The Artificial Intelligence & Machine Learning (AIML) team develops cutting-edge artificial intelligence and machine learning tools to help researchers and fire managers make better decisions about prescribed burns, smoke impacts, and wildfire risk. As a cross-cutting thrust within SMART FIRES, AIML supports all other research teams by turning complex data into actionable insights.
What does the AIML team do?
We build and apply AI/ML models to:
- Predict wildfire spread and smoke emissions
- Assess risk and optimize prescribed fire planning
- Analyze hyperspectral imagery to classify fuels and burn severity
- Model public sentiment and decision-making behavior
- Fill in missing data from damaged or obstructed sensors
- Support real-time, resource-efficient data processing
Why is AI/ML important for wildfire science?
Wildfire and smoke data are:
- High-volume (from drones, sensors, satellites)
- Complex (multi-modal, noisy, and dynamic)
- Time-sensitive (needed for rapid decisions)
AI/ML helps us process this data efficiently, uncover patterns, and make predictions that improve safety, resilience, and environmental outcomes.
Who is on the AIML team?
Led by Dr. John Sheppard (Montana State University), the AIML team also includes:
- Bradley Whitaker (MSU)
- Jesse Johnson (University of Montana)
- Lucy Owen (UM)
- Michael Wojnowicz (MSU)
- Tim Price (Flathead Valley Community College)
- 2 new faculty hires
- 1–2 postdoctoral researchers
- 7 graduate students
- 18 undergraduate students
What kinds of models are being developed?
We use a range of advanced techniques:
- Neural networks and sparse learning for risk modeling
- Continuous Time Decision Networks (CTDN) for fire planning
- Surrogate models for emissions prediction
- Deep learning and NLP for sentiment analysis
- Probabilistic interpolation for missing sensor data
- Multi-modal learning to integrate diverse data sources
How does AIML support other SMART FIRES teams?
- Works with Smart Optical Sensors (SOS) to fuse sensor data
- Collaborates with Fire & Smoke Science (FSS) to model emissions
- Supports Social Psychology, Economics & Ethics (SPEE) in analyzing public sentiment
- Partners with the Cyberinfrastructure (CI) team to manage and share data
What are some key accomplishments so far?
- Developed baseline models for wildfire spread and air quality
- Published multiple conference papers on hyperspectral image classification
- Created a machine learning framework for predicting fire perimeters
- Built tools to assess social media sentiment during fire events
- Contributed to the SMART FIRES data lake and research visibility efforts
How is AIML helping with risk assessment?
We’re building models that:
- Simulate fire spread under different conditions
- Estimate emissions and air quality impacts
- Optimize prescribed burn timing and location
- Incorporate uncertainty and community response into planning
How does AIML address ethical and social concerns?
- Models are designed to be interpretable and responsible
- We study how trust in AI-generated data affects decision-making
- We collaborate with social scientists to ensure equitable outcomes
- We develop tools for communicating risk and benefits to diverse audiences
What’s next for the AIML team?
- Expand real-time classification tools for field deployment
- Integrate foundational climate models (e.g., Microsoft Aurora)
- Improve air quality predictions using Gaussian processes
- Support new faculty and graduate students in AI/ML research
- Continue publishing and sharing tools with the broader community