhered to a food source or arepresent in the gut of the food source being ingested by areceptor. Daily intake of a contaminant through soilingestion (EDIs) can be estimated using Eq. (1),

substituting soil as a food type. Direct ingestion of shotwas not considered, as shot in the size range potentiallyconsumed by receptors (i.e. bird shot) is not used at thestudy sites.

In risk assessments, the Hazard Quotient (HQ) isfrequently used to estimate potential risk to target organ-isms. The HQ is defined as the quotient of exposureconcentrations divided by a toxicological benchmark(Suter et al., 2000). HQs for receptor organisms werecalculated as follows:

HQ ¼ ðEDIf þ EDIsÞTDI

ð2Þ

where TDI is the tolerable daily intake of a specificcontaminant (mg kg−1 d), from literature values (Sampleet al., 1996). Given that site contaminant characteristicswere well known, we used the lowest observed adverseeffects level (LOAEL) as the TDI. In literature-based, orscreening-level ERAs, HQN1 indicates possible risk tothe receptors, and thus warrants further exploration. Incases where site-specific data are used, the magnitude ofthe HQ is also important, where large HQs indicatepotentially larger effects, and/or more certainty thateffects will indeed occur (Suter et al., 2000).

Accounting for uncertainty is an important aspect ofrisk assessment in general (Hammonds et al., 1994), and isparticularly important in firing ranges because ofhigh uncertainty in contaminant concentration results(USEPA, 2003). Uncertainty in risk assessment para-meterswasmodeled byMonte Carlo simulation, using thesoftware @RISK© V4.5.5 (Palisade Corp., Ithaca, NY).

2.4. Incorporation of foraging ranges

In most screening-level ERAs, the assumption ismade that a receptor attains 100% of its food from themost contaminated area of a site (Suter et al., 2000). Thisassumption, while conservative, does not account for thefact that receptors may spend only a portion of their timeforaging in the contaminated area, particularly if this arearepresents a small fraction of their home ranges. Theissue is compounded in RP ranges, where soil Pb con-centrations may reach levels N90000 mg kg−1 at small“hot spots” in bullet pockets and then attenuate quicklywith distance (ITRC, 2003, 2005). Extrapolation of HQscalculated for a “hot spot” does not provide a realisticestimate of use patterns for receptors typically consid-ered at RP ranges, and is therefore of limited utility inguiding remedial action.

In order to account for foraging ranges of receptororganisms, we computed HQs for all soil sample

Fig. 2. Least squares regression of ln-transformed bioaccessible Pbconcentrations in vegetation versus soil Pb concentrations.

96 J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

locations, and then used ordinary kriging with a linearvariogram to estimate HQs for the receptor foragingranges. Ordinary kriging was chosen because it appearsto provide a good estimate of site-wide contaminantconcentration distributions (and by extension, inferredHQs) for firing range sites given proper data transfor-mation and adequate sample coverage (Thayer et al.,2003). HQs were ln-transformed to account for posi-tively skewed contaminant patterns found at RP ranges(USEPA, 2003). Transformed HQs for 1-m2 pixels overthe extents of contaminated areas were calculated bykriging, using modeled values at sample locations. Thisprocedure was repeated using +2 standard deviation(SD) HQs resulting from risk calculations. Back-transformed HQ pixels were then overlain on “worst-case” (i.e., maximum exposure) foraging ranges for thereceptors, and an average, or range-normalized HQ(RNHQ) was calculated. Receptors were assumed toforage equally across their ranges, as there was noevidence for differential use of the backstop berms andthe surrounding areas, and the receptor organisms arehabitat generalists. Areas that were uncontaminated byfiring range activities were assumed to have HQsequivalent to background metal levels. This techniqueallowed a realistic computation of modeled and +2SDRNHQs, and also allowed us to investigate the influenceof targeted remediation on RNHQs.

3. Results and discussion

3.1. Plant Pb accumulation

Vegetation samples collected at Range C showed aclear pattern of Pb bioaccumulation. Natural-log (ln)[Pb] in unwashed plants was highly correlated with ln[Pb] in soil (r2 =0.71; pb0.0001); ln [Pb] in washedplants was also highly correlated with ln [Pb] in soil(r2 =0.76; pb0.0001). In the ORNL generalized modelfor Pb accumulation in plants (Bechtel Jacobs Ltd.,1998), ln [Pb] in plants is considerably less highlycorrelated with ln [Pb] in soil (r2 =0.24), possibly due tothe fact that the ORNL generalized model integratedresults from a variety of contaminant sources as well assoil and ecosystem types.

