ANDREW FARNSWORTH: Thank you all for being here. I really appreciate it. This is a fun opportunity, especially living in Manhattan. Nice to be able to talk to you about this.
So what I'm going to be telling you about today is about a really neat new project that the lab has for forecasting bird migration. And when I say lab, I mean Lab of Ornithology at Cornell. It's a collaborative effort. And I'll tell you a little bit about it.
I'll give you some background on the project, and some of what we're thinking about, and show you some examples, and then sort of tie it together with where this is all going. To start, though, I definitely want to sort of orient you to thinking about bird migration a little bit and that a huge percentage of it happens at night.
So what you're hearing now are the calls of birds migrating at night. Each one of these sounds is unique to a species. In this case, there were a lot of Swainson's thrushes and some other songbirds. There are some herons in here, American bittern. But a huge percentage of bird migration that happens, spanning the hemisphere, occurs at night.
So there are a number of ways that we study this. But it can be a very challenging thing to get at. So this BirdCast project is interested in understanding, really at a continental scale, how to predict bird migration.
And as I'm saying, it's a collaborative project. It's not just Cornell. It's a joint effort between Cornell, Oregon State University. It's funded by the National Science Foundation and Leon Levy Foundation. There are also some other partners in Microsoft; and at University of Massachusetts, Amherst; the National Oceanic and Atmospheric Administration; and additionally, Cornell University Computer Science.
So it's really a blend of biology and computer science. The computer scientists are interested from the perspective of saying, OK, there's a lot of messy data here. Try to understand where birds are, maybe how weather influences migration, a suite of other data sets that are difficult to use to predict. Computer scientists are interested in understanding that. The biologists, such as myself, are interested in understanding bird migration and understanding patterns of migration.
So the partnership is really designed to think about how to take all this messy information, what we know about bird migration, and make predictions based on it. Figure out where birds are going to migrate; when they're going to migrate; how many; how far they're going to go; what species might appear on a given day, in a given location, anywhere in the US. So some really fun stuff from people that are interested in birds.
And it's not just about sort of these predictions, not just about the information in and of itself. There's also some conservation ideal behind it as well, with the idea that providing this kind of a forecast and this kind of information might be useful for people that are interested in understanding, OK, if I operate or site wind turbines in a given location, is that going to be problematic or hazardous for birds passing an area? If a city, for example, New York, has lots of buildings that are lit up at night or projects that involve light at night, are there going to be certain times during the year when it might not be a good idea to have those lights on extensively throughout the night because birds are attracted to light when they're migrating at night. So there are some conservation connections here.
And then also, there are some broader applications that relate to, as I said, the study of the biology of migration, that relate to climate change and understanding how animals, specifically birds, might respond to climate change, also about how birds use habitats around, for example, the New York City area or anywhere where you're monitoring. As birds are passing through, they're going to be stopping over at times. And they're going to use certain habitats. So understanding something about these patterns of occurrence, where birds are going to show up, when, and how many, and so on, relates back to that.
And also, simply for understanding, birds are great bioindicators. They're really great, sort of monitors for environmental health. So understanding something about movements of birds and migration of birds across a broad scale, like the US, could be very powerful for understanding population trends. Again, all related back to conservation.
A little more background about migration. So with sort of a New York-centric view. The New York state is highlighted in yellow. This is a list of species that are commonly occurring in New York, that breed in New York, that pass through New York. And the point of this slide is to generate that-- or to illustrate, excuse me, that there are connections going on here between, say, New York state and Central America, South America. That birds that are breeding in New York state are migrating much, much farther afield than simply New York or even the southern US.
Birds are moving thousands of miles, to the Andes. For example, this pattern-- you can't see the color so well. But this is a strong link, in this area of the map, between New York and northern South America, for a couple of species that are represented here. There's an olive-sided flycatcher; black-billed cuckoo; a few warblers, blackpoll warbler, bay-breasted, species that breed in the Adirondacks, for example, in New York; scarlet tanager; a bird-- a beautiful bird of forested landscapes in Upstate; Caspian terns; Swainson's thrush.
These are birds that, as I said, are breeding or passing through New York. But there's a connection between these species and what's going on elsewhere, in the hemisphere even. So there's a direct link between these populations that are in New York and wintering areas in the tropics. So, yeah, this is all sort of background.
AUDIENCE: If I could just interrupt for just second.
ANDREW FARNSWORTH: Sure.
AUDIENCE: I know this is going to make for TV.
ANDREW FARNSWORTH: Yeah. We can fix that.
AUDIENCE: But, um--
ANDREW FARNSWORTH: Lights?
AUDIENCE: Yeah. I'm going to have Edward turn off this light. Is that OK with you, Jake?
AUDIENCE: Thank you very much.
ANDREW FARNSWORTH: Oh, perfect. It's a little easier to see now. OK.
AUDIENCE: OK. Great. I appreciate that.
ANDREW FARNSWORTH: Sure. Sure.
AUDIENCE: Thank you, Edward.
AUDIENCE: You're welcome.
ANDREW FARNSWORTH: And so to get back to BirdCast here, thinking about this linkage and thinking about the fact that birds are moving from New York to places much farther afield, we're interested in studying those movements and in understanding how to forecast them. There are a couple of different ways that we can do that, remembering again that most of this migration happens at night. Obviously, not a lot of it is visible directly to the human eye or to observers that are out there. But there are some ways we can get at it.
