Latest / Internet of Things with Fexingo: Connected Devices, Sensors, and Industrial IoT / How IoT Sensors Are Saving Endangered Species from Poachers
Transcript
- Lucas: So you know how we've spent the past twenty-nine episodes talking about IoT sensors in pipelines, warehouses, wineries, bridges, parking lots — pretty much everywhere you can bolt a sensor. But there's one application I've been wanting to get to, and it's probably the most urgent use case I've come across. Luna: You're going to say wildlife conservation, aren't you? Lucas: Exactly. Specifically, using IoT sensors to stop poaching of endangered species — rhinos, elephants, pangolins. And it's not hypothetical. There's a nonprofit called Rouxcel Technology that's been working with the Ol Pejeta Conservancy in Kenya, and they've deployed connected collars on rhinos that can detect a poaching event in under a minute. Luna: That's wild — literally. How does the collar know something's wrong? Lucas: It uses a combination of sensors you've actually heard us talk about before: a three-axis accelerometer and a heart-rate monitor. The accelerometer tracks the animal's movement patterns — walking, grazing, resting — and builds a baseline. If the rhino suddenly bolts at high speed, or if the collar detects a sudden impact, like a bullet strike, the algorithm flags that as anomalous. The heart-rate sensor adds another layer; a rhino at rest has a heart rate around thirty to forty beats per minute, but under extreme stress it can spike to over a hundred. Luna: And all that data gets transmitted in real time? In a remote conservancy with no cell towers? Lucas: Right — that's the engineering challenge. You can't rely on cellular or Wi-Fi. So Rouxcel uses LoRaWAN — Long Range Wide Area Network — which is a low-power, long-range radio protocol. One gateway can cover up to fifteen kilometers in open savannah. The collar transmits small packets of data — just the accelerometer summary and heart rate — every few seconds. If the onboard algorithm detects an anomaly, it sends an alert immediately. Otherwise, it batches data to save battery. Luna: Battery life must be a huge constraint. These collars have to last years, right? Lucas: Yeah, you can't be tranquilizing a rhino every six months to swap a battery. That's stressful and dangerous. Rouxcel's collars use solar panels integrated into the collar itself — small photovoltaic cells that trickle-charge a lithium-ion pack. In sunny Kenya, they get about three to five years of continuous operation. The real breakthrough is the edge computing piece: instead of sending raw accelerometer data every millisecond, the collar processes the data on-device and only transmits when something interesting happens. Luna: So it's the same principle as the smart building sensors we talked about in episode thirteen — where the sensor does the thinking locally instead of cloud-washing everything. Lucas: Exactly the same. Edge computing for conservation. And the results are striking. According to Rouxcel, their system has helped rangers intercept poachers before they could kill the animal in over a dozen documented cases at Ol Pejeta alone. The alert goes to a central dashboard and to rangers' phones, often within ninety seconds of the event. That's fast enough to deploy a helicopter or a rapid-response team. Luna: How does the collar distinguish between a rhino that's just startled by a lion and one that's actually being poached? Lucas: Great question. The algorithm is trained on years of behavioral data — including events where poaching was confirmed. A lion chase typically lasts a few minutes, then the rhino settles down. A poaching event is different; you often see a sudden deceleration — the animal falls — and then the heart rate drops rapidly. The collar can also detect the acoustic signature of a gunshot if a microphone is included, though that's more common on elephant collars. The accelerometer can measure the shockwave from a bullet impact. Luna: That's terrifyingly precise. But it also means false positives are possible. If a rhino trips and falls, does the system trigger a full-scale response? Lucas: It can, but the rangers would rather check a false alarm than miss a real poaching. The system has a tiered alert: a 'high confidence' event — multiple sensors agreeing — gets an immediate response. A 'medium confidence' event triggers a notification to verify via camera trap or drone. Rouxcel says their false-positive rate is under five percent for high-confidence alerts. Luna: And the cost? These collars aren't cheap, I imagine. Lucas: Each collar runs about two to three thousand dollars — not trivial, but compared to the cost of losing a rhino, which is a keystone species and a tourism draw, it's a bargain. Ol Pejeta has about two hundred rhinos, so equipping all of them would be around half a million dollars. That's a fraction of the revenue from wildlife tourism. Luna: If today's conversation gave you something usable — maybe