When applied to Range C data, the ORNL general-ized model (Bechtel Jacobs Ltd., 1998) also under-predicted every value of [Pb] in plants, predicting onaverage∼12% of actual [Pb] in washed plants and∼6%of actual [Pb] in unwashed plants. Cao et al. (2003)found accumulation of Pb in washed plant shoots at afiring range of ∼87 to ∼800 mg kg−1, corresponding tosoil Pb from ∼150 to ∼6800 mg kg−1, values which

compare favorably with our results, but not with thegeneralized model. Indeed, both our observed plant [Pb]as well as those of Cao et al. (2003) were outside theupper confidence limits of the generalized model,indicating that accumulation of Pb in plants at RP rangesis greater than at sites considered by Bechtel Jacobs Ltd.(1998). The additional input of dust may be the primarycause of this difference; the Bechtel Jacobs Ltd. (1998)model did not use sites with aerial input of Pb.

The pattern of washed and unwashed plant [Pb] atRange C also revealed important information aboutsources. In washed plants, [Pb] averaged ∼42% lowerthan in unwashed plants, indicating that adhered dust(likely originating from bullet impacts) contributedsubstantially to the Pb burden of in situ plants. However,washed plants still contained Pb levels much higher thanpredicted using the ORNL generalized model (BechtelJacobs Ltd., 1998). Laboratory experiments by Rooneyet al. (1999) using TS range soils found uptake by radishroots of up to 3825 mg kg−1 Pb corresponding with soil[Pb] of∼5900 mg kg−1 Pb, while stem uptake remainedN100 mg kg−1. However, in an active firing range(where dust would presumably be a factor), Cao et al.(2003) found only slightly higher [Pb] in roots (average∼750mg kg−1) than in shoots (average∼420mg kg−1).The relative contributions of dust versus root uptake to[Pb] in plant shoots at RP ranges need to be resolved, asthe elevated concentrations are clearly an important con-sideration where herbivores are potential ecologicalreceptors.

3.2. Bioaccessibility

Bioaccessibility analyses indicated that only a portionof the Pb in the soil and unwashed plants is bioaccessible.

Table 2Hazard Quotients (HQ) for receptors at Ranges A and B, including maximum HQ found in a single sample and HQs normalized for “worst case”foraging ranges; +2SD values are italicized

Receptor Foragingrange(hectares)

Maximum single-sample Range-normalized HQ

HQ (+2SD) (+2SD)

Range A Range B Range A Range B

American robin a 0.15 36.7 (86.3) 23.8 (55.2) 4.10 (9.24) 1.84 (4.13)Short-tailed shrew b 0.35 2.03 (4.57) 1.35 (3.01) 0.113 (0.243) 0.054 (0.117)Eastern cottontail c, d 3.24 15.8 (123) 0.109 (0.703)Red-tailed hawk d, e 671 0.494 (1.35) 0.006 (0.013)

a Summer adult foraging range when feeding nestlings in Ontario, from Weatherhead and McRae (1990).b Average of foraging ranges presented in USEPA (1993).c Average of spring and fall values from USEPA (1993), as assumed not to be foraging on firing ranges during winter.d Receptor's foraging range encompasses both Ranges A and B.e Average from Luttich et al. (1971), which used northern portion of species distribution.

97J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

Ln [Pb] in soil extracts using mammalian and avianin vitro bioaccessibility methods were both stronglycorrelated with ln [Pb] in the original soil samples(r2N0.95, pb0.0001), allowing us to estimate bioacces-sibility across the range of soil [Pb]. Bioaccessibilityin soil averaged ∼66%, and ranged from ∼100% insamples with low [Pb] to ∼13% in a sample with∼21900 mg kg−1 Pb. These values correspond well toother bioaccessibility analyses on soil (e.g.Oomen et al.,

Fig. 3. a, b. Isopleths of pixilated HQs for the American robin, and foraging raconstructed using mean HQs modeled using Monte Carlo simulation, while

2003), and the relationships with soil [Pb] provide esti-mates of the amount of Pb that is absorbed by receptorsconsuming soil at RP ranges.

