Generating Narrative on the Immersive Forest with Knowledge Graphs and Conversational AI. In Data Science & Engineering Master of Advanced Study (DSE MAS) Capstone Projects

Wildfires are growing more destructive and costly each year—threatening lives, infrastructure, and ecosystems. To combat this, land managers and burn bosses need smarter tools to plan and execute prescribed burns that reduce hazardous fuels. This project integrates vegetation characteristics and fuel LiDAR data/metrics with weather data into a spatial knowledge graph using GraphDB and SPARQL, enabling precise, location-based insights. The graph includes both spatial and temporal properties, allowing users to examine before-and-after effects of prescribed burns over time. Combined with large language models (e.g., OpenAI, LLaMA) and orchestrated through LangGraph, the system allows users to interact with complex fire and fuel data through a natural language chat interface. It also includes a web search tool for retrieving external fire-related content to enrich responses. The platform is accessible via a web-based chatbot, can integrate with the Immersive Forest Unreal Engine simulation, and supports advanced research use through the NDP platform or direct API access. The ultimate goal is to minimize the devastation of wildfires by enabling proactive, data-driven fuel management at landscape scale.

The project utilized a diverse set of wildfire-related datasets, including aerial and terrestrial scans plot metrics alongside sensor readings. These datasets were integrated and aligned with our geospatial and temporal knowledge graph, enabling advanced queries and analysis for wildfire risk.

This project was conducted as part of the Data Science and Engineering Master of Advanced Study (DSE MAS) program.