259878

Environmental Forensics (2001) 2. 141-15?doi:10.1006'enfo. 2001. 0046. available online al http://www.idealibrary.coin on

Protocol Development for Assessing Arsenic BackgroundConcentrations in Florida Urban Soils FPAR«, CDEPA Region 5 Records Ctr.

Tail Chirenje*, L. Q. Ma and A. G. Hornsby

Soil and Water Science Department, University of Florida, Gainesville, FL 32611-0290, U.S.A.

K. Portier

Statistics Department, University of Florida, Gainesville, FL 32600-0339, U.S.A.

W. Harris

Soil and Water Science Department. University of Florida, Gainesville, FL 3261 1-0290, U.S.A.

S. Latimer

Department of Urban ami Regional Planning, University of Florida. Gainesville, FL 326-5705, U.S.A.

E. J. Zillioux

Florida Power and Light, Environmental Services Department, PO Box 14000, Juno Beach, FL 33408, U.S.A.

(Received 20 February 2001, Revised manuscript accepted 26 Marc/i 2001)

Knowledge of arsenic background concentrations in urban soils is important Tor making remediation decisions. The soilcleanup target level (SCTL) for arsenic in Florida lies within the ranee of arsenic background concentrations. Theresidential SCTL is also near the practical quantification limits using analytical procedures. Currently no standardprotocols are available for determining arsenic background concentrations in urban soils, apart from site-specific cases.Therefore, a pilot study was conducted to develop and employ appropriate protocols to determine arsenic distribution inurban soils. This involved: site selection (e.g. size and sampling frame), sample collection (e.g. sampling technique), andstatistical considerations (e.g. design). Factors such as ease of sample collection and maintaining anonymity of privateproperties were also considered as they influence the successful implementat ion of the study. Forty surface soil sampleseach were collected from five categories in three land use classes (residential-yard and right-of-way, commercial andpublic land-parks and public building), digested using EPA method 3051 a and analysed using graphite furnace atomicabsorption spectrometry. Experiences from the pilot study (e.g. complications during sample selection, digestion, datacensoring etc.) were used in the development of the final protocol to be used in determining the distribution of arsenic inurban areas. t 2001 AEHS

Keywords: anthropogenic: background: urban soils: stratification.

Knowing background concentrations of arsenic inurban soils provides a yardstick against which theimpacts of human ac t iv i ty can he measured. Back-ground concentrations of arsenic in relatively undis-turbed Florida soils are established and thc> vary fromO . U l to 61.1 mg/kg, with a geometric mean (GM) ofn 11 mg kg (Chen. Ma and Harr i s . |QOO). Typical soilarsenic concentrations range between 0.1 and 40 mg kgworldwide, with an ar i thmetic mean concentration of5 6 mg kg (KabaUi-Pcnduis and Pendias. IW2). How-ever , l i t t l e information is ava i lab le on arsenic back-ground concentrations in urban soils.

M.i\ ':'•: v>: l'>n: li-m.iil ichircni,Mitl.'.\ln

Urban soils, created and/or modified in theprocess of urbanizat ion, are complex and hetero-geneous in t h e i r structure and composition (Craul .l l ix .s ; l ) : i \ i c v Wal l and Thornton. 19X7). In contrast tonatural auen tM such as wind, water, ice. gruvilv andheat, h u m a n ac t iv i ty is the predominant active agent inthe modi f ica t ion of these soils (Barret t . I1W7).There is ,i h igher probability of historic anthropo-genic con tamina t ion , vertical mixing during develop-ment , use of fill from different geologic areas,deposi t ion and or contributions from the use ofpesticides or amendments from other sources inu r b a n soils t h a n m undeveloped areas ( C ' r a u l . I ^ X s ;T h o r n t o n . i ' lS" ) .

141):; nl i O ' i l - l : • ;.- S?5 nil on Hi in i .\i:ns

142 Chirenje ft al.

Arsenic concentrations in urban areas vary con-siderably over short distances, but the collection of alarge number of samples to overcome this problem isprohibitive both in terms of expense and time. Theextent of anthropogenic influence on soil arsenicconcentrations depends on both current and previousland use in most urban areas. For example, the impactsof heavy construction in many commercial areas andintensive management in "golf greens" on soil arsenicconcentrations are higher than that of relativelyundeveloped areas such as parks. Hence the differencesin land use can serve as the basis for stratification whendetermining arsenic background concentrations inurban areas. Stratification is a word that is usedextensively by both geologists and statisticians. Statis-ticians commonly use it to refer to the division of astatistical population into groups or blocks by anybasis (i.e. samples from one stratum have at least onecommon characteristic). Geologists, on the other hand,use the term to refer to the arrangement of rocks(e.g. sedimentary rocks) in layers. In this manuscript,the word strata will be used to refer to blocks ofsamples that have at least one common characteristicamong them. While the differences between arsenicconcentrations among these land uses may not besignificant in smaller cities like Gainesville, they areexpected to be significant in larger metropolitan areassuch as Miami. In smaller cities, the differencesbetween land use categories (e.g. residential andcommercial areas) in terms of general human disturb-ance, e.g. construction, is much lower than that inlarger and more developed cities.

Assessment of elemental concentrations in urbanareas facilitates the distinction of na tura l from anthro-pogenic background levels, and the distinction ofnatural and anthropogenic background levels fromthose of impacted (contaminated) areas. Most pub-lished protocols deal with site-specific backgroundconcentrations for known polluted sites on a smallscale (Brcckcnbridgc and Crockett. 1995). However,there are significant differences between site-specificdeterminations and non-site specific determinationswhich make it difficult to adapt standardized soilsampling procedures from site-specific determinationsto studies done on a regional scale, i.e. large scale.

Various researchers have carried out studies onbaseline concentrations of trace metals using differentmethods (Schwar el til., 1988; Ti l le r . 1992; Br inkmann .1994; Br inkmann and Ryan. 1997; Fields a al.. 199?;Thornton, 1987; Kaminsk i and l.anclsberger. 2000).For example, Chen ct til. (19961 sampled surface soils(0 5 cm) from representative areas (2 3 compositescomprised of eight sub-sample-.! m their assessmentstudy of trace metal d i s t r ibu t ion and ccnUiminut ion insurface soils of Hong Kong. Samples close to site-specific pollut ion sources (e.g. l a n d f i l l s , refuse stations,industr ia l plants, gasoline stations etc.) were excludedexcept in urban areas where soils were genera l ly sampledfrom the roadside. In their s tud\ i>l heavy metal con-lamination in Bangkok. Wilcke i-i ul. \ I W7i sampled 15soils (0 5 cm) in the metropol i tan Bangkok area alonga transect (at 2-km intervals) using a tolerance of about100 m around selected sampl ing points .

