Our advice for most accidental encounters with bats discovered outside will be to leave the animal where found; do not handle it, take it in, or attempt to care for it.
Just because you noticed a bat roosting someplace "odd" does not necessarily mean it is injured. It may simply be waiting out and unanticipated cold night or some other unusual circumstance. Do not "take the bat in" as you may well be doing -far- more harm than good for a perfectly healthy animal!
Please note that grounded bats should never be rescued with bare hands. See this Locate a Rescuer page for information on how to safely contain bats in trouble, as well as a state-by-state list of bat rescuers.
"In Stock" items may ship same day to up to 3 business days after placing your order. Some items are already in boxes ready to ship where other items may require packing lead time depending on sales and workload.
Note: Please see the item listing for the best estimate on shipping times.
"Out of Stock" items may still be purchased, but realize these items may be built-to-order or we may be at the mercy of our vendors for parts of your order or product. Shipping for out of stock items may be days or weeks after placing your order. Some speciaity items may have estimated availability mentioned in their product description page.
All international orders must be paid in fullvia credit card or international money order before shipping. We currently accept American Express, Visa, MasterCard and bank/wire transfers.
For international orders, any duties, tariffs or broker fees are not included in the order amount and will be collected by the shipper upon delivery. Any such fees are the responsibility of the customer.
Please ensure the shipping address you provide can accept UPS deliveries. BCM is not responsible for any fees or delays incurred from an incorrect shipping address.
In Spanish: Todos los pedidos internacionales deben pagarse en su totalidad mediante tarjeta de crédito o giro postal internacional antes de su envío. aceptamos American Express, Visa y Mastercard. No somos capaces de aceptar banco.
Para los pedidos internacionales,Â los impuestos, aranceles u honorarios de agente no están incluidos en el orden cantidad y serán recogidos por el cargador a la entrega. estas matrículas son responsabilidad del cliente.
Por favor garantizar la dirección del envío que usted proporcione puede aceptar envíos ups. bcm no se hace responsable por el pago de honorarios o retrasos de una incorrecta dirección de envío.
We can only replace the exact item and version you ordered, and you are responsible to return the damaged item within 30 days. If you want to exchange for a different item, you will need to return the item and place a new order. Shipping is not refundable.
All detectors are tested by BCM staff before shipping to customers. Please contact BCM first detailing the issue you are having with the hardware and obtain a Return Authorization. Should email or phone troubleshooting fail to resolve the issue, you should send the unit in suitable packaging to BCM, with the RA number written on the packaging.
BCM will confirm the unit is defective was not abused or water damaged. BCM will replace the unit at our discretion usually within one week. Units requiring repairs may take 6-8 weeks.
If you are sure your detector needs service, it is best to send the device directly to Pettersson by following the directions here: https://batsound.com/service/
Shipping the unit direct to Pettersson will greatly speed up servicing your unit and will be at a lower cost to you.
All microphones are tested by BCM staff before shipping to customers. Please contact BCM first detailing the issue you are having with the Microphone and obtain a Return Authorization. Should email or phone troubleshooting fail to resolve the issue, you should send the unit in suitable packaging to BCM, with the RA number written on the packaging.
DOA microphones are replaced immediately at no charge. There is a $150 for refurbishing used microphones, shipped immediately if in stock.
Electronics purchased from BCM are easily returned within 30 days of receipt of shipment. These items must be in new condition with original packaging and accessories. All bat detectors sold by BCM are supported by the device's manufacturers. In the unlikely event a detector appears to be defective, please contact the customer support network for the manufacturer to have it serviced under warranty. Pettersson D500x devices have specific policies, see above. Shipping is not refundable.
Software titles purchased from BCM are easily returned within 30 days of receipt of shipment. These items must be unopened and still in their plastic wrap. Digital downloads with software keys are sold as-is and cannot be returned or refunded.
Other items purchased from BCM can only be returned if:
The return is packaged and shipped in the original packaging.
The items are still in new condition.
You ship your return within 30 days of receipt of the original shipment.
Shirts, hats, and mugs are made on demand and are not returnable.
Any returns that do not meet the above qualifications may receive a reduced refund or be returned to the purchaser. Some items may incur a restocking fee as these items must be inspected, tested, and repackaged. If a restocking fee applies it will be noted on the product information page.
Just because you noticed a bat roosting someplace "odd" does not necessarily mean it is injured. It may simply be waiting out and unanticipated cold night or some other unusual circumstance. Do not "take the bat in" as you may well be doing -far- more harm than good for a perfectly healthy animal!
Please note that grounded bats should never be rescued with bare hands. See this Locate a Rescuer page for information on how to safely contain bats in trouble, as well as a state-by-state list of bat rescuers.
Because bats need an adequate
amount of warmth during the day, solar exposure is very important when setting
up a bat house. However bats can also get too warm, so coordinating bat house
color with your geographic location can aid in having a successful bat house.
Refer to the temperature zone map located on our Four-Chamber bat house page.
The wood thickness has
something to do with it, how many screws are used to hold it all
together, and how well the bat houses have been finished (how many coats of
paint used, etc.). So, material is actually not a big player in this in my
opinion. True cedar may be more insect resistant, but either box will probably
fall apart from normal weathering long before insect damage comes into play.
BCM pre-scratches all
interior surfaces of all kits in multiple directions, so bats have the maximum
ability to hold on. We do not use screen that bats can become trapped under.
There are some conflicting observations about the usefulness of pup catchers. PA Biologists observed pups never being recovered on pup catchers at all, while others have observed pups climbing back into the bat house. It probably depends how old the pups are, and the reason behind them being out.
The two pieces of wood are “pup ledges”. Nursing females will leave their young in the roost for short periods to forage for insects. Although nursing pups rarely loose their grip, it does sometimes happen when they awkwardly move around within the bat house during this time or are knocked loose by other bats in the colony. The pieces of wood act as a small ledge on either side of the baffle to give the pups something to grab onto if they lose their grip at the top of the bat house. Please refer to the PDF assembly manual for instructions about attaching the Pup Ledges.
We offer another product called a Pup Catcher as an accessory if you observe a chronic pup issue. This is much larger, more robust "basket" intended to go below the entire bat house. Pup Ledges in contrast are small wooden ledges inside the bat house between certain baffles.
As long as enough limbs are trimmed off so that the bat house receives direct sun most of the day (especially in the morning), that snag would be just fine. Mounting a bat house is a little trickier because the tree is round and uneven. If using our pole mount kit or something similar, attach the pole mount brackets straight and flush with each other on a flat surface using a few pieces of scrap wood. The mounts can then be lag-bolted onto the tree, and the scrap wood discarded. The two brackets should then be on the tree in the proper spacing and position for the bat house. In addition, stainless steel lag bolts should not corrode and kill the tree like other materials.
The Phoenix area is rather extreme for bat roosts.
You might try bat houses that are completely shaded. At the desert museum in Tucson, they have bats in bat houses that are under a pavilion/covered garage. Completely shaded attached to a brick wall, under a garage roof.
You might also try putting out hummingbird feeders. You will probably get nectar -feeding bats such as Mexican long-tounged bats, which are quite fascinating to watch as they hover for a split second while sipping. Normally they are pollenating agave and following the bloom cycle, but they will happily take to the feeders when in your area.
If you understand the principals of a modern bat house… (it needs to be wide, tall, airtight roof, and have a combination of 3/4’’ and 1’’ crevices), you can take our plans and just expand or build your own. To do 6’ wide bat house, consider screwing a full sheet of plywood to the side of a wood or brick structure, and that’s an instant single chamber bat house (add 3/4’’ spacers to precent warping and seal the top) and could be painted dark and even used as signage.
For a free -standing bat roost 6’ wide, you’ll want to incorporate a 4x6 post on either end for support.
Sorry I do not have plans for a structure with 2 posts. Some people have simply put 2 posts in, framed the top, and hung bat houses on both sides, so it’s modular.
I don’t have time to learn all of the calls via a class or anything. I need whatever it takes to have our calls analyzed and nailed down. Whatever that takes is what I need. End of story.
There just is no system that you can order that can provide a complete certainty and process everything for you, but SonoBat comes closest to that ideal than any other system. Bats have considerable plasticity in the calls they produce and within each species' repertoire there can be some data space that overlaps in characteristics with parts form the repertoire of another species. Definitive species recognition relies upon the unique subset of each species' repertoire. The acoustic table that you received provides some guidance for that. This also means that no one can identify each and every recording; for reasons of call plasticity and call quality. Noise and distance from the microphone greatly affect signal quality and classification in the same way that distance and backlight can make it difficult to identify birds. Unfortunately that's the nature of the beast. It's not so different than most other wildlife monitoring methodologies in that it requires some level of expertise.
