Tip:
Highlight text to annotate it
X
Thank you Peter.
Yeah, I'm back for only a week this time and then I'm heading off to Melbourne, and Tatura,
and New Zealand and parts unknown.
And I'm on a big driving trip, so I have my wife with me and we're just gonna go back
and reminsce about our historic days in Victoria when we were back in '83/'84.
So we'll revisit Victoria.
As Peter said I want to just give you an update on what I've been up to.
And this is only a tiny little part.
I've got new projects in Alaska and upper great lakes, and a whole bunch of new things,
but we'll dispense with that, and just talk about the Samford work.
This is just the front page of our website, so for those of you who'd like to go there
and dabble in soundscape, that website is an open access website, so you can go there.
Look at the project or any of the other projects that are online.
It's probably most of you know, or some of you know that I've been around for a few years
trying to promote the important stuff: sound as an ecological attribute.
I won't go into the gory details about that, because we've been there before.
We've had, oh I don't know, 5 or 6 workshops over the last 7 years, Peter, I suppose, and
we've gone through a lot of the details and so on.
So I'd just like to give you a little bit of an update on the work that has been going
on in Samford.
Jason and I and others have been collaborating to put up some sensors there.
He had his group of sensors, and I put up a group of sensors to do, kind of a test,
and I locked into a technology that I like working with from wildlife acoustics.
They're little sensors that look like this, little boxes tied to trees.
And they have little ears on them, these little ears are the microphones.
And this is the technology we're using, sampling from 22 kilohertz and collecting Wave files
to record for 30minutes for a 2 minute duration.
And that just goes on and on, and they record every 1/2 hour.
We set them up almost exactly this time of the year last year, and I stayed out at Samford
for 17 days, and was able to kinda get things set up on my own, hobble around on a little
cane along the dirt tracks and try and avoid the snakes down there.
And this is the process that we go through.
I want to first of all recognise Ole, who is not here anymore, but he did a lot of the
leg-work, and of course Michelle and Jason: major players in the logistics of all this,
helped fix things that I had broken.
Or fix things that made them better when I hadn't done that.
So they certainly deserve a lot of the congratulations for keeping these things going, particularly
during a nasty time of the year.
Because about the day I left, it started raining.
-and didn't stop.
(chuckle) - it never stopped.
And I came back and brought the drought with me or something. (Laughter)
Anyway the process goes like this.
You program the recorders, you deploy them in the field then you collect the recordings
on a little SD card.
You FTP them to the regional server up in Michigan State University.
There they run an Auto-Load program which loads them up into the digital library.
You compute some metrics and then you can access the data in the metrics and you can
browse and search and do all kinds of filtering.
And then you can analyse and interpret the information.
That's basically the flow of work.
These are where the 8 sensors were located.
Trying to give them a little bit of diversity.
Some open field areas.
One near one of Peter's gas chambers here, next to a little farm.
One near the barracks.
Couple on the roads going in.
One down by the creek and a couple on the walking tracks, long way from the centre.
These are the sites.
Obviously the barracks site, but behind there is some interesting vegitation.
Here's the one down by the gas shed, down by the creek, and then one on the forest walking
tracks.
The other sites were located on the paperbark, adjacent to one of the old drives, one adjacent
of the new drives, and along the walking track.
So just kinda trying to capture little differences in habitat that are on the site.
Little bit of background so here's what these things sound like.
While this is playing, I'll kind of give you an idea about how we look at the world from
a stand-point of most of the biological information ranges between 2-11 kilohertz.
There's machines and other, what we call technophony or human voices, tend to be down on the lower
frequencies.
And we divided up the sound into these different sound levels to try and look at the kinds
of patterns of sounds at these different frequencies.
And certainly we have a long time series so we can look at these over time.
But here's a, for example 4:00 in the morning, and then, 1/2 an hour later, the whole world
changes.
That's an easy transition during just that period, and of course this is in, up on the 18 November, kind of
the late Spring.
And so the butcher birds and other critters wake up and start doing their work.
