Modeling Resource Selection

1
Modeling Resource Selection

• Dana L. Thomas
• University of Alaska Fairbanks

2
Outline

• What is resource selection?
• What factors impact resource selection?
• Why study resource selection?
• Defining a population of resource units
• How might we quantify selection?
• Two single factor examples
• A more complex forestry example

3
What Is Resource Selection?

• When resources are used disproportionately from
  their availabilities, use is said to be
  selective.

4
Common Types of Resource Selection Studies

• Forage Studies
• Types such as prey species, color or shape
• Attributes such as prey size

• Habitat Studies
• Types such as old growth, new growth or clear cut
  or substrate type in fisheries
• Attributes such as shrub density, distance to
  nest, canopy height

5
Resource selectionis a function of many factors

• Individual factors
• Resource unit factors
• Population factors
• Scale
• Seasons or time periods

6
Why Study Resource Selection?

• Identifying long term resource requirements for a
  species
• Projecting the impact of habitat change
• Why Are You Interested in Studying Resource
  Selection?

7
A Population of Resource Units
8
A Sample or Census?Sampling Protocols

• A A census or sample of available units is
• taken and used units are sampled.
• B A census or sample of available units is
• taken and unused units are sampled.
• C Unused units and used units are
• independently sampled.

9
Common Study Designs Design I

• Use and availability are desribed for the
  population of animals as a whole no information
  on individuals is collected.
• e.g., Use a map or GIS of the study area to
  describe availability. Conduct one or more
  aerial or ground surveys and record the location
  of animals within the study area.

10
Common Study Designs Design II

• Use is recorded for each animal availability is
  presumed the same for all animals and is recorded
  at population level.
• e.g., Use a map or GIS of the study area to
  describe availability as in Design I. Record
  relocations of uniquely marked or radio-collared
  animals within the study area to describe use.

11
Common Study Designs Design III

• Use and availability are recorded for each
  individual.
• e.g., Identify individual home ranges Use a map
  or GIS of each home range to describe
  availability. Record relocations of uniquely
  marked or radio-collared animals within the home
  range to describe use.

12
Early Analyses Indices

• - Scott (1920) Average number of each prey
  species per fish stomach divided by the number
  found in plankton hauls per unit area
• - use/available called Forage ratio, selection
  ratio or preference ratio

13
Common Analyses for Selection

• Chi-square tests
• Goodness of fit
• Tests of homogeneity
• Maximum chi-square
• Univariate nonparametric tests
• Friedman test
• Multivariate tests
• Johnsons Prefer program
• Logistic Regression
• Discriminant Analysis
• Compositional Analysis

14
A Unified Theory

• The resource selection function (RSF) is the
  relative probability that an animal uses a
  particular resource unit, assuming equal
  availability of competing units.

15
Resource Selection Modeling

• rsf f(elevation, habitat type, distance to
  nest, age, sex)
• Once we find a good function describing
  selection, we estimate the odds of selecting one
  unit over another when the units differ by only
  one characteristic.

16
Beer Example

• Suppose I have a cooler of beer.
• It contains
• 12 bottles of Coors
• 6 bottles of Ballard Bitter
• 2 bottles of Guinness Stout
• 1 bottle of Odouls (non-alcoholic)
• Pick one!

17
Beer Example
18
(No Transcript)
19
How Can Resource Selection Functions Be Useful
An Example

• The Chickahominy case
• USFWS filed for an injunction to stop West Coast
  Forest Resources from harvesting a 100 acre unit.
  
• USFWS claimed harvest would harm a pair of
  northern spotted owls.
• Harvest unit was within the home range of both
  owls.
• Judge ordered a 1 year telemetry study. WEST,
  Inc. helped analyze the data.

20
After One Year of Telemetry

• Relative probability of the owls selecting a
  point on the landscape (the rsf) was related to
  the following environmental variables
• Forest type (non-forest, young forest, old
  forest)
• Distance from the nest
• Elevation
• Here rsf f(forest type, distance to nest,
  elevation)

21
(No Transcript)
22
Interpretation

• All else being equal,
• These owls relative probability of selecting old
  forest was twice that of young forest.
• These owls relative probability of selecting
  young forest was 5 to 12 times that of clearcut/
  non-forest.
• Probability of selecting a point more than 2 km
  from the nest was 80 smaller than for a point
  near the nest, assuming the two points were
  otherwise identical.

23
(No Transcript)
24
(No Transcript)
25
Assessing Potential Impacts

• Use the volume under the RSF surface in risk
  assessment as an index to harm.
• Large areas in rarely selected habitat could have
  same harm as small areas in heavily selected
  habitats.
• Compute harm for a number of different harvest
  units.

26
Candidate Chickahominy Units
1
2
3
27
Harm Indexes

• Conclusion Unit 1 does least harm of
• these three

28
Summary

• Resource selection modeling relates probability
  of resource use to factors like age, sex,
  distance, elevation, etc.
• It can be used as a management tool to assess
  harm
• Requires real data (often telemetry and GIS or
  collected animals for stomach analysis)
  therefore, can be expensive.

29
Issues to Keep in Mind

• Is availability known or estimated?
• Is availability presumed the same for all animals
  or is it measured for each individual?
• What happens if availability changes with time?
• What is the impact of errors in describing
  availability or use ?

30
Issues to Keep in Mind

• Is the sample of animals for which we are
  recording resource use representative of the
  population?
• Are animal relocations really independent?
• Does it matter if animals are in groups?
• Should the analysis of selection be conducted for
  each animal or using data aggregated across
  animals?