a fresh angle on how IoT sensors can be applied beyond industry — the way these episodes stay ad-free is listener support. You can find us at buy me a coffee dot com slash fexingo. Lucas: Appreciate that. So back to the tech — one of the cleverest parts is how they handle data transmission when there's no LoRaWAN coverage. Some reserves are just too remote, or have terrain that blocks the signal. In those cases, the collar stores data locally and offloads it when the rhino comes within range of a gateway — sort of like a 'data mule' approach. There's even a pilot using drones that fly over the reserve and collect data from collars via short-range radio, then relay it to the cloud. Luna: That's basically a flying USB stick. Lucas: Exactly. And it's not just rhinos. Elephants are being collared with similar tech, and there's a project in Sumatra using vibration sensors on trees to detect the sound of chainsaws — illegal logging is often linked to poaching. The same LoRaWAN infrastructure can serve multiple purposes. Luna: One thing I keep coming back to is the scalability. You mentioned earlier that these sensors are essentially the same ones used in industrial IoT. So the manufacturing scale is already there — the components are cheap. But the deployment and maintenance in a national park — that's a different beast. Lucas: Right, the bottleneck isn't the sensor, it's the ecosystem. You need trained rangers to fit collars, maintain gateways, interpret alerts. Rouxcel provides training and a software platform, but the conservancy has to commit to it. That said, the technology is becoming more off-the-shelf. There are now commercial LoRaWAN gateways that cost under a thousand dollars and can cover a whole reserve. The collars are assembled from standard parts — the same accelerometer you'd find in a smartphone. Luna: So the unit economics are improving fast. Lucas: Yeah, and that's driving adoption. In 2024, the Kenyan Wildlife Service started a program to collar all rhinos in key parks. South Africa's Department of Forestry, Fisheries and the Environment has similar initiatives. And there's a startup called Smart Parks based in the Netherlands that's doing open-source collars — they publish the hardware designs and firmware so any reserve can build their own. Luna: That's the kind of democratization we saw with industrial IoT tools like Arduino and Raspberry Pi. Lucas: Exactly. And it's not just mammals. There's a project in the Galápagos using IoT sensors to track giant tortoises — they're endangered and their migration patterns are poorly understood. The collars are smaller, lower-power, and use GPS plus accelerometer to map their routes. That data helps create protected corridors. Luna: So the same sensor stack — accelerometer, heart-rate, GPS, LoRaWAN — is being adapted across species. Lucas: Right. And the machine learning models are getting better. One challenge is that each species has its own baseline behavior. A rhino's gait is different from an elephant's, and both differ from a tortoise's. So you can't just copy-paste the algorithm. But the training process is similar: gather labeled data, build a classifier, deploy on the edge. Luna: What about privacy concerns? These collars are essentially tracking animals — but the same technology could be used to track people. Is that a conversation in the conservation community? Lucas: It's definitely discussed, especially when you talk about camera traps and drones. But the collars are specifically designed for wildlife, and the data is owned by the conservancy. There's no commercial data-sharing. Rouxcel and Smart Parks both have strict data governance policies — they don't sell data, and they only share anonymized metadata for research purposes. It's a far cry from the surveillance capitalism we see in other IoT domains. Luna: That's reassuring. And it's a good reminder that the technology itself is neutral — it's how you apply it. Lucas: Absolutely. And I think that's the takeaway: the same accelerometer that detects a vibration in a pipeline can detect a bullet strike on a rhino. The same LoRaWAN gateway that monitors a smart building can cover a savannah. The engineering principles are identical — just the context changes. Luna: And the stakes change dramatically. Lucas: Right. When the alert goes off, it's not about preventing a leak or a fire — it's about preventing extinction. That's a pretty powerful mission for a sensor. Luna: So if someone listening wants to dig deeper — maybe they're a student or an engineer looking to get involved — where should they start? Lucas: Smart Parks has an open-source GitHub repository with all the collar designs and code. Rouxcel Technology also offers internships and field training. And if you're just curious, the Ol Pejeta website has live data from some of their collars — you can see the accelerometer readings in near real time. It's a fascinating window into how IoT is saving lives. Luna: And saving species. Lucas: Exactly.