Ln [Pb] in unwashed plant samples analyzed usingthe mammalian in vitro bioaccessibility method wasalso strongly correlated with ln [Pb] in soil samples(r2 =0.72, pb0.0001). This relationship between soil[Pb] and unwashed plant bioaccessible [Pb] (Fig. 2) wasused in risk calculations for the eastern cottontail.

nge (inset) centered to maximize normalized HQ. Isopleths in (a) wereisopleths in (b) were constructed using +2 standard deviation HQs.

98 J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

3.3. Risk characterization using foraging ranges

HQ calculations using foraging ranges indicate thatPb contamination at ranges A and B appears to pose thegreatest risk to the American robin (Table 2). Potentialrisk was also found for the short-tailed shrew and theeastern cottontail. Risk to the red-tailed hawk appearedto be negligible from firing range Pb contamination,with a worst-case (+2SD) RNHQ of ∼0.01.

The large difference between the highest HQ for asingle sample and range-normalized HQs illustrates thepotential problem in using a site-maximum value toestimate risk, especially in sites where the area of a sitemaximum represents only a very small portion of re-ceptors' foraging ranges. Risk assessment experts recog-nize that accounting for receptor foraging ranges isdesirable in sites with small and/or spatially heteroge-neous contaminated areas (e.g. Suter et al., 2000). Severalapproaches to this issue have been proposed. Sample et al.(1997) and Suter et al. (2000) recommend partitioningforaging ranges into several discrete sub-areas of identicalcontaminant levels. The US Environmental ProtectionAgency (USEPA) 3MRA risk modeling system uses aseries of bins, whereby receptors are assigned to one offour generic habitat sizes used for risk calculations

Fig. 4. a, b. Isopleths of pixilated HQs for the short-tailed shrew, and foragingconstructed using mean HQs modeled using Monte Carlo simulation, while

(USEPA, 1999). Marinussen and van der Zee (1996)propose an approach using a moving window of speciesranges across a contaminated site. Our method is moreaccurate than the former two approaches, because itallows both foraging ranges and contaminant distributionsto be more precisely represented. And it is more ap-propriate for smaller sites than the approach of Mar-inussen and van der Zee (1996), since remediation of thearea affecting the worst case foraging range would bringall other foraging ranges below the HQ criterion.

Despite the fact that the American robin and short-tailed shrew shared a diet of earthworms, and earthwormswere assumed to comprise only 36% of the robin's diet(versus 100% for the shrew), the robin was more at riskfrom Pb contamination than the shrew. This is because: 1)the published LOAEL for Pb is greater for the shrew(175.83 mg kg−1 d) than for the robin (11.3 mg kg−1 d)(Sample et al., 1996), and 2) the robin's foraging range issmaller than that of the shrew. In our case, the estimatedforaging range for breeding robins in Ontario fromWeatherhead and McRae (1990) represented the bestconservative estimate; risk assessments in other areascould with justification use different range estimates.

Isopleths of pixilated HQs (Figs. 3, 4, and 5) showedclearly that the highest HQs were concentrated at the

range (inset) centered to maximize normalized HQ. Isopleths in (a) wereisopleths in (b) were constructed using +2 standard deviation HQs.

Fig. 5. a, b. Isopleths of pixilated HQs for the eastern cottontail, and foraging range (inset) centered to maximize normalized HQ. Isopleths in (a) wereconstructed using mean HQs modeled using Monte Carlo simulation, while isopleths in (b) were constructed using +2 standard deviation HQs.

99J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

bullet pockets. However, areas contributing to elevatedrisk extend throughout the berms, and in some casesbeyond. This is likely a result of both errant shots andthe accumulation of contaminated dust generated asbullets strike the pockets. Indeed, subsurface [Pb]profiles (unpublished data) indicate that migration ofPb contamination in the horizontal plane along thesurface is more significant than along the vertical planethrough the soil column. The results of several otherstudies (e.g. Astrup et al., 1999; Chen and Daroub,2002; Cao et al., 2003) indicate that this situationcommonly occurs at RP ranges, and is partly due to theformation of relatively insoluble weathering products inthe soil (Vantelon et al., 2005). However, under certainsoil conditions, including low/high pH and high organiccarbon, Pb may become mobile in soils (Cao et al.,2003; Hardison et al., 2004).