This is one of them in particular. So this should look familiar, probably if you've ever watched the Weather Channel or ever seen weather on your local news station. This is a mosaic of data produced by radars, a network of weather radars that span the continental US. There are about 140 of them, maybe a little bit more. And there's pretty good coverage, really to monitor weather phenomena for the most part.
The radar was developed to help meteorologists forecast severe weather and understand the movements of meteorological phenomenon. As it turns out, radar is also extremely good at detecting other things in the atmosphere, notably birds. So what I want you to see here is that there are a couple places in the map where there are some very bright yellows, and reds, and greens, very blocky looking patterns. But then there's a lot of other area of the map, with sort of this light-stippled blue, maybe some lighter, more uniform green colors.
What you're seeing on this map is not only the rain, the very blocky sort of intense color patterns. But in these blues and more uniform patterns, you're seeing bird migration as it happens. This is an animation of, as I said, all 140 of the radar stations in the US, that I took from the internet last night. So this represents-- it ends at 4:41 AM this morning. And it started at about 7:41 PM the previous night.
And so what you're seeing in the animation is, as the Sun sets and birds begin to move into the atmosphere at night, they all depart about 30 to 45 minutes after sunset. They take off. They fly. They get detected by radar.
Some of them fly for a few hours. Some may fly all night. But the idea is that using this network that was created for something totally different, that you have a window into this incredible phenomenon that is actually occurring across the continent. And just to give you a sense of the scale-- so some of these areas, in particular the green colors over New York and the green colors in the Upper Midwest, that happen around maybe 11:00 or 12:00 at night, those green pixels represent probably something like 1,500 to 2,000 birds in a cubic kilometer. So if you think of a city block, and you think of the volume in a city block, and you put about 2,000 birds in there, that's a very, very dense movement.
And one pixel in this case, one green pixel, represents a kilometer of space. And you think about how many kilometers of space there are in the continental US, you're talking about tens of millions of birds migrating on a given night, in particular last night, over parts of the Upper Midwest, the Northeast, all through the Southeast, even across into the West. So incredibly powerful tool.
And one of the things BirdCast is doing with this type of data is actually saying, all right, well, we'd like to take out all of the rain. We'd like to take out all the other things that radar detects and talk about only birds. So we built some models to start to do this. The first stage of the project really is to say, all right, what if we subtracted out all the rain, and all of the insects, and all of the other things that appear in radar imagery and talk only about birds? What kinds of information might we be able to extract?
This is a simplified map, actually from some data collected around this same time last year, three hours after the Sun set, on September 8th last year. And what we've done is basically done a very simple set of-- through a simple set of algorithms, factored out rain, factored out other things that might be moving, with the wind, like insects, and bats, and other things radar might detect, and come up with numbers, for three radar stations around New York state, for densities of birds moving at this time and in the direction that the arrows are pointing.
So in this case, between a hundred and 200 birds per cubic kilometer, sort of Upstate New York, in Binghamton, moving generally in a southerly direction; in Albany, moving sort of in a west/southwesterly direction, and New York, to a lesser extent, a little bit lower numbers, in a southwesterly direction. So the idea here is, as a very preliminary sort of first stage in this project, that we go through a radar image-- set of radar images like this one and extract out all of the information that is just birds, and get a density and direction of migration. So super cool stuff.
The only thing about radar though is it's just pixels on a screen. It's representative of birds that are migrating. But we know nothing about the species from radar. Radar is really good at telling you where things are, and how many of them, and in what direction, and so on. But it's very bad at telling you what the species are.
So thankfully, we're going to employ another set of information, that we're starting to gather. Those sounds, like the sounds I played you in the first slide of the presentation, of all of these vocalizations that birds use when they're migrating at night. So with each of these calls, there's a species identity, that we can then start to relate back to the radar and understand, OK, not only how many and in what direction are birds moving, but what are the species? What's the composition?
And just to give you a sample of what some of these things sound like, actually with the species represented-- so it's not just a sound, like what you heard originally. If you could actually see what was migrating, you'd be able to put a face on the sound. There are a few blackbird-type birds here, like dickcissel--
--and bobolink. Bobolink is a grassland breeding species, that breeds in upstate New York. Actually, probably in the next couple of days would be a good time to go out and look for this species as it passes through the area.
Red-breasted nuthatch-- some of you may have bird feeders and have heard this sort of cute, little, tiny toy horn sound. The red-breasted nuthatch is actually a species that appears in this area, sometimes irregularly, in very large numbers. And this looks like one of those years where that might happen. A couple other examples-- white-throated sparrow, a little bit harder to hear, a very high frequency call.
So these are flight calls, as I was saying. They are species specific. Most birds use them in social migratory contact. So when they're migrating at night, they're trying to stay in touch with one another. In a sense, trying to say, where are you, what direction are you going, are you going down when you land? I want to have safety in numbers. So I want to stay close to species like me or things that look like me.
