A growing archive of photos, videos and recordings shared by the public is transforming conservation science.
We know almost nothing about the subspecies of White-faced Storm Petrels living on a remote islet off the coast of New Zealand, except that they're extremely rare.
Often referred to as Kermadec Storm Petrels, their range is a mystery, their breeding patterns are poorly understood, and most of what we know has been assumed based on similar seabirds.
They were first described in 1951, but their similarity to other White-faced Storm Petrels means they're hard to reliably identify at sea, posing a major challenge for conservation scientists.
"If you can't identify something, you can't know where it exists and where it doesn't," says Simon Gorta, a PhD candidate at UNSW.
"Everything in conservation is compromised if we can't identify the things we are trying to conserve."
Gorta calls this the "identification gap", and it's a problem that vexes conservation efforts for countless species, he writes with co-author Peter Allen, Monash University PhD student, in the journal Diversity and Distributions .
For a long time, the problem was a lack of data.
Now, researchers are using citizen science to answer questions that were once impossible.
Tapping into this vast public archive, Gorta and Allen have, for the first time, confirmed a reliable way to identify this seabird subspecies.
And they think their technique could be used by other scientists to identify all sorts of threatened, rare, or little-understood species.
"Having a reliable way to identify Kermadec Storm Petrels, and potentially countless other species, should give conservationists another tool to help respond to the threats they face," says Gorta.
Little bird in a big world
Scientists think there are about 100 to 300 pairs of Kermadec Storm Petrels living in a single colony on Haszard Islet in the Rangitāhua/Kermadec Islands.
The Kermadecs are a string of islands and islets about 1000 kilometres north-east of Aotearoa New Zealand's North Island and are among the most important seabird breeding sites in the South Pacific, with millions of birds calling the islands home.
Kermadec Storm Petrels weigh around 70 grams - about the same as a small apple - and spend their lives skipping over the choppy waters of the Coral and Tasman seas.
Haszard Islet itself is about 200 metres across and is so steep and rugged that it is only accessible during rare good conditions, making gathering data extremely tricky.
And because these birds potentially travel over vast, isolated stretches of ocean, it's almost impossible to conduct field surveys for them at sea.
"It's hard work, especially out at sea," says Gorta.
So Gorta and Allen turned to a growing archive of photographs taken by citizen scientists - everyday members of the public armed only with a camera and a keen eye.
"Lots of bird-watchers and bird photographers are much better with a camera than we scientists," he says.
This means that scientists can access research-quality data from vast image libraries with just a few keystrokes.
The team analysed a dataset of more than 1000 individual White-faced Storm Petrels across the South West Pacific, drawing on thousands of photographs taken by members of the public and archived on citizen science platforms like Macaulay Library , via eBird .
These birds represented individuals from three different subspecies with subtle differences in feather colours, patterns, and shape, which the team formally tested and confirmed using photographs for the first time.
There are many ways to tell these birds apart, but they vary, and until now no one has properly tested how well they actually work.
The team complemented these photographs with others from the internet and scientific literature, as well as specimens from the Australian Museum.
From these images, the researchers manually scored subtle physical traits - things like rump colour and tail shape - and used statistical modelling to work out which features reliably distinguish the species.
Once classified, those records could be mapped, revealing birds travelling across the Tasman and Coral seas, and Kermadec Storm Petrels occasionally visiting Australia's east coast, offering a clearer picture of how and where they move.
That's vital information that could inform conservation efforts, says Allen.
"As we now know they're crossing the Tasman and Coral seas, we can start to work out what threats they might be facing away from their breeding grounds, such as the impacts of climate change."
Bridging the 'identification gap'
There are plenty of organisms like the Kermadec Storm Petrel that we can't identify, and Allen thinks the process the researchers are outlining here could help.
"You could also use it for other animals, plants, anything really, as long as you can measure traits from media," he says.
The team here measured the colour, patterns, and shape of the feathers of birds, but if you can measure any trait, you could adapt their tool to identify practically any organism.
And researchers are not limited to just images.
"You could use recordings of vocalisations that animals make, or photographs of flipper patterns in the case of some penguin species.
"We currently rely on flipper patterns, among other features, to identify rockhopper penguin species, but it's very inconsistent, and there seems to be a lot of overlap among species.
"An approach like this would probably help resolve that one way or the other."
A big data problem
Type "Australian Magpie" into eBird and it shows up 1.61 million observations, including nearly 30,000 with photos.
Over on iNaturalist - another citizen science database - for the same bird there are nearly 60,000 identifications covering Australia, New Zealand, Fiji, and New Guinea.
And that's just one bird.
Around the world, databases such as these host hundreds of millions of individual observations of everything under the sun.
Members of the public see a cool bird, bug, plant, or animal (or mushroom), snap a picture, and upload.
That's a lot of observations, and it's resulting in a quiet revolution in conservation biology.
Scientists are working out how to tap into this archive, to fill in the blanks where field studies aren't possible.
"We are getting quite good at it," says Gorta.
"We've got the computing and statistical capacity to do it, and in universities we're increasingly trained on how to work with these kinds of datasets - there's a lot of expertise."
Approaches like Gorta and Allen's are changing how scientists work, opening up remote and hard-to-study environments through vast public datasets.
As that archive grows, Gorta says, the chance to fill in the gaps grows alongside it.
"We hope this approach will be used to solve other long-standing identification problems that affect the way we conserve so much of our wildlife," he says.
