Eyes on Recovery

Woman adjusting equipment on tree

Abby Hehmeyer checks a camera trap in a fire-impacted area.

Assessing the Devastation with Data

The 2019-2020 Australian bushfire season was the most catastrophic in the country’s history and caused widespread damage to people and nature, the impacts of which will be felt for years to come. More than 24.7 million acres of land burned and an estimated 3 billion animals were harmed or killed, including important native wildlife such as kangaroos, wallabies, koalas, wombats and echidnas.

Map showing Australia fire locations

Map of the Eyes on Recovery project sites in southern and eastern Australia.

In the wake of the fires, it was unclear if, and when, Australian ecosystems would recover. In response, WWF-Australia, WWF-US, and Conservation International, supported by a $1 million grant from Google’s philanthropic arm Google.org, launched Eyes on Recovery, the most extensive post-fire monitoring program ever initiated in Australia.

To undertake a monitoring effort at this scale, WWF-Australia convened a wide range of on-the-ground project partners, including Parks and Wildlife Services, university researchers, environmental non-profit and natural resource management organizations. By combining many distinct projects together under a single, overarching initiative, the Eyes on Recovery team was able to deploy cameras in eight fire-affected regions across southern and eastern Australia, and bring those data together to provide a more complete picture of the impacted ecosystems.

For the past two years, these cameras have been our eyes on the ground, allowing us to answer questions about the fire impacts and recovery for species like the Kangaroo Island dunnart in South Australia, brush-tailed rock wallabies in Queensland and koalas on the north coast of New South Wales.

  • A wombat and its baby look at the camera

    Wombat and joey

  • A swamp wallaby looks at the camera

    Swamp wallaby

  • Dingo pups wander in a small open area in the forest

    Dingo pups

  • Three short beaked echidnas walk toward the camera

    Short-beaked echidnas

  • A koala captured by a camera trap on a sunny day


The Power of Artificial Intelligence

Camera trap monitoring programs like Eyes on Recovery generate incredibly large datasets. More than 1,100 cameras have been deployed across Australia as part of this project, resulting in over 7 million images! These photos contain valuable information about wildlife recovery, but the sheer quantity of data make it slow to process and analyze, which can prevent quick and effective management actions on the ground.

Therefore, images captured by the cameras were sorted and analyzed in Wildlife Insights, a groundbreaking cloud platform that uses artificial intelligence to identify the animals detected in each photo. 

For this project, Wildlife Insights implemented an updated AI model, specifically trained to identify more than 150 different Australian species. Like humans, AI models get better at recognizing and identifying animals, once they’ve looked at hundreds or thousands of images. This model was trained on over 3 million Australian images!

“The new Wildlife Insights AI model has dramatically reduced the time ordinarily required to process camera trap images,” said Emma Spencer, WWF-Australia program coordinator for Eyes on Recovery. “Having a computer do the arduous task of removing blank images from the data, and identifying important species, many of our partners are reporting at least a five- to 10-times increase in their image processing speeds. That means data which usually takes 500 hours to sort, is now being processed in around 100 hours.

The Wildlife Insights AI accurately detects a threatened brush-tailed rock wallaby.

Visualizing Impacts and Recovery

After data are collected, it is important that land managers and researchers have a means to quickly visualize and understand fire impacts and wildlife recovery in post-fire environments. Therefore, the Eyes on Recovery team built an online tool, which is integrated with Wildlife Insights, that lets you easily explore and compare data across the project sites. This tool allows users to visualize the number of species present in a site and the monthly species detection rates—the average number of independent observations for a species during the study period each month. It also provides a graph of each species’ daily activity patterns—how often a species is observed at different times of day—and lets users quickly compare between two different species or two different time periods.

So far, many threatened species have been detected across the project, including those that had a high proportion of their habitat burnt, such as Kangaroo Island dunnarts, long-nosed and long-footed potoroos, brush-tailed rock wallabies, and superb lyrebirds.

“While species recovery generally takes longer than two years in fire impacted environments, the project has already detected threatened species such as the Kangaroo Island dunnart and brush-tailed rock wallaby in encouraging numbers,” said Tracy Rout, conservation analyst for WWF-Australia.

Graph showing the detection rate for the superb lyrebird across multiple project sites, in burnt and unburnt habitat.

Management Actions

Armed with this information, project partners have hit the ground running. For example, in many sites, monitoring data have informed when and where to manage invasive species. Data have also informed how fire will be managed, such as the timing and location where prescribed burning will be done. At some sites, threatened species have been detected for the first time, allowing managers to plan for their recovery.

On Kangaroo Island in South Australia, land managers from the Kangaroo Island Landscape Board are now using Eyes on Recovery cameras to test artificial shelters for Kangaroo Island dunnarts. This critically endangered species had 98% of its habitat burnt in the 2019-20 fires. Now, cameras are being used to monitor the popularity of different types of artificial shelters to protect dunnarts from predators when vegetation is sparse following a wildfire. Shelters, made of tiles or sheets of corrugated iron, are distributed on the ground and cameras are used to see which ones the dunnarts prefer to hide under. The hope is that this research will enable effective shelters to be deployed both during and after wildfires and hazard reduction burns.

“In the long-term, data from Eyes on Recovery will inform management actions to increase the resilience of native species to wildfires, such as increasing habitat connectivity to allow species to easily move from burnt areas into unburnt habitat,” said Abby Hehmeyer, program officer for WWF-US. “Our hope is that these actions will help to ensure the conservation of Australia’s wildlife, even in the face of future fires.”