Koalas are quite vocal mammals, especially in the breeding season. Surveying for koalas through audio recordings is non-invasive and effective, especially for low density populations. It is also fairly easy to do with inexpensive equipment, and really fun.
We took recordings at two Koala Clancy Foundation planting sites in Victoria for 37 nights at one, and 14 nights at the other, in peak southern koala breeding season: mid November to mid December. Only three koala bellows were recorded, two from one location 10 days apart, and one from another location at the same property, and a koala was found but not at the location of any of the calls.
Good scientific literature can be found here, but starting out with passive audio surveys can be daunting for citizen science groups. Here’s a step-by-step guide to using audio recorders for koala surveys, with the aim of helping other groups use this technology.
Our main intention is to monitor these planting sites over 5-10 years to see if koala records increase as a result of planting. There have been no known Koala sightings on the two sites since 2021 (or earlier – landowners can’t remember when a koala was last seen), but survey effort has been low and we know koalas can go unreported. Both sites had a small amount evidence of recent koala use. This monitoring is intended to start developing a baseline. As the planted trees grow, koala sightings and recordings could increase.
Choosing the koala audio survey sites
Two early-stage revegetation project sites were chosen, both on private property. Woolbrook was planted with 10,000+ koala trees & plants in 2022 & 2023, and Moranghurk was planted with 3600+ koala trees & plants in 2023.
Moranghurk, on the Moorabool River at Lethbridge, was chosen because the planting project was funded by the International Fund for Animal Welfare (IFAW) and a component of the funding was dedicated to monitoring. It was also the site where koala presence was most strongly suspected – it is adjacent to a large continuous block of forest that links to the Brisbane Ranges NP, a known koala area. Koala scratches were found on trees on the property, and koalas had been sighted and reported on iNaturalist in 2021, 2022 and 2023 both upstream and downstream, some within 2km of the site.
Audio recorders were placed along the river in 4 locations. Location 1 was in River Red Gum woodland, and judged to be the most likely location for koala presence due to a large block of suitable forest to the north and west. Location 2 was amongst Yellow Gums, on a well-vegetated slope. Locations 3 & 4 were on cleared slopes at the 2023 and 2024 (future) planting sites respectively.
Woolbrook, on the Native Hut Creek at Teesdale, was chosen due to evidence of koalas on the property (tree scratches) but no recent sightings. The property has 90 ha of contiguous old growth remnant forest, dominated by River Red Gum (Eucalyptus camaldulensis), Manna Gum (Eucalyptus viminalis) and Yellow Gum (Eucalyptus leucoxylon). The site is adjacent to the Inverleigh NR, which is known to have a small population of koalas.
Audio recorders were placed in the woodland with recent evidence of koalas, and the most distant from houses.
Buying the equipment
We bought 10 Audiomoths from Group Gets and they arrived within 6 days. In total we paid $1440AUD for the units and $75AUD for postage.
Initially we bought 10 micro SD cards (SanDisk Extreme 32GB) from Officeworks. But soon realised that you need at least twice as many cards as recording units, as it takes days (at best) to analyse and remove the data on the cards, and you need another card to keep recording. We got a better deal on cards at Cheap Chips.
We used ordinary NiMH AA batteries that we already had, and bought new rechargeable AAs. Each unit takes 3 x AA batteries, and as with the cards, you need twice as many batteries as units. They are costly, I would recommend buying these from an online supplier.
So all up the out of pocket costs were:
Audiomoths 10 @ $144.50ea (+ postage) = $1515.00
SD cards 20 @13.50ea (+GST & postage) = $287.90
Batteries 60 @ $22 for 4 pack = $330
TOTAL: $2132.90 AUD
Configuring the audiomoth recorders
The manufacturers notes on setting up the audiomoths are fairly easy to follow. Start here: https://www.openacousticdevices.info/getting-started
Follow the instructions, download the configuration app to your computer (ignore the AudioMoth Flash app – I don’t think that is necessary any longer), and install the audiomoth app from the app store on your phone.
When you open the configuration app the time on it will make no sense, but don’t worry. (#1 on image below)
Recording tab: We left everything on the Recording tab as it was in default.
Schedule tab: this is where you add your recording period in UTC (#2 on image above). Convert UTC to local time here: https://savvytime.com/converter/utc-to-australia-melbourne
Change the start & end time (yellow arrow) and hit Add recording period. Recording period will appear in red on the bar (yellow arrow) You can only do max 4 of these. Leave everything else there as is.
