Hot on the tracks of feral animals in the Top End

Two men in cart with drone beside it

It’s easier to find a feral buffalo than a mobile signal in Arnhem Land, but that hasn’t deterred a team of researchers and Indigenous rangers from connecting a bunch of the region’s feral animals to the internet.

Buffaloes, along with feral pigs, are having a devastating impact on wetlands and important cultural sites in the Djelk Indigenous Protected Area in east Arnhem Land. Feral animals can really wreak havoc, destroying wetlands and creeks, causing declines in water quality, spreading diseases that can be passed on to livestock, and even destroying infrastructure like roads and fences, which are so important in remote areas.

Past attempts at controlling these introduced pests across the Top End have had limited success but recent advances in digital technologies might be about to change that.

Hot under the collar

Hand holding electronic device

A new technology we’ve developed with JCU and Indigenous Ranger groups in Arnhem Land and Cape York, is bringing invasive species monitoring into the internet age and allowing for real-time monitoring of the animals. The system uses new electronic GPS trackers attached with animal collars and an array of environmental sensors embedded in the landscape. Telecommunication infrastructure then receives the signals from the trackers and sensors, to deliver information about the feral animals and the habitats they frequent.

The GPS trackers, developed in-house by our Data61 researchers, were made at a fraction of the cost of commercially-available alternatives. This has opened up the prospect of conducting large-scale tracking projects that simply haven’t been economically feasible before. They have an on-board wireless communication system, called LoRa, which is low-power and long-range. Both the GPS trackers and the environmental sensors upload their data to our Senaps cloud service in near real-time, as the animals are roaming around.

Shadow of helicopter and animals running

Working on country

All the elements of the tracking system are things that work readily in our well-connected cities, like in our Brisbane workshop where the initial testing took place. But what happens when you take it out into some of the remotest areas of Australia?

When the team took their equipment to the Djelk Indigenous Protected Area in Arnhem Land in November 2017, they found that testing technology out in the scrub of the Northern Territory presented a unique set of challenges. While everything worked in the town of Maningrida (500 km east of Darwin), as soon as the team ventured out of town to the vast floodplains where Buffalo and pigs can be found, they found themselves out of range of mobile networks. A hill that was picking up 3G signal seemed a promising place for their communications base station, but upon further testing discovered the signal was too intermittent to work.

Group of people

This is where the local expertise of the Djelk Rangers came in. The system would only work if it could connect to communications infrastructure and if their own infrastructure could be placed close enough to where the animals are. The rangers, with their intimate knowledge of the landscape, are able to advise on the places where the feral animals are most frequently found, allowing for optimal positioning of sensors and communications equipment. They also have the unique on-the-ground skills required to track down and collar the animals.

Horned buffalo in the bush

In the end, the team were able to connect to Maningrida’s WiFi signal and use a series of relay nodes to connect up the environmental sensors and GPS trackers with the communications base station. Ecologist Eric Vanderduys took a GPS tracker for a trek through the bush to confirm everything was online, with the team watching on as his tracks appeared on their monitoring equipment. Eric’s trek would prove a valuable rehearsal run for the main event, which came last week, when the Kalan Rangers of eastern Cape York managed to capture and collar a feral pig with the help of the project team. We are pleased to report that we are now able to follow the movements of Nic the pig, named after one of Data61’s developers of the tracking technology, Nic Heaney.

Satellite tracking image

Hitting the hot spots

The data that will be collected by this system will also feed into a model that is able to ‘learn’ the behaviour of the animals over time, using machine learning, and eventually to predict what their future behaviour will be. The model uses the data from the environmental sensors, which capture things like the temperature and humidity of particular parts of the landscape, together with the data about the animals’ movements to learn where the animals are likely to go under specific weather conditions. It might tell us, for example, that in the heat of the day feral pigs head for a particular cool, shady spot.

Working closely with environmental managers like the Indigenous Ranger groups is ensuring that the new tracking system is co-designed by the people who will need to use it to manage the impacts of feral animals. This means that the technology and the insights it delivers will be ready to use in the areas where it is most needed, sooner, and has the potential to change our long-term management strategies for feral animals impacting the outback.

Aerial view of muddy landscape

And why limit it to remote Australia? There’s no reason why this system can’t go global too, helping manage ecosystem threats around the globe.

Original article –

How researchers are mapping an invasive species advancing across an entire region

computer image of trees
 High-resolution 3D mapping of gamba grass invasion front with terrestrial LiDAR.

BY definition, the notion of invasive species is overwhelming, where one species, by dint of human interference, has an unnatural advantage over a finely-tuned ecosystem. How, then, do you confront an introduced species taking over thousands of square kilometres of country?

Northern Australia is vast and home to the largest intact savanna region in the world. It is across this region that the proliferation of an invasive weed is so extensive researchers are now looking to satellites to track its spread.

Gamba grass (Andropogn gayanus) was originally introduced to Northern Australia from the savannas of Africa in the 1930s as a pasture grass, but has since proliferated across approximately 15,000 square kilometres.

It has the potential to cover over 380,000 square kilometres of the Northern Australia tropical savanna, including parts of the Northern Territory, Western Australia and Queensland.

Researchers are currently working on the most efficient way to halt the spread of the invasive species, including using break-through technologies to map where it occurs and the rate at which Gamba grass is advancing across Northern Australia.

Changing the environment

Gamba grass is a perennial grass that forms dense tussocks up to 70cm in diameter and can grow up to four metres tall. It is a Weed of National Significance, and deemed a weed through Northern Australia. Its greatest threat is in the intense fires the dense biomass generates, dramatically increasing fire fuel load and posing a major threat to biodiversity and carbon storage in northern savannas.

CSIRO Principal Research Scientist and Associate Professor at Charles Darwin University Dr Shaun Levick has witnessed the dramatic ecosystem effects of Gamba grass invasion in both Brazil and Australia.

tall grasses
               Gamba grass grows up to four metres tall. Image : Btcpg/CC BY-SA 3.0

“Gamba grass is changing the behaviour of the fire regime in the tropical savanna regions it invades. Native grasses are typically much shorter and burn at a lower intensity, but Gamba can grow up to four metres tall and carries flames into the tree canopies, leading to extensive mortality,” says Levick.

The high-intensity fires caused by Gamba are also a major threat to savanna burning emissions abatement initiatives in the region.

“Once it spreads and occupies dense stands, it’s incredibly hard to remove, and manual clearing and chemical spraying are expensive and time-consuming work,” he says.

Fieldwork in Gamba-infested sites is formidable, the stands tower above and the wall is almost impenetrable. Fire risk is ever-present and is a serious threat to properties and lives in rural blocks.

“We need a better understanding of its current spatial distribution and of the rate at which it is spreading to coordinate control measures.”

Modelling the landscape through lasers

Levick joined CSIRO early in 2017 and brings an international collaboration with the German Max Planck Institute for Biogeochemistry, with cutting-edge technology and computer vision methods transforming traditional field-based ecology.

He is currently working with CSIRO’s Dr Garry Cook to monitor landscapes and vegetation across Northern Australia using a remote sensing system called terrestrial LiDAR (Light detection and ranging).

man with laser equipment in bush
Formidable field work: Dr Shaun Levick operating the RIEGL VZ-2000 terrestrial LiDAR system for vegetation monitoring in the Kimberley region of Northern Australia.

LiDAR systems can be operated from the ground, light aircraft/UAVs, and even satellites, but ground-based sensors offer by far the greatest resolution for extremely detailed reconstruction of ecosystem state.

LiDAR sensors emit energy in the form of a pulsed laser that bounces off objects that it strikes to provide an in-depth 3D model of the target area up to 2,000 metres away.

LiDAR sensors produce a 3D point-cloud of the surrounding area and can be used to map the structure of terrain, trees and even Gamba grass due to its large size.

high res 3D laser image of two trees
3D rendering of Boab trees in the Kimberley, WA. Colours represent height above ground level. Point-cloud data collected with high-resolution terrestrial LiDAR system, resolving fine detail in branches and grasses.

Levick says LiDAR shows good potential for monitoring the invasive species because of its high-resolution and versatility as a surveying tool compared to previous field-based methods.

The difference between LiDAR and more common optical image remote sensing systems, which capture the sunlight reflected off an object, is that the LiDAR sensor generates its own pulse of light energy, allowing for greater control of data to be collected, and operation at night.

computer generated image of tree canopy
The advantage of scanning at night is there is reduced wind movement of tall thin trees, and grasses, which sway more in the breeze. This image shows a canopy where if the wind blows the trees sway side to side which interferes with measurements, night time scanning gives cleaner results.

By analysing the shape of the pulse returned to the LiDAR system, researchers are able to distinguish differences in the surfaces it’s hitting from the reflective properties of the return, of which colour, shape and distance can be determined.

This data is georeferenced from a high precision GPS and provides accurate coordinates from where the data was collected, allowing researchers to accurately assess conditions in specific areas, explore changes through time, and to cross reference it with data from other sources – like satellites.

The future of LiDAR through satellites

Given the vastness of northern Australia, the only feasible approach for regular, wall-to-wall mapping of ecosystems is via satellite imagery.

Levick’s research is currently focusing on upscaling datasets to allow satellites to detect the Gamba grass through the use of RADAR remote sensing.

“We have seen some exciting developments in RADAR and optical sensors in recent years, including the current development of NovaSAR – an S-band RADAR satellite from Surrey Satellite Technology Limited that CSIRO is in partnership with,” says Levick.

computer generated image of satellite
NovaSAR-S will image Earth in all weather conditions, both day and night (computer generated image). Currently, the expected launch window for the NovaSAR Satellite is Q1 2018. Image: SSTL

Through a process known as machine learning, Levick aims to train satellites to detect the Gamba grass and provide a much-needed edge to researchers in the region where satellites can provide regular wall-to-wall datasets at regional scales.

Dr Levick says, “The terrestrial LiDAR scanner will be used to reconstruct ecosystems with varying degrees of Gamba invasion in 3D, and these models will be used to train satellite-based RADAR sensors to detect the presence of high-biomass grasses, such as Gamba.”

This process is done through modelling areas which are populated with the Gamba grass species, then overlaying it with satellite data from the same area to “teach” the satellites to recognise the structural properties of the vegetation in the landscape. This research forms part of a collaboration with Associate Professor Samantha Setterfield, from the University of Western Australia, and Dr Natalie Rossiter-Rachor, from Charles Darwin University, through the Northern Australia Environmental Resources Hub of the Australian Government’s National Environmental Science Program (NESP).

“The success of any satellite mission revolves around the training and validation of space-borne measurements with ground-truthed data in the field. The high precision 3D measurements that we are collecting will prove extremely valuable in this context, and will provide critical validation of both current and upcoming spaceborne missions such as NASA’s GEDI and NovaSAR,” says Levick.

The results

He adds, the development of the multisensory modelling system is critical to the success of the project as the size of Northern Australia means that current capacities within the monitoring system are insufficient in providing the data needed to manage the species efficiently in the region.

The accuracy and scope of data provided from the LiDAR system is allowing researchers to gain a more in-depth understanding of the areas they’re monitoring and can drastically reduce the time and resources required to combat the species.

Other uses for the technology include monitoring carbon and fire management, mining legacy rehabilitation, rangeland condition assessment, and biodiversity and habitat conservation.