Even with the high quality data on which the US west classifier was built (11,000 call samples), species classification rates range from just below 90% to 100%, with acceptable classifications (those that come out with an acceptable level of confidence, e.g., in discriminating data space) range from 27% to 97%. So statistically, even with a 98% correct rate (and that's ideal- for good data), the overlapping characteristics of calls will bring up some misclassifications for any substantial data set, particularly if recorded under less than ideal conditions.
But, that is the state of the art, and most agencies that require this sort of survey work understand that, but do expect best possible practice.
Understand this is related to a combination of factors, including microphone sensitivity and the frequency, amplitude and distance of the bat call. A further large part of the complicating factors not on your list involves the orientation of the bat toward the microphone. Bats emit their vocalizations with most of the energy directed forward. So the orientation of the bat relative to the microphone will profoundly affect the effective distance of perception. So, even if you could get some data on the other factors, the position of the bat will throw in additional stochasticity.
If you really need to get a handle on this, I suggest an empirical approach. Do some recording when you can visually track and assess bat positions and vectors and develop some calibrations.
You can certainly detect the lasiurines a couple of hundred feet away at least, especially when they are calling towards the detector. Some whispering bats like the northern myotis, 50' is lucky, but they won't be up at balloon height. It would be exciting if you detected free tailed bats and grey bats that -would- be at balloon height!
Of course then there is a subjective assessment of "how good" of a bat pass do you have for excellent chances of a species ID. I suggest you need about a dozen pulses including one that exhibits a harmonic in the recording. Most of the files you record don't meet that criteria.
The higher above the ground, the better. Consider that the microphone picks up good recordings from bats up to say, 20 ft. Then you might as well get the mic up 20 ft to pick up the bats 20 ft below it, and 20 ft above it. That way you'll cover more volume of flight space and minimize ground echoes.
See the Acoustic Microphone Mast for a lightweight, telescoping solution that goes 20': https://batmanagement.com/collections/detector-and-acoustic-survey-accessories
BCM's Single High mist net poles are heavier duty and come in 3' sections:
In a perfect world detectors would have logic to recognize when a bat pass begins and ends and to save individual files that cover each bat pass. So in this imperfect world we compromise with a file length that covers most pass durations. One second is far too short and will distribute a single pass into multiple files. By convenience we like our passes in single files as that lets us count activity (number of passes) as number of files. In my experience, 4 seconds provides a good compromise for recordings near the ground, and 5 seconds for high up in the sky (met tower heights and such) for open air foraging bats.
We deploy a couple different brand detectors in the desert SW many weeks (months?) out of the year. The main trick is to be sure the detector case is shaded; stick it in a bush, put some trash wood on it, or worst case bring along something to cover it. Some have put it in a hole. If it really must be exposed, you should probably drill holes in the ammo box enclosure and spray paint it white. So far we’ve alway been able to microsite it somewhere with the case is shaded, and once found a discarded barrel or something we sheltered one in.
As for rain, all microphones left for long term should be mounted horizontal, and raised up into the airspace as high as practical. This usually is a 20’ pole lashed to a ~15’ tall shrub of some sort for support. If you have no suitable shrub to use as a pole support, a pole is slid over rebar then tied off in 3 directions will do.
For a Pettersson D500x, plan to service the devices about once every 8 days with the C batteries installed, though you may get twice that with the external D pack we also have. You really don’t want to leave any detector more than 10-14 days as often other issues arise with mounting, etc.
All of them! Acoustic monitoring is becoming more popular with citizen science because of advances in technology and the fact that you don’t need a permit. I do have two favorites I find most useful. The BatScanner Stereo is my favorite “simple” bat detector that gives me an idea of what species guild is passing by, and with headphones you can actually tell what direction the bat is coming from, which will rock your world. It is also quite robust and won’t break when kids drop it. https://batmanagement.com/collections/bat-detector-buyers-guide-active-detectors/products/elekon-batscanner-stereo
The professional bat detector is without a doubt SonoBatLIVE. https://batmanagement.com/products/sonobatlive This is a bat capture app that runs on a Windows 10 device (a ~$400 Surface Pro). The software is designed to properly organize files for long term monitoring, record only bat calls, and also assign species ID to the bat calls, with very little post processing hassle as compared to other available workflows. While SonoBatLIVE records and IDs bats consistantly better than any other system we've tested, to manually review your recordings you'll need the full complete SonoBat software package. Besides this software a microphone is needed, and fortunately one of the best ones is also the least expensive… the Pettersson M500 https://batmanagement.com/collections/pettersson-collection/products/pettersson-m500-microphone
You still need to use a hacked cable that disables the Heterodyne channel, and it is easiest to use it in continuous record mode. Note the time when you begin recording, and then parse out the triggered events in post processing with the SonoBat AutoParser, a legacy utility that is included with SonoBat.
You may run into trouble though depending on the version of Windows that you have. iriver ended support for the file management software that you need to offload the files from the IFP units. You may need to do that on a legacy computer running an older version of Windows.
As an alternative, you could keep your D240x units going with a Samson Zoom recorder. Or a Tascam DR-08 recorder. With the DR-08, (or with the Zoom), to avoid the signal level trickery to enable triggered recording, just record continuously and then use the SonoBat AutoParser to parse out the bat pass events as separate files during post processing. More info in the Guidance doc for SonoBat utilities:
These are legacy devices that was one of the few ways to record FS before specifically designed long term monitoring bat detectors existed, so we preserve these methods as they may be helpful to someone finding this equipment and would like to get it working.
That provides setup info fro the D240x, and how to set up Samson Zoom H2 recorders to work with the D240x. Those instructions provide for a single night of recording, but if you want you can take it further. The Zooms apparently accommodate larger CF cards than they say. I have used 8 GB cards fine, and I suspect they would accept 16 GB, although 8 would likely get you through two weeks. You would need to supply external power to the Zoom, and it has a jack for that- 9V (peculiar, as it runs off two batteries = 3V, so I suspect some tolerance with it's voltage supply); note polarity. You can power the D240x directly from a 12 V supply (Pettersson-approved). Attach the power supply to the D240x using a 9V battery clip, making sure to get the correct polarity as it would come from a single 9V battery.
You can not schedule the Zoom or D240x to cycle on or off. However, we have accomplished that using programmable lawn irrigation timers. These take a single 9V battery that lasts for months. Take the signal from the timer and use that to control a relay to open and close the power to the D240x. When off during the day, the Zoom gets nothing to record, and keeping power on the Zoom maintains its clock setting. When power returns to the D240x it goes right back to operating as set.
For an alternative, higher tech approach, you can use Binary Acoustic Technology FR125 programmable recording units. They can be set up to trigger from D240x units.
Digital recording from the D240x amounts to something of a hack because these older device were not set up to do this. If you have a large project ahead of you I would recommend a D500x for realtime recording (no wait from time expansion data offloading) better microphone deployment, weatherproofing, data management, and longer term automated field recording.
I have not used the H2next, but I presume it should work like the earlier model. Just find the mode that accepts a single microphone. That will probably be a 2 channel mode, for right and left stereo. Zooms record in stereo, although you want the right channel for the full-spectrum signal that comes out of the D240x. Also follow the cable recommendations in this PPT. Actually, you may not have to make a custom cable if you use the method below, but be sure to specify the right channel when using the AutoParser.
Rather than futzing with the autorecord and trigger settings, I suggest using an 8 or 16 GB memory card (enough for one night), and just begin a recording and let it run continuously during your evening or night recording session (Note the time that you start the recording). Then use the AutoParser to parse out the bat pass events as separate files during post processing. More info in the Guidance doc for SonoBat utilities: http://www.sonobat.com/download/SonoBatUtilities.pdf
Unfortunately, you would find disappointing results using an older D240x for a mobile survey. We ran some trials at Lava Beds NP, alongside a D500x with remote microphones. You can readily get the D240x attached to a pole and rigged above a vehicle, but the delay from offloading data by time expansion makes it unfeasible. Mobile surveys tend to have a lot more noise that trigger recordings. That does not affect a real time recording device like a D500x as it can get right back to recording after a noise-triggered event, but with the D240x that spends ten times as long offloading the data, the D240x just misses too much on a mobile transect.
