Distribution and Sampling of Agricultural Pests

1
Distribution and Sampling of Agricultural Pests
2
Sampling

• Important for collection of any ecological data
• Objective estimate pest population of field

3
Why sample?Because crop yields are related to
pest population numbers
4
Objective of Sampling
Cant sample everything !
5
Sample Unit a single measurement or estimate
6
Sample Units

• Impractical to count entire field, so select a
  smaller unit ( sample)
• Counts of pest numbers per unit
• Weeds/row, weeds/m, insects/leaf, diseased
  plants/row, nematodes/volume of soil
• Of course, can use for plant and soil
  measurements like yield, plant height, nutrients

7
Basic Statistics

• Sample data set 5, 3, 8 weeds/row
• Mean
• Standard deviation
• Variance
• CV Coefficient of variation

8
Basic Statistics

• Sample data set 5, 3, 8 weeds/row
• Mean arithmetic average of the data x 5.33
• Our sample mean is an estimate of the true mean
  density in the field
• Standard deviation
• Variance
• CV Coefficient of variation

9
Basic Statistics

• Sample data set 5, 3, 8 weeds/row
• Mean
• Standard deviation measure of the average
  deviation of data from the mean s 2.5
• SD measures dispersion or spread of the data
• SD different from the range (high-low value)
  here range is 8-3 5
• Variance
• CV Coefficient of variation

10
Basic Statistics

• Sample data set 5, 3, 8 weeds/row
• Mean
• Standard deviation
• Variance s2 similar to standard deviation
• CV Coefficient of variation (s/x) x 100

11
Objective of Sampling
Field has true mean, standard deviation
Our sample estimates mean, standard deviation
Powers and McSorley, 2000
12
Precision vs. Accuracy

• Accuracy - how close the sample mean (x) is to
  reality (?).
• Maybe measured mean was 5.3 but should be 8.5 if
  more efficient method was used.
• Strongly affected by carelessness and/or correct
  choice of best sample unit and methods for
  collection, see literature for best methods.
• Precision - how close measurements are to one
  another measures variability.
• Precision measured by standard deviation or CV.
• Measurements of 4, 5, 6 (s1.0) more precise than
  measurements of 3, 5, 8 (s2.5).

13
Precision vs Accuracy

• Accuracy can become a serious problem if pests
  are microscopic, hidden in soil or plant tissue,
  highly mobile, etc.
• Poor precision unpredictable results

14
Patterns of Spatial Dispersion or Distribution

• Uneven or irregular distribution of items
  (plants, pests, etc.) in field causes serious
  problems in sampling precision!
• Spatial dispersion patterns (Southwood, 1978).
• Test for pattern by finding x and s from
  preliminary samples.

15
Spatial Dispersion Patterns
Ludwig and Reynolds, 1988
(Regular)
16
Spatial Dispersion Patterns, based on preliminary
samples of s2 and x

• Regular s2 lt x
• Machine-planted crop plants in field (plants/m2)
• Random s2 x
• Model Poisson distribution (similar to bell
  curve)
• Ideal case, crop yields, plant growth.
• Clumped aggregated contagious
• s2 gt x
• Model contagious distributions such as negative
  binomial.
• Most agricultural pests, many soil
  characteristics follow this pattern!

17
Negative Binomial DistributionApplies to Clumped
Data
Not a typical bell-shaped curve !
Elliott, 1979
18
Negative Binomial DistributionApplies to Clumped
Data
High frequency of low counts
Some extremely high counts possible
Not a typical bell-shaped curve !
Elliott, 1979
19
Spatial Dispersion Pattern

• To check dispersion pattern of a pest
  collect several samples calculate x, s2
• Aggregated dispersion pattern causes problems for
  sample precision good chance of getting extreme
  high (near clump) or low (away from clump) values
• Plant data (height, yield, etc.) expected to be
  random clumped values for yield indicate
  underlying problem related to soil, pest, etc.

20
Most agricultural pests show aggregated (clumped)
dispersion due to

• Dispersal patterns (large numbers of seeds or
  young from point limited mobility, etc.)
• Microhabitat (prefer certain spots within plant
  rows, etc.)
• Physical or environmental differences within
  field (soil type, topography, etc.)
• Agricultural practices (crop history,
  cultivation, old row sites, etc.)

21
Size of Sample Unit

• Check literature
• Not too large or too small (affects total amount
  of data collected, time and cost of sampling)
• More small units better than few large units
  (better chance to visit clumps in field)

22
Sampling Pattern (arrangement of samples in
field

• Random basis for sampling statistics, hard to
  set up in practice.

23
Sampling Pattern -- Systematic

• Systematic samples collected at regular
  intervals, often used

Systematic sample Gives some positional
information on the pest
Transect series of samples along straight line
path
24
Sampling Pattern - Stratified

• Stratified - heterogeneous area divided up into
  portions, or strata.
• separate sample(s) collected from each stratum.
• sample mean for area must be weighted depending
  on size of strata.
• can be stratified random or systematic.

Often based on natural strata like soil type or
vegetation
25
Sample Size (number of samples)

• Sampling always has error associated with it
• No error Sample everything !!

26
Sample Size (number of samples)

• How many samples to collect?
• No definite answer -- depends on level of
  precision, which is arbitrary.
• Indices of precision
• Coefficient of variation

High CV Low Precision High Error
27
How many samples to collect?

• Check literature on specific pest
• Seek advice of statistician
• Trial and error (based on previous sampling for x
  and s) some preliminary sampling always
  essential
• Must set desired limits of precision !!

28
How many samples to collect?

• Cant answer directly stat formulas will
  require some set level of precision (e.g., CV
  50, CV 20, etc.)
• Low number of samples less certain result
• Sample number always at least 2

29
Measures of Precision

• Coefficient of variation (CV) a common measure
  of precision
• Confidence interval interval around the sample
  mean that has a 95 probability of containing the
  true field mean (e.g., 20-5 vs 20-15)

30
Relationship Between Sample Error (or Precision)
and Number of Samples

• Sampling error and precision show opposite trends.

CV
31
Relationship Between Sample Error (or Precision)
and Number of Samples

• Cost of sampling is important limitation
  Diminishing returns at high sample numbers

CV
CV
Cost
32
Sample Error and Number of Samples (some rough
numbers)
CV
23456789
33
References

• Text, Ch. 1, pp. 8-11 Ch. 11, pp. 222-226.
• Elliott, 1979. Some Methods for the Statistical
  Analysis of Samples of Benthic Invertebrates.
  Freshwater Biological Association, Windermere,
  UK.
• Ludwig and Reynolds,1988. Statistical Ecology.
  John Wiley, NY.
• Southwood, T.R.E. (1978 many editions).
  Ecological Methods. Chapman and Hall, London.