How we learned to listen to elusive, threatened frogs

The Cape peninsula moss frog is smaller than 20mm and is, therefore, hard to monitor. Francois Becker

Scientists the world over have a frog problem: we have little idea of how many frogs there are in each species in the world. This means that we are unable to predict how many there will be in the future due, for example, to the effects of climate change.

There have been very few studies recording the size of amphibian populations. So those of us studying amphibians have very little with which to work. What we do know is that many frogs are under threat mostly from habitat change, but also from disease. In southern Africa they are particularly vulnerable to invasive species.

Male frogs advertise their presence with species specific mating calls. Determining their presence has traditionally entailed listening for these calls. If you want to know how many frogs are calling, then stand with your hands behind your ears and try to count all the animals you hear. It sounds simple, but it’s not that easy. I’ve tried.

Listening to 10 calling animals is taxing. More than 10 and it’s possible to get muddled. Choruses of over 50 sound like noise.

Luckily the revolution in digital media has helped those of us studying amphibians. We’re able to monitor vocalising species and record a large number of sounds from the environment. Automated computer software, like voice recognition on smart phones, can then pick out particular species from their calls. Acoustic monitoring is ideal for monitoring the abundance of a species over time, as it has minimal impact on the species being monitored.

But what then? Interpreting the number of calls is problematic for several reasons. First and foremost is the fact that the area a microphone listens to is not defined. The vagaries of the acoustic environment mean that the slightest wind, or even a change in humidity, can affect the distance over which sound propagates. This means that microphones listen to different areas every time they record.

Recently published research by my colleagues and I could change this. It provides a methodology for acoustic monitoring that calculates the area listened to by an array of microphones, as well as estimating the number of calls made in that area. This has important implications for acoustic monitoring – and for attempts to catalogue just how many frogs are “ribbiting” around us.

A new methodology

We used microphones to monitor the Cape peninsula moss frog (Arthroleptella lightfooti) in Table Mountain National Park, South Africa. These frogs are endemic to the area and occur nowhere else on the planet. They are listed as “Near Threatened” on the IUCN list of threatened species. Any data we can glean will contribute to their conservation.

We used an array of six microphones to monitor populations of Cape peninsula moss frogs over their winter breeding season from May to October.

Researchers have previously used time of arrival of sound at each microphone in an array to determine the position of calling animals.

We used a novel statistical technique developed by co-author statisticians Stevenson and Borchers (Spatial Capture Recapture: SCR) to analyse the automated call data.

Listening for calls of the endangered Arthroleptella subvoce to monitor its abundance in the Groot Winterhoek Wilderness Area, South Africa. John Measey

The analysis takes into account both microphones which hear calls and those which don’t hear calls, as well as the distance between microphones to build up an estimate of the number of calls and the area from which the calls come. For an encore, the statisticians combined the information from time of arrival and call amplitude difference with Spatial Capture Recapture to build the first ever statistical estimate of the density of calling male frogs from an acoustic array.

We also recorded rainfall and temperature and were surprised to find that the number of frogs calling at each site didn’t relate to either of these factors. Instead, it started off with relatively few animals calling early in the season (May), built to a fine crescendo in July and then tailed off toward October.

More surprising was that the area in which the microphone array could detect frog calls nearly doubled (from 400 to 800 m₂) during the winter breeding season. But because the technique accounted for this change in the size of the sampling area we could effectively monitor the calling density of the species without having to worry about the changing areas that the microphones recorded.

The new direction

The idea of using an array of microphones is not new. Using the time of arrival of sounds to each microphone to determine the presence of a calling animal, such as a frog, is old hat. The technique we’ve pioneered provides a new direction for acoustic monitoring because we were able to define the area in which the microphone array was detecting calls.

The study has produced a robust technique for estimating call density. This is because it combined the statistical wizardry that allowed estimates from Spatial Capture Recapture with the time of arrival and signal strength. That combination makes it possible to increase the accuracy of the estimate of call density.

Researchers wishing to take advantage of the wonders of digital media to monitor species at risk of climate change, like frogs, can now use a technique that will give them call density which can be compared across recording occasions.

This will contribute to our understanding of how threatened species which vocalise are faring on our changing planet.