Topic 3: Sampling.

1
Topic 3 Sampling.
Objective to take from an object (the
population) an amount of material (the sample)
that can be analysed in a laboratory in such a
way that the analytical result for the sample is
of such quality that it represents the
concentration of the analyte in the object.
object
homogeneous
heterogeneous
discrete
continuous
discontinuous
Well mixed gases, liquids or solutions, pure
metals.
Ores in soils or rocks, particles in
suspension.
Change as a function of distance from source
(plumes)
Change with time reactive solutions or gases
settling suspensions, varying inflows.
examples
2
Sampling an object.
object
Quality control samples
Parts of the object
increments
sample

• Related samples
• co-located (replicates)
• Spiked in field
• Field blanks
• Control site sample(s)

Representative part of the object
Gross sample
Representative parts of the sample.
Sub-samples
storage
analyses

• Laboratory sub-samples
• duplicates
• Spiked
• blanks
• RMs

(x y)units
The analytical result(s) can only be as good as
the sample(s) collected.
3
Types of Errors.
Gross Errors arise from mistakes dont make
them. Systematic Errors lead to poor accuracy
can be identified and allowed for. Random
Errors lead to poor precision cant be avoided
but can be minimized.
Errors in sampling
Random errors - sample is not representative
too small and thus dont contain all of the
components of the object.

• Systematic errors
• biased towards certain components of the object
• eg sediment from a river where fine particles
  have been lost while collecting sample,
• Samples taken from within concentration gradients
  (object boundary effects such as haloclines in
  coastal waters, sides of rivers, surfaces of
  piles, tops of barrels )
• ?? poor definition of the object ??,
• changes that occur after sampling
• eg redox reactions, evaporation, biological
  changes, separation of particles, adsorption to
  surfaces (particles of containers)
• contamination of the collected sample
• eg from preserving agents, dust, container
  rinsing solutions, sampling equipment,

4
Sampling errors.
For systematic errors (biases) Bt B1 B2 B3
(detect and correct) For random errors Vt
V1 V2 V3 V s2 , the variances.
(minimize)

• Consider the random errors arising from two
  activities of an analysis
• sampling (ss) and measurement (sm).
• Neither sm or ss are significant - where the
  precision of the result is not of concern.
• sm is but ss is not significant - where the
  object sample are stable and homogeneous and/or
  the analytical method is not very precise.
• sm is not but ss is significant - the major
  contribution to the total error comes from
  sampling. The most common scenario frequently
  ss 3sm.
• Both sm and ss are significant - the sampling
  error is under control.

Total error st2 ss2 sm2. Which errors
should we try to minimize, if any, to ensure
that the result (x y) units is
appropriate? Eg If ss 9 and sm 3 then st
v(81 9) 9(.48) If ss 4.5 and sm 3
then st v(20.2 9) 5.4 If ss 9 and sm
1.5 then st v(81 2.25) 9.1 Improve
precision by focusing on the least precise
component usually sampling.
5
Sampling Protocols
Shall means do and document.
ISO 17025 5.7 Sampling. 5.7.1. The
laboratory shall have a sampling plan and
procedures for sampling . The sampling
plan and sampling procedure shall be available at
the location Sampling plans shall,
whenever reasonable, be based on appropriate
statistical methods. The sampling plan shall
address the factors to be controlled to ensure
the validity of results.
5.7.2. .. Client requires deviations from the
documented procedure, these shall be recorded
and communicated to appropriate personnel.
5.7.3. The laboratory shall have procedures for
recording relevant data and operations relevant
to sampling (sampler, conditions, locations,
statistical basis if appropriate)
5.8 Handling of test and calibration items. 5.8.1
shall have procedures for transporting,
receiving, handling, protect, dispose, .. 8.8.2
shall have a system for identifying 8.8.3
shall record abnormalities,
6
Sampling Protocols
The data generated from the samples must be
accurate and adequately precise so that the
hypothesis can be tested.
The protocol must define
1. sample type and the analytes
3. sample containers cleaning methods
2. sampling equipment cleaning methods
5. sampling locations
7. amount of each sample
6. number of samples
4. time of sampling
8. sample collection procedures
9. sample labeling
10. sample preservation
11. quality control samples
12. required recording of data observations
13. sample transport
14. training needs for the samplers
(?) sample pretreatment
7
Sampling Protocols cont.

