A Visibility-Based Historical PM2.5 Estimation for Four Decades (1981–2022) Using Machine Learning in Thailand: Trends, Meteorological Normalization, and Influencing Factors Using SHAP Analysis

Aerosol and Air Quality Research, Год журнала: 2025, Номер 25(1-4)

Опубликована: Март 27, 2025

Abstract Introduction PM 2.5 pollution is a significant environmental and health concern in Thailand, with levels intensifying during the dry season. However, lack of long-term PM2.5 data limits understanding historical trends meteorological influences. Objective This study aims to reconstruct from 1981 2022 analyze influence various contributing factors across six key provinces Thailand: Chiang Mai (CM), Lampang (LP), Khon Kaen (KK), Bangkok (BK), Chonburi (CB), Songkhla (SK). Methods A Light Gradient Boosting Machine (LightGBM) model was developed using aerosol-related variables Thai Meteorological Department MERRA-2. The trained on spanning 2012–2022, depending availability for each province. Model performance evaluated diurnal, monthly, annual scales then used reconstruction data. SHAP analysis determine important predictor affecting prediction. Results LightGBM accurately predicted all provinces, showing better daily prediction than hourly accuracy higher clean hours haze hours. Good agreement between observed found different time (diurnal, annually). CM shows non-significant trend, limiting insights into effects, while LP exhibits decreases PM2.5_emis, indicating positive weather impacts air quality. In contrast, regions like KK, BK, CB display worsening influences, or increasing despite declines _emis. SK, removing effects reveals decreasing underscoring critical role meteorology. identified visibility, gridded , specific humidity at 2 m as common over along additional that were not consistent provinces. Conclusion effectively reconstructs provides insight influences Based findings study, some policy implications have also been provided. Graphical abstract

Язык: Английский

---

Predicting Ozone Concentrations in Ecologically Sensitive Coastal Zones Through Structure Mining and Machine Learning: A Case Study of Chongming Island, China

Atmosphere, Год журнала: 2025, Номер 16(4), С. 457 - 457

Опубликована: Апрель 15, 2025

Elevated O3 concentrations pose a significant threat to human health and ecosystems, but little research has been performed on coastal wetlands near large cities. This study focuses investigating the key factors affecting formation in ecologically sensitive Dongtan Wetland (Chongming District, Shanghai, China) area. By comparing performance of concentration prediction multiple machine learning models, this found that random forest model achieved highest accuracy (R2 = 0.9, RMSE 11.5). Feature importance structure mining showed peroxyacetyl nitrate (PAN), nitrogen oxides (NOx), temperature, wind direction, relative humidity were main drivers formation. Specifically, PAN exceeding 0.1 ppb temperatures above 3 °C have impact levels, especially spring, summer, autumn. Trajectory analysis westward urban pollution emissions transported from ocean highlights need for targeted emission control strategies, precursors generated by ships NOx industries, providing important insights improving air quality areas.

Язык: Английский

---

Machine learning-based quantification and separation of emissions and meteorological effects on PM2.5 in Greater Bangkok

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 28, 2025

Язык: Английский

---

Chemical Composition and Source of PM2.5 during Winter Heating Period in Guanzhong Basin

Atmosphere, Год журнала: 2023, Номер 14(11), С. 1640 - 1640

Опубликована: Окт. 31, 2023

An intensive field campaign was carried out from December 2022 to March 2023 at six different sites across five major cities (Xi’an, Baoji, Xianyang, Weinan, and Hancheng) in the Guanzhong Basin, China, covering most of heating period there, which is characterized by high PM2.5 pollution levels. During campaign, mean concentrations these exceeded 24 h standard (75 μg m−3), except site Hancheng, with 57.8 ± 32.3 m−3. The source apportionment varied significantly sites, vehicle exhaust being dominant urban located Xi’an coal combustion suburban comparable contribution industrial emissions Xianyang Weinan. Compared clean condition, secondary inorganic sources (SIs) were largely enhanced during heavy periods, while biomass burning (BB) dust decreased all sites. Combined an analysis meteorological parameters, study further found that higher contributions SIs generally associated relative humidity (RH). In addition, related lower wind speeds, could be explained stagnant condition favoring accumulation local as well formation pollutants. contrast, (e.g., Xianyang), more strong influence slightly speeds.