Fields ct a! ( 1 9 9 ^ ) collected tuo to six samples percounty in thei r s tudy of selected soil con taminan t s at

background locations in New Jersey, U.S.A. Samplescollected from suburban areas were taken from areaswith a moderate amount of human activity (e.g. parks).Urban samples were taken from parks, densely popu-lated areas and other developed areas of the state.Kaminski and Landsberger (2001) sampled primarilythe top 8 cm but also collected samples at lower depthsat some sites. Chen ct nl. (1999) did a more compre-hensive study, analysing over 440 archived surface soilsin Florida, but they did not consider urban or suburbansoils. In a study of natural background soil metalconcentrations in Washington State, U.S.A. Juan(1994) collected only 60 samples from 12 main geologicand urban regions. McGrath (1986) was more thoroughin his sampling protocol to determine the range of metalconcentrations in topsoils of England and Wales. Soilsamples in the top 15 cm were collected on a regular5-km grid over England and Wales. A 2.5-cm-diameterauger was used to take 25 cores at 4-m intervals on a20 x 20 m grid centered on a coordinate (from theNational Grid) for a composite sample. This type ofstudy requires a tremendous amount of resources that isnot commonly available to most researchers. None-theless, these studies revealed the diversity of ideas andmethods for studying background concentrations oftrace elements and the need for definitive protocols tofacilitate comparison of results.

This protocol, developed and tested in Gainesville,Florida, seeks to balance a workable number of sampleswith acceptable spatial resolution in an attempt toaddress this apparent deficiency. The protocol devel-opment study was divided into a literature review phase,the implementation of known methods in a pilot study,and a refinement phase. The main objectives of thestudy were to (1) develop a sampling and analysisprotocol for determining arsenic background concen-trations in urban soils; and (2) test the efficacy ofsampling based on three land uses to adequatelydescribe background soil concentrations in a diverseurban environment. The final product details specificprocedures to collect representatives samples to providedata of known precision and accuracy on arsenicbackground concentrations in urban soils. The factorsconsidered in developing this protocol and how theyinfluenced the final product are outlined at each step ofthe study. Thus, the study is presented as a "how-tomanual" with specific reference to the experiences fromthe Gainesville pilot study in all relevant sections.

MethodologyThe sampling protocol developed from the Ga i i i e s \ i l l cpilot study and the experiences tha t influenced the f ina lproduct arc discussed in this section. Careful a t t e n t i o nwas paid on the pract icabi l i ty , cost and a b i l i t y of thefinal product to answer specific research questions. I liedraft technical guidance tile "Determination of na tu ra lbackground concentrations of contaminants" preparedby Dade County of Florida was used as a guidel ine forsampling because guidelines for background concen-trations of urban soils are not yet avai lable. Recom-mendations from a technical report on soil samplingfor contaminat ion assessment prepared by the Centerfor Envi ronmenta l and H u m a n Toxicology al theUniversity of Florida l l l a l i n e s ct t i l . . l t ) 9X) and var ious

Assessing Arsenic Background Concentrations in Florida Urban Soils 143

researchers working in association with the FloridaDepartment of Environmental Protection (FDEP)were also considered. An important consideration inthis study was the randomization procedure. Dataanalysis in background studies depends on the type ofpopulation distribution. Hence, care was taken toavoid experimenter induced skewness, multiple popu-lations, or an excessive number of outliers so that thenature of the population distribution can only beexplained by its intrinsic characteristics and not theexperimental design.

Sampling site selection

As defined by the U.S. Census Bureau (1997), anurbanized area comprises one or more central placesand adjacent densely settled surroundings (urbanfringe) that together have a minimum of 50,000 people.The urban fringe generally consists of contiguousterritory having a density of at least 1000 people persquare mile (256 ha). Using this definition, the city ofGainesville is considered an urbanized area.

/. Development of the Sampling Frame. The first step indetermining arsenic background concentrations inurban soils is to define the sampling frame. Accuratecurrent and historical (digital and print) maps, satelliteimages, relevant aerial photos and associated back-ground information represent the best and mostefficient means of making this determination. Whilethese readily-available maps and images are very useful,it is important to check their accuracy against the latestmaps or through actual visits to selected locations(ground-truth assessment). It was discovered that eventhe most recent maps were inaccurate or misleadingabout 5% of the time in the Gainesville pilot study.

The complex nature of urban soils has to be takeninto account when assessing arsenic levels in urban soils( B r i n k m a n n , 1997). Intensive human act ivi ty signifi-cantly alters the original native soils, making it diff icultto describe these soils using typical soil classificationschemes. Knowing current and previous land usehistory of sampling sites is critical in order to obtain arepresentative soil sample for the determination of trueurban background concentrations. Natural areas suchas parks and nature preserves in urban settings rarelyrepresent totally undisturbed soils, hence they wereincluded in determining background levels of arsenic inurban soils. On the other hand, it is appropriate toexclude certain areas. A significant feature in manyurban landscapes is the occurrence of sites contami-nalcd as a resul t of specific commercial, recreational(e.g. gol! courses), or indus t r i a l (e.g. b a t t e r y manufac-t u r i n g ) ac t iv i ty . For the purposes of t h i s s ludy i t uasconcluded tha t such sites should be avoided since theydo not reflect '"the t rue background conceiur.iliuiis"but rather point source contaminat ion. This wasincorporated into our sample collection strategy bydeveloping specific criteria for exclusion. These exclu-sion cri ter ia were applied w i t h i n a temporal l i m i t of theprevious 30 years of land use changes in order tomaintain a reasonable budget. This time-frame waschosen because it is impor t an t to unders tand the landuse changes tha t have taken place before and al ter mostenv i ronmenta l regulations came in to effect af ter 1975.

All the critical distances proposed in the protocol wererevised after going through the Gainesville pilot studyin order to better reflect the actual reality in most urbanareas. In general, not all sample sites had sufficient areawithin the public utility right-of-way to collect a samplethat meets all the predetermined criteria. The presenceof CCA-treated wood poles and fences within and closeto the right-of-way in most areas also led to problems inselecting sample plugs within a sample site.

