Athabasca University · VIP Research Group · Alberta Innovates

Catching the spark
before the fire

LeoSpark is an AI driven wildfire risk prediction system designed to help identify wildfire prone conditions using environmental, geospatial, and historical fire data.

Landscape
Environmental
Observation for
Smart
Prediction,
Analytics,
Risk &
Knowledge
Explore the Research Watch Demo
Multi-Source
Environmental Data Integration
Alberta, Canada
Primary Focus Region
Machine Learning
Random Forest Focus
Wildfire Risk Prediction
Research Objective
About the Project

AI powered wildfire risk prediction

LeoSpark explores how artificial intelligence and machine learning can support proactive wildfire management by analyzing environmental, geospatial, weather, and historical wildfire data, with a primary focus on Canada.

Wildfire Risk Prediction

Identifies conditions associated with elevated wildfire risk using weather patterns, vegetation conditions, terrain, and historical wildfire activity.

Machine Learning Focus

Explores Random Forest and other machine learning approaches to better understand wildfire risk patterns and improve predictive accuracy..

Canadian Focus

Uses Canadian wildfire datasets and environmental data sources to support regionally relevant wildfire risk analysis.

LeoSpark wildfire detection
Methodology

How LeoSpark works

A multi-stage machine learning workflow for understanding and predicting wildfire risk using environmental and historical fire data.

01

Data Collection

Gathering wildfire, environmental, weather, terrain, vegetation, and satellite derived datasets, with an emphasis on Canadian wildfire conditions.

02

Data Preparation

Cleaning, organizing, and transforming multi-source datasets into structured formats suitable for machine learning and wildfire risk analysis.

03

Machine Learning Modeling

Training and evaluating Random Forest to identify wildfire risk patterns and improve predictive performance.

04

Risk Prediction & Visualization

Generating wildfire risk predictions and interactive visual outputs to support wildfire awareness, preparedness, and future decision support systems.

The People

Our Team

A collaborative research effort between student and faculty at Athabasca University.

Dr. Maiga Chang
Dr. Maiga Chang
Research Supervisor

Full Professor and Associate Dean, Research & Innovation at Faculty of Science and Technology, Athabasca University, Canada. He is IEEE Senior Member.

Leonel Navarrete
Leonel Navarrete
Graduate Researcher · MSc CIS

Master of Science in Computing and Information Systems student at Athabasca University, focused on AI driven wildfire risk prediction to support smarter wildfire management. He is IEEE Student Member.

See it in Action

Demo

Explore the evolving LeoSpark prototype for wildfire risk prediction and visualization.

▶

Demo video coming soon

Questions

Frequently Asked Questions

LeoSpark (Landscape & Environmental Observation, Smart Prediction, Analytics, Risk & Knowledge) is an AI driven wildfire research project conducted through Athabasca University and the VIP Research Group. Supported through Alberta Innovates graduate research funding, the project explores how machine learning, environmental data, and historical wildfire information can support wildfire risk prediction, with a focus on Alberta, Canada.
LeoSpark evaluates a combination of Canadian and international environmental and wildfire related datasets, including wildfire history, weather, terrain, vegetation, and satellite-derived information. Dataset selection is ongoing and prioritized based on compatibility with machine learning approaches such as Random Forest.
LeoSpark is currently an active graduate research project and prototype under development. Access is presently limited to research and development activities. Broader availability may be explored in future project phases.
Model performance is currently being evaluated as part of ongoing research. Different machine learning approaches and evaluation metrics are being explored to better understand wildfire risk prediction performance. Findings will be shared through research outputs as the project progresses.
LeoSpark currently focuses on machine learning approaches for wildfire risk prediction, with an initial emphasis on Random Forest models. Additional methods may be explored as the research evolves.