Air Pollution Modelling and Source Inference: My Dissertation Journey
I never knew I would be doing data in environmental science. In fact, I did not even put this project in my top choices.
I have always loved working with data, analysing patterns and drawing insights. I ended up with a dissertation on Air Pollution Modelling and Source Inference.
I have always had a soft spot for environmental science, but I never thought I could contribute to it as a Computer Science student.
The Framework
I worked with a cutting-edge research model developed by my lecturer and his team — the Advection Gaussian Process (Advection GP) framework.
This model had only ever been tested on synthetic data, where everything can be controlled. My job? See if it could work on real-world data.
That sounds simple — but it was not.
Since this had never been done before, I had to figure out a lot on my own. Along the way, I ended up providing valuable insights for my supervisor's research.
Challenges of Real-World Data
Applying theory to reality meant dealing with:
- Satellite data uncertainty — AOD reanalysis vs raw satellite data
- Column-integrated AOD — no fixed height, yet wind varies by altitude. I had to explore ways to convert AOD into PM2.5
- 2D was not enough — vertical transport mattered, so we needed 3D
- Map projections — the Earth being round complicates things
- Understanding the limits of NASA reanalysis data
- Reading and adapting complex research code
- Figuring out how to validate and visualise outputs
At times, it felt like I was doing everything: from learning pollution transport physics, to finding validation data, doing literature reviews on different methods of source inference, visualisation, and understanding satellite-derived measurements.
Case Study
My case study focused on the 2019–2020 Australian bushfires, modelling how pollution spread across Victoria.
Along the way, I learned so much — from uncertainty in environmental modelling to the messiness of real-world data.
Outcome
I achieved a Distinction (75) for this dissertation.
But more than that, I am proud of the learning journey. This went far beyond the classroom; it involved self-driven research, weekly planning, and constant learning.
I would not have made it through without the support of my amazing supervisor, Dr. Michael Smith, who encouraged me at every step. This project gave me a real sense of purpose in my degree, like I was actually contributing to society.
Looking Forward
When people ask what I want to do after graduation — I think it is something like this. Insha Allah, I would love to return to this field one day. But first, I am looking for a role where I can continue growing — especially in using data to solve real-world problems.
The code is adapted from my lecturer's original model, which was tested only on synthetic data. My forked repo builds on it to apply the framework to real-world conditions.