Artificial Neural Networks, Sequence-to-Sequence LSTMs, and Exogenous Variables as Analytical Tools for NO2 (Air Pollution) Forecasting: A Case Study in the Bay of Algeciras (Spain)

Abstract

This study aims to produce accurate predictions of the NO2 concentrations at a specific station of a monitoring network located in the Bay of Algeciras (Spain). Artificial neural networks (ANNs) and sequence-to-sequence long short-term memory networks (LSTMs) were used to create the forecasting models. Additionally, a new prediction method was proposed combining LSTMs using a rolling window scheme with a cross-validation procedure for time series (LSTM-CVT). Two different strategies were followed regarding the input variables: using NO2 from the station or employing NO2 and other pollutants data from any station of the network plus meteorological variables. The ANN and LSTM-CVT exogenous models used lagged datasets of different window sizes. Several feature ranking methods were used to select the top lagged variables and include them in the final exogenous datasets. Prediction horizons of t + 1, t + 4 and t + 8 were employed. The exogenous variables inclusion enhanced the model’s performance, especially for t + 4 (ρ ≈ 0.68 to ρ ≈ 0.74) and t + 8 (ρ ≈ 0.59 to ρ ≈ 0.66). The proposed LSTM-CVT method delivered promising results as the best performing models per prediction horizon employed this new methodology. Additionally, per each parameter combination, it obtained lower error values than ANNs in 85% of the cases.

Modelling multiple seasonalities with ARIMA: Forecasting Madrid NO2 hourly pollution levels

“…On the other hand, we have statistical techniques establishing relations and patterns among historical data. Here we find different methods, such as using linear regression (Agirre-Basurko et al, 2006), ARIMA (Garg and Jindal, 2021), Prophet (Shen et al, 2020) and (Garg and Jindal, 2021), Artificial Nerural Network (ANN) (Agirre-Basurko et al 2006 andGonzález-Enrique et al 2021), Convolutional Neural Network (CNN) (Garg and Jindal, 2021), Long Short Term Memory (LSTM) (Garg andJindal 2021 andGonzález-Enrique et al 2021) or an ensemble of multiple models predictions (Medrano et al, 2021). None of these methodologies addresses the multiple seasonalities modelling, although they obtain good forecasting performance.…”

“…In many cities and regions continuous monitoring of NO 2 levels are carried out and there are action protocols in case high concentration levels are observed or expected (see, for instance, González-Enrique et al 2021 andMedrano et al 2021). The main objective of this study is to predict the time series of hourly levels of NO 2 , which would serve as a complementary tool to the action protocols.…”

Data-Driven Framework for Understanding and Predicting Air Quality in Urban Areas

“…The pollutants in our proposed framework will be monitored through reliable sensors. Inclusion of weather parameters in air quality modeling has shown promising results (Kalajdjieski et al, 2020;Gonzlez-Enrique et al, 2021). Deep Learning models mainly LSTM based RNNs are being popularly used for both univariate and multivariate (with exogenous features) time series pollutant data.…”

“…Deep Learning models mainly LSTM based RNNs are being popularly used for both univariate and multivariate (with exogenous features) time series pollutant data. Different configurations of LSTM mainly cross-validation procedure for time series (LSTM-CVT) were compared with basic (Artificial neural networks) ANNs by Gonzlez-Enrique et al (2021) for NO 2 in the Bay of Algeciras (Spain). It was found that exogenous variables like weather parameters have shown considerable improvement in performance.…”

Air Pollution Forecasting Using Deep Learning

“…A study was made in the Bay of Algeciras, Spain to produce precise forecasts of the NO2 concentrations [22]. In order to create the forecasting models, ANN and LSTM were used in the forecasting process.…”