Simulating Space Use of Animals from RSF and SSF Johannes Signer ( - - PowerPoint PPT Presentation

▶
simulating space use of animals from rsf and ssf
SMART_READER_LITE
LIVE PREVIEW

Simulating Space Use of Animals from RSF and SSF Johannes Signer ( - - PowerPoint PPT Presentation

Dec 26, 2022 8 likes •341 views

Simulating Space Use of Animals from RSF and SSF Johannes Signer ( signer_j) Wildlife Sciences, Georg-August-Universitt Gttingen Movebank Workshop SCENE 2019-05-02 Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 1/18

slide-1
SLIDE 1

Simulating Space Use of Animals from RSF and SSF

Johannes Signer ( signer_j)

Wildlife Sciences, Georg-August-Universität Göttingen Movebank Workshop SCENE

2019-05-02

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 1/18

slide-2
SLIDE 2

Problem: How to quantify and predict space use by animals?

  • 1. Space use: usually summarized in terms of a 2-D (or 3-D)

utilization distribution that captures the relative frequency of time spent in different locations.

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 2/18

slide-3
SLIDE 3

Problem: How to quantify and predict space use by animals?

  • 1. Space use: usually summarized in terms of a 2-D (or 3-D)

utilization distribution that captures the relative frequency of time spent in different locations.

  • 2. How to obtain accurate estimates of space use?

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 2/18

slide-4
SLIDE 4

Problem: How to quantify and predict space use by animals?

  • 1. Space use: usually summarized in terms of a 2-D (or 3-D)

utilization distribution that captures the relative frequency of time spent in different locations.

  • 2. How to obtain accurate estimates of space use?
  • 3. Is it possible to predict space use of animals in novel or altered

landscapes?

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 2/18

slide-5
SLIDE 5

Why not use home ranges?

  • Traditional home-range concept1 is complex and nontrivial to

quantify.

1Burt, W. (1943). Territoriality and home range concepts as applied to mammals. Journal of mammalogy,

24(3), 346-352.

2Signer, J. et al. (2017). Estimating utilization distributions from fitted step-selection functions. Ecosphere,

8(4), e01771. Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 3/18

slide-6
SLIDE 6

Why not use home ranges?

  • Traditional home-range concept1 is complex and nontrivial to

quantify.

  • Most home range estimators do not provide a mechanistic

model linking space use to habitat characteristics and movement → prediction.

1Burt, W. (1943). Territoriality and home range concepts as applied to mammals. Journal of mammalogy,

24(3), 346-352.

2Signer, J. et al. (2017). Estimating utilization distributions from fitted step-selection functions. Ecosphere,

8(4), e01771. Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 3/18

slide-7
SLIDE 7

Why not use home ranges?

  • Traditional home-range concept1 is complex and nontrivial to

quantify.

  • Most home range estimators do not provide a mechanistic

model linking space use to habitat characteristics and movement → prediction.

  • Simulations from integrated Step Selection Functions (iSSFs)

are an interesting alternative to home ranges to quantify space use2.

1Burt, W. (1943). Territoriality and home range concepts as applied to mammals. Journal of mammalogy,

24(3), 346-352.

2Signer, J. et al. (2017). Estimating utilization distributions from fitted step-selection functions. Ecosphere,

8(4), e01771. Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 3/18

slide-8
SLIDE 8

Integrated Step Selection Functions (iSSFs)

  • Estimate distribution for step lengths and turning angles.
  • Pair each observed step with J random steps.
  • Extract covariate values at the end of each step.
  • Estimate selection coefficients β with a conditional logistic

regression.

1Avgar, T. et al. (2016). Integrated step selection analysis: bridging the gap between resource selection and

animal movement. Methods in Ecology and Evolution, 7(5), 619-630. Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 4/18

slide-9
SLIDE 9

Integrated Step Selection Functions (iSSFs)

  • Estimate distribution for step lengths and turning angles.
  • Pair each observed step with J random steps.
  • Extract covariate values at the end of each step.
  • Estimate selection coefficients β with a conditional logistic

regression.

  • iSSF: including movement related covariates (e.g., step length

and turning angles) is equivalent to fitting a biased correlated random walk to the data1.