Язык: Английский

---

Estimating visibility and understanding factors influencing its variations at Bangkok airport using machine learning and a game theory–based approach

Environmental Science and Pollution Research, Год журнала: 2024, Номер unknown

Опубликована: Авг. 5, 2024

Язык: Английский

---

Joint estimation of PM2.5 and O3 concentrations using a hybrid model in Beijing-Tianjin-Hebei, China

Atmospheric Pollution Research, Год журнала: 2024, Номер 15(8), С. 102174 - 102174

Опубликована: Май 7, 2024

Язык: Английский

---

Mining of dynamic traffic-meteorology-atmospheric pollutant association rules based on eclat method

Atmospheric Pollution Research, Год журнала: 2024, Номер unknown, С. 102305 - 102305

Опубликована: Сен. 1, 2024

Язык: Английский

---

Accurate and efficient prediction of atmospheric PM1, PM2.5, PM10, and O3 concentrations using a customized software package based on a machine-learning algorithm

Chemosphere, Год журнала: 2024, Номер 368, С. 143752 - 143752

Опубликована: Ноя. 1, 2024

Язык: Английский

---

Industrial Heat Source-Related PM2.5 Concentration Estimates and Analysis Using New Three-Stage Model in the Beijing–Tianjin–Hebei Region

Atmosphere, Год журнала: 2024, Номер 15(1), С. 131 - 131

Опубликована: Янв. 20, 2024

The prevalent high-energy, high-pollution and high-emission economic model has led to significant air pollution challenges in recent years. industrial sector the Beijing–Tianjin–Hebei (BTH) region is a notable source of atmospheric pollutants, with heat sources (IHSs) being primary contributors this pollution. Effectively managing emissions from these pivotal for achieving control goals region. A new three-stage using multi-source long-term data was proposed estimate atmospheric, delicate particulate matter (PM2.5) concentrations caused by IHS. In first stage, region-growing algorithm used identify IHS radiation areas. second third stages, based on seasonal trend decomposition procedure Loess (STL), multiple linear regression, U-convLSTM models, IHS-related PM2.5 meteorological anthropogenic conditions were removed 2012 2021. Finally, study analyzed spatial temporal variations BTH findings reveal that areas higher than background areas, approximately 33.16% attributable activities. decreasing observed. Seasonal analyses indicated industrially dense southern region, particularly during autumn winter. Moreover, case Handan’s She County demonstrated dynamic fluctuations concentrations, reductions periods inactivity. Our results aligned closely previous studies actual operations, showing strong positive correlations related indices. This study’s outcomes are theoretically practically understanding addressing regional quality IHSs, contributing positively environmental improvement sustainable development.

Язык: Английский

---

Estimating visibility and understanding factors influencing its variations at Bangkok airport using machine learning and a game theory-based approach

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Апрель 2, 2024

In this study, a range of machine learning (ML) models including random forest, adaptive boosting, gradient extreme light cat and stacked ensemble model, were employed to predict visibility at Bangkok airport. Furthermore, the impact influential factors was examined using Shapley method, an interpretable ML technique inspired by game theory-based approach. Air pollutant data from seven Pollution Control Department monitoring stations, visibility, meteorological Thai Meteorological Department's Weather station Airport, ERA5_LAND, ERA5 datasets, time-related dummy variables considered. Daytime ((here, 8–17 local time) screened for rainfall, developed prediction during dry season (November – April). The boosting model is identified as most effective individual with superior performance in three out four evaluation metrics (i.e., highest ρ, zero MB, second lowest ME, RMSE). However, SEM outperformed all both hourly daily time scales. seasonal mean standard deviation normalized are lower than those original indicating more influence meteorology emission reduction on improvement. analysis RH, PM_2.5, PM₁₀, day year, O₃ five important variables. At low relative humidity (RH), there no notable visibility. Nevertheless, beyond threshold, negative correlation between RH An inverse PM_2.5 PM₁₀ identified. Visibility negatively correlated moderate concentrations, diminishing very high concentrations. year Julian day) (JD) exhibits initial later positive association suggesting periodic effect. dependence values equal step size method understand effects, suggest effect hygroscopic growth aerosol Findings research feasibility employing techniques predicting comprehending influencing its fluctuations. Based above findings, certain policy–related implications, future work have been suggested.

Язык: Английский

---