EPA/600/R-93/073
March 1993
AN X-RAY FLUORESCENCE SURVEY OF LEAD
CONTAMINATED RESIDENTIAL SOILS
IN LEADVILLE, COLORADO:
A CASE STUDY
by
C.A. Kuharic and W.H. Cole
Lockheed Environmental Systems and Technologies Company
Las Vegas, Nevada
and
A.K. Singh and D. Gonzales
University of Nevada, Las Vegas
Harry Reid Center for Environmental Research
Las Vegas, Nevada
Project Officer
Kenneth W. Brown
Technology Support Center for Monitoring and Site Characterization
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89193-3478
Printed on Recycled Paper
-------
NOTICE
The information in this document has been funded wholly or in part by the United
States Environmental Protection Agency under Contract No. 68-CO-0049 to the
Lockheed Environmental Systems and Technologies Company. It has been subjected
to the Agency's peer and administrative review, and it has been approved for
publication as an EPA document. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
ii
-------
TABLE OF CONTENTS
NOTICE «
TABLE OF CONTENTS .."*
LIST OF FIGURES v
LIST OF TABLES vi
ACKNOWLEDGEMENTS vii
FOREWORD 1
INTRODUCTION 2
Background 2
History 2
Site Description 3
Migration of Waste Materials 3
The Role of EMSL-Las Vegas 4
PRINCIPLES OF X-RAY FLUORESCENCE 4
THE X-MET 880 5
THE KEVEX 770 5
INSTRUMENT CO-CALIBRATION : 5
SAMPLE HANDLING 8
FIELD SAMPLING 8
SAMPLE PREPARATION 8
The First 850 samples 8
The Rest of the Samples 9
DATABASE MANAGEMENT 9
QUALITY CONTROL 9
INTERNAL CONSISTENCY 10
QUALITY CONTROL CHARTS 10
CLP VERSUS X-METs 11
PRECISION, ACCURACY AND DETECTION LIMITS 13
TOTAL VERSUS ESTIMATION ERROR 16
»
FALSE POSITIVES/FALSE NEGATIVES 16
-------
POPULATION TESTING
CONCLUSIONS
REFERENCES ...
17
20
21
IV
-------
LIST OF FIGURES
Figure 1. Calibration results for the Kevex and X-Mets 1, 2, and 3 7
Figure 2. Correlation plots between Kevex and moist/dry X-Met. 10
Figure 3. Quality control charts for Kevex and X-Met #1 . 12
Figure 4. Correlation plot between CLP and combined X-Mets, and Kevex 13
Figure 5. Frequency distribution of %RSD for moist field duplicates 15
Figure 6. False negatives and false positives 18
V
-------
LIST OF TABLES
Table 1. False positives and false negatives 17
Table 2. List of results of t-test and Kolmogorov-Smirnov (KS) test 19
vi
-------
ACKNOWLEDGEMENTS
The authors are indebted to Mr. Ken Wangerud, the Remedial Project Manager,
personnel from the Roy F. Weston, Inc., Denver office, and to Dr. Rex Bryan and Dr. John
Drexler for their valuable technical suggestions and advice.
Vll
-------
-------
FOREWORD
Continuing concern over the adverse impacts to human health due to exposure to
lead has prompted characterization efforts at numerous sites across the United States. One
of the primary potential exposure routes is through ingestion and inhalation of lead
contaminated soils. This problem can be serious at old mining and smelting sites, especially
in the western United States.
The California Gulch Superfund Site in Leadville, Colorado was added to the
National Priority List (NPL) in 1983. It is an historic mining and smelting site that is
currently the focus of extensive studies on soil lead contamination and bioavailability.
Studies of ground and surface water impacts are also in progress. Personnel from the
Environmental Protection Agencies (EPA) Environmental Monitoring Systems Laboratory
at Las Vegas (EMSL-LV), the Denver office of Roy F. Weston, Inc., and Geostat Systems,
Inc. (GSI) used field-portable X-ray fluorescence (FPXRF) to determine the spatial
distribution of lead concentrations in residential soils.
This report details the FPXRF program sample collection, preparation, and analysis
procedures, database management, and program quality assurance efforts at Leadville. The
program clearly demonstrates that small, field portable XRF instrumentation can produce
large quantities of acceptable quality data in a timely and cost-efficient manner when used
properly.
When combined with the results of blood lead level and bioavailability studies, this
data can help to develop a true assessment of the risks posed by lead in the residential soils
of Leadville.
