🔥 Alberta Wildfire Data Story (2006-2025)

A comprehensive data science investigation of 26,551 wildfire incidents across two decades

---

📊 Project Overview

This project analyzes 26,551 wildfire incidents from Alberta, Canada (2006-2025) using advanced data science techniques to answer critical questions about fire patterns, causes, and predictability.

🎯 Research Questions

#	Question	Methods	Key Finding
1	Are wildfires increasing?	Linear regression, trend analysis	High variability, no simple trend
2	Where do fires concentrate?	Geospatial clustering (EPSG:3403)	Three distinct regions identified
3	What causes fires by region?	Chi-square test, contingency analysis	Causes vary significantly N→S
4	Does fast response reduce size?	Correlation analysis	Weak correlation (r≈0.3)
5	What weather predicts fire behavior?	Pearson correlation, scatter analysis	Combinations matter most
6	Can ML predict fire types?	K-means, Random Forest	87% accuracy, 4 fire types

💡 Key Insights

✅ 87% ML prediction accuracy for fire size classification
✅ 4 distinct fire behavior types identified through clustering
✅ Regional differences support tailored management strategies
✅ High year-to-year variability dominates temporal patterns
✅ Weather combinations predict risk better than individual variables

---

🚀 Quick Start

Prerequisites

• Python 3.12 or higher
• Jupyter Notebook
• 4GB+ RAM recommended

Installation

# Clone repository
git clone https://github.com/yourusername/alberta-wildfire-analysis.git
cd alberta-wildfire-analysis

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Launch Jupyter
jupyter notebook Wildfire_DataStory_Enhanced.ipynb

Get the Data

Option 1: Download from Source

1. Visit Alberta Wildfire Historical Data
2. Download complete dataset (2006-2025)
3. Save as data/wildfire_data.csv

Option 2: Use Sample Data

• Sample dataset available in /data folder (10% random sample for testing)

---

📁 Repository Structure

alberta-wildfire-analysis/
│
├── Wildfire_DataStory_Enhanced.ipynb    # Main analysis notebook ⭐
├── README.md                             # This file
├── requirements.txt                      # Python dependencies
├── LICENSE                               # MIT License
├── .gitignore                           # Git ignore rules
│
├── data/
│   ├── README.md                        # Data source info
│   └── wildfire_data.csv                # Dataset (download separately)
│
├── images/                              # Visualizations
│   ├── eda_*.png                       # Exploratory charts
│   ├── q1_*.png                        # Question 1 visuals
│   ├── q2_*.png                        # Question 2 visuals
│   └── ...                             # All generated charts
│
└── docs/                               # Additional documentation
    ├── methodology.md                  # Detailed methods
    └── data_dictionary.md              # Variable definitions

---

🛠️ Technical Stack

Data Science & ML

Library	Purpose	Version
pandas	Data manipulation	2.0+
numpy	Numerical computing	1.24+
scipy	Statistical analysis	1.10+
scikit-learn	Machine learning	1.3+

Visualization

Library	Purpose	Version
matplotlib	Static plots	3.7+
seaborn	Statistical graphics	0.12+
plotly	Interactive charts	5.14+

Geospatial

Library	Purpose	Version
geopandas	Geographic data structures	0.13+
pyproj	Coordinate transformations	3.5+
contextily	Basemap tiles	1.3+
shapely	Geometric operations	2.0+

Coordinate System: EPSG:3403 (NAD83 Alberta 10-TM Forest)

---

📈 Analysis Workflow

1. Data Loading & Profiling

• Import 26,551 fire records
• Assess data quality (completeness, types, distributions)
• Identify missing data patterns

2. Data Preparation

• Handle missing values appropriately
• Engineer features (Fire Weather Index, periods, regions)
• Convert dates and categorize variables

3. Exploratory Data Analysis

• Visualize distributions
• Identify temporal and spatial patterns
• Compute initial correlations

4-9. Six Research Questions

• Each question follows: Motivation → Methods → Analysis → Findings → Implications
• Statistical rigor: hypothesis tests, significance levels, confidence intervals
• Multiple visualization types for each question

10. Machine Learning

• Unsupervised: K-means clustering (k=4) to discover fire types
• Supervised: Random Forest to predict fire size categories
• Validation: Silhouette scores, confusion matrices, precision/recall

11. Synthesis & Conclusions

• Connect findings across questions
• Identify actionable insights
• Acknowledge limitations
• Recommend next steps

---

📊 Sample Visualizations

Temporal Trends (Question 1)

Annual fire frequency shows high year-to-year variability with extreme years (2016, 2019, 2023) rather than a consistent linear increase.

Spatial Clustering (Questions 2 & 6)

Geographic analysis reveals three distinct fire environments: Northern boreal (remote, lightning-caused), Central transition zone, and Southern grassland (human-caused).

Machine Learning Results (Question 6)

K-means clustering identified 4 fire behavior types with 87% Random Forest classification accuracy.

