Ivory-billed Woodpecker (formerly updates) (3 Viewers)

emupilot said:

So the answer is, there isn't even one example of Pileateds sustaining a rapid wingbeat (sample size of zero). Your positive identification is based on video artifacts and ignores a wingbeat never documented in Pileateds? Talk about farcical...

Click to expand...

Docmartin's paper:
The woodpecker in the Luneau video maintains a steady rapid wingbeat rate of 8.6 beats s-1 for at least 8 wingbeats [1], a figure that was confirmed by independent analysis during preparation of this paper. The Pileated woodpeckers in DN's video do not do this – after initial rapid flapping immediately after take-off, they settle into a more relaxed level flight. As shown in Tables 1 and 2, although the mean wingbeat frequencies of the Pileated Woodpeckers in DN's video are slower than the 8.6 s-1 recorded for the bird in the Luneau video [1,3,5] the first four wingbeats, the initial escape response, are faster than those claimed for Pileated Woodpeckers in the literature [1,3,5]. For the four escape flights, the mean frequency values for the first four wingbeats are 7.1, 6.7, 8.6, and 8.0 s-1, respectively. The 8.6 beats s-1 of the bird identified in the Luneau video, while consistent with the limited data (n = 1; see Discussion) for Ivory-billed Woodpecker, is equally consistent with Pileated Woodpecker in its initial escape flight. The bird in the Luneau video maintains a frequency of 8.6 s-1 for the next four wingbeats too, whereas the Pileated Woodpeckers recorded here all slowed their flight as they prepared to land in nearby trees. There are no data to suggest whether Pileated Woodpeckers can maintain a wingbeat frequency approaching 8.6 s-1 for eight or more wingbeats, like the bird in the Luneau video. It remains possible that the flight pattern of the bird in the Luneau video is unusual for Pileated Woodpecker, but a frequency of 8.6 s-1 is consistent with a Pileated Woodpecker gaining initial speed and height in escape flight, and by itself cannot be taken as strong evidence that the Luneau video bird was an Ivory-billed Woodpecker.

The bird is consistent with PIWO too. It's very simple, really. And the 'sample' of one for IBWO could be of a bird flapping at a nest or anything! It does sound rather odd - and the spectrogram shows an abrupt stop after 8 flaps and so does the audio... where's the trailing off?

Looks shaky, dunnit?

Tim

Ooh - where's that paper?

Tim Allwood said:

Docmartin's paper:
The woodpecker in the Luneau video maintains a steady rapid wingbeat rate of 8.6 beats s-1 for at least 8 wingbeats [1], a figure that was confirmed by independent analysis during preparation of this paper. The Pileated woodpeckers in DN's video do not do this – after initial rapid flapping immediately after take-off, they settle into a more relaxed level flight. As shown in Tables 1 and 2, although the mean wingbeat frequencies of the Pileated Woodpeckers in DN's video are slower than the 8.6 s-1 recorded for the bird in the Luneau video [1,3,5] the first four wingbeats, the initial escape response, are faster than those claimed for Pileated Woodpeckers in the literature [1,3,5]. For the four escape flights, the mean frequency values for the first four wingbeats are 7.1, 6.7, 8.6, and 8.0 s-1, respectively. The 8.6 beats s-1 of the bird identified in the Luneau video, while consistent with the limited data (n = 1; see Discussion) for Ivory-billed Woodpecker, is equally consistent with Pileated Woodpecker in its initial escape flight. The bird in the Luneau video maintains a frequency of 8.6 s-1 for the next four wingbeats too, whereas the Pileated Woodpeckers recorded here all slowed their flight as they prepared to land in nearby trees. There are no data to suggest whether Pileated Woodpeckers can maintain a wingbeat frequency approaching 8.6 s-1 for eight or more wingbeats, like the bird in the Luneau video. It remains possible that the flight pattern of the bird in the Luneau video is unusual for Pileated Woodpecker, but a frequency of 8.6 s-1 is consistent with a Pileated Woodpecker gaining initial speed and height in escape flight, and by itself cannot be taken as strong evidence that the Luneau video bird was an Ivory-billed Woodpecker.

