11-19, 11:30–11:55 (Pacific/Auckland), WG308 TE IRINGA
An application case of combining open-source technology and AI to develop a Wildfire Spread Prediction platform that supports on-site decision-making.
The frequency and intensity of wildfires are increasing globally due to climate change impacts, and the Republic of Korea is no exception. Particularly during spring seasons, the combination of dry weather and strong winds leads to large-scale wildfires, resulting in national disasters.
In March 2025, simultaneous large-scale wildfires that broke out across South Korea were recorded as one of the worst in the nation’s history, burning approximately 100,000 hectares of forest an area 1.7 times the size of Seoul, the capital city. It lasted nine days before it was fully extinguished due to strong winds with maximum instantaneous wind speeds of over 25 m/s, making it extremely difficult to extinguish.
To effectively respond to these large-scale wildfires, which are difficult to predict and cause immense damage, it is essential to have technology that can analyze and predict their paths by receiving real-time data on the three key elements of wildfire spread: topography, fuel, and weather.
Accordingly, we have developed a system that improves upon existing services based on empirical algorithms, precisely predicting wildfire spread using AI deep learning technology. The system processes and analyzes real-time data on the three elements of wildfire spread and utilizes open-source technology to predict, analyze, and visualize the path and speed of the fire.
We have actively utilized a proven open-source tech stack—including QGIS, PostGIS, GeoServer, OpenLayers, and CesiumJS—to implement an integrated development environment that covers everything from the analysis of prediction data to its 2D and 3D visualization.
In this presentation, we aim to introduce a practical case study where this technological foundation was used to effectively support on-site decision-making during wildfire events.
This study was carried out with the support of 'R&D Program for Forest Science Technology '(Project No. "RS")' provided by Korea Forest Service(Korea Forestry Promotion Institute).
Hanjin Lee is a GIS Developer at the Gaia3D Inc. He has been working on application development and education using open source GIS for many years. Worked in data processing, visualization, and communicating information intuitively. More recently, he's been interested in GeoAI and hopes to use it to develop applications that capitalize on emerging technology trends.
SW Developer (Gaia3D corp, Inc)
Hyun-Woo Jo is a Research Professor at the OJEong Resilience Institute of Korea University and a Postdoctoral Fellow at the International Institute for Applied Systems Analysis (IIASA). He received his B.S. in Environmental Science and Ecological Engineering and his Ph.D. in Environmental Planning and Landscape Architecture from Korea University in 2018 and 2023, respectively.
His research focuses on integrating remote sensing and deep learning in agriculture and forestry, with a strong emphasis on combining domain-specific knowledge with machine-learnable models. In 2022, Dr. Jo was selected for the prestigious IIASA Young Scientists Summer Program (YSSP), where his project on optimizing IIASA’s FLAM wildfire model for Korean conditions earned an honorable mention. He later expanded this work by embedding process-based wildfire algorithms into neural networks as part of his doctoral research.
Dr. Jo has also developed AI models for crop and land cover monitoring using satellite data, with a focus on cross-regional generalization through transfer learning. These techniques have been tested in South Korea and applied to international contexts, and the results were presented at NeurIPS Workshop on Climate Change and AI.
He is also interested in building user-friendly software tools for AI-driven environmental analysis, such as Platform Dryad, which supports scalable modeling and decision support in ecological systems.