2. Stratification. The sampling frame resulting fromStep 1 still included areas with different levels of arsenic.To improve the precision of the overall sample distribu-tion and gain insight into the distribution character-istics of arsenic in different land use categories, thesampling frame was stratified. Stratified randomsampling based on three land uses, residential (yardsand right-of-way), non-industrial commercial andpublic land (public parks, and public buildings) wasselected as the most appropriate sampling strategy.These three land uses were chosen because theyrepresent the largest proportion of area in most urbanareas. For residential areas, samples were split into twocategories, yard and public utility right-of-way samples.The rationale for splitting residential areas was todetermine if right-of-way samples could be used torepresent yard samples. Section 5 discusses this in moredetail. Samples from public land were split into twocategories, public parks and public buildings and/orproperties because samples from public buildings wereexpected to be different from those from parks, whichare relatively less impacted by construction.

The type of stratification described here has thepotential to provide more homogenous sub-popu-lations of background arsenic concentrations andprovide data sets with smaller variance than if simplerandom sampling is used. In other words, stratificationhas the potential to provide a better estimation ofoverall urban average concentrations with higherconfidence compared to what would be obtained in asimple random sample. It is important to note that thetotal area divided among the four categories and theexcluded areas are urban-area specific. The relativeproportion of area should be used as mixing weightwhen the individual strata distributions for each landuse category are eventually combined to produce anoverall urban concentration distribution of arsenic.

3. Number of samples per land-use type. The number ofsoil samples needed depends on the overall precisionspecified in the study objectives. Precision is typicallyspecified for the mean or average concentration level,and several methods are a \a i lablc for selling samplesi/c based on target precision an<! associated confi-dence level exist .

For the Gainesvil le pilot s iud \ . the number ofsamples collected was based on soil heterogeneity anddetermined using the following equation:

N = [(S x t,)/R] :( 1 )

where N is the number of samples. S is the estimatedstandard deviation of the a r i t hme t i c mean of all singlevalues (in our case. S was calculated from the 25samples collected from the Unive r s i ty of Florida

144 Chirenje et at.

campus), t., is the Student r-value for a given confidenceinterval (1.96 for the 95% confidence interval) and R isthe accepted variability in mean estimation (usually 10-20% depending on the scale and budget of project). Avalue of 20% was used in our case and the minimumnumber of samples needed for Gainesville wasdetermined to be 130.

Forty samples were collected from each category (i.e.right-of-way, yards, public buildings, public parks andcommercial areas), resulting in a total of 200 samples.One out of every five samples taken from each categorywas duplicated, bringing the total number of samplesto 240.

It was later determined at the end of the pilot studythat the focus of such background studies shouldproduce a good estimate of the overall concentrationdistribution in each stratum without primarily focusingon the central tendency of each stratum. Therefore theprecision target would be set on an upper percentile ofthe concentration distribution. This assures that thebody of the distribution would be well represented whileat the same time assuring a high probability that the tailof the distribution would be represented as well.Conover (1980) described a method for calculating theminimum number of samples needed for a givenpercentile of a distribution to be exceeded by themaximum observed sample value with a given confi-dence level for environmental samples. For example,the sample size needed to assure exceedence of theupper 95th percentile with 95% confidence is 59, whilethe sample size needed for exceedence of the 75thpercentile is 11. These sample sizes would need to beapplied to each stratum to assure adequate estimationof the stratum distribution. Based on these compu-tations, it was recommended that 60 randomly-selectedsamples be obtained in each urban area stratum forfuture studies, yielding a total of 240 samples from fourcategories per city. Based on our study, no significantdifference was observed in arsenic concentrationsbetween soils in residential-yard and residential-right-of-way, thus the latter was used to represent residentialsoil, reducing land use categories to four.

4. Site selection. I n i t i a l l y the area should be stratifiedinto the four strata previously defined, i.e. residentialareas (right-of-way next to yards), commercial (right-of-way close to businesses), and public land (publicparks and public buildings), and random samplingsites selected. In practice, the random site selectionshould be done for 60 sites wi th in each stratum with anadditional six (or 10%) samples added in each stratumto provide a l t e rna t i ve sampl ing >i tes to selected areaswhere samples cannot be collected due to exclusioncriteria or olhcr un!'ore.>-tvn circumstances. The sameprocedure was used in the pi lot stiuh except that 40samples were selevt-.d hvni e.i\.l: c . t -cgorv (in c;icl\ case44 sites, i.e. 40 + Id" , , ) .

5. Obtaining permissnin in ; nlk-ct soil wimples. A signifi-cant fraction of the areas t h a t were selected for samp-ling in the residential and commercial areas were privateproperty. Thus prior to sample collection, permissionfrom the property owner was obta ined. In the Gaines-ville pilot study, approximately 65% of commercialpro-pcrty owners and I5" i , of rc t idcn t ia l property owners

denied permission to sample. Since the results fromGainesville suggested that there was no significant diffe-rence in arsenic concentrations between samples fromresidential yards and right-of-way, it may be acceptableto take soil samples in the public right-of-way. Sub-sequent discussions with other researchers revealed thatthe relationship between residential right-of-way andyard samples may not be consistent among cities.

Right-of-way v. yard sample. As stated earlier, therationale for splitting residential areas was to deter-mine if right-of-way samples could be used to representyard samples. Commercial right of way samples werealso compared with samples collected adjacent tobusinesses. Collecting samples from the right-of-wayis time-efficient and easier since no permission fromproperty owners is required. It is very difficult to obtainpermission from private (residential and commercial)property owners due to the liability issues that mayarise if high concentrations of arsenic are discovered.This is particularly true in Florida where the "Sunshinelaws" require the disclosure of results for areas whereconcentrations of arsenic are higher than the SCTL.However, sampling from the public ut i l i ty right-of-wayoften leads to the revision of the exclusion criteria usedin deciding which areas are suitable for samplecollection (e.g. sampling from close to roadways).Nonetheless, right-of-way sampling is proposed herebecause it eliminates some of the complications arisingfrom sampling from private yards, hence simplifyingthe sample collection process.

6. Sample collection. At the sampling site, the soilsample was collected using an Arts ManufacturingSupply (AMS) stainless steel soil recovery probe havingdiameter of 2.5 cm (Forestry Supplies Inc. Jackson,MS, U.S.A.). A probe with such a small diameterminimized soil disturbance associated with sampling,in some cases facilitating obtaining permission forsampling. A mallet was used to drive the probe into thesoil to the specified depth and the probe was pulled outby hand. Three plugs, located within half a meter ofeach other in a triangular pattern, were collected andcomposited at each site. Potting soil and or plugs wereused to fill any holes created during sampling. Properlylabeled soil samples were stored in an ice chest prior tobeing transported back to the laboratory for analysis.

