Research on nonstationary univariate time series modelling for missing value imputation on air quality datasets.
Weather and air pollution monitoring around the world are fundamental to the study of air quality. Having a solid infrastructure of monitoring stations can help reveal how pollution travels throughout regions and also assist in defining pollution hot spots or raising alerts for poor air conditions.
However, a good air monitoring infrastructure is not a given for every country worldwide. While there are currently more than 30,000 known air quality monitoring stations and developed countries tend to have full coverage within their major cities, some have lacking or non-existent facilities to measure air pollutants, especially in African and Middle East countries (World-wide Air Quality Monitoring Data Coverage, 2022).
This research aims at presenting a reliable univariate missing data imputation method for complex air pollution time series data. When the infrastructure is lacking, the data also tend to be of lower availability, higher volatility, and of poorer quality in general. Therefore, the goal is to present an alternative to enhance the available data so it can better serve its purposes.
This repository contains exploratory and development notebooks for the methods in study, as well as master datasets of currently explored air quality index stations.