Old smartphones, rainforest ecology and illegal logging. This could be a “which one doesn’t belong” list, but in fact all three terms are connected. As it turns out, old smartphones used to power cutting-edge artificial intelligence (AI) technology are great tools to combat illegal logging in rainforests and monitor biodiversity.
Before delving into the tech, it may be worth revisiting a lesson most of us learned in biology class: Trees absorb carbon dioxide as they grow and convert it into oxygen. In fact, there is no greater carbon-capture technology than photosynthesis. Keeping forests intact is, therefore, a no-brainer in the fight against climate change.
With COP26 around the corner this topic will be at the top of the global agenda, and rainforests are an essential ally in this battle. Scientists and climate policy experts have confirmed that saving and restoring forests – especially tropical rainforests – is crucial to warding off the worst effects of global warming.
Sadly, illegal deforestation is still a persistent problem. According to the UN, up to 90% of logging in tropical rainforests is illegal. This in turn destroys the natural habitats of flora and fauna, driving diverse species to extinction.
The issue is that illegal activity can be arduous to detect. Tropical rainforests span massive areas where the surrounding environment impairs vision and sound. But what human eyes and ears cannot pick up, computers can. This is precisely how old smartphones hanging in trees – coupled with AI technology – have revolutionised the fight against illegal logging and poaching.
Listening to nature
It may seem like an odd pairing, a California-based non-profit working to combat illegal deforestation and a Chinese tech conglomerate, but the partnership between RFCx and Huawei has proven to be very fruitful.
Founded by ITER physicist and software engineer Topher White in 2014, RFCx is an NGO that uses acoustic monitoring technology to combat illegal deforestation and document ecological changes.
White first had the idea of using technology to detect the sound of chainsaws or gunshots in vast rainforest areas.
“It struck me that even though the forest is really loud, and no one can really hear and see what’s happening just a few 100 metres away, computers can pick that out pretty well.”
The initial idea was to build devices from scratch to record illegal activities in rainforests. However, it soon became clear that it was much easier and cheaper to upcycle old smartphones, for instance, old Huawei models.
While these devices may be considered outdated for the general public, they formed the core of RFCx’s essential technology: the so-called Guardians. These are the gadgets the organisation uses to record the sound of chainsaws, gunshots etc and to monitor the movement of certain animal species.
Essentially, we are talking about a black box surrounded by small leaf-shaped solar panels designed to catch as much sunlight as possible. These Guardian devices are equipped with microphones and antennas that collect and transmit audio data via cell phone networks.
Luckily, the mobile network in tropical rainforests is generally strong enough to transfer audio data in real-time
“There are few things more amazing on earth right now than the extent to which we’ve built infrastructure that is actually reliable in the middle of nowhere. It’s really incredible, the number of cell phone networks that exist around the world,” White tells Verdict.
However, this is not where the partnership between RFCx and Huawei ends. In fact, it was only the beginning. When Huawei approached White to see how they could help the NGO, the discussion was not centred around hardware but rather on big data analytics and building AI models.
In terms of hardware, cooperation has evolved as well. RFCx started out using upcycled Huawei phones, but now the Guardians consist of bespoke devices designed for bio-acoustic monitoring.
Over the years, the partnership between Huawei and various NGOs has expanded. The Guardian systems are now present in nature reserves around the world, from Costa Rica’s Osa Peninsula to Palawan in the Philippines and Sarawak in Malaysia. The devices are also used in regions closer to home, such as Greece, where they are deployed to identify gunshot sounds to combat illegal poaching and even in Ireland to monitor the sound of whales, dolphins and porpoises.
As to why the focus on sound in particular, White explains that this remains a largely underdeveloped avenue.
“Sound is not really an inherently human focus and yet it is really the way so much happens in the forest. Because you can’t see several metres in front of you, but you can hear stuff that’s happening a kilometre away, and these computers can. So that really became our speciality.”
Smartphones and intelligent models
RFCx found out that the microphones in old Huawei phones were strong enough to pick up relevant sounds from the get-go. However, White explains that they pick up certain sounds better than others.
“They’re more powerful for some things than others. Low sounds like chainsaws travel through forests very very far. High frequencies like bats and certain monkey calls don’t travel as far,” he points out.
This is where AI models can help. Microphones may pick up signals that are not audible to the human ear, “but the AI can pick it out,” White says.
The help from tech experts in companies such as Huawei is therefore much needed. Henk Koopmans, the CEO of research and development for Huawei in the UK, points out that he is by no means a specialist in anything wildlife-related. What Huawei can provide is expertise in building AI models and processing data.
Meanwhile, White says that RFCx’s contribution hasn’t so much been in training the data and dealing with machine learning, but more in speeding up the training process.
“It used to take months of work to add a single species, but by stacking up different filtering technologies, we can very quickly add new species over the course of an hour or two to the training data set.”
The combination of expertise has opened new doors in the fight against climate change and documenting biodiversity.
“In [using AI], it’s almost as if we’ve unlocked this multi-dimensional depth of the forest that people aren’t even able to figure out on their own. So when you combine AI with this passive acoustic monitoring that we’re working on, that’s where you begin to realise that this is how nature asserts itself,” White says.