The bird is consistent with PIWO too. It's very simple, really. And the 'sample' of one for IBWO could be of a bird flapping at a nest or anything! It does sound rather odd - and the spectrogram shows an abrupt stop after 8 flaps and so does the audio... where's the trailing off?

Looks shaky, dunnit?

Tim

Click to expand...

Collinson's paper looks shaky indeed. The data it cites indicates that Pileateds do not maintain 8.6 beats per second for more than 4 beats, and yet we have 11 beats worth in the Luneau video. Collinson himself points out the obvious:

There are no data to suggest whether Pileated Woodpeckers can maintain a wingbeat frequency approaching 8.6 s-1 for eight or more wingbeats, like the bird in the Luneau video.

Click to expand...

His own data indicates that Pileateds either can not or do not maintain that frequency. If the hypothesis is that Pileateds can maintain that wingbeat frequency, the only conclusion that can be drawn is that they do not. It's ridiculous to say the hypothesis has been shown true in spite of the data saying otherwise.

colonelboris said:

Ooh - where's that paper?

Click to expand...

yertiz

http://www.biomedcentral.com/1741-7007/5/8

Tim Allwood said:

And the 'sample' of one for IBWO could be of a bird flapping at a nest or anything! It does sound rather odd - and the spectrogram shows an abrupt stop after 8 flaps and so does the audio... where's the trailing off?

Click to expand...

The bird is described as flying off. If you listen to it, you find the volume trails off as the bird flies away.

Tim Allwood said:

yertiz

http://www.biomedcentral.com/1741-7007/5/8

Click to expand...

Awesome - thanks Mr Allwood!
Makes the paper I'm working on seem pretty dull now, even after today's excitement...

edit: I'll save it for work tomorrow as it looks suitably academic to pass a quick glance at the computer from the boss...

emupilot said:

Collinson's paper looks shaky indeed. The data it cites indicates that Pileateds do not maintain 8.6 beats per second for more than 4 beats, and yet we have 11 beats worth in the Luneau video.

Click to expand...

If Collinson's data is shaky, surely basing the beat rate of Ivorybills on one sample is just as shaky? The point is that Pileateds can reach 8.6 bps-a documented fact, thus the video cannot be definitely IBWO on that evidence.
11 beats makes less than 2 seconds of video, hardly conclusive really.

Ilya Maclean said:

The Bayesian approach isn't correct because of more than one potential confusion species - i.e. there's more than two possible outcomes. .....

Click to expand...

Even if there are more than one confusion species the Bayesian approach is the proper one. The similarity of the problem with medical testing is helpful to understand this. There you have hundreds of diseases (confusion species), but the point is you are testing only for one particular disease (IBWO) and there are only two possible answers, IBWO or not IBWO, regardless of the number of confusion species. There is also multidimensional Bayesian analysis.

Ilya Maclean said:

Lets assume 50% match based on features, but that there are no Piliated woodies. Bayesian approach would yield probability of bird being an IBWO as 100%, my approach as 50% (i.e. 50% chance it's something else entirely). .....

Click to expand...

This is a logically inconsistent assumption leading to wrong quantitative conclusions. If confusion is not possible, one cannot have a 50% match based on features.

Fundementally though, the fact that one is much less likely to encounter a IBWO over a potential confusion species does mean that one needs to be exceptionally cautious when dealing with IDs that are not 100% .....

Click to expand...

