Toward an Operational Machine-Learning-Based Model for Deriving the Real-Time Gapless Diurnal Cycle of Ozone Pollution in China with CLDAS Data DOI

Nanxuan Shang,

Ke Gui, Fugang Li

et al.

Environmental Science & Technology Letters, Journal Year: 2024, Volume and Issue: 11(6), P. 553 - 559

Published: April 29, 2024

An operational real-time surface ozone (O3) retrieval (RT-SOR) model was developed that can provide a gapless diurnal cycle of O3 retrievals with spatial resolution 6.25 km by integrating Chinese Land Data Assimilation System (CLDAS) data and multisource auxiliary information. The robustly captures the hourly variability, sample-based (station-based) cross-validation R2 0.88 (0.85) RMSE 14.3 μg/m3 (16.1 μg/m3). additional hindcast-validation experiment demonstrated generalization ability is robust (R2 = 0.75; 21.9 Compared previous studies, performs comparably or even better at daily scale fills gaps in terms missing within 24-hour cycle. More importantly, underpinned RT availability CLDAS data, concentration be updated RT, which expected to advance our understanding pollution China.

Language: Английский

Simulation and prediction of PM2.5 concentrations and analysis of driving factors using interpretable tree-based models in Shanghai, China DOI
Wei Qing, Yongqi Chen, Huijin Zhang

et al.

Environmental Research, Journal Year: 2025, Volume and Issue: unknown, P. 121003 - 121003

Published: Feb. 1, 2025

Language: Английский

Citations

2
Estimation of near-surface ozone concentration and analysis of main weather situation in China based on machine learning model and Himawari-8 TOAR data DOI
Bin Chen,

Yixuan Wang,

Jianping Huang

et al.

The Science of The Total Environment, Journal Year: 2022, Volume and Issue: 864, P. 160928 - 160928

Published: Dec. 17, 2022

Language: Английский

Citations

41
Synergistic observation of FY-4A&4B to estimate CO concentration in China: combining interpretable machine learning to reveal the influencing mechanisms of CO variations DOI Creative Commons
Bin Chen, Jiashun Hu,

Yixuan Wang

et al.

npj Climate and Atmospheric Science, Journal Year: 2024, Volume and Issue: 7(1)

Published: Jan. 6, 2024

Abstract Accurately estimating the concentration of carbon monoxide (CO) with high spatiotemporal resolution is crucial for assessing its meteorological-environmental-health impacts. Although machine learning models have predictive ability in environmental research, there are relatively few explanations model outputs. Utilizing top-of-atmosphere radiation data China’s new generation geostationary satellites (FY-4A and FY-4B) interpretable models, 24-hour near-surface CO concentrations China was conducted (resolution: 1 hour, 0.04°). The improved by 6.6% when using all-sky dataset (cloud-contained model, R 2 = 0.759) compared to clear-sky (cloud-removed model). interpretability analysis estimation used two methods, namely ante-hoc (model feature importance) post-hoc (SHapley Additive exPlanations). importance daytime meteorological factors increased 51% nighttime. Combining partial dependency plots, impact key on elucidated gain a deeper understanding variations CO.

Language: Английский

Citations

11
Estimation of Atmospheric PM10 Concentration in China Using an Interpretable Deep Learning Model and Top‐of‐the‐Atmosphere Reflectance Data From China’s New Generation Geostationary Meteorological Satellite, FY‐4A DOI Creative Commons
Bin Chen, Zhihao Song, Jianping Huang

et al.

Journal of Geophysical Research Atmospheres, Journal Year: 2022, Volume and Issue: 127(9)

Published: April 18, 2022

Abstract The rapid urbanization in China and the long‐range transport dust (LRTD) from arid semi‐arid areas has resulted an increase of PM 10 concentration. In this study, interpretable deep learning model [deep forest (DF)] with FY‐4A top‐of‐the‐atmosphere reflectance (TOAR) data were used to obtain hourly China. optimal average R 2 10‐fold cross validation can achieve 0.85 (13:00 Beijing time); (RMSE, μg/m³) daily, monthly, annual averages 0.82 (24.16), 0.97 (6.53), 0.99 (2.30), respectively. Using TOAR data, DF performed better than other machine models. feature importance TOAR‐PM showed that meteorological elements both contributed significantly model. spring, northern was greater southern China, which may be related LRTD. Excluding weather periods, high values mainly cities their suburbs, where correlated human activities. During a process, LRTD increased by 80.4%. mixture haze 130.2% led 73.7%. sources (from Taklimakan Desert China) transmission paths these two processes similar. contribution intensity conditions. results local pollution important periods.

Language: Английский

Citations

29
Spatiotemporally continuous estimates of daily 1-km PM2.5 concentrations and their long-term exposure in China from 2000 to 2020 DOI Open Access
Qingqing He, Tong Ye, Weihang Wang

et al.