-------
INTRODUCTION
Background
During the summer of 1991, over 3700 soil samples were collected and analyzed for
lead content. These samples were collected from the edge of alley right-of-ways (alleyways)
because of access difficulties, and from individual yards. The samples were analyzed using
X-ray fluorescence spectrometry. A laboratory-grade Kevex 770 spectrometer was used to
corroborate the analyses from three field-portable X-Met 880 spectrometers, which were
then used to rapidly generate data of known quality.
Forty-five characterized Leadville residential soils were available at the onset of the
survey. Ten of these samples were used to co-calibrate the Kevex 770 and three different
X-Met 880 instruments, and the remaining 35 were used to confirm the calibrations. The
first 850 samples were measured field moist after being brought to the field laboratory.
They were then oven-dried and sieved, and approximately six grams of each subsampled into
polyethylene X-ray cells. These prepared samples were analyzed on the Kevex and the X-
Met 880s. Some of these samples were randomly subsampled for Contract Laboratory
Program (CLP) analysis. Data comparability was demonstrated by correlating X-Met and
Kevex results to each other and then to the CLP results.
All remaining field samples were analyzed field moist on one of the three X-Met 880
instruments. One sample from each preparation batch of 30 samples was selected for Kevex
analysis and samples for CLP analysis were randomly selected from the entire sample suite.
Results of the survey indicated that the X-Met 880s produced good quality data.
From the beginning, two major issues were of concern. First, was whether a single
matrix model for the X-Mets would be adequate to analyze all soils within the area to be
surveyed. Comparison of the X-Met and CLP data for the same samples indicated this to
be the case. Second, was whether the alleyway samples and the nearby residential yard
samples were of a single population. The non-parametric Kolmogorov-Smirnov procedure
and the paired-wise sample t-test indicated no significant differences in the alleyway and yard
samples; thus inferences could be made in unsampled yards from nearby alleyway easement
samples.
History
Mining activity in the vicinity of Leadville was traced to the 1859 discovery of placer
gold in California Gulch. The decline of the placer deposits led to the search for and
discovery of lode gold in 1868. This discovery failed to reverse the decline, but in 1874 a
heavy mineral that had been interfering with the placer operations was identified as silver-
bearing lead carbonate. The ensuing silver boom was such that by 1880, there were enough
mines to keep a number of smelters in operation.
-------
Eventually, the mining and processing turned from the oxidized, carbonate ores to the
less desirable sulfide ores of the more mundane metals, lead and zinc. Activity since has
been cyclic, following trends in the base metals markets. Total metal production to 1966 was
about 24,000,000 tons. There is currently only one active major operation in the area,
ASARCO's Black Cloud Mine. The last smelter ceased operations in 1960.
The extensive underground mining activity left a legacy of drainage tunnels, perhaps
most notably the Yak tunnel, and numerous mine waste rock and tailings piles dotting the
landscape, almost entirely to the east and upslope of the town. The processing activities,
which were scattered around the town, left behind large amounts of slag. Owing to the
relative inefficiency of the recovery methods of the time, much of the waste material
contains high amounts of lead and zinc, along with some other metals.
A preliminary EPA site evaluation took place in 1982 and 1983, and the site was
placed on the NPL in 1983. The initial concern was on the effluent from the Yak tunnel.
A surge pond and treatment system have since been constructed to deal with this aspect of
the mining legacy. However, as is sometimes the case around old mining districts, material
has been taken from abandoned dumps for use as fill, slag has been used as railroad ballast,
and it has been crushed for use on streets in the winter. This use, taken in conjunction with
the spread of smelter dust and wind transport of fine grained particles from various dumps,
prompted concern about the effects of lead on people living in the town, especially the
children.
Site Description
The town of Leadville lies about 100 miles southwest of Denver. It is situated on the
western slope of the Mosquito Range, just east and upslope of the Arkansas River, at an
elevation of approximately 10,200 feet. The mining activity was predominantly on the
eastern side of town, along the drainages of California, Evans, Stray Horse, Oregon, Malta
and Georgia Gulches. The area lies within the Colorado Mineral Belt and is highly
mineralized(1).
The soils in the affected area are sandy loams to gravelly sandy loams, with varying
amounts of admixed slags and tailings. The Leadville sandy loam predominates, with lesser
amounts of Pierian and Troutville gravelly sandy loams present(2). Large piles of slag and
tailings are also scattered around the site.
Migration of Waste Materials
Contaminant transport occurred from natural processes, such as wind and water
dispersion, but in Leadville as previously stated, the problem was exacerbated by human
impacts, ranging from smelter stack emissions to slags that were crushed and spread on icy
roads and used for railroad ballast. Tailings and mine waste rock were thoroughly mixed
into local soils by over a century of mining and commercial activities where these materials
-------
were often used for fill material.
