- Cameras and microphones are capturing images and sounds of the world’s largest rainforest to monitor the Amazon’s species and environmental dynamics in an unprecedented way.
- The Providence Project’s series of networked sensors is aimed at complementing remote-sensing data on forest cover change by revealing ecological interactions beneath the forest canopy.
- Capable of continuously recording, processing and transmitting information to a database in real time, this high-tech experiment involves research institutions from three countries and the skills of biologists, engineers, computer scientists and other experts.
- The monitoring system will connect to a website to disseminate the forest biodiversity data interactively, which the researchers hope will contribute to more effective biodiversity conservation strategies.
Monitoring technology deployed in a remote rainforest over time subjects it to high heat, humidity, and rainfall, as well as potential damage by insects and other wildlife. When you add an annual 10-meter (33-foot) flood to the mix, you’re taking on nature in a big way.
Despite these challenges, an interdisciplinary research effort aims to connect a wireless network of 1,000 automated multi-sensor biodiversity monitoring modules in the western Amazon basin over the next five years.
An international team of researchers believes these units will provide an unprecedented understanding of the basin’s natural diversity, through collection and sharing of information from deep in the forest.
“Remote sensing satellites and science aircraft provide a wealth of data about broad changes in forest cover, deforestation and land use, but these methods reveal almost nothing about the true story of biodiversity beneath the forest canopy,” Emiliano Esterci Ramalho, from the Mamirauá Institute in northern Brazil, said in a statement.
Initial testing
Ramalho coordinates the Providence Project, which installed its first 11 prototype units in April 2018 in the Mamirauá Sustainable Development Reserve. Researchers set out 10 units — each containing a camera trap, a thermal sensor, and a microphone — to capture images and sounds of animals in tall trees. To capture underwater audio, they placed another unit, with underwater microphones, along the Amanã River that cuts through this 11,000-square-kilometer (4,250-square-mile) protected reserve, one of the largest floodplain areas in the world.
New sensors and better processing have greatly improved passive underwater acoustic wildlife monitoring in recent years, said Michel André, project collaborator and director of the Laboratory of Applied Bioacoustics of the Technical University of Catalonia, BarcelonaTech (UPC), in the statement. “For the first time, this technology is being applied to a large-scale environment in the Amazon for the conservation of terrestrial and aquatic creatures,” he said.
The researchers placed the test modules several kilometers away from one another. Each also contained components to process the data, harvest solar energy, manage power use, and transmit images and text data via wireless modem. The researchers assembled a network of solar panels and long-life lithium batteries to run the modules and process information locally. They also successfully transmitted the data to off-site researchers and managers using radio, GSM/cellular, or satellite networks from equipment installed in the treetops.
The water level in the Amazonian floodplain fluctuates some 10 meters between its high point from May through July to its low from September to November. The Providence researchers, therefore, placed the initial monitoring modules in large trees, some higher than 30 meters (98 feet), to ensure that the modules would not be submerged during the flooding period. One tree, the apuí, also serves as a shelter for jaguars during Mamirauá’s several-month flood period.
The initial phase, completed in October 2018, proved that the tech installations were viable, despite the logistic difficulties of the dense forest.
The project has initially focused its monitoring on fauna in the Mamirauá reserve. The first target species to be monitored are a cross-section of vertebrates, including jaguars (Panthera onca), large-headed capuchin monkeys (Sapajus macrocephalus), wattled curassows (Crax globulosa), and pink river dolphins (Inia geoffrensis).
The Providence researchers expect to gradually add species, to better understand the ecological community, how different species interact, and their vulnerability to various threats.
The modules installed in the first half of 2018 in Mamirauá are improved versions of previously tested modules. The researchers had worked for 18 months to collect data and test transmissions from the earlier versions to a receiving antenna at a floating lodge inside the reserve. As a result, all 11 units successfully processed and transmitted image and sound data to the antenna, which, in turn, sent the data to an online center accessible to the researchers.
Ross Dungavalle, an engineer at the Commonwealth Scientific and Industrial Research Organization (CSIRO), the Australian research institution that created the commercial wireless network and the structure of Providence’s data communications, said the module development phase was needed to ensure the technology could withstand the challenging conditions of the Amazon forest, including animals that might attempt to disassemble the equipment.
