photo by: Boone Bradley / Kansas News Service

The BeeMachine app identifies a bee visiting a suburban flower garden in Lawrence as a two-spotted bumblebee. The app uses artificial intelligence to distinguish among species, some of which are very difficult to tell apart.

Just a few decades ago, bee enthusiasts across much of North America could count on spotting one of the continent’s most common bumblebee species buzzing from flower to flower.

Today the American bumblebee is in trouble. Its numbers have dropped sharply, and it has vanished entirely from large swaths of its range.

Yet the fact that biologists even know of this pollinator’s plight marks a key step toward helping it, because population trends steer conservation efforts.

By contrast, scientists remain in the dark about how most of the other estimated 4,000 bee species in North America are handling habitat loss, pesticides, global warming and other challenges.

A new smartphone app called BeeMachine harnesses artificial intelligence to tackle a key hurdle to figuring this out: Right now, experts struggle to tell many species apart.

“It’s a huge problem,” said entomologist Brian Spiesman, the app’s creator and a professor at Kansas State University. “We bring back a few hundred specimens (from fieldwork) and we spend much longer identifying them in the lab than we do actually collecting them.”

Bee ecologists mail tricky specimens — many species are nearly identical and tiny as gnats — to specialized taxonomists.

But these taxonomists are in short supply, so Spiesman and his collaborators are training artificial intelligence to help. As an added bonus: The app lets the public participate in documenting bees, too, by snapping photos when they spot one.

“This type of citizen science has the potential to get more eyes out there sighting bees than any single study could ever hope for,” Spiesman said. “Better tools for crowdsourcing are really important.”

The public’s sightings can offer valuable intel on which bees live where.

In the Midwest, for example, a hiker wandering trails or a gardener scouting their flower beds could find a Southern Plains bumblebee or an American bumblebee, both of which are currently under review by the U.S. Fish and Wildlife Service for potential listing as threatened or endangered.

A separate project from Cornell University, E-Bird, has already proven the power of large-scale public participation by turning passionate birdwatchers into a wellspring of data for avian research and conservation efforts.

So far, BeeMachine can distinguish between more than 350 kinds of bees at the species or genus level.

The goal: teaching it to identify all of the estimated 20,000 bee species worldwide.

That will require international collaboration, though, because Spiesman and his colleagues need high-quality photos of accurately identified specimens to train BeeMachine.

Bees without borders

Spiesman and his colleagues corral photographs from museums and other reliable sources around the world.

They’ve taken thousands of images themselves and have pulled others from international projects such as the Global Biodiversity Information Facility, a multigovernment-funded online repository of species data.

Natural history museums hold a treasure trove of specimen collections that have already undergone painstaking identification. Those specimens let BeeMachine correctly learn the minutiae of tricky species.

Yet getting images of these specimens remains a challenge. It can require a lot of specimen handling and imaging work.

“Museum staff often don’t have the time to pull specimens and definitely don’t have time to photograph them for us unless we have an existing collaboration that is already funded,” Spiesman said.

As BeeMachine grows and gets smarter, it has the potential to collect and accurately identify sightings on every continent and make the information available to researchers globally.

Already, the project partners with data collection efforts in Japan and Argentina, for example.

Ultimately, Spiesman hopes that BeeMachine will let scientists identify more bees in the field without needing to kill the creatures and scrutinize them under microscopes.

Researchers could, for example, aim a camera at a flower and leave it there to gather images of foraging pollinators for BeeMachine to analyze.

So far, data gathered by BeeMachine isn’t viewable online, but that will change soon — likely this month. Users will be able to view each other’s sightings on the project’s website.

Popular naturalist apps, such as Seek, do a good job of identifying the largest and most common bees, but Spiesman says ecologists need a more powerful tool to tell apart the many tiny, near identical species that exist.

“We are not replacing taxonomists at the rate that they’re retiring,” he said.

Taxonomy requires specialized expertise and rigor. Taxonomists say their field is widely underappreciated and misunderstood, exacerbating the shortage of professionals.

How many kinds of bees exist?

About 90% of plants can’t reproduce through wind pollination. They depend on animals to do the work, and bees rank among the most important of pollinators.

Yet bees remain as mysterious as they are important.

Scientists are still figuring out how many species of them exist. Estimates vary, but they commonly hover around 20,000 worldwide and 4,000 in North America.

That staggering biodiversity is typical of invertebrates. For comparison, North America has fewer than 500 mammal species.

But what scientists know so far skews toward bigger species that are easier to observe and identify. Thousands of small bee species remain poorly understood.

Case in point: About 40% of the bee species assessed so far by the International Union for Conservation of Nature are bumblebees. This is despite the fact that bumblebees constitute a tiny sliver of the bee universe.

Based on the documented plight of some of the world’s best studied bees, though, scientists worry that their smaller counterparts could struggle, too.

Habitat loss and climate change have hit many insects hard. Insecticides inadvertently poison beneficial pollinators, and herbicides lead to fewer wildflowers on many farms.

Microscopic parasites spread to wild bee species from infected European honeybee hives that people ferry from region to region to pollinate crops.

Introduced honeybees compete for food in areas without enough of it to go around, and wild, native bees face undernourishment.

On some other continents, introduced bumblebees cause similar problems for native pollinators.

— Celia Llopis-Jepsen reports for Kansas News Service.