Read the full article at https://www.tandfonline.com/doi/full/10.1080/17538947.2025.2458688
Powered by the emerging large language models (LLMs), autonomous geographic information system (GIS) agents can perform spatial analyses and cartographic tasks. However, a research gap exists in enabling these agents to autonomously discover and retrieve the necessary data for spatial analysis. This study proposes an autonomous GIS agent framework capable of retrieving required geospatial data by generating, executing, and debugging programs. The framework, with an LLM-driven decision core, selects data sources from a predefined list and fetches data using source-specific handbooks that document metadata and data retrieval details. Designed in a plug-and-play style, the framework allows human users or automated data crawlers to add new sources by creating additional handbooks. A prototype agent based on the framework is developed and released as a QGIS plugin and a Python program. Experiment results demonstrate its capability of retrieving data from various sources, including OpenStreetMap, administrative boundaries and demographic data from the U.S. Census Bureau, satellite basemaps from ESRI World Imagery, global digital elevation model (DEM) from OpenTopography.org, weather data from a commercial provider, and the COVID-19 case data from the NYTimes GitHub. This study is among the first attempts to develop an autonomous GIS agent for geospatial data retrieval.
