According to the WHO wildfires and volcanic activities affected 6.2 million people between 1998-2017 with 2400 attributable deaths worldwide from suffocation, injuries, and burns, but the size and frequency of wildfires are growing due to climate change. They may also affect food security if they happen on farms or displace communities if they happen near residential areas. These problems are the inspiration for a concept to use remote sensing data from satellites to help with prediction of wildfires before they happen to enable effective response.

What it does

Predict the chance of a fire based on the following variables, (NDVI: Normalized Difference Vegetation Index), meteorological conditions (LST: Land Surface Temperature) as well as the fire indicator “Thermal Anomalies” (BURNED_AREA).

How we built it

Using python and scikit-learn inside the jupyter notebook on Amazon's Sagemaker studio lab.

Challenges we ran into

Hyper-parameter tuning of the XGBoost algorithm

Accomplishments that we're proud of

Building the prediction model and running a simulation.

What we learned

Imageio a Python library that provides an easy interface to read and write a wide range of image data, including animated images

What's next for Ignis - Wildfire Prediction

Use reinforment learning to help with the response in the case of a wildfire.

Built With

Share this project: