A Comparative Study of Deep Learning Models on Tropospheric Ozone Forecasting Using Feature Engineering Approach DOI Creative Commons
Reza Rezaei, Behzad Naderalvojoud, Gülen Güllü

et al.

Atmosphere, Journal Year: 2023, Volume and Issue: 14(2), P. 239 - 239

Published: Jan. 25, 2023

This paper investigates the effect of architectural design deep learning models in combination with a feature engineering approach considering temporal variation features case tropospheric ozone forecasting. Although neural network have shown successful results by extracting automatically from raw data, their performance domain air quality forecasting is influenced different analysis approaches and model architectures. proposes simple but effective time series data that can reveal phases evolution process assist to reflect these variations. We demonstrate addressing when developing architecture improves models. As result, we evaluated our on CNN showed not only does it improve model, also boosts other such as LSTM. The development CNN, LSTM-CNN, CNN-LSTM using proposed improved prediction 3.58%, 1.68%, 3.37%, respectively.

Language: Английский

Air pollutant prediction based on ARIMA-WOA-LSTM model DOI Creative Commons
Jun Luo,

Yaping Gong

Atmospheric Pollution Research, Journal Year: 2023, Volume and Issue: 14(6), P. 101761 - 101761

Published: April 21, 2023

The problem of air pollution has always plagued people's lives, and the management cannot be achieved without prediction assessment concentration various pollutants. In this paper, we propose a method to accurately predict pollutants with aim ensuring efficiency management. proposed ARIMA-WOA-LSTM model uses ARIMA extract linear part data output nonlinear part, while WOA-LSTM is used where whale algorithm find perfect hyperparameters for LSTM, objectives search include number neurons, learning rate batch length. To prove excellence developed in article compared ARIMA-LSTM, CEEMDAN-WOA-LSTM, WOA-LSTM, ARIMA, LSTM. results show that performs better than other models three aspects: pollutant accuracy, stability; combined also much single aspects; excellent five which important error reduction model. high reference

Language: Английский

Citations

66
The Explainable Potential of Coupling Metaheuristics-Optimized-XGBoost and SHAP in Revealing VOCs’ Environmental Fate DOI Creative Commons
Luka Jovanovic, Gordana Jovanović, Mirjana Perišić

et al.

Atmosphere, Journal Year: 2023, Volume and Issue: 14(1), P. 109 - 109

Published: Jan. 4, 2023

In this paper, we explore the computational capabilities of advanced modeling tools to reveal factors that shape observed benzene levels and behavior under different environmental conditions. The research was based on two-year hourly data concentrations inorganic gaseous pollutants, particulate matter, benzene, toluene, m, p-xylenes, total nonmethane hydrocarbons, meteorological parameters obtained from Global Data Assimilation System. order determine model will be capable achieving a superior level performance, eight metaheuristics algorithms were tested for eXtreme Gradient Boosting optimization, while relative SHapley Additive exPlanations values used estimate importance each pollutant parameter prediction concentrations. According results, are mostly shaped by toluene finest aerosol fraction concentrations, in environment governed temperature, volumetric soil moisture content, momentum flux direction, as well hydrocarbons nitrogen oxide. types conditions which provided impact aerosol, temperature dynamics distinguished described.

Language: Английский

Citations

59
VAR-tree model based spatio-temporal characterization and prediction of O3 concentration in China DOI Creative Commons
Hongbin Dai, Guangqiu Huang, Jingjing Wang

et al.

Ecotoxicology and Environmental Safety, Journal Year: 2023, Volume and Issue: 257, P. 114960 - 114960

Published: April 26, 2023

Ozone (O3) pollution in the atmosphere is getting worse many cities. In order to improve accuracy of O3 prediction and obtain spatial distribution concentration over a continuous period time, this paper proposes VAR-XGBoost model based on Vector autoregression (VAR), Kriging method XGBoost (Extreme Gradient Boosting). China used as an example its simulated. paper, data monitoring sites are obtained, then mass established, finnally influencing factors analyzed. This concludes that features highest correlation with PM2.5 lowest SO2. Among measurement factors, wind speed temperature most important affecting pollution, which positively correlated pollution. addition, precipitation negatively 8-hour ozone concentration. performance evaluated ten-fold cross-validation sample, site comparison results XGBoost, CatBoost (categorical boosting), ExtraTrees, GBDT (gradient boosting decision tree), AdaBoost (adaptive RF (random forest), Decision tree, LightGBM (light gradient machine) models conducted. The result shows better than other models. seasonal annual average R2 reaches 0.94 (spring), 0.93 (summer), 0.92 (autumn), (winter), 0.95 (average from 2016 2021). show applicability simulating concentrations performs well. Chinese region obvious feature high east low west, strongly influenced by topographical factors. mean clearly winter summer within season. study can provide scientific basis for prevention control regional China, also new ideas acquisition

Language: Английский

Citations

55
Air-quality prediction based on the ARIMA-CNN-LSTM combination model optimized by dung beetle optimizer DOI Creative Commons

Jiahui Duan,

Yaping Gong,

Jun Luo

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: July 26, 2023

Abstract Air pollution is a serious problem that affects economic development and people’s health, so an efficient accurate air quality prediction model would help to manage the problem. In this paper, we build combined accurately predict AQI based on real data from four cities. First, use ARIMA fit linear part of CNN-LSTM non-linear avoid blinding in hyperparameter setting. Then, dilemma setting, Dung Beetle Optimizer algorithm find hyperparameters model, determine optimal hyperparameters, check accuracy model. Finally, compare proposed with nine other widely used models. The experimental results show paper outperforms comparison models terms root mean square error (RMSE), absolute (MAE) coefficient determination (R 2 ). RMSE values for cities were 7.594, 14.94, 7.841 5.496; MAE 5.285, 10.839, 5.12 3.77; R 0.989, 0.962, 0.953 respectively.

