Our idea originated from the bushfire crisis in Australia. Massive wildfires have burned over 15 million acres of land across Australia in the current fire season. Two dozen people have been killed. Tens of thousands have been displaced. Hundreds of millions of animals from across Australia's unique ecosystems have lost their lives. The environment has been impacted with increased CO2 levels contributing to increased air pollution and climate change. This increased air pollution has posed a risk to public health by exposing people to more toxic fumes.

What it does

Our solution consists of three parts.
1) Machine Learning Algorithm: Using our developed ML algorithm to predict wildfires based on certain parameters like Fine Fuel Moisture Code (FFMC) , Wind speed, Humidity, Duff Moisture Code (DMC) etc.
2) Web application: Embedding our algorithm into a web application making it easier for stakeholders (eg. governments who would want to monitor the situation actively) to enter input parameters of a given time in order to view accurate predictions. Alongside, makes it easier to demo and pitch the concept to key stakeholders.
3) Twilio Messaging Interface: Using the Twilio messaging interface to alert communities within 5 km radius about a possible wildfire.

Impact of our solution

1) Decreased contribution to Climate Change and Air Pollution: By being able to control wildfires, the potential negative impact to climate change is being cut. Air pollution levels are decreased.

2) Improved Public health: By decreasing air pollution levels, people inhale fewer toxic fumes.

3) Saving animal lives: Saving the lives of animals by allowing authorities to take needed action (such as rescue) at the right time.

4) Allowing people to be better-prepared: Allowing communities of people to be better prepared in terms of resources and travel

How we built it

We built the ML algorithm using Python. We built the web app using bottle and flask in Python too. The twilio messaging interface was built using twilio, javascript and node.js.

Challenges we ran into

We had to seek lots of online support to learn how to embed our ML model into a web app.

Accomplishments that we're proud of

Being able to create a web app and learning the concepts behind the code

What we learned

We learnt how to use Twilio and how to make a web app!

What's next for Predicting Wildfires and Alerting Communities

As of now, our ML algorithm is not perfect. Our RMSE score is fairly large. This could be attributed to various problems such as overfitting of data or the data itself being very noisy. Another factor could be easy human intervention in spreading forest fires, which is hard to account for in data. So, our aim would be to gather a clean dataset and optimise our ML model accordingly. We would also love to beautify how our web app looks.

Share this project: