My background is primarily in research but my interests lie in the realm of programming. I believe in making my work easily accessible which is where software comes in - I try to build to packages and frameworks that researchers can use to test my research on their own work. Whilst I come from a physics background, I have tailored my degree to use some of its founding principles to drive my studies with big data and programming. You can read more about my work on the Gulf Stream 'Big Data' below.
The Gulf Stream is a warm ocean current that flows off the east coast of the US towards Northern Europe (into the North Atlantic Drift). It is thought that up to 80% of the warm climate in northern Europe can be directly attributed to the Gulf Stream & the North Atlantic in general. In particular the large amounts of heat transported by strong flowing waters is considered to be the most direct pathway for warm European climate.
In addition to this, the Gulf Stream facilitates the development of storms which travel across the North Atlantic towards Europe. This is another way in which the Gulf Stream directly causes European climate variability. Many authors have noted that Gulf Stream influences storms that travel along this 'storm track' towards Europe, but an exact, accepted mechanism for their enhancement and variability are unknown.
→ read my literature review for a more detailed overview of recent physics research on the Gulf Stream
Whilst these mechanisms are unknown, more fundamental questions are yet to be rigourously answered. It is hoped that answering these question will provide a better working knowledge to develop a full physical mechanism that can be later parameterised and incorporated into the next generation of climate models. One of these question is regarding the time variability of the Gulf Stream itself
My current work involves looking at extremely large datasets for the entire planet, ocean and atmosphere, in order to try and find patterns of variability and work out their root physical causes in order to make better predictive climate models. Whilst I am at an early stage in this research, machine learning has become an invaluable tool. The figure shown above is an example of one typical machine learning technique commonly employed, principle component analysis (PCA) or in meterology, empirical orthogonal function (EOF) analysis.
The figure on the right showing the variation of the Gulf Stream in time (from Jan 2004 to Jan 2016) can be summarised by a list of these orthogonal modes and the variation in time of these modes - it's very similar to a Fourier analysis in some ways. Another 'newer' machine learning algorithm that is employed is called the 'diffusion map and spectral clustering' (DMSC) algorithm which is used to track the Gulf Stream meandering.
→ find details of one way machine learning is used in this paper
During a research internship, I was working with numerical models and analysising their output. The aim was to isolate the physics of how the Gulf Stream contributed towards the mean climate, both locally and upstream towards Northern Europe. A series of idealised, aquaplanet simulations were run and a breakthrough in physics behind extratropical cyclones and the Gulf Stream was discovered.
→ see this academic poster for some information about the preliminary results of these simulations - more details are to follow