Towards a NUWRF and AI-based Mega Wildfire Digital Twin (WDT)
The Wildfire Digital Twin (WDT) project aims to implement and evaluate a regional digital twin model for the simulation of fire, its spread, and smoke impacts on air quality, regional deforestation, and human health in local and distant communities. Such a regional model would enable mega wildfire smoke impact scenarios at various spatial scales and arbitrary locations over North America. WDT will utilize AI machine learning to provide ensemble wildfire forecasts in real time, and implement a satellite data assimilation scheme to conduct impact scenarios on wildfires and wildfire smoke.
Persistent droughts and extreme heatwave events over the Western US and the Canadian and the Eurasian Boreal Forests are creating highly favorable conditions for mega wildfires that are generating CO2 emissions having positive feedbacks on global warming. Smoke from these wildfires are adversely impacting cities on both coasts. Further, intense wildfires are destroying large areas of forests and homes even impacting US travel.
This WDT will embed within the NASA Unified Weather Research Forecasting (NUWRF) model, a chemically interactive SFire atmospheric fire spread resolving module (SFIRE2 at 30m). Ensemble regional AI emulations will be conducted with a deep, dense, physics aware neural operator emulation of an interactive microphysics/chem model compatible with high resolution rapid refresh hourly predictions.
- The NUWRF model performance will be accelerated to enable high resolution of coupling SFire modules for emulating the wildfire microphysics and chemical atmospheric interactions.
- A regional AI/ML model driven by remote sensing data (radar, satellites, drones) will conduct near-real-time digital twin ensemble forecasts and impact inferences of smoke and the rapid spread of fires on forests and neighboring homes.
- Visualization and animations, computer graphic algorithms, and tensor decomposition analytics will validate wildfire model spread with Landsat and Sentinel data sets.
Milton Halem is a research professor in the UMBC Computer Science and Electrical Engineering department. He also holds an Emeritus position as Chief Information Research Scientist to the Director of the Earth Sciences Directorate at the NASA Goddard Space Flight Center. Prior to retiring in 2002, Dr. Halem served in the joint capacity as Assistant Director for Information Sciences and Chief Information Officer for the NASA Goddard Space Flight Center. Dr. Halem provided the strategic information science and technology focus and oversight for the entire mission critical programs and projects at the Center. He is most noted for his ground breaking research in simulation studies of space observing systems and for development of four dimensional data assimilation for weather and climate prediction. Over the years, his achievements have earned him numerous awards including the NASA Medal for Exceptional Scientific Achievement, the NASA Medal for Outstanding Leadership, and NASA’s highest award; the NASA Distinguished Service Medal in 1996