In addition, many homes are built among the waste rock/tailings piles, or immediately
adjacent to them. The tailings piles in particular are used as recreational areas by many off-
road motorcycle riders and are criss-crossed with trails. Children also have access to these
areas as unsanctioned playgrounds.
The Role of EMSLr-Las Vegas
At the request of EPA Region 8, an intensive effort was mounted by the EMSL-LV
Technology Support Center to gather a large amount of data on the surficial (0-4") spatial
distribution of lead in the residential soils of Leadville during the 1991 field season. Owing
to the rapid and economical approach afforded by FPXRF equipment and methodology, the
EMSL-LV FPXRF team became an integral part of the program design group, along with
personnel from Roy F. Weston, Inc., Denver, and Geostat Systems, Inc., of Golden,
Colorado.
The study was designed to achieve three primary goals. First, to demonstrate the
ability of the FPXRF instrumentation to generate quantitative data of known quality.
Second, to gather sufficient data to allow the use of geostatistics to determine the optimal
sample spacing for further sampling (to be determined by Geostat Systems, Inc.). Third, to
define areas of contamination in specific concentration ranges of <500 mg/kg, 500-1500
mg/kg, and >1500 mg/kg.
To achieve these goals prior to field work, EMSL-LV participated in the writing of
a Sampling and Analysis Plan and a Quality Assurance Project Plan as part of the Workplan
for XRF(3). Once the plans were written and accepted, the EMSL-LV mobile XRF
laboratory was moved to Leadville, and a temporary field laboratory was set up for sample
preparation and analysis. In addition, a database management system was designed and
implemented to handle the large amount of data generated by the XRF program.
Once on site, the next step was to run a feasibility study to demonstrate that the data
generated by the FPXRF instruments would meet the data quality objectives of the project.
The remainder of this report is dedicated strictly to XRF aspects of the program.
PRINCIPLES OF X-RAY FLUORESCENCE
XRF spectrometry is based on the principle that photons produced from an X-ray
tube or radioactive source bombard the sample to produce fluorescence. The incident
photons impinge on the electron cloud of the atom. Among other events, this process
creates vacancies in one or more of the inner shells. The vacancies cause instability within
the atom. As the outer electrons seek stability by filling the vacancies in the inner shells, the
atom emits energies as X-ray photons. The emitted energy (fluorescence) from a particular
shell is characteristic of the atom in which it was produced and is equal to the difference in
bonding energy between the outer shell electron and the vacant shell. Most elements under
-------
the photon bombardment fluoresce simultaneously to produce a spectrum of characteristic
radiation. It is this spectrum that the XRF detector senses and counts.
There are two types of XRF spectrometers, energy dispersive and wavelength
dispersive. The principal differences are in the method of detection of the fluorescent
energies of the specimen and the method of quantifying the analytes of interest. For more
detailed information on X-ray fluorescence, the reader is referred to Jenkins et_al.(4).
THE X-MET 880
The X-Met 880 used on the Leadville site is a field-portable, energy dispersive XRF
spectrometer marketed by Outokumpu Electronics, Inc., Langhorne, PA. The unit is self-
contained, battery powered, microprocessor based, and weighs 8.5 kg. The surface analysis
probe is specifically designed for field use. The X-Met 880 is hermetically sealed and can
be decontaminated with soap and water. The probe includes two radioisotope sources,
Americium-241 and Curium-244, a proportional tube counter, and the associated electronics.
The source is protected by a Nuclear Regulatory Commission approved safety shutter. The
electronic unit has thirty-two calibration memories called "model". Each model can be
independently calibrated for as many as six elements. Using a multivariate regression
procedure with the proper isotope sources, the instrument can be calibrated to measure
elements from silicon to uranium. Unknown sample intensities are then compared to the
calibration curves to yield quantitative concentrations.
THE KEVEX 770
The Kevex 770 is a laboratory-grade, energy dispersive XRF instrument marketed by
Fisons Instruments, San Carlos, CA. The instrument contains a 198 watt, Rh anode, liquid-
cooled X-ray tube, and a cryogenically cooled, lithium drifted silicon solid state detector.
Optimal excitation conditions for analytes ranging from Na to U can be achieved using tube
direct excitation with filters or using secondary targets for essentially monochromatic
excitation. The higher resolution of the Kevex (over five times that of the X-Met) allows
for excellent qualitative scans for analytes that interfere with the analytes of interest.