At that setting you cannot resolve any frequency content above 96 kHz. This is very important for certain species, such as Myse. We suggest a sampling rate of at least 250 kHz or higher to increase your frequency resolution to 150 kHz. This is a mathematical reality in signal processing. See https://en.wikipedia.org/wiki/Nyquist_rate and related articles.
What happens when a frequency above the Nyquist frequency is sampled and played back? Do these frequencies simply disappear?
Unfortunately not. Frequencies above the Nyquist frequency cause aliasing (also called foldover or biasing). The spurious frequencies they produce are predictable, in that they are mirrored the same distance below the Nyquist frequency as the originals were above it, at the original amplitudes.
Also think of it this way: if you keep you eyes closed, and only open them once a second to peek at a swinging pendulum, if it always appears on the upper part of its swing to the right, you would conclude that the pendulum swings once per second, even though it may swing twice that rate but you don't see the extra swing when you have your eyes closed. Your visual sampling technique aliased the pendulum to make it appear to swing at a lower rate.
Digital sampling takes snapshots of a signal at regular intervals, say 300,000 or 500,000 times a second. But it can't "see" what happens between those snapshots.
Yes, sampling at 500 kHz will produce less aliasing of real frequency content than 300 kHz, but will not prevent artifactual alias harmonics from signal overloading.
I have a 18” parabolic dish I use for bird recordings. I was wondering if that could be adapted to enhance the signals to the SMX mic, or would the design and signal reflection diminish/warp the signals to the point there would be no advantage in using it?
If done properly it should not distort the signals. However, when you reach further through the air you will enhance the differential frequency attenuation effect of propagated ultrasound. That is, higher frequencies get more attenuated than lower frequencies. So I suspect you would detect bats at greater distances, but your rate of effective species ID may diminish from the selective loss of higher frequency content.
Wildlife acoustics offers a directional horn for focusing bat signals to the SMX mic. Have you heard how well that works? Any distortion etc.?
A glance at a graph provided by Wildlife Acoustics suggests it’s a great accessory. However, the graph is carefully titled “on axis response”. This means it works great when a bat is positioned perfectly in front of the microphone and horn, which in reality, it never is. In real life use, the horn attachment generally leads to fewer files being able to be classified.
Two polar plots showing the frequency response of a Wildlife Acoustics SMX U1 microphone, without any special attachment, and then with the directional horn.
The "on axis response" graph that looked great only applies when the bat is directly in front of the microphone, which is almost never is. 20° on either side of this mic and you’ve lost 20 dB, and it gets worse from there.
The manufacturer has tested
CF card sizes up to 32GB. However, the Pettersson D500x has a total of (4)
storage slots, allowing a total of 128GB of combined storage. Things like
expected nightly bat activity, insect noise, recording length, and length of
time between detector service checks are just a few of the main things to
consider when it comes to making sure you have enough storage. As a general
rule, try to stick with a single 16GB card for deployments lasting a few nights or a 32GB card for a week. You’ll have more flexibility in
terms of when you’ll need to offload your data, and are usually large enough to
accommodate a few busy nights of bat activity. One large card is better than 4 tiny cards; one large card will be faster to post-process.
The external mic is essentially weatherproof so long as you do not point it up so that it can collect water. So pointing horizontally should be fine. The only problem with vertical would be that you can't record through the water- there is less concern that the water would hurt anything. The directional horn can be removed and would allow water to not collect in the vertical position.
The connection to the cable should have some protection from soaking rain, but I think a minimalist configuration as shown in the attached will serve to protect and clear the airspace around the microphone. I made this one from 1/16" x 1" aluminum angle. I bent the angle at about a third the length to give me the option of mounting with a shorter or longer horizontal reach. You can get releasable cable ties to attach the mic to the angle.
And BCM has been shipping mic protector tubes for some time now. These cleverly slip over a D500x mic and have a cap to minimize dust during storage, then the tube is reversed for actual deployment.
Begin by charging your
external battery. While your battery is charging it is a good idea to check to
make sure your detector will power up when using internal batteries. (C or AA
depending on the year your D500x was made). If your detector powers
up, the problem most likely lies in the external battery. Once charged, test
your detector to see if it will power on when using the newly charged external
battery. Also, we recommend having a
multi-meter on hand in order to test the voltage of your external batteries.
Often times if a SLA (Sealed lead-acid) 6V battery has been left alone for too
long without being charged, it may have lost is ability to power your device
without further maintenance. Occasionally, a neglected battery can be
“resurrected” by means of a trickle charger used for small vehicle batteries.
With regular use and or/maintenance, a standard 6V battery charger such as the
Battery Tender brand offered by BCM is more than sufficient for keeping your
batteries in top shape.
This is highly dependent upon
your record settings! Things like recording length, high insect volume, display
mode, and trigger level are all things that can contribute to increased battery
usage. Consider that the more often your detector wakes up to record, the more
battery power is being used. How much mileage you get out of a deployment will
also be influenced by whether or not you are using internal or external
batteries, or solar array for power.
However, it would be foolish
to leave -any- bat detector unattended for more than week or so, as other
factors such as spiders in the mic, cables being chewed on, etc. will disrupt
All that aside, the
"AA" batteries in the origional D500X last about 2-3 nights. The
"C" batteries last for about 5-7 days in real life project use.
External battery packs using "D" cells can last 10-14 days, perhaps
more. External 20 aHr batteries can last a month or more. Adding a small solar
panel would allow it to run indefinitely.
There are currently 10
different “USER” profiles (USER0 – USER09) that you may customize/pre-program
in order to save different recording settings.
There are also 10 different “PROFILE” (PROFILE0 – PROFILE9) that may be
selected as well. However the recording settings within these profiles
(PROFILE0 – PROFILE9) are predefined and cannot be changed like the “USER”
profiles can be. Example: If you are customizing your recording settings for
“USER3” under “F1>USER PROFILES”, be sure that you have “USER3” selected
before deploying your detector if you would like to use the settings you just
While on the home screen of
the detector (the first line on the home screen will indicate which user or
profile is active. Ex: “USER01”), press the ENTER key. A new screen will appear
that lists the “USER” or “PROFILE” number at the top, followed by a short list
of settings. Hitting the left or right arrow button will allow you to navigate
to different user or profile. Hit the ENTER key again to select, and you will
be brought back to the home screen.
Because the D500x is designed
for bats, there are not many settings that the user needs to worry about. Edit the USER0 profile to reflect these settings:
500 (if you desire the
highest quality recordings and disk space is not an issue) or 300 (when disk
space is a consideration)
LENGTH: 4 seconds
High Pass (15kHz rolloff)
filter: Y (or N if you expect bats below 15 kHz)
TRIGGER SENSITIVITY: LOW (use
very low if you want only the in range bat passes. Use Medium/High/Very High if
you want to record more out of range bat passes. The out of range passes will
not be classified automatically by any software and are arguably of no use,
since bat detectors cannot be used to provide numbers of individuals.)
Save this profile and be sure
the D500x is set to the correct USER0 profile on the "ready" screen.
Press ENTER and then the L/R arrow keys to change profiles. Note that USER0 is
completely different from PROFILE0.
The final RECORDING SETTINGS
(when you push the REC button to start the session) could be:
GAIN: 45 (possibly use
60/70/80 for whispering bats)
TRIGGER LEVEL: 160 (possibly
use 120 in very quiet areas, like deserts)
INTERVAL: 0 (unless possibly
you are in a high bat traffic area)Don't forget to press ENTER to start the
recording session. The D500x will not turn off and on according to your TIMER
Set up a USER1 profile for MANUAL recordings. (We use USER0 for the passive recordings)
For those options:
SAMPLING FREQUENCY: 500 (because if you are bothering to manually record, might as well be the
highest sample rate)
Length=MAN (you may have to first set Autorecord=NO before you can set this to
HP=YES (if you are somewhere)
TSENSE: LOW (doesn’t matter, not used in this application))
Now set your gain to whatever you want (45 is good) by tapping the L-R arrow
buttons on the “ready” screen.
Switch user profiles by tapping ENTER, then the L-R arrow keys to USER1
Now when you press REC, there is about a 1 second pause, then you are recording
full spectrum until you press the REC button again. Remember SonoBat likes
files less than 30 seconds, and more like 5-10 seconds are much better
(however, you can simply cut up the wav file later using the “save zoomed
section” in SonoBat' realtime view). I do try to count silently to
myself and limit to under 10 seconds in the field in order to reduce post-processing time.