1Avgar, T. et al. (2016). Integrated step selection analysis: bridging the gap between resource selection and

animal movement. Methods in Ecology and Evolution, 7(5), 619-630. Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 4/18

slide-10
SLIDE 10

Integrated Step-Selection Functions (iSSF)

  • −8

−6 −4 −2 2 4 −6 −4 −2 2 4 6 x y

  • Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner)

5/18

slide-11
SLIDE 11

Integrated Step-Selection Functions (iSSF)

  • −8

−6 −4 −2 2 −6 −4 −2 2 x y

  • Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner)

5/18

slide-12
SLIDE 12

Integrated Step-Selection Functions (iSSF)

  • −8

−6 −4 −2 2 −6 −4 −2 2 x y

  • Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner)

5/18

slide-13
SLIDE 13

Integrated Step-Selection Functions (iSSF)

  • −8

−6 −4 −2 2 −6 −4 −2 2 x y

  • Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner)

5/18

slide-14
SLIDE 14

A case study: red deer in Germany

  • 24 red deer collared in northern Germany from 2008 to 2013
  • 6 hours sampling rate (the number of relocations range from

430 to 3600)

  • Each observed step was paired with 9 random steps
  • iSSF as mixed Poisson Regression1 with package amt2

1Muff, S. et al. (2018). Accounting for individual-specific variation in habitat-selection studies: Efficient

estimation of mixed-effects models using Bayesian or frequentist computation. bioRxiv, 411801.

2Signer, J. et al. (2018. Animal Movement Tools (amt): R-Package for Managing Tracking Data and

Conducting Habitat Selection Analyses. arXiv preprint arXiv:1805.03227. Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 6/18

slide-15
SLIDE 15

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 7/18

slide-16
SLIDE 16

With the following covariates

  • Land cover (forest or open)
  • Distance to urban areas
  • Distance to home-range center
  • Step length
  • Interactions with time of day

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 8/18

slide-17
SLIDE 17

Results

Fixed effects: Term Estimate Forest (time of day = day) 2.36∗∗∗ Forest (time of day = night) −3.42∗∗∗

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 9/18

slide-18
SLIDE 18

Results

Fixed effects: Term Estimate Forest (time of day = day) 2.36∗∗∗ Forest (time of day = night) −3.42∗∗∗ Distance to urban (time of day = day) 0.26∗ Distance to urban (time of day = night) −0.39∗∗

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 9/18

slide-19
SLIDE 19

Results

Fixed effects: Term Estimate Forest (time of day = day) 2.36∗∗∗ Forest (time of day = night) −3.42∗∗∗ Distance to urban (time of day = day) 0.26∗ Distance to urban (time of day = night) −0.39∗∗ Distance to center (time of day = day) −3.36∗∗∗ Distance to center (time of day = night) 3.28∗

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 9/18

slide-20
SLIDE 20

Results

Fixed effects: Term Estimate Forest (time of day = day) 2.36∗∗∗ Forest (time of day = night) −3.42∗∗∗ Distance to urban (time of day = day) 0.26∗ Distance to urban (time of day = night) −0.39∗∗ Distance to center (time of day = day) −3.36∗∗∗ Distance to center (time of day = night) 3.28∗ log(step length) (time of day = day) −0.11∗∗∗ log(step length) (time of day = night) 0.46∗∗∗

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 9/18

slide-21
SLIDE 21

Underlying step-length distribution differs between day and night:

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 10/18

slide-22
SLIDE 22

Simulate and predict space use from fitted iSSF

  • 1. A typical animal (fixed effects only)
  • 2. Use random effects of a specific animal
  • 3. For prediction: random effects of a similar animal (in

environmental space)

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 11/18

slide-23
SLIDE 23

A typical animal (fixed effects only)

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 12/18

slide-24
SLIDE 24

A typical animal (fixed effects only)

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 12/18

slide-25
SLIDE 25

This animal (random effects)

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 13/18

slide-26
SLIDE 26

This animal (random effects)

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 13/18

slide-27
SLIDE 27

Predict space use in a novel environment

Find animal that is closest to the new environment in environmental space...

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 14/18

slide-28
SLIDE 28

... and predict space use in novel environment.

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 15/18

slide-29
SLIDE 29

... and predict space use in novel environment.

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 15/18

slide-30
SLIDE 30

... and predict space use in novel environment.

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 15/18

slide-31
SLIDE 31

Summary and outlook

  • Space use depends on time of day and the environment.
  • iSSFs provides a simple but powerful mechanistic movement

model, that allows simulations.

  • We are working on more sophisticated simulations (time

varying covariates).

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 16/18

slide-32
SLIDE 32

I Appendix

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 17/18

slide-33
SLIDE 33

Model

yntj = Poisson(λntj) log(λntj) = αnt + β1nforest + β2ndist_urban + β3nlog_sl + β4nforestnight + β5ndisturbannight + β6ndist_cent + β7nlog_slnight + β8ndist_cent + β9ndist_centnight With

  • n = 1 . . . N individuals
  • t = 1 . . . Tn time points (= strata)
  • j = 1 . . . J steps per stratum.
  • yntj = 1 for observed steps and 0 for random steps.
  • αnt ∼ N(0, 106)

Random effects were uncorrelated.

Johannes Signer ( jsigner@gwdg.de, signer_j, jmsigner) 18/18