New technologies are generating far more information than ever before to help scientists assess and predict the health and behaviour of species and ecosystems, as well as the threats they face.
By Sue Palminteri
These include cryptic cameras, acoustic sensors, satellite imagery and citizen science apps. Now, researchers and conservation practitioners analyzing large data sets are exploring artificial intelligence, or AI—the ability of a machine or a computer program to think and learn—to help them process, analyze and interpret data to monitor ecosystems and predict results.
Computer systems already exist that can host huge amounts of data, use AI with increasingly “smart” algorithms to classify data from the various types of sensors used by scientists, apply modeling results to create reproducible code, and create user interfaces to allow people to monitor natural systems and make predictions with high accuracy.
By training computer algorithms with a subset of available data, machines can now learn what they should do for a given challenge—such as classify photos by the species found in them, identify areas of a satellite image containing water or intact forest, or translate speech from one language to another —based on human feedback and data collected from previous experience.
The applications for AI in conservation science are myriad: Fei Fang, an assistant professor at Carnegie Mellon University in Pennsylvania, leads a group applying game theory and machine learning to optimize patrol routes to catch poachers. At Wild Me, a nonprofit based in Portland, Oregon, director Jason Holmberg’s team applies AI-infused computer vision to identify individual animals of endangered species while reducing fieldwork and image processing time, under the Wildbook open-source project. And at Santa Cruz, California-based Conservation Metrics, chief executive Matthew McKown and his team are using AI and remote-sensing to automate wildlife surveys and machine learning to classify animals’ acoustic activity.
Cloud-based machine learning
A panel moderated by Mongabay founder Rhett Butler at the AI for Earth meeting at Microsoft headquarters in Seattle, Washington highlighted the work of Fang, Holmberg and McKown. The meeting, from May 16 to 18, gathered some 40 researchers, engineers and conservation professionals from over 25 institutions to network and improve their capacity to apply AI tools to specific conservation challenges.
The participants were all recipients of seed grants from Microsoft’s AI for Earth program. The grants provide between $5,000 and $15,000 worth of access to various resources on Microsoft’s Azure cloud-based computing platform. These include data, data storage, and tools for modeling, predictive analytics, machine learning, spatial analysis, and visualization, as well as training on using the various tools.
Over the two-day meeting, the grantees attended classroom sessions to learn about cloud-based machine learning tools on the Azure platform. Topics included modeling with AI, cognitive services (e.g. vision or speech), deep learning, project management, and data visualization. The meeting also introduced the Microsoft R Workflows tool that combines AI and the R statistical analysis tool, and the GeoAI tool that integrates AI with a geographic information system (GIS) to perform and automate spatial analyses, conduct predictive analytics, and integrate statistical and location information.
The AI for Earth grantee projects aim to classify images to map the distribution of species, vegetation types or spread of disease; predict extreme weather events, snow pack or crop cover; and assess risks or impact from climate change, geophysical hazards or wildlife poachers in a given area.
‘You know the problem you’re trying to solve’
Since its launch in December 2017, the AI for Earth program has provided assistance to more than 110 individuals and organizations in 27 countries working to address complex environmental and conservation challenges. Grantees include students, universities, private nonprofit organizations, government agencies, land trusts, and other groups.
It continues to accept applicants on a rolling basis, reviewing applications from around the world several times each year.
These collaborations earn Microsoft both positive publicity and new customers in the environmental sector, and grantees welcome the potential gains that AI-related technologies can bring to conservation science.
Fang and her students at Carnegie Mellon, for instance, expect that machine learning tools and game theory will help them more accurately and efficiently predict where wildlife poaching will happen next. Holmberg wants to speed up the Wildbook project’s processing of images, where AI can automate the scanning of social media sites for photographs of a particular species. Algorithms help identify each individual animal, together with where and how often it’s photographed, and suggest new ways to produce global population estimates for endangered species.
At Conservation Metrics, McKown wants to understand impacts of conservation actions, such as excluding domestic animals from islands, on seabirds and grassland nesting habitat. He and his team are beginning to use AI to help analyze acoustic activity data see if such actions actually help restore wild animal populations.
AI for Earth awards grants in four broad thematic areas: addressing climate change, protecting biodiversity, improving agricultural yields, and lessening water scarcity. It particularly seeks projects focusing on a question that lends itself to an AI answer.
“At this stage in our program, we’re trying to cast a very broad net,” said Josh Henretig, senior director and co-lead of AI for Earth. “There are a range of different technology solutions out there.”
AI for Earth can be most helpful to groups that have defined the problem they want to solve and have validated the problem with research or a target community to identify, and ideally gather, the data they need to solve it, Henretig said.
“Ultimately,” he said, “through our grant program … we’re starting with the premise that you’ve done that background work, you know the problem that you’re trying to solve, so let’s start with an initial investment in Azure infrastructure that you can begin to shape some part of the project that you’re trying to drive forward.”
Grantees at last week’s meeting were familiar with AI or machine learning but had a range of experience with the Azure platform. Lucas Joppa, Microsoft’s chief environmental scientist and AI for Earth’s other co-lead, said the initial grants of access to cloud infrastructure, AI software, and services in the Azure cloud were intended “to get people up and running to address issues with machine learning.”
“We do have fairly introductory-level content for people,” Henretig said, “like how to access and set up your Azure account, how to organize resources on that Azure account, [so] you know how to get started.”
Researchers can apply for AI for Earth support online here. Grant applications should include a proposal that describes the environmental problem or challenge being addressed, the datasets used for analysis, proposed methods to solve the problem, and the expected impact the project will have on the issue. Henretig said his team had a particular interest in supporting innovative water and agriculture projects, though they welcome proposals for projects in all four thematic categories.
FEEDBACK: Use this form to send a message to the editor of this post. If you want to post a public comment, you can do that at the bottom of the page.
(Feature Image – Sun sets over the Tsitsikamma National Park in South Africa. Photo by Charl van Rooy on Unsplash)