Presented - Environmental Law and Litigation

Uncertainties in Trace Analysis
Presented by:
Dr. George Duncan, P. Geo., C. Chem., MCIC,
Environmental Consultant Performing Phase 1 & 2
Environmental Site Assessments
What are the issues?
Phase 2 Environmental Site Assessments (ESA)
include sampling and analysis of soil, sediment
and ground water.
The “Qualified Person” (QP) must interpret the
analytical results to determine if the site meets
the “generic” (Province-wide) criteria or sitespecific, “risk-assessed” criteria.
Exceedence of a criterion---even by the slightest
amount can result in very expensive clean-ups.
Defining “Uncertainty”
What kinds of uncertainty are we dealing
What are the sources of these?
What do they mean in PRACTICAL terms to
the “Qualified Person” (QP) who must
interpret the lab results to assess the
environmental condition of the site?
What kinds of uncertainty?
To the Analytical Chemist..
Uncertainty: A non-negative parameter
associated with the result of a measurement
that characterizes the dispersion of the values
that could reasonably be attributed to the
measurand. (International vocabulary of
metrology – Basic and general concepts and
associated terms; ISO/IEC Guide 99:2007
(VIM 2007)).
What kinds of uncertainty?
Standard Uncertainty: Uncertainty
components are evaluated by the appropriate
method and each is expressed as a standard
deviation and is referred to as a standard
What kinds of uncertainty?
Combined Standard Uncertainty:
Standard uncertainty components are
combined to produce an overall value of
uncertainty known as the combined standard
uncertainty. It is an estimated standard
deviation equal to the positive square root of
the sum of variances of all uncertainty
This uncertainty is what the labs report.
What kinds of uncertainty?
Expanded Uncertainty: Expanded
uncertainty (U) is obtained by multiplying the
combined standard uncertainty by a coverage
factor “k” to provide an interval within which the
value of the measurand is believed to lie, with a
specified level of confidence (e.g. 95%).
The Sources of Uncertainty
These exist at every stage of the analysis:
In the balance used to weigh the standards
and samples;
In the volumetric pipets used to deliver and
dilute the solutions;
In the instrument’s readings;
In Practical terms…
The uncertainties quoted by the laboratories to
their clients are typically reasonably “tight”
(narrow range) for most parameters:
Arsenic @ 18 mg/Kg = +0.5 mg/Kg (3%)
Cadmium @ 2 mg/Kg = +0.08 (4%)
Benzo-a-pyrene @ 0.3 mg/Kg =+0.022(7%)
Benzene @ 0.32 mg/Kg = +0.063 (20%)
TCE @ 0.55 mg/Kg = +0.061 (11%)
NOTE: trace organics are much higher than metals
In Practical terms…
Uncertainty of Measurement is NOT a
satisfactory measure of the
Accredited laboratories generally are excellent
at minimizing uncertainty in the measurements
they make---they have to be to stay accredited!
Uncertainty of measurement as described
above is a measure of the unavoidable
variability inherent the analytical method BUT
In Practical terms…
QP’s generally accept the analytical results on
the C/A as “correct” and representative of the
site condition in the area sampled.
Under the current “Pass/Fail” system of O.
Reg 153/04 (December, 2009), any lab result
that exceeds a limit by ANY amount must be
considered as correct and dealt with either
through remediation or risk assessment--both of which can be very expensive.
In Practical terms…
Problems arise when the C/A shows results
just above the allowable limit.
One or two repeat analyses on the SAME
sample (if enough sample is left) often show
some exceedences becoming “Non-detects”,
some falling below the limits and some rising
even higher, well beyond the uncertainty in
the analytical method:
In Practical terms…
In Practical terms…
Real Uncertainty…inhomogeneity
To the lab chemist, uncertainty in measurement
is a measurable quantity that can be reported
but it represents only a small part of the REAL
uncertainty in the analytical result.
The QP must also consider the uncertainty in the
reported result created by two important criteria:
Sample inhomogeneity
Field inhomogeneity
Real Uncertainty…Sample
Sample Inhomogeneity relates to how (un)well
the component being analyzed is distributed
throughout the sample. Where the sample (e.g.
metals), can be ground to -355 μm (0.355 mm)
this alleviates the problem but only a little.