Filtering tab: We left everything on this tab as it was in default.
Advanced tab: click “Always require acoustic chime on switching to CUSTOM. (#3 on image, yellow arrow, blue tick. If you don’t do this, you’ll have to reconfigure the unit every time you change the batteries.)
We left everything else on this tab as it was in default.
Put batteries in the Audiomoth and you’re ready to configure.
The biggest hurdle was the cable required to connect the unit to the computer. It must be a Micro USB data cable. Initially I tried to connect using a charging cable, which looks the same – it fit all the ports, and I didn’t realise it didn’t carry data. The only difference is the number of wires inside, and apparently some manufacturers are misleading about this on the packaging. The Click 1m Micro USB cable from Bunnings works.
If the cable is correct, you’ll hear a bell when you connect the unit to your computer, and the time on the configure app will start to change. Click Configure, and the time will change to the UTC correct time, and the green light on the unit will come on. That’s it, its ready.
If the cable is not a data cable, you’ll hear no bell and the time won’t change.
Based on this paper by Hagens et al (2018) and our perceived ability to analyse the recordings manually, we decided to only record for one hour each night for 7 nights, but to set the unit to record the southern koala peak calling time from 12midnight to 1am (beginning 4 hours after sunset).
Part way through we discovered that an automatic recogniser is available that works for surveying Victorian koalas, so we changed to recording period from 10pm to 2am.
Installing & replacing the equipment
On site, switch the audiomoth to CUSTOM and open the audiomoth app you installed on your phone. Play chime. The green light should flash. The unit is now ready to record. (Occasionally the chime doesn’t work and the red light flashes or nothing happens. Keep trying until it does.)
We installed the audiomoths using two ziplock bags: an inner bag inside a slightly larger outer bag. Both were sealed tightly. We attached the top seam of the outer bag to a thin branch using a removable cable tie.
The tree branch chosen was a living branch where possible, thin and tending to horizontal to reduce attenuation effects on the recording. Audiomoths were installed at about shoulder height for easy access, and above the height of curious sheep.
Some other bags did take some water in, but in all cases except one the water didn’t penetrate the inner bag.
This method proved effective in most cases, even after several days of heavy rain and strong winds. However, one inner bag sustained tears, probably from the audiomoth unit rubbing against it in strong winds, and some water penetrated. It didn’t damage the recordings on the card.
After 37 days in use, the bags were showing signs of wear, and I would recommend replacing them monthly.
The rechargeable batteries lasted 11 nights @ 4hours per night.
We numbered all the cards and the audiomoth recorders, and kept a spreadsheet record of what was where and when.
Analysing the audio recordings
Our first survey design generated 420 files/7 hours on each recorder, each week. Analysing that manually was a big challenge, even with a willing crew of volunteers. We gave each volunteer the SD card directly, as that seemed the quickest option.
Our second survey design (with recogniser) generated 240 files/4 hours per recorder per night, or 1680 files/28 hours per week.
Analysing those, for 3 units for 12 days (144 hours) and 3 units for 14 days (168 hours), with a koala recogniser, still takes time.
There are several ways to ‘listen’ to the recordings.
Audiomoth Play https://play.openacousticdevices.info/ is online and generates a really good spectrogram of the whole 55second recording. But you have to click into every file and that takes time. Koala audio looks like this:
AviaNZ https://www.avianz.net/ is a program you download and it can take the koala recogniser.
The NSW koala recogniser can be found here: https://www.dpi.nsw.gov.au/forestry/science/forest-ecology/fauna-identification-service Thankyou Dr Brad Law for alerting me to this, and the NSW DPI for making it publicly available. I tested it on known Victorian koala recordings from You Yangs and Raymond Island and I can confirm that it works for Victorian koalas.
The AviaNZ program is relatively simple to use in its basic form, but I am still getting my head around aspects of it. It doesn’t create a spectrogram that is easy to interpret though, and the user manual online seems to be out of date. I’ve tended to use it just for the recogniser, and then run good recordings through Audiomoth play.
To run 1 card of 1680 files/28 hours through the AviaNZ recogniser takes my computer ~3 hours. Sometimes the recogniser returns 7 to 10 possibles, sometimes up to 30. Manual reviewing takes less than a minute each. False possibles occur from thunder, wind, vehicles, dogs barking, loud music and Common Brushtail Possums. It is recommended that you check randomly for false negatives too – according to Law (2018), faint bellows can be missed by the recogniser.
