Systematic Approach for the Prediction of Ground-Level Air Pollution (around an Industrial Port) Using an Artificial Neural Network
Abstract
The prediction of air pollution levels is critical to enable proper precautions to be taken before and during certain events. In this paper a rigorous method of preparing air quality data is proposed to achieve more accurate air pollution prediction models based on an artificial neural network (ANN). The models consider the prediction of daily concentrations of various ground-level air pollutants, namely CO, PM10, NO, NO2, NOx, SO2, H2S, and O3, which were measured by an ambient air quality monitoring station in Ghadafan village, located 700 m downwind of the emissions of Sohar Industrial Port on the Al-Batinah coast of Oman. The training of the models is based on the multi-layer perceptron (MLP) method with the Back-Propagation (BP) algorithm. The results show very good agreement between the actual and predicted concentrations, as the values of the coefficient of multiple determinations (R2) for all ANN models exceeded 0.70. The results also show the importance of temperature in the daily variations of O3, SO2, and NOx, whilst the wind speed and wind direction play significant roles in the daily variations of NO, CO, NO2, and H2S. PM10 concentrations are influenced by almost all the measured meteorological parameters.
Figures
- CitationsCitations8
- ReferencesReferences27
- "In these studies, black-box models for ozone prediction – obtained with a range of linear and nonlinear regression methods from Principal Component Regression to Takagi-Sugeno fuzzy models – are used for different geographical regions and for different objectives of the ozone prediction. It is possible to find models for the prediction of hourly ozone values, e.g., (Al-Alawi et al., 2008; Duenas et al., 2005; Feng et al., 2011; Lin and Cobourn, 2007; Petelin et al., 2013; Solaiman et al., 2008), daily maximum ozone values, e.g., (Baawain and Al-Serihi, 2014; Chelani, 2010; Cheng et al., 2011; Fontes et al., 2014; Grašič et al., 2006; Moustris et al., 2012; Nebot et al., 2008; Sundaramoorthi, 2014), or different average ozone values, e.g., (Fontes et al., 2014; Garner and Thompson, 2013; Sun et al., 2013; Sundaramoorthi, 2014), to list only a selection of recent publications. These models use various pollutants and various meteorological variables, together with their lagged values, as the regressors. "
[Show abstract] [Hide abstract] ABSTRACT: It is important to be able to predict high concentrations of tropospheric ozone and to inform the population about any violations of air-quality standards, as defined by international regulations. Although first-principle models that cover large geographical regions and different atmospheric layers are improving constantly, they typically still only cover geographical regions with a relatively low resolution. Such model predictions can be problematic for the micro-locations of a complex terrain, i.e., a terrain with a large geographical diversity or urban terrain. For such micro-locations, statistical models can be utilised. This paper presents a modelling and prediction algorithm that can be used in, or in accordance with, a mobile air-quality measurement station. Such a mobile station would enable the set-up of a statistical model and a relatively rapid access to the model's predictions for a specific geographical micro-location without a large quantity of historical of measurements. Uncertainty information about the model's predictions is also usually required. In addition, such a model can adapt to long-term changes, such as climate changes. In the paper we propose Gaussian-process models for the described modelling and prediction. In particular, we selected evolving Gaussian-process models that update on-line with the incoming measurement data. The proposed algorithm for the mobile air-quality measurement and the forecasting station is evaluated on measurements from five locations in Slovenia with different topographical and geographical properties. The obtained evaluation results confirm the feasibility of the concept.- "Air quality forecasting using statistical techniques has been performed in several studies. Statistical models including autoregressive models (Zennetti, 1990) and neural networks (Gardner and Dorling, 1998; Baawain and Al-Serihi, 2014) have been extensively used in the air quality forecasting. knearest neighbour method of forecasting is a simple machine learning based method, which determines the nearest neighbours of an object in question and use those in estimating the object. "
[Show abstract] [Hide abstract] ABSTRACT: Air quality forecasting using nearest neighbour technique provides an alternative to statistical and neural network models, which needs the information on predictor variables and understanding of underlying patterns in the data. k-nearest neighbour method of forecasting that does not assume any linear or nonlinear form of the data is used in this study to obtain the next step forecast of PM10 concentrations. Various function approximation techniques such as mean, median, linear combination and kernel regression of nearest neighbours are evaluated. It is observed that kernel regression of nearest neighbours outperforms the other individual models including bench mark persistence model for obtaining the next step forecasts. As the data may involve both linear and nonlinear patterns and any individual model cannot capture both types of patterns, combination forecasting is suggested as an alternative. The forecast error showed the outperformance of combination forecasting over individual forecast, which is quite obvious as it assigns more weightage to the model with minimum error. The study is useful when the data on predictor variables that influence the air pollutant concentrations is not available. The assumption on the underlying distribution of the data is also not required for the approach.- [Show abstract] [Hide abstract] ABSTRACT: This paper presents a design of models for common air quality index prediction using computational intelligence methods. In addition, the sets of input variables were optimized for each air pollutant prediction by genetic algorithms. Based on data measured by the three monitoring stations of Dukla, Rosice and Brnenska in the Czech Republic, the models were designed to predict air quality indices for each air pollutant separately and, consequently, to predict the common air quality index. Considering the root mean squared error, the results showed that the compositions of individual prediction models significantly outperform single prediction models of the common air quality index. The feature selection procedure indicates that the determinants of air quality indices were strongly locality specific. Therefore, the models can be applied to obtain more accurate one day ahead predictions of air quality indices. Here we show that the composition models achieve high prediction accuracy for maximum air quality indices (between 50.69 and 63.36%). The goal of the prediction by various methods was to compare the results of the prediction with the aim of various recommendations to micro-regional public administration management.
Data provided are for informational purposes only. Although carefully collected, accuracy cannot be guaranteed. Publisher conditions are provided by RoMEO. Differing provisions from the publisher's actual policy or licence agreement may be applicable.
This publication is from a journal that may support self archiving.
Learn more