Ambient air pollution is well known to affect humans and the environment severely. Some primary sources of air pollution are industrial processes, motorized traffic, wildfires, and volcanic activities. Estimating future air pollution concentrations can lower the effects on humans and the environment by taking appropriate measures that, for example, restrict or distribute air pollution sources.

The proposed research will evaluate air pollution predictions in three different scenarios.
First, a single ground-level sensor forecast of different pollution concentrations will be performed using historical air pollution concentrations, meteorological data and the forecast of the established forecasting model for Europe. Second, the current pollution concentration and other variables from neighbouring measurement sensors are also included as inputs to evaluate if the performance increases. Third, the historical pollution data and station-relevant parameters are excluded from the target station during training and only used for evaluation. All forecasts will be performed for the next 24 hours.

A dataset will be collected, analyzed and preprocessed to be fed to state-of-the-art models identified during the literature review. The results of the different models will be compared with the established forecasting system to evaluate if it can be improved for an urban study area, predicting in near-real time.

This presentation will give a short overview of the topic, possible challenges and the planned steps to answer the research questions mentioned above.