Wildfires continue to pose a significant threat to the environment, public health, and the economy. In this study, we introduce an AI-Based Digital Twin for Wildfire, an advanced computational framework designed to enhance the precision, efficiency, and real-time responsiveness of wildfire forecasting, response, and analysis. This wildfire Digital Twin employs a comprehensive set of technologies, including AI-based forecast systems with high-resolution predictive models to predict wildfire spread and behavior in the future and its downstream impact on air quality. The wildfire Digital Twin includes a unified data platform to ingest, integrate, and process the latest available information from satellite and ground-based observations to visualize the current status of wildfires and air quality (What-Now), and prepare the data for predictive models (What-Next) and impact assessment (What-If).
The predictive core of the wildfire Digital Twin is driven by deep learning models specifically designed to capture the complex spatiotemporal dynamics of wildfire progression. The models utilize fire observations from the VIIRS and GOES. Additional key inputs include terrain features (elevation, slope, aspect), meteorological variables (wind speed, wind direction, temperature, relative humidity, precipitation), soil moisture, and vegetation indices from MODIS, specifically the NDVI and EVI with a spatial resolution of 250 meters.
This wildfire ESDT can offer important insights into the current and future status of fire and air quality. This information empowers responders to strategically allocate resources and establish effective containment strategies, maximizing their efficiency and effectiveness in combating wildfires. By capturing both spatial and temporal dependencies effectively, this system offers predictions and impact assessment crucial for disaster management and mitigation.