The Internet of Wild Things: How the IoT joined the battle against climate change

From earth-observation satellites to smartphones listening out for chainsaws in the forest, cutting-edge technology is helping to record and protect the natural world.
Written by Charles McLellan, Senior Editor

This article was originally published as a TechRepublic cover story.

The interrelated issues of biodiversity loss and climate change are rising fast up the popular and political agenda. One reason is that the world increasingly appears to be -- on fire.

In August 2019, wildfires -- many started deliberately -- consumed large areas of Amazonian rainforest, reducing the Earth's 'lung capacity', rendering indigenous people and wildlife homeless, and releasing copious amounts of greenhouse gases. In September, on the other side of the world, forests in Indonesian Borneo and Sumatra burned. Again, human agency is widely suspected, as palm oil planters clear the jungle to make way for their crop. Massive bushfires are currently raging in eastern Australia, which experienced its hottest recorded summer in 2018/19.


Wildfires are also occurring in the far North with increasing frequency and intensity: in June 2019 (the hottest on record in the region), fires in the Arctic emitted 50 megatons of carbon dioxide -- equivalent to the total annual CO2 output from Sweden. Evidence that Arctic permafrost is melting faster than previously expected only exacerbates the carbon release problem.

Why is the world apparently fiddling while Rome burns?

The tendency for national governments to have a short-term focus, addressing immediate problems and deferring longer-term issues for successive administrations or generations, is not helpful when confronting planet-scale problems like climate change and biodiversity loss. That's because incremental 'business-as-usual' activities can run into irreversible tipping points that flip systems unexpectedly into new and undesirable states (the Amazon being an increasingly urgent example).

Although supranational bodies such as the UN and the EU try to take a wider view of such issues, recent years have seen a rise in nationalism around the world, leading to suspicion and even rejection of such bodies, often accompanied by the denigration of scientific evidence and expertise.

The Internet of Things, or IoT, is an area of science and technology that can help in the fight against biodiversity loss and climate change. In this article we'll outline the current state of the IoT and examine some examples of its use in vulnerable ecosystems.

The Internet of Things

The Internet of Things comprises sensor-equipped devices ('things') that can capture data, perform varying amounts of local processing, and connect to the internet to pass data on for further processing and/or storage. The latest estimate from analyst firm IoT Analytics puts the number of connected IoT devices at around 9.5 billion worldwide at the end of 2019. 

There is a wide variety of internet connectivity for IoT devices: wireless personal area networks (e.g. Bluetooth), wireless local area networks (e.g. wi-fi), low-power wide-area wireless networks (LPWAN, e.g. Sigfox), wired networks, or cellular (including, increasingly, 5G). Use cases cover 'smart' homes and offices, factories, cities, transportation systems, and many points in between -- including monitoring what remains of the natural world, where satellite connectivity will become increasingly important.

Download this article as a PDF (free registration required).

By 2025, IoT Analytics forecasts that there will be 28 billion connected IoT devices worldwide -- that's about 3.4 devices for every person on the planet (UN world population estimate for 2025: 8.18 billion). Other forecasts are available, of course, but the consensus is clear enough: IoT devices will proliferate, generating vast amounts of data, which will become actionable if backed up by a robust architecture comprising appropriate connectivity, gateways, analytics (including machine learning and AI), and storage/archiving.

2019 report from IoT Analytics includes an 'on the radar' guide to emerging IoT technologies, using a five-level classification: 'Fairly mature'; 'Nearing maturity'; 'Coming up'; 'Years out'; and 'Far on the horizon'. Technologies nearing maturity include IoT platforms, edge analytics, IoT-based streaming analytics, supervised and unsupervised machine learning, containers, low-power wide-area networks, and pub/sub messaging. Many of these technologies are already in use in biodiversity and climate change-related projects.

Among its top ten IoT developments in 2019, IoT Analytics highlighted nanosatellites, such as Eutelsat's ELO constellation, that are dedicated to IoT connectivity, and also the increasing role of IoT technology in helping organisations achieve environmental goals, as summarised in a September 2019 report from IoT specialist Libelium.

Technology versus biodiversity loss and climate change

Non-Government Organisations (NGOs) and charities

Instant Detect 2.0

The Zoological Society of London (ZSL), founded in 1826 by Sir Stamford Raffles, famously runs London Zoo, but is also an important international conservation charity. One of ZSL's current technology projects is Instant Detect, a monitoring system that combines sensors, cameras, low-power radio networks, and satellite technology to capture and transmit real-time information on wildlife and human activity anywhere in the world. The aim is to remotely monitor wildlife behaviour and habitat changes, and give timely warning of illegal poaching activity.

The first incarnation of Instant Detect, which was deployed in 2014, used cameras and metal-detecting sensors to identify poachers in protected areas, sending images or alerts over an FSK radio link on ISM frequency bands to a base station. From here, data was transmitted via the Iridium satellite network to a command centre, alerting the authorities in near-real-time to detected threats.

With the concept proven, ZSL and Cambridge Design Partnership set out to create the more ambitious Instant Detect 2.0 (ID 2.0). Key elements of the ID 2.0 specification were: affordability; maximising the number of sensor devices that can connect to a base station; reliable transmission between devices and the base station; good system diagnostics; usability; low power consumption; a modified camera; and cloud-based data and alert management.


Instant Detect 2.0 components (l-r): base station, sensor endpoint, camera.

Image: Sam Seccombe / ZSL

The Instant Detect 2.0 system comprises a base station that talks to the Iridium satellite constellation, plus cameras and sensor endpoints that relay images and alerts to the base station. To improve on the original FSK-based system's transmission range and power consumption, ID 2.0 uses LoRa, a low-power wide-area network, for wireless communication between cameras and endpoints and the base station. 

According to Sam Seccombe, ZSL's Technical Project Manager, "When we field tested LoRa radios in 2018 we managed radio transmissions of 10km when sending small packets of data through bushy scrub and up to 1km range when sending through dense rainforest." A custom LoRa protocol was developed to handle images, which are not normal fare for this traditional low-data-rate IoT-focused standard.

The base station can receive data from up to eight devices concurrently, and can queue transmissions if this number is exceeded. Testing in June 2019 achieved successful image transmissions over more than a kilometre  and improvements to the custom LoRa protocol are continuing, Seccombe says.


Left: An Instant Detect 2.0 base station ready for burial. Right: A camouflaged ID 2.0 camera. The camera can be triggered by a metal-detecting sensor rather than infrared, so that only images of humans carrying metal (poachers) are transmitted.

Images: Zuzanna Reymer, Sam Seccombe / ZSL

As well as improved range, LoRa also offers reliable transmission thanks to its ALOHA-type protocol, in which communication is initiated by the end-device and uplinks can be sent at any time. These are followed by two short downlink windows that provide the opportunity for bi-directional communication with the base station — a resend request for an interrupted transmission, for example. The ID 2.0 devices are ruggedised and send daily reports on battery and memory status, signal strength, and image/sensor event frequency. Although cost considerations rule out GPS tracking, if a device is moved — due to human or animal activity, for example — a tamper alarm will trigger an alert.

Download this article as a PDF (free registration required).

System management is handled via a browser-based OAM (Observation, Administration, and Maintenance) tool, accessed over a wi-fi link. ID 2.0 devices, which are designed to sleep whenever possible, are powered by internal Li-ion batteries and optionally by external (possibly solar-powered) batteries. ZSL is currently working to reduce the power draw in sleep mode to prolong battery life, Seccombe says. The ID 2.0 camera has a 5MP sensor and optics that deliver wide-angle or zoomed fields of view, and runs a Linux OS hosting image-recognition software that determines whether captured images should be transmitted or not. A cloud-based interface is also planned for Instant Detect 2.0, which will support image and alert handling, and remote management of devices in the field. 

Field trials with Instant Detect 2.0 are currently underway and production is planned for early 2020.

"One of the big issues we are currently facing is passing the tests needed to CE mark the system so that it can be provided legally to users in the European Economic Area [EEA] and many other countries that use the EEA's CE marking standards," Seccombe told ZDNet. "I know this is not as fun to report on as elephants ripping sensors out of the ground by their antennas or poachers discovering cameras and chopping them off trees with machetes, but I don't think this issue is well known to potential developers of wildlife conservation technology solutions -- and if it's not budgeted for or thought about early enough it might act as a critical stumbling block for a lot of people out there," he added.

ZSL is currently redesigning some parts of Instant Detect 2.0 to eliminate unwanted electronic noise and comply with stringent EMC and RED requirements for CE certification.

TrailGuard AI

Another camera-based anti-poaching system with satellite connectivity and image processing at the edge is TrailGuard AI, developed by Washington DC-based non-profit RESOLVE in partnership with Intel and Inmarsat.


The TrailGuard AI camera has on-board computer vision intelligence courtesy of Intel's Movidius Myriad 2 VPU chip (inset).

Images: Intel

As with ZSL's Instant Detect, the TrailGuard AI project needed to find a way of reducing false-positive threat alerts from its first-generation cameras, which were funded by the Leonardo Di Caprio Foundation. It did this by incorporating Intel's Movidius Myriad 2 VPU (Vision Processing Unit) into the camera. This low-power (1W) SoC (System on Chip) adds an additional layer of image signal and vision processing intelligence, allowing it to determine when a person or a vehicle is present, rather than something harmless such as a moving cloud or an animal.

The Myriad 2-powered TrailGuard AI camera is designed to perform in the field for up to 18 months on battery power — a big improvement over the much larger original version, which had a separate compute unit requiring time-consuming and potentially dangerous field maintenance every 4-6 weeks.


The compact TrailGuard AI camera (ringed, left) is easily concealed, and therefore much less likely to be vandalised or stolen than a traditional wildlife camera trap (right).


Near real-time alerting is achieved by transmitting images over cellular, LoRa, or satellite connections, depending on what's available in the particular protected area. Satellite connectivity for the TrailGuard AI cameras is provided by Inmarsat, using its L-band network and BGAN (Broadband Global Area Network) satellite modems.


The military-grade Explorer 540 BGAN satellite modem (left) is also easily concealed -- under a fake rock, for example (right). Also shown (arrowed, left) is a small WildTech gateway that receives images from multiple TrailGuard AI cameras via LoRa and transmits them to the modem.


"Wildlife poaching in Africa is at epidemic levels, but despite the best efforts of dedicated rangers, the large park boundaries and rough terrain mean that they often only find out about poaching when it's too late," said Dr. Eric Dinerstein, Director of WildTech and the Biodiversity and Wildlife Solutions Program at RESOLVE, in a statement. 

With TrailGuard AI in place, ranger teams should be able to respond rapidly, intercepting poachers before they strike. Dr Dinerstein told ZDNet: "The biggest advantage of TrailGuard is with its early warning system: it can stop poachers before they lay down snares or shoot wildlife."

The original non-AI version of TrailGuard detected 50 intruders that led to the arrests of 30 poachers representing 20 different poaching gangs, Dinerstein said. "We seized over 1000 kilograms of bushmeat, motorcycles, snares, weapons, and so on in the space of a few months."

Now it's full steam ahead for the Intel-powered AI version. The first 300 hand-built units have been shipped to parks in Africa, Dinerstein told ZDNet, while the next 700 TrailGuards  -- all of which are spoken for -- will be built in China by an Intel ODM.

Download this article as a PDF (free registration required).

Rainforest Connection

Treetops of Dense Tropical Rainforest With Morning Fog Located N

Rainforest Connection repurposes Android smartphones, adding a solar array and a microphone, and installs them in the forest. Audio from these 'Guardians' is streamed to the cloud, where it's analysed for suspicious activity, raising real-time alerts.

Image: Getty Images/iStockphoto

Rainforest Connection (RFCx) is the brainchild of Topher White, a San Francisco-based engineer and developer, who in 2011 was inspired to join the fight against rainforest destruction and climate change after visiting a gibbon reserve in Borneo and witnessing the illegal felling of an old-growth teak tree. The downed forest giant was no more than five minutes' walk from a ranger station, and yet the chain saw went unheard.

As an experienced 'maker', White set about creating an early-warning system based on repurposed Android smartphones that could be installed in the rainforest canopy, run on solar power, listen for suspicious activity (chain saws, vehicle engines) and alert the authorities to a positive match in near real time. Perhaps surprisingly, many rainforest areas -- particularly at the vulnerable edge -- have sufficient mobile phone coverage to make this feasible.

