The problem

Oil spills cause severe ecological and economic damage. Once oil is in the water it moves with the surrounding flow and can spread across large areas — threatening marine life, coastlines and fishing zones. The practical questions are simple to ask and hard to answer:

• Where does the oil go?
• How fast does it spread?
• How much of it reaches a sensitive area?
• Can we show this in a way decision-makers understand?

Approach

I represented the bay as a triangular mesh loaded from a .msh file. Every triangle is a small computational cell that stores how much oil it currently holds. At each time step, neighbouring cells exchange oil through their shared edges, driven by a velocity field and the geometry of the edge — a finite-volume style upwind scheme.

1. Load the mesh and build cell objects
2. Initialise oil around a source point
3. Pre-compute neighbour relationships and edge normals
4. Step the oil distribution forward in time
5. Integrate the amount of oil inside the fishing-grounds polygon
6. Write frames, an animation and a log for analysis

The goal was not a perfect ocean model, but a clear, testable one that shows how computation can support environmental risk decisions.

Engineering

The code is split into small, testable modules — mesh.py for geometry, cells.py for cell types, Simulator.py for time stepping and Visualizer.py for plots and video. Runs are configured with TOML files and launched from a CLI with argparse; every run writes a structured log. A pytest suite covers mesh loading, cell geometry and the simulator.

python main.py --config input.toml

What I learned

• Turning a physical process into a discrete numerical scheme
• Mesh-based modelling and neighbour graphs
• Time-stepping stability and conservation
• Writing scientific code that is configurable, logged and tested
• Communicating results visually — frames and video, not just numbers