Soil sampling followed current state regulations(FDEP. 1999) that require soil samples to be taken toa depth of two feet or 60 cm (wherever possible). Eachcore sample was split into three sub-samples dependingon depth: 0 20 cm. 2(1- 4(1 cm. and 4<i <>0 cm and thethree cores were composited for e.ich depth. Thesesampling depths were ma in t a ined t h r o u g h o u t the s tudyexcept in areas where distinct diso 'ni aunties wereobserved w i i h i i i a g i v e n 2 i i - em i n t e r v a l . i - < -.uch c.tsCvthe samples w i t h i n the 20-cm i n t e r v a l were split in totwo or more samples depending on the nature ofdiscontinuit ies . Some sites, especially in the commer-cial areas, had extremely hard subsurfaces, a conditionwhich was exacerbated by the spring early summerdrought of 2000. Thus, only ihe top 20 cm of soil (orthe accessible d e p t h ) were collected from these sites.The sampling depths were later refined to 0 10. 10 30.and M> 60 cm so tha t the surface samples w o u l d heller

Assessing Arsenic Background Concentrations in Florida Urban Soils 145

reflect the conditions at the surface without mixing orincluding too much subsurface soil. Data from the 1-30 and 30-60 cm depth would be composited later(using simple proportion) with surface soil (0-10 cm)data for comparison with State regulations. Differentstates have-different regulations pertaining to samplingdepth and appropriate adjustments are recommendedin these cases. However, it is recommended that the top10 cm of soil be taken separately as these represent thesoil to which land life is continuously exposed.

The precision of our sampling technique waschecked by collecting duplicates every fifth sample.Location information (including the logging of pos-itions using GPS) was taken at the time of sampling toavoid misrepresenting the sample locations. The valueslogged using the GPS receiver had a unique identifierthat allowed only the sample collectors to identify thesites. This was done to maintain the anonymity of thesample locations.

Specific exclusion criteria were applied and differentadaptation techniques were used in each land useduring sampling (Appendix 1). For example, most ofthe commercial areas are covered by paved surfaces.Samples were collected from the exposed soil surfaces todetermine the impact of these developments on soilarsenic levels. However, most of the soils in these areasare actually potting soils from commercial sources.Exposure to humans in these areas may not be a majorissue as most of the surface is covered. In all cases,contingency plans were in place in the event that permis-sion to sample was denied. In cases where permissionwas denied or sampling could not be done for otherreasons, samples were collected from the right-hand-side of the properties as the first option (unless theywere at the end of the street, in which case samples werecollected on the left-hand side). Permission was soughtin the same way for all alternate sites.

A data collection sheet was used during sampling todetail the location, longitude and latitude, age andhistory of the site, vegetation, and presence anddistance to chromated copper arsenic (CCA) treatedpoles and fences. All samples were coded so that onlythe principal investigators would be able to identify thelocation of any given property. This was done byhaving different labels on the sample collection sheetsand bags for each sample. A separate file tha t related totwo sets of labels was kept in a discrete place.

7. Sample digestion and analyses. It is important tofollow an approved QA/QC plan from the beginning ofthe study. In Florida, there are set protocols (approvedhy I DEP) thai must be .followed in oulei u> produceacceptable data for these types oi an . i l>>cs . In general,q u a l i t y assurance procedures start at the beginning ofthe study during the selection of field samples and(.oiUmuo in to the labora iory . S l .md . i i v ! U . ' i ' t ' i a l o i yprocedures included spikes, blanks, dupl icates andstandard reference materials representing approxi-mately 20% of total samples in the pilot s t u d y . Anapproved QA/QC plan was also used (( hen. Ma andHarris. 1999). All samples were air-dried, passedthrough a 2-mm sieve, and digested using USEPAmethod 3051 a (comparable to USEPA method 3050.the hotplate digestion method). In summary , 0.5-g soilsamples were digested with double acid (9 mL cone.

HNO, and 3 mL cone. HCI) in a CEM MDS-2000microwave sample preparation system (CEM, Mat-thews, NC. U.S.A.). The resulting solution was filteredthrough a Whatman No. 42 filter paper and diluted to100 mL. Concentrations of arsenic in the digestateswere determined using a SIM AA 6000 graphite furnaceatomic absorption spectrophotometer (GFAAS, Per-kin Elmer, Norwalk. CT, U.S.A.). In addition, soilproperties that have been shown to affect arsenicconcentrations (pH, clay content, total organic carbon,and total Fe and Al) were measured. Spike recovery, R.was used to determine matrix effects. It was calculatedas: R = [(xs — x)/.xs]100% where xs is the measuredconcentration of the spiked reference sample, and x isthe concentration of the soil sample. Batches in whichspike recovery varied by more than +10% from thetarget were redigested and reanalysed.

Due to low concentrations of arsenic and the sandynature of the soils in the study area, the proposed weightof soil to be used in the digestion procedure wasadjusted from 0.5 to 1 g for future studies to reduce thenumber of samples falling below the method detectionlimit. It is also proposed that, apart from the standardQA/QC protocols, five samples each of high and lowarsenic concentrations be sent to an independent labo-ratory for analyses for inter-laboratory comparison.

Accuracy was assessed by comparing the measuredconcentration (xs) of the reference standard material toits known value (x,) and computed on a relative basisas follows [(xs — x,)/x,]100%. Batches of samples forwhich the difference between the measured concen-tration of the standard reference material and itsknown value were greater than ± 20% were redigestedand reanalysed.

The precision in measurements was assessed usingthe coefficient of variation (CV), which is defined asthe standard deviation divided by the mean. Therelative percentage difference (RPD) between duplicatesamples, which was computed as: RPD = [2(xi — x^)/(x, + x:)] 100%, may also be used (USEPA, 1986) andthis value was set not to exceed +20%.

The instrument detection limit (IDL) of the instru-ment used in the analyses was determined as threetimes the standard deviation of 20 method blanks. Themethod detection limit (MDL) was determined as threetimes the standard deviation of seven samples that haveconcentrations approximately, four times higher thanthe method blanks. Breckenridge and Crockett (1995)give a good discussion on this topic. It is anticipatedthat the adjustments that were proposed in the weightof soil used in the digestion step and an increase in theAA aulos.tmplcr injection \olumes would reduce thenumber ol samples that fal l below the MDL. A moredetailed discussion on how these samples uerecensored is done by Porlier ( 2 < > i H ) .

