A habitat mask serves four main purposes in spatially explicit
capture--recapture. Firstly, it defines an outer limit to the area of
integration; habitat beyond the mask may be occupied, but animals there
should have negligible chance of being detected (see pdot
and below). Secondly, it distinguishes sites in the vicinity of the
detector array that are `habitat' (i.e. have the potential to be
occupied) from `non-habitat'. Thirdly, it discretizes continuous habitat
as a list of points. Each point is notionally associated with a cell
(pixel) of uniform density. Discretization allows the SECR likelihood to
be evaluated by summing over grid cells. Fourthly, the x-y coordinates
of the mask and any habitat covariates may be used to build spatial
models of density. For example, a continuous or categorical habitat
covariate `cover' measured at each point on the mask might be used in a
formula for density such as D $~$cover.
In relation to the first purpose, the definition of `negligible' is
fluid. Any probability less than 0.001 seems OK in the sense of not
causing noticeable bias in density estimates, but this depends on the
shape of the detection function (fat-tailed functions such as `hazard
rate' are problematic). New tools for evaluating masks appeared in
secr 1.5 (mask.check, esa.plot), and
suggest.buffer automates selection of a buffer width.
Mask points are stored in a data frame with columns `x' and `y'. The number of rows equals the number of points.
Possible mask attributes
| Attribute |
| Description |
| type |
| `traprect', `trapbuffer', `pdot', `polygon', `clusterrect', `clusterbuffer' (see make.mask) or `user' |
| polygon |
| vertices of polygon defining habitat boundary, for type = `polygon' |
| pdotmin |
| threshold of p.(X) for type = `pdot' |
| covariates |
| dataframe of site-specific covariates |
| meanSD |
| data frame with centroid (mean and SD) of x and y coordinates |
| area |
| area (ha) of the grid cell associated with each point |
| spacing |
| nominal spacing (metres) between adjacent points |
| boundingbox |
| data frame of 4 rows, the vertices of the bounding box of all grid cells in the mask |
Attributes other than covariates are generated automatically by
make.mask. Type `user' refers to masks input from a text file
with read.mask.
A virtual S4 class `mask' is defined to allow the definition of a method
for the generic function raster from the raster package.
make.mask, read.mask,
mask.check, esa.plot,
suggest.buffer, secr.fit,
secr density models