Still, White faced plenty of obstacles: the phones had to be stripped down to a bare minimum of components, an extra microphone attached, and the manufacturer-installed Android OS replaced by (now discontinued) CyanogenMod to ensure that the listening software -- which was created from scratch -- could run. Crucially, a system of solar panels had to be designed to cope with transient spots of direct sunlight in the forest canopy called 'sunflecks'. The result was a petal-shaped configuration of seven panels, spaced to the average diameter of a sunfleck, that could reliably deliver the voltage (5V) required to charge a mobile phone.

Subsequent developments have moved the analysis of the audio captured by these canopy-mounted 'Guardian' devices to the cloud, where Google's TensorFlow machine-learning framework is used to identify sounds of illegal activity and adjust audio inputs to minimise the number of false positives.

The AI system uses 'binary classifiers' to determine whether a section of audio contains signs of logging, and also a 'cognition layer' that extracts intelligence from the entire data set. 

Rainforest Connection currently has 'Guardian' projects in northern Brazil (Guama, Tembé Territory), Ecuador (Cerro Blanco), Peru (Alto MayoTambopata), Romania (Carpathian Mountains), Costa Rica (Osa Peninsula), and South Africa (Cradle of Humankind). You can listen to audio streams from these projects via the Rainforest Connection mobile app on iOS or Android.

As well as detecting illegal deforestation, RFCx seeks to recognise patterns of activity related to poaching, and to pioneer 'bio-acoustic monitoring' of wildlife populations.

Download this article as a PDF (free registration required).

BTO Cuckoo Tracking Project


The UK breeding population of the Cuckoo has crashed in recent decades; the British Trust for Ornithology (BTO) has been tagging and satellite-tracking birds as they migrate to and from Africa to find out why.

Image: Getty Images/iStockphoto

The Common Cuckoo (Cuculus canorus) is the traditional harbinger of summer in the UK and Europe, arriving in late March/April, mating, and laying its eggs in the nests of a range of host species, and leaving before summer's end to overwinter in Africa. During the past 20 years, the UK's breeding population of Cuckoos has more than halved, for reasons unknown. To discover more, the British Trust for Ornithology (BTO) began a tagging and satellite tracking project in 2011.

The BTO doesn't use GPS trackers to keep tabs on migrating Cuckoos because, although accurate, they are currently too large and power-hungry. Instead solar-powered PTTs (Platform Transmitter Terminals) weighing about 5g are used (male Cuckoos weigh around 130g). Solar PTTs transmit short-duration messages to the Argos satellite constellation, which estimates their location using the perceived Doppler shift between successive transmissions. This gives a typical location accuracy of within 500 metres (GPS, by contrast, has an accuracy of a few metres). As satellite tracking technology develops, the BTO hopes to use smaller tags with GPS accuracy, allowing the tracking of females (which weigh around 110g) and finer-grained investigation of habitat use.

To date, the BTO has discovered that after breeding in the UK, male Cuckoos fly south to Africa via one of two routes: to the west through Spain and Morocco or to the east through Italy and the Balkans, before crossing the Sahara and converging on their wintering grounds in the Congo rainforest. On the return journey the following spring, the Cuckoos fly to a previously unrecorded 'stopover' area in West Africa, where they fuel up for the northward Sahara crossing.

Mortality rates have been found to be higher on the western southward migration route, a finding that correlates with trends in the breeding population in the UK. Droughts, wildfires, and habitat change in Spain may be the cause, although food shortages during the breeding season are also implicated. Going forward, the BTO will look further into this differential southward mortality, and also examine the Cuckoo's dependence on resources associated with the Inter Tropical Convergence Zone (ITCZ) during the West African stopover.

The BTO's satellite-tracking work elucidates the natural history of this charismatic bird, and highlights the importance of habitat conservation across its entire migratory life cycle. Furthermore, the Cuckoo's presence in an area has been found to be an effective surrogate for several aspects of biodiversity, making it a good candidate for citizen science projects.

Read more on the BTO Cuckoo Tracking Project: Population decline is linked to migration route in the Common Cuckoo, a long-distance nocturnally-migrating bird.

Citizen science and deep learning


The ability to capture and disseminate large amounts of image and other data across the internet has enabled the rise of citizen science, where the collective pattern-recognition abilities of non-scientists helps with the classification and analysis of large data sets. Zooniverse is a prominent citizen science portal, with 105 projects underway at the time of writing, covering a wide range of fields -- arts, biology, climate, history, language, literature, medicine, nature, physics, social science, and space.