X. Ciiuipini'i pruci'Miin;;. Numerous software appli-cations are available for statistical analyses. The mostcommonly used a IK! most powerful for our purposes isthe Statist ical Analysis Software (SAS", 2000). Thelatest version of SAS" (Version 8) allows the user to doQQ plots, perform tests for normal i ty , performtrans format ion and carry out both parametric andnon-parametric analyses. The generalized linear model(GI.M) was used in preference to the analysis of

146 Chirenje et at.

variance (ANOVA) procedure to account for theunequal number of samples within each class and QQ(quantile quantile) plots in SAS were used to eliminateoutliers from our sample population. The tests fornormality of both transformed and untransformeddata were carried out using the Shapiro-Wilks test.

Based on the geometric mean (GM) arsenic concen-trations, the 95th percentile concentration (95% of alldata fall below this value) and the 95% upper confidentlevel (UCL) of mean (the GM mean concentration fallswithin this range 95% of the time) for each land usewere calculated. These values were compared to thoseof relatively undisturbed soils obtained in a previousstudy (Ma, Tau and Harris. 1997; Chen et at., 1999).The 95th percentile and 95% UCL were determined inthis study because these values can be used asreferences in determining whether soil arsenic concen-trations are likely to be enriched or not.

Various software are also available for spatialanalyses and the most popular in this category isArcview M (ESRI K , Broadlands, CA, U.S.A.), whichcomplements very well with the very powerfulArclnfo", another ESRI" package. Version 3.2 ofArcviewR allows the user to do spatial analyses of thevariances to determine the relationship between thevariation of variance and other layers of spatialattributes. Another software package is usuallyrequired to transform the logged positions on theGPS unit into spatial data readable by Arcview^ or thespatial analysis software of your choice. This usuallydepends on the make and model of the GPS receiverused (e.g. Trimble Pathfinder").

Results and DiscussionIn general, sampling methods (based on objectives) andsample distribution (measured by skewness) determinehow the background concentrations are calculated. Fornormally distributed discrete samples, the backgroundlevel is calculated using arithmetic mean (AM) of thedistribution plus two standard deviations (Halmesci at.. 1998). The concentration obtained using this

method (mean -I- 2) is equivalent to the 95th percentileof the samples. The data in our case were lognormallydistributed. In such cases, the 95% upper confidencelimit (UCL) of the mean (n = 200) is calculated usingthe H-statistic from equation 2:

UCL,_, =exp(u> + 0-5S-+ 6-

x H,_, / [n- l ]°- s ) (2)

where ;i is the arithmetic mean of the log-transformeddata, (5 is the standard deviation of the log-transformeddata, n is the number of samples, H |_ 3 and H., are theH-slatistics for the upper confidence limits. The UCLdepends on f t v , n and the chosen confidence limit(Gilbert, 1987)".

Singh, Singh and Engelhardt (1997) provide anexcellent discussion on the treatment of environmentaldata from lognormal distributions using non-para-metric statistical procedures. They compared fivemethods (Jackknife, Bootstrap, Central Limit Theo-rem, Chebychev Theorem and the H-statistic) ofestimating the UCL and concluded that when thesample size was large (n > 100), all methods producedsimilar results. For smaller sample sizes such as thesizes of our individual strata (n = 40), they recom-mended the use of the Jackknife, Bootstrap or theCentral Limit Theorem to estimate the UCL becausethe other two methods (the Chebychev and H-statistic)tend to overestimate the UCL. The Jackknife andBootstrap procedures as discussed by Efron (1982) andMiller (1974) are recommended for studies similar tothe current study. Both methods do not requireassumptions about the dis t r ibut ion (they work forboth normal and lognormal distributions).

Results from the Gainesville pilot study

The summary statistics for arsenic concentrations inGainesville soils in the five strata analysed aresummarized in Table 1 and the general distributionand cumulative frequency curves are shown inFigures 1 and 2. The kurtosis (relative peakedness or

Tahlc I Summary <>/ ITAII//.V /nr I/if five categories unj their camhined effects

Commercial Public buildings Public parks Residential swales Residential yards Combined

No. of samplesAMASDMin imumMedianM a M u i ' i n icvSkc« nc>>.K u r i l »i.v" ' • I I S ' , ! • _ • K_'

",. -• Min."Loc- t rans lo i mod data

c,M(iSI)cvSkew IK-SIKur tos i s

I C L95th percentilc

401 . 1 92 2 3006II V*

( .</ ,

I . S 71 7I

' ; 4

ion

n 600 8N1 441 . 1 4.1.40

1 595 6S

440.570.540.01li ^4

in"1

0.971 844 'iK

^" i

4119

0.14I . . 1.1.1.94

-0980 9 1

1 011 65

380.520.670010 351.761.293.27

14623 "•57 9

0.232.58

11.42-0.70-0.36

I . . IS1 86

.190.680.500020 >()21170.740.980 24

1", t

41 2

0 50O.Sfr1 74

-1 142 66

1 051.6S

400680 560.01n >•)' w.

(I S21 4-)- is

n. 122 074 4 4

-1 763 .1.1

1 451 80

2010.7.31 1.10010 5')

d%1 556.39

57.626 l>41 8

0 19

1 574.01

-1 .171 98

099

3.53

o

I

Assessing Arsenic Background Concentrations in Florida Urban Soils 147

1000

100

10

1

0.1

0.01

50

40

30

20

10

0

(a)

«

• «

• •*

* * ••

30 60 90 120 150

Sample number

180 210

O Oc o o q

t D O o o o j ^ t q o q i - i Id o r H ^ ~ ^ ~ ° 3

Class (mg/kg)

6o

Class (mg/kg)

Higurc I (a) As concentrations in the study area, (b) Distribution of As in all areas, (c) Cumulative frequency curve for [As| in all areas

flatness of a distribution compared with the normaldistribution) of the data sets were all positive, withdistributions from residential areas (both yard andright-of-way samples) showing a tendency towardsflatness (Table I). The skewness of the data sets, thedegree of asymmetry of a dis t r ibut ion around its mean,were all positive indica t i iu 1 (rut all our d is t r ibut ionshad asymmetric 'ails • • v i . - n < ! > n - : ' ! , > • . > . > - d nioiv positivevalues (posi t ively skewed) . Th'.s i1- rv>i expected inbackground concent ra t ions of '.race clement studiesK'-i lber ' . . 1'iS"). t h u s t h c G M ar.1 GSH uerc used in "Hsubsequent calculat ions of the I 't'L. and 95th percen-ti le.