Language: Английский

Citations

40
Multi-objective optimal dispatch strategy for power systems with Spatio-temporal distribution of air pollutants DOI
Hongbin Dai, Guangqiu Huang, Huibin Zeng

et al.

Sustainable Cities and Society, Journal Year: 2023, Volume and Issue: 98, P. 104801 - 104801

Published: July 17, 2023

Language: Английский

Citations

39
Heavy metals contamination status and health risk assessment of indoor and outdoor dust in Ahvaz and Zabol cities, Iran DOI

Seyed Reza Asvad,

Abbas Esmaili‐Sari,

Nader Bahramifar

et al.

Atmospheric Pollution Research, Journal Year: 2023, Volume and Issue: 14(4), P. 101727 - 101727

Published: March 22, 2023

Language: Английский

Citations

28
Predictive modeling of air quality in the Tehran megacity via deep learning techniques DOI Creative Commons

Abdullah Kaviani Rad,

Mohammad Javad Nematollahi, Abbas Pak

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 8, 2025

Air pollution is a significant challenge in metropolitan areas, where increasing amounts of air pollutants threaten public health and environmental safety. The present study aims to forecast the concentrations various pollutants, including CO, O3, NO2, SO2, PM10, PM2.5, from 2013 2023 Tehran megacity, Iran, via deep learning (DL) models evaluate their effectiveness over conventional machine (ML) methods. Key driving variables, temperature, relative humidity, dew point, wind speed, pressure, were considered. R-squared (R2), root-mean-square error (RMSE), mean absolute (MAE), mean-square (MSE) used assess compare models. This research demonstrated that DL typically outperform ML forecasting pollution. Gated recurrent units (GRUs), fully connected neural networks (FCNNs), convolutional (CNNs) recorded R2 MSE values 0.5971 42.11 for 0.7873 171.40 0.4954 25.17 respectively. Consequently, FCNN GRU presented remarkable performance predicting NO2 (R2 = 0.6476 75.16), PM10 0.8712 45.11), PM2.5 0.9276 58.12) concentrations. In terms operational model exhibited most efficiency, with minimum maximum runtime 13 28 s, feature importance analysis suggested O3 SO2 are affected by Thus, temperature humidity primary factors affecting variability pollutant conclusions confirm achieve accuracy serve as essential instruments managing pollution, providing practical insights decision-makers adopt efficient quality control strategies.

Language: Английский

Citations

1
Methods for Urban Air Pollution Measurement and Forecasting: Challenges, Opportunities, and Solutions DOI Creative Commons

Elena Mitreska Jovanovska,

Victoria Batz, Petre Lameski

et al.

Atmosphere, Journal Year: 2023, Volume and Issue: 14(9), P. 1441 - 1441

Published: Sept. 15, 2023

In today’s urban environments, accurately measuring and forecasting air pollution is crucial for combating the effects of pollution. Machine learning (ML) now a go-to method making detailed predictions about levels in cities. this study, we dive into how settings measured predicted. Using PRISMA methodology, chose relevant studies from well-known databases such as PubMed, Springer, IEEE, MDPI, Elsevier. We then looked closely at these papers to see they use ML algorithms, models, statistical approaches measure predict common pollutants. After review, narrowed our selection 30 that fit research goals best. share findings through thorough comparison papers, shedding light on most frequently predicted pollutants, models chosen predictions, which ones work best determining city quality. also take look Skopje, North Macedonia’s capital, an example still working its prediction systems. conclusion, there are solid methods out measurement prediction. Technological hurdles no longer major obstacle, meaning decision-makers have ready-to-use solutions help tackle issue

Language: Английский

Citations

19
Improving PM2.5 prediction in New Delhi using a hybrid extreme learning machine coupled with snake optimization algorithm DOI Creative Commons
Adil Masood, Mohammed Majeed Hameed, Aman Srivastava

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: Nov. 29, 2023

Fine particulate matter (PM2.5) is a significant air pollutant that drives the most chronic health problems and premature mortality in big metropolitans such as Delhi. In context, accurate prediction of PM2.5 concentration critical for raising public awareness, allowing sensitive populations to plan ahead, providing governments with information alerts. This study applies novel hybridization extreme learning machine (ELM) snake optimization algorithm called ELM-SO model forecast concentrations. The has been developed on quality inputs meteorological parameters. Furthermore, hybrid compared individual models, Support Vector Regression (SVR), Random Forest (RF), Extreme Learning Machines (ELM), Gradient Boosting Regressor (GBR), XGBoost, deep known Long Short-Term Memory networks (LSTM), forecasting results suggested exhibited highest level predictive performance among five testing value squared correlation coefficient (R2) 0.928, root mean square error 30.325 µg/m3. study's findings suggest technique valuable tool accurately concentrations could help advance field forecasting. By developing state-of-the-art pollution models incorporate ELM-SO, it may be possible understand better anticipate effects human environment.

Language: Английский

Citations

19
A novel approach for forecasting PM2.5 pollution in Delhi using CATALYST DOI
Abhishek Verma, Virender Ranga, Dinesh Kumar Vishwakarma

et al.

Environmental Monitoring and Assessment, Journal Year: 2023, Volume and Issue: 195(12)

Published: Nov. 11, 2023

Language: Английский

Citations

17