Using acoustic monitoring devices to collect data about ecosystem health based on the sounds of different animals living in an area is a powerful and cost-effective way of measuring biodiversity and should be one of the primary data collection methods in new biodiversity crediting systems, say industry sources.

The field of ‘bio-acoustics’ – using sounds made by animals to understand more about the fauna and flora within a landscape – has been around for many decades, but it is only with more recent advances in artificial intelligence (AI), machine learning (ML), and data processing/storage that it has developed into a scalable way to track biodiversity.

Placing small acoustic devices across isolated points in a landscape allows ecologists to record the multitude of sounds of the many animals living there, which they can then feed into AI models to automatically detect where and when particular species are occurring across an entire dataset using their specific acoustic signatures.

“Acoustics is a tremendous dataset, because most species make noise … it’s an incredible way to be able to scale up and really track what’s happening with biodiversity from a presence perspective, ecosystem health perspective, and behavioural perspective,” said Chrissy Durkin, vice president of growth and impact at Rainforest Connection.

Through more advanced data analysis and aggregating similar sounds, scientists can gain a greater understanding of how animals use ecosystems – whether species return when landscapes are restored, whether they are reproducing there, or just passing through, she told Carbon Pulse.

As a result, bio-acoustics is a great way to measure biodiversity for biodiversity crediting mechanisms and is a much more cost-effective and scalable form of measurement than conducting point count surveys on the ground, for example, said Durkin.

The non-profit originally used acoustics as a tool to detect threats of illegal logging in rainforests, developing the technology required to pick out the sounds of chainsaws. It has since realised the power of acoustics for understanding more about species health.

Combining this acoustic data with other monitoring methods such as camera traps and environmental DNA (eDNA), together with information about environmental and climate factors, enables scientists to build a nuanced picture about how an ecosystem is changing over time, and how it responds to restoration efforts, for example.

No single technology or data collection method should be relied upon to measure change in biodiversity over time, as each comes with its advantages and disadvantages, but rather bio-acoustics should be used in conjunction with other complementary collection methods such as cameras and eDNA, depending on the ecosystem in question and the scientific aims, sources said.

METHODOLOGIES

The latest draft of Plan Vivo Foundation’s PV Nature crediting methodology, developed in partnership with Pivotal, does not stipulate the kinds of technology and data-collection methods that should be used for biodiversity measurement. Instead, it leaves the choice open to project developers to choose the best monitoring approach, said Cameron Frayling, co-founder at Pivotal.

However, the partnership intends to flesh out more detail around this, in alignment with the ongoing development of PV Nature.

Plan Vivo told Carbon Pulse: “The PV Nature methodology will be using digital data collection methods such as camera traps, high resolution imagery, and acoustic monitors, which allow the data to be independently checked, audited, and quality controlled”.

The benefits of digital data collection such as bio-acoustic devices are that it “allows any community member, field technician, or project coordinator to deploy the sensors, which can record data at scale for a fraction of the time and cost”, said Plan Vivo.

“This makes the costs and complexity much lower but importantly democratises access to biodiversity credits markets. This means that those who can’t afford a team of ecologists can still collect the data required to issue high-quality credits”.

Nature and carbon standard Plan Vivo launched a second consultation on PV Nature in August, which ended on Sep. 4, following a first consultation earlier in the year.

The methodology uses a percentage change per hectare and species-based approach to generate biodiversity credits. The organisation also plans to launch guidance on the pricing of biodiversity credits within a year, Carbon Pulse reported earlier this month.

Initial scoping showed prices would be “not far off the range we see in voluntary carbon markets, around $20” per unit, said Plan Vivo CEO Keith Bohannon.

The Plan Vivo Biodiversity Standard will launch later this year with details of digital data collection methods, the organisation said.

Fellow standard Verra says it is unlikely to prescribe specific monitoring technologies for its Nature Framework, which is out for consultation.

“Instead, Verra will develop criteria for appropriate monitoring approaches. Projects will likely be able to use bio-acoustics … for monitoring their selected indicators,” a Verra spokesperson told Carbon Pulse.

The consultation feedback and pilot process will help to inform Verra’s decision on the prescriptiveness of the Nature Framework regarding specific metrics or indicators, it said.

There isn’t a single scenario where bio-acoustics monitoring wouldn’t be useful for measuring biodiversity, even in marine sites, as it is an “incredibly cost-effective way” of collecting a rich set of biodiversity data, and is a powerful tool that can be deployed by anyone, claimed Frayling.

The use of automated analysis of animal sounds was recently highlighted by researchers, in the journal Nature Communications, for monitoring the return of biodiversity in abandoned agricultural areas of Ecuador.