Calibration curves were chosen over Fundamental Parameters (FP)1 for quantitating
the analytes of interest. It was decided in committee that calibration curves based on a suite
of site specific calibration standards would be a more robust approach than using FP.
INSTRUMENT CO-CALIBRATION
Forty-five residential soil samples from a 1990 sampling effort by Walsh and
Associates, Denver, Colorado, were available with corresponding EPA CLP analytical results.
^Fundamental Parameters is a mathematical approach which resolves analyte interferences and
produces quantitative results for all analytes in the sample.
-------
Each sample was poured from its eight ounce jar into a pan, dried overnight at 100 °C, and
then sieved through a two mm sieve. The sample was then rolled 20 times corner-to-corner
on a piece of kraft paper down the long axis of the subsequent ellipsoid shape of the pile
of soil. Approximately ten, half gram samples were extracted from the pile of soil and
placed in a 31 mm polyethylene X-ray cell which was sealed with 0.2 mil thick polypropylene
film. Ten of these samples which spanned the full concentration range were used to "co-
calibrate" the four XRF instruments (site specific soil standards). The remaining samples
were used to verify the accuracy of the calibration procedures. There were four instruments
used in this study: one Kevex 770 and three X-Met 880s. Correlations between CLP results
and the results from the four instrument verifications are shown in Figure 1.
Site specific soil standards were used to compensate for physical (particle size, bulk
density, heterogeneity) and chemical (spectral) matrix effects that impact instrument
response when analyzing soils. Calibration curves for one matrix usually give incorrect
results when used to analyze samples of a different matrix.
-------
LEADVILLE PROJECT: INSTRUMENT CO-CALIBRATION
n=35; R-squar«=0.89
CLPPbwt*
0.8
0.7
o.e
0.5
0.4
0.3
0.2
0.1
0.0
I I
--i 1
! |
f~
1 «
;_
i
+ * ^ %*
j£ | +
* +?" *
T : :
:
.1- -t
+
i-
:
I
*
0.0 0.1 0.2 0.3 0.4 0.5 0.8 0.7 0.8
KEVBCPbWCK
CLP VS. KEVEX FQH LEAD
LEADVILLE PROJECT! INSTRUMENT CO -CALIBRATION
n=35; R square=0.83
OPPbwM
0 S
. 0.7
0.6
0.5
0 4
0.3
0.1
n rt
i i
+
*
%
!
f-
1 *
^ *±*
+ ^
=f-'
4*-
+
>
i i
+
:
:
"'"' t
ViJ i i l 1 1 1 j 1
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 <3
8
XMET1 Fbwtti
CLP VS. XMET rfl FOR LEAD
LEAD1
CLPPbwM
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0.
iflLLE PROJECT: INSTRUMENT CO-CALIBRATION
n=35; R square=0.85
^
- 1
w
^Jl.
.
+ 4
+
............
4-
^
, _
...
II
4
.
,,,
0.0 0.1 O.Z 0.3 0.4 0.5 0.6 0.7 0.8
XMEI2Pb*rt%
CLP VS. XMET « FOR LEAD
LEAD1
CLPPbwt*
0.8
0.7
0.6
0.5
0.4
0.3
0.1
VILLE PROJECT: INSTRUMENT CO -CALIBRATION
n=35; R-square=0.87
+..
+
]
t -
: 4-
|
I
* 1
**" * ]
{ -
+
O.O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
XMEI2Pb«vt%
CLP VS. XMET #3 FOR LEAD
Figure 1. Calibration results for the Kevex and X-Mets 1, 2, and 3.
-------
At Leadville, at least four different matrices were expected to occur: a carbonate
tailings matrix, a sulfide tailings matrix, slag, and local soil. An assumption was made that
since human activity over the last century thoroughly mixed these matrices in the residential
portion of the town, a single set of calibration curves could be used. The ten calibration
samples came from locations scattered across the townsite, and the success of the calibration
confirmation and subsequent CLP corroboratory analyses indicated that the assumption was
valid.
SAMPLE HANDLING
It was originally intended that once X-Met correlation to the Kevex was clearly shown
to be acceptable (i.e., R2 > 0.7) on a significant number of samples (800-1000), the X-Mets
would be taken into the field for in situ analysis. This was to be done for the sake of
expedience, but the field sampling crews proved so efficient that sample analysis continued
in the laboratory for the duration of the program. Access to residential lots was
extremely limited at the beginning of the program, so transects were surveyed on public
properties which were mostly along the unpaved shoulders and edges of alley rights-of-ways,
usually only a few feet from backyard property lines.