Since risk calculations were performed for conserva-tively estimated home ranges of receptors whose popula-tions are not under threat, there is likely no significanteffect on receptor populations at Ranges A and B.However, if protection of individuals was crucial (e.g. ifanalogous protected species were present), risk toreceptors with diets similar to the American robin, short-tailed shrew or eastern cottontail could be unacceptable,

especially if the receptors' foraging ranges were equal toor smaller than those presented here. Lewis et al. (2001)found evidence of Pb toxicosis in individual small birdsand mammals at a large shooting range complex, andpostulated that it was perhaps caused by inadvertentingestion of Pb shot ormacroscopic fragments. Our resultsindicate that consuming large particles is not necessary toproduce elevated risk, and that risk to individual receptorsmay occur based on normal foraging behavior.

If necessary, targeted remediation of the areas ofhighest risk (as indicated by isopleths) could be used toreduce the risk within the worst-case foraging ranges.For example, in order to reduce HQ to b1 for theAmerican robin, 170 m3 of soil (or 308 m3 if +2SDresults were used) would need to be remediated tobackground Pb levels, assuming remediation of the mostcontaminated surface soil down to 30 cm depth. Thedifferences between calculated HQs and +2SD HQsargues for a conservative approach that accounts for theconsiderable uncertainty associated with risk calcula-tions at RP ranges. Periodic remediation at active rangescould reduce risk to more conservative levels.

It should be noted, however, that periodic remedia-tion at active ranges would need to account for dustgenerated by bullet impacts as a potential contaminant

100 J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

vector. The accumulation of dust on plants and surfacesoil would continue to increase Pb concentrations at theunremediated periphery of the contaminated area, andwould increase range-normalized HQs, thereby neces-sitating subsequent remediation of larger areas. Givenan estimated ∼1.5% abrasion of bullets to b2 mmparticles on impact (Hardison et al., 2004) and usepatterns of Ranges A and B, ∼3.7 kg of b2 mmparticulate Pb is added to each firing range per year,much of it potentially susceptible to aerial transportduring bullet impacts. We recommend that this issue beconsidered in RP range monitoring programs.

The results of this study illustrate the importance ofconsidering specific characteristics of RP ranges, not onlywhen performing risk assessments, but also in guiding themanagement of active ranges. For example, where in-dividual receptors analogous to the American robin areimportant, discouragement of nesting and foraging (e.g.through decoy predators) may be necessary. Ourobservation of a bird nest in the shooting gallery ofRange B indicates that intermittent firing is not enough todiscourage continued foraging at firing ranges.

Our results for plant [Pb] as well as those of Cao et al.(2003) also clearly indicate that livestock should not begrazed near RP range berms. Dust suppression techni-ques that do not encourage increased Pb solubility (e.g.ITRC, 2005) may be needed to minimize contaminantspread and reduce risk to receptors. In addition, if grasseson firing ranges are mowed, the mulch should be treatedas being potentially highly contaminated with Pb.

Acknowledgements

This research was supported by the Natural Sciencesand Engineering Research Council of Canada (NSERC)through the award of a Discovery Grant to KJR as well asthrough theMetals in the Human Environment (MITHE)Research Network. We also acknowledge funding fromthe Department of National Defence Academic ResearchProgram (ARP). Laura Stewart, (EnvironmentalSciences Group, Royal Military College of Canada)assisted in creating the figures. Additional technicalassistance provided by David Dougherty (ConsultingServices Canada), and helpful comments from twoanonymous reviewers are gratefully acknowledged.

References

Astrup T, Boddum JK, Christensen TH. Lead distribution and mobilityin a soil embankment used as a bullet stop at a shooting range.J Soil Contam 1999;8:653–65.

Baker JA, Brooks RJ. Distribution patterns of raptors in relation todensity of meadow voles. Condor 1981;83:42–7.

Bechtel Jacobs Ltd. Empirical models for the uptake of inorganicchemicals from soil by plants. Oak Ridge: US Department ofEnergy; 1998. 38 pp.