After a short pause, you can hit REC again and repeat as many times as the bat
is in view.
If you have the D500 with the headphone jack, get headphones with volume
control, and you can hear the FD output in real time and know if the bat is in
range for recording.
If you don’t have the headphone jack version, just rubber band a heterodyne
detector (I suggest the Batscanner Stereo and you’ll be able to tell if your
target is left or right) to the D500 and use it to listen for bats in real
time, then you know when to trigger the D500 for manual rec.
Finally, note that the file count on screen is the LAST file recorded. So, if
you let a bat go and made 3 recordings of it flying around, I rubber band a
notepad to the back of the D500 and note that “M0003, M0004, M0005 = MYSE” etc.
etc. This is very important and makes the post processing matching up as easy
as it can be. DO NOT write down any mistakes, that is, if M0006 was a
dud, I don’t write it down, and just delete those files later, only keeping the
ones I specifically noted. Maybe you want to note those duds…it can be a little
confusing late at night especially if you hand the detector to someone else to
take over the note taking and they don’t have a clear idea how the D500 is
showing you the last file recorded onscreen, etc.
In this scenario, the error message that people often see is a
variation of “CF* NOT READY”. This generally happens under the following
combined circumstances: Two or more CF cards are being used in the detector.
The detector batteries died during deployment. CF cards were pulled before
being turned on again. A single CF card is inserted back into the detector and
the detector is powered back on. The reason that this combination of events may
cause such an error is that since the detector lost power while it was in
“Record” mode and is now being powered back on, it is trying to resume where it
left off when it initially lost power. Since you powered the detector on again
with only one CF card, and the detector lost power while there were multiple
cards inserted, the unit is expecting to see those additional cards. To solve
this error, turn off your detector, replace the number of CF cards that were
used during the previous deployment, and power the device back on. The
error should be gone, and you can now “wake” your detector by holding the power
button to resume use.
The Pettersson D500x uses
either AA or C type batteries, depending on the manufacture date.
Unfortunately, due to the high number of variables (record time and length,
sensitivity settings, screen brightness, nightly bat activity, insect noise, humidity,
temperature, etc.), it’s difficult to give an all-inclusive estimate of battery
life. Due to these factors, this timeframe can range anywhere from a few days,
to well over a week. However, we generally recommend checking your detector at
regular intervals to keep an eye on remaining battery life. Advice that we find most useful: Don’t leave
your detector unattended any longer than you can afford to lose data!
The d500x was possibly improperly shut down with more than one CF card in it. Try booting
it with 1, 2,3 then 4 cards inserted. When you have the right number that was
inserted before, it will boot up and still be in record mode.
batteries died in record mode, it will still be in record mode looking for the
same number of cards that was in it before. Always wake up the detector
and get it out of record mode by holding down the soft power button on the
keypad. Only then turn off the hard switch under the cover. Otherwise the D500x
is always in record mode and may confuse the next user!
Be sure the cards are
put in the right order: top left, bottom left, top right, bottom right, in that
order. Use one large card instead of many small cards to reduce your post
Similar to the question
regarding expected battery life, it’s highly dependent upon how often your
detector is recording. If one night’s deployment yields 2000+ recordings on
Detector A, and Detector B records only 100 files, it becomes clear that
Detector A is using more power to make all of those recordings, resulting in an
increased demand for battery power and ultimately, less time before you will
need to swap out your batteries.
Not necessarily. First
determine that the microphone cable is securely attached to the detector.
Likewise, ensure that the external microphone is fully connected to the cable.
If this has been confirmed and your D500x is still not picking up sound, detach
the microphone from the cable and plug it directly into the detector. If you
can suddenly pick up sound, this would be indicative of a bad microphone cable.
If the problem still persists, it may be your microphone or the mic port on
your detector. If you have a spare functioning microphone, you can determine
whether or not it is the microphone port. If you suspect that this is the
problem, it may be time to send your detector in for repair.
I would consider 100m to be
effectively out of range of the D500X, though you may still document the
presence of "some bats" at that range. 50m will let you identify
high-flying "loud" bats like Hoary, Red, Free-Tailed, Silver-Hairs and
Of the most significant
variables, species is at the top of the list. Recording in a wide open meadow
with zero clutter heavily biases your sampling toward the above
"loud" bats. Next on that significance
list is your detector settings: notably gain and trigger level. If you want to
"reach out", crank up the gain to maximum, and lower trigger level
until it is just barely above ambient noise level.
However, if you tailor
recording settings to maximize your detection distance, you will most
definitely lower your ability to classify those recordings. Ultrasound
attenuates through the atmosphere at a different rate for different
frequencies. The farther away you are from the bat, the less bandwidth and call
detail you receive; you may be able to say that you've detected "a
bat", but may not be able to say anything more than that. Plus, if you've
cranked the gain, you will overload those rare perfect bat passes which are
actually near your mic.
I suggest putting the mic on a pole, then wedge the pole between the driver-side rear door and the back seat of your vehicle. The mic should be 2’ or higher above the roof of the vehicle…not flush or sitting on top of the vehicle. You might need to use small rope to secure the pole to a luggage rack, etc. Angle the mic approximately 45°; direction probably doesn’t matter. Remove the D500x directional cone if desired. You may get better recordings with it on, but fewer recordings overall. (BTW, never, ever, use the Wildlife Acoustics directional horn on -any- microphone.)
The slower you drive your transect the more recordings you will make, and the better the recordings will be. If you are not committed to a protocol I would avoid speeds more than 10 mph. Also, point surveys -consistently- outperform mobile transects, so consider just moving to designated spots quickly and remain stationary for x-minutes, rather than constantly moving. This way you at least get usable call files for analysis rather than trashy call files that can’t be identified.
Mobile transects are probably the hardest kind of bat survey to conduct, and introduce additional layers of bias on top of an already biased survey method; good luck!
The main trick is to be sure the detector case is shaded; stick it in a bush, put some trash wood on it, or worst case bring along something to cover it. Some have put it in a hole. If it really must be exposed, you should probably drill holes in the ammo box enclosure and spray paint it white. So far we’ve alway been able to microsite it somewhere with the case is shaded, and once found a discarded barrel or something we sheltered one in.
As for rain, all microphones left for long term should be mounted horizontal, and raised up into the airspace as high as practical. This usually is a 20’ pole lashed to a ~15’ tall shrub of some sort for support. If you have no suitable shrub to use as a pole support, a pole is slid over rebar then tied off in 3 directions will do.
Plan to service the devices about once every 8 days with the C batteries installed, though you may get twice that with the external D pack we also have. You really don’t want to leave any detector more than 10-14 days as often other issues arise with mounting, etc.
Of course your best solution would involve recording away from the insects or during a season when the insects don't makes as much sound! Impractical, I know. But so consider that for longer term planning. This complication affects acoustic bat work in many parts of the world and we also suspect that it must also affect the bats, too.
As a compromise the best that you can do involves reducing the trigger sensitivity of the detector. Also, you will gain some improvement by locating the microphone as away from vegetation and the ground as you can to gain some distance from the insect sound sources and get it closer to the bats in the air that can trigger it.
Gain: 45 (80 when targeting quieter whispering bats)
Trigger level: 160
In the User Profile:
Trigger Sensitivity: very low (use low or medium in quieter environments)
Note: the somewhat awkwardly named Trigger Sensitivity setting controls the duration of the sound that triggers the detector. By setting it to very low you avoid short duration sound sources (like rain drops) from triggering the detector, but "longer" ones like 2 msec bat calls will still trigger the unit. This is the D500x onboard noise rejection control. The default settings for each level of threshold (very low, low, medium, high, very high) can be modified using a hidden menu. Contact John Chenger at BCM if you would like to tweak these setting beyond the default.
We cannot recommend going
more than “quad-high” with those poles, that is, add perhaps one more 6’ pole
section to each side. If your crew takes special attention with rigging guy
lines, you might be able to add 2 poles to each side, giving you a total height
of 36’. When the nets are collapsed, the poles will be very flimsy and will
want to bend in the middle and possibly snap. (When nets are raised, additional
guy lines would be placed on the white rings about 1/2 way up and would keep
the poles straight well enough, but when collapsed the poles would be very
All of the hardware and ropes on the 3H, 1H, and harp trap systems are perfectly fine to soak in hot water. If you use any chemicals, you’ll want to rinse it off as the metal may tarnish or corrode if left in storage with chemical residue on it.