FIELD SAMPLING
Intrusive samples were collected at 25-foot intervals along the transects. At each
sample location, a volume of soil approximately six inches in diameter and four inches deep
was thoroughly disaggregated with a pick and shovel. The soil was turned onto itself seven
times to reduce heterogeneity and to allow a surface measurement technique to represent
a volume. This was the surface upon which in situ analysis was originally intended to be
performed. Approximately 300 to 500 grams of soil was scooped into a large plastic bag,
labeled and custody sealed, then double bagged and placed in a cooler for transport to the
field laboratory. All samples were obtained under chain of custody. Upon arrival at the
field laboratory, all samples were logged in and remained under locked custody.
SAMPLE PREPARATION
The First 850 samples
Each bag was opened and the field moist samples were analyzed in the bag three
times (by placing the probe in the bag), thoroughly shaking the sample in the bag between
analyses. This mode of analysis was deemed analogous to in situ analysis, if the program
had gone to in situ analysis.
Each sample was then transferred to a Pyrex loaf dish and oven dried overnight at
100 °C. The dried sample was passed through a ten mesh (2 mm) sieve onto a three by
three foot piece of kraft paper where the sample was rolled onto itself 20 times to reduce
heterogeneity. Approximately six grams taken in 10-12 subsamples was placed into a 31 mm
-------
diameter polyethylene X-ray cell and sealed with 0.2 mil polypropylene film. The cupped
sample was then analyzed on the Kevex and again on one of the three X-Mets. The
remaining dried soil was put into archive, still under chain-of-custody. All Pyrex loaf dishes
were decontaminated with an Alconox/distilled water solution, rinsed with distilled water, and
air dried.
All X-Met measurements were done in triplicate and the values averaged partly due
to the relatively short acquisition time (30 seconds livetime), but mostly to obtain a more
representative composite value of each sample. A sample is measured only once on the
Kevex due to the long acquisition time (200 seconds livetime2).
The Rest of the Samples
Once good correlation was established between X-Met and Kevex analysis (see Figure
2), only one in ten samples was dried, cupped, and analyzed on the Kevex. The X-Met
procedure continued in the same manner throughout the rest of the program; that is,
samples were brought in by the field crews, analyzed in triplicate in the plastic bag, and then
archived.
DATABASE MANAGEMENT
A database management system was developed using the Statistical Analysis Software
supported by the SAS Institute Inc., Gary, NC. Screens were designed to look exactly like
the data recording forms. Along with all the analytical data (all triplicate measurements and
associated QC samples), other data including easting, northing, analytical date and time,
sample type, instrument ID, and a unique serial number for each sample were tracked by
the management system.
Data were entered in batches (30 samples per batch) and a$ each batch was
completed, it was printed out and 'hand' checked against the data recording forms for entry
errors. All changes made to the database were tracked by a comparison procedure, and a
printout of each set of changes made to the data set was put into the hard copy dataset for.
documentation purposes.
QUALITY CONTROL
As the database grew, comparison plots were periodically generated between the
Kevex and the X-Mets to assess the degree of correlation, as was done with the co-
calibration, thus demonstrating internal consistency within the data collection effort. Once
the CLP corroboratory data were available, the same type of correlation was generated,
confirming the accuracy of the data.
2At an average deadtime of 40 percent, this equates to 280 seconds acquisition time per sample.
9
-------
INTERNAL CONSISTENCY
Figure 2a shows the correlation between dry Kevex results and moist X-Met results
and 2b shows the correlation between dry Kevex and dry X-Met results. Note that when the
samples are dried and cupped and reanalyzed on the X-Met, the correlation improves only
five percent. The X and Y axis have been cut off at 1.0 weight percent for illustrative
purposes (21 points in excess of 1.0 weight percent are removed from 2a and the R2 drops
from 0.90 to 0.85; 17 points are removed from 2b with no change in R2). These plots show
excellent correlations, and provide assurance of good internal consistency throughout the
program.
LEADVttlE PROJECT DRY KEVEX VS MOIST XMET RESULTS
n = 1314; Rsquare value = .85
reVEXPbwflC
1,0
0.0
O.O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
LEWvriuwhwtK
LEADVILLE PROJECT: DRY KEVEX VS DRY XMET RESULTS
n - 1100; R-aquaro value - J»
0.0!
O.O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
LEAD valUM ki wt%
(a)
(b)
Figure 2. Correlation plots between Kevex and moist/dry X-Met.