Overcoming these challenging local factors, he added, showed that Providence’s modules could also work well in other forested areas.
Operator-free data harvesting
José Reginaldo Hughes Carvalho, a professor at the Institute of Computing at the Federal University of Amazonas (UFAM), developed the machine-learning system the project uses to recognize animal species from images captured with camera traps. “This system is based on Google Tensor Flow, an open platform,” Carvalho said. “Our work consisted of three activities: installing Tensor Flow on the equipment hardware, integrating the camera and the presence sensor into the classifier software, and training the classifier from a deep neural network.”
Carvalho said some groups were training machine-learning algorithms for automated recognition of camera trap images. However, he added, “We do not know one off-the-shelf camera trap with a built-in classifier that transmits both the image and the label wirelessly and in real time, 24/7.”
He also said that even offline systems were expensive. “The current Providence Project prototype is based on cheap off-the-shelf components, including Raspberry Pi V3, Raspberry PI camera, and LoRa.”
Setting up the system was not easy, Carvalho said, as the highly qualified Brazilian-Australian technical team had just four days to set up and test equipment under the stress of high heat and humidity, buzzing insects, and the lack of an internet connection to download documents or ask for support. Despite the difficulties, the team successfully installed the devices, sensors and solar panels; set up modems and antennas; integrated hardware and software; and tested the system in the forest.
The flood cycle hinders long stays in the field, Carvalho said, one of the motivations for exploring the capacity of automatically functioning systems to monitor wildlife.
The project’s solution is to install the various devices far above the water line and keep them there for months. “Energy comes from solar panels, and the sensor automatically manages memory and data storage and transmits the information to a server,” Carvalho said.
The team of researchers in Mamirauá periodically inspects and maintains the project’s equipment, said Ramalho, the Providence Project coordinator. They make the repairs and occasional adjustments needed to keep equipment running and, if necessary, exchange antennas, solar panels and other components of the system. They also assess devices for interference by animals.
Looking to the future
Ramalho said the next challenges would involve the improvement of a website they created to make the field data accessible to decision-makers and the general public over the next two years.
“In the future,” he said, “this communication system will allow people to learn from the data interactively.”
Ramalho said he expected this virtual environment — with images and sounds of the animals, maps and others research results — to broaden the understanding among researchers, managers, authorities, and civil society groups about what happens to the biodiversity of the Amazon in the face of pressures such as increased deforestation, infrastructure construction, and climate change effects.
Ramalho said the research would explore the dynamics of biodiversity under the treetops in a way that could not be done now, since monitoring efforts are often conducted at specific sites within small forest fragments. He added that more complete analyses, generated by data collection carried out over long periods and at a broad scale, would enable more informed decision-making regarding the conservation of the Amazon’s biodiversity.
Examining data on water level variation in floodplain forests over the next few years, based on information collected daily, will be one way the project will broaden scientists’ understanding of the potential effects of climate change on Amazonian ecosystems.
Ramalho also said that maintaining a fully functioning Amazon forest would require more commitment from policymakers. “The government’s lack of understanding that it is possible to develop the country [while] preserving the forest represents a challenge,” he said. “We work in a sustainable development reserve, and we know that it is possible to maintain cultural plurality and forest diversity.”
Ramalho also said the lack of significant investments in biodiversity conservation and other scientific initiatives was a major obstacle in Brazil.
The Providence Project is funded by the Gordon and Betty Moore Foundation. Initial costs were $1.2 million. Ramalho said he expected that installing 1,000 monitoring modules would require an additional investment of $2 million to $3 million. As a first step, the Moore Foundation recently approved funding for developing and monitoring with 100 new modules.
The partner institutions are also looking for more resources to support the project. Representatives of the Mamirauá Institute and the Foundation of Reference Centers for Innovative Technologies (CERTI), based in southern Brazil, recently began discussing a possible partnership to replicate the Providence system elsewhere in the country. The partners’ initial idea is to create a startup to make less costly versions of imaging and sound modules, reducing the operation’s cost and making the technology more accessible to environmental agencies and other groups interested in monitoring remote natural environments.
Disclosure: The Gordon and Betty Moore Foundation was at one time a Mongabay funder but provides no editorial input in our reporting.
FEEDBACK: Use this form to send a message to the editor of this post. If you want to post a public comment, you can do that at the bottom of the page.