GeoNatureAgent evaluates AI tools on environmental map tasks
GeoNatureAgent (GNA): A Framework and Benchmark for Pre-Production Evaluation of Tool-Using Agents on Geospatial and Environmental Tasks
Artificial Intelligence
Summary
Environmental teams want to trust AI agents to pick the right tools for mapping and analysis tasks using real geographic data. The authors created GeoNatureAgent, a system that tests different AI models on a fixed set of 16 geographic tools and 103 tasks related to environmental info in Spain and Portugal. They found some models perform better but none match general GIS benchmarks when strict scoring is used. Their framework is open source and can be adapted for other geographic areas or tools. This helps teams know how well an AI agent might work before using it in real projects.
What this means in practice
- •For environmental data teams: Assess the reliability of AI agents using real APIs before deploying them in geospatial environmental mapping workflows.
- •For geospatial software developers: Use the GeoNatureAgent framework to benchmark and improve AI models that interact with geospatial tools and datasets.
Authors
Gabriel Diaz-Ireland, Diego Prieto-Herráez, Mario García Peces, Javier Velázquez, Benjamin Zaitchik, Devika Jain
Abstract
Before tool-using LLM agents are deployed in environmental and geospatial workflows, teams need evidence that an agent reliably selects the right operations against real APIs. We introduce GeoNatureAgent (GNA), a framework for pre-production evaluation of tool-using agents: a fixed sixteen-tool geospatial interface published as a Model Context Protocol (MCP) server, so the agent under test is the only variable, scored against an identical tool layer, task suite, and deterministic scorer. Its flagship instance is a 103-task benchmark (a 93-task main suite across 18 categories plus a ten-task comparison expansion) evaluated against an open, self-hostable geospatial API serving three environmental indicators across Spain and Portugal. We evaluate nine LLMs under three temperature-1.0 seeds, reporting capability and per-case cost as orthogonal axes. (1) Claude Sonnet 4 achieves the highest capability (61.7% +/- 0.7% on all 103 tasks; 60.8% on the main suite), followed closely by DeepSeek V3.2 (57.9%), while no other model exceeds 53%; (2) the cost-accuracy Pareto frontier is mostly open-weight, with DeepSeek V3.2 offering 93% of Claude's capability at 11.3x lower list-price cost; (3) under strict all-checks scoring the best model sits 24-36 points below the 85-97% reported on general-purpose GIS benchmarks, whereas per-check partial credit for the top four models (86-90%) is comparable, so much of that gap reflects scoring strictness rather than task difficulty alone. The MCP server, evaluation harness, benchmark, and API are publicly available; swapping the tool executors and task suite instantiates an equivalent benchmark for any geospatial domain.