• sample type and the analytes
• soil- horizon,
• water- surface, depth, tap, ground, river,
  marine, rain, vent
• air- inside, outside, plume, ambient,
• Blood, urine, sweat,

• sampling equipment cleaning methods
• Equipment use recommended procedures if
  possible - soil corers, air pumps, sediment
  traps, water bottles, commercially available
• Clean - to avoid contamination with the analyte
  rinse with de-ionized water, rinse with sample
  water, brush with a clean brush, sterilized
  stainless steel needles,

• sample containers cleaning methods
• Containers plastic bags for soils, plants
  plastic or glass bottles for liquids vapour
  losses samples often in containers for
  considerable times.
• Clean to avoid analyte additions or losses
  adsorption to surfaces, particles or liquids from
  container surfaces soak in acids (1N HNO3, 1N
  HCl, ) and the DI water, dry in ovens, cap, seal
  in bags for storage, rinse filters with DI water
  and oven/furnace dry,

• time of sampling
• Depend on hypothesis and sampling demands
• Pb in air a function of traffic,
• O2 in water a function of photosynthesis,
  pollution from industry, sewage treatment, ..
• Na in canned meat a function of production
  schedules,

8
Sampling Protocols cont.

• sampling locations and points of sampling

Defined by regulations urine or blood for drug
testing well studied systems

• Random (statistical) where all parts of the
  object have an equal chance of being sampled.
  Used when prior knowledge is lacking.
• Divide the object into a fixed number of parts,
  number them and then pick a predetermined number
  using a random number generator.

eg soil in a farmers cultivated lot when
determining available P 1 hectare (100m x 100m),
divide into 100 100m2 parts and randomly select
20. Also need to determine where within the
10m x 10m part to take the sample (use the same
random numbering method).
or Pb in street dusts, contamination of coastal
waters, a shipment of pimento berries in bags,
Need to define action to take if cant sample
at predetermined position eg rock in the soil.
9
Sampling Protocols cont.

• sampling locations and points of sampling cont.

Systematic where the object has a known
structure or behavior a plume from a point
source entering a river through a halocline in
an estuary parts of a plant (roots, stems,
leaves, flowers), Sample at fixed distances,
salinities, defined parts.
Sequential when there is some regularity to the
object. Products on a production line (take ever
100th or take a predetermined set of randomly
numbered samples)
Selecting sampling sites randomly or sequentially
must be done prior to sampling but systematic
sampling site selection will be done following
observations of the object.
Ad hoc sampling as opportunity arrives
forensic, only small amounts are available.
The interpretation of the data generated from a
sampling programme must be interpreted while
bearing in mind how the samples were selected.
10
Sampling Protocols cont 6. number of samples
Depends on the purpose of sampling, the
heterogeneity of the object and the confidence
level at which you wish to make your decision.
Often need to do some preliminary sampling
before making the final sampling plan.

• comparing two means when si are equivalent.
• (m1 m2) tsv(1/n1 1/n2) and
• s2 ((n1 1)s12 (n2-1)s22)/(n1 n2 - 2)
• the number of samples dominates the first
  expression and
• the population standard deviation dominates the
  second

Example Sampling a hectare to determine
available P. If mo 5mgP/kg, so
0.5mgP/kg, (mo ms) 10 mo at the 90
CL (ms mo) tso/vn (ms sample mean,
mo object mean) vn (0.51.64/0.5) n
2.7 or 3 samples. (8 DOF ? t 1.64) Divide the
plot into 100 squares, number them, select 3
using a random number generator
(???)
Then assume that other plots are similar and ?
sample similarly. After completing the
exercise were the assumptions reasonable?
Consider ms and ss. If not you may need to
resample.
11
Sampling Protocols 6. number of samples cont.
Acceptance sampling (MM, 7th, P101).
The concentration of a component in a product
should be mo (quality of the product,
regulations, ) and analyses show that the
concentration actually averages mo then 50 of
the determined concentrations will be greater
than mo. Is the product acceptable when ms ? mo?
This depends on the confidence level you wish to
make your decision at.
At the 95 CL a value greater than mo is
acceptable when (ms mo) tso/vn. t 1.64
(DOF 8 one sided question ? use 90 t value
from t-table two sided tables).
Also a client wishes to be 90 confident that a
product with m1 is of the required quality then
(ms m1) 1.28so/vn. (DOF 8 80 t-table
value).
eg (MM P103) mo 1.00g/kg and m1 1.05g/kg
and s0 0.05g/kg, the manufacturers and
customers risks are 5 and 10. (ms 1.00)
1.64(0.05)/vn (ms 1.05) -1.28(0.05) /vn
subtracting 0.05 (1.64 1.28)(0.05) /vn n
8.5 ? 9 samples and ms 1.027.
It could well be that such considerations could
be used to set the UAL on production quality
control charts.
12
Sampling Protocols 6. number of samples cont.
eg a sewage discharge must not exceed 2140µM
total dissolved N (NEPA standard). If the
operator accepts a 10 risk of exceeding the
limit and the total (sampling measurement)
uncertainty is 140 µM and it is decided that the
plant must shut down if the concentration exceeds
2200, then (2200-2140) t(140)/vn. t 1.28.
So take (8.9) 9 samples within the specified
time frame.
7. Amount of each sample. This relates to the
heterogeneity of the object, the sample must be
big enough to include all components in their
right ratios.
Ingemells criteria wR2 Ks.
w weight of sample to be taken
Ks sampling constant the weight of sample
required to ensure a 1 sampling error at 68 CL.
R2 required relative standard deviation ss/ms
x 100
mo 1s
13
Sampling Protocols 7. amount of sample cont.
Ingemells criteria wR2 Ks. Can determine
Ks by taking several 1g samples and analysing for
the analyte.
eg. If on 1g samples of a soil the following Cu
were found 65.1, 71.4, 38.9, 48.5, 62.5, 51.3,
58.9, 62.6, 47.6, 45.1 How much sample should be
collected to ensure that the sampling error is
lt3?
Then n 10, ms 55.2, ss 10.4. 1 x (10.4 /
55.2 x 100)2 355 Ks W Ks/R2 355/32
39.4g. (must have enough to do all the analyses
required)
8, 9, 10. Sample collection, labeling and
preservation procedures Document what the
samplers are to do when collecting the samples
Use the identified equipment, pre-cleaned before
going to the sampling site. Air pump where to
put it, what height above ground, away from
buildings, secure from pampering, how long it is
to run for, how to put the filter in, how to take
the filter out, addition of preservatives, label
(date, site number, samplers name, ), storage.
14
Sampling Protocols 8, 9, 10. Sample
collection, labeling and preservation procedures
cont.
Water sample from a river at centre of river
from a bridge, weighted, pre-numbered,
pre-cleaned container, stoppered, lower to 15cm
below water level, remove stopper, wait until
bubbles stop, recover sampler, empty with
efficient washing of stopper, repeat 2 times,
collect sample, tightly replace stopper, store
on ice.
Sample preservation depend on analyte metals
acidify to pH 1, Bacteria cool to 4C,
... Standard methods.
13. Sample transport. In the field and
returning to lab. Keep sample from contamination
and change.
12. Required recording of data and observations
sample ID, site ID, time, samplers, conditions
at site, information needed to interpret the
data
15
Sampling Protocols cont.

• Quality control samples.
• Pre-decided upon field replicates (co-located
  samples) to test for total uncertainties (ss
  sm). A decided upon percentage of collected
  samples.
• Intra- and inter-comparison samples split or
  replicate samples. previously arranged with
  other analysts, labs, of collected samples.
• Field spiked samples add a know amount of
  standard to a replicate (or portion of a) sample

Control site samples from an area expected to
be free of analyte being investigated (component
of site selection) eg low traffic density part of
town for Pb in atmospheric particles, site
up-river of pollution source, productive
agricultural plot to compare with a poorly
productive plot Field blanks deionized water
carried from the lab. Soil sampling equipment
rinse sampling equipment between samples and
analyse washings. Water samples rinse and fill
a sample bottle with DI water while in the
field. Air filter filter left in the pumping
system for normal time without pumping. be
innovative but defined.
14. Training needs for the samplers all aspects
of the protocol must be adhered to and understood.
16
C 30J The Analysis of real samples Problem
Sheet.

• A farmer grows coffee on 10 hectares of land. He
  finds his caffeine in raw beans content is 0.9
  by weight, approximately 0.3 below that
  considered ideal. He suspects that this is due
  to a Cu deficiency in his soils. For normal
  growth coffee should be grown on soils with 55ppm
  Cu. Devise a sampling protocol to a) test the
  soil in the plot for Cu and b) test the 100 bags
  of coffee produced per crop for caffeine. Assume
  that analytical and sampling errors are 10 of
  the determined values. Clearly state and other
  assumptions you make.
• SWH, 7th edition, Chapter 32 problems 32-5,
  32-6. Chapter 34 Example 34-1 (p766), Example
  34-2 (p768), Example 34-3 (p771), Problems 34-1,
  34-4, 34-8, 34-9.
• SWH, 7th edition, Chapter 24 example 24-1 (a)
  (p572), example 24-2 (p575), problem 24-16
  (p598).
• C 30J examination papers 1999 questions 5 and 6
• 2000 question 5 (a)
• 2001 questions 5 and 6
• 2002 question 5
• 2004 question 5, 6 B