This is indeed true! It worthwhile to note that these Bayesian probability calculations would apply to the identification of any rare or vagrant bird. It is surprising to learn that this is less known in the birding community, in contrast to medical testing, where it is common knowledge that screening for rare diseases would yield disastrous results even if the sensitivity of the test is very high.
For example, one is 99.9% sure of the identification of a bird, but the abundance of this bird is one in 100 000 (rare vagrant). The probability calculations in this case reveal that the probability that the bird is indeed the rare one would be only 0.99%. To increase this value to 90% one should be 99.9999% certain about the identification.

ortolan10 said:

Even if there are more than one confusion species the Bayesian approach is the proper one. The similarity of the problem with medical testing is helpful to understand this. There you have hundreds of diseases (confusion species), but the point is you are testing only for one particular disease (IBWO) and there are only two possible answers, IBWO or not IBWO, regardless of the number of confusion species. There is also multidimensional Bayesian analysis.



This is a logically inconsistent assumption leading to wrong quantitative conclusions. If confusion is not possible, one cannot have a 50% match based on features.



This is indeed true! It worthwhile to note that these Bayesian probability calculations would apply to the identification of any rare or vagrant bird. It is surprising to learn that this is less known in the birding community, in contrast to medical testing, where it is common knowledge that screening for rare diseases would yield disastrous results even if the sensitivity of the test is very high.
For example, one is 99.9% sure of the identification of a bird, but the abundance of this bird is one in 100 000 (rare vagrant). The probability calculations in this case reveal that the probability that the bird is indeed the rare one would be only 0.99%. To increase this value to 90% one should be 99.9999% certain about the identification.

Click to expand...

What happens when you add the ability of the observer in-for example if you are 80% sure the observer is able to ID the bird to 99.9% confidence?

any formula that produces a result of 10.7% that the large unidentified woodpecker is the Luneau video is an IBWO needs, er 'reworking'... or are Steve Howell, David Sibley and the rest of the mainstream birding community etc 89.3% idiots?

ortolan10 said:

Even if there are more than one confusion species the Bayesian approach is the proper one... The similarity of the problem with medical testing is helpful to understand this. There you have hundreds of diseases (confusion species), but the point is you are testing only for one particular disease (IBWO) and there are only two possible answers, IBWO or not IBWO, regardless of the number of confusion species. There is also multidimensional Bayesian analysis.


This is a logically inconsistent assumption leading to wrong quantitative conclusions. If confusion is not possible, one cannot have a 50% match based on features.

This is indeed true! It worthwhile to note that these Bayesian probability calculations would apply to the identification of any rare or vagrant bird. It is surprising to learn that this is less known in the birding community, in contrast to medical testing, where it is common knowledge that screening for rare diseases would yield disastrous results even if the sensitivity of the test is very high.

For example, one is 99.9% sure of the identification of a bird, but the abundance of this bird is one in 100 000 (rare vagrant). The probability calculations in this case reveal that the probability that the bird is indeed the rare one would be only 0.99%. To increase this value to 90% one should be 99.9999% certain about the identification.

Click to expand...

I agree my original calculations were erroneous, but the possibility of multiple confusion species is important, because not all are equally likely. To see this consider this example. Imagine a population comprising only of IBWOs, PIWOs and Anhingas. Confusion with both PIWO and Anhinga is possible, so one has to consider the ratio of IBWOs to Anhingas and PIWOs to calculate the probability. Consider another population comprising of IBWOs and PIWOs (in the same abundance as previously), but instead of Anhingas there are Ruby-throated Hummingbirds. Confusion between IBWO and Ruby-throated Hummingbird is so unlikely that one can effectively dismiss the possibility of doing so. One therefore only needs to consider the ratio of IBWOs to PIWOs.

However, lets use the Bayesian approach and assume that Pileated is the only confusion species. Population of PIWOs: 930,000. Population of IBWOs (probably 0, but lets be generous and say 50). Chances of it actually being an IBWO if you’re 99% certain: 0.53%. To make it more likely that the bird is an IBWO rather than a PIWO one would have to be 0.9999% certain about the ID. So far none of the evidence I've seen comes anywhere close. I won't get started on the number of tree-branches, deer, nuthatches etc that could produce interesting calls

Imaginos said:

What happens when you add the ability of the observer in-for example if you are 80% sure the observer is able to ID the bird to 99.9% confidence?

