Watch videos with subtitles in your language, upload your videos, create your own subtitles! Click here to learn more on "how to Dotsub"

BITC / Biodiversity Diagnoses - Inventory Completeness 3

0 (0 Likes / 0 Dislikes)
Most of the time when we go to the field to sample, we do things by effort. OK? Which is to say I know if I stay away from home for more than about 4 weeks, I won't have a home when I want to go back. Which is to say my permission from my wife is a month. Or maybe you say I'm going to sample for 5 days at each site and then move. Or maybe you say I'm going to accumulate 500 trap-nights at each site or like I just said each camp lasted about a month. So notice that we are staying at a site or we are sampling for a set amount of effort. OK? And that's a very artificial thing to do. Let me illustrate this for you I'm gonna take you to sample in this forest. Notice that is a nice and beautiful trail. The trees are spaced out very beautifully. There is not a lot of underbrush. You can set your traps or your nets or you can do your observation, whatever you need, very easily This is a very pleasant place to spend time. And I go out and I work for a certain amount of time, I'm accumulating my species and when I get to my effort-based cut off That's the number of species that I've detected. Now, these are easy places to spend time. We could sample this area. Looks like where you could take your family and have a picnic. Or I could take you to this place. You guys have been to some of these places, right? There are places where you would take your family for a picnic, there are places where you do not want to go at all My unfavorite habitat is mangroves, right? I hate mangroves, I truly hate mangroves. Kate loves mangroves because she loves crocodiles. But if you had to spend the rest of your life in one habitat So, forget about comfort. Sampling efficiency, right? Where are you going to be more efficient? Remember, I work with birds. So, you trying to see a bird perched on the top of this tree versus trying to see a bird perched on the top of this tree. A little bit different. So, in the first example, we had our amount of effort, but in our second example, our inventory may be quite different just because it's harder. And so even know this site has many more species than this site, it's just easier to sample and detect even access the first site. And if we sample for a long time, our true number of species here is whatever that number is. But if we sample the same amount of time, notice that in the end, the second site is more diverse. But if we cut off based on effort, we get the opposite picture. This is a made up example, but it's very true. I've done this results-based sampling in local habitat matrices in Mexico and there are things, these are habitat type called "foreign forest". And you can't walk more than a couple of meters into thorn forests before you're just basically immobilized by these trees. Or any bamboo forest if you've worked in bamboo, is miserable. So, the idea of results-based sampling is that you continue your sampling until your results achieve a certain level of precision. Not until you fit some effort-based criteria. Not until you accumulate 500 net-nights or net-days, but rather until your inventory gets up to some level of precision. And so, you can develop what we in that paper we termed "stopping rules". So, remember we talked about S observed, the number of species known to be in the site And we talked about S expected, number of species estimated to be at the site And you take the ratio of that and you call that completeness. Which we use the abbreviation C. And so, we can say something like we want to observe at least 95% of the species that are projected or predicted to be at this site And then we want to maybe add another criterion like let's go five days without adding species to the inventory Or you could say, you know a day, or 5 days without adding more than 10 species, whatever, depends on your organism. But you can set up stopping rules that kind of fit the dimensions of the accumulation curves that fit your group. Now, I'm perfectly perfectly conscious of the fact that this maybe impossible You know, the plane is coming back for me on the first of the month, I can't leave before that. But this is an idea that really can improve the inventories. So, if you look back at these plots from that paper that we developed. Here is S observed, this is the truth by the way, but that's not the point. The point is let's take this estimator, it comes down So, initially, my estimates are very high. So here, C is going to be about 0.6, right? I've seen this many species and my estimator says I've got this many more to discover By the time I get here, maybe it's 90, 92% because I'm comparing this number to this number. By the time I get out here, I'm somewhere around 100%. Now, if I got 20 days then my stopping rule could be 100% and no new species added. That's very stringent. That might work for you know, wild palms in Benin, where