Hurricane Forcasting Using Deep Learning

This project aims to use machine learning and deep learning for hurricane forecasting. This mainly includes generating forecasting products that are useful to decision-makers, such as track (position + intensity) forecasts.

To evaluate the forecast, I compare the forecast errors from my models to the forecast errors of the official National Hurricane Center (NHC) forecasts. The goal is to achieve a relatively similar or smaller error than the official NHC forecast error.

Traditionally, the National Hurricane Center (NHC) models are simulation-based and compute-intensive. If I can achieve comparable performance but with much fewer computing resources, then there may be some interesting insights that can be discovered.

Current Results

Below is some current results of all the models I have run. This is all evaluated on the test data split.

Plot of the track forcast error of various models
Objective Function Plots with Solution