Orkanskader

Harnessing data to forecast hurricane impact and financial losses.

About

orkanskader, a term for "hurricane damage" in Norwegian, is a project intent on analyzing the radial impact of hurricanes based on historical data.

The project consists of a couple of segments,

1. estimating the radial impact of hurricanes based on historically observed data.
2. mapping the affected areas based on damage severity via proximity to the eye of storm.
3. estimating the overall and categorical financial losses incurred by owners of residential properties affected by the storm.
4. if we're lucky, we may add another components to estimate loss to insurance companies!

Team Members

• Venkata Datta Adithya Gadhamsetty
• Matthew Brownrigg
• Beatrix Wen
• Richard Zhang

How to build

This project was built using conda environment manager so it is recommended you use a conda distribution. With conda you can use the orkan.yml file with:

$ conda env create -f orkan.yml

This will install all the requisite libraries to run our files.

Running the Hurricane Range & Location Predictive Model

• Rmax prediction:
  Run Rmax_Prediction.ipynb
• Rmax specific to Florida region:
  Run Rmax_Prediction_FL.ipynb
• Hurricane location prediction:
  Run Deep_Learning_Hurricane_Location.ipynb

Results from the Hurricane Range Prediction

 

Results from the Hurricane Location Prediction

 

 

Running the Damage Model

Use the BASH script run.sh located in src/ to generate a hurricane damage model. To properly run run.sh you need the following parameters:

$ cd src
$ bash run.sh <longitude [-79.0,-84.0]> <latitude [24.0,31.0]> <radius_64 (km)> <radius_max (km)>

The diagram will be generated in data/ as hurricane@(<long>,<lat>&<radius_max>). The estimated damage will appear as printed text on the console.

Results from Running Damage Model

$ bash run.sh -80.1918 25.7617 120 60