title	author	date	output
Hurricane Track & Intensity Investigation	Henry Blassingame	02 OCTOBER 2020	html_document

Capstone Project

Henry Blassingame

September 2020

---

Problem

• Review of over 100 years of Hurricane Data from the National Weather Service. Create some visualizations to look at the increase in number of storms and increase in storm intensity over time.

• Compare Linear Regression to Random Forest to see which one performs better at predicting the number of storms per year.

Data Source:

The project is based on "Hurricanes and Typhoons, 1851-2014", provided by the National Oceanic and Atmospheric Administration (NOAA) and the National Hurricane Center (NHC.) The data can be found at: https://www.kaggle.com/noaa/hurricane-database.

Project Rationale

The reason I selected hurricane data as my project idea is simply that I have always been fascinated by weather, especially in its more severe forms. My Father, who was an Amateur Radio Operator, would go to the Emergency Operations Center in Baytown whenever storms threatened the upper Texas Gulf Coast, and as a young boy I would often accompany him. I was fascinated by the radio equipment and the teletype, and having access to the same information as the weather forecasters on television. Another reason is that I have always been intrigued by the 1900 Storm that destroyed Galveston TX, and which remains to this day the worst natural disaster to strike the United States in terms of lives lost. Possibly another reason was that at the time we were thinking about what to choose for a project, we were in the midst of one of the most active hurricane seasons on record.

Approach

The data set includes storms from 1851 to 2015, with varying degrees of completeness (and presumably accuracy, especially for the older storms.) Furthermore, there are problems with the formatting of dates in the dataset, as well as observation times, which are stored separately from the dates. Another problem is that Latitude and Longitude are stored as strings, and need to be re-formatted as numbers and adjustment made for longitudes East and West of Greenwich.

Due to the incompleteness of much of the data before 1900, this was removed before visualization and analysis.

Tools Used in the Project

Data Cleanup and Manipulation

• Excel
• RStudio

Data cleanup was performed in two phases: one in Excel, and another in R. One problem with data cleanup in Excel is that it does not work with dates before 1900. So, while many tasks (re-formatting times, fixing Latitude and Longitude, calculating storm categories based on wind speed) are very simple to perform in Excel, the date manipulation was best left to R.

Due to the incompleteness of much of the data before 1900, this was removed before visualization and analysis.

Visualization

• RStudio
• Tableau

The visualizations for landfall on the US coast and the bar and whiskers plot of landfall wind speeds by state were based on Giovanni Maccioni's Kaggle page (https://www.kaggle.com/gi0vanni/analysis-on-us-hurricane-landfalls/report), but modified based on my own data cleanup and manipulation, and to modify several other aspects.

The graph of number of storms per year broken down by category/strength is a link to a previous project I completed in Tableau.

Machine Learning

The challenge with this data was to try to find a good fit for Machine Learning, and to decide on which method made the most sense. The data did not lend itself to predicting direction (without some heavy manipulation to derive this from the coordinates and next/previous observations) or even storm strength. At the end, the simplest metric to determine was how many storms there would be in a given year based on the data.

This was modeled using both Linear Regression and Random Forest models, and the residuals from both were plotted, showing a higher correlation between the data and the Random Forest model.

Repository Contents:

Filename	Description
README.md	This document.
Atlantic_Updated.csv	Hurricane data, cleaning performed in Excel.
Atlantic.csv	Hurricane Data in its raw format, prior to any scrubbing
BLASSINGAME_CAPSTONE.rmd	R Markdown file for the project
BLASSINGAME_CAPSTONE.html	knitted output of the project