Space-Time Prediction of PM2.5 Concentrations in Santiago de Chile Using LSTM Networks

Abstract

Currently, air pollution is a highly important issue in society due to its harmful effects on human health and the environment. The prediction of pollutant concentrations in Santiago de Chile is typically based on statistical methods or classical neural networks. Existing methods often assume that historical values are known at a fixed geographic point, such that air pollution can be predicted at a future hour using time series analysis. However, these methods are inapplicable when it is necessary to know the pollutant concentrations at every point of the space. This work proposes a method that addresses the space-time prediction of PM2.5 concentration in Santiago de Chile at any spatial points through the use of the LSTM recurrent network model. In particular, by considering historical values of air pollutants (PM2.5, PM10 and nitrogen dioxide) and meteorological variables (temperature, wind speed and direction and relative humidity), measured at fixed monitoring stations, the proposed model can predict PM2.5 concentrations for the next 24 h in a new location where measurements are not available. This work describes the experiments carried out, with particular emphasis on the pre-processing step, which constitutes an important factor for obtaining relatively good results. The proposed multilayer LSTM model obtained R2 values equal to 0.74 and 0.38 in seven stations when considering forecasts of 1 and 24 h, respectively. As future work, we plan to include more input variables in the proposed model and to use attention-based networks.

Deep-learning architecture for PM2.5 concentration prediction: A review

“…[ 62 ] 2022 Ningxia, China PCA-Attention-LSTM D/S/T+1 7.57 4.93 - 0.91 Peralta et al. [ 48 ] 2022 Santiago, Chile LSTM H/S/T+1 9.85 4.40 - 0.74 Liu et al. [ 71 ] 2022 Jing-Jin-Ji Region, China MGC-LSTM H/S/T+1 2.91 2.16 12.96 - Hu et al.…”

A Control-Theoretic Spatio-Temporal Model for Wildfire Smoke Propagation Using UAV-Based Air Pollutant Measurements

“…It is also possible to learn such models offline using more powerful computers and then compute the model online onboard UAVs using their limited computational power. Machine learning techniques, such as deep learning, are the current state-ofthe-art for modeling highly nonlinear, high-dimensional and complex real-world systems that cannot be conveniently described by physical laws [33][34][35][36][37][38][39][40]. However, the downside to these data-driven models is the large amounts of data required to learn the models.…”

Advancing air quality forecasting in Abu Dhabi, UAE using time series models

“…Similarly, (Katsoulis and Pnevmatikos, 2009), successfully applied ARIMA models to predict daily PM10 concentrations in Athens, Greece, showcasing the model's adaptability across diverse environmental settings. Comparative analyses, such as the study by (Peralta et al, 2022), which evaluated neural networks against ARIMA models for air pollution forecasting in Santiago, Chile, revealed that despite neural networks' marginally better accuracy, ARIMA models' simplicity and interpretability render them a practical option for air quality prediction. The employment of statistical models for air quality assessment in developing nations has been particularly noteworthy.…”