You can adjust the white ring spacing to suit by pulling the white rope thru the white rings. Depending on how much extra rope is left over below the bottom white ring, you might start at the top and adjust all of them slightly down, or move them all up, or half-and-half. You’ll probably want needle-nose pliers to help pull the rope thru the white rings, as they are intended to not move on their own and require a little force to move.
The key features of the ATN OTS-HD 640 for bat monitoring use are:
30 Hz frame rate for smooth video of bats in flight
640x480 sensor size, which is considered high-end for a sub-$5000 device
1x lens, which is about as wide angle as these devices come. You will still need to be a fair distance from a mine entrance, for example, to capture the entire scene
Capable of external power battery pack (Thermal cameras eat AA batteries)
Records video onto a memory card (no additional hardware/cables needed)
What the device -doesn’t- do:
- Manual exposure/gain/contrast, so some situations the image is sub-par
-Automatically NUC (reset/clear the image). This thermal camera requires the user to periodically cover the lens and “clear” the image sensor (push a button) to obtain the best image quality. How frequently depends on if the temperatures in the environment are changing and the device is cooling down/warming up. For our current project in AZ, we are setting up the cameras an hour before sunset. During recording we are clearing the image about once every hour to 90 minutes.
- The file size is limited to about 1.5 hours, then the device stops recording. You must manually start a new recording, so it can’t be left unattended forever. This could be easily fixed with a firmware update if customers complain enough (it should just close one file and start another automatically)
The Cavity Peeper Camera is either white light or IR… both lights are not on the same camera; and the camera is different. The transmitter might be the same, so you might only need to buy 2 different cameras, and the receiver/monitor would be the same, so you could save a little there.
This requires using your iPhone or android device as the monitor, but if you don’t need IR, this is the way to go. The camera is not as compact as the peeper but probably bends on an angle enough to get into your entrances. FYI I noticed a software bug with these awhile ago where it wouldn’t save video unless you first took a still photo, but that may be resolved by now. Like everything else, you’ll want to make sure it works before taking it to the field for real.
I believe the bird folks use the white light with good success and almost exclusively. Probably anything you stick into a roost/nest is going to cause disturbance, wether its white or IR is moot at that point, considering all the banging around you’ll do just getting the camera into the hole. If you wanted to leave a camera in there for hours or days or weeks, you’d want IR and also some other system, because these run off batteries and are not designed for long term out of the box, though probably could be modified or rigged.
How far it can detect a bat is unclear. You will have to experiment with trigger sensitivity settings at your location and toss simulated bat around to get an idea how far from the sensor will a bat sized object trigger the camera.
Independent test have begun to show SonoBat as much more accurate and higher performing. The SonoBat classifier is built upon a larger and more thorough collection of reference files acquired by expert bat biologists carefully tracking known bats through a variety of microhabitat environments over the past 20 years.
The Kaleidoscope classifier relies on crowd sourcing of files and so has biases toward hand release type of recordings (different than free-flight bats you record), and consequent unknown quality control. Note that Kaleidoscope makes claims about its speed, not accuracy. It achieves that speed by processing all data by zero-crossing and thus forfeiting much of the fine scale information content. SonoBat uses the full information content inherent in full-spectrum data to more robustly and reliably track and interpret bat calls.
Additionally, the SonoBat classifier uses a hybrid machine learning hierarchical classifier with additional programmed redundant logic implementing what we have learned about the species (see attached). Finally, all bat classification depends upon manual confirmation of the automated results, and SonoBat's legacy of facilitating viewing calls in detail, in multiple ways, enables that process far better than any other acoustic analysis software available.
Much help information is accessed from the ? icon in the upper right corner of the SonoBat Viewer.
SonoBat also has help-contextual pop-ups available throughout the program. Activate with Ctl-h (orCommand-Shift-h on Mac) and then wherever you have your cursor positioned, you will access explanatory messages and tips.
SonoBat is intended for analyzing full-spectrum signal data. A divide by ten frequency division essentially discards all but 1 in ten parts of the data and prevents the high resolution analysis that is the foundation of the SonoBat approach. You can use SonoBat to inspect frequency division audio signals, but you will get a very smeary-looking image because you only have 10% of the signal resolution of a full-spectrum signal.
So the answer is yes, you can use SonoBat with frequency division, but how well you could use that information for bat identification will be limited, you will have a lack confidence for anything more than sorting of the most basic call characteristics, and you will not have access to the subtle discriminations that SonoBat can support with full-spectrum data.
To interpret recordings and make classification decisions, SonoBat needs quantitative data from the calls and sequence. This process entails the most processor-intensive part of the procedures. SonoBat must recognize the individual calls in the recording, extra file snippets of the calls, generate a high resolution 2048 frequency bin sonogram with 0.025 msec time resolution from which it then uses an intelligent call trending algorithm to follow the track of the call from low amplitude details through noise and distortion. Then use that trend to evaluate about a hundred time-frequency and time-amplitude measures, both discrete values and multiple form fitting shape measures of various call parts. Rather than performing those time consuming tasks to fill a vetting table each time, SonoBat saves these extracted call measurements in the metadata of each file and that enables rapid display of call information content and sequence decisions when viewing in the Vetting Table, and enables switching a regional classifier to seek potential overlap species at range edges to point to individual files for purview.
Synopsis: you must first run the files through SonoBatch to preload this data in the files to have it revealed in the Vetting Table and work with it. See also the note on the SonoBatch panel when you first launch it.
First things to check: have you already SonoBatch processed the files? The Vetting table depends upon the fine scale analysis, extraction, and loading of call parameter data into file metadata to render results. That’s the most processor and time demanding step; the scheme here is to perform that once, load the files with that data, then use that variously with the Vetting table. If recorded using SonoBatLIVE the files would already have that data embedded. (See first page of attached.)
Also check: has the Vetting scroll bar on the table become moved downward and hidden any files displayed? Sometimes drag and dropping does not register and you may need to use the navigate buttons to load files.
As a workaround fix, run them through the legacy Attributer and set the TE = 10x (select do not change filenames) and then SonoBat can properly interpret them. If you have file sets with notes in them, copy that note to the note field to have it included in the new (TE fixed) files.
The color mapping in SonoBat intends to graphically represent the trend from highest to lowest amplitude. We urge caution to depend upon it only so far in a quantitative sense as the values you get depend heavily upon microphone type (frequency response and dynamic range effects), distance of the bat from the microphone (differential frequency attenuation with distance), other atmospheric distortion effects, and any post processing with preemphasis and frequency filtering. That said, SonoBat maps the amplitude range over 256 values, typically having about a 48 dB range for most bat calls.
SonoBat has a print function (lower right button), but for graphics, I usually just set up the display as I like and then take a screen grab of the area I want. Use the Snipping tool in Windows, or Command-Ctl-Shift-4, the Grab or Screenshot utility on a Mac.
SonoBat has the sonogram screen colors and brightness optimized for night or desk work, so for graphics I usually tweek the image a bit in PhotoShop or anything else that lets you adjust the contrast, light levels, and saturation to make it look better on a printed display.
We’re using Sonobat 3 to identify bat calls and we have used the Sonobatch tool to run the analysis, and now I am attempting to go back and re-check classifications. However, I am unable to move around the high frequency, low frequency, start, end, and knee points, even though I am in standard and analysis view.
An early, manual, version of SonoBat relied on users to position the parameter cursors, but that created much discrepancy in species classification as it depended upon an inevitable subjective, and non-repeatable process. To enable automated analysis and classification, SonoBat automatically and objectively positions the cursors to display the selected measures. That is, the cursor positions you see on the screen serve as indicators rather than user-input controls of the process that generates their positions, and that best match the process by which SonoBat generated the data that went into the known data sets that serve as the basis for the SonoBat automated classifiers.
We do sometimes see calls on which I disagree with how the automated system interprets the parameter positions, but that typically happens on distorted, weak, or otherwise compromised signals best rejected for analysis anyway (and SonoBat may also automatically reject, too). In short, that was a legacy feature that has been replaced with automation.
I would say still more accurate and better representation of CPS with the CalcCPS. Here’s why: yes, seems a more refined measure that rejects approach phase etc. ought to come out with lower CPS. It does not though because the crude measure that gets displayed in compressed view only divides the total number of call parts discriminated (to construct the compressed view) by the time of the file, and that can include other empty non-bat time intervals. The CalcCPS works from calls that have undergone a selective process of recognition and separation. For example, if you rendered the sequence in realtime and counted search phase calls across a good section of the sequence, you would get a better count than the crude screen-displayed CallsPerSec, and one that comes out closer to the more refined CalcCPS. I will probably update the screen to display CalcCPS, but that can only follow after performing a classification to recognize, select, and group the data results from that process.