Results from G a i n e s v i l l e MIL 'LVSI t h a t r i gh t -o f -waysamples can be used in place of yard samples(F igure 3). However, dilferenl cities may not exhibi tthe same pat tern . R i g h i - o l - w a y samples are recom-mended because they are not only more practicable andeasier to sample, but are also equa l ly representative ofresidential areas ( s p a t i a l l y ) . There were a few outl iers in

the public building [four values (107, 79, 37 and36 mg/kg As, Figure l (a))] and commercial area strata[one value (656 mg/kg As, Figure l (a»] , which wereexcluded from the mean computations Outliers (in ourcase, areas suspected to have high concentrations dueto contamination) were excluded because they wouldhave shifted the mean u p w a r d s , exagge ra t i ng thevar ia t ion, confidence l imiK and l *^ih per^eniile >M~ th%

populat ion. The methods for i i l e n i i f y i n g o u t l i e r * arcmostly s ta t i s t ica l (e.g. Shapiro W i l k s lesi or QQ plots ,l - ' iynrc 4>. but there are often »I|KT fac 'or - . i n i ' h i d t t i L 1

site h is tory , thai could he used in de te rmin ing whethera site should be considered an out l ier or representativesample. I 'or l icr ( 2 < M . ) l l discusses v a r i o u s methods ofcensoring values both on the high end (outliers) and onthe lower end (below MDL) .

An average of ~27°o ( f a b l e 1) of the t o t a l samplesfrom Gainesvi l le were above the Florida SCTL of0.8 mg kg a l though 90% of these samples were below1.4 mg kg (Figure Ho). Onl\ 3% of all samples (67°, ,

148 Chirenje et al.

OOK

OO'OOI-tO'Ol

oo'oi-ioz:

OO'Z-18'l

09'i-iM 3?B.

02'I-IO'I O

OOl-18'O

080-19'0

ISO-0

o o oIO Tf PO

83|duiB8 jo jaqiunfvj

o o o o o o10 •* co N i-isajduiBS jo .raqumN

D•g

i!c1

[J

dcz

cd

tzz:

f<*»>' -ii

r— 1i^ n

OOK

OOOOl-TO'OI

OO'OI-IO'g

00 2-18' t

O8'l-t9'l

09'T-ua ^

oz-i-io-i G

OO'I-18'O

08'0-I9'0

09'0-Kf'O

[

C

d

:cz:

[Z

cz?

f1

[

[—1

•^ , '— -, — , •• ,

00l<

OO'OOT-IO'OI

OO'OI-IO'Z

00'2-I8'I

08'T-I9'I

09't-H-'I ,|

Wl-Wl |

02'T-IO'T 0

08'0-I9'0

C.

Ic"

n1

0

•f.73

-

CC

O

n

.>

f .'1

-.C

IllC

,1

1 1

g

Assessing Arsenic Background Concentrations in Florida Urban Soils 149

0.5 1 1.5

[As] (mg/kg) Yards

Figure _V The correlation between concentration of arsenic in yard and righl-of-way samples

100

0.1 10.0

Log total [As] (mg/kg)

1000.0

I

(b) R2 = 0.97

O O

R2= 1.00

-4 -2 0

Normal quantile

•al;nr! :>f ,'rsenic in .ill lour •ilr;il;i Parks (.IXH '" i. pnbl'.c b-"' 00 I'lv! of 'I'.'' cunihuK-d il.it.i

•••_•! .ir.Mi ( 4 1 ) ( '.. I rcsuk-nt ' i !

of ihcsc samples came from the commercial areas) wereabove the SCTL of 3.7 mg kg for commercial areas.An imp ' . i r i an l lesson from the pilot study was thepresence of a large fraction of samples lying below theMD1. (--40%). This led to the revision of the sampleweight that was used in the digestion step. While thisw i l l nut guarantee the complete elimination of below-MDL samples, it wi l l def ini te ly reduce the number ofsamples thai fall below the MDL.

Porher ( 2 l ) i i l ) discusses the implications of variouss t a t i s t i c a l techniques on the f inal results of backgroundstudies in a companion paper in this issue. There aremul t ip le objectives in the analysis of sample-baseddata . I ' I t ima lc ly the sampling program musl providean est imate of d i s t r ibu t ion characteristics of back-ground concentrat ions of arsenic in urban soils. It isimportant to point out that background concentrations

ISO Chirenje et at.

0.5

0.4

0.3

&

0.2

0.1

0.00 2 4 6 8

Natural log concentration (mg/kg)

Figure 5. Histogram of Gainesville data with smooth density estimate.

10

describe widely distributed chemicals present in theenvironment due to non-point sources. It is impossibleto establish directly the natural background concen-trations of arsenic in today's environment since all soilsin the urban environment are influenced by humanactivities to some extent. How this distribution isestimated will depend on a number of factors, not leastof which is the fraction of samples that occur belowMDL. In addition, we need to discuss how thedistributions constructed for the strata are combinedinto an overall distribution that reflects the relativefraction of total urban area represented by eachstratum.

The approach to estimate the dis t r ibut ion character-istics of background arsenic concentrations is to firstapproximate the probability density function of theconcentration by applying a non-parametric densityestimate to the observed data. This smooth estimate ofthe density function allows us to i d e n t i f y where the bulkof the background distribution lies as well as examinehow below MDL values affect the distribution. Inaddition, we can look at the distribution to determine ifthe observed concentration values might have comefrom a site that we would classify, in retrospect, assuspected of enrichment with arsenic. This primarilygraphical view of the density curve provides usefulguidance as to how the data might be censored prior tousing cen.sored dat;i es t imat ion ' - L : i n i v . | i : j - . 10 lit aformal parametric distribution function. normal' to log-transformed concentrations. This f i na l t i l l ed parametricd i s t r i bu t ion function is llv K.^1 e s t i m a t e o'~ -ir^ 'f i1 . -background concentration d i s t r i bu t i on Details of thisapproach, with graphical support for the discussion,follow.

Estin\uiin% a tlixtrihniiiin time lion In theory, if \\chave a set of concentration values, say X, . X,Xn. collected at random from the same populat ion, aroueh estimate of the density of the d i s t r ibu t ion is

given by a histogram. Given an origin X0 and a positivewidth h, we define the histogram intervals to be thepartition of [X0 + mh, X0 + (m + l)h) for positiveintegers m = 1 , 2, 3 ... The histogram is defined by:

/(.v) = — - (no. of Xj in same int as .v) (3)tnh

A histogram for data from the Gainesville pilot study isgiven in Figure 5. Another way of writing thehistogram function is to represent it as a weightedaverage of observed data within some distance h of thex for which the density is to be computed. Thus eqn (3)can be rewritten as:

where

w(x) =- if |x| < ]

0 otherwise(5)

This essentially constructs intervals of width 2h.assigns weight (2nh)" ' to each observation, thensums up each observation that is within h units of x.