Click to expand...

The answer to this question cannot be provided by stochastics. Your personal assessment of the capabilities of the observer does not have a bearing on the probability of correct identification or misidentification by the observer. Strictly speaking, these probabilities should be determined experimentally, rather than just saying ...I am 99.9% sure....The question is indeed how one can obtain these data. This is practically impossible for real observations, since every observation would be different from each other and, the true identity of the bird is a priori not known. Yet, with videos and images there might be a way of quantifying probabilities. Let us suppose the identity of the bird is known to the supervisor of the experiment. Videos and images might be taken with varying quality and birders then will be asked to confirm the identity of the bird (They should not be asked to identify). The confusion species will be likewise documented and birders will be asked to confirm (falsely) the identity of the bird. The outcome of such experiment might be interesting. A single birder can be tested with a larger selection of images as well as many birders can be tested with a limited number of images.

Ilya Maclean said:

However, lets use the Bayesian approach and assume that Pileated is the only confusion species. Population of PIWOs: 930,000. Population of IBWOs (probably 0, but lets be generous and say 50). Chances of it actually being an IBWO if you’re 99% certain: 0.53%. To make it more likely that the bird is an IBWO rather than a PIWO one would have to be 0.9999% certain about the ID. So far none of the evidence I've seen comes anywhere close. I won't get started on the number of tree-branches, deer, nuthatches etc that could produce interesting calls

Click to expand...

I think you need to consider the density because not all USA PIWO live in the IBWO habitat. Consider the Big Woods. From earlier posts I learned that the area is about 859 sqm = 2225 km². Density of PIWO 1 pair/200 ha. This is one bird per km². Let us assume 3 birds/km² including non-breeding birds, youngs. On total we would have 6675 PIWO. Let us further assume 2 IBWO in the whole Big Woods. Assuming 99% certainty I arrive at 2.9%. With 99.9 certainty of observation the probability of having indeed seen an IBWO is 23%. The likeliness that five such observations were all PIWO but none IBWO is only 14%.

ortolan10 said:

I think you need to consider the density because not all USA PIWO live in the IBWO habitat.

Click to expand...

True, but we actually don't know what IBWO habitat is.

ortolan10 said:

Consider the Big Woods. From earlier posts I learned that the area is about 859 sqm = 2225 km². Density of PIWO 1 pair/200 ha. This is one bird per km². Let us assume 3 birds/km² including non-breeding birds, youngs. On total we would have 6675 PIWO. Let us further assume 2 IBWO in the whole Big Woods. Assuming 99% certainty I arrive at 2.9%. With 99.9 certainty of observation the probability of having indeed seen an IBWO is 23%.

Click to expand...

Unfortunately evidence is a lot less convincing than 99%, more like 70-80%.

ortolan10 said:

The likeliness that five such observations were all PIWO but none IBWO is only 14%.

Click to expand...

I don't think the sightings are all independent of another in the sense that the first spurred the other observers into wishful thinking. This is actually my major reservation about the whole debacle.

those folks over at Nelson's blog aren't half as polite as we are...:

(saves the rest of us time stating the obvious over and over again though...)


Over at BirdForum, emupilot is trying to dredge up the old chestnut that Pileateds have never been documented flapping at 8.6 beats/sec. Putting aside the absolute idiocy of the belief that every flight behavior of any bird has been videotaped, particular every possible escape flight of a startled bird, and putting aside that flap rate for ID purposes was tried and tossed as useless sometime back in the 1980s by the top field birders of the day, I decided to do a little more looking.

Dave Nolin provided a video that showed that PIWO can flap at a rate of 8 beats/sec, but it did not continue for a complete second. Still the rate was right.

Nolin has another video that can be downloaded here. Using frame advance, it appears to me that this bird is flapping at 8 beats per second (looks like 30 frames/sec). Feel free to check my counts.