So these calls, as you can tell, are, in some cases, very faint, very high pitched. They tend to be single notes. And almost all of them are less than a quarter of a second long. Some of them are really only 20 to 30 milliseconds long. So extremely short, extremely simple sounds that birds are using to communicate, very different from birdsong, territorial song, that you might hear when birds are nesting.
So by thinking about taking radar data with this sound data, the acoustic data, you have a more complete picture. And building these things into models, we're hoping to take this information about the species composition, and where, and how many, and when, and in what direction birds are migrating, and start to assemble these things into a forecast. Now, that said, there are some challenges involved.
This is a very cool representation, I think, of all of the warblers, a very colorful group of birds, diverse group, 48 species here, that occur in North America, that migrate. So there's an image associated with each of the species, that shows you the plumage of those birds, what they look like.
Now, there's also a visual representation of some of those sounds that I just played, their flight calls. And, for an example, a really complex example, where there's sort of a real challenge in identifying these species, I'd like to play these in real time. And then I'll play you a slowed down version, just to give you a flavor for what these flight calls are all about and some of the challenges that we might have in using them to try to model migration, and to identify these sounds, and assign them properly to species. So here's the real-time version.
[WARBLER FLIGHT CALLS]
So you guys got all those, right? You know all the species. You can tell them all apart. Again, each one is unique to a species.
AUDIENCE: That was all 48 of them?
ANDREW FARNSWORTH: That was all 48. You get the idea that it's--
--very fast. These are the kinds of calls that are 50 to at most a hundred milliseconds in length, so very, very short. If you slow it down, things get a little more interesting. And I'll follow along with the pointers. You can keep track of what species are what.
ANDREW FARNSWORTH: I might.
[WARBLER FLIGHT CALLS]
So you get the idea that when you-- this is slowed down to a 10th speed, that there's a lot of information here. We may not be able to hear it terribly well. But birds are getting information here. This is, as I said, one of the ways that they communicate.
And we are trying to take data like these and automate the process of extracting these kinds of sounds from, say, thousands of hours of recordings, and then build that, together with the radar data, to make these forecasts in the future. So just a sample of what the kinds of data are about.
And finally, it helps that when we're thinking about nocturnal migration, obviously the radar and the flight calls aside, those are all about what's happening at night, when the birds are migrating. It would be nice to have some kind of ground truth for what happens at night in building these models, and in thinking about where birds are, and where they're starting from, and where they end up. And we have a very unique opportunity these days to access some data like that.
It's from a project called eBird. And right now, it's getting maybe 150,000 to 250,000 checklists every month. So birders that are going out and going birding during the day, are reporting their information to eBird. And it's being saved in a large database. And it's over 100 million observations now, over the last 10 years.
So incredibly large amount of information about where birds are, when, how much people-- excuse me-- how much effort people have put forth in finding those birds, some comments on the birds, photographs of the birds. All sorts of useful information to tell us actually, OK, well, you say nocturnal migration happened last night, what does it actually look like on the ground when birds arrive?
So taking the migration as it happens, with radar and acoustic data, and then building in this component of ground-truth data from birders that are going out and submitting their checklists, we think is going to be a really powerful suite of tools to start answering these questions about, OK, what birds are going to show up where, when, how many? How far birds are going to fly?
When are we going to see scarlet tanagers in Central Park in May? When are we going to see the first white-throated sparrows appear in the fall? Questions like this-- what night is going to be a big night for thousands of birds per cubic kilometer? What might be a good night to turn off building lights, or so on?
These kinds of questions are the kinds of things that we're going to want to get at. This project is in its first year. And what we're starting to do with some of this type of information is to think about, all right, we have radar, we have acoustics. What happens-- we want to understand bird distribution over the course of the year as well, sort of a bigger picture. So from eBird, we're able to take a year-- let's say an annual sample of the distribution of this species, white-throated sparrow.
So by March or April, you see that the birds start to move north into the US. By May, late May, June, they're in areas where they're breeding. Again, this is where people are seeing the species. Birders are out.
And then later in the summer, maybe beginning around August, the end of August or September, you start to see the birds move south again. And in this area, there's a large movement in October. Birds pass through. And some of them spend the winter. But by the middle of November, the species is back in the area where it's wintering again.
And this kind of information we think is going to be very powerful for understanding how to forecast where birds are going to appear, when. Looking backwards in time right now, to build these models is really the stage where we're thinking, OK, we have this unique set of information from radar, a unique set of data from the sounds of birds migrating at night, and this incredibly large data set from citizen scientists going out and collecting bird information.
And what we're going to do is present this on the web in the future, probably a few years out. Hopefully with an app that will tell you if, in a given place, OK, you're going to see the following species tomorrow. If you go out at night, you're likely to experience these species calling or this much migration if you look at the radar.
So this is our website. We've just launched. We're going to be adding forecasts right now, that are really done by me and a couple of other people sort of in advance of letting the models do it. And we're going to be posting them on a weekly basis here. Birdcast.info is the website.
And really the stage we are now, just to bring it back a little bit, if you think about the evolution of meteorological forecasting, the early days of it were very much about looking at the sky, and thinking about relative humidity, and understanding, OK, well, when I see these patterns, I expect that it's going to rain tomorrow. And after that, after the rain, it's going to clear up, and it's going to get cooler, and so on. It was very much about observational experience to make the forecasts.