Snapshot Serengeti

One of Zooniverse's longest-running projects is Snapshot Serengeti, which has been classifying images of animals captured by a grid of camera traps in Tanzania's Serengeti National Park since 2010. Multiple users view each image and record the species, number of individuals, associated behaviours, and presence of young with the help of an identification guide. An algorithm then aggregates these classifications to achieve a consensus, a process that has been validated against a 'gold-standard' subset of images classified by experts. After 10 seasons, the Snapshot Serengeti data set contains some 6.7 million images (around 75% of which are empty), with labels provided for 55 animal categories -- the most common being wildebeestzebra, and Thomson's gazelle.

This 'wisdom of crowds' project provides the opportunity to study multi-species dynamics in an ecosystem of world importance, particularly the interactions between large predators and their herbivorous prey. Key to these dynamics is the seasonal rainfall that drives the annual migration of around 1.3 million wildebeest and 250,000 zebra in search of the best grazing.

Large human-tagged sets of image data such as Snapshot Serengeti are also perfect for training deep learning algorithms, which can then be used to automate species detection and classification. Another Zooniverse project demonstrates the value of this approach.

Serengeti Wildebeest Count

The great annual migration

The Serengeti ecosystem is dominated by the annual migration of some 1.3 million wildebeest. Counts are based on aerial photographs, which are analysed by experts, citizen scientists and now deep learning algorithms.

Image: Getty Images/iStockphoto

As noted above, wildebeest, along with zebra, are a keystone species in the Serengeti ecosystem, so it's important to know about their population dynamics. Wildebeest are counted every two or three years by flying transects over the herds in March-May, when the population is mostly on short-grass plains (where they are most easily seen from above), and taking aerial photographs under controlled conditions. Manual counts of these images can take several experts several weeks, so the process has been outsourced to Zooniverse's Serengeti Wildebeest Count project, where images are counted multiple times by citizen scientists, tallies combined into a statistical model, and a final population estimate calculated. 

In pursuit of even greater efficiency, a paper by Colin J Torney and others used this data to train automated machine learning algorithms, comparing the resulting counts with those generated by citizen scientists and by experts.

The study was based on 1,584 georeferenced images with a resolution of 7,360 by 4,912 pixels. These large images were divided into 12 tiles and uploaded to the Serengeti Wildebeest Count project, where 2,212 citizen scientists counted the number of wildebeest in each tile 15 times. For automation, the authors employed the YOLOv3 object detector and the Keras and TensorFlow open-source deep learning packages, using 500 randomly selected images for training. This resulted in a list of locations for potential objects in each image, which were then filtered by discarding detected objects that did not match up to an identification in the Zooniverse data.

Training took 34 hours on a powerful Nvidia Quadro GP100 GPU, whereupon 1,000 randomly selected survey images were processed by the (slightly modified) algorithm — a process that took two hours on the same Nvidia GPU platform. The deep learning and citizen science counts were then compared to a 'gold standard' estimate made by a single human expert, which was taken to be the true number of wildebeest in each image.

Download this article as a PDF (free registration required).

The citizen science and deep learning methods can both deliver 'highly accurate image counts', the researchers found. The best citizen science fit to the expert count was achieved when the 5 lowest of the 15 counts for each image were discarded and the mean taken of the remaining 10, suggesting a systematic tendency to undercount (in this project, at least). The deep learning algorithm delivered its results much faster, and within 1% of the experts' -- 1.7 miscounts per image on average, recording a total 20,631 wildebeest compared to the expert estimate of 20,489.

"The 1,000 images can be processed in under 2 hours, meaning every future census could be counted within 24 hours. Hence, a process that currently takes 3-6 weeks, involving 3-4 wildlife professionals and countless cups of tea, can potentially be replaced with an automated system that runs overnight," the researchers said.

Earth observation projects

Landsat and forest loss/fragmentation

Perhaps the broadest definition of a sensor-laden IoT 'thing' is a satellite that observes changes on Earth from orbit, beaming data back to ground stations for distribution to scientists and policy-makers around the world. 

Landsat, a NASA/USGS project, has been collecting high-resolution multispectral images of the Earth's surface to aid decision-making on land use practices since 1972 with Landsat 1. The current satellite is Landsat 8, with Landsat 9 due to launch in late 2020. 


