Query regarding Multi-session SECR analysis

[Aishwarya Karanth]

Hello everyone,

I am currently working on the multi-session SECR analysis for estimating leopard density in southern India.

 There are two sites (Cauvery and MM hills). They technically belong to one single landscape but due to administrative purpose they are divided as two separate sites. We have 4 year leopard data for both the sites and want to run multi-session SECR. However sampling period for both the site in each year for both sites are different (Ex: Cauvery-14 Jan to 14 March 2014 and MM hills-14 Oct to 14 Dec 2014). Also we have got 36 common individuals between the sites.  I wanted to know if its logical to combine both the sites and do the analysis.

[Murray Efford]

I think it's hard to choose a closed SECR model that strictly matches the sampling. Common individuals really require that the two sites are analysed together. 

Assuming you use 'secr' I suggest 

1. combining the two sites in one session with a habitat mask that covers both, and total number of occasions s1 + s2,

2. setting 'usage' to zero for occasions and sites that were not sampled (e.g. occasions 1:s1 for site 2) (see helper function below), and

3. modelling within-year density differences with a mask covariate that separates the mask into sectors corresponding to the two sites.

In the combined model, the within-year differences may be temporal or spatial - you can't tell.

The alternative is an open population model in which each site in each year contributes a single session. This would in theory allow for the overlapping sets of known individuals (both between sites and over years), but there are some restrictions (you would use a single, inclusive, mask throughout). For interest I would also try a closed model ignoring the overlap (8 separate sessions); that sacrifices some within-year recaptures and assumes independence where there is none.

I hope these thoughts are useful. Either way, setting up sessions and usage is going to be a bit fiddly.

Murray

# combine two traps objects, assigning non-overlapping usage

trapmerge <- function (traps1, traps2, noccasions){
    if (!is.null(usage(traps1)) | !is.null(usage(traps2)))
        warning("discarding existing usage data")
    if (length(noccasions)!=2)
        stop ("two occasion numbers required")
    site2 <- rep(0:1, noccasions)
    usage(traps1) <- matrix(1-site2, byrow = TRUE,
                            nrow = nrow(traps1), ncol = sum(noccasions))
    usage(traps2) <- matrix(site2, byrow = TRUE,
                            nrow = nrow(traps2), ncol = sum(noccasions))
    rbind(traps1, traps2)
}