QUALITY CONTROL CHARTS
In order to monitor instrument stability, quality control check samples (QCCS) were
measured before and after each block of ten samples on each X-Met, and with each batch
of 13 samples on the Kevex3. A low- and a mid-calibration range of QCCS selected were
used to verify the calibration curves.
Kevex 770 has a 16 place autosampler. Each batch of samples consisted of 1 tube flux monitor
standard, 2 QC standards, and 13 routine samples.
10
-------
The low-range sample was used to calculate detection limits (discussed in section entitled
"PRECISION, ACCURACY AND DETECTION LIMITS" below), and the mid-range was
plotted in a control chart to monitor instrument stability. These control charts are shown
in Figure 3.
There are notable differences in the charts that need explanation. There were
considerably more samples run on X-Met Number One than on the Kevex, thus the
difference in a horizontal sense. There was only one set of QCCS run on the Kevex;
however, there were at least three sets run on X-Met Number One due to breakage of the
polypropylene film and loss of sample. What was not foreseen, mainly because the authors
had not engaged in an XRF survey of this magnitude, was that time would wreak havoc with
the QCCS. The 0.2 mil polypropylene film, through which the samples are X-rayed,
developed small holes with time and the soil leaked out. Consequently, the film had to be
replaced periodically and, being highly electrostatic, it extracted portions of the very fine
fraction from the soil, thus biasing the standard. Occasionally, the film would simply break
and another QCCS would have to be used. The bias between the different QCCS samples
could be seen as several distinct concentration groupings, the first break being at run
number 117 (for X-Met Number One). All of these potential sources of error manifested
themselves, particularly in the control chart for X-Met Number One, because the majority
of the samples were analyzed on this instrument. Also, the QCCS were analyzed in triplicate
on the X-Mets and were shaken and tapped between each measurement. The least
variability is evident from the Kevex chart because considerably fewer samples were run on
this instrument and samples were run only once, thus much less abuse of the polypropylene
film occurred.
In retrospect, the QCCS should have been pulverized and pressed to pellets,
alleviating the biases with time. The accuracy of the X-Met is clearly displayed in the
correlation between CLP and X-Met data (see Figure 4). Ninety-six of the 140 samples sent
to the CLP were run on X-Met Number One (R2 = 0.86).
CLP VERSUS X-METs
Figure 4 shows the correlation plots between combined X-Met and CLP and between
the Kevex and CLP. These plots show very good agreement between XRF measurements
and those of the CLP, thus supporting the high degree of accuracy achieved in this field
survey. The higher R2 value for the less sophisticated X-Mets is probably a result of the
triplicate measurement average accounting for the heterogeneity of the sample better than
the single, longer measurement used on the Kevex.
11
-------
LEADVULE PROJECT INSTRUMENT CONTROL CHAHT
KEVBCPbXt*
0.160
0.140
0.120
0.100
o.oso
0.060
0.040
0.020
0.000
I
60
l 1 1 1
12O 180 2*0 3OO
CONSECUTIVE HUN NUMBER
360 420
LEAD veluBi for KEVEX
Broken linos ot Oio mean and +/- 2 standard deviations
LEAEMLLE PROJECT: INSTRUMENT CONTROL CHART
XUETI PbvrtK
0.160
0.1 40 :
0.100
O.OBO
0.060
0.040
0.020
0.000
0 60 120 180 240 300 360
CONSECUTIVE RUN NUMBER
LEAD values for XMETJ1
Broken lines trt mean and +/ 2 standard deviations
420
LEADVU
0.160.
0.140
0.120.
0.100
0,080
0.060
0.040
0,020
0.000
Br
LE PROJECT: INSTRUMENT CONTROL CHART
hffljn,jjrj
rwpM
i 1 r 1 r i i
3 60 120 180 240 300 360 420
CONSECUTIVE RUN NUMBER
LEAD vtJuM for XUET #2
oken lines at mean and +/- 2 standard deviations
LEACML
0.160-
0.140-
0.120.
0.100.
0.080.
0.060.
0.040-
0.020
0.000
Br
US PROJECT: INSTRUMENT CONTROL CHART
j-i ITPTI
__(i . I . .
ir~~
T^IT
3 60 120 180 240 300 360 420
CONSECUTIVE RUN NUMBER
LEAD vriuM ferXMET #3
oken lines ai mean and +/ 2 standard deviations
Figure 3. Quality control charts for Kevex and X-Met #1.
12
-------
LE
cu>R>wt%
1.4
1.2
1.0
o.a
0.6
0.4
0.2
0.0!
