Accelerate Global Conservation Decisionmaking

The Latest

• The Geospatial LLM-Enabled Navigator (GLEN) is an open source AI assistant that allows decisionmakers to interrogate global data on threatened lands and waters in order to inform conservation policy and action. 
• Users can ask the interactive model more complex analytical questions than current LLMs can answer, and will receive reproducible answers in minutes.
• GLEN runs locally on a small desktop server, which minimizes cost, privacy concerns, and energy demands.
• There are over a dozen active use cases. For example, policymakers are currently using GLEN to inform proposals for the United Nations High Seas Treaty to establish new global marine protected areas. In California, scientists and land managers are employing the tool to improve gray wolf recovery.

GLEN: Advancing Geospatial Analysis for All

Decisionmakers working to address biodiversity loss increasingly need answers to scientific questions that require analyzing dozens of global, geospatial datasets – such as protected area boundaries, migratory patterns, carbon stocks, and threatened watersheds – and they need those answers fast enough to inform a negotiation, policy, or regulatory filing. Commercial LLMs are often unable to query authoritative geospatial data and may invent plausible-sounding (but incorrect) results. At the same time, custom analyses across disparate datasets can take an expert weeks or months to perform.

Led by Carl Boettiger, Professor in the Department of Environmental Science, Policy, and Management and Schmidt DSE faculty advisor, DSE developed an AI assistant that closes this gap. GLEN is built on open source components (including the GeoJupyter platform, developed within DSE’s broader effort to democratize geospatial data analysis). The tool runs on a local machine and combines the reasoning ability of frontier language models with rapid, rigorous, and reproducible analytical results. 

How it Works

With GLEN, users with and without deep technical expertise can:

• Ask analytical questions about global conservation data in plain language and receive country-, region-, or watershed-level answers in minutes
• Layer well-known global datasets onto a single, interactive global map. These include the World Database on Protected Areas (WDPA), Ramsar wetland sites (wetlands of international importance designated under the Ramsar Convention), data on vulnerable sites with vast carbon stores, threatened watersheds, and natural places of importance to communities
• Generate reproducible queries that others can interrogate, modify, and build upon

Case Studies

Inform the United Nations’ High Seas Treaty

Together, our oceans represent the largest unprotected surfaces on the planet. In 2025 the United Nations finalized the Agreement on Marine Biological Diversity of Areas beyond National Jurisdiction (also called the High Seas Treaty), which opened the door for nations to propose new marine areas for protection. GLEN compiles and visualizes data on existing protected areas, ocean floor geology, biodiversity, and global fisheries activity and links this data to UN policy text. Using the chatbot, users can ask questions like:  “if I created a new marine protected area that was 100 kilometers wide in the high seas right next to my country’s border, how many undersea mountains and how many whale migration pathways would be included?” and will receive scientifically-rigorous answers in minutes.

Treaty discussions will be a focus at the upcoming IUCN World Conservation Congress in fall 2026, and we look forward to continuing to socialize the tool leading up to these important discussions. 

Improve California Gray Wolf Recovery

Gray wolves are returning to California after nearly a century. State agencies, scientists, ranchers, and conservation groups face complex questions about conservation and coexistence with livestock. For example, which movement corridors for wolves should be prioritized for protection? Where is wolf-livestock conflict most likely? The state is using GLEN as part of its 30x30 strategy to more rapidly analyze species occurrence data, land cover, land ownership boundaries, and livestock density so that managers can identify opportunities to restore natural prey habitat to reduce conflict. At the same time, land owners and the general public can use the same tool to track where wolves are spending time and/or their movement across lands. 

Democratizing & Dececentralizing AI

Commercial LLMs require a data center to deliver sufficient (and massive) power for the models to function. In contrast, GLEN can be downloaded onto a laptop or desktop computer and runs locally on that same machine. As a result the model uses 100x less power than a commercial counterpart (no data center required!). Moreover, GLEN is fully open source. By definition open source models can be viewed, modified, and distributed by any user, and offer more control over your data.

GLEN is one of many examples of a local, open source AI model in science that achieves similar (or better) performance to models like ChatGPT or Gemini, while avoiding the need for incredibly resource-intensive data centers. Since our inception in 2022, DSE has co-developed nine open source, decision-support tools for policymakers and scientists, and we are increasingly pivoting to developing local AI models for stakeholders. This includes GLEN and tools to help reduce global greenhouse gas pollution and improve vegetation recovery after a wildfire. 

Future Vision

Over the next year, we will scale the tool for more case studies and more user communities. Near-term milestones include deeper engagement with High Seas Treaty working groups, continued partnership with wildlife managers on wolf recovery, and an emerging project focused on conservation in the Greater Yellowstone Ecoystem. 

Looking further into the future, we see GLEN as a template for how researchers can create nimble yet robust AI tools with diverse use cases, in order to address increasingly complex conservation and climate challenges.