Weather forecasts have gotten tremendously better over the last several decades, with computer technology and with modeling of when certain systems are going to appear, where hurricanes track, when is it going to rain, what are conditions going to be like tomorrow, the next day, et cetera. We want to do the same thing with the forecast of bird migration.
So right now, it's in that very early stage of taking the knowledge that people like me and my colleagues have, about what we know about bird migration and when we expect certain birds to be in places. And we're making forecasts based on that. And we're saving all that information. We're saving the radar data.
All of these forecasts we're making are going to go back into the modeling that we're doing, to try to assess the models and evaluate, well, are these features about weather, the temperature, or the wind speed, the wind direction, are they important in predicting when birds move in certain times, in certain places? Having this massive amount of information and a set of computer scientists that are interested in making sense of it and helping us to build models that look forward, is all about what this project is doing.
So right now, we're talking in really very human terms. In a couple of years, probably beginning next year or the year after, we're going to start automating this process. So we're going to use the models to tell us what's going to happen tomorrow. And then probably we're going to have some opportunities where I go against the machine and say, all right, I think the model is going to-- the model's wrong in these places. Here's what I think is going to happen.
It's going to evolve in ways like that. And again, the output at the end is really information about where birds are going to be when, how many, what species, through the web, hopefully through smartphone apps, and really give us a new powerful sense of understanding how birds are moving in space and time, that then we can use for conservation purposes. That's really the end goal here.
There's biological information. There's a coolness factor for people that are interested in birds and interested in biology. There is a coolness factor for people that are interested in computer science and data, and so on.
But the bottom line is-- and this is really a mantra of the lab-- is that we're thinking about biodiversity. We're thinking about conservation, specifically about conservation of birds and understanding patterns of birds, to try to make sense of how can we better understand when birds are in trouble, or what things are working, to set aside habitats to save birds, et cetera, that kind of thing.
And so as I said, there are a number of partners here. This is definitely not just me in this project. This is funded by National Science Foundation and Leon Levy. It's a collaborative effort. And there are a couple of different sites I'd like to point you to.
The first one is that birdcast.info site. That's the one that will have the weekly forecast. It has more background information about the project and all of the people involved in it. That's going to certainly be updated at least weekly with forecasts, not only for the New York City area, but also for the entire US.
And, additionally, I want to clue you in to a couple of-- one of the-- I mentioned eBird before and the ground truthing. Ebird.org and a smartphone app called BirdLog, which is a handheld data entry app that allows you to submit your sightings to this big eBird database, are two things that are going to be critically important for this project in terms of the ground truthing of what birds are actually in a location in a given time.
So with those two things, and with the radar information, and the acoustics, we're looking forward to some really, really fun stuff, I think, in the next couple of months, but certainly in the next couple of years. So with that, I think I'll stop talking and take some questions, if there are any.
ANDREW FARNSWORTH: All right, yeah.
AUDIENCE: How do you collect the acoustic data?
ANDREW FARNSWORTH: How do we collect it? So-- good question-- so there are a couple of different ways that we're doing it now, with a very simple setup. Some people are using microphones that are connected directly to home computers, running some sound analysis software. And basically, you just put a microphone outside. You have it set up.
You start recording at night. And there are a couple of different ways to automatically extract, say, the sounds of these flight calls from that incoming audio stream.
We're also doing it in places where, obviously, there aren't so many people. Were using these, what we call autonomous recording units. They're basically just programmable recording devices, to which you connect a microphone and can go put it out wherever you like. Point the microphone up, have it on a preprogrammed schedule to collect data, bring the data back in the next day or-- presently, it's not quite so real time-- the next week or something.
And start analyzing it, again with this sound analysis software and some of the algorithms that we've developed to go through and extract the relevant information. As you can imagine, it's easy to collect a lot of recordings. I mean, those of you that are sound oriented, thinking about making an audio recording of eight or 10 hours, you immediately start to think, well, if these sounds are all less than 100 milliseconds long, and they're probably not continuous throughout the night, that's a lot of dead space that you have to go through.
So we're looking at automating the process with computers, so that humans don't have to listen through all thousand and thousands of hours of recording. But that's the approach, using microphones and recording devices, some connected to computers, some connected to portable recorders that you could deploy out into the field.
AUDIENCE: So you said, simple algorithms to filter out insect, rain, and bat. How is that simple/
ANDREW FARNSWORTH: It's not simple. It's not. It's not simple, yeah, yeah. The goal is to have it be simple. Right now, it actually is--
ANDREW FARNSWORTH: Yeah. The insect and bat problem, biological stuff showing up on radar, for meteorologists is just a simple problem because that stuff can easily be eliminated--
AUDIENCE: From the weather data.
ANDREW FARNSWORTH: --from the weather data. But separating insects, bats, and birds is primarily about speed of each of those things. So birds are incredibly directed powered fliers at night. And they're often about, maybe 5 to 6 meters per second above whatever the wind speed is, or moving across the wind speed. So that's one of the ways, using that simple depiction of birds per cubic kilometer and direction of movement.
