Photo by NASA on Unsplash

Introducing G.AI.A: Artificial Intelligence and Planetary-Scale Environmental Management

Can we build an AI-driven system for managing the planet? Perhaps — but let’s be realistic about the challenges & what it can achieve.

Dave Thau
Dave Thau
May 24 · 8 min read

First in a series on the promise and challenges to using AI and machine learning to create a planetary environmental management system.

Thanks to my esteemed reviewers, commenters, and contributors: Adia Bey, Aurélie Shapiro, Azalea Kamellia, Bob Lalasz, Dan Morris, Debora Pignatari Drucker, Diana Anthony, Erik Lindquist, Gregoire Dubois, Holly Grimm, Karen Bakker, Karin Tuxen-Bettman, Johanna Prüssmann, Mayra Milkovic, Nasser Olwero, Nicholas Clinton, Sophie Galloway Tanya Birch, and Tyler Erickson.

From climate change to ocean acidification to conversion of habitat for agricultural use, humans are having a massive impact on the Earth.

With humanity slated to add 2 billion more people to the planet by 2050 — and with people using more natural resources than ever before — we need to better manage our use of those resources and find ways to reduce our impact on the planet.

A planetary environmental management system — one that a) collects the right data about the state of Earth, its ecosystems and populations; b) turns that data into useful information; and c) helps us leverage that information to inform better decision making — could help us meet our environmental and societal goals.

But there are big challenges to making such a system reality. One such challenge: sifting through and making sense of the sheer amount of Earth observation and other relevant data being collected from existing sensors (satellites, river level gauges, cell phones, motion-activated cameras, just to name a few) — not to mention the exponentially greater amount of data coming in from new sensors continually being designed and deployed.

When faced with enormous amounts of rapidly accumulating data, analysts in governments, businesses, and academia are increasingly turning to artificial intelligence and machine learning. Could AI and machine learning help us build a system that brings us the insights we need to meet our goals?

Exploring that question — both the opportunities and the challenges to leveraging AI and machine learning to create an AI-powered environmental management system — is the objective of this new series on Medium.

I am a computer scientist by training, and have spent decades applying those skills to address questions about biodiversity, ecology, and conservation. I’ve worked in academia, industry, and non-profit organizations and have seen where each of these contributes to the broader goal of achieving sustainability. Throughout, I have helped build tools that apply machine learning and “Good Old Fashioned Artificial Intelligence” to topics like detecting deforestation from satellite imagery, tracking species names as they change over time, and identifying wildlife from camera trap images.

Don’t assume from my background, however, that I’m a techno-utopian. I am allergic to hype about AI, and this series is not going to be about a post-human world in which we’ve uploaded our minds and turned the Earth into a nature park. Instead, I’ll talk about the promises and pitfalls of using AI in conservation, show how it’s being used now, and point to ways to tailor its use in ambitious ways to help manage our environment more effectively.

Why Build a ‘Planetary Management System,’ Anyway?

The past 50 years have seen a drastic decrease in global species populations and a significant increase in greenhouse gas emissions and global temperatures. The trends are all pointing toward twin crises in ecosystem health and climate. Both of these crises stem from our current mismanagement of global resources.

One thing is clear: Given the current trends, if we don’t do a better job of managing our climate and environment (and ourselves), our planet will become, on average, considerably less hospitable to humans and many other species.

As we will see below, other industries use resource management systems to improve their bottom lines. The human species needs to do the same for our natural resources. Other industries are also building AI enhanced resource management systems. We must do the same to manage and protect our global environment.

What Do I Mean By a ‘Planetary Management System’?

Before you jump to conclusions about the phrase “planetary management system” and its tyrannical or dystopian connotations, let’s take a deep breath. Just like there’s no single place where the world’s financial systems are managed, or the world’s healthcare systems, there will be no single entity responsible for managing all the world’s environmental data and decision making. (Nor am I advocating one.)

No global organizations, in fact, even come close today to being in a position to manage global data even for a single issue. For instance, the World Trade Organization, the World Bank and the World Health Organization all play a supporting role when it comes to global data collection and decision support for their respective topics and issues. Similarly, the United Nations Environment Programme’s (UNEP) mission isn’t to maintain global datasets about the planet, but instead to “provide leadership and encourage partnership in caring for the environment by inspiring, informing, and enabling nations and peoples to improve their quality of life without compromising that of future generations.” UNEP provides many datasets and tools, but these are generally aggregations of data produced by other agencies.

When I talk about an AI-powered environmental management system, I don’t mean a single monolithic system. Instead, I mean a collection of tools and approaches that leverage artificial intelligence to better understand and manage our environment for the benefit of people and the other species on Earth. Some of these tools and datasets may be automatically interoperable; many will need to be combined by people in the context of addressing specific needs. In aggregate, the tools and methods covered in this series of blog posts will move us to a future in which we can better manage our environment with the help of artificial intelligence to combine data, generate new insights, and help us convert data into knowledge and ultimately into impactful actions. I would imagine users of this environmental management system would include (among others) UNEP, global conservation NGOs, corporations seeking to become more sustainable, countries looking to add sustainability to their national accounting practices, and local actors.

A Planetary Monitoring & Management System: Beyond Data Collection

The big gorillas in the computer industry and conservation world are already envisioning how to use technology to better measure our present and future impact on the planet and build systems to help create positive impactful change. Some key examples on the industry side are Google technology for sustainability, Microsoft’s Planetary Computer, Amazon Earth on AWS, ESRI’s Living Atlas of the World, and IBM’s Green Horizons, to name just a few. Big conservation organizations and governments are engaging as well. See for example, UNEP’s discussion of a Digital Ecosystem for the Environment, the European Union’s DestinE, the Ocean Data Platform, and the conservation and technology community, which is a partnership of some of the world’s largest conservation non-profit organizations.

Many of the applications developed under these projects use AI and machine learning to coalesce large amounts of data into actionable information. This process includes using satellite imagery to detect and predict tree cover loss, using communications equipment to detect illegal fishing, and managing data coming from air quality sensors to detect the sources of problematic emissions. Examples like these show the power of AI and machine learning to make sense of the massive amount of data we currently collect. However, they barely scratch the surface of what we will need to create a truly AI-powered environmental monitoring and management system.

Some conservation problems, such as identifying species in camera traps, are perfectly suited to the kinds of machine learning that have been so successful in recent years. Other challenges are a less natural fit. These include how to use AI to help drive decision making, how it can be used to analyze systems as well as data streams, and challenges around using AI to predict future scenarios. All of these areas are ripe for further development and should be supported with further research and application.

Domains outside of conservation are already leveraging artificial intelligence to make significant gains:

  • For example, Accenture estimates that AI applications will save the United States healthcare industry $150 billion annually starting in 2026. These savings come from more efficient processing of medical records, faster medical image processing, and accelerated drug discovery.
  • MHI and Deloitte report that 12% of businesses are currently using AI for supply chain management, applying it to warehouse management, supply chain planning, and procurement. Most of these businesses report revenue increases and many report cost savings from the efficiencies gained with AI.

In all of these examples, AI is making already existing processes more efficient as well as accelerating new discoveries.

These business-based examples cast successes in terms of maximizing monetary profits and minimizing losses. But the central problems they all address are ones of resource management. Conservation is faced with similar questions, although the currency here is in terms of intact habitat, ecosystem services, and biodiversity instead of cold hard cash. The critical questions in conservation include:

  • How can we support our growing human population in a way that sustains the environment upon which we depend?
  • How can we manage our impact on the climate to best ensure a continuation of the favorable conditions we have enjoyed during most of our species’ history?
  • How can we decrease the extinction rate down to its historical baseline rate?

Given humanity’s global impact, all of these questions turn into questions of resource management. Put another way: How can we best manage the ecosystems and biodiversity on the planet?

What comes next?

Much of my work to date has focused on environmental monitoring and data management. These are important pieces of any kind of management system, but over the years it has become increasingly obvious that these are necessary but not sufficient when it comes to building a system that will actively help us achieve our sustainability goals. The famously questionable line “If you build it, they will come” has an equally questionable parallel: “If you show it, they will act.”

The next post of this series will flow from this observation: Awareness of what is happening globally to the environment is not enough to drive action toward more sustainable practices. I’ll give examples of where global monitoring systems have revealed specific acts of illegal resource exploitation, how the monitoring has helped and how it still falls short. Then I’ll describe what researchers, advocates, and businesses have been doing to turn information into action and delve into current and possible future roles of artificial intelligence to help address this need.

Stay tuned!


Nature and Artificial Intelligence