Environmental Sampling Designs

1
Environmental Sampling Designs

• Brief Introduction

2
Sampling Designs

• Non-probability sampling
• Haphazard sampling
• Judgement sampling
• Probability Sampling
• Simple Random sampling
• Stratified random sampling
• multistage sampling
• cluster sampling
• systematic sampling
• double sampling
• search sampling
• composite sampling

3
Goal of Sampling

• To estimate the mean with good precision. This
  can then be compared to a control site or to
  before impact data.
• Equations for estimating the for each sampling
  scheme are given in Surveying Sampling books e.g.
  Elementary Survey Sampling by Scheaffer, et al
  (1996).
• Need another course just to cover this topic
  thoroughly.

4
Sampling and Sampling Design
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
Summary of sampling designs and when they should
be used for estimating means and totals

• Haphazard sampling
• A very homogeneous population over time and space
  is essential if unbiased estimates of population
  parameters are needed. This method is NOT
  recommended due to difficulty in verifying this
  assumption.
• Judgement sampling
• Target population should be clearly defined,
  homogeneous, and completely assessable so that
  sample selection bias is not their unique value
  and interest rather than for making inferences to
  a wider population.

24
Probability sampling

• Simple random sampling
• The simplest random sampling design. Other
  designs will frequently give more accurate
  estimates of means if the population contains
  trends or patterns of contamination.
• Stratified random sampling
• Useful when a heterogeneous population can be
  broken down into parts that are internally
  homogeneous.
• Multistage sampling
• Needed when measurements are made on subsamples
  or aliquots of the field sample.
• Cluster sampling
• Useful when population units cluster together
  (schools of fish, clumps of plants, etc) and
  every unit in each randomly selected cluster can
  be measured.

25

• Systematic sampling
• Usually the method of choice when estimating
  trends or patterns of contamination over space.
  Also useful for estimating the mean when trends
  and patterns in concentrations are not present or
  they are known a priori or when strictly random
  methods are impractical.
• Double sampling
• Useful when there is a strong linear relationship
  between the variable of interest and a less
  expensive or more easily measured variable.
• Search sampling
• Useful when historical information, site
  knowledge, or prior samples indicate where the
  object of the search may be found.
• Composite sampling
• Useful when field samples are easy and cheap to
  collect but laboratory costs are high.
  Compositing physically average the results. But
  measure of variability may be lost in the
  composite sample.

26
Some 2-D probability designs for sampling over
space
A) simple random sampling
B) Stratified random sampling
Primary units
C) Two-stage sampling
27
Clusters
D) Cluster Sampling
E) Systematic Grid Sampling
F) Random Sampling within Blocks
28
Some probability sampling plans for sampling over
time or along a transect to estimate a mean or
total

• Simple random sample - samples chosen at random
  over whole line
• Stratified random sampling - samples chosen at
  random with strata. Strata are usually different
  in size and based on a priori information about
  variation and the site.

strata
29

• Two-stage sampling - primary units selected at
  random samples chosen at random within selected
  primary units.
• Cluster sampling - All possible samples taken
  within randomly selected clusters.

Primary units
Second stage units
Two clusters
30

• Systematic sampling - samples are evenly spaced
  after a random start position A is chosen
• Random sampling within segments - one or more
  samples chosen at random within each segment
  segments may be different in sizes, not based on
  a priori information about variation or the site.

A
31
Other sampling plans

• Composite sampling - see handout on simple
  composite sampling
• Gradient sampling design- best when dealing with
  pollution from a point source. Do regression of
  concentration Y vs. Distance from source plus
  other explanatory variables. If regression slope
  is significant, then there is impact. Can use
  ANCOVA for comparing from year to year or
  baseline data vs. impact data.

Y
Impact period
baseline
Distance from source
32
Summary

• A basic idea of sampling from a population is
  given
• Different sampling schemes used in environmental
  studies were discussed
• Purpose was to estimate the mean or total to
  compare with before impact or to a control site
  (BACI type)
• Best to keep sampling scheme as simple as
  possible
• easier to explain to authorities
• data can be used for other purposes (not foreseen
  earlier)
• data easier to analyze
• Equations for estimating means - refer to
  references.
• Must know how to analyze the data before deciding
  on a scheme.