The algorithms that we started with, sort of the standard, is to take that reflectivity data, that tells you how many things are up in the atmosphere on radar, overlay that with the velocity for the same information, overlay that with wind data, actual wind speed recorded by balloons, and then threshold. And say, all right, everything that's above a given wind speed by 5 to 6 meters per second, we're confident is birds.
So it's sort of a step by step process to extract, not only from just everything, well, OK, what's meteorology? We're not interested in that. What's biological? We're interested in birds. We want to get rid of the things that are just floating with the wind, like insects or bats, which tend to be less directed at night.
AUDIENCE: So would you actually be able to separate the warblers?
ANDREW FARNSWORTH: No. You can--
AUDIENCE: You can't tell by listening, it's warblers.
ANDREW FARNSWORTH: You tell by listening, yes.
AUDIENCE: You can?
ANDREW FARNSWORTH: Yes.
AUDIENCE: At 4:48 AM?
ANDREW FARNSWORTH: Yeah, probably. I mean, it's close to it. If you hear it well, I mean this is-- there are some challenges, no question. There are some where you were you could say, well, you know what, it's one of these five species. And there-
ANDREW FARNSWORTH: There are some--
AUDIENCE: Some were discernible. But some were [INAUDIBLE]
ANDREW FARNSWORTH: Yes. Some are extremely similar, absolutely. One of the things that I've learned very quickly-- so these flight calls are what I studied during my graduate work and before that, just as someone who's interested in birds, and has gone out since a very young age.
It became clear very early that there is no way to put a name to everything. It's not possible because there are situations where you just cannot identify a sound. You hear it too quickly. There is background noise. Maybe you don't know what it was.
AUDIENCE: [INAUDIBLE] clean as the chart. You can pretty much get that.
ANDREW FARNSWORTH: Yes, that's right. That's right.
AUDIENCE: So this is a little off point. But do different species of warblers, let's just take as an example, that are so close in sound, do they ever interbreed? Are they absolutely a distinct species?
ANDREW FARNSWORTH: No. Sometimes they do. Sometimes they hybridize.
It's mostly a function of-- it's a whole other topic. You know, the genetics of this whole group, in particular warblers, are very interesting. But there are times where there are hybrids.
In fact, some pairs, some sister, closely related species, hybridize so much that it's difficult to tell when it's one or another sometimes. Blue-winged warbler and golden-winged warbler are two-- what probably were warblers that were diverging over the last several million years. But then came back together as habitat and climate changed.
AUDIENCE: [INAUDIBLE] the cowbirds in Mexico and the cowbirds in [INAUDIBLE].
ANDREW FARNSWORTH: Similar. Yeah. That's right. That's right.
So definitely a digression, but a very interesting story. And the flight calls, they likely don't relate really to the hybridization so much. You may be able to detect it. But we don't know that yet. We actually don't. There's not enough information from hybrids.
AUDIENCE: [INAUDIBLE] species [INAUDIBLE]
ANDREW FARNSWORTH: Could be. Yeah. That's right. That's right.
AUDIENCE: Thank you.
ANDREW FARNSWORTH: Sure. Sure.
AUDIENCE: Any other questions?
AUDIENCE: Yeah. Can you tell us a little more about what you envisage doing to track climate change, and how it affects birds and adaptation? You know, there's a theory that says that birds are most affected by climate change, than amphibians. And I'm wondering if you can disprove that?
ANDREW FARNSWORTH: I don't know if we can disprove it. Climate change and migration go hand in hand. I mean really bird migration is, in fact, at a very fundamental level, a response to climate change. So some species are, in fact, able to deal with changing climates very well.
If there is a large enough population of birds, they can actually respond. If there's enough variation in the timing that birds move and so on, they can actually respond to climate change. But that's not the case for everything.
So the idea with how we might use the information we get, to think about studying climate change, is all about arrival dates, and when birds are appearing. And looking for patterns, where over the last 50 years, maybe we see certain species appearing much earlier, certain years. Over the course of time, on average, maybe a few days earlier every year. That kind of pattern, especially in eBird data, is something we're expecting to detect. So look at arrival times and across a large number of years is a place where we expect to be able to see some of these changes.
Sometimes, we're going to see marked differences between years. This year, for example, there were patterns in March and April, where birds were arriving, like, two or three weeks ahead of typical average times. And the problem with that is that if birds are arriving so far in advance of the, quote, "average time," is there can be a mismatch between the food that they're really trying to target, and the areas they're going, and when they arrive.
So if the temperatures are much warmer three weeks ahead of schedule, and leaf-out occurs three weeks ahead of schedule, and the bloom of caterpillars and insects peaks three weeks ahead of schedule, but birds are arriving at the time they normally do, this asynchrony can develop. And without proper resources, birds are not going to breed well. Populations are not going to increase each year. And you're going to start to see declines.
So one of the things that we're hoping to be able to look at here is, say, on a daily kind of a level or a weekly level, what sorts of patterns are birds responding to in local meteorology, in weather? How do they respond to a cold front, or a warm front, a shift in direction in the wind, distribution of rain, that kind of stuff? Over the course of time, that becomes relevant for thinking about, well, what happens if, in a climate change scenario, we expect more severe storms, or fewer more severe storms, or really changing patterns of circulation?
