Roadside Air Pollution Analysis

A series of Jupyter notebooks assessing the impact of Covid-19 lockdown on air pollution readings from UK roadside sensors.

An in-progress exploratory analysis of air pollution data from roadside sensors. Initial data processing pipeline has been completed by combining external package functions imported and consumed in Jupyter notebooks. I hope to identify and visualise impacts of Covid-19 lockdowns on air pollution readings at roadside sensors. For example, any shifts in distribution of pollution levels across weeks; general increase/decrease; rise/fall in particular kinds of pollutants; etc.

I'm also interested to see whether a ML time-series analysis can predict future unseen years of air pollution data with appreciable accuracy. If so, it may be interesting to compare predicted vs. actual air pollution measurements for 2020.

Getting Started

1. Clone or download the repository:

git clone https://github.com/Simon-Lee-UK/air-pollution.git  # Grabs the code from GitHub
cd air-pollution  # Navigates into the top level of the repository

2. Using the Conda environment and package manager, create the environment required by the analysis notebooks and custom packages. First, create the environment from the environment.yml file:

conda env create -f environment.yml

Next, activate the newly created environment:

conda activate air-pollution-dev

3. Run Jupyter Lab from the repository root:

jupyter lab

4. Follow the instructions within notebooks to download, process and explore the example set of data or analyse data from a new roadside sensor.

Extensions

Currently, data preview and read-in pipelines have been defined in custom package modules. These are generalised, making it simple to apply future notebook logic to new sets of air pollution data from different sensors.

Still need to:

• Create new package module to handle local saving of processed data.
• Create exploratory graphs for example site data set incl. weekly distribution shifts pre-/post-C19 lockdown.
• Perform a time-series analysis to give a predicted set of data for 2020 based on previous years; this can then be compared to the actual readings gathered in 2020.

License

The code in this project is licensed under the MIT License.

The air pollution data accessed in this project is licensed under the Open Government Licence (OGL).

See LICENSE.txt for full licensing details.