A Landsat 8 Operational Land Imager (OLI) image of pastures in South America's Gran Chaco plain, east of the Pilcomayo River near Tezén in Paraguay, captured on 14 August 2016. The rectangular clearings in the dry forest are created by large-scale cattle-ranching operations.


Landsat data is freely available, and several projects have used it to investigate changes in forest cover around the world.


Time-series (2000-2012) analysis of Landsat images showing forest extent (green) and change (red=forest loss; blue=forest gain; purple=loss and gain). 

Image: University of Maryland

A landmark 2013 study led by the University of Maryland's Matt Hansen used Google's Earth Engine cloud platform to map the extent of global tree cover, plus losses and gains, between 2000 and 2012 at a resolution of 30 metres. This was a big job for Earth Engine, to say the least: 20 terapixels of data from 654,178 Landsat 7 images were processed using one million CPU-core hours on 10,000 computers. "When they ran it, the lights dimmed," Hansen told the New York Times.

When the lights came back on, Hansen et al found that the world had lost 2.3 million square kilometres of forest between 2000 and 2012 and gained just 0.8m km2, with 0.2 million km2 experiencing both loss and gain.

At the time of the 2013 study, losses in the Brazilian rainforest had been declining for a decade, while the opposite was true in Indonesia. Since then, as documented by Global Forest Watch (a WRI-funded initiative to which Hansen's research group contributes), the trend has reversed with Brazil showing a marked increase and Indonesia a dramatic decline in primary forest loss.

Images: Global Forest Watch

In an April 2019 blog post, GFW noted that the decline in forest loss in protected Indonesian forests was especially dramatic, reflecting the success of recent government policies. However, it warned that 2019, an El Niño year, would likely see dry conditions and a prolonged fire season. In October 2019, GFW reported that 2019's fires were the worst since 2015 -- the last time Indonesia experienced an El Niño weather pattern.

GFW estimates that annual gross carbon dioxide emissions from tree cover loss in tropical countries averaged 4.8 gigatons per year between 2015 and 2017. This means that, if tropical deforestation were a country, it would rank third behind China and the US in CO2 emissions.

Image: Global Forest Watch

To keep tabs on these issues, Global Forest Watch collates several data sources into an interactive environmental early warning system. These include deforestation alerts from the University of Maryland's Global Land Analysis and Discovery (GLAD) lab, and VIIRS fire alerts from infrared sensors on weather satellites.


Early warning system: Deforestation alerts from the past year (purple) and a week's worth of fire alerts (yellow/red/black) in South America.

Image (from 11 October 2019): Global Forest Watch

Deforestation tends to go hand-in-hand with the fragmentation of remaining areas into ever smaller patches. The level of fragmentation affects wildlife in several ways: key species will have minimum habitat area requirements; reduced connectivity among remaining fragments means a smaller chance of recolonisation should a local population go extinct; and edge effects tend to be more severe in smaller fragments.


Landsat images of Amazon rainforest in Rondônia, Brazil, from 1975 (left) and 2012 (right). A major north-south road spawned secondary roads at right angles as settlers cut and burned the forest and established farms, creating a distinctive pattern of fragmentation. 

Image: Landsat/NASA


Large-scale Earth-observation projects like NASA's Landsat and the ESA's Sentinel programme deploy very large and very expensive satellites with relatively long repeat cycles (16 days in the case of Landsat 8, for example). At the opposite end of the scale are CubeSats -- constellations of small, inexpensive satellites, often built with off-the-shelf components and open-source software, launched as secondary payloads or via the International Space Station (ISS) -- which can deliver high-resolution imagery on a much shorter cycle.


Planet's CubeSat satellites, called 'Doves', measure just 10x10x30cm and weigh 4kg. The company currently has a constellation of 120+ Doves in sun-synchronous orbit at an altitude of 475km.

Images: Planet

Planet is a leading exponent of what co-founder and CEO Will Marshall, an ex-NASA scientist, calls 'agile aerospace'. Founded in 2010, Planet numbers Google among its equity stakeholders following the acquisition of Terra Bella and its SkySat constellation in 2017. The company currently has over 150 satellites in orbit, comprising 120-plus PlanetScope 'Doves' (10x10x30cm, 4kg, 3m resolution), 15 SkySats (60x60x95cm, 110kg, 72cm resolution), and 5 RapidEye (<1m3, 150kg, 5m resolution) devices. Orbiting the poles every 90 minutes, Planet's constellation can image the entire land surface of the Earth every day, providing data suitable for use cases including mapping, deep learning, disaster response, precision agriculture, and temporal image analytics.

