There is a discussion group on SCR: https://groups.google.com/forum/#!forum/secrgroup which is probably the better place to ask this question.

From experience with large mammals (elephant, bear, cheetah, wolf) across continents, I doubt that "home" range is a species characteristic. The actual species range differ over geographic space depends on the environment including dynamics and developing human wildlife conflicts.  

My answer can only be more targeted if you tell us what species over what extent you intend to study.

Hi Thomas,

I think the answer to your question depends on three main factors: species, scale & habitat. Home range could be a first appropiate approach for some species (i.e, some specialist species with small behavioural and territorial variations), the scale you are planning to answer you question (i.e. local, landscape, global) will surely affect the outcome, and; the habitat you will be working at (i.e. some generalist species such as the white-tailed deer may have very different home-ranges depending on the habitat they live in). So, I guess you need to be more specific about your research question in order to get better feedback. Good luck!

Very interesting research your are planning. I am assuming the fecal samples are for DNA identification of individuals, right? What is the main question/objetives of your research?

Whoops, thought I'd answered.  Tom, I think the reference you should look at (beyond the Royle et al. 2014 book chapter dealing with area searches) is by Thompson et al. 2012 in JWM (something like a framework for SCR with unstructured sampling, involves scat-detecting dogs).  You also might look at Murray Efford's 'secr' vignettes for area searches or transect searches.  These will be the most trustworthy sources. 

In short, I think the grid cell must be smaller than a home range.  in a traditional area search (e.g., the Royle and Young 2008 paper), the (single) cell is being exhaustively systematically searched, so that the sampling allows for (imperfect) animal detection essentially anywhere.  In your situation, you are sampling along a one dimensional-transect or route, and animals can only be detected along that route.  For a number of reasons, transect-based samples or linear routes cause some difficulty related to the continuous space being sampled. One way of dealing with this is to treat the center of a grid cell as something like a static camera or hair-snag; if this is the route you follow, then you want grid cells smaller than a home range so that you actually generate spatial recaptures!  My memory is that the guidelines were something like 1-2 x sigma for cell dimensions, but check the references. 

And just to follow up quickly: although approximating the survey effort into a discrete grid is likely to cause some loss of efficiency when estimating detection parameters, if telemetry information is available concurrently, you could integrate the two with no problem.

Hi Tom,

If you are concerned about the recaptures (uneeded, you said), I assume your goal is the identification of ALL the individuals in your sampling area. If the average home-range in your study area has been established as 7 square km, I would say that your grids should be at least the size of one home-range area. Otherwise, by simple logic, you might miss some individuals, assuming that pandas are highly territorial and home-ranges do not overlap (I am just assuming this, I do not really know). Scale is always a huge problem in large carnivore research, given the size of their home ranges.

One thing that concerns me, is that you say " I was planning on rather opportunistically hiking in good habitat, along animal paths, etc". As I am sure you know, many carnivores use defecation as a marking behaviour. Some use latrines and then feces might be grouped and even mixed from different individuals. Do not know if there is information available on defecation behaviour and defecation rates of pandas, but, if there is not, your should be aware of these, because it might heavily bias your research. On the other hand, sampling only "good habitat along animal paths" automatically brings a bias to your sampling design, as "poor habitat" might be as important for unaccounted reasons.

Good luck!

Right, the overall effort should be larger than a home range, but the sub-sampling grain should be smaller.

Well, I did some reading about panda ecology. Here some highlights:

1. Home-ranges between 1 - 60 Km2 (averages 5 - 15 km2), but small core-areas of activity (usually +/- 30 Ha)

2. Vertical migrations well documented (bamboo phenology related)

3. Home-ranges overlap little within sexes, but males usually overlap several female home-ranges.

4. Some information about scent-marking (pandas use urogenital and anal marking), but not much about defecation.

___________________

Given this info and the one you have provided about your research, Tom,  I am pretty sure you can use home-range as a starting point for your sampling design.

I found these two interesting papers:

Khon & Wayne. 1997. Facts from feces. TREE (old but nevertheless interesting review on the topic)

Wei He et al. 2008. Genetic diversities of pandas... (with feces). Cons. Gen

Cheers!

Hi Thomas. For use a SECR model you will need: 1) enough capture-recapture data (maybe at least 5-10 recaptures). If you have only few data, the bias could be very high. 2) Some of the recaptures must be spatial (in different sites/trap/grid). For a unstructured sampling could be useful to use the total length of sampling line in the grid as covariate of detection.

You have another google group about spatial capture-recapture.

https://groups.google.com/forum/#!forum/spatialcapturerecapture

Sure you know this paper:

Hello @Jose, fancy meeting you here. I just joined this website a few days ago, almost under duress.

In comment to Jose's post, I usually associate bias with unrepresentative sampling whereas low number of captures more often than not inflates the variance of the estimates.

What I found interesting in my simulations is that if you're able to encompass the whole population, which perhaps might be doable in a reserve of a rare animal (?), you wouldn't benefit from a spatially explicit model at all.