In this blog post I will provide an insight into how we conducted one of the world's first, large-scale, joint acoustic and camera trap survey, and how the acoustic data is helping to unearth a secretive side of large carnivores.
Entering the hidden world of large carnivore vocalisations
Good conservation is rooted in evidence-based decision-making. For conservation practitioners, one key bit of evidence is the knowledge of where and how many individuals of a species exist in a landscape. Conservation technologies such as camera traps (cameras which are triggered by movement), GPS collars, or autonomous recording units (microphones which record sounds) can help. These record data which can be used to calculate population sizes, determine where a species occurs across a landscape, or understand how a species might be impacted by human-wildlife conflict issues. However, biodiversity monitoring surveys rarely integrate different types of technology and autonomous recording units have never been used to study the occurrence of large African carnivores across a landscape. So, in August 2023 we set out to change this. In Nyerere National Park, Tanzania, an area with globally important populations of lion, leopard, and spotted hyaena, we conducted the first large-scale acoustic monitoring survey - 50 devices across 300km2 - for African carnivores alongside a camera trap survey and the collaring of 4 lions, with the goal of recording what lions sound and look like.
This picture is formed from 2 separate camera trap images fused together to demonstrate the technology which we deployed in Nyerere National Park. Nb this lioness was collared by the Tanzanian Wildlife Institute for Research (TAWIRI), not Lion Landscapes.
What are autonomous recording units/acoustic devices?
Autonomous recording units, which is just a fancy way of describing a series of microphones encased within a plastic enclosure, record both ambient sounds and those originating several kilometres away. As an interesting side note, and as a general rule of thumb, acoustic devices have a similar ability to ‘hear’ sounds as humans do.
On the left: an autonomous recording unit, named CARACAL. Each unit contains a circuit board with 4 microphones and a GPS unit. Cockroaches are generally not included. On the right: an example of the installation process of a CARACAL.
Usually, in terrestrial habitats, these devices are installed several metres high onto a solid substrate (e.g., a tree), are powered by a battery source, and record acoustic data onto an SD card. Our devices were powered by 6 D-cell batteries and recorded acoustic data onto a 256 GB microSD card. This enabled us to continuously record the highly diverse sounds of the African bush for three weeks, before the card became full and the batteries died. Because data were recorded continuously, in addition to large carnivore vocalisations, we also captured brand-new sounds emitted from birds, insects, and other mammals. In total, the acoustic survey ran for 60 days (this required a set-up, two checks and a takedown - these efforts were supported by the Frankfurt Zoological Society, Lion Landscapes, TAWIRI and the Tanzania National Parks Authority - TANAPA - for which I am immensely grateful) and racked up a massive 75,000 hours of acoustic data. So now that I’ve recorded lions, what exactly am I going to do with the data?
What do you mean - ‘lions have unique roars?’
A little-known fact is that lions have unique roars. A lion’s roaring bout is actually made up of 4 different stages and their unique full-throated roar is only one section of this full bout. We can use computational software to see how a lion's full-throated roar differs per individual. But admittedly, this is a weird concept - how can we see sound?
We can view sound using something called a spectrogram. A spectrogram is a visual representation of frequencies within an audio signal as it varies with time. Below we can see the roaring bout of a lion. A lion's roaring bout can be separated into four distinct stages. Initial soft moans develop into full-throated roars which then switch to intermediary roars before the bout is finished off by a series of grunts. We can see that actually the majority of a roaring bout is made up of grunts and that the full-throated roars (which are the really important section because these roars are unique per individual) are usually restricted to 2-4 per bout.
This is a spectrogram of a lion’s roaring bout. Black lines indicate the different vocalisations within the 4 different stages of a bout. A bout can be broken down into four stages. A) moans B) full-throated roar C) intermediary roar D) grunts.
The full-throated roar of three different lions. We can see how the contour varies per individual.
One focus of my PhD is to use the fact that a lion's full-throated roar varies per individual to produce a density estimate by looking at the number of unique roars across the landscape. The idea being that the number of unique roars is related to the number of vocalising individuals in Nyerere National Park. Any estimate that we calculate can then be compared to the results from the camera trap survey - a more standard and proven method of estimating the size of lion populations - to determine the usefulness of acoustic data in estimating the density of lions.
But what about other species? Surely spotted hyaenas and leopards make cool noises too?
Whilst this blog has specifically looked at lion vocalisations, my research will also hopefully incorporate the fascinating noises produced by spotted hyaena and leopard too. Via research or anecdotal/heuristic knowledge these species are also thought to produce vocalisations which are unique to an individual. But these have their own challenges.
Spotties (spotted hyaena) have incredibly complex vocal repertoires. This is really cool, but from an analytical point of view, a nightmare. So, what this means is that for any analysis which attempts to differentiate between individuals I will need to select one very specific type of vocalisation. Even across a spottie’s whooping bout there are several variations of a whoop. Research has suggested that using the symmetric whoops are the best way of differentiating between individuals. This means though that there are many other very cool and interesting vocalisations that I will not use for this study e.g., hyaena laughter or predator-prey vocal interactions.
Different vocalisations within a hyaenas whooping bout. Black lines indicate the different vocalisations within each stage. A bout can be broken down into 4 stages. A) preliminary whoop B) symmetric whoop C) asymmetric whoop D) terminal whoop.
Leopards produce incredibly distinctive ‘sawing’ roars. If this statement leaves you confused, check out this link - and be ready to discover nature’s own chainsaw. Currently, there is no baseline study which proves that leopard roars differ per individual. This makes it difficult to conduct an acoustic density survey. However, given the volume of data that we have collected we are hopeful that these more basic questions can perhaps now be answered, and therefore lay the groundwork for future acoustic surveys of large carnivores.
When applying the notion that an animal's vocalisations are unique to themselves, an important caveat and absolute necessity is that there is greater interindividual vocalisation variability than intraindividual. I.e., there is more variation between the vocalisations of different individuals than the natural variation that occurs from the same individual.
The discovery of carnivore vocalisations
Something which I find incredibly exciting is that large carnivore vocalisations (including the globally recognisable lion roar) are very much an under-researched field. But we are entering the gold-rush phase of acoustic monitoring; by combining large-scale passive surveys with newly acquired AI tools which allow us to process data more quickly, we will start to break into and discover a world that in the past we have only been a listener to. Combining audio data from the acoustic devices, visual data from the cameras and GPS from the collars will surely only be a good thing if we want to succeed in our goal to develop a more holistic understanding of large carnivores across a landscape.
About the author:
Jonathan Growcott is a PhD student working with Lion Landscapes. He is integrating tech, AI, and on-the-ground fieldwork to improve the monitoring of large African carnivores with a particular focus on lions. You can follow his work on Twitter/X.
Kommentare