If I open a file with a sequence of calls and choose the "std view" display, how can I assess which call out of the sequence was selected? Does SonoBat automatically choose the call that suits best for analysis of parameters?
If I choose a single call manually the standard view often seems to be less noisy and of a higher quality.
Yes, SonoBat processess the selected call into a separate high resolution sonogram. The notion of standard view length keeps them to a fixed time scale so that you always view calls in a similar scale and get used to the appearance of different call types and species to support recognition. See the slides beginning with number 12 in this linked guide
Basically, standard view selection works slightly different in compressed view mode and realtime mode. The latter gives you complete control over the time selection and positioning of the call within the frame.
Change the filter setting to unfiltered and the preemphasis setting to none, and the view setting to realtime.
That will enable you to open the large files with the minimal amount of processing overhead. You can also set the max segment to process to its largest setting of 32 sec to view the largest chunks of the files at a time.
You can then work through file sections, do zoom-selections of passes, and save them as separate files using the "save zoomed selection" option (only available when zoomed in an part of a sonogram in Realtime View).
SonoBat works with all full-spectrum data (including time expanded full-spectrum data) including that from Wildlife Acoustics SM2 recordings. But do note that the lower priced SM2 equipment can produce somewhat compromised recordings compared with other ultrasound recording equipment, particularly for lower amplitude signals.
That will reduce the overall number of usable recordings you acquire, but you should still manage to record many good samples from bats that fly close enough to your microphone to discern what you have. Overall these differences should have less affect on the European suite of species that tend to have more acoustic differences than some North American Myotis species that have ambiguous and subtle differences.
When I analyse a call and exported data to a file, the field PrecedingInterval is regularly set to zero. In consequences the field CallsPerSec is improper too. What can I change in settings or selecting to extract this parameter correctly
Perhaps you may be attempting to extract call data in realtime mode. SonoBat only calculates call intervals when it performs call discrimination to display in compressed view mode. In realtime mode, call parameterization will always output a zero for PrecedingInterval for that reason. Sorry for the trouble; that minor point is buried in the manual.
Also note this from the mean calls/sec popup help message:
Displays the mean calls per second of the recording or section of recording displayed. The accuracy of the reported value depends both on the quality of the recording and the absence of other bats in the recording. Any other signal components other than the bat of interest that pass thtrough the discrimination logic will be counted as calls and contribute to (and reduce the accuracy of) the calculation.
Two hours worth of recording bat pass events, i.e., individual files for each pass, or a single two hour recording? If the latter then that's the trouble.
The design of SonoBat processing and its utilities optimize it to handle files with durations typical of individual pass events. And it considers each file you feed it as one event, and will only output a single decision for any file it processes. It is not designed to sort out the individual passes in a continuous recording- it leaves that up to the recording hardware. Most detector hardware will trigger and output individual files of bat pass duration length, .e.g., of 3, 5, or 8 seconds of duration or up to about 5–8 MB. If you throw bigger files than that at SonoBat it will likely choke with a memory error because of the processor-intensive work it does to interpret the signals at a high level of scrutiny. If you record in a way that produces files greater than 10 seconds or so, you will not resolve individual bat passes (i.e., such long files would likely have more than one bat pass). In addition to subverting activity counts of passes, that will also subvert the classification process that considers the strongest signal in any one file.
SonoBat works best when fed files of about 3, 5, or 8 seconds of ultrasound data. Even with 4 GB of RAM (depending on other CPU processes) a 24 MB file might choke it. Files bigger than 5 or 8 seconds would probably have more than one bat pass on them anyway, and that could cause confusion for identifying sequences. Try pulling out the bigger files, perhaps editing them down to individual sequences, and have another go at it.
This is of course less of a problem with newer, faster hardware. Because a long file may and usually will contain more than one bat pass, for autoclassifing files you should not make recordings longer than ~4 to ~15 seconds, just because you can!
You can never get more than +/- 1 or 2 kHz precision with bat calls because of Doppler shift effects from their movement relative to the microphone, and we can't control that. The bats also vary their calls considerably. They are quite flexible, in fact. Precise analytics do not apply here, rather overall ensembles of features and trends. And if two conspecifics fly together, then they will shift frequency to accommodate each other to parse out bandwidth, and if the other one is beyond the range of your microphone then you only get an odd sequence.
That means that you have files longer than your "max segment to process" setting (lower blue control to right of select subsection bar). That should have the default setting of 8 sec, and that will handle all normal 3 or 5 sec long recordings (use the pop up Help messages for more info- invoke with Ctl-H). If you have files longer than 8 seconds you may want to change the settings on your recording equipment or in post-processing as files that long will not cleanly discriminate discrete bat passes, and will interfere with effective classification.
Newer computers can open files much larger than 8 seconds. Set this to 32 seconds if you have newer hardware. You can set the default max segment to process in the preference panel. This is a control that helps prevent people from opening files too large to process, should you find yourself with file lengths ranging in minutes instead of seconds.
The SM2 MEMS microphones (now thankfully discontinued, but samples still remain in the wild) have a strong peak in sensitivity in the 40-60 kHz range. That overlaps with the strongest part of 40 kHz myotis calls, and unfortunately tends to deemphasize the lower amplitude beginning and ending details of call structure that can otherwise help with classification.
I'm giving a presentation tonight and would like to be able to play the sound files as I hear them when I click the "play real sound" button within Sonobat. Copying and playing the raw .wav file gives me everything (including lots of katydid noise), which is not what I'm after. Is it possible to play the sound file as it's heard within Sonobat, without having Sonobat on the computer that is hosting the presentation? (And how exactly would I do that?)
Sonobat's "play real sound" function not only applies any high-pass filter you've chosen, but also manipulates the ultrasound down into the audible range so you can hear the cadence in realtime.
This means you can't simply play back the .wav file outside of Sonobat even if you'd applied a high-pass using audio manipulation software to get the same effect. And Sonobat unfortunately doesn't let us save this special version of the audio to file.
Your best bet, assuming you use Windows, is likely this:
You can use the Windows "Sound Recorder" program to digitally record sound destined for your speakers, simply by selecting the "Stereo Mix" device as its input rather than your microphone. So you'll need to configure the alternate input device; hit the record button in Sound Recorder; hit "play real sound" in Sonobat; then stop recording in
Here are instructions for making the Stereo Mix device visible, and then selecting it as your recording device:
If all that sounds incredibly difficult, a different option would be to use audio editing software (for example, Audacity
http://audacity.sourceforge.net) to apply a filter (Effects -> Equalization) and then convert the .wav to a time-expanded version (filename dropdown -> Samplerate -> remove one digit off the end for 10x time expansion). Of course, a time-expanded version would not have the same fast cadence of a realtime bat pass, and it sounds like that's what you're attempting to demonstrate.
Go to preferences, and first select a Region, then a Subregion as appropriate. You can learn what exactly is in each region by reviewing the help info under the "?" button.
If you do not have all of the regions you are looking for, you can download them for free, then drop them into the Data folder that is located in the SonoBat 4 NA program folder. Restart SonoBat and all those new regions are available. For example Download the Great Basin classifier here from the SonoBat website: https://sonobat.com/product/sonobat-4-gb/ under the link that says “already own?" Repeat for other regions you are interested in, downloading those small classifier files from each region.
Ran same file thru NE and NNE, got different results...just wondering if this is "normal" because of the slightly different decision trees??
Yes, different decision processes. When you can exclude ambiguous species of concern then you gain much greater performance in classifying the species that you expect to be present in a geographic area. I mention that in the Help panels:
Applying this classifier to a geographic region outside the range of these species may result in some misclassifications of the out of range species (see auto classification note).
Because Myso and Mylu overlap so much in data space any normal Mylu call will often share similarity with calls that a Myso could make, and that gets represented in the SonoBat classification result. So if you know you can exclude one species or another, then best to use a classifier that excludes an ambiguous species when you can and you will eliminate the ambiguous probabilistic part of the classification.
By virtue of the call variation from all species, the process of classification does produce false positives on species ID. The species ID process essentially amounts to a probabilistic process because the plasticity and resulting call repertoires of species results in so many areas of overlapping data space.
Although some species do show a 100% correct rate for some situations, do note that rate describes the classifier's ability to recognize ideal reference (i.e., good quality) files of the type on which the classifier was built, and a finite data set. You should expect actual field data to present unknown call types and sounds outside of the known library, and any such signals can potentially result in misclassifications.
