In this case study, we speak with PJMF partner, Earth Genome, on how they went about defining the problem when first developing Earth Index, a tool that empowers environmental defenders with AI to understand local trends.
<aside>
AI Journey Phases: Exploration; Adoption
Module(s): Problem Definition; Product Development
About Earth Genome:
Domain: Climate
Organization Size: 11-50
Region: Global
Website: https://www.earthgenome.org/
</aside>
Founded in 2015, Earth Genome is harnessing their deep expertise in data and AI to address some of the world’s most pressing environmental challenges. With satellites, sensors, and other technology generating an ever-increasing amount of data about the Earth, we’ve never been better equipped to analyze what’s happening to our planet. But the communities who often benefit most from these insights still face a barrier to entry when it comes to accessing that data and the AI tools that can make sense of it.
Earth Genome saw the opportunity to make analysis more understandable and affordable with the goal of enabling more sustainable use of our planet’s natural resources. They realize this impact through building open, accessible tools and through collaboration with partners across the conservation space — from researchers, policy makers, journalists, builders, and more.
In this case study, we’ll be diving into how Earth Genome went about understanding the problem and embarking on the product development of Earth Index, an AI-enabled search tool that allows users to obtain insights from satellite data. At its core, Earth Index provides access to earth observation data. With the use of machine learning models, it enables ongoing monitoring and change detection so that users of the tool, like journalists or Indigenous community organizations, can identify and compare changes to the earth’s surface in near-real time.
Earth Index is an AI-enabled search tool that allows users to obtain insights from global satellite data.
In sharing about Earth Genome’s process, we’ll go beyond how we strictly define problem definition and show how the step of deeply understanding the problem at hand is not always linear. It’s often interwoven with the design, prototyping, and iteration processes that shape product development. But what remains true throughout Earth Genome’s work is how they remain steadfastly rooted in the needs of their users rather than being carried away by what the technology is capable of.
Well before the development of Earth Index began, a seed of the idea had been planted with the Earth Genome team — how could data be used to help people see for themselves how the earth’s surface is changing?
In 2021, they kicked off a project, called Amazon Mining Watch, in partnership with the Pulitzer Center´s Rainforest Investigations Network to bring together the power of machine learning and investigative journalism to shed light on illegal mining in the Amazon. Using remote sensing data — global data about earth’s systems obtained via satellite — and data gathered by journalists on the ground, the team was able to uncover critical insights. They found that illegal gold mining in the Yanomami Indigenous Territory in the Brazilian Amazon was significantly more prevalent than previously reported. One in three Yanomami villages were being affected by mining-related deforestation happening within 10km of their village. These insights were a result of a lot of deep data science work on the part of the Earth Genome team and no real tool existed yet to repeat this type of investigation, for instance in another region or in an effort to identify another environmental challenge. But the team immediately realized how valuable it would be to have a more replicable way to enable this work.
An example of what mining operation detection looks like via satellite imagery on Earth Index.
An example of what mining operation detection looks like via satellite imagery on Earth Index.
Others around the globe came to the same conclusion as the project set off a cascade of interest. Other journalists and organizations came to Earth Genome with their own questions relating to conservation issues in their communities. From illegal logging in Albania to rainforest loss due to coco distribution (the source of cocaine) in Peru, there was no shortage of problems. These organizations often had a hunch for what was going on in their communities, but the sheer scale of where to look to find evidence of the problem was overwhelming. And for many, the technical capabilities required to gather a suitable dataset and build a tool to search the region themselves was simply out of the question.
The team saw that there was a huge world of need, and they recognized that a generalized tool that democratized access to remote sensing data and their AI model had the potential to revolutionize conservation action across the globe.
Earth Genome let the success and learnings from these early projects lead the way and define what they were building. From the start, they recognized the importance of a human-centered approach to product development. The data and AI model alone were certainly going to be critical to whatever product they built, but they would not be enough to carry it. They wanted to build something that would truly be accessible to the communities that need this tool most. They knew this would require a thoughtful product experience that would get users in the driver’s seat to be able to answer their own questions, no matter their technical know-how or the geographical scope of their issue.
To help shape what that solution might look like, they sought to deeply understand who their core users were, what use cases they were turning to the tool for, and how they would be interacting with a tool in order to achieve their goals.
Some of the key ways they went about doing this included: