High-resolution global air pollution mapping has significant social and academic impacts, but is a tremendously challenging task especially in terms of data assimilation and analytics. In this workshop, I will introduce most recent status in global air pollution modelling and evolvement in data (from social science, Earth observations, numerical models), with a focus on explaining various machine learning algorithms (e.g. ensemble trees, deep convolutional neural networks) and overfitting-controlling strategies (e.g. regularization, post-processing), and how they could contribute to global air pollution mapping.