Cao X, Ma LQ, Chen M, Hardison DW, Harris WG. Weathering oflead bullets and their environmental effects at outdoor shootingranges. J Environ Qual 2003;32:526–34.

Chapman JA, Hockman JG, Ojeda MM. Sylvilagus floridanus.Species account, vol. 136. American Society of Mammalogists;1980. 8 pp.

Chen M, Daroub SH. Characterization of lead in soils of a rifle/pistolshooting range in Central Florida USA. Soil Sediment Contam2002;11:1–17.

Environmental Sciences Group (ESG). Phase II ecological riskassessment of correctional service Canada small arms firingranges. Kingston: Royal Military College of Canada; 2005. 59 pp.

Fitch HS, Swenson F, Tillotson DF. Behavior and food habits of thered-tailed hawk. Condor 1946;48:205–37.

Hammonds JS, Hoffman FO, Bartell SM. An introductory guide touncertainty analysis in environmental and health risk assessment.Oak Ridge: Oak Ridge National Laboratory; 1994. 34 pp.

Hardison DW, Ma LQ, Luongo T, Harris WG. Lead contamination inshooting range soils from abrasion of lead bullets and subsequentweathering. Sci Total Environ 2004;328:175–83.

Ireland MP. Metal accumulation by the earthworms Lumbricusrubellus, Dendrobaena veneta and Eiseneilla tetraedra living inheavy metal polluted sites. Environ Pollut 1979;13:201–6.

ITRC. Characterization and remediation of soils at closed small armsfiring ranges. Washington: Interstate Technology and RegulatoryCouncil Small Arms Firing Range Team; 2003. 67 pp.

ITRC. Environmental management at operating outdoor small-armsfiring ranges. Washington: Interstate Technology and RegulatoryCouncil Small Arms Firing Range Team; 2005. 87 pp.

Kelley ME, Brauning SE, Schoof RA, Ruby MV. Assessing oralbioavailability of metals in soil. Columbus: Battelle Press; 2002.124 pp.

Kendall RJ, Lacher TE, Bunk C, Daniel B, Driver C, Grue CE, et al.An ecological risk assessment of lead shot exposure in non-waterfowl avian species: upland game birds and raptors. EnvironToxicol Chem 1996;15:4–20.

Levengood JM, Skowron LM. Use of a simulated gizzard to measurebioavailability of metals and other elements to waterfowl.Ecotoxicol 2001;10:299–304.

Lewis LA, Poppenga RJ, Davidson WR, Fisher JR, Morgan KA. Leadtoxicosis and trace element levels in wild birds and mammalsat a firearms training facility. Arch Environ Contam Toxicol2001;41:208–14.

Luttich SN, Keith LB, Stevenson JD. Population dynamics of the red-tailed hawk (Buteo jamaicensis) at Rochester, Alberta. Auk1971;88:75–87.

Marinussen MPJC, van der Zee SEATM. Conceptual approach toestimating the effect of home-range size to the exposure oforganisms to spatially variable soil contamination. Ecol Model1996;87:83–9.

Oomen AG, Rompelberg CJM, Bruil MA, Dobbe CJG, PereboomDPKH, Sips AJAM. Development of an in vitro digestion modelfor estimating the bioaccessibility of soil contaminants. ArchEnviron Contam Toxicol 2003;44:281–7.

Peddicord RK, LaKind JS. Ecological and human health risks at anoutdoor firing range. Environ Toxicol Chem 2000;19:2602–13.

Peijnenburg WJGM, Jager T. Monitoring approaches to assessbioaccessibility and bioavailability of metals: matrix issues.Ecotoxicol Environ Saf 2003;56:63–77.

101J.R. Bennett et al. / Science of the Total Environment 374 (2007) 91–101

Reid BJ, Watson R. Lead tolerance in Aporrectodea rosea earthwormsfrom a clay pigeon shooting site. Soil Biol Biochem 2005;37:609–12.

Reinecke AJ, Reinecke SA, Musilbono DE, Chapman A. The transferof lead (Pb) from earthworms to shrews (Myosorex varius). ArchEnviron Contam Toxicol 2000;39:392–7.

