An STM32WL microcontroller running a gas sensor powered by the sun could be the most effective solution to the early detection of wildfires. The question remains whether countries and companies will invest in it. The unique ecosystem at the heart of this approach is Silvanet by Dryad, a company that recently raised more than 10 million euros in Series A funding. Silvanet nodes and gateways are the first to offer solar-powered LoRa connectivity in wild environments. They can last 10 to 15 years outdoors to build a mesh network in the middle of nowhere without any operator infrastructure. Performance-wise, the system can detect smoldering within minutes, making it the quickest and most sensitive detection system yet.
Dryad will showcase its Silvanet wildfire sensors and gateways at the Things Conference, which takes place on September 22 and 23 in Amsterdam. ST will also participate in this event; attendees may use the code FRIENDS-ST to get 20% off their tickets.
Why is early detection so critical and so complex?
Increasing costs of wildfires
Summer 2022 holds the record for some of the worst wildfires globally. The earth observatory of the National Aeronautics and Space Administration (NASA) published satellite images of heat waves and fires in Europe, Africa, and Asia. Moreover, a new analysis from the World Resources Institute estimates that forest fires today burn nearly twice as much land as they used to in 2001. To put these numbers in perspective, wildfires in 2021 caused a global loss of 9.3 million hectares, a land area approximately the equivalent of Hungary. Worst of all, droughts and record-breaking heat waves stemming from climate change will continue aggravating matters for years.
Wildfires are a problem with such astronomical impacts that they are difficult to quantify. According to a 2017 National Institute of Standards and Technologystudy, wildfires represent an annual economic burden potentially exceeding 300 billion dollars in the United States alone. Moreover, the paper doesn’t consider indirect costs, like food shortages caused by the loss of agricultural land or the long-term health impact of smoke inhalation over the coming decade. When Moody, a credit agency, estimates, according to Reuters, that wildfires could potentially disrupt the economy of an entire nation (Greece) if the situation worsens, it’s hard to overstate the urgency and severity of the situation.
The shortcomings of early detection systems
Early detection is one of the most critical measures in preventing wildfires. However, current solutions aren’t precise or fast enough. Lookout towers have been around for over a century, but they must be manned and only work when there’s already a substantial amount of smoke. Satellite systems automate surveillance but only detect fires that have already spread. Others have tried using cameras, but they need a powerful energy source, which isn’t often available in nature, and they detect fires one to three hours after they’ve started. Silvanet provides a new solution that solves the power consumption and early detection challenge.
Why is Silvanet a unique answer to previous challenges?
The power consumption challenge
Carsten Brinkschulte, one of Dryad’s cofounders, explained that power consumption is the most critical technical challenge to solve in this situation and a paper from Michigan State University agrees with him. Indeed, the scholars explain that using solar panels is challenging because trees provide shade and cover, which blocks a significant portion of the sunlight. The researchers thus tried to harvest energy from moving branches. Dryad took a different approach: design a system that could run despite those dire conditions. And as Carsten told us, his company could only do it by using an STM32WL.
By combining the MCU and the radio transmitter, the STM32WL limits the number of components on the PCB and optimizes the overall efficiency. For instance, it’s possible to use only one crystal for the high-speed clock of the MCU and the radio. Additionally, the STM32WL’s LDO and SMPS allow certain operations to run faster on wake, thus performing tasks quicker, which saves even more energy. The STM32WL also has low-power modes fit for such a scenario. Dryad worked with ST to extract as much optimization as possible from the STM32WL to run on solar power in a forest.
The computational challenge
Another challenge that wildfire detection mechanisms face is scalability. Not all fires look alike, especially during the smoldering stage. Silvanet relies on a gas sensor, and not every fire will register the same way. For example, pine emits a different smoke profile than eucalyptus. As a result, Dryad implemented a machine learning application that adapts to different forests and got a powerful microcontroller to run it. It’s the other reason why the Silvanet sensors use an STM32WL. The MCU holds up to 256 KB of Flash and can run at up to 48 MHz. Thanks to hardware optimizations, the device can process data from the sensor and run a machine learning algorithm locally.
The communication challenge
Detecting the start of a forest fire is only one of the problems. The other challenge consists of sending that information to a monitoring system in the cloud, although connectivity is often non-existent. Silvanet thus cannot rely on cellular networks like NB-IoT or technologies that require an operator like Sigfox. Hence, Dryad chose LoRaWAN because it works in the middle of nowhere. Indeed, Silvanet uses custom gateways that run on an STM32U5. One is a LoRa mesh gateway, which extends the LoRaWAN network into the depth of the forest. The other is a border gateway that provides backhaul using a 4G/LTE modem with a 2G/GPRS fallback. There’s even built-in satellite connectivity using Swarm by SpaceX.
One wildfire sensor node typically covers one hectare of forest, and it is possible to use the mesh gateway to extend the network as needed while using the border gateway with connectivity to send messages to the cloud. The sensors are rated IP67 to weather storms and support temperatures ranging from -40ºC to +85ºC. This is possible because the Silvanet sensors and gateways have no ports or connectors. They all use rugged plastic housing. As a key requirement, Dryad implemented a firmware update over-the-air mechanism and used ST’s library to facilitate its implementation. Consequently, the products can receive patches to fix bugs or update the machine learning application to improve fire detection.
The last challenge
The last challenge remains to convince governments, industries, and others that there’s a new solution to wildfire detection. A sensor costs 48 euros, a mesh gateway costs 371 euros, and a border gateway costs 549 euros. When covering hundreds of thousands of hectares, the investment isn’t trivial. However, the outlay pales in comparison to forest fires’ direct and indirect costs. Dryad is at the Things Conference to show that thanks to new optimizations by STM32WL, a mesh of LoRa devices can run for 15 years in a forest. There are already Silvanet networks in Germany, Spain, Greece, Portugal, Turkey, the US, and South Korea, among others, and the company plans to manufacture 230,000 units in 2023.
