Re: [unmarked] Abundance value in relation to determining population size

[Jeffrey Royle]

hi Sarah,
 if you are doing a fixed radius point count, then it might be reasonable to use the estimated abundance per point and divided it by the area of the fixed radius point count, for purposes of extrapolation.
 However, there are a lot of caveats related to non-closure and the length of time between sample intervals, etc.. You should read a few of the cited papers on this before investing too much.

 I would use a capture-recapture type of method -- in what sense are these not feasible?

regards
andy

On Sun, Mar 17, 2013 at 8:49 AM, <saver...@hotmail.com> wrote:

Hi,

I am from the Mauritian Reptile Recovery Programme, Mauritius. I am working on designing a monitoring method to determine the population size of a translocated population (to an offshore island) of Guenther gecko, Phelsuma guentherii. This is a highly crypic species at low density, however the population has increased to the point where Capture-Mark-Recapture methods are no longer feasible  Therefore a point count methodology is being used, and I am using the program Unmarked using 'pcount' to determine the abundance of these geckos at each point count.

Sorry if it seems a simple question, but I was wondering is the value for 'abundance' the abundance per area (i.e. density)? I understand this model works at a metapopulation level, but basically what I need to work out from this is a value of the population size for the whole island. How can I extrapolate the values from the point counts to determine the population size for the whole island? 

Thanks in advance for your help,

Sarah

--
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[saver...@hotmail.com]

Thanks for your reply Andy,

Before collecting the data I marked out a fixed radius using flagging tape so the area of each point count is known. I have treated it as a closed population (using pcount rather than pcountopen) due to the population being on a small offshore island (24 hectares) and collected the data in a tight time frame (all the data-with 3 repeats of the sites-was collected within 2 weeks) to limit the effect of demographic processes (immigration between sites etc.). Each point was separated more than the known home range of the species.

There is no possible way of marking the individual geckos. Originally PIT-tagging was tried but the skin of guenther geckos tears easily, therefore this method cannot be used. Due to the nature of the skin no other method of marking can be used either. The markings are also very subtle, therefore (as far as I am aware) the computer software available to recognise markings is not sensitive enough to work for these animals. The original geckos that were translocated in 2010 are known individually by scars, markings etc. therefore capture-recapture methods were originally used, however we believe (from observation) the population is now between 100-150 individuals and there are too many unknown individuals for the software to work any more. The only method we know of that could work is toe clipping, but we do not wish to use this method as it is not ideal for this long-term study.  

 It was due to this that we looked into using unmarked, but I have found little in the literature in relation to determining population size rather than just the relative abundance between sites. Do you know of anyone who is trying/done something similar that I can get in contact with? 

So (sorry for the basic questions!) from the abundance of each site I can work out the density per area. Obviously this varies across the island, so I wanted to also check that extrapolation of this will be possible to gain an accurate estimate of population size?

Thanks again for your help,

Sarah

[Richard Chandler]

Hi Sarah,

Yes, you can use the N-mixture model to estimate island-wide abundance. In the most basic case, the model estimates the parameter "lambda" which is the expected number of individuals per "site". If you want to estimate density instead of abundance, you can include log(area) as an offset. If the island is composed of multiple habitat types, the syntax might look something like this:

fm <- pcount(~1 ~habitat + log(area), umf)

You can used the fitted model to predict abundance at any habitat patch on the island, and then sum those up. 

Richard 

Sarah

--

[CHSalvador]

Hi Andy and Richard,

I have similar sampling design and question of Sarah. I´m also interested on estimation of density with pcount. Sorry to bring this topic again, but I searched the emails’ archives and the information became quite confuse for me with different comments about this topic. I found three different comments in this group and I´d like to check their difference to estimate density (individuals/unit area).

My umf is 90 sites x 30 occasions with number of individuals of wild boar counted by camera trap/day. In this case, 1 day = 1 occasion. The sites had fixed distance of 3 km for independence during 30 days. It was closed population design.

Let me resume the comments about density estimations I found:

1.       From Royle’s email above: “doing a fixed radius point count, then it might be reasonable to use the estimated abundance per point and divided it by the area of the fixed radius point count”. Translating to my situation, the density estimation would be “lambda/(28km²)”. The lambda is the estimation from predict pcount. Example: predict(pcount(~1~habitat,umf),type="state")
   
2.       From Richard’s email above: pcount(~1 ~habitat + log(area), umf)
3.       From Matt´s email: https://groups.google.com/d/msg/unmarked/x9QVRk6DTew/Lrf2wZRMIC8J    …with some modification to compare: pcount(~1 ~habitat+offset(log(area)),umf)

It would be very helpful to understand the difference, please.

Please, let me also check the content of “area” in “log(area)” command line. Is it a value (e.g., 28) or a covariate vector : siteCovs(umf)["area"]

Thanks,

Carlos

[Richard Chandler]

Hi Carlos,

I wouldn't recommend estimating density with N-mixture models when you are using "plotless" sampling methods such as camera traps. There has been a lot of discussion of this topic on this list, but you should really consult the primary literature on the topic rather than relying on the various opinions that get thrown around online... in my opinion.

Richard

For more options, visit https://groups.google.com/d/optout.

--

Richard Chandler

Assistant Professor

Warnell School of Forestry and Natural Resources

University of Georgia

706-542-5815

Lab website

[CHSalvador]

Richard,

many thanks.

Carlos