ADMLLE PROJECT: CLP VS ALL 3 XMET RESULTS
n=140; R-square=0.88
M
-(..-^M
jfft.
4-
++
: -f +**
*>« +
^*'-t
+
4.4.^. . .
p+4f
i-
+
+.
...*...
I-
+
h
+
+
+
0-0 0-2 0.4 0.6 0.8 1.0 1.2 1.4
XMETFbwt^
LEAD valuae h wHt
LJ
CLPPbwt*
1.4
1.2
1.0
0.8
0.6
0.4
0.2
EADVILUE PROJECT: CLP VS KEVEX RESULTS
n=140; R-squ
-------
Using the American Chemical Society (ACS)(5) definition of detection limits4, the low-
calibration level of QCCS estimated a quantitation limit (QL) of approximately 0.16 wt%
lead. Given the errors introduced by the unstable QCCS standards, this detection limit is
probably artificially high. This assumption is supported by Figure 2a. At the origin of both
these plots, a "take off' point for X-Met detection can be seen on the Y-axis at or below 0.1
wt% lead. Note that the QL goes down slightly for the dried samples. Based on Figure 2a,
an empirical QL for lead measurements in this program is.set at approximately 0.1 wt%
(1000 ppm).
*The ACS defines a minimum instrument detection limit as 3 times the standard deviation of a series
of nonconsccutive blanks or low level check samples, and a quantitation limit as 10 times the standard
deviation.
14
-------
FREQUENCY
i
220 +
i
!
J
200 +
i
i
i
180 *
i
i
i
i
i
160 +
i
i
i
i
140 -!
i
120 +
i
i
i
i
i
100 +
1
1
1
80 +
|
1
J
i
60 +
i
i
i
40 +
i
j
i
20 +
1
1
!
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
****
2.5
****
****
****
****
****
****
****
****
****
***'*
****
****
****
**** ****
**** ****
**** ****
**** ****
**** ****
**** ****
**** **** ****
**** **** ****
**** **** ****
**** **** ****
**** **** ****
**** **** ****
**** **** ****
**** **** ****
7.5 12.5 17.5
Moments
N 531
Mean 8.8
Variance 80.8
Std Dev 9.0
.Skewness 2.1
Kurtosis 5.5
Coef Var T02.0
Quantiles(Def=5)
100% Max 58.3 99% 43.8
75% Q3 11.9 95% 26.8
50% Med 5.9 90% 19.5
25% 01 2.6 10% 1.0
0% Min 0 5% 0.5
1% 0.1
****
****
**** ****
**** **** **** **** ****
22.5 27.5 32.5 37.5 42.5 47.5 52.5 57.5
RSD_PB MIDPOINT
Figure 5. Frequency distribution of %RSD for moist Held duplicates.
15
-------
TOTAL VERSUS ESTIMATION ERROR
The above precision and QL estimates seem high when compared to those obtained
from a laboratory, but there were in excess of 3700 samples obtained in this program.
Decisions on the spatial distribution of lead in Leadville will not be based on the individual
samples, but rather on a geostatistical model of those samples. Consider the following
equation: . .
-------
Action Level (ppm)
500
1000
1500
2000
2500
3000
False Positive
2 (8.7%)
9 (22.0%)
13 (22.4%)
21 (26.6%)
13 (14.6%)
8 (8.4%)
False Negative
3 (2.5%)
2 (2.0%)
3 (3.6%)
6 (9.5%)
4 (7.5%)
8 (17.0%)
Table l. False positives and false negatives generated at six
action levels.
POPULATION TESTING
^ explained above, residential access was extremely limited
at the beginning of the field season, so the first 50% or more of
the samples came from shoulders of alleys in direct proximity with
residences. The obvious question was whether the contaminant
concentrations along the alleyways were significantly different
trom contaminant concentrations in nearby residential soils (i e
are alley samples representative of yard material?). To answer
this question, 38 'cells' were located where samples were taken
both within alley rights-of-way and in the nearby residential
yards. The paired-sample t-test and the nonparametric Kolmogorov-
Smirnov procedures of the SAS software were used to show that
with very few exceptions, there is no significant difference
between contaminant concentrations in the alleys and in the nearby
.
The results of the tests on the individual cells are shown in
Table 2. The test results are shown for test size 0.05.
