| possum {secr} | R Documentation |
Data from a trapping study of brushtail possums at Waitarere, North Island, New Zealand.
possumCH
possumarea
possumremovalarea
possummask
possum.model.0
possum.model.Ds
Brushtail possums (Trichosurus vulpecula) are an unwanted invasive species in New Zealand. Although most abundant in forests, where they occasionally exceed densities of 15 / ha, possums live wherever there are palatable food plants and shelter.
Efford et al. (2005) reported a live-trapping study of possums in Pinus radiata plantation on coastal sand dunes. The 300-ha site at Waitarere in the North Island of New Zealand was a peninsula, bounded on one side by the sea and on two other sides by the Manawatu river. Cage traps were set in groups of 36 at 20-m spacing around the perimeter of five squares, each 180 m on a side. The squares (‘hollow grids’) were centred at random points within the 300-ha area. Animals were tagged and released daily for 5 days in April 2002. Subsequently, leg-hold trapping was conducted on a trapping web centred on each square (data not reported here), and strenuous efforts were made to remove all possums by cyanide poisoning and further leghold trapping across the entire area. This yielded a density estimate of 2.26 possums / ha.
Traps could catch at most one animal per day. The live-trapped animals comprised 46 adult females, 33 adult males, 10 immature females and 11 immature males; sex and/or age were not recorded for 4 individuals (M. Coleman unpubl. data). These counts do not sum to the number of capture histories - see Note. One female possum was twice captured at two sites on one day, having entered a second trap after being released; one record in each pair was selected arbitrarily and discarded.
The data are provided as a single-session capthist object
‘possumCH’. ‘possummask’ is a matching mask object - see Examples.
Two fitted models are provided for illustration.
The dataframe possumarea contains boundary coordinates of a
habitat polygon that is used to clip possummask at the shore
(from secr 1.5). possumarea comprises a single polygon
representing the extent of terrestrial vegetation to the west, north and
east, and an arbitrary straight southern boundary. The boundary is also
included as a shapefile and as a text file (‘possumarea.shp’ etc. and
‘possumarea.txt’ in the package ‘extdata’ folder). See Examples in
make.mask.
The dataframe possumremovalarea contains boundary coordinates of
another polygon, the nominal removal area of Efford et al. (2005 Fig. 1)
(from secr 2.3).
| Object | Description |
| possumCH | capthist object |
| possummask | mask object |
| possumarea | habitat perimeter |
| possumremovalarea | nominal boundary of removal region |
| possum.model.0 | fitted secr model -- null |
| possum.model.Ds | fitted secr model -- distance to shore |
A significant problem with the data used by Efford et al. (2005) was noticed recently. Five capture histories in possumCH are for animals that had lost a previous tag. A further three histories may also have been animals that were tagged previously or mis-recorded. Analyses that treat each previously tagged animal as a new individual are in error (this includes the published analyses, the pre-fitted models described here, and those in the vignette secr-densitysurfaces.pdf). All eight questionable histories are now indicated in possumCH with the logical covariate ‘prev.tagged’.
Methods have not yet been developed to adjust for tag loss in SECR models.
Landcare Research, New Zealand.
Borchers, D.L. and Efford, M.G. (2008) Spatially explicit maximum likelihood methods for capture-recapture studies. Biometrics 64, 377–385.
Efford, M. G., Dawson, D. K. and Robbins C. S. (2004) DENSITY: software for analysing capture-recapture data from passive detector arrays. Animal Biodiversity and Conservation 27, 217–228.
Efford, M. G., Warburton, B., Coleman, M. C. and Barker, R. J. (2005) A field test of two methods for density estimation. Wildlife Society Bulletin 33, 731–738.
plot(possummask)
plot(possumCH, tracks = TRUE, add = TRUE)
plot(traps(possumCH), add = TRUE)
lines(possumarea)
summary(possumCH)
## compare & average pre-fitted models
AIC(possum.model.0, possum.model.Ds)
modelAverage(possum.model.0, possum.model.Ds)
## Not run:
## Roughly estimate tag-loss error by dropping dubious histories
## i.e. restrict to "not previously tagged"
NPT <- !covariates(possumCH)$prev.tagged
possum.model.0.NPT <- secr.fit(subset(possumCH,NPT), mask =
possummask, trace = FALSE)
predict(possum.model.0)[1,2]/ predict(possum.model.0.NPT)[1,2]
## ...about 9%
## End(Not run)