we have 8 species. It might not work for birds in Benin and certainly, wouldn't work for beetles. But you can invent stopping rule that works for your group. I'm gonna move on to this. So, here is the idea, let's imagine we have different kinds of stopping rules. And let's imagine I'll just need a certain amount of time to detect each species. This is a simple world. So, for this site, that, in reality, has 120 species, maybe I need 12 hours to detect that. I shouldn't say hours because I'll get it wrong. I need the certain amount of hours to detect that 120 species and I need a much lower number, in fact one sixth of that time to get these 20 species. But if I go in and I say I'm gonna do a set amount of time per site. Maybe 7 hours each. Then I'm undersampling this site and oversampling this site. Which is to say the accumulation for this site is still going up and the accumulation curve for this site leveled off a long time ago. and these two sites were done OK. If I sample the longest time which is to say I know I need at least 12 hours to sample any of the sites at this region. Then I wasted time at three of my sites. But if I have the possibility of putting 12 hours to this site and two hours to this site and 7 and 7, then I've optimised the use of my time. [Lindsay]: I have a question ... [Town]: Yes ma'am? hold on, hold on [Lindsay]: When you are #### the times for these sites, how long you are going to spend at each site? Do you take into account the ecology of the site? [Town]: No, that's what I call this results-based sampling [Town]: I don't choose amounts of time at all. What I do is as my inventory proceeds, I am guessing at how done I am. So, I'm essentially watching that accumulation curve come up and maybe every day or every hour, I'm doing a completeness analysis And when that completeness analysis gets up to my criterion and any other criteria that I might have imposed, I stop, OK? So, I don't have to guess in advance about where it would be easier and where it would be hard [Arturo]: That pretty much works very well if you are progressing on specimen by specimen basis. In bird watching which is correct. So, you get one new specimen which may or may not be recorded a species [Arturo]: However, in some cases, you would have to go work with actual samples in which you have to count what is in one sample. And then decide when to take the next sample or not. [Town]: That's true. [Arturo]: In that case, the percentage that the accumulation curve that we affected by the order in which the new sample is out there. [Town]: Right. [Arturo]: So you have to, somehow work, however, this looks quite efficient for the specimen by specimen accumulation. [Town]: And you could even use this for post processing, which is to say, you know, you go out and take soil samples and then you're gonna invest all that time with Berlese funnels, and things like that. Or sitting down at microscope to identify the taxa that are present Maybe you get 15 samples from each site, but you only analyze until you've met result-based criterion [Arturo]: That's what we do [Arturo]: We can get the average [Town]: So, definitely definitely, we haven't mentioned this to everybody, but when we do these calculations , for example, using Chao's indicator, the order does matter And so we do a bootstrap, a sub-sample and we will randomize the order. So, the accumulation curve becomes a kind of #### accumulation curves that get you out of what would essentially be temporal autocorrelation. So, in bird work that's very common. For example, in March, if I go to the Cameroon mountains I don't know any of the bird songs there. And so, I may spend three days walking around the forest and not detect the species. But then, the species pops up in front of me and sings. And I then learn that song and now I'm thinking OK, if I hear that song, that is that species. And so I get my occurrence matrix looks like a string of zeros before I knew the song and the string of ones after I knew the song. Or even worse sometimes, I realize OK there is a nest of that species on that branch. And so, all the days before I discover that nest, I never see the species. And all the days after I discovered that nest, I just look up there, there is the bird. Done. So you get temporal autocorrelations in your data and so a randomization procedure is very important to breaking up that autocorrelation.

Video Details

Duration: 13 minutes and 50 seconds
Language: English
License: Dotsub - Standard License
Genre: None
Views: 0
Posted by: townpeterson on Jul 26, 2016

This talk was presented in the course on National Biodiversity Diagnoses, an advanced course focused on developing summaries of state of knowledge of particular taxa for countries and regions. The workshop was held in Entebbe, Uganda, during 12-17 January 2015. Workshop organized by the Biodiversity Informatics Training Curriculum, with funding from the JRS Biodiversity Foundation.

Caption and Translate

    Sign In/Register for Dotsub above to caption this video.