This paper analyses the impact of urban mobility (UM) on air pollution by studying the effects of an intervention on local air quality. The study focuses on the PM2.5 levels in Kampala, the capital of Uganda, and considers COVID-19 as an unintentional intervention. The PM2.5 level of the city was obtained from a network of low-cost calibrated sensors, while UM is characterized by open-access Google reports. The period under consideration excludes the weeks immediately before and after the first lockdown. PM2.5 data were deweathered using the machine learning technique of random forest (RF) to exclude the variation of meteorological factors, seasonality, and weekday-weekend effect, and then the impact of the pandemic was parametrised. The traffic pattern is discussed, and air mass clustering and pollution polar plots are used to analyse the distribution of long- and short-range sources, respectively. The percentage change from the baseline (PCfB) of the average of UM dimensions is then assessed against that of deweathered PM2.5 level to investigate the impact of UM on the PM2.5 level. Our analysis shows a strong correlation between urban mobility and roadside PM2.5 levels and a weaker relationship with urban PM2.5 levels. The profile of long-range emission sources was consistent over the study period, with more than 61% of the modelled air masses that arrived in Kampala first passing over Kenya and Tanzania. Overall, the COVID-19 pandemic reduced PM2.5 levels in Kampala by about 10%, which is relatively small compared to many other cities that have been studied around the world.