Conflict, Confusion, and RoboBirders

For the remainder of their time in Tennessee, the Cornell crew only made it out in the field for one more full day, April 5th. On the 6th they did put in a half day, but that was mostly dedicated to retrieving Reconyx cameras. My info about these days is second hand, passed to me via Scott, as I was 160 miles away at the time. Apparently on the 5th Leighton heard a double knock while he was in the woods east of Rhodes Lake, with the sound coming from the south. According to Scott, Leighton said that if he had been in the tropics he would have identified what he heard as a Pale-billed Woodpecker, a sound he is familiar with. In Tennessee, on the other hand, he didn't know what to make of it. Do any of us? All told for their week in Tennessee, the augmented Cornell crew had three full field days plus one short take-down day. The crew ranged from 4 to 7 people each day, in 4 to 6 boats. Two double knocks were reported, no sightings, no kentings. And so it goes, on and on and on... It seemed that nothing would ever change.

Then the ARU results came back, which changed things, but not necessarily for the better.

Earlier in March, Scott and Melinda had deployed two automatic sound recorders (ARUs), borrowed from Cornell, for about a week at several places within the "hot zone." The data had been sent to Ithaca for processing, which is time-consuming. The analyses were finally complete right as everyone was returning from the latest excursion to Moss Island. Evidently one of the units malfunctioned, so there was only one machine's worth of data. I was not actually privy to any of what transpired about this first hand, but I gather over the course of a couple of conference calls, some things were said, misstated, and/or misunderstood that lead to some heated discussions.

The starting point is this: the only ARU that recorded data yielded a large collection of mechanical banging sounds but nothing that passed the screens as a candidate double knock under the criteria developed by Cornell. None of the human double knock detections were in places or at times that put them near the operating ARU, so no cross checking was possible. This was disappointing, but given a single deployment by only one machine this isn't really surprising. But then it all got complicated. Somehow things were said or construed, and then related to me on down through the telephone game, that left me with the impression that the Cornellians had concluded that we couldn't tell a double knock from a hole in the ground, and all we'd been hearing was these random mechanical noises that we were imagining were double knocks. This was, to say the least, kind of insulting. When I asked about the double knock that had been heard by one of their own crew, the (again second or third hand) answer I got was that Leighton had said he didn't know what he had heard. Apparently it was never quite as bad as it sounded to me, and it was mostly straightened out in a few days, But it reminded me of something that my wife's boss at a Federal environmental research lab had told her about science in Academia and government. Fair warning: I am going to use a time-honored four-letter Anglo Saxon word, unaltered and undisguised. The story really doesn't work as well if you sanitize it.

If you are a scientist at any sort of institution, you have three crocks of shit in front of you. One of them is your own shit, one is your friends' shit, and one is everyone else's shit. No matter what kind of scientist you are, your official position has to be that your own shit is gold, your friends' shit is silver, and everyone else's shit is shit. But if you are in Academia, you have to do more than just say this. You have to actually *believe* it.

Before anyone thinks I am accusing Cornell of being unusually and inordinately difficult, proprietary, or dismissive, think about your own dealings with groups, teams, and institutions under situations of pressure and frustration. These sorts of conflicts are extremely common, downright normal. It's just the way people are. The Man Behind the Curtain at the FWS (yes, there is one; let's call him "Jacob") and I talked over this situation a fair bit. He reminded me that there was a lot of frustration all across the region, with every crew feeling stuck in the rut of finding "something but not enough." Still, even though there were disagreements, on the whole the projects were doing a pretty good job of cooperating and sharing information internally, even if this was still being kept mostly out of the public eye. He also reassured me that as far as he could tell, Scott and Bob had successfully clarified the situation and the Tennessee position. To this day, though, there have been no further field collaborations between the Arkansas and Tennessee projects since all this came down.

Putting all the institutional drama aside, let's look at some of our own shit from Moss Island. I've heard 12 of the ARU samples. Most of them are clearly what I now call the "river noises," the off-site banging sounds that seem to mostly be associated with barge traffic on the Mississippi. Some of this is also from truck traffic on Highway 88 and Great River Road. A typical example is this (click the sonogram for a larger image, click the link below it to hear the audio):

Listen to the audio clip

Three things to note here:

1. The bangs themselves: They are generally long-echoing. They are very loud for the most part, and several kilometers away, so there's a lot of reverberation going on en route to your ears or your microphone. On the sonogram this shows as a long rightward smear of the signature. Also note how they are fairly "tall," covering a broad range of frequencies.

2. The pattern of the bangs: They tend not to occur in isolation. One loud BANGang is going to be preceded and followed by a bunch of other bangs in other patterns.

3. The other sounds: Associated with the bangs are various whining and churning sounds, which show in the sonogram as horizontal lines that fade in and out. These are motors and such.

This specific example is some of the "river noises," but similar things apply to human-made noises in general, be they gunshots, construction sounds, vehicles, etc. They tend to be associated with other human noises and recur in patterns that make their origins clear. These sounds are very familiar to us; they are not what we have been calling double knocks. Of the 12 clips I was provided, 9 represent these "river noises." All were recorded between 9:09 and 9:26 a.m. on March 6, 2008; in other words they all represent one single 17-minute long event. I've heard many of these events. They all come from the same direction (west), are nearly always accompanied by audible engine sounds, and are very easy to identify and eliminate as being of no interest at all.