Here's an example of Planet's imagery documenting illegal gold mining in the Peruvian rainforest:


In 2016 the 'La Pampa' gold mine illegally expanded into the protected Tambopata National Reserve in Peru. The Amazon Conservation Association used Planet imaging data to publish a series of alerts which resulted in government intervention.

Images: Planet

What next?

The current focus on biodiversity loss and climate change is the latest manifestation of fears, traceable back to Malthus in the 18th century, that the combination of human population growth, resource consumption per capita (particularly in developed countries), and consequent degradation of the natural environment will so reduce the planet's carrying capacity that we risk apocalyptic consequences -- war, famine, disease, extinction.

Technology is often held up as a potential 'fix' for the negative effects of such developments, and the Internet of Things -- in its broadest sense -- can play an important role by providing timely and actionable information on the state of the natural environment.

From earth-imaging satellites mapping land use changes, to citizen scientists and deep learning algorithms monitoring species' population changes, to smart cameras detecting poachers in the bush, to sensor-tagged animals revealing details of their life histories, to smartphones listening out for chain saws in the forest -- the potential to install a regime of benign surveillance over the natural world is immense.

As biodiversity loss and climate change rise up the global agenda, businesses are beginning to take notice -- and sometimes action. A 2018 study led by Oxford University's Department of Zoology found that nearly half (49) of the top 100 companies from the 2016 Fortune 500 mentioned biodiversity in their reports; 31 made clear commitments, although only 5 were 'specific, measurable, and time bound'. 

Meanwhile, on climate change, Science Based Targets -- a collaboration between CDP, the United Nations Global Compact (UNGC), World Resources Institute (WRI), and the World Wide Fund for Nature (WWF) -- reports that 732 companies are taking science-based climate action and 312 have approved science-based targets (as of mid-January 2020). CDP's 2019 A-List of 179 companies with "transparent and comprehensive disclosure of climate data, thorough awareness of climate risks, demonstration of strong governance and management of those risks, and demonstration of market-leading best practices" includes many global brands. (CDP runs similar lists for forests and water security.)

Governments must play their part too, of course. As of January 2020, 169 Parties (out of 196) have submitted NBSAPs (National Biodiversity Strategies and Action Plans) under the UN's 1992 Convention on Biological Diversity (CBD). Of the 169 NBSAP-submitting Parties, all but 13 take into account the CBD's 2011-2020 Strategic Plan for Biodiversity, which includes 20 Aichi Biodiversity Targets. Target 11 is among those aimed at improving the status of biodiversity:

"By 2020, at least 17 per cent of terrestrial and inland water, and 10 per cent of coastal and marine areas, especially areas of particular importance for biodiversity and ecosystem services, are conserved through effectively and equitably managed, ecologically representative and well connected systems of protected areas and other effective area-based conservation measures, and integrated into the wider landscapes and seascapes."

A notable absentee from the list of countries with an NBSAP is the United States

The Paris Agreement of December 2015 builds on the UN Framework Convention on Climate Change (UNFCC) to 'accelerate and intensify the actions and investments needed for a sustainable low carbon future'. Its central aim is to keep global temperature rise well below 2 degrees Celsius over pre-industrial levels, with a target of 1.5 degrees. All Parties to the agreement are required to make Nationally Determined Contributions (NDCs), with global stocktakes occuring every five years. To date, 187 out of 197 Parties have ratified the Paris Agreement

The ten absentees are: Angola, Eritrea, Iran, Iraq, Kyrgyzstan, Lebanon, Libya, South Sudan, Turkey, and Yemen. On 4 November 2019, the United States notified the UN of its decision to withdraw from the Paris Agreement, which will take effect on 4 November 2020.

There's plenty of technology and expertise available to help combat biodiversity loss and climate change. However, implementing that technology and expertise effectively around the world is another -- increasingly urgent -- matter.

Download this article as a PDF (free registration required).

Also see

Editorial standards