Next, I looked at the Luneau video. What is being used to support the 8.6 beats/sec "sustained" flight? 11 flaps. 11 lousy flaps! And after that, the flaps appear to SLOW DOWN, just like Nolin's 2nd Pileated video.

So we have a video of a PIWO that is not being startled flapping less than 10% slower than the Luneau bird (a.k.a. Luneau's Pileated or Cornell's Folly). As for the "sustained" argument, it's 8 wingbeats vs. 11. That would be what an intelligent and semi-knowledgeable birder would call "insignificant".

I probably shouldn't have bothered, but I just felt the need to combat willful ignorance.

ortolan10 said:

I think you need to consider the density because not all USA PIWO live in the IBWO habitat. Consider the Big Woods. From earlier posts I learned that the area is about 859 sqm = 2225 km². Density of PIWO 1 pair/200 ha. This is one bird per km². Let us assume 3 birds/km² including non-breeding birds, youngs. On total we would have 6675 PIWO. Let us further assume 2 IBWO in the whole Big Woods. Assuming 99% certainty I arrive at 2.9%. With 99.9 certainty of observation the probability of having indeed seen an IBWO is 23%. The likeliness that five such observations were all PIWO but none IBWO is only 14%.

Click to expand...

Top maths and very interesting stuff. Not sure how relevant to IBWO it is since the output figure rests on the assumed population, particularly whether it is >0, which is of course the issue at hand. Petitio principii, IMHO

I'm interested in how this applies to claimed sightings of other rare species confusable with common ones. I can see one flaw, which is the tacit assumption that all individual birds are likely to be sighted in direct proportion to their abundance, regardless of species. The failings of this are evident when one considers a reed bed holding, say 4 Bitterns, 2 Marsh Harriers and 500 Reed Warblers. The probability of seeing a Bittern is not double that of seeing a Marsh Harrier, nor is a Reed Warbler sighting 125 times more likely. In short, likelihood of sighting a bird is dictated by the habits of the bird and the observer. Of course with genuine confusion species this problem is not as great as in the hypothetical example but it can be pronounced as with, say, Whitethroat vs Lesser Whitethroat in the UK where the latter species is much more skulking. Given the conjecture about the elusiveness or otherwise of the IBWO, this must have a bearing in this case for advanced mathematicians who can divide by zero.

Graham

emupilot said:

Collinson's paper looks shaky indeed.

Click to expand...

Emu, I think you're looking for things in my paper that weren't there. Remember as of 2005 we had a video that looked totally unlike our preconceptions of what a PIWO would look like on video, but quite like how we might expect an IBWO to look. In 2007 my paper pointed out that in both wingbeat and plumage the Luneau bird was, unexpectedly, pretty damn close to a PIWO, and the remaining discrepancies are not sufficient to upset a basically compentent birder who knows about how birds fly. Together with the Sibley et al papers which showed why the Luneau bird could almost certainly not be an IBWO (plumage) the data were pretty convincing that the Luneau bird was realistically likely to have been a PIWO, and at least that any claim that the bird was definitely Id'd as an IBWO were verging on delusional.

There are a number of people who are pinning their hopes for the Luneau bird as an IBWO on the rate of wingbeats 4-11 but they are clutching at straws that cannot hold. jus gunnae no, as they would say here.

Never having set foot (or canoe) in the territory in question or seen a large woodpecker in the US, I've nothing to say on IBWO, except that I'm enjoying the soap opera.

But I'll throw my oar in on the statistics part.

Ortolan is probably right that the Bayes approach is most appropriate when determining if an IBWO sighting is genuine. But to do that you need the following:

(A) The prevalence of IBWO as compared to other large woopecker-like birds.
Bittern's point about visibility could be factored into this figure.

(B) The proportion of sightings by the observer of large woodpecker-like birds incorrectly identified in the past as NOT IBWO.

(C) The proportion of sightings by the observer of large woodpecker-like birds incorrectly identified in the past as IBWO.