The research group Reassembly (funded by the German Research Foundation) worked in northern Ecuador on abandoned pastures and former cacao plantations where the forest is gradually being reestablished to monitor how autonomous sound recorders and AI can recognise the species mix returning to the area.

A GREAT INDICATION

Bio-acoustic data is a great tool for measuring ecosystem health because many species indicative of a thriving ecosystem are vocal, such as lemurs that are endemic to Madagascar, and primates in Tanzania, said Durkin of Rainforest Connection.

Rainforest Connection has a project on the Ecuador mainland where bio-acoustics have been used to identify several new species of frog. In the Galapagos, it is working with the Charles Darwin Foundation to identify the presence of the critically endangered mangrove finch, of which only 100 or so known individuals remain globally.

Detecting critically endangered species with bio-acoustics can be a good indicator of a really healthy ecosystem, she said.

Rainforest Connection has been building AI models of biodiversity hotspots around the world using bio-acoustic data and makes its datasets publicly accessibly through its online cloud-based platform ARBIMON (Automated Remote Biodiversity Monitoring Network).

Its bio-acoustic datasets are so far mainly concentrated in the Tropics and are spread across Indonesia, Malaysia, Tanzania, Kenya, Western Amazon, Ecuadorian cloud forests, Panama, Mexico’s Yucatan, and Puerto Rico, said Durkin.

The data-logging effort is still far from having all species globally logged with specific acoustic signatures in the AI models, but the model does have the ability to flag any sounds it doesn’t recognise and is constantly being improved and added to, she explained.

Pivotal’s Frayling stressed the importance of using human quality control when deploying AI and ML, in order to spot any errors and improve the model with new sounds not yet registered.

Thousands of scientists are so far using the ARBIMON platform to store, manage, and analyse their acoustic datasets, and Rainforest Connection is also working with Google.org – the philanthropic arm of tech giant Google – to make the acoustic dataset more widely available to the scientific community, said Durkin.

Example of a spectrogram for animal sounds, provided by Pivotal

LIMITATIONS

Yet there are limitations to using bio-acoustic data to track ecosystem health, such as its inability to monitor non-noise-making species or those that don’t vocalise consistently, such as large cats and microorganisms.

Some bird species can also throw off AI models altogether by mimicking other species in their calls, which is why human experts are needed to double check bio-acoustics data, Frayling pointed out.

It is important to place sensors at high density in an area but “at least several 100 metres apart” to avoid overlaps in sound, said Durkin.

The number of sensors required will vary depending on the scientific question you want to answer and the ecosystem in question, but generally, sensors can be placed strategically in locations, so statistical survey models can be used to extrapolate beyond the direct area surveyed and make estimates, she said.

For example, “you can estimate that a species occupying a particular area will also live in another similar area exposed to the same environmental and climatic factors, without having sensors on the ground,” she told Carbon Pulse.

NEW HARDWARE

New hardware and software are also under development in the realm of bio-acoustics, as interest in biodiversity crediting picks up.

Pivotal is working with Swedish acoustics design and engineering firm Above to improve the hardware for bio-acoustics and make the technology more affordable and useful for a range of environments.

The venture is currently testing a new battery-powered acoustics device with four integrated microphones, which will be suited to land-based applications, and intends to make future devices that will also be able to be deployed underwater in marine environments, Frayling said.

These devices are about the size of a fist and can be strategically attached to a tree, post, or similar structure.

Pivotal’s focus is always on improving evidence-based data collection, so it hopes the new bio-acoustics device will be used in many biodiversity crediting projects to come. There is also potential for the device to have built-in GPS and time recordings to enable better data traceability, he said.

When it comes to effective data sampling for bio-acoustics, it is always better to have higher density sampling over shorter periods of time (using small and relatively inexpensive, battery-powered devices) than lower density sampling over longer periods of time, said Frayling.

“The most expensive version of bioacoustics is where you have a sensor that’s left out in the field, and some kind of transmission every day or regularly back to some kind of base. That’s very power hungry and expensive, and so will be unaffordable to most and doesn’t necessarily deliver better data,” he said.

UNDERWATER SOUNDS

Recent research has shown that changing ocean temperatures due to climate change could impact the way sound travels underwater and, in some cases, enable sound to travel much farther, which may have implications for marine acoustic data collection.

Researchers from the Royal Netherlands Institute for Sea Research in Texel found that colder surface water emerging in the North Atlantic Ocean as a result of climate change will open up a “sound channel”, enabling sounds to travel much farther and producing a fivefold increase in underwater sound if moderate carbon emissions continue.

This could enable more effective and widespread acoustic data collection in marine environments but could also pose an issue to the many marine species, particularly whales, which rely on sound for communication, finding mates, and detecting prey.

By Bryony Collins – bryony@carbon-pulse.com