Birds that are following certain pathways, that we may be able to identify in this project, may be more susceptible to problems on route during migration than others. So it's really an opportunity to think about perhaps not only what species are at risk now, but what species might be at risk in the future, given what different attributes they might be cuing on, given what patterns they actually use in migration. Because migration is, even though it seems like this-- it's an incredible feat that birds are doing. But it also is a very stable way for birds to do things.
It's not like it may be stepping out on a limb. It's something that works on an evolutionary time scale. Birds have been doing it for millions and millions of years. The question is, how does it respond to super-rapid changes? Some birds will respond well. Some birds will not.
We think there is a lot of information that may relate to thinking about climate change and monitoring it, that can come out of this project.
AUDIENCE: And do you have a favorite bird migration feat you can tell us?
ANDREW FARNSWORTH: There are some spectacular ones. One that I like in particular-- I showed a species earlier called the blackpoll warbler. It's a small [INAUDIBLE] Canada, up into Alaska. And there are individuals in that species that choose a migration strategy that sees them departing from, say, Nova Scotia, or Maine, or Massachusetts, departing southeast in the fall, over the northwest Atlantic. And staying out over the Atlantic for several days, maybe as much as a week sometimes. And not landing until they reach somewhere, reasonably far into South America.
So you're talking about a multi-thousand kilometer trip for a bird that's like 15 grams, 19 grams. And they're not stopping. They're not feeding on route. They're doing this straight shot.
There are other birds, some shorebirds in particular, Bartel godwit and whimbrel, which are larger-bodied shorebirds. One, really, the Bartel godwit's a European and Asian species primarily. Whimbrel is found around the world.
But some of these species are doing the same kinds of things, and going even farther. Bartel godwit has made an 11,000-mile jump, in basically one stop. Take off-- take off somewhere in Asia or Alaska. Land somewhere in the South Pacific, a week or 10 days later. So absolutely incredible.
And the other thing is that, as I said, much of this is happening at night. So birds are not using the typical cues that we might to navigate and orient. They're using things like stars and the magnetic field. They may be using acoustic cues different from even their own calls. But listening to the approach of a thunderstorm, or listening to the pattern of waves breaking on a coastline, to understand where they are.
So there are some really incredible things that go on with migration. There is a list of them. But those are some of my favorites, I think.
AUDIENCE: Is there actually evidence that they use magnetic and polar--
ANDREW FARNSWORTH: There's good evidence. Actually for the last couple of decades, there's been increasing acceptance of this. And I think there are very few people in the ornithological community that would say that there's any reason to think that it's not something birds do all the time.
AUDIENCE: But it's empirical. It's not a [INAUDIBLE]
ANDREW FARNSWORTH: They've actually done--
ANDREW FARNSWORTH: They actually have. They found a molecule in the eye, a macromolecule that helps birds actually sense the magnetic field.
ANDREW FARNSWORTH: And there was some thought for a time that there might be a particular organ involved, which there likely still is. But this molecule in the eye that seems to be the most likely culprit, actually is something that's been known for a number of years-- a few years, not too long. Real cutting edge kind of research, that actually is, for the first time, not just empirical evidence, that shows, OK, if you change the magnetic field, and you have a bird in captivity that's displaying sort of migratory behaviors, that it will change those based on the way you shift the magnetic field.
This is actually looking at the way the brain works and the neurobiology of it. And they actually found what they think is the source of the way birds use it. So really, really neat stuff. Yeah.
These feats of migration, there are many. They relate to both distance traveled and time traveled. They also relate to the way they do it.
And I think just the notion of-- the number-- simply the magnitude of migration that can happen on a given night-- The forecast that I made for New York City area in the next couple of days suggests this. But probably on Sunday or Monday night, if you were to go up to the Empire State Building after sunset, maybe an hour or two after sunset, and you spent a couple of hours up there, you'd actually see birds flying past at night. And on good nights, heavy migration nights, you're going to see that sort of tens-- hundreds of thousands of birds per cubic kilometer.
So birds passing there, tens of birds per minute sometimes. Some of them will fly around the Empire State Building at night, depending on the conditions. But they're illuminated by the city. So you see these streaking, bright songbirds flying past. It can be quite spectacular.
So I expect that there's going to be a good night for that early in the week, probably Sunday night or Monday, depending on when the weather, the nice, cool weather that they're forecasting, actually shows up. So lots of cool migration stories to tell. Yeah?
AUDIENCE: Is any of the acoustic data that you're going to be using [INAUDIBLE]
ANDREW FARNSWORTH: It's not available in the same kind of a nice packaged format that the eBird data are. But it's absolutely something that we send it out when people request it. There are places sometimes where we post it.
It's not quite to the level of refined sort of display and visualizing the data. But it's absolutely something that's available. And if you're interested in it, I can certainly send you a copy of it. There's no problem doing that. And all the sounds that I have here in this presentation, I can easily send.
The presentation itself I'm going to post on the birdcast.info website, either later today or tomorrow. But, yeah, it's no problem to send those kinds of sounds. I warn you, for audio files, some of these recordings are not the cleanest recordings. There's a lot of information, other noise, that gets into them, planes, cars, insects, lots of background signals.
So they're raw recordings. But there are some with really, really good, cool, like call stuff in them too.
AUDIENCE: Because the people that you have out there recording them, are trained, right? They're not just random people, right.
ANDREW FARNSWORTH: Not necessarily. Some of it is just very much a do-it-yourself approach.
There's a website that a friend of mine runs, called oldbird.org, that shows you how to build microphones for maybe $20 to $50. And you can build these, and attach them to your computer, and use them. So it's not necessarily skilled recordists that are going out there and making these recordings.
It could very much be a simple-- a homemade kind of microphone that's producing this information. And, in fact, we're looking for that kind of-- if anything, we're more interested in that approach because of the potential to, say, have a lot of people that are just interested in having microphones on their roofs, that are reasonably cheap, that could be sent out to hundreds or thousands of people. We'd definitely be much more interested in that kind of broad-scale approach, than even a handful of people that are maybe the best recordists around, going on recording.
We want both data. But for a project like this, we're really interested in the amateur kind of recordists, that just get inspired to build the microphone, and put it up on top of their house, and send us the information.
So one of the other components to this project, that probably will happen in the next year or two, is this notion of creating a bit more of a network, so that people that do this can provide us the data in a timely fashion. So that as the data are getting recorded-- let's say you have a microphone up on top of your roof. And it's plugged into your computer. And you've got some Cornell software, like Raven, which is a sound analysis package.
You have it running and extracting flight calls, as the microphone detects them. And those go into a folder. The data from that folder gets sent to Cornell, to a server, where we are processing those things. And then by morning, we actually have a histogram of what calls and what species occurred last night, that would go into the forecast for the next day.
That's where we're thinking that this notion of amateurs, that by the hundreds or thousands, could be involved in the collection of the acoustic data. That's the direction that we're going with this. And I think it's very much a real one.
We showed, well, a couple of years ago, that something like this could work. There are other projects that do similar things. We're very positive about it. So I think it's going to be an exciting way of creating a whole new network.
And one of the other possibilities is that-- even the smartphone community, we haven't done it yet, per se. But the notion of using an iPhone to record nocturnal bird migration absolutely works. There's no question that if you put an iPhone out on a window sill, in Columbia County, or in Westchester, or wherever, and you have a night of migration, you're going to record flight calls on it.
So the idea of potentially engaging that sort of approach is huge. We're excited about it. It hasn't happened yet. But there are a couple of people that are working on it and thinking about it. So it's a possibility that I think is very real.
And it's going to be something that I think happens, at least in a pilot form, probably in a more expansive form, during the lifecycle of this project. And I should mention this is-- so this project is funded through 2015. We're just in the end of the first year of it now. Yeah?
AUDIENCE: So obviously, we're all here because we're interested in bird migration. And you're an ornithologist. But I'm curious, since you keep saying that you guys need to filter out the sounds of the bats or insects, has anyone approached you and said, like hand over that insect data? Is anyone else interested, in a different field?
ANDREW FARNSWORTH: There are some people that are interested in the insect data. They have not come out of the woodwork yet. But there have been enough people that have made inquiries about it, that it's very clear that saving all the data is going to be important. Because in the future, understanding when, like a fork-tailed tree katydid, or something, first started calling-- there is actually climate change information there as well, absolutely.
So we're saving all of this information. We plan to save it all. And from the perspective of analyzing the sound for birds, we know that the technology to analyze this automatically using computers is going to get better over the next 10 years. We assume that the same thing is going to happen for other sounds that we recorded in the data set, whether it's insects, or wind, or other things that may be of interest to people.
So the people that have expressed interest so far are kind of on the leading edge of thinking, like, oh, you guys are going to have a neat data set that we could piggy-back on. That's cool. And we want to do that.
And it's funny how this kind of thing happens. It actually just happened to us, when it comes to sort of another-- a little bit of a side component to this project. One of the other things we're interested in is-- amateur astronomers are using cameras now, low light video cameras, to track meteor showers and to understand how high, and where they're going, and so on.
And in talking with some of them, they said we have this problem that we keep getting birds in the cameras. And immediately the light went off, like, oh, boy, here's a great way to ground truth radar data with low light video cameras and actually count the numbers of birds. You can't tell what the-- it's a very, very difficult identification to identify black and white pixels flapping across a screen, in five seconds or something. But you can tell at least how many birds there are, and what direction they're going, and so on.
So we're going to do a pilot actually, I think in a couple of weeks, in Manhattan, and then later again in October, where we try using these cameras to count birds that are flying at night, as a sort of an add-on to this project. It's not specifically built into the way we're thinking about the project right now. But it could be.
And it's another one of these things, like, OK, well, here's another unique data set that might actually be very informative, in the same way that sound data might be informative to someone else.
AUDIENCE: What kind of cameras are we talking about for the astronomy?
ANDREW FARNSWORTH: These are super simple, closed circuit television cameras, like the typical night cameras that you see. I mean even a camera like this, that's in this room, a webcam type thing, could do it. They may be a little bit more specialized. But not much more specialized than, say, like, $100 to $200 per camera it looks like, which is a lot less expensive than a million dollars for a radar unit or something like that.
So this is a sort of check back in a month or so and see. We're going to experiment with it. I've seen some of the data. I haven't actually played with the technology or the technique myself. But in two weeks, I will. And in another three or four weeks, I will. So it's sort of an ongoing thing.