A histogram is a nai've estimator of the probabilitytLnsiU func t i on of a cont inuous random v a i i a h i ^ ,\belter e s t ima to r is obtained bv replacing the weight (orKernel i funct ion in eqn (5) \v i th a smoother f u n c t i o nt h a t is defined for the complete range of x. In t h i s \ \a \[ l i L i c s i i i i i ng cu rve looks like the smooth i>l ihohistogram as presented by the smooth line in l-'igurc 5.The constant h can be adjusted to produce ver\smooth to ver\ rough density estimates hence it iscalled the smoothing parameter (also wi th referencesto time series analysis from which these techniquesderive, h is also referred to as the bandwidth of thesmoothing func t ion ) . This smooth density es t imate issometimes called a non-parametric es t imate . Thechoice of the smoothing funct ion. w(x) is U p i c a l l y

Assessing Arsenic Background Concentrations in Florida Urban Soils 151

0.15 -

0.10 -

0.05 -

0.00 -

Concentration

0.30

0.25

0.20'

-a a15

o.io

0.05

0.00-

(b)

- 4 - 2 0 2 <Concentration

Figure 6. Mixtures of distributions. The solid top line is the smoothed distr ibution for the mixture.

not that important. For il lustrative purposes, aGaussian function is used.

w(x) = (6)

The smoothed density curve illustrated in Figure 5 istypical of what we can expect from the data resultingfrom the sample survey proposed here for the followingreasons. Conceptually we can envision the urbanarsenic population as a mixture of three populations.Centered more to the left on the scale of arsenicconcentrations is the natural background population.These are areas that , for what ever reason, have beenl i t t l e affected by the activities of humans and hencedisplay very low coii^nualuii i lc \ols . l. 'nfui t t in . i lc l )(or fo r tuna te ly) Ihcsc levels arc o f ten so |»v\ a^ u> bebelow ihe MDL lor the iLV/nmjLic-, typical!) u.scd tomeasure arsenic in soil. This equipment- induced leftcensoring of l i io d i s t r i b u t i o n i iukc-, n \ e r \ d i l l i c u i l toget a good estimate of the n a t u r a l backgrounddistribution.

On the right side of the scale of arsenic concen-trations arc values from areas t h a t have been enriched(poten t ia l ly contaminated) beyond background levels.While the sampling study has been designed to avoidsuch sites as much as possible, there is no avoiding thefact tha t we do not and cannot know of all these sites,hence there is a lways a posi tuc probabili ty of

observing high concentration levels in the smootheddistribution. Since we go to a lot of trouble to keepthese sites from being sampled, there would only be afew such data points in the observed data, hence itwould be very difficult to estimate this distr ibution wi thany precision.

Of direct interest in this study are the arsenic-concentration values that are in the "middle" of thedistr ibution. Because of the study design, these valuesshould be represented by the middle mode of thedis tr ibut ion. By definition, we will call th is middledis t r ibu t ion the "distribution of anthropogenic arsenicconcentrations" for the urban environment. Thediff icul ty is determining where this middle part beginsand ends On its left side, the anthropogenic values arcnii\ . ,! wi th the upper range ol lK n a t u r a l backgroundd i s i t t i ' M t i o n . On its right side, the .mihropogenn:values mix w i t h the lower range of the enrichedpopi;!:mon. If 'hese three p o n ' i l a i i o n s \ \ .TC 10 besampled in equal proportions, the conceptual under -lying d is t r ibut ions and the observed data d i s t r i b u t i o nwould be like that presented in Figure N J ) . As thefraction of anthropogenic to na tu ra l and enrichedsamples increases, the d i s t r ibu t ion becomes more l ikethai in Figure 6 ( b ) .

The statistical goal in this s tudy is how lo gel a goodq u a l i t y est imate of the anthropogenic d i s t r i bu t i onfrom a sample "contaminated" wi th n a t u r a l and

152 Chirenje et al.

enriched data. The approach we propose is discussedby Porticr (2001) and includes the following:

(1) How does left censoring of concentrations belowMDL and reassignment to 1/2 the MDL affectthe estimation of the distribution? [This can beapproached by sensitivity analysis, e.g. startingwith MDL = 0, MDL = 1/2 MDL, ANDMDL = MDL];

(2) How do we get a good estimate of the centercomponent of the distribution;

(3) What are Type I and Type II errors in thisprocess? How would they be estimated? What arethe ramifications of making these errors?

(4) How will the data from the four strata becombined to estimate the overall anthropogenicbackground distribution (i.e. incorporating thesampling weights)?

Summary

The results form this study help us understand theexpected distribution of arsenic in other small citiesthat are on the same type of geological material asGainesville and have limited sources of contamination.Although these results may not be used to predictarsenic concentrations likely to be found in Miami orother big cities, the experience of going through thetasks for this study helped in developing a protocolthat can be used in carrying out other backgroundstudied in cities of different sizes and locations.Although the methods presented in the proposedprotocol are dependent on the objectives of thebackground study in question, the protocol can beeasily modified to accommodate different objectives.The improvement in accuracy of digital data and therapid development of GJS techniques of both digitaland non-digital data will make it easier to characterizesampling sites more efficiently and add to the reliabilityof the results. In the meantime, careful search ofrecords should be done so that meaningful inferencestha t are not influenced by unknown site history, can bemade about the distribution characteristics of arsenicconcentrations in urban areas.

ReferencesBarren. I 1987. Research in urban ecology. Rcporl to the Nature

Conservancy CouncilBreckenridgc. R P. and Crockett. A.B 1995. Determination o/

btiikground toncfiiti-titions of inorganic* in soils and sediments atI,.,-,':,--!,;,. u,/.v ,./,•, TPA 54f) S-96 SOO

Hi ' t > . • • • i - . . u i''"1 ! vj pollution in soil- ndjacciil to hews inl.:mr\:. ! !i>r:.!;: h.i:vinat Ginc/u-m Health 16. 59 6-4.

K n n k m . i i m . K .irni Kv.n i . J 1997. Background cunieinninoiis nl;i.v.pi/.'. "! ii.niJu \ i ' i l < Draft report.

Chen. M . Ma. L Q. and Harris, W.G. 1999. Baseline concentrationsof 1? trace elements in Florida surface soils. J. Environ. Qua/. 28.1173-1181

Chen. TB. Wong. J.W.C . Zhou. HY and Wong, MH 1996Assessment of trace metal distribution and contamination insurface soils of Hong Kong. Environ. Poll. 96, 61-68.

