Forecasting Smog Clouds With Deep Learning: A Proof-Of-Concept

Valentijn Oldenburg, Juan Cardenas-Cartagena, Matias Valdenegro-Toro

Keywords: Time series, Deep Learning, Pollution, Meteorology, Recurrent Neural Networks

TL;DR: This study investigates deep learning models, particularly LSTMs and GRUs, for multi-location air pollution forecasting of NO2, O3, PM10 & PM2.5, with a hierarchical GRU emerging as the most effective method.

Abstract: In this proof-of-concept study, we conducted a multivariate time-series forecasting for concentration of nitrogen dioxide (NO2), ozone (O3), and (fine) particulate matter (PM10 \& PM2.5) with meteorological covariates between two locations in the Netherlands using various deep learning models, with a focus on long short-term memory (LSTM) and gated recurrent unit (GRU) architectures. In particular, we propose an integrated, hierarchical model architecture inspired by air pollution dynamics and atmospheric science that employs multi-task learning and is benchmarked by unidirectional and fully-connected models. Results demonstrate that, above all, the hierarchical GRU proves itself as a competitive and efficient method for forecasting the concentration of smog-related pollutants.

Submission Number: 48