Only a smaller set of species-discriminating call varieties can provide confident spp id, e.g., the short, flat, above 26 kHz call types of Lano, and these are the type of calls that you would use to make confident decisions about presence . However many other call types from spp fall into overlapping data space, e.g., the longer curved call types of Epfu and Lano, and SonoBat will output what can only be considered a probabilistic decision, the correctness of which will depend upon quality, the distance of the bat from the microphone, and probably the bat itself. For some spp like Myso and Mylu back east we have yet to even recognize a species-discriminating call variant and the entire call repertoire apparently overlaps, even though the classifier shows some statistical difference.
Please read through the recommendations in the Classification Notes document and that text that comes up when you launch SonoBat. i.e., please don't trust SonoBat or any automated classification system without manually checking/verifying decisions. Those bats (and distorting effects) do all sorts of unexpected things that can result in spurious results
I’m in the process of analyzing a number of bat monitoring surveys I performed in the last month. I have come across some issues with Sonobat failing to identify the calls.
Due to variations in recording conditions, quality of recordings, and especially variations in the calls that the bats themselves make (with overlap in characteristics with other species), you cannot expect to successfully classify every recording. SonoBat performs a number of stringent quality control procedures to only output a classification when confident. Please review the Classification notes to recognize and interpret some of the challenges inherent in this work. Other classifiers will output more decisions, but with a consequence of higher error rates. Just as Siri can't always understand what we say, we should consider any automated acoustic classifier as assistive technology- it still requires some overview, and we humans can often do better.
Low signal level that don’t rise much above the ambient sound level is a problem with whispering or out-of-range bats. In most signal analysis work, usable signals have a signal to noise ratio of 5–10 or better. You can get an estimate of that by looking at the oscillogram trace below the sonogram. The height of it for the bat calls represents the signal level, and compare that to the height of the ambient sound level (i.e., noise) between the calls.
Most of your recordings will come out as low signal levels. However, you should get some proportion of them, perhaps 5, maybe 10% as good strong signals from bats making closer approaches to your microphone. Those fewer but good files provide your best samples for confident species presence decisions. They may have harmonics that indicate you have acquired the low amplitude components of the calls and thus have all the details.
If everything you record looks poor files then you may have a faulty microphone or have it set back in a way that it can only record bats at a distance, or something else impeding the signal.
You can (and should!) learn to manually recognize bat by their calls to manually ID many files that lack signal strength sufficient for acceptable auto ID, and to confirm those auto IDs. That’s the way we did it for the first 15 years or so with SonoBat before developing automated assistance. Learning and knowing the call structures, types, and when to and when not to make decisions through that process supported what we built into the auto process. Part of that gets reflected in producing “indefinite results” through recognizing that such signals lack sufficient data for confident decisions. Review and use the special characteristics in the attached acoustic table as a guide. Once you learn to notice them you start to calibrate your search image and ingrain your pattern recognition process for them.
Most detectors lack the smarts to discriminate between bats and other ultrasound sources and so will trigger on anything above a certain amplitude threshold and thus leave you with many or few noise files relative to your bat files depending on the ambient sound and richness of the bats at any site. Have a look again at the example SonoBat workflow.
With its original intent for determining species presence, SonoBat operates rather conservatively to only output a species decision when it achieves a confident decision that indicates clear disambiguation from other species and passes a number of recognized signal characteristics that can result in unreliable results, e.g., distorted, corrupted, or weak signals. Due to variations in recording conditions, quality of recordings, and especially variations in the calls that the bats themselves make (with overlap in characteristics with other species), you cannot expect to successfully classify every recording. SonoBat performs a number of stringent quality control procedures to only output a classification when confident. Many of the files you record amount to the acoustic equivalent of birds at too great a distance to indentify. You can recognize it as a bird, and perhaps even within a group such as a flycatcher, but you may still lack sufficient information to make a complete and confident identification. We have no control over how close the bats approach our microphones and we end up with many recorded sequence with just fragments of complete calls and SonoBat has logic steps to recognize these situations with unreliable information content that otherwise can lead to misclassifications.
SonoBat rejects calls deemed unreliable based on several factors, one of which is a quantified quality measure described in the popup as,
SonoBat calculates the signal quality based on the total points of the sonogram above a threshold value, and uses this value internally to assist in the call trending analysis of strong and weak call signals.
Generally, the best and most reliable calls to use for species recognition will have a quality value above 0.90 and calls with values below 0.90 will become increasingly less reliable with lower and lower quality ratings.
That is a synthesized measure found to work well for discarding unreliable signals. It is a basically a tweak of a combining signal to noise ratio and dynamic range measurements. It has a user defined threshold in the Prefs panel.
SonoBat also rejects calls based on poor call trends, i.e., too much small scale bouncing around of the trend line indicates noisy, unclear, or poor calls that produce unreliable results. E.g. a jumping trend can happen at the end of a call from overlapping echoes and this can result in the trend missing the downward toe that may possibly be essential for discriminating an Eptesicus from a low freq Myotis or a Myotis from a red bat.
And finally SonoBat rejects calls after a species determination if they do not meet minimum characteristics found to provide reliable results for a given species. One of those that can reject a call is too little bandwidth as that can indicate a call fragment. Under quiet ambient conditions calls can come out with a high quality rating but the bat can still be too far from the microphone to provide full call content essential for a confident ID and eliminating the chance of misclassification with the fragment of another spp, e.g., the fragment from the body of the call from a M. evotis falls into the data space of short calls of Eptesicus. With the full bandwidth of a M. evotis call and details like resolving its toe, the discrimination becomes confident.
Low quality equipment such as the SM2 with now discontinued microphones record calls that fail at that last stage because of the truncated bandwidth. A lot of other calls fail because of even great equipment set near the ground and with lots of objects around the microphones to generate lots of poor quality signals. Externally mounted microphones with good microphones and quality recording equipment allow much more data to be successfully analyzed.
Bats exhibit considerable plasticity in their calls and this and any other classification system must be considered fallible and serve merely as a guide and inference to classification. Please see again the first few paragraphs in the note below. Only a subset of each species' repertoire provides confident species discriminating characteristics, and the rest may exhibit ambiguous characteristics with another species. Call quality (and in particular distance from the microphone) greatly influence classification results.
For example, an out of range Epfu call will present or be absent of species discriminating features and may only exhibit a simple curved signal that falls into the data space of Labo, that only exhibit a simple curved call morphology. SonoBat uses many signal logic tests to attempt to recognize and discard lower quality signals. One of the Labo-classified calls may have later been discarded by the logic. If any of the call quality tests fail, SonoBat will display the classification result, but display it as grayed out to indicate an unreliable result. Such results will not contibute to the sequence decision.
Early reports of regional variation in the literature were not common garden comparisons, e.g., bats in clutter in one region compared to bats in the open in another region, and then concluding the differences were due to geographical locale. Upon discerning the full call repertoires of spp in each region, the geographical variations essentially evaporate into the intraspecific repertoire variation.
The expectation for regional variation follows from birds, who select their songs as identifiers, and have no constraint on complexity. Bats select their calls to optimize information acquisition to suit their tasks, and those tasks remain similar within a species. The physics of sound do not change across the continent, and as such the selective forces on call morphology operate to maintain consistent species call repertoires across regions.
The files included in the example files folder represent just a small sample of the thousands of sequences from species-known bats that we have tracked and recorded since 1991. They provide some views of different species, and with the intent that users of SonoBat can work with them to learn operations of the software. As such they provide an insufficient and unrepresentative sample of the call repertoires made by different species as covered by the complete archive that I incorporate into the SonoBat classifiers.
It may help you to know, and save you some effort, that the SonoBat classifiers work along the lines of what you suggest, i.e., each call that gets included in a sequence decision (after quality control acceptance to reject call fragments and other known indicators of issues that lead to misclassifications) receives an assigned probability and that contributes to reaching the overall sequence decision, along with known field-tested rates of confusion with calls of other species. Sequence decisions follow from an adapted sort of MLE decision from the set of individual calls along with other logic, redundant checks, and thresholding to reduce situations that lead to erroneous results.