Chen, M . Ma. L Q.. Hornsby. A.G. and Harris. W.G. 1999.Background concentrations of trace metals in Florida surface soils:la.xonumic and geographic distributions of total-total and totalrecoverable concentrations <>/ selected trace metals. Florida Centerfor Solid and Hazardous Waste Management 99-7

Conover. W.J. 1980 Practical Nonparametric Statistic*. New York,John Wiley & Sons.

Craul, PJ 1985 A description of urban soils and their desiredcharacteristics J Arboriculture II. 330-339.

Davics, D J A . Walt, J M and Thornton, I 1987. Lead levels inBirmingham dusts and soils. Sci. Tot. Environ. 67, 177-185.

Efron. B 1982 The Jackknife. lite Bootstrap and Other ResamplingPlans Philadelphia. SI AM.

FDEP (Florida Department of Environmental Protection). 1999.Contaminant Cleanup target levels. Chapter 62-777. Tallahassee.FL.

Fields, T W , McNcvin. T.F., Markov, R.A. and Hunter, J.V. 1993. Asumrnarv of selected soil constituents and contaminants at hack-ground locations in New Jersey. Interim Final Draft New JerseyDepartment of Environmental Protection and Energy. Siteremediation program and division of science and research.

Gilbert, R.O 1987 Statistical Methods for Environmental PollutionMonitoring. New York. NY, John Wiley & Sons.

Halmcs. N.C., Tonner-Navarro. L.E . Porticr, K.M. and Roberts,S.M. 1998 Soil sampling /or contamination assessment Technicalreport 97-04 Center for Environmental and Human Toxicology.Universi ty of Florida.

Juan, C S. 1994. Natural background soil metal concentrations inWashington Stale Publication Number 94-115. Toxic CleanupProgram. Olympia. Washington.

Kabata-Pcndias, A. and Pcndias. H. 1992. Trace Elements in Soilsand Plants Boca Raton, CRC Press.

Kaminski . DM and Landsbcrger. S. 2000. Heavy metals in urbanareas of East St Louis. IL. Part 1: Total concentrations of heavymetals in soils Air & Waste Manage. Assoc. 60. 1667-1679.

Ma. L Q . Tan. F and Harris, W.G. 1997. Concentrations anddistributions of eleven elements in Florida soils. J. Environ. Qual.26. 769-775.

McGrath. S P 1986. The range of metal concentrations in topsails ofEngland ami H'a/es in relation to soil protection guidelines. Proc.Trace Suhst Environ. Health Columbia. Missouri

Miller. R 1974 The Jackknifc a review. Biomelrika 61, 1-15.Portier. K 2001 Statistical issues in assessing background concen-

trat ion of arsenic in urban areas Environ Forensics 2.Singh. A . K . . Singh. A. and Engclhardl. M 1997. The lognormal

distribution ill einirorimi'titiil applications. EPA Technology Sup-port Center Issue EPA.600 R-97 006.

Schwar. M.J R . Moorcroft. J.S., Laxcn. D.P.H.. Thompson. M. andAmorgie. M. 1988 Baseline mctal-in-dust concentrations inGreater London. St i Tot. Environ. 68. 25—43.

Thornton, I 19X7 Metal contaminat ion of soils in urban areas. In:Soils in the i'rbun Emironment. pp. 47-85. (Bullock, P. andGregory. P J.. Eds) London, Black well Scientific Publications.

Tiller. K G 1992 Urban soil contamination in Australia. Ausl JSoil Res. 30. 937 957.

USEPA (U S Envi ronmenta l Protection Agency). 1986. Field manualIn, ••nl •..I'lK-li,,,, ,.i /<( 'K </W/ -.Hi- in ii-ri ' i ,h\inu/> EPA5M1 S-S6-I)!7 O!ti'.: "! ' . - v - • --M!,.; ..,„.-, \V:«hin-:li'll IX

\Vikkc. \V . Mul'.r. S . Kancli.i:-,;;k.<..)l. N. ;:nd /cch. \V. I<M7. Urbansoil conijMiiUjniMi in R.ingkok. hca\> metal and aluminumpartitioning (n--i.k • • • i t - . i 86 21 I 228

Assessing Arsenic Background Concentrations in Florida Urban Soils 153

Appendix 1

Criteria for exclusion

Sample site selection within the selected sampling units were subject to these elimination rules:

(1) An area within I m (3.3 ft) of a paved surface (concrete pavement or tarmac):(2) An area within 3 m (10 ft) of a CCA-treated wood pole (arsenic from the CCA treatment may have leached

from pole into the surrounding soil);(3) An area wi thin 2 m (6.6 ft) of a CCA-treated fence (see 2);(4) An area within 3 m (10 ft) of a CCA-treated wood deck (see 2);(5) An area currently or formerly under a CCA-treated wood structure (see 2, background or historical records

used to determine previous land use);(6) An area wi th in 100 m (330 ft) of the perimeter of a gas station and obvious pathways for runoff (pollution

from automobile exhaust and petroleum fuel carried away by runoff);(7) An area within 200 m (660 ft) of the outer limits of the boundaries of an identified former cattle dip site and

obvious pathways for runoff;(8) An area within 100 m (330 ft) of the boundaries of an identified "contaminated area" (determined using

background and historical information);(9) An area within 500 m (0.3 miles) of the boundaries of an identified site of hazardous waste dumping or

processing;(10) An area within 100 m (330 ft) of the boundaries of an identified waste treatment site;( 1 1 ) An area within 100 m (330 ft) of an industrial area (operationally defined as an area where manufacturing or

processing activities are taking place);(12) An area within 20 m (66 ft) of a major road as defined by the Florida Geographical Data Library (11 m

[33 ft] width). [Most roads from which right-of-way samples are collected are classified as minor roads(average 6 m width)];

(13) An area that is a part of a runoff flow depression;(14) An area that is a part of a school (schools are politically sensitive areas and collecting and publishing data

from them can engender public or media attention that detracts from the main purpose of the study);(15) An area that is a part of an urban playground (see item 14);(16) An area within 100 m (330 ft) of established places of worship (such as churches, see 14);(17) An area w i t h i n 100 m (330 ft) of a gold course (intensive management practices likely lead to enhanced

arsenic concentrations);(18) An area within 100 m (330 ft) of power substations (likely to have enhanced arsenic concentrations);(19) An area within 100 m (330 ft) of a current or former railroad corridor (likely to have enhanced arsenic

concentrations).