True, Bayes considers the sensitivity of a test rather than the competence of an observer. This complicates things, because a test is deemed to be the same for all the data gathered, whereas observers' competence may vary.

In any case, we don't know (A) because that's what we're trying to find out!

We don't know (B) because we don't have stats on unremarkable sightings of PIWO or other birds.

And we definitely don't know (C)!

So I'd say the stats approach may be a non-runner!

Docmartin said:

Emu, I think you're looking for things in my paper that weren't there. Remember as of 2005 we had a video that looked totally unlike our preconceptions of what a PIWO would look like on video, but quite like how we might expect an IBWO to look. In 2007 my paper pointed out that in both wingbeat and plumage the Luneau bird was, unexpectedly, pretty damn close to a PIWO, and the remaining discrepancies are not sufficient to upset a basically compentent birder who knows about how birds fly. Together with the Sibley et al papers which showed why the Luneau bird could almost certainly not be an IBWO (plumage) the data were pretty convincing that the Luneau bird was realistically likely to have been a PIWO, and at least that any claim that the bird was definitely Id'd as an IBWO were verging on delusional.

Click to expand...

As for "plumage" analysis of the Luneau video, Bill Pulliam pointed out quite clearly that if you and Sibley want to claim the bird in the video has a black trailing edge to the underwing, you need to also claim that the bird has a black leading edge, since that is equally apparent. If you don't want to claim the black leading edge (uncharacteristic of Pileated), you can't claim the black trailing edge either. I don't think there are nearly enough pixels to make an identification claim either way, as (to state the obvious) the video is poor quality and the artifacts are so significant relative to the image as to make it impossible to divine very much in the way of plumage.

As for the wingbeat, your data contradicted your hypothesis! You claim that a bird maintaining 8.6 beats per second could be a Pileated, but in none of your test cases does this actually occur. I suspect you are not aware of any case where a Pileated has actually been recorded maintaining the Luneau flap rate, otherwise it would have been in your paper. You can not claim the Luneau bird to be "pretty damn close" to a Pileated wingbeat without a statistical analysis, which I doubt would be friendly to your hypothesis.

Docmartin said:

There are a number of people who are pinning their hopes for the Luneau bird as an IBWO on the rate of wingbeats 4-11 but they are clutching at straws that cannot hold. jus gunnae no, as they would say here.

Click to expand...

As I've stated before, I don't think many are or should be "pinning their hopes" on the Luneau video, as the observations of many experienced observers is and has always been far better evidence for the Ivory-bill's existence in my opinion. I am surprised that you think people are clutching at straws with wingbeats 5-11, but your data only served to bolster Cornell's case! You fall short of others by not definitively identifying the bird as Pileated, only calling it "probable", but I don't think there is a basis to make any claim on the identity of the bird. Your own wingbeat analysis demonstrates why the Luneau bird does not appear to be a Pileated, so that calls into question any conclusion about the bird actually being a Pileated. The video analysis fails to account for artifacts.

You put alot of time and thought into the paper, and I welcome academic analysis. I appreciate that you carefully pointed out the Luneau video has no bearing on the sightings in Arkansas or Florida, which your fellow skeptics easily forget. I strongly disagree with your conclusions, however, as I don't think anything conclusive can be said of the Luneau video.

The Luneau bird is a Pileated flaping at 8.6 bps

I thought these IBWO stringers were gonna call it a day...

the video is not an IBWO
the audio is hopeless
Hill couldn't document the bird in a couple of square miles, despite finding one in an hour of arriving in the Choctawhatchee and predicting nine pairs...Collins reckons Rufous-headed Peckers are as elusive and i've seen those every day for three weeks nearly... hmmm

and a handful of fleeting glimpses that are mis-identifications

a couple of longer sightings are the real thing, mis-id or worse. So where are they?

Cornell are hoping it will fade away. And must be thanking the lord for Hill

Still, it's good fun for us. If not the birds and habitats the world over starved of attention and cash