Predicting Impact and Severity of Wildfires in Alberta

Inspiration

We are inspired by the following important questions:

Which top FSA regions are more vulnerable with wildfires?
What are the main reasons causing wildfires near each vulnerable FSA region?
How to identify the impacts of wildfires on indigenous population and vulnerable population?
How to predict the severity and the final size of area burned by wildfires?

The issues addressed here are important because decision-makers in different industries such as forestry, agriculture etc. can have insights on what strategies to employ in order to preserve indigenous areas and protect life and property.

What it does

Analyze the data provided by the Alberta Fire Department to find out the main causes of wildfires, assess the vulnerable population/regions, and predict the final size of burned area using machine learning.

How we built it

The report began by identifying vulnerable regions in central Alberta based on the frequency of wildfires and the size of the area burned. These regions included Calgary, High Level, Slave Lake, and Fort McMurray. For assessing vulnerable populations, our team considered demographics such as ages 0-14 and 65 and over, as well as Indigenous populations, due to their susceptibility to wildfire impacts. The report detailed steps in data cleaning, engineering, and model development while providing interpretations and comparisons of different models for business insights.

Challenges we ran into

We faced challenges in the data cleaning and feature engineering steps. Fortunately, data for most of the important variables is fairly clean and only a small fraction has missing values which are excluded from the model.

Accomplishments that we're proud of

We are proud of being able to finish every step from end to end in data project including data cleaning, data engineer, and model development. Moreover, the most important accomplishment is being able to deliver insightful information based on available data and external evidence to help decision-makers have more insights regarding wildfires.

What we learned

The key findings are that lightning is the main cause of wildfires and the vulnerable population consists of demographics from 0 to 14 years old, 65 years old and over, and Indigenous people. Furthermore, we also see that the final size of area burned has extreme outliers which affect the predictive model in development.

What's next for Predicting Impact and Severity of Wildfires in Alberta

Potential improvements for the methodologies used in this report could be finding better criteria to define vulnerable FSA regions/populations, coming up with a better algorithm that balances interpretability and predictive power, or performing more comprehensive feature engineer etc.