Journal of Environmental Management, Journal Year: 2023, Volume and Issue: 342, P. 118145 - 118145

Published: May 19, 2023

Language: Английский

Citations

22
Predicting PM2.5 Concentrations Across USA Using Machine Learning DOI Creative Commons

P. Vignesh,

Jonathan H. Jiang, P. Kishore

et al.

Earth and Space Science, Journal Year: 2023, Volume and Issue: 10(10)

Published: Oct. 1, 2023

Abstract Economic growth, air pollution, and forest fires in some states the United States have increased concentration of particulate matter with a diameter less than or equal to 2.5 μm (PM ). Although previous studies tried observe PM both spatially temporally using aerosol remote sensing geostatistical estimation, they were limited accuracy by coarse resolution. In this paper, performance machine learning models on predicting is assessed linear regression (LR), decision tree (DT), gradient boosting (GBR), AdaBoost (ABR), XGBoost (XGB), k‐nearest neighbors (K‐NN), long short‐term memory (LSTM), random (RF), support vector (SVM) station data from 2017 2021. To compare all nine models, coefficient determination ( R 2 ), root mean square error (RMSE), Nash‐Sutcliffe efficiency (NSE), ratio (RSR), percent bias (PBIAS) evaluated. Among RF (100 trees max depth 20) (SVR; nonlinear kernel, degree 3 polynomial) best for concentrations. Additionally, comparison metrics displayed that had better predictive behavior western eastern States.

Language: Английский

Citations

20
Identification of particle distribution pattern in vertical profile via unmanned aerial vehicles observation DOI

Zhiheng Chen,

Bowen Li, Bai Li

et al.

Environmental Pollution, Journal Year: 2024, Volume and Issue: 348, P. 123893 - 123893

Published: March 29, 2024

Language: Английский

Citations

7
Combining Himawari-8 AOD and deep forest model to obtain city-level distribution of PM2.5 in China DOI
Zhihao Song, Bin Chen, Jianping Huang

et al.

Environmental Pollution, Journal Year: 2022, Volume and Issue: 297, P. 118826 - 118826

Published: Jan. 8, 2022

Language: Английский

Citations

28
Interpretable Machine Learning Approaches for Forecasting and Predicting Air Pollution: A Systematic Review DOI Creative Commons
Anass Houdou, Imad El Badisy, Kenza Khomsi

et al.

Aerosol and Air Quality Research, Journal Year: 2023, Volume and Issue: 24(1), P. 230151 - 230151

Published: Nov. 30, 2023

Many studies use machine learning to predict atmospheric pollutant levels, prioritizing accuracy over interpretability. This systematic review will focus on reviewing that have utilized interpretable models enhance interpretability while maintaining high for air pollution prediction. The search terms "air pollution," "machine learning," and "interpretability" were used identify relevant published between 2011 2023 from PubMed, Scopus, Web of Science, Science Direct, JuSER. included assessed quality based an ecological checklist maximizing reproducibility niche models. Among the 5,396 identified studies, 480 focused prediction, with 56 providing model interpretations. 20 methods identified: 8 model-agnostic methods, 4 model-specific hybrid Shapley additive explanations was most commonly method (46.4%), followed by partial dependence plots (17.4%), both which are methods. These important features, enhancing researchers' understanding making outcomes more accessible non-experts. can prediction prevention adverse weather events pollution, benefiting public health.

Language: Английский

Citations

16
A Two-Stage Machine Learning Algorithm for Retrieving Multiple Aerosol Properties Over Land: Development and Validation DOI

Mengdan Cao,

Ming Zhang, Xin Su

et al.

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2023, Volume and Issue: 61, P. 1 - 17

Published: Jan. 1, 2023

Satellite-based aerosol optical property retrieval over land, especially size-related parameters, is challenging. This study proposed a novel two-stage machine learning (ML) algorithm for retrieving depth (AOD), Ångström exponent (AE), fine mode fraction (FMF), and AOD (FAOD)) land using MODIS observed reflectance. The new ML consists of three steps: (1) first, all samples extracted from AERONET measurements were used to train the model, (2) then, reduce extreme estimation bias divided low-value high-value models, respectively, (3) finally, models integrated into final based on weight interpolation. Independent site network validation results show that has Pearson correlation coefficient (R) 0.894 (0.638, 0.661, 0.865) root mean square error (RMSE) 0.146 (0.258, 0.245, 0.153) (AE, FMF, FAOD) retrieval, which significantly outperforms metrics operational products, with RMSE 0.130-0.156 (0.536-0.569, 0.313, 0.191). inter-comparison products shows spatial patterns AOD, AE, FAOD are in good agreement those POLDER products. These illustrate performance transferability indicate ability methods be applied multispectral instruments (such as MODIS) retrieve multiple properties.

Language: Английский

Citations

15