Yes, SonoBat would enable you to build reference libraries and compilations for manual inspection and comparison for learning the bats in your area using SonoBat's standard view approach and append reference views. More info here on the RefCompiler. SonoBat will also process calls to output parameters that you can use for building classifiers. If you can assemble sufficient samples of known species in free flight conditions we could asist you in building automated classifers for your species. I recommend this Chapter for some suggestions on collecting reference recordings
Parsons, S., and J.M. Szewczak (2009) Detecting, Recording, and Analyzing the Vocalizations of Bats, in Ecological and Behavioral Methods for the Study of Bats, 2nd Edition, T.H. Kunz, ed. Johns Hopkins University Press (920 pp).
Regarding microphones and recording equipment: It will require a lot of effort to acquire your recordings, do please get the best quality of data that you can to make that effort worthwhile. I you would feel disappointment to later see the essential details of some species' calls that you might miss with lower fidelity recording equipment.
I have endeavored over the years to base the SonoBat reference library on as much a natural set of recordings as possible. Early on we did a lot of hand releases and ziplining but soon realized that those recordings only represent approach phase type variants and not what you record from unrestrained open air flying bats. So most of the last 15 years of effort have focused on acquiring known calls from species confirmed open air flying bats, and basing classification on those. We have used a variety of ways to acquire those: light-tags, spotlighting, recording downlight from known roosts, bats before hitting a net, and some special situations where you could presume a species such as recording an acoustically distinctive species far out of the range of another acoustically similar species, or using rosetta sequences that go through a transition of call types that include some species-discriminating variants that we can use to know with confidence that the other calls in that sequence come from that species.
The current reference library I use to build the classifiers consists of probably 90% or better sequences from unrestrained free-flight bats (that does include light tagged bats, but sequences from when the bats have settled into stready flight or returned again engaged in apparently normal flight activity. Restrained and hand please calls only contribute to filling in some of the short duration call varieties for species’ repertoires.
For a first look, plot all the data in dur v. Fc plots and you should see a nice cloud, or clouds of points. Then the outliers from the clouds become apparent and one must manually inspect those calls. That may reveal an errant trend from a noisy or low quality call, or a different bat that got in the mix. If something like that then consider it quality control and remove that sample.This process is repeated for some other parameters as well, usually with dur as the independent variable again.
MLEs provide a probabilistic inference of presence based on the known and expected ambiguity of species (i.e., confusion table) and the numbers in your analysis. For example, if your analysis comes up with 95 Epfu and 5 Lano, but we know that the process misclassifies 5% of Epfu sequences as Lano, then we can probablistically expect your 5 Lano to just be misclassified Epfu, and it would require more Lano to reach a confident presence for that species. The corrected counts reflect that same sort of logic. In the case above you would have perhaps zero for the corrected count of Lano. Of course in practice you can manually establish presence based on one unambiguous species-distinctive sequnce, despite the probabilistic expectation. That of course requires knowing what to look for, and not everybody using this stuff can do that. So yes, the automated process comes out rather conservative in establishing presence, but a knowledgeable human can do better.
One of our biologists contracted to do NABat acoustic survey work in Maryland had asked me about what MLE percentage to use? I told him that I think it depended on what the goals of the acoustic survey were and the risk assessment of errors of commission or omission. Are there any published bat studies that used Sonobat and reported MLEs less than 95%? I think if you are using acoustics to identify places to mist net for target species then a lower percentage would be acceptable? Any guidance on that?
MLEs (maximum likelihood estimates) provide a probabilistic inference of presence based on the known and expected ambiguity of species (as presented in a confusion table, e.g., in the SonoBat region tabs of the Help panel) and the numbers in your analysis. For example, if your analysis reported 95 accepted Epfu files and 5 Lano, but we know that the process misclassifies 5% of Epfu sequences as Lano, then we have a probabilistic expectation that your 5 Lano results may just result from misclassified Epfu. A probabilistic inference of Lano would require more Lano results than an expectation of possible misclassifications when you have that many Epfu results. The MLE calculation quantifies this expectation in the form of a statistical null result from 1 (likely not present) to 0 (likely present); SonoBat expresses this relationship in a more naturally interpretable form from 0–100% likely present.
Of course, in practice those 5 Lano could very well have come from Lano, and some of those Epfu results could have resulted from misclassified Lano. Despite the probabilistic expectation, you can manually establish presence based on one unambiguous species-distinctive sequence. That of course requires knowing what to look for, and not everybody doing acoustic analysis has the expertise to perform such manual vetting. The automated process and MLE estimates come out rather conservative in establishing presence, and a knowledgeable human can do better for species that have some distinctive call variants (e.g., Epfu vs. Lano) but not better for species lacking appreciably distinctive call variants such as Mylu vs. Myso.
Also appreciate that MLE estimates do not provide an absolute metric. To do so would require assessing the species classification ambiguities for each particular recording location and microhabitat elements, conditions, equipment used, and foraging modes of the bats.
To address your question regarding guidance on using the MLEs: In my experience with manual vetting I find that SonoBat rarely stumbles on presence when species achieve a 90% MLE or better. That goes with the caveat of the above paragraph and for microphones properly deployed to provide clean, well interpretable recordings (see SonoBat 4 Workflow and Recording Primer). Despite that, I still review data by putting it in the Vetting table, sort for vetting (right click on a column header to select that option), then checking the top results for each species. That said, depending on location, I may skip the species with high MLEs and focus on the lower ones to ascertain presence.
If you want to keep it just purely automated though, I would review some subset of your data and develop some rubrics, such as Epfu MLE>90% and more than 12 files assume presence, etc. Your suggestion of accepting 95% will probably* always work.
* There’s a statistical inference to that word ; )
The likelihood estimates provide a probabilistic estimate and have no direct connection to certainty. The likelihood calculation depends upon the absolute and relative counts of each species, and if there was a chance of their species count falling within the error rate of other classified species that sometimes get misclassified as those species. In other words, not a strong enough sample of a species' results for confident presence. Of course that's the probabilistic result, but you could possibly look at the files and accept or reject presence.
The Corrected count just considers the numbers of counts of each species and their known respective ambiguity from the classification matrix and decrements the counts for species that might likely produce misclassifications accordingly to provide a more confident count.
For example, having a lot of Epfu counts will probablistically generate some Lano, and an estimate of the number of Lano misclassifications that result from Epfu will subtract from the Lano count to make the "corrected count."
Even the corrected counts can still output misclassifications for difficult to discriminate call types. I work from the Consensus decisions and check or reject enough files to confirm presence or absence of species.
The likelihood estimate follows in the routine way from a confusion matrix of classifier performance. However, such an approach uses idealized results from data used to generate a classifier. The SonoBat "estimated" likelihood calculation uses adjustments based on experience of classifier performance on real world data sets.
In SonoBat 3, the workflow was that you got a results spreadsheet immediately after the analysis processing. You can still get this if "output legacy spreadsheets" is checked on the SonoBatch panel.
In SonoBat 4, the concept is more advanced. The batch process now embeds calculation results in the metadata, you only do this process once! It is not necessary to re-batch them ever again (unless you want to change the species decision threshold and call quality settings to force more classifications onto your data).
In SonoBat 4, the workflow is this:
1) SonoBatch classify a folder of files
2) Open the SonoVet tool, and add the folder(s) to the “project”
3) Use the Vetting Table tab to edit/sort your data to suit your needs
4) Use the Export tab to export your project into a number of different formats (the "current layout" is exactly what you see on the Vetting Table, but other options like Project Summary and Night-Site Summary are useful for Indiana bat surveys. (You may want to insert a “monitoring night” column on the Vetting Table before export if that info is useful for you.)
There’s lots of cool things you can do in the Vetting Table, mostly accessed by right-clicking the column headers…this is how you can reveal (add) or hide columns/metadata useful for sorting. … try to read all the help tips that appear and toggle the contextual help on and off using or .
SonoBat writes the batch data output sheets as tab-delimited spreadsheets at the same directory level as the first directory (folder) that you drop into the SonoBatch job field (see pop-up Help message for the SonoBatch button). The sheets you will find at that level will cover all of the directories you had in your batch. SonoBat names those files after the first director you drop.
For example if BatSpring you will find an output sheet named BatSpring-BatchClassify.txt and BatSpring-BatchSummary.txt. You will find separate output sheets for individual directories at their directory level. When building your SonoBatch jobs, SonoBat will look one directory level deep and include any directories of files in that directory with the exclusion of directories named Deleted Files or Scrubbed Files.
You can find more info on opening and using the output sheets here
When I click the output document for my batch summaries, the columns are shifted making it difficult to interpret my results without reformatting the document. With all the data I am entering, reformatting every output document would be very time consuming.<