I recommend caution in siting the audio recorders near human activity – many of the Woolbrook files were of loud music and dogs barking, even though the nearest houses were 400m away. At Moranghurk the River Red Gums were flowering, and many of the files were very loud Grey-headed Flying-foxes. However I don’t believe any of these other noises would have drowned out a koala bellow.
Results of koala audio surveys
In total we ended up with 231 hours of recordings at Moranghurk, and 168 hours of recordings at Woolbrook.
To date, we have recorded koala bellows on four occasions. Two were single calls from a recording device at Location #1 Moranghurk, separated by 10 days. The other two were single calls at Location #3 Moranghurk 5 days after the the first call and at Location #2 Moranghurk 7 days after the first recording. We have also recorded a koala visually on one occasion at Location #4.
The first call was recorded on 15 November 2023 at 12.12am at Location #1. Listen: https://inaturalist.ala.org.au/observations/192038956 The minimum temperature was 9.4C
The second call was recorded on 20 November 2023 at 12.33am at Location #3. Listen: https://inaturalist.ala.org.au/observations/197580036 The minimum temperature was 10.9C
The third call was recorded on 23 November 2023 at 12.37am at Location #2. Listen: https://inaturalist.ala.org.au/observations/197367905 The minimum temperature was 11.6C.
The fourth call was recorded on 26 November 2023 at 12.49am at Location #1. Listen: https://inaturalist.ala.org.au/observations/195211590 The minimum temperature was 12.7C. There was 15.8mm of rain.
It is interesting to note that the first three calls were not repeated in the one hour recording period of those nights. The fourth call was on a recorder set up from 10pm to 2am but was not repeated either.
Followup site visits showed recent koala scratches on a tree 70m away from the audio recorder at Location #1.
Koala scat less than 7 days old was found on 8/12/23 near Location #4, 0.9km from Location #1. No calls were recorded on that audiomoth, that was in place for 10 nights before the scat was found. Scat: https://inaturalist.ala.org.au/observations/193430826
On 19 December 2023 at 2.15pm a male koala was found 50m from Location #4. The audiomoth was recording for 21 nights before that, and did not record any koala calls. https://inaturalist.ala.org.au/observations/194379490 The minimum temperature was 11.3C.
Following this, a male koala was seen by the owners on 22 January 2024 in another part of the property. It is impossible to tell from the photo whether it is the same koala.
Very wet weather occurred during the 37 day recording period at Moranghurk: 1 day over 30mm rain , 2 days over 15mm rain and another 5 days over 5mm rain at SheOaks weather station near Moranghurk. But the overnight minimum temperatures were not particularly high over the period.
Conclusions & Recommendations
I was surprised at how few koala calls were recorded, and that the ones found were single calls. Did he call once, and move out of the area? Or did he stay, and only call once because there are few koalas in the area and there’s no need to call more often?
It is clear that male Koalas can be present and not be recorded. The results also suggest that the koala population at these sites is fairly low, as I expected, but there is so much more to learn about interpreting these results for abundance.
This method is suitable for citizen science groups. Managing the large volume of data collected is a challenge, but it can be done. The automatic recogniser does streamline the process, but it still requires some people-power.
We changed batteries and cards every week on average, but on reflection I think we could stretch it to 14 days without missing any recordings. Regular visits to a site that include a visual survey are highly recommended.
If you find a koala, don’t remove the audiomoth. I made an error by removing it, assuming there would be at least one call recorded on it! There wasn’t, and now I’ll never know whether he called that night.
NOTES & REFERENCES
Hagens, S.V., Rendall, A.R. and Whisson, D.A., 2018. Passive acoustic surveys for predicting species’ distributions: Optimising detection probability. PLoS One, 13(7), p.e0199396. https://doi.org/10.1371/journal.pone.0199396
Law, B.S., Brassil, T., Gonsalves, L., Roe, P., Truskinger, A. and McConville, A., 2018. Passive acoustics and sound recognition provide new insights on status and resilience of an iconic endangered marsupial (koala Phascolarctos cinereus) to timber harvesting. PLoS One, 13(10), p.e0205075. https://doi.org/10.1371/journal.pone.0199396
Law, B., Kerr, I., Gonsalves, L., Brassil, T., Eichinski, P., Truskinger, A., & Roe, P. (2022). Mini-acoustic sensors reveal occupancy and threats to koalas Phascolarctos cinereus in private native forests. Journal of Applied Ecology, 59, 835–846. https://doi.org/10.1111/1365-2664.14099