05/31/18

Data analysis of the sinking of SS El Faro by hurricane Joaquin.

https://github.com/raybellwaves/SSElFaro

Update Dec 2019: Paper published https://www.tandfonline.com/doi/full/10.1080/1755876X.2019.1684136

On October 1st 2015 the cargo ship SS El Faro (which translates to the lighthouse) sunk amidst dangerous seas caused by hurricane Joaquin during a routine crossing from Jacksonville, FL to San Juan, PR. For general information about the sinking I recommend the wikipedia page and the National Transport Safety Board (NTSB) youtube video.

I wasn't aware of this event as I had not moved to the US yet, but did stumble across some chatter regarding the sinking on the tropical storms list. A couple of weeks ago Diana Udel the director of communications at RSMAS mentioned the discovery channel were visiting RSMAS to talk to researchers about the hurricane for an upcoming documentary. She gave them my details but eventually they only spoke to Prof. Dave Nolan about the hurricane.

I have a particular interest in this kind of work as I studied hurricanes for my PhD and this did a post-doc on extreme waves with BP. Part of my work with BP looked at Vessel response to sea-states. You can calculate the six degrees of freedom of a vessel if you know the wave spectrum and the shape and weight of a vessel (see here for a bit more detail).

I was curious as what kind of environmental conditions did the crew found themselves in. Due to a series of unfortunate events they found themselves in the eye of a slow moving category 4 hurricane. The slow moving nature made the seas more dangerous as energy is continually being transferred from the winds to the waves allowing the waves to grow larger.

A vessel responds to forces from the winds, ocean surface waves and ocean surface currents. I wanted to obtain data of the all these for this case study.

I obtained data for the winds and waves from ERA5 which is ECMWF's new re-analysis product (a reanalysis is a model run of previous weather when known observations are incorporated so it is as close to reality as possible). ERA5 is fairly high resolution (1/4 degree) and provides hourly data. It provides the 10-m U and V winds and a variety of ocean wave parameters. The ocean data forcing ERA5 (ORAS5) is not available yet so I used HyCOM data.

I wanted to explore the data interactively. I was not making publication plots! There are some great tools to do this using python such as ipython, jupyter-notebook, jupyterlab, holoviews and xvelmap. I can also run the code on the cloud using mybinder.org making the animations easy to share.

The tools worked fairly flawlessly apart from xvelmap which I had to improve slightly. Such is the beauty of open source: someone else can help you with your work.

The first gif below showed the last location of the vessel as the white dot and the the position of the hurricane from September 29th to October 3rd. The next gif shows the ocean currently velocity field. Couple of things are worth pointing out here. The hurricane went further south than the forecast, hence the captain thought he would miss it. The stalling of the hurricane is also very obvious which created large waves (see picture below of a time series of maximum wave height at the last known location). For the ocean currents, the Antilles current is reversed (it is going east instead of west due to the hurricane). This caused a dangerous situation when they lost power they were being 'sucked' into the storm by the winds and the currents.

You can play the data yourself and view the code on github and I may write this up a natural hazards paper at some point.