If the sample cannot be ground (trace
organics), serious problems arise. Slight, or
even major exceedence of a limit may not be
“real” and repeat analyses can bounce wildly.
Real Uncertainty…surrogates
Surrogates are used for organic tests. All
samples are spiked with compounds (usually
deuterated analogs) representative of the analytes
being determined but not found in environmental
samples. The surrogates are spiked into the
sample prior to any sample preparation steps and
carried through the entire analytical process.
Surrogate Recovery = ([measured concentration]
/ [theoretical concentration]) x 100
Real Uncertainty…what is real?
Question: If the acceptable surrogate
recovery is 50 – 150%, what is the “acceptable”
deviation from the TRUE concentration in the
For PAHs & VOCs, percent surrogate recoveries
in soil and water should be between 50 – 140%;
for hydrocarbons 60 – 140%, for mercury 70 130% (matrix spike).
Real Uncertainty…the
Most QPs are engineers and geoscientists
(non-chemists) and are unaware of these
issues so the reported lab results are
accepted at face value and any exceedences,
including the slight ones, are treated as
“real”. A “Record of Site Condition” cannot be
registered for any site with an exceedence of
ANY allowable limit.
No RSC?---no building permit will be granted
Real Uncertainty…the
O. Reg 153/04: Data quality objectives outline
the overall level of uncertainty that a QP will
accept in collecting field data in order to develop
a CSM. Data quality objectives are set to
determine precision, accuracy, reproducibility,
representativeness, and completeness for field
data (i.e., relative percent difference (RPD),
matrix spikes, matrix spike duplicates, data
qualifiers, etc.).
Real Uncertainty…the
“Data quality objectives for sampling and
analysis plan are to be set beforehand”
The data quality objectives for any field data
collected during the phase two ESA must ensure
that the decision making is not affected and the
overall objectives of the investigation are met.
O. Reg. 153/04 gives no instructions on how to
achieve this. For example:
Real Uncertainty…the
Stockpiles: “Sampling locations must be chosen
so as to ensure uniformly distributed and
representative sampling collection throughout
the stockpile”.
Question #1: How many samples should be
gathered and what size should they be (Weight?
Volume?) ?
Question #2: How do you sample the
Real Uncertainty…the
Real Uncertainty…the
Lab sample soils
are screened
through a 2 mm
sieve before
analyses is begun
(except for
organics such as
F2-F4 hydrocarbs
and PAHs).
Real Uncertainty…the
Real Uncertainty…the
Question #1(b): ……….what size (Weight?
Volume?) should they be?
Reg 153/04 Answer: “Any sample(s)
collected and analyzed must be representative
samples collected at locations and frequencies,
following a sampling plan determined by the QP
The lab answer: Fill the 100 mL soil jar to the
top and we’ll use that for metals, F2-F4, and
PAHs, etc. For VOCs put 2 – 3 g in a VOC vial!
Real Uncertainty…reducing it
Reducing Uncertainty in the lab
The labs are already doing all they can to produce
accurate results. The major source of real
uncertainty is sample inhomogeneity. Fine
grinding of the sample helps but this is not
allowed for organic COC analyses. The few
grams of sample actually analyzed is often not
representative of the 100 g of sample in the soil
Real Uncertainty…reducing it
If one of the large
ptles is lead paint
and the rest are
native soil from the
site, would the
analysis result be
close to the “true
value” if the true
value is 1000 ppm?
Real Uncertainty…reducing it
Reducing Uncertainty at the site
This is currently an impossible task because each
tiny sample submitted for analysis is forced to
represent many tons of soil. Soils are notoriously
heterogeneous and many contaminants occur as
discrete particles. The Reg pays only lip-service to
collecting REPRESENTATIVE samples but truly
representative samples will be far too large (>10
- 100Kg) to be analyzed at the lab.
Real Uncertainty…reducing it
Real Uncertainty…reducing it
Real Uncertainty…reducing it
For COCs in coarse soils at low ppm levels (e.g.
As, Be, Cd..), unless you grind ALL the sample
very fine (e.g. <-200 mesh or 74 μm) and
perform repeat analyses, the C/A can be way off
the mark. It explains why results at near-limit
values can bounce wildly, making cleanup
decisions very difficult.
Consider the following site investigation of a
sand-blasting site used to remove lead paint.
The problem is statistics
The site soils are coarse sand
of average particle size 1 mm.
Let’s assume the ACTUAL Pb
concn = 1000 mg/Kg (O.Reg
limit = 120 mg/Kg)
In 1 cc of soil (~1.4 g) there
are 1000 ptles max. Actually
it’s only ½ this due to the airspace between particles.
The problem is statistics
A 1.4 g soil sample (1 cc) CANNOT PRODUCE
THE CORRECT ANSWER since it will contain 0, 1,
2 or more 1 mm ptles of lead paint! Each paint
ptle weighs 8x more than a sand ptle so if the
sample has no paint particles, the C/A indicates 0
mg/Kg but if the sample has 1 ptle of paint, it will
produce a lab result of ~8000 mg/Kg (a 1mm3
paint ptle weighs ~11 mg)!
The problem is statistics
The only way to get the correct result is to take
a much larger sample!
Remember: In the 1.4 g of soil the grain size is
1 mm so there are only ~500 ptles in the
sample. The true Pb concn. is 1000 mg/Kg so
you need ~8x1.4 g sample (and be lucky
enough) to catch one ptle of lead paint and get
the correct result!
The problem is statistics
You can grind the 11.2 g sub-sample as fine as
you like and get very reproducible analysis
results but they will all be equally wrong if:
Your sub-sample has only 5 or 6 ptles of paint
or 12 or 14 ptles of paint!
Remember: O. Reg. 153/04 screens the whole
soil sample through a 2 mm screen and the 5 10 g sub-sample is taken from this.
What happens if the Pb ptles are 1.9 mm?
The problem is statistics
The limit for lead in soil is 120mg/Kg
The cardinal rule in sub-sampling is: once you
lose sample representativeness, you can NEVER
regain it---not matter how fine you grind.
If the <2mm sub-sample taken for grinding is
non- representative of the whole lab-sample,
grinding is useless!
We haven’t yet discussed field sampling!!
The problem is statistics
Consider the highly toxic metals such as arsenic,
beryllium, cadmium, etc. The allowable limits for
these are in the low ppm range (1 – 20 mg/Kg).
1 ptle of arsenic in 1 cc = ~37 mg/Kg (assuming
no density difference). The limit is 18 mg/Kg for
Table 2 soils. How much sample are you going
to have to collect and send to the lab to be
The problem is statistics
NOTE: none of the above discusses the much
more serious issue of how to get a 3g sample of
methanol-preserved soil to represent 50 m3 (3
truckloads or ~ 65 Tons) from a soil pile or a
borehole you just drilled. YOU CAN’T!
Statistically you are not even in the ballpark.
The mining exploration industry would never
base any decision to build a mine on such paltry
numbers and sizes of samples.
Major changes to O. Reg
153/04 are needed
Much larger sample sizes are needed to begin
to be representative but labs cannot handle
Leach-testing of large samples may provide a
more reliable and a more relevant estimate of
a site’s condition.
For MEANINGFUL lab analyses to be done,
the sample size MUST be established at the
SITE, not at the lab!
Major changes to O. Reg
153/04 are needed
The polling industry (Gallup, etc.) first calculates
the # and location of people to be polled
(“sample size and distribution”) before
commenting on the accuracy of the poll
(“correct within 3% points, 19 times out of
The ESA industry has this backwards by limiting
the size of the sample no matter the size and
distribution of the population!
O. Reg 153/04 is currently
under Review
Hopefully, we will see the changes necessary to
correct these issues.
Until then we will continue to struggle with C/As
showing dubious exceedences of regulatory
limits resulting in many $ millions spent on
needless investigations and cleanups.
REMEMBER: If the C/A shows an exceedence of
any limit---clean it up or risk assess it!

similar documents