AUDIENCE: Yeah. Anything interesting to report yet with this migratory season [INAUDIBLE]?
ANDREW FARNSWORTH: Yeah. There are some interesting patterns that are already looking like they're appearing. Something that happens for-- that we see in the ornithological community are these eruptions of species sometimes. So species that may have irregular movements, that are tied very closely to, say, pine cones and pine cone production, or something related to seed banks and other food like that.
And that these eruptions happen every few years. Or mice, with certain kinds of hawks and owls. And these eruptions are irregular. They're based on the pattern of where certain plants produce a lot of food one year and produce them the next year. And it looks like this year that there are a number of species-- red crossbill is one, which is one of these eruptive or nomadic kind of species, that's tied very closely to certain kinds of pine cones. They're staging a major eruption.
In fact, there were red crossbills in Central Park last weekend, which is an extremely rare bird in New York. It's not only something that is not seen every year, it's probably not seen every five or 10 years. There are really, really not many of them. And there were 10 to 15 of them in the park last week.
And based on some other information we're gathering from birders that are out in the Midwest and a few other places, they're starting to see large numbers of crossbills in Minnesota and Wisconsin. So it looks like there's an eruption building this year for that species.
The same thing for red-breasted nuthatch. I mentioned that one, that funny sort of toy horn sounding little guy. That's another species that every few years erupts en masse. And it moves much farther south into the US than that it's typically found and in much larger numbers. It looks like there may be something like that happening this year.
And there are a few species like that, that are sort of exhibiting the same-- those eruptive types. It also looks like there are some other-- maybe a few other species, like palm warbler, for example, which is-- I showed a picture of that. It's another bird that breeds across a lot of the boreal regions of North America and Canada.
That species are showing up a bit earlier than usual. There are some patterns of that occurring as well. Some people have suggested that some patterns of early occurrence for these noneruptives may indicate some problems with the breeding season, maybe because of drought conditions or other things. It's hard to know if that's true yet. But maybe a pattern like that is starting to appear.
That one's much less clear. The eruptive pattern, though, for the crossbills and a few other species does look like it's showing up.
AUDIENCE: And when you say [INAUDIBLE] for red crossbill and [INAUDIBLE]
ANDREW FARNSWORTH: Yeah. There's some really interesting, cutting edge work going on in red crossbill right now. And it relates to the flight call of the species, which relates back to the food that they're eating.
It looks like that species may, in fact, have as many as 10 species represented in what is now currently called red crossbill, maybe more, that are all identifiable by the flight call types. And each of them has a unique bill morphology for a-- a bill structure, a bill configuration, for a specific kind of pine.
So there are a couple of people who are working on that. There's somebody at Cornell that's interested in it, whose name is Matt Young. And these crossbills, for those that haven't seen them, have this unique structure of actually-- the mandible and maxilla of the upper and lower parts of the bill cross over. So it almost looks sort of like this.
So when they go into pines, and they can squeeze out the seeds that are in these tightly closed cones, it's an adaptation to extracting the seed out of the cone. So around here, I'm not sure exactly what they're going to be targeting in terms of the cone crop. But they're certainly going to be targeting conifers specifically.
And in Central Park, they were in the area with some of the white pines and a few others, right around the Shakespeare Garden. So in general, if you wanted to look for crossbills, look to try to see this eruption as it happens, and also for nuthatchers, coniferous areas would be the place to look.
So the South Shore of Long Island, places like Jones Beach, that have lots of pines. Within the city, there are a few parts of Central Park that have a lot of pines, and so on. Those would be the places to target. The same thing for red-breasted nuthatch.
AUDIENCE: So we'll see red crossbills now?
ANDREW FARNSWORTH: You could.
AUDIENCE: They're so amazing.
ANDREW FARNSWORTH: Yeah. Well, they've been-- I don't think they've been around for a few days. But if what I think is going to happen, and what people have suggested might be happening, does, then over the course of the fall, we're going to see increasing numbers of crossbills. So that this is just the very early part of the invasion, if you will. That's the sort of the terminology that birders use for these eruptions of birds. It's sort of an invasion year.
Last year, it happened to be an invasion year for snowy owls. Snowy owls were found far south into the US and lots of different places where they don't normally occur because their food source, these rodents, crashed up north. And the birds were moving south, looking for food.
So these cycles, these irregular cycles, are going to be things that we're interested in, in BirdCast. It'll be really interesting to see if we can get at those. We're first trying to tackle the more regular cycles of birds that migrate every fall and return every spring. These crossbills and nuthatches may move every four falls or every six falls, and come back again two springs later, or something.
So there are these sort of odd patterns that we'll definitely want to get at with BirdCast. But they'll be the sort of second stage, second stage kind of stuff.
That said, I'm still going to be forecasting that stuff. If I think that crossbills are going to be coming, I want to make the statement that I think it's going to happen.
We've received your request
You will be notified by email when the transcript and captions are available. The process may take up to 5 business days. Please contact firstname.lastname@example.org if you have any questions about this request.
Andrew Farnsworth, of the Cornell Lab of Ornithology, provides insight on fall bird migration in the Northeast and New York City, at the "Inside Cornell" event on September 6, 2012.