#9 Habitat selection III

Sink

An area where a population declines over time (λ < 1)

• habitat where mortality exceeds reproduction

Source

An area where a population increases over time (λ > 1)

• habitat where reproduction exceeds mortality

Ecological trap

A mismatch between evolved cues and actual habitat quality leading to harm

• The cues no longer match actual habitat quality

What is often related to reproduction when mistakes become very costly?

Ecological trap

Proximate cue

A stimulus used by animals to evaluate habitat quality

What is a cue in ecology?

an evolved stimulus that animals use to make decisions based on the assumption that the cue reliably signals habitat quality

What defines a high-quality habitat?

When the habitat allows for population growth

• the population's growth rate (lambda, λ) is greater than 1 (λ > 1)

What is a low-quality habitat?

A habitat where lambda (λ) is less than 1

• meaning the population declines over time.

What happens if an animal avoids a high-quality habitat?

This indicates a cue mismatch

• the animal misinterprets the habitat as low-quality due to unreliable or outdated cues

• potentially leading to missed opportunities for survival or reproduction.

What happens if an animal chooses a low-quality habitat?

indicates a mismatch

• The animal is misled by the cue and ends up in a habitat that doesn't support its survival or reproductive success

• possibly an ecological trap

Is cue response evolution fast or slow?

slow

• Environmental cues take time to evolve

What do we use to make a study design for habitat selection?

The hypothetico-deductive method

The hypothetico-deductive method

• Define a research question based on observed patterns or phenomena

• Conduct a literature review 

• Identify specific objectives, hypotheses

• Formulate testable predictions from hypotheses

• Design research methodology

• Write proposal and get that good money

• Collect data!

What are the two parts to build a hypothetico-deductive structure?

• Hypothesis

• Predictions

Hypothesis

A possible explanation for a phenomenon, a speculation on how something operates

• Example: Nesting Cinnamon Teal hens use vegetation height as a cue to select nesting sites

Predictions

logical, testable extension of a hypothesis (if/then)

• Example: If nesting teal cue in on vegetation height, then observed vegetation height at occupied nest sites should be taller than the average available height

What are the 2 statistical hypotheses?

• Null hypothesis→ no relationship or no effect between the variables being tested

  • Example: Habitat type does not affect population growth.

• Alternative hypothesis→ what you want/ think should happen (is a relationship or an effect)

  • Example: Habitat type does affect population growth.

What are Johnson’s (1980) orders of habitat selection?

• 1st order: selection of physical / geographic region (species range)

• 2nd order: home ranges (individual or population)

• 3rd order: selection of habitat components for specific uses within the home range 

• 4th order: selection of specific resources within habitat components

What are 4 different study designs?

• Design 1: Comparing collective use and collective availability

• Design 2: Comparing individual use to collective availability

• Design 3: Comparing individual use to individual availability

• Design 4: A special case– pairing occupied and random sites

What is the study design for Comparing collective use and collective availability?

• Record presence/absence (fine to use non-identifiable individuals) in study area

• Measure available habitat in the entire study area (must be representative!)

What is the study design for Comparing individual use to collective availability?

• Record habitat use by known individuals (e.g., through remote telemetry or visual marking)

• Measure available habitat in entire study area, assume availability is equal for all individuals in your study population

What is the study design for Comparing individual use to individual availability?

• Record habitat use of individually marked animals

• Measure habitat available to individuals (usually only that within or near its range)

• Can then choose to amalgamate individuals depending on questions

Why is it not a good idea to just tag a few animals while thinking they have equal access to the whole habitat?

• Small animals are limited in its home range so its not good to assume they have access to the whole large habitat

• You want to focus it on what it can actually use

  • The individual availability 

  • Make it fine scale

What is the study design for pairing occupied and random sites?

• Compare known occupied habitat to random nearby site to assess difference

• Instead of assuming everything is available-> survey for where the animal is

  • Why is it here and not here -> paired site

  • Find the occupied wetland habitat to a site that is not occupied/unknown

    • What are the specific differences between the two sites

• Increase statistical power

What do paired experimental designs offer?

• stronger statistical power to detect selections and are often preferable if possible

• involve comparing used habitat sites with available habitat sites that are closely matched (or "paired") based on specific criteria like location, time, or habitat type. 

• Do not require marked individuals of known identity

What do paired experimental designs control for?

controls for variation

• helps isolate the effect of habitat features on selection

Why is it important to ask, "What is actually available?" when studying habitat use?

Animals can only choose from habitats that are available to them, not everything in the landscape is going to be available

• Misjudging what’s available leads to misleading conclusions about preference or selection

What does it mean for your habitat sampling to be representative?

sample should reflect the true variety and proportions of habitat types available to the population

• If not, your results may be biased and not generalizable

What does it mean to measure the right predictors?

You need to choose variables that are biologically relevant and connected to your research question

• Measuring irrelevant ones adds noise and increases the risk of overparameterization.

pooling

combining data from multiple individuals or groups

Why can pooling be a problem?

• can mask inter-individual variation

  • Can hide specializations, unique groups, etc

  • Cancel out and mask some variation

  • Loose nuance when pooling data together

Overparameterization

When a model includes too many variables (predictors), making it:

• Harder to interpret

• More likely to detect false positives (Type I errors)

• Less statistically powerful

Why does testing too many predictors increase the risk of a false positive?

• With an alpha level (α) of 0.05, there’s a 5% chance of a false positive for each test.

• If you test 20 variables, the chance of getting at least one false positive is about 64-65%, even if none are truly significant

How does the Bonferroni correction help?

It adjusts the alpha level by dividing it by the number of tests to reduce the likelihood of false positives.

• Example: If α = 0.05 and you're testing 10 predictors → use α = 0.005 for each.

Why are models with many predictors problematic?

• Increase the chance of random significance (false positives)

• Make the model more complex than necessary

• Often include correlated or unimportant variables

What does independence of observations mean in ecological studies?

Each data point (observation) should be statistically independent

• not influenced by another

• If one observation affects another, your analysis could be invalid

How is independence violated in social animals?

• In social species (e.g., flocks, herds), individuals are not independent

  • seeing one increases the chance of seeing another.

• So the group, not the individual, should be treated as the unit of analysis

What is the most common and straightforward test for comparing use/availability?

χ2 tests

• No apparent habitat selection vs what is observed

  • No apparent selection-> chance should be 50/50 for use of different habitat

X2 test

• compares observed data vs what actually happens

  • Does this thing happen because its available or because of other reasons