This is a little video that I produced for Michelle.
I think she liked it.
I think Jason liked it too.
But that's the idea of trying to take a visual view of the world at Samford, in trying to
see if we can understand how the sounds are dynamic over a day, so this is just looking
at a single day, recording midnight, and just going through the time.
I'm going to show some spectograms, just recording the first 7 seconds, and then switch to the
next one.
I just want you to see the dynamics of how these work, so this is a spectogram of the
sound, going over time.
And so, you probably can't read this but we're at 1:00 in the morning, 1:30, and you can
see changes, you can hear the crickets and some of the frogs croaking.
Maybe I'll try and speed this along... and so it's sort of a fascinating view of, drag this along, and
here is the transition I was talking about, going from 4:00 to 4:30.
You can see the soundscape just changes dramatically in that time because of the beginning of the
dawn chorus.
So these are just some examples of the different types of sounds.
This is taken, this is what I do with select information from this Real Library.
You can search for particular frequency values, or you can pick out sensors, or you can pick
the year, the month, the day.
The time of day, the sensor unit and so-on.
And then you can just query those and it will just pull that information up in unique subsets.
So here's an example of sounds.
That happens to be at 6:00 in the morning.
Here's one at 5:00.
So a lot of dynamics going on.
You probably recognise...
So I got very interested in these patterns and how these patterns change over time, and
I'm looking at the patterns, not from the species standpoint about who's in the recording,
but actually from the amount of energy coming out of the ecosystem.
So it's really, my perspective is kinda from an energy perspective.
And as you'll hear from Jason and others, Mike I think too, we'll talk about some of
the species identification from that.
I'm looking at the amount of energy, so this little bar graph here is the idea of looking
at the amount of energy in each one of the different frequency levels, and you can see
right here, between 4 and 5 kHz, that's where most of the energy is in this particular sample
of the soundscape.
With a nice, you know, song, this is represent in this low frequency area here, you can see
as we go up you can see different levels.
But most of the energy is occuring right around this area right here.
You can actually see the sound scape, "bip, bip, bip, bip" that's those little dots right
there, notice that? Who knows what that animal is? Anybody?
Michelle's koala isn't it?
It's a koala. (laughter)
So onto the library then, we put information into this digital library and these are the
recordings, these are the number of recordings, from each one of these sensor types in the
library, 50,000 recordings.
Amounting to almost 1/2 a terabyte of observations, and this is the process that we go through
to do it.
We got to collect the data, which is just down here.
You upload the observations into a data bank, sort of temporary storage spot, then we compute
some indicies, and that happens up here.
Then we archive those and then we put them into the acoustic library.
And then, we can access the information in a number of different ways: archive access,
analyse, and then interpret.
And then the observations can be interpreted on the web in real time, if you want, or download
them and put them into a program like SAS, or SPSS or MiniTab, or whatever you're happy
or familiar with in terms of statistical analysis.
So what I was quite interested in taking a look at for this particular presentation is,
what about the time of day when the soundscape changes its energy profile.
What time of the year does most of the energy occur in, what are the dominant sound frequencies
that are happening in the soundscape, and then what are the energy sources.
And I'm gonna leave the energy sources up to Jason.
He'll tell you about his amazing project to actually identify the sources of the soundscape.
So, I got interested in some of these precipitous changes in the sound energy, as we go from
dawn till dusk or from night till dawn, to look at some of those changes.
Looking at the night vs the day sound energy shifts and then looking at the total energy
pattern for patterns.
So these are some question that I don't think I'm going to answer, but perhaps they could
lead me to some interesting discussions some time.
But what's going on in an ecosystem? What is the meaning of sound, of sound energy output
relating to ecosystem function? A lot of people will approach those of us interested in sound
and say,
"well you know the organism that I study doesn't make any noise so why should I bother and
be interested in sound?"
Well the point is that there are trophic interactions, so if you hear a butcher bird, there's meaning.
That butcher bird's working - it's doing something.
What do butcher birds do? Well they're predators and they're busy preying on the species that
they work on and so there are some translations.