Note: All visualizations are generated automatically when running the notebook

---

🎓 Methodology Highlights

Statistical Methods

• Linear Regression - Trend detection (coefficients, R², p-values)
• Pearson Correlation - Association strength and significance
• Chi-Square Test - Independence testing (categorical variables)
• Hypothesis Testing - α = 0.05 significance level throughout

Machine Learning

• K-Means Clustering

  • Optimal k selection via elbow method and silhouette scores
  • Feature standardization (StandardScaler)
  • 9 variables: weather, location, timing, fire characteristics

• Random Forest Classification

  • 70/30 train/test split
  • Hyperparameter tuning via grid search
  • Performance metrics: accuracy, precision, recall, F1
  • Feature importance analysis

Geospatial Analysis

• Projection: EPSG:3403 (NAD83 Alberta 10-TM Forest)
• Grid Resolution: 5km × 5km cells
• Smoothing: Gaussian kernel (σ=2.5 cells)
• Density Mapping: 2D histograms with interpolation

---

💡 Key Findings & Implications

For Fire Managers

✅ Pre-position resources based on identified geographic hotspots
✅ Use cluster profiles for initial fire risk assessment
✅ Differentiate strategies by region (North vs. Central vs. South)
✅ Peak suppression capacity needed June-August

For Policy Makers

✅ Evidence supports regional (not province-wide) strategies
✅ Invest in northern detection (helicopters, remote sensing)
✅ Invest in southern prevention (public education, fuel mgmt)
✅ Climate adaptation: prepare for high-variability future

For Researchers

✅ Demonstrates ML feasibility for fire classification
✅ Identifies data gaps (fuel moisture, suppression effort)
✅ Provides baseline for climate change studies
✅ Methodology transferable to other regions

---

⚠️ Limitations & Caveats

Data Quality

• 60% missing environmental data (weather measurements)
  • Small fires receive abbreviated assessments
  • May over-represent larger fires in correlations
  • Complete case analysis is valid but introduces bias

Analytical Scope

• Correlation ≠ Causation - We show associations, not proven causes
• 20-year window - May be too short for climate trend detection
• Suppression effects - Final fire size reflects both behavior AND firefighting
• Missing variables - Fuel moisture, suppression effort, economic costs

Model Limitations

• 87% accuracy - Means 13% error rate (168 large fires misclassified)
• Cannot replace experts - Models support, don't replace human judgment
• Temporal validity - Patterns may shift with climate change

See notebook for complete limitations discussion

---

🤝 Contributing

Contributions welcome! Please follow these guidelines:

How to Contribute

1. Fork the repository
2. Create a feature branch (git checkout -b feature/improvement)
3. Commit your changes (git commit -m 'Add improvement')
4. Push to branch (git push origin feature/improvement)
5. Open a Pull Request

Areas for Contribution

• 🔬 Additional analyses - Temporal forecasting, fuel type deep-dive
• 📊 New visualizations - Interactive dashboards, animated maps
• 🤖 Model improvements - Alternative ML algorithms, ensemble methods
• 📝 Documentation - Data dictionary expansion, methodology details
• 🐛 Bug fixes - Code optimization, error handling

Code Style

• Follow PEP 8 guidelines
• Add docstrings to functions
• Include comments for complex logic
• Update requirements.txt if adding dependencies

---

📧 Contact & Support

Questions? Open an issue

Project Maintainer:

• Name: Ifeanyi Njoku
• Email: ifeanyinjoku2@gmail.com
• LinkedIn: www.linkedin.com/in/ifeanyi-e-njoku
• Portfolio: https://github.com/cnero101/alberta-wildfire-analysis.git

Want to collaborate? Reach out directly or open a discussion!

---

📜 License

This project is licensed under the MIT License - see LICENSE file for details.

Data License

Alberta Wildfire data is public domain (Government of Alberta).
Attribution required: "Data provided by Alberta Wildfire Management"

---

🙏 Acknowledgments

Data Source

• Alberta Wildfire Management for maintaining comprehensive public records
• Government of Alberta for open data commitment

Inspiration & Support

• Fire management professionals whose expertise keeps communities safe
• Data science community for tools and best practices
• Open source contributors for excellent libraries

Tools & Technologies

• Python ecosystem (pandas, scikit-learn, matplotlib, geopandas)
• Jupyter Project for notebook environment
• GitHub for version control and collaboration

---

📚 References & Resources

Data Source

• Alberta Wildfire Historical Data: https://wildfire.alberta.ca/resources/historical-data/
• Canadian Wildland Fire Information System: https://cwfis.cfs.nrcan.gc.ca/

Relevant Literature

1. Flannigan, M., et al. (2013). Global wildland fire season severity in the 21st century. Forest Ecology and Management.
2. Rodrigues, M., & de la Riva, J. (2014). An insight into machine-learning algorithms to model wildfire susceptibility. Environmental Modelling & Software.
3. Tymstra, C., et al. (2010). Development of Prometheus: Canadian Wildland Fire Growth Model. Natural Resources Canada.

Related Projects

• FireSmart Canada: https://www.firesmartcanada.ca/
• NASA FIRMS (Fire Information for Resource Management): https://firms.modaps.eosdis.nasa.gov/

---

📈 Project Stats

---

🌟 Support This Project

If you found this analysis useful:

⭐ Star this repository
🔀 Fork for your own use
📢 Share with colleagues
💬 Provide feedback
🤝 Contribute improvements

Every star helps make data science research more visible!

---

📅 Version History

v1.0.0 (February 2026)

• Initial release
• Complete 6-question analysis
• Machine learning implementation
• EPSG:3403 geospatial visualization
• Comprehensive documentation

---

🔥 Analyzing wildfires with data science to build a more resilient Alberta

Last Updated: February 2026