## Using SECR to figure out individual home range sizes

questions concerning anlysis/theory using program DENSITY and R package secr. Focus on spatially-explicit analysis.

### Using SECR to figure out individual home range sizes

Dear all,

My interest is about the influence of temperament on the reproductive success in a squirrel population. For 9 years, the population has been monitored using Capture-Mark-Recapture procedure, 8 months a year (3 or 5 consecutive trapping days). The trapping grid consists of 104 live traps.

To access temperament traits indices, we want to classify each squirrel based on traps informations and home range size (activity center). All indices have been found biased due to the edge of the trapping grid (as individuals living close to the edge have underestimated temperament indices).
Then, I would like to input the spatial detection histories in SECR in order to estimate i) individual activity center (area) and ii) the part of individual home-range size which is outside of the trapping grid. This area will be useful as a (an edge effect) bias correction of temperament traits indices.

1) Does this approach (i.e using SECR at individual scale) make sense? I am concerned that the calculated home ranges truly match observed individuals.
2) Using fxi.contour and extractmode functions allow me to plot the contour of individual home-range and barycenter. Is there a way to obtain the value of each home-range size (square meter)? If not, are there any functions in SECR to calculate the area based on coordinates from individual polygon?
3) As an alternative way, what about using RPSV value (not pooled over individual indeed) as measurement of individual home-range size?

C. Le Coeur
C. Le Coeur

Posts: 13
Joined: Mon Feb 13, 2012 9:36 am

### Re: Using SECR to figure out individual home range sizes

You are right to think the spatial scale parameter from SECR (usually 'sigma') can be used to indicate home range size free of effects from the trap layout. However, it is probably too much to expect a measurement of sigma for each individual, and the 'secr' package is not set up to do that except in the following sense.

In a model fitted by maximizing the conditional likelihood (CL = T), you can model sigma as a function of one or more individual covariates. If an independent temperament score was one of your covariates, AIC comparison of models with and without such an effect would be a natural way to go. Presumably you might also use an outcome like reproductive success as a covariate. It is possible to predict sigma from the fitted model for each combination of covariates, and potentially for each individual (to do this you set up the 'newdata' dataframe argument of predict, one row per level at which you want a prediction).

[If you had a great deal of data on each individual it might be possible, in principle, to extract a unique sigma for each one - I'm not sure - but the covariate approach is probably more elegant in any case.]

You do have a lot of data, over many sessions. In principle this is great, but I can see some of these models taking a long time to fit if you try to do it in one go!

fxi.contour concerns the most likely location of the home-range centre of a detected animal, given the data and the fitted model. It does not describe home range per se.

RPSV is a pooled statistic that depends on trap layout: it won't give you what you want.

The scale parameter sigma loosely translates to range area (for a halfnormal detection function, 95% of activity is within a 2.45 sigma circle - see help for circular.r).

I hope this is enough to get you moving.
Murray
murray.efford

Posts: 614
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand

### Re: Using SECR to figure out individual home range sizes

I'd suggest that--regardless of the computational limitations involved with estimating an individual sigma--SECR (the approach rather than the package!) may not provide a real accurate measure of an individual's home range size. My experience has been that 2-3x the sigma value provides a roughly accurate measure of home range size vs. telemetry data, but there's been huge individual variance, and estimated home range centers disagree with radio-collared locations (this is another topic altogether, but I find that secr provides higher density estimates than ad-hoc MMDM). This depends on detectability, of course, but Andy Royle's recent papers incorporating landscape resistance/RSF also make conceptual sense (though I haven't found that those models fit really plausibly with my own data).

I guess if you have a pretty high rate of individual detection and null g0 and sigma models seem to fit well you could assume a uniform home range size (?).

Just my two cents. Good luck!

John
JDJC

Posts: 10
Joined: Fri Sep 28, 2012 4:05 pm

### Re: Using SECR to figure out individual home range sizes

Dear Murray,
for a halfnormal detection function, 95% of activity is within a 2.45 sigma circle. How about for exponential function? By which factor should I multiply sigma if not 2.45?
Many thanks,
Valeria
valeria.boron

Posts: 7
Joined: Mon Aug 13, 2012 6:34 am

### Re: Using SECR to figure out individual home range sizes

Code: Select all
`library(secr)circular.r(detectfn='HHN')# [1] 2.447747circular.r(detectfn='HEX')# [1] 4.743868?circular.r`
murray.efford

Posts: 614
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand

### Re: Using SECR to figure out individual home range sizes

Thank you so much
valeria.boron

Posts: 7
Joined: Mon Aug 13, 2012 6:34 am