What AI has really highlighted is its ability to sift through significant amounts of data very quickly. In the past, the monitoring of wildlife literally entailed that hundreds of volunteers would set off into nature reserves, holding clipboards and spend days or weeks documenting everything they saw or heard.
Modern technology has effectively enabled monitoring of an area 24 hours a day, seven days a week. Moreover, automation can crunch this data in split seconds, which has been transformative for science and research.
AI-powered monitoring technology has also made it easier to detect and react to illegal activities more quickly.
Eleanor Flatt from Osa Conservation explains that the ability to react in real-time is a game-changer.
“Rangers can actually monitor and respond in real-time. That’s one of the biggest challenges. How do you actually monitor and stop illegal activities to save the rainforest and stop hunting in real-time.”
White explains that it is also safer for the rangers. However, the reality is not always that simple.
“The collaboration and consensus on the reality of what is happening on the ground is more difficult than simply synthesising data and handing it off to somebody,” he explains.
Beyond stopping illegal loggers, real-time detection is also valuable to save endangered animals. Knowing where wildlife is moving gives rangers a better chance of protecting them.
“Hearing a chainsaw or a gunshot or getting an AI image that has someone with a gun or someone with a chainsaw near a threatened species can be super viable,” Flatt emphasises.
This real-time approach has also helped in providing evidence for illegal activity. It is one thing to catch illegal loggers and poachers and another to see them actually being prosecuted. Flatt explains that the usage of monitoring devices has significantly changed this endeavour.
“Normally, a lot of people know that it’s going on, but to actually have the evidence and get people prosecuted is another thing.”
Leaving a legacy
Beyond detecting illegal loggers and poachers, technology also plays a central role in the long-term strategy.
RFCx’s ambitions reach much further into the future. White explains that, apart from detecting and monitoring real-time changes, technology now enables data preservation for future research. Even if we are unable to preserve the land as it is today, these long audio recordings can document the land in its current state.
“Archiving this moment in history, I think, might be our legacy,” he said.
Arguably, it is now or never when it comes to documenting ecosystems.
“Climate change is already very well underway, and nature is in the process of adapting to it,” White pointed out. And with this change, animal species are changing how they interact and how they move.
Modern technology’s ability to collect and store raw data has become a central part of RFCx long-term strategy.
“I believe that there’s never been a moment since humans have existed on Earth that there’s been this much change so quickly in the ecology of the planet, so this might be the most pivotal interesting time,” White asserted.
However, the ambitious goal of capturing hours upon hours of data, and more importantly, analysing that data is not easy and would require a massive amount of human resources. This is where AI comes into play.
White compared the usage of AI technology in understanding complex ecosystems to the invention of the microscope.
“When they first invented the microscope, this piece of technology allowed us to peer into worlds that we’d never knew existed.”
“AI, compared with long-form audio capture, is like a microscope for ecology,” he said. Adding that AI can pick up the nuances in how animals interact based on behavioural patterns, which otherwise would be lost to human senses.
“I think that we’re going to see in the near future a renaissance in ecology and knowledge and understanding as relevant as the invention of the microscope, simply through the combination of AI, acoustic recording and making that available all together,” he argued.
Perhaps scientists are not yet able to fully grasp the extent to which the planet’s ecology is changing, but White believes that this data will only become more valuable in the future. “This is a moment in history that we can’t possibly grasp on our own right now. We have to save it and archive it and make it possible for people 100 years from now to look back at this singularly critical moment in the history of the planet.”
“That’s part of our mission, almost like the war photographers of climate change.”
Of course, AI is not the panacea for all of earth’s problems. The possibility of false alarms still exists. Essentially, the quality of an AI model reflects the quality of data that it has been given. “In terms of AI, you’re always thinking about to what degree is the data biased and to what degree are the algorithms, therefore, biased,” Koopmans agrees.
To ensure that the AI and the underlying algorithms are fed with correct data, cooperation between Huawei and different organisations is essential. The data needs to be constantly monitored to make sure that the right audio signal is identified correctly.
Especially when it comes to dealing with threats in real-time, the margin of error for false alerts is essentially non-existent. That is why cooperation and communication between the NGO and partners on the ground is essential, White explains. “You can really only send a few false alerts before you degrade someone’s confidence in taking a very risky move and responding to [a threat].”
White points out that the computer is not the one that makes the final decision. “AI is great at sifting through the data, curating it and then presenting actual information to people upon which they can make a decision,” he emphasised.
“I believe that decision making is the best when it’s semi-collaborative, so we want to make sure that if anyone in the field is getting information that we’re paying attention to it as well.”
Flatt also agrees that cooperation and mutual trust are essential. “Part of the role that Osa Conservation plays is getting that conservation technology on the ground, implementing it and training the right people in the community. The park rangers and putting the technologies in their hands, but then also working with the people who are designing the technology to make it easier for them to use.”
AI can clearly be very powerful in helping humans comb through large amounts of data. However, a blind following of technology – which may be faulty or biased – could lead to adverse and even dangerous situations. This is a point that RFCx, Osa Conservation and Huawei agree on, as White points out:
“There’s always going to be human input to machine learning models and making that efficient in an inherently inefficient medium, which is temporal audio, that has been our big breakthrough.”