This next clip is a "bridge whump," the sound made by a heavy vehicle passing over the expansion joints on one of the many highway bridges within a couple of miles of Moss Island. It was recorded on March 8, 2008 at 6:41 a.m.:

Listen to the audio clip

As the sonogram shows, this is actually a fairly mushy sound. It isn't crisp and sharp like the "real thing" would be. This mushiness is noticeable to the ear as well -- they sound distant and sort of muffled. You can also hear, and faintly see, the long trailing echo again, typical of loud, distant human-caused mechanical sounds and gunshots.

One more clip illustrates some other interesting features. This was recorded on March 11, 2008 at 3:35 p.m., the afternoon of the day on which I heard my first double knocks. Scott, Melinda and I were all sitting quietly at Hushpuckett Lake at this time, hearing nothing of interest:

Listen to the audio clip

This is the only one of the 12 clips I was provided with that even required a second listen. Sometimes, depending on your headphones and software, you will hear something that sounds rather like a woodpeckery double knock in the middle. Two things to note about this clip:

1. The spectrum of the knocks or bangs is different than the mechanical bangs shown before. It is focused in a narrower range of frequencies, centered in this case around 400 Hz. It also lacks the long, smeary trailing echo. This low, narrow frequency range is what gives real double knocks their resonant, sonorous, woody quality. Just to make matters confusing, a similar effect can also be produced by whacking on the hull of a boat, or banging your boat in to a tree. It's a predominance of lower frequencies with a fairly distinct peak at a specific tone, as compared to the mechanical bangs (and gunshots) that include much more high-frequency sound and not such a strong dominant pitch. The two loudest knocks in the middle of the sonogram are actually not that bad of a match for a "real" double knock, in terms of quality and timing, though the second knock is a bit high in pitch compared to the first one.

2. There are multiple other knocks in this clip, all within a little more than a half second. The pair that can sound like a double knock are the second and third notes. About 0.5s before them (just to the right of the vertical bar marking 6.0s in the sonogram) there is a fainter but distinctly audible knock; right after the pair are two other faint knocks that to the ear get swallowed by the middle louder sounds: "knock......KNOCKNOCKockknock." It is impossible to tell if these sounds are all from the same source, since the ARU gives you no information at all about direction and no real clues as to distance. If you isolate just this one second of the clip and play it in a continuous loop with the volume turned up, you can hear all five of these knocks and the superficial similarity to a double knock fades quickly. To a live observer hearing the sound in person, I suspect all that extra banging would have been more evident and it would not have come across as a double knock at all. My guess as to the source of these sounds is either one or two loudly foraging Pileateds, or a distant boat banging through the woods.

I have never been a big fan of Robobirders, like ARUs and automatic cameras. To me the absence of all the extra contextual data you get from a real, live person is a near-fatal flaw. The advantage of continuous surveillance and an "objective" record to me is largely outweighed by the time- and resource-consuming nature of the data processing and the ambiguous nature of all these "objective" results, mandating a massive amount of "subjective" (and controversial) interpretation. A large investment in time and resources has gone into robobirding for Ivorybills in recent years, resulting in only a modest amount of data the significance of which no one can agree on. For the most part it has yielded exactly what the human beings have produced: mysterious sounds of indefinite origin, and visuals that might suggest but come nowhere near proving. I realize they have their use as a compliment to the in-the-flesh field observations; but when the robots begin to take precedence over the humans that just isn't right.

Some might argue that this very ability to look at the sonograms and analyze their detailed characteristics is exactly why they are more valuable that reports from observers. But in fact, the Cornell analysts were not able to identify the sounds in those first two clips as anything but "mechanical sounds" of undetermined origin. I can identify them as "river sounds" and "bridge whumps" because of my actual "subjective" human experience at the site. And I would argue strongly that all these distinctions that can be picked out from the spectrogram are even more obvious to the ear of an experienced, skilled, discriminating observer (the only kind of people we should be fielding on these expeditions). All of these sounds are vastly bigger, richer, and more complex than what gets picked up by the RoboBirders and visualized on the sonogram. The live observer perceives, processes, and incorporates all this additional information. Ideally s/he also manages to capture a recording of what s/he observed; but lacking this does not mean that his/her observation is meaningless, or even less valuable than an ARU clip or crappy blurry video. I could just as well argue that the ARU data are meaningless because they are not corroborated by the experience and judgement of a skilled live observer.

Though I will get in to this more towards the end of my saga, we did eventually manage to get a recording of an example of the sounds that we live observers had been calling "Campephilus-like double knocks." Though disagreement continues about the origin of those sounds, I would like the record to reflect that all who have heard the recording have agreed that they are indeed dead ringers for the "real thing" as best as can be determined, both to the ear and on the sonogram. So it seems that even we Tennesseans with our fallible human ears actually can tell the difference between potential double knocks and the plethora of other possible confusion sounds.

Other posts in this series:
First
Previous
Next