For the first time in four centuries, it’s good to be a beaver. They have long faced persecution for their fur and were considered pests, but the rodents that build dams are now hailed by scientists as environmental saviors. Their ponds and marshes store water in the face of drought, filter pollutants, provide habitat for endangered species, and combat wildfires. In California, the value of Castor canadensis is so great that the state has committed hundreds of millions to its restoration.
The Benefits of Beavers
Although the benefits of beavers are undeniable, our knowledge is still rife with gaps. We do not know how many there are, or in what direction their populations are moving, or which watersheds are in dire need of beaver flow. Few states have systematically surveyed them; moreover, many beaver ponds are located in remote streams far from human settlements, making it nearly impossible to count the beavers. “There is so much we don’t understand about beavers, partly because we don’t have a knowledge base of where they are,” says Emily Fairfax, a beaver researcher at the University of Minnesota.
Learning the Algorithm to Detect Beaver Ponds
But that is starting to change. Over the past several years, a team of beaver scientists and Google engineers have been teaching an algorithm to detect beaver structures on satellite images. Their innovation has the potential to transform our understanding of these tail-wagging engineers and help states affected by climate change like California in their recovery efforts. And although the model has not yet been made public, researchers are already excited about its potential. “All of our efforts in the state should benefit from this powerful photographic tool,” says Kristen Wilson, the chief forest scientist at The Nature Conservancy. “It’s really exciting.”
The Beaver Mapping Model
The beaver mapping model is the brainchild of Eddie Corwin, a former member of Google’s real estate sustainability team. Around 2018, Corwin began thinking about how his company could improve its stewardship of water, especially many of the coastal streams that run through its offices in the Bay Area. In the course of his research, Corwin read a book titled “Water: A Natural History” by Alice Outwater. One chapter focused on beavers, whose abundant wetlands, Outwater notes, “can hold millions of gallons of water” and “reduce flooding and downstream erosion.” Corwin was intrigued, and he began reading other books and articles about beavers, soon evangelizing to his friend Dan Ackerstein, a sustainability consultant working with Google. “Both of them fell in love with beavers,” says Corwin.
Training the Algorithm
Corwin and Ackerstein were welcomed by a company culture that famously encourages its employees to devote time to their passion projects, which produced Gmail. Corwin decided that his passion was beavers. But how could that help improve the work of white-toothed engineers? Corwin knew that beaver structures—their meandering dams, expansive ponds, and spiderweb-like channels—are often visible from space. In 2010, a Canadian researcher discovered the world’s longest beaver dam, a structure of sticks and mud stretching over half a mile across a park in Alberta, by browsing Google Earth. Corwin and Ackerstein began to wonder if they could contribute to beaver research by training a machine-learning algorithm to automatically detect beaver dams and ponds in satellite images—not one by one, but thousands at a time, across the surface of an entire state.
Training the Algorithm to Discover Beaver Ponds
Training EEAGER to detect beaver ponds wasn’t easy. The American West is full of features that seem practically designed to fool a model looking for beavers. Curving roads remind EEAGER of meandering dams; edges of human-built reservoirs register as beaver ponds. Strangely, the most confusing features are closed roads in residential neighborhoods, where asphalt circles surrounded by gray pavement strips resemble beaver ponds encircled by dams. “I don’t think anyone expected suburban America to be full of what the computer thinks are beaver dams,” says Ackerstein.
Uses
EEAGER
According to Fairfax, there are many uses for EEAGER. The model can be used to estimate beaver populations, monitor population trends, and calculate the ecosystem services provided by beavers such as water storage and fire prevention. It can help identify areas where beavers should be reintroduced, focusing on restoring waterways and wetlands, and creating conservation areas. It can also allow researchers to track beaver spread in the Arctic as the climate changes, or their movements in South America, where beavers were introduced in the 1940s and have spread since. “We really can’t handle all the requests we receive,” according to Fairfax, who serves as a scientific consultant for EEAGER.
EEAGER Application in California
The most promising application of EEAGER may be in California. The Golden State has a complicated relationship with beavers: for decades, the state generally denied that the species is native, being a product of large-scale fur trade that removed beavers from the West Coast before biologists could properly survey them. Although recent historical research has proven that beavers belonged almost everywhere in California, many water managers and farmers still consider them a nuisance, often killing them to block road culverts and interfere with irrigation structures. However, these entrenched attitudes are changing. After all, no state needs the water storage services of beavers more than fire and drought-prone California. In recent years, thanks to hard-hitting publicity from a campaign called “Bring Back the Beaver,” the California Department of Fish and Wildlife has started to renew its outdated policies regarding beavers. In 2022, the state allocated over $1.5 million for beaver restoration and announced it would hire five scientists to study and support beavers. It also adjusted its official approach to dealing with beaver conflicts to prefer coexistence over trapping.
EEAGER Experiment in California
It is quite fitting that California is where EEAGER will undergo its first major test. The Nature Conservancy and Google plan to run the model across the state sometime in 2024, conducting a comprehensive survey of every beaver dam and pond. This should give the state’s wildlife management a good idea of where beavers are and their approximate numbers, as well as where they could be most useful. The model will also provide California with strong baseline data against which future populations can be compared to see if its new policies are helping beavers to recover. “When you have images that repeat over time, it gives you the chance to understand change over time,” according to Kristen Wilson of the Nature Conservancy.
Future of EEAGER
What is in store for EEAGER after its trial in California? The main thing, according to Akrestine, is training it to recognize beaver structures in new places. (Although beaver dams and ponds look similar in every state, the model also relies on contextual clues from the surrounding landscape, and the High Plains of Wyoming look very different from the deciduous forests of Massachusetts). The team also needs to determine the long-term fate of EEAGER: Will it remain a Google-hosted tool? Will it spin off into an independent product? Will it become a service operated by a university or nonprofit organization? “That’s the challenge for the future — how do we make this more accessible and usable globally?” according to Corwin. The beaver revolution may not be broadcast on television, but it will certainly be documented via satellite.
This story first appeared on wired.com.
Leave a Reply