Development and comparative analysis of tropospheric ozone prediction models using linear and artificial intelligence-based models in Mexicali, Baja California (Mexico) and Calexico, California (US)

Salazar-Ruíz, Enriqueta; Ordieres, J.; Vergara, Eliseo P.; Capuz-Rizo, Salvador F.

doi:10.1016/j.envsoft.2007.11.009

Cited by 69 publications

(42 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The index of agreement was 0.92. In a similar study, Salazar-Ruiz et al [37] have used meteorological data, precursor concentrations, and persistence information as inputs to predict next-day maximum tropospheric ozone levels. R, d 2 , and RMSE for the MLP model developed were 0.74, 0.85, and 9.43 ppb (1 ppb ¼ 1.96 mg/m 3 at 258C), respectively.…”

Section: Resultsmentioning

confidence: 99%

Artificial Neural Network Prediction of Tropospheric Ozone Concentrations in Istanbul, Turkey

İnal

2010

CLEAN Soil Air Water

View full text Add to dashboard Cite

Artificial Neural Network Prediction of Tropospheric Ozone Concentrations in Istanbul, TurkeyTropospheric (ground-level) ozone has adverse effects on human health and environment. In this study, next day's maximum 1-h average ozone concentrations in Istanbul were predicted using multi-layer perceptron (MLP) type artificial neural networks (ANNs). Nine meteorological parameters and nine air pollutant concentrations were utilized as inputs. The total 578 datasets were divided into three groups: training, crossvalidation, and testing. When all the 18 inputs were used, the best performance was obtained with a network containing one hidden layer with 24 neurons. The transfer function was hyperbolic tangent. The correlation coefficient (R), mean absolute error (MAE), root mean squared error (RMSE), and index of agreement or Willmott's Index (d 2 ) for the testing data were 0.90, 8.78 mg/m 3 , 11.15 mg/m 3 , and 0.95, respectively. Sensitivity analysis has indicated that the persistence information (current day's maximum and average ozone concentrations), NO concentration, average temperature, PM 10 , maximum temperature, sunshine time, wind direction, and solar radiation were the most important input parameters. The values of R, MAE, RMSE, and d 2 did not change considerably for the MLP model using only these nine inputs. The performances of the MLP models were compared with those of regression models (i.e., multiple linear regression and multiple non-linear regression). It has been found that there was no significant difference between the ANN and regression modeling techniques for the forecasting of ozone concentrations in Istanbul. IntroductionOzone (O 3 ) is a secondary pollutant and formed in the lower atmosphere (troposphere) by the complex reactions of nitrogen oxides (NO x ) and volatile organic compounds (VOCs) in the presence of solar radiation. Ozone formation in the troposphere is a rapid photochemical cycle and a non-linear process depending on the concentrations of precursors, meteorological parameters, and the sunlight intensity and spectral distribution [1,2]. Briefly, it involves the photolysis of nitrogen dioxide (NO 2 ) by solar radiation to form nitric oxide (NO) and a ground-state oxygen atom (O).The major reaction forming O 3 in the troposphere includes the reaction of oxygen atom with oxygen molecule (O 2 ). Collision of recently formed O 3 with a third body (i.e., a molecule from the surrounding air) removes the excess energy of ozone and allows it to stabilize.O 3 is removed by the reaction with NO to reform NO 2 ;However, reactions between NO and reactive radical species that are formed by the oxidation of reactive VOCs can also oxidize NO to NO 2 without the involvement of ozone, NO ÀÀÀÀÀÀÀ!

show abstract

Section: Resultsmentioning

confidence: 99%

Artificial Neural Network Prediction of Tropospheric Ozone Concentrations in Istanbul, Turkey

İnal

2010

CLEAN Soil Air Water

View full text Add to dashboard Cite

show abstract

“…On the other hand, few hidden neurons decrease the learning ability of the training sets; however, the network is generalized well. In Elman networks, in order to achieve better performance in the problems with considerable complexity, the hidden (recurrent) layer may possess any number of neurons [35][36][37].…”

Section: Modeling and Structure Of Ernnmentioning

confidence: 99%

“…In addition, to learn spatial patterns, an Elman network is also able to learn time-varying patterns because it can store information for the future. Through an iterative procedure, an Elman network will be able to train and produce these two patterns [34,35].…”

Section: Elman Recurrent Neural Networkmentioning

confidence: 99%

“…This network uses the hyperbolic tangent sigmoid (tansig) and linear (purelin) functions in the hidden and output layers, respectively [34,35]. The activation function which is the tansig function (f) is given by:…”

Section: Elman Recurrent Neural Networkmentioning

confidence: 99%

“…1 and 2) [31]. In NN modeling for the limited sets of data, usually 70 % (2/3) of the experimental data are randomly selected for the training step and the rest of them 30 % (1/3) is used for generalization of the trained network [35][36][37][38].…”

Section: Modeling and Structure Of Ernnmentioning

confidence: 99%

See 2 more Smart Citations

Modeling the Presence of Humic Acid in Ultrafiltration of Xenobiotic Compounds: Elman Recurrent Neural Network

et al. 2011

View full text Add to dashboard Cite

Predicting the rejection of pesticides in ultrafiltration (UF) processes in the presence of common components of dissolved natural organic matter would be taken into consideration as a principle for surface water treatment. This paper presents the application of the Elman Recurrent Neural Network (ERNN) model, which has been trained with previously-obtained experimental data so as to predict the rejection of a class of xenobiotic compounds (nitrophenols (NPs)) dynamically, in the absence and in the presence of humic acid at neutral and acidic conditions. For each trained network, the training function, number of neurons in the hidden and output layers, number of epochs, train and test MSE (mean square error) and MRE (mean relative error) were compared to find the best ERNN. The trained MRE and test MSE for all NPs at the neutral condition was, respectively, less than 1.03 % (4.9 % at acidic condition) and 2.4 % (2.01 % at acidic condition), which showed high network reliability.

show abstract

Artificial Neural Network Modeling of the Environmental Fate and Ecotoxicity of Chemicals

Devillers

2009

Ecotoxicology Modeling

View full text Add to dashboard Cite

Development and comparative analysis of tropospheric ozone prediction models using linear and artificial intelligence-based models in Mexicali, Baja California (Mexico) and Calexico, California (US)

Cited by 69 publications

References 42 publications

Artificial Neural Network Prediction of Tropospheric Ozone Concentrations in Istanbul, Turkey

Artificial Neural Network Prediction of Tropospheric Ozone Concentrations in Istanbul, Turkey

Modeling the Presence of Humic Acid in Ultrafiltration of Xenobiotic Compounds: Elman Recurrent Neural Network

Artificial Neural Network Modeling of the Environmental Fate and Ecotoxicity of Chemicals

Contact Info

Product

Resources

About