Rooney CP, McLaren RG, Cresswell RJ. Distribution and phytoavail-ability of lead in a soil contaminated with lead shot. Water Air SoilPollut 1999;116:535–48.

Sample BE, Opresko DM, Suter II GW. Toxicological Benchmarks forWildlife: 1996 Revision. Oak Ridge: US Department of Energy;1996. 43 pp.

Sample BE, Aplin MS, Efroymson RA, Suter II GW, Welsh CJE.Methods and tools for estimation of the exposure of terrestrialwildlife to contaminants. Oak Ridge: US Department of Energy;1997. 148 pp.

Sample BE, Beauchamp JJ, Efroymson RA, Suter II GW, AshwoodTL. Development and validation of bioaccumulation models forearthworms. Oak Ridge: US Department of Energy; 1998a. 45 pp.

Sample BE, Beauchamp JJ, Efroymson RA, Suter II GW. Develop-ment and validation of bioaccumulation models for smallmammals. Oak Ridge: US Department of Energy; 1998b. 41 pp.

Sauer JR, Hines JE, Fallon J. The North American Breeding BirdSurvey, Results and Analysis 1966–2005. http://www.mbr-pwrc.usgs.gov/bbs/. July 2006.

Stansley W, Roscoe DE. The uptake and effects of lead in smallmammals and frogs at a trap and skeet range. Arch EnvironContam Toxicol 1996;30:220–6.

Suter II GW, Efroymson RA, Sample BE, Jones DS. Ecological riskassessment for contaminated sites. Boca Raton: Lewis Publishers;2000. 438 pp.

Talmage SS, Walton BT. Food chain transfer and potential renaltoxicity to small mammals at a contaminated terrestrial field site.Ecotoxicol 1993;2:243–56.

Thayer WC, Griffith DA, Goodrum PE, Diamond GL, Hassett JM.Application of geostatistics to risk assessment. Risk Anal 2003;23:945–60.

Tripathi RK, Sherertz PC, Llewellyn GC, Armstrong CW. Leadexposure in outdoor firearms instructors. Am J Publ Health1991;81:753–5.

U.S. Environmental Protection Agency (U.S. EPA). Ecologicalassessment of hazardous waste sites: a field and laboratoryreference. Washington: Office of Research and Development;1991. EPA/540/R-92/003.

U.S. Environmental Protection Agency (U.S. EPA). Wildlife expo-sures handbook, vol. 1. Washington: Office of Research andDevelopment; 1993. EPA/600/R-93/187a.

U.S. Environmental Protection Agency (U.S. EPA). Data Collectionfor the Hazardous Waste Identification Rule, Section 13.0 —Ecological Receptors and Habitats. Office of Solid Waste,Washington, 1999. http://www.epa.gov/epaoswer/hazwaste/id/hwirwste/pdf/risk/data/s0043.pdf.

U.S. Environmental Protection Agency (U.S. EPA). Best managementpractices for lead at outdoor shooting ranges. New York: Divisionof Compliance and Enforcement; 2001. EPA/902/B-01/001.

U.S. Environmental Protection Agency (U.S. EPA). TRW recommen-dations for performing human health risk analysis on small armsshooting ranges. Washington: Office of Solid Waste andEmergency Response; 2003. OSWER 9285.7–37.

U.S. Environmental Protection Agency (U.S. EPA). Estimation ofrelative bioavailability of lead in soil and soil-like materials usingin vivo and in vitro methods, draft final. Washington: Office ofSolid Waste and Emergency Response; 2004. OSWER 9285.7–77.

Vantelon D, Lanzirotti A, Scheinost AC, Kretzschmar R. Spatialdistribution and speciation of lead around corroding bullets in ashooting range soil studied by micro-X-ray fluorescence andabsorption spectroscopy. Environ Sci Technol 2005;39:4808–15.

Vyas NB, Spann JW, Heinz GH. Lead shot toxicity to passerines.Environ Pollut 2000;111:135–8.

Weatherhead PJ, McRae SB. Brood care in American robins:implications for mixed reproductive strategies by females. AnimBehav 1990;39:1179–88.

Wheelwright NT. The diet of American robins: an analysis of U.S.biological survey records. Auk 1986;103:710–25.