TW< testing methods were employed, one a parametric
lPPl°a?? (2-sainple t Test) , the other a nonparametric
test (Kolmogorov-Smirnov Test). The two approaches
were tried because the distribution of data values was tending
toward the lognormal, but not strongly so. Thus in order to cover
all the bases, both methods were tried. Of the 38 pairwise
comparisons of yard and nearby alleyway samples, five were
significantly different using the parametric test, while only two
were significantly different using the nonparametric " test . This
confirms the more nearly lognormal distribution of values. Both
c°nclusi°"' that «» *» Populations are
17
-------
£
S
§
3
S
(fl
H
0)
s-
g
H
ts
ti
0)
"
I
4J
(tf
0)
?
S
0)
tp
01
s
!
e £
i
1 i
IS
§
§
3
8
H
4J
H
(0
O
04
0)
10
w
I
0)
c
0)
(fl
vo
0)
I
H
18
-------
Cell
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
t
Test
NSD
NSD
NSD
SD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
SD
NSD
NSD
NSD
NSD
NSD
KS
Test
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
Cell
Number
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
t
Test
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
SD
NSD
NSD
NSD
NSD
NSD
SD
SD
KS
Test
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
NSD
SD
NSD
NSD
NSD
NSD
NSD
SD
NSD
Table 2. List of results of t-test and Kolmogorov-Smirnov (KS) test. "NSD" means no
significant difference, "SD" means significant difference.
19
-------
CONCLUSIONS
The lead concentration data generated by FPXRF on residential soils in Leadville is
optimal for producing concentration isopleth maps that depict gross contamination
patterns across the townsite. The large number of sample points analyzed minimizes
errors in estimating values at unsampled points, yielding a more representative
depiction of contaminant distribution than a smaller data set of higher data quality.
A very large number of samples (>3700) were analyzed in a period of about three
months, allowing for cost- and time-effective determination of spatial patterns of
contamination distribution.
Precision and accuracy of the data generated are good, being ±27% and ±6%
respectively. The levels of false positives and negatives are also quite low.
The initial assumption that residential soils in Leadville were of a very similar matrix
(with respect to XRF) appears to have proven valid. The close agreement between
XRF results and randomly chosen samples analyzed by CLP methods indicates that
the single matrix model used with the X-Met 880 instruments was a reasonable
approach to dealing with a soil whose initial components were distinctly different, but
presumably well mixed by residential and mining activities over a long period of time-
Lead data from alleyway samples can be used to infer lead concentrations in the soil
of nearby yards. Statistical analyses showed no significant differences between lead
values in backyards and values in nearby alleyways.
As a recommendation, any long-term future work (greater than two weeks) using
FPXRF should use pelletized QCCS. It is virtually impossible to prepare identical
QCCS from loose soil, which is .falsely interpreted as inter-instrument bias. With
extended use, the fine soil fraction leaks through holes in the polypropylene biasing
the sample, which falsely translates to instrumental drift (intra-instrument bias).
Pulverization of the QCCS material prior to splitting minimizes inter-instrument bias,
and pressing the pulverized soil into a pellet produces a stable, homogeneous check
standard, thus removing the intra-instrument bias.
20
-------
REFERENCES
1.
2.
3.
4.
5.
6.
7.
Tweto, Ogden. 1968. Leadville District, Colorado, in Ore Deposits in the United
States. 1933/1967. J.D. Ridge, ed. American Institute of Mining, Metallurgical, and
Petroleum Engineers. New York, New York. pp. 681-705.
United States Department of Agriculture, Soil Conservation Service, in cooperation
with the Colorado Agricultural Experiment Station. 1975. Soil Survey of Chaffee-
Lake area, Colorado and Parts of Chaffee and Lake Counties. United States
Government Printing Office, Washington, D.C.
Workplan Soil Sampling and X-ray Fluorescence Analysis, Volume 1, Leadville
Colorado. September 1991. Roy F. Weston, Inc. Denver, Colorado.
Jenkins, Ron, R.W. Gould, and D. Gedcke. Quantitative X-ray Spectrometry. 1981.
Marcel Dekker, Inc. New York and Basel. 586 pages.
American Chemical Society. 1983. Principles of Environmental Analysis. Analytical
Chemistry, Vol. 55. pp 2210-2218.
Flatman, G.T., E. Englund, A. Yfantis. 1988. Geostatistical Approaches to the
Design of Sampling Regimes, in L.H. Keith, ed. Principles of Environmental
Sampling. American Chemical Society, pp. 73-84.
United States Environmental Protection Agency. March 1987. Data Quality
Objectives for Remedial Response Activities - Development Process. EPA 540/G-
87/003. Washington, D.C.
21
&U.S. GOVERNMENT PRINTING OFFICE: 1993 - 750-002/80242
-------
------- |