Mike, Actually I would argue that a sampling scheme like that for trapping webs (it could be called radial or polar systematic sampling) is not appropriate for plot characterization. The density of points is higher near the center, such that the center of the plot is sampled more intensively than the outer rings and so biased toward the characteristics of the center of the plot.
It is this exact property of trapping webs or 'polar systematic sampling' that imposes the detection function on the captures. Buckets or traps near the center have a higher prob of capturing a animal because of their higher density. And lower det prob in each succeeding ring out. While I like the polar systematic sampling for its elegance, it's not appropriate. Maybe there is a more elegant way of choosing random or systematic subplots within a circular plot but I don't know it. I would like to learn more about various vegetation sampling methods. I like Mr. Tesslers idea of basically reporting presence of habitat types because as he says the sampling methods require enough replicates. In my limited experience it can be difficult (read: many subplot replicates) to get a reasonably precise estimate of density or abundance of a plant on a plot. Recently I was involved in a study to document p/a or distribution of a rare-ish plant (a hostplant for an endangered butterfly). We wanted some estimate of density for each p/a plot as well. It was my idea to use 4-5 subplots and count all the plants of interest in them and extrapolate from the subplots to the larger plot. Unfortunately, the plant turned out to have a very clumped distribution. So subplot abundance would be, for example, in one case, 0, 0, 0, 0, 111, for 5 subplots. Run a confidence interval on the mean of that series and you can see why we stopped doing the subplots. We settled for a rough count/estimate of how many plants in the plot (ranging from very low, e.g. 10, to over 1,000). We don't put much confidence in this index, but it gives us an order of magnitude idea of how many plants were in plots. It is precise enough for our needs at this stage. As Mr. Tessler also touched on, the important thing is to decide what you can document/measure that is relevant to your study question. Are all of the things being measured thought to have some relevance to the owls or whatever the study question is? Proponents of information-theoretic analysis using model selection argue that in measuring tons of variables and indiscriminately testing them with a stepwise multiple regression you are likely to increase your chances of finding a spurious result. Of course, finding the balance between testing variables indiscriminately and leaving out potentially informative variables is what I struggle with. It helps to think in terms of exploratory and confirmatory analyses. Hope this heps. Given the fact the study is already designed its probably mostly moot. Tyler Grant ----- Original Message ----- From: "Michael Fuller" <[EMAIL PROTECTED]> To: "Tyler Grant" <[EMAIL PROTECTED]> Cc: <[email protected]> Sent: Monday, June 04, 2007 7:05 AM Subject: Re: Sampling vegetation for wildlife habitat purposes > Thanks Tyler, > You make several important points about distance sampling. And I agree > that Amy's study is not a distance sampling problem. But are you saying > that the use of a web layout for sampling points (spokes radiating from a > center) is inappropriate for taking veg samples? I don't see why. One can > use any layout regardless of whether it involves a distance sampling > problem. I was not suggesting Amy use distance sampling techniques, only > that she consider the web as an approach to systematic sampling. > > Mike > > On Jun 4, 2007, at 1:58 AM, Tyler Grant wrote: > >> I've used distance sampling in several situations, including trapping >> webs for flat-tailed horned lizards. Distance sampling isn't >> appropriate for the situation below. Distance sampling estimates >> density. Is that what is wanted? And distance sampling is best for >> objects of interest that have low detectability and/or are more or less >> sparsely distributed. It's a way to deal with detection probability >> issues. If det prob is not an issue, like detection of p/a of sedges in >> a plot is probably not an issue, then you don't need to use distance >> sampling. You could cram the situation below into a distance sampling >> framework but it would not be worth the effort. You'd probably find det >> prob equaling 100%. Amy's situation sounds like a simple choose your >> plots and count situation. To choose plots she could simpling choose >> random bearings and distances from the center of the plot if a few need >> to be chosen. Otherwise could divide the plots up somehow into 10cm >> square sub-quadrats, number them, and randomly choose from the numbered >> list. >> >> Distance sampling has come far since the 1983 paper listed below. The >> standard references are Buckland et al. 2001 Intro to Distance Sampling >> and Buckland et al. 2003(?) Advanced Distance Sampling. Program >> DISTANCE is available for analysis. Numerous papers are out there using >> different flavors of distance sampling. >> >> Tyler Grant >> Fish and Wildlife Biologist >> >> >> >> ----- Original Message ----- From: "Michael Fuller" >> <[EMAIL PROTECTED]> >> To: <[email protected]> >> Sent: Sunday, June 03, 2007 6:12 AM >> Subject: Re: Sampling vegetation for wildlife habitat purposes >> >> >>> Hi Amy, >>> Ahh, thanks for clarifying your approach. :-) As you are using >>> circular plots (and, it seems, a form of distance sampling), you >>> might consider adapting the "trapping web" design to your needs. >>> Trapping webs are often used for capture-recapture studies. The web >>> design uses spokes radiating out from the center of the circle. >>> Samples are then taken at specific intervals along the spokes. This >>> approach yields several subsets of samples that are equidistant from >>> the center (i.e. rings of samples, like pearls on a necklace). The >>> following paper describes the design as applied to capture-recapture >>> studies, along with statistical analysis. >>> >>> Anderson, D.R., K.P. Burnham, G.C. White, and D.L. Otis. 1983. >>> Density estimation of small-mammal populations using a trapping web >>> and distance sampling methods. Ecology 64:674-680. >>> >>> The above is an early paper on the topic. The authors have written >>> several related papers, including one that compares the web design to >>> grid designs (Parmenter et al. 2003. Ecological Monographs 73:1-26). >>> >>> All the best, >>> Mike >>> >>>> Date: Sat, 2 Jun 2007 11:57:37 -0700 >>>> From: Amy Williams <[EMAIL PROTECTED]> >>>> Subject: Re: Sampling vegetation for wildlife habitat purposes >>>> >>>> Thank you all for your many insightful responses. Perhaps I can >>>> clarify a >>>> little bit. The research is already designed to gather plot-level >>>> information on vegetation physiognomy (canopy height and density, >>>> DBH, tree >>>> condition, percent of plot that is meadow, forest, bare ground, >>>> downed wood, >>>> etc.), as well as soil attributes and hydrology. In addition, I did >>>> intend >>>> to use the 10-cm square sub-quadrats to represent single points to >>>> gather >>>> presence/absence of either species or type (sedge, forb, grass), >>>> and average >>>> height. Is there a systematic way to place a large number of these >>>> in a 15-m >>>> square circular plot, besides limiting them to orthogonal cross- >>>> transects? I >>>> will make sure to consult the texts recommended to me; maybe I will >>>> find my >>>> answers there. >>>> >>>> Again, thank you for the help! >> >> > >
