Get started with a free trial of Domino→

×
Industry

Non-profit

Location

Headquarters: Manhattan, New York

Data Science Team: 70 percent working remotely across 11 states

Use Cases
  • Research to understand the impact of climate change on bird species
Impact
  • Conducted the most in-depth research to date on the effects of climate change on birds in North America
  • Research encompasses both broad-scale and local data
  • Empowers prioritized conservation efforts across 604 species
Data Science at Scale
  • 160 machine learning models for 604 bird species, covering two different seasons (winter and summer) and three warming scenarios for a total of over 180,000 models
  • 140 million observations from scientists and community members analyzed
Users

Quantitative researchers

Solution

Data Science Tools: R, Python, cURL, Maxent

IT Infrastructure: Amazon Web Services

Data Science Platform: Domino

Data Science at the National Audubon Society

The National Audubon Society has been working to protect America’s birds and the places they need since 1905. Back then, one of the greatest threats to birds came from plume traders who hunted birds for their feathers. Today, climate change stands as the biggest threat to bird populations. Audubon applies data science to outline the risks birds face with a changing climate so they can mobilize their one million members and supporters to take action.

In particular, the Audubon Birds and Climate Change Report documents where bird species live in North America and predicts the effect different warming scenarios will have on them. Using the Domino data science platform, Audubon has created 180,000 models to analyze more than 140 million observations from scientists and community members, and to predict the impact of different warming scenarios on 604 bird species.

The prognosis isn’t good: according to Audubon’s latest report, Survival by Degrees, climate change threatens nearly two-thirds of North American birds. But with the deep insight gained on each species, Audubon and other conservation groups can better target policy recommendations and conservation efforts to make a difference.

Challenge

Audubon’s first Birds and Climate Change Report, published in 2014, found that 314 species, including well-known species like the bald eagle, brown pelican, and common loon, were severely at risk due to global warming. Ten states were at risk of losing their official state bird.

This inaugural report drew on climate models and observations from thousands of “community scientists” who contribute to Audubon’s annual Christmas Bird Count and other bird surveys. Ecologists have long observed that people thrive where birds thrive, making this a watershed report in understanding the impact of climate change on both birds and people.

Despite the project’s success, Audubon’s scientists saw an opportunity to take their work to the next level. “The volume of information available has dramatically increased through community science, web services, and mobile phones,” said Audubon’s Interim Chief Scientist and Vice President of Conservation Science, Chad Wilsey. “We wanted to scale our quantitative analyses to incorporate the growing volume of community science data into our models and build more individualized models for each species.”

Domino has allowed our quantitative researchers to scale-up our analysis faster and with less effort.

Chad Wilsey, Interim Chief Scientist and Vice President of Conservation Science at Audubon
Solution

In early 2017, Audubon started working with Domino Data Lab to amplify its data science work and analyze how climate change will expand or contract bird habitats. The organization first built and ran individualized models for 40 species, and ultimately scaled its project more than 10X to understand the impact of different warming scenarios on 604 species.

“Domino has allowed our quantitative researchers to scale-up our analysis faster and with less effort,” said Wilsey. “We can generate hundreds of models, run diagnostics, and select the best-performing model in relatively little time, and we’ve been able to crunch years of work into months. This is critical, as the models we’re building incorporate huge areas of land at high spatial resolution, and need lots of processing time.”

Additionally, Domino offered several key differentiators, including:

  • Openness. The platform is language-agnostic, allowing data scientists to use the tools they’re most familiar with, such as the R and Python programming languages, and specialized tools such as cURL and Maxent
  • Reproducibility. Domino’s sophisticated repository maintains version control across an entire project, allowing data scientists to review all steps leading to a given result, change inputs, and track the impact of changes on models and their predictions.
  • Reduced deployment friction. Domino’s Kubernetes-native compute grid makes it fast and easy to deploy models as low-latency, high-availability APIs or to publish interactive Shiny, Flask or Dash apps at the click of a button. “Using Domino, we’re looking to do more on-the-fly deployment of models to create real-time outputs for the organization,” said Wilsey.
Use Case: Protect the birds and protect the earth

In building its models, Audubon data scientists wanted to understand the impacts of global warming and nine climate-related threats, including fire weather, sea-level rise, and droughts, on North American birds and the places they need to survive. For example, would a species expand into a new area if their existing range was no longer habitable, or would they face extinction? The most vulnerable species are ones that face significant loss of their current range and are not likely to move to new areas.

Additionally, the team is using models to map out areas that will see multiple impacts from global warming, such as false springs and fire weather, depending on the degree of change.

“With Domino, data scientists can explore different parameter settings to ensure we have the best performing model for each species and collaborate as efficiently as possible."

Chad Wilsey

“We built our models using a workflow that we deployed in Domino,” said Wilsey. “We built 160 machine learning models for each of the 604 species, covering two different seasons (winter and summer) and three climate change scenarios (representing 1.5, 2.0, and 3 degree Celsius increases in global mean temperature), for a total over 180,000 models.”

One of the most critical capabilities for scaling this work has been Domino’s built-in reproducibility. “With Domino, data scientists can explore different parameter settings to ensure we have the best performing model for each species and collaborate as efficiently as possible,” said Wilsey.

The Domino Effect

Identifying the most vulnerable species and habitats. One of the project’s unique contributions is its analysis of both broad-scale and local data on hundreds of bird species. “Thanks to the horsepower of Domino, this is the most in-depth research ever conducted on the effects of climate change on birds in North America,” Wilsey said.

Targeting conservation efforts. While the report shows that nearly two-thirds of bird species in North America are highly vulnerable, it also illuminates a path forward. "This work helps direct on-the-ground conservation efforts and informs advocacy to reduce the impacts of climate change,” said Wilsey. “If we keep global temperatures down to a 1.5 degree Celsius increase, we can significantly improve the chances for up to 76 percent of those at risk.”

“Thanks to the horsepower of Domino, this is the most in-depth research ever conducted on the effects of climate change on birds in North America.”

Chad Wilsey

Better measuring success. As a non-profit, Audubon measures its success by how much habitat it has protected and how many birds it has saved. “With our data science work, we can show with hard data trends that we’re actually having an impact,” said Wilsey.

Maximizing data science productivity. Using Domino, Audubon data scientists can easily collaborate, iterate faster, and quickly reproduce results to scale its work. “We’re a 700 person non-profit,” said Wilsey. “We have 16 people developing models for the organization. We started our work with Domino looking at 40 species in particular and within a year have been able to scale that to 604 species.”