Even though we don't know exactly what they're eating, we know that there are other animals
in the ecosystem because the butcher bird is there.
And you can make that interpretation in terms of many many different animals, whether it's
sulphur crested cockatoos or the scarlet king fisher.
Scarlet king fisher? Sacred king fisher.
Scared king fisher
(laughter) Scared king fisher.
Ok so you can then make inferences about the ecosystem function by the sounds that occur.
So a long time ago, Mike Towsey and I had lots of conversations about how we build metrics
to try and compute the amount of energy in the soundscape.
I came upon this application using what's called a power spectral density analysis that
was derived from a person who studied sound in a way to come up with pulling the amount
of energy out of sound.
Then I normalised the power spectral density to enable a comparison between, sort of within
between places.
So here is the raw power.
Here is what it looks like.
It can range from huge numbers, like this, in the low frequency to very very small numbers,
like this.
So you can have this huge range, and when you normalise it, it helps to minimise that
variance.
But lets, Jason and I were looking at this yesterday.
He said, "what is that thing up there?" What's that dot up here? Why is that noise so dominant?
Well I went and dug up the information out of the digital library.
(loud cricket sound) That's what it is.
So, in some cases there are very very loud noises that come out of the ecosystem and
in other cases there are very quiet noises.
Now here is the normalised values.
So this is looking at just a single day, sort of the raw data coming out of the soundscape.
And that's what it looks like when you normalise between 0 and 1.
So all these value that I'm going to show you in the next few slides are ranging from
0 - 1.
And that's basically what this says, so I'm looking at all the different frequencies from
1 to 2 to 3 to 4, up the scale of frequency, up the pitch, to look at the comparison between
night and morning, day and evening.
I can't go into the details too much about the interpretation but I want you to just
take a look at these patterns and see what you think.
These are the 4 sites I concentrated on.
Here's again, the spectogram, looking at these different levels for going up in pitch, up
in frequency, take slices and looking at each one of these energy slices.
Here's a day.
Now these patterns, now here's the difference between these 4 sites.
Very low amount of energy in the 1-2 kilohertz range in this site, compared to this site,
compared to this site, and similar to site 3, and you can see all those changes occur.
Now here's the beginning of the changes from the evening to the dawn, through the day.
So you can see, high values to low values.
So this is looking at time of day, and this is looking at the energy.
Patterns, patterns, patterns, patterns, patterns...
We'll talk about site number 1, which is one of the sites down by that little farm and
where Peter was doing some of his gas monitoring.
Collected a fair number of sounds from each one of those, from that site for periods of
months so I averaged these values over time, to look at changes in the soundscape.
So this is just a little premer so the frequency values here range from 0 - 1.
Here are some day sounds, you can see right here.
No sound or not much energy to higher amounts and then down again.
These are the dawn sounds, increases in energy and then decline as evening progresses.
And here's some night sounds, high values here, and then dropping off during the day,
and increasing again, and then rising again till dusk occurs, and the crickets begin to
increase their sound.
So you can see some of these patterns.
Right here - this is like total energy, and this is an index used over time to try and
look at the ratio of biological sound to technilogical sound.
Let's listen to a few of these.
This is midnight. (crickets and frogs sound)
So if you go out to Samford and sit out in the dark and listen.
And if you do that, about now, this is probably what you're gonna hear.
If we look at 6:00 in the morning, crows at least have woken up , busy doing their business.
You still hear some buzzing insects and others.
By noon the cicadas have started coming online.
This is in the fairly distant background, but you can see the patterns of the cicadas.
A little bit of wind.
And then, here's another one, interesting one (kids sound).
Who's that?
Jason's kids
Is that Jason's (laughter).
Jason and his kids.
That was the sensor near the old farmstead, so I imagine they were having a party, invited
the kids over for a little soccer game or whatever.
But again you can see the sounds of the log that humans have sounds in the lower frequency
bands here.
These little flecks are kid sounds and then we got crickets and other animals sitting
there.
So I'm not going to make a lot of interpretation about these patterns, just look at these changes.
But what I was quite interested in is the consistancy of some of these patterns over
time.
Really the amazing changes in these energies as we go over time with these different frequency
levels.
So you can see nighttime, at the lower this is 3-4 kilohertz, lots of sound at night and
disappearing and then increasing again.
Here we have very low sounds at night.
This is at the 5-6 frequency.
This is nearly in the bird range, so we won't expect to see a lot of high frequency bird
sounds occuring at night.
As soon as the dawn chorus begins to, the birds begin to wake up around 4:00 in the
morning, sounds start to increase and then they stay and disapate.
So these are all the different frequency levels, and I'm gonna walk through November through
to May so you just see these changes.
Here's December.
Lots of night sounds.
Not much going on in the day time.
Lots of base sounds with chattering at night depending on which frequency band you look
at.
January, February, March...
Now these averages, these are averages of all of the 6:00 in the morning for the 30
days, so that's what these little bars are here, so they're averaged at that particular
time, so every 1/2 hour we're averaging across the month.
And I didn't know what period to pick, so I thought a month would be reasonable, just
to kind of get the changes in the seasonal dynamics.
April, May and so on.
And then I took the different sensors, the eight sensors, and I did the same thing.
Look at Sensor number 1, 2, 3, 4, 5, 6, and 7...
Now they're all quite different.
They're all reflecting what that habitat is like, and that's why I've been so keen to
do vegetation analysis out at Samford.
So that we actually know: what vegetation type, what habitat type is there and how do
we classify those habitat types so we can coralate these sound characteristics to habitat
type.
And then this is just a bar chart series, showing the different sensors, different locations,
for each one of these different frequencies.
And you can see some, there's some: this one happens to be sensor number 6, which is dominant
in the low frequency.
Here's sensor number 5, dominant in the very high frequencies.
So you can see differences between the different types of sensors.
So I started getting more statistical in my thinking about how to interpret this and I
came up with an actually subsampling sounds within these different frequencies.
And I don't think I'll go through any particular interpretation except to say there's lots
of work to be done in terms of the quantification and how we interpret these patterns that we
see.
Certainly there's a wonderful opportunity to do time series analysis with all of this.
I've got people interested in how we... (turns off phone).
How do we interpret these patterns looking at the temporal dynamics? So time series - it's
an eventuary.
So there is a lot of changes in the soundscape.
There are major acoustic energy shifts.
There are acoustic patterns that are dynamic, and the patterns makes sense to me.
I can see them.
I can interpret them.
I know what's going on and one of the things I'm really anxious to do now is to merge these
energy patterns with the kinds of work that Jason's been doing to try and come up with
the actual identification of who is in the soundscape.
So we have lots of work to do.
I've been relatively busy this year.
Probably had one of my better years out of all the years that I've been professor, after
I retired and started publishing.
But we got some, this is a PhD student that recently graduated who did a study of the
interpretation of the variability of the soundscape in urban areas, trying to look at what does
the cityscape look like, and how does that vary over time.
This is a study of ensemble extraction, trying to classify and detect the species.
This is an old study that we did in Sequoia National Park and then Bernie Krause who's
a wonderful, historic soundscape ecologist and so on.
So there's lots of things that are just starting to emerge, and I am just so excited about
what's happening here at QUT and ISR and beyond with respect to the interest there is in the
soundscape and I think that interest is going to continue as all of you working in the sound
area knows there's more and more interest in this area as time goes on.
And so I'm still very excited about this.
I want to thank Peter for allowing me to do what I enjoy, this kind of integrating biology
and technology to try and understand the soundscape in this case.
And of course Michelle and Jason, and Kristine have helped tremendously in terms of just
making my life easy getting here, and being here and having fun.
And of course Ole.
I know he did a lots of work in terms of the nasty data collection stuff, and gone out
slogging around in the rain, pullling the sensors out of the Songmeters, and uploading
the data with Jason's help.
So thanks to everybody. (Applause)