sim.capthist: Simulate Detection Histories

In secr: Spatially Explicit Capture-Recapture

Simulate Detection Histories

Description

Create a set of capture or marking-and-resighting histories by simulated sampling of a 2-D population using an array of detectors.

Usage

sim.capthist(traps, popn = list(D = 5, buffer = 100,
    Ndist = "poisson"), detectfn = 0, detectpar = list(),
    noccasions = 5, nsessions = 1, binomN = NULL, exactN = NULL,
    p.available = 1, renumber = TRUE, seed = NULL,
    maxperpoly = 100, chulltol = 0.001, userdist = NULL, 
    savepopn = FALSE)
sim.resight(traps, popn = list(D = 5, buffer = 100, Ndist = "poisson"), ...,
    pID = 1, unmarked = TRUE, nonID = TRUE,  unresolved = FALSE, unsighted = TRUE, 
    pmark = 0.5, Nmark = NULL, markingmask = NULL)

Arguments

traps	traps object with the locations and other attributes of detectors
popn	locations of individuals in the population to be sampled, either as a popn object (see sim.popn) or a list with the named components.
detectfn	integer code or character string for shape of detection function 0 = halfnormal etc. – see detectfn
detectpar	list of values for named parameters of detection function
noccasions	number of occasions to simulate
nsessions	number of sessions to simulate
binomN	integer code for distribution of counts (see Details)
exactN	integer number of telemetry fixes per occasion
p.available	vector of one or two probabilities (see Details)
renumber	logical for whether output rows should labeled sequentially (TRUE) or retain the numbering of the population from which they were drawn (FALSE)
seed	either NULL or an integer that will be used in a call to set.seed
maxperpoly	integer maximum number of detections of an individual in one polygon or transect on any occasion
chulltol	numeric buffer (m) for polygon around telemetry locations
userdist	user-defined distance function or matrix (see details)
savepopn	logical; if TRUE then the popn (input or simulated) is saved as an attribute
...	arguments to pass to sim.capthist
pID	probability of individual identification for marked animals
unmarked	logical, if TRUE unmarked individuals are not recorded during ‘sighting’
nonID	logical, if TRUE then unidentified marked individuals are not recorded during ‘sighting’
unresolved	logical, if TRUE then individuals of unresolved mark status are not recorded during ‘sighting’
unsighted	logical, if TRUE and sighting only then capthist includes all-zero histories
pmark	numeric probability that an individual is ‘pre-marked’ (see Details)
Nmark	number of individuals to be ‘pre-marked’ (see Details)
markingmask	mask object

Details

If popn is not of class ‘popn’ then a homogeneous Poisson population with the desired density (animals/ha) is first simulated over the rectangular area of the bounding box of traps plus a buffer of the requested width (metres). The detection algorithm depends on the detector type of traps. For ‘proximity’ detectors, the actual detection probability of animal i at detector j is the naive probability given by the detection function. For ‘single’ and ‘multi’ detectors the naive probability is modified by competition between detectors and, in the case of ‘single’ detectors, between animals. See Efford (2004) and other papers below for details.

Detection parameters in detectpar are specific to the detection function, which is indicated by detectfn. Parameters may vary with time - for this provide a vector of length noccasions. The g0 parameter may vary both by time and detector - for this provide a matrix with noccasions rows and as many columns as there are detectors. The default detection parameters are list(g0 = 0.2, sigma = 25, z = 1).

The default is to simulate a single session. This may be overridden by providing a list of populations to sample (argument popn) or by specifying nsessions > 1 (if both then the number of sessions must match). Using nsessions > 1 results in replicate samples of populations with the same density etc. as specified directly in the popn argument.

binomN determines the statistical distribution of the number of detections of an individual at a particular ‘count’ detector or polygon on a particular occasion. A Poisson distribution is indicated by binomN = 0; see secr.fit for more. The distribution is always Bernoulli (binary) for ‘proximity’ and ‘signal’ detectors.

If exactN is not specified or zero then the number of telemetry fixes is a random variable determined by the other detection settings.

p.available specifies temporary non-availability for detection in multi-session simulations. If a single probability is specified then temporary non-availability is random (independent from session to session). If two probabilities are given then non-availability is Markovian (dependent on previous state) and the two values are for animals available and not available at the preceding session. In the Markovian case, availability in the first session is assigned at random according to the equilibrium probability p2 / (1 - p1 + p2). Incomplete availability is not implemented for sampling lists of populations.

detectpar may include a component ‘truncate’ for the distance beyond which detection probability is set to zero. By default this value is NULL (no specific limit).

detectpar may also include a component ‘recapfactor’ for a general learned trap response. For ‘single’ and ‘multi’ detector types the probability of detection changes by this factor for all occasions after the occasion of first capture. Attempted use with other detector types causes an error. If recapfactor x g(d) > 1.0, g(d) is truncated at 1.0. Other types of response (site-specific bk, Markovian B) are not allowed.

If popn is specified by an object of class ‘popn’ then any individual covariates will be passed on; the covariates attribute of the output is otherwise set to NULL.

The random number seed is managed as in simulate.

chulltol is used only when simulating telemetry locations. By default, a new 'traps' polygon is generated as the convex hull of the simulated locations, with a slight (1 mm) added buffer to ensure boundary points are within the polygon. Buffering is suppressed if chulltol is NA or negative.

userdist cannot be set if ‘traps’ is any of polygon, polygonX, transect or transectX.

sim.resight generates mark-resight data. The ‘markocc’ attribute of ‘traps’ indicates the occasions which are for sighting-only (0) or marking and recapture (1). The number of occasions is determined by markocc. sim.capthist is first called with the arguments ‘traps’ and .... The detector type of ‘traps’ should be ‘proximity’ or ‘count’ for sighting occasions (markocc = 0). The detector type need not be the same for marking and sighting occasions ('multi' is allowed on marking occasions). If ... includes a non-null ‘seed’ the random seed is reset in sim.resight and not passed to sim.capthist.

A special case arises when all occasions are sighting-only. Then it is assumed that individuals in the population are marked prior to the start of sampling with a known spatial distribution (i.e. marking does not follow a spatial detection model). By default, animals throughout the buffered area are pre-marked with probability pmark. If Nmark is specified then a sample of size Nmark will be selected for marking, overriding pmark.

The marked population may be restricted to a subset of the space spanned by popn by specifying markingmask, which may have a further covariate ‘marking’ to vary the intensity of marking.

Value

For sim.capthist, an object of class capthist, a 3-dimensional array with additional attributes. Rows represent individuals and columns represent occasions; the third dimension, codes the number of detections at each detector (zero or one for trap detectors (‘single’, ‘multi’) and binary proximity detectors.

The initial state of the R random number generator is stored in the ‘seed’ attribute.

For sim.resight, an object of class capthist for which the traps object has a markocc attribute (marking occasions), and there are further attributes Tu (sightings of unmarked animals) and Tm (sightings of marked but not identified animals).

Note

External code is called to speed the simulations. The present version assumes a null model, i.e., naive detection probability is constant except for effects of distance and possibly time (using vector-valued detection parameters from 1.2.10). You can, however, use rbind.capthist to combine detections of population subclasses (e.g. males and females) simulated with different parameter values. This is not valid for detector type "single" because it fails to allow for competition for traps between subclasses. Future versions may allow more complex models.

truncate has no effect (i) when using a uniform detection function with radius (sigma) <= truncate and (ii) with signal strength detection (detectfn 10, 11). Note that truncated detection functions are provided for de novo simulation, but are not available when fitting models with in secr.fit or simulating from a fitted model with sim.secr.

maxperpoly limits the size of the array allocated for detections in C code; an error results if the is number is exceeded.

Prior to 2.10.0 sim.resight interpreted length-2 vectors of detection parameters as referring to marking and sighting occasions; this feature has been discontinued.

References

Borchers, D. L. and Efford, M. G. (2008) Spatially explicit maximum likelihood methods for capture–recapture studies. Biometrics 64, 377–385.

Efford, M. G. (2004) Density estimation in live-trapping studies. Oikos 106, 598–610.

Efford, M. G., Borchers D. L. and Byrom, A. E. (2009) Density estimation by spatially explicit capture-recapture: likelihood-based methods. In: D. L. Thomson, E. G. Cooch and M. J. Conroy (eds) Modeling Demographic Processes in Marked Populations. Springer, New York. Pp. 255–269.

Efford, M. G., Dawson, D. K. and Borchers, D. L. (2009) Population density estimated from locations of individuals on a passive detector array. Ecology 90, 2676–2682.

See Also

sim.popn, capthist, traps, popn, Detection functions, simulate, Tu, Tm

Examples

## simple example
## detector = "multi" (default)
temptrap <- make.grid(nx = 6, ny = 6, spacing = 20)
sim.capthist (temptrap, detectpar = list(g0 = 0.2, sigma = 20))

## with detector = "proximity", there may be more than one
## detection per individual per occasion
temptrap <- make.grid(nx = 6, ny = 6, spacing = 20, detector =
    "proximity")
summary(sim.capthist (temptrap, detectpar = list(g0 = 0.2, 
    sigma = 20)))

## marking on occasions 1, 3 only
temptrap <- make.grid(nx = 6, ny = 6, spacing = 20, detector = 'proximity')
markocc(temptrap) <- c(1,0,1,0,0)
CH <- sim.resight (temptrap, detectpar = list(g0 = 0.2, sigma = 20))
summary(CH)

## multiple sessions
grid4 <- make.grid(nx = 2, ny = 2)
temp <- sim.capthist (grid4, popn = list(D = 1), nsessions = 20)
summary(temp, terse = TRUE)

## unmarked or presence types
# grid <- make.grid(nx = 10, ny = 10, detector = "unmarked")
# CH <- sim.capthist (grid, noccasions = 5)
# CH
## "presence" and "unmarked" data are stored as "count" data
## behaviour is controlled by detector type, e.g.
# detector(traps(CH)) <- "presence"
# CH

secr documentation built on Nov. 4, 2024, 9:06 a.m.