Spatially explicit mark resight models

[Andreu Rotger Vallespir]

Hi group, 

I need hel to excute a SECR sighting-only model using 20 camera traps identifiyind individuals ONLY with natural marks to estimate density of feral cats . 

I have 26 known animals recognized by their natural patterns. All animals with patterns have been indentified, I mean, I don't have no unknown marked animals, and I have 19 unmarked detections (no patterns black cats) within the study area. 

Data it is updated correctly, but when I ran the model warning messages always appear and I don't know why these messages appears:

> CH <- read.capthist("capthist.txt", "Cameras.txt", detector =

+                              "proximity",fmt = "trapID", markocc = c(rep(0, 59)),

+                            trapcovnames="usage")

No errors found :-)

> CH <- addSightings(CH, unmarked = 'unmarked.txt', verify = TRUE)

No errors found :-)

> fit.HN <- secr.fit (CH, mask = Pianosamask, detectfn = 'HN', 

+                     trace = FALSE,fixed = list(pID = 1.0),details = list(knownmarks = TRUE))

Warning messages:

1: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 5

2: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 22

3: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 27

4: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 32

5: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 37

6: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 42

7: In (function (data, sessnum)  :

  invalid distribution for sighting at Eval 47

8: In secr.fit(CH, mask = Pianosamask, detectfn = "HN", trace = FALSE,  :

  possible maximization error: nlm returned code 3. See ?nlm

I don't how to deal with this warning messages. However, the model run and provide estimates.

Any help?

Thanks,

Best

Andreu

[Murray Efford]

Andreu

The case of 'sightings only when the number marked is unknown' (Section 4.2 in Efford & Hunter 2017) can be problematic. Warnings 1-7 arise because the number of individuals observed (n) is more than actually present in the model for the parameter values at that step in the likelihood maximisation. It can probably be ignored if the algorithm goes on to produce estimates. The last warning (8) often reflects trouble in estimating the variances, so to be safe I would recalculate those with

fit.HN2 <- secr.fit (CH, mask = Pianosamask, detectfn = 'HN',  start = fit.HN, method = 'none',

                     trace = FALSE,fixed = list(pID = 1.0),details = list(knownmarks = TRUE))

This applies a better algorithm.

Murray

[Andreu Rotger Vallespir]

Thank you Murray, in this study the number of marked individuals in Known because cats presented clear pelage patterns, except the black ones, I considered black cats (without patterns) as unmarkedl is it correct?
I tried run with this code and it worked.

Best,

Andreu