Prediction models of CO, SPM and SO2 concentrations in the Campo de Gibraltar Region, Spain: a multiple comparison strategy

Abstract

The 'Campo de Gibraltar' region is a very industrialized area where very few air pollution studies have been carried out. Up to date, no model has been developed in order to predict air pollutant levels in the different towns spread in the region. Carbon monoxide (CO), Sulphur dioxide (SO 2 ) and suspended particulate matter (SPM) series have been investigated (years 1999-2000-2001). Multilayer perceptron models (MLPs) with backpropagation learning rule have been used. A resampling strategy with two-fold crossvalidation allowed the statistical comparison of the different models considered in this study. Artificial neural networks (ANN) models were compared with Persistence and ARIMA models and also with models based on standard Multiple Linear Regression (MLR) over test sets with data that had not been used in the training stage. The models based on ANNs showed better capability of generalization than those based on MLR. The designed procedure of random resampling permits an adequate and robust multiple comparison of the tested models. Principal component analysis (PCA) is used to reduce the dimensionality of data and to transform exogenous variables into significant and independent components. Short-term predictions were better than medium-term predictions in the case of CO and SO 2 series. Conversely, medium-term predictions were better in the case of SPM concentrations. The predictions are significantly promising (e.g., d SPM 24-ahead =0.906, d CO 1-ahead =0.891, d SO2 1-ahead =0.851).

Time effects of high particulate events on the critical conversion point of ground-level ozone

“…The results showed that the O 3 diurnal variation of each site exhibited a similar pattern, whereas the magnitudes of variations during HPE and non-HPE differed. The diurnal pattern of O 3 for each site is characterised by a maximum concentration in the afternoon and a minimum concentration during nighttime (Turias et al, 2008;Jones and Kirby, 2009) parallel to variations in solar radiation intensity during the day, which is a favourable condition for promoting photochemical reactions. In normal non-HPE days, solar radiation with a wavelength of less than 400 nm has sufficient energy to photolyse NO 2 into NO and oxygen atom (O) (Seinfeld and Pandis, 2006;Ghazali et al, 2010).…”

Experimental Study and ANN Dual-Time Scale Perturbation Model of Electrokinetic Properties of Microbiota

“…In agreement with the previously mentioned concepts, some non-linear Machine Learning methods can be combined with time-series analysis methods, such as two of the most known examples of each area of Autoregressive Integrated Moving Average (ARIMA) and non-linear Artificial Neural Network (ANN). ARIMA and ANN algorithms have been compared to each other in time-series problems (Turias et al, 2008). However, the ARIMA and non-linear ANN can be further integrated to seek hybrid ARIMA–ANN models for different time-series issues (Babu and Reddy, 2014).…”

Using wavelet–feedforward neural networks to improve air pollution forecasting in urban environments

“…The applied methods are diversified from statistical methods, artificial intelligence (AI) techniques, and probabilistic approaches to hybrid algorithms and complex models. Statistical methods (i.e., Multiple Linear Regression, Auto-Regression including ARMA, non-seasonal and seasonal ARIMA, Vector Auto Regression, and Autoregressive Conditional Heteroskedasticity-ARCH) and various stochastic algorithms (i.e., Hidden Markov Model, Singular Spectrum Analysis, and Cross Spectrum Analysis) have been applied successfully for real-time control of emissions showing good short-term forecasting capabilities for several air pollutants (Nunnari et al 2004;Zolghadri and Henry 2004;Modarres and Khosravi Dehkordi 2005;Turias et al 2008;Wu and Kuo 2012;Iordache and Dunea 2013).…”