HurriCas

Inspiration

Communication and politics are almost more important than Science in Climate Change, so we wanted to build a tool that can help predict direct consequences of an extreme meteorological event on the population.

What it does

Takes data on a storm :

• its path,
• its size,
• its duration

Returns :

• Number of Casualties (death count)
• Damages in dollars

How we built it

We joined the _ ExtremeWeather Dataset _ with the _ The Internation Disaster Database _ to link each event registered in the first one with its consequences in the second one. We then tried to build a model that learned the dependencies between the event and its consequences.

Challenges we ran into

Curating data is always difficult. It was hard to link the two dataset as the events did not have nor common keys neither common geographical index. We achieved a way to link

Accomplishments that we're proud of

We achieved a non trivial merge of the two datasets, giving the scientific community an opportunity to train models to perform what we initially wanted to do.

What we learned

Working with real world data is often complicated. More communication between database maintainer and a more transdisciplinary approach would be necessary to tackle Climate Change. It's a real scientific challenge but is also a key to our survival as a specie, hence worth the struggle.

What's next for HurriCas

We want now to evaluate different regressor models. If necessary we could try to include other data for each meteorological events such as mean temperature of the storm (and other meteorological features aggregated on the storm's path). Adding information on environment, such as GDP or population density of each region could also leverage any bias left in our data.

Built With

• jupyter-notebooks
• python