Create a set of capture or marking-and-resighting histories by simulated sampling of a 2-D population using an array of detectors.
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)traps object with the locations and other attributes of detectors sim.capthist mask objectFor sim.capthist, an object of class capthist, a matrix or
3-dimensional array with additional attributes. Rows represent
individuals and columns represent occasions; the third dimension, used
when detector type = `proximity', codes presence or absence at each
detector. For trap detectors (`single', `multi') each entry in
capthist is either zero (no detection) or the sequence number of
the trap.
The initial state of the R random number generator is stored in the `seed' attribute.
For sim.resight, an object of class capthist, always a
3-dimensional array, with additional attributes markocc (marking occasions),
Tu (sightings of unmarked animals) and Tm (sightings of marked but not
identified animals).
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 a numeric code (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.
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.
sim.popn, capthist, traps,
popn, Detection functions, simulate## 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
Run the code above in your browser using DataLab