A review of machine learning for modeling air quality: Overlooked but important issues

Atmospheric Research, Год журнала: 2024, Номер 300, С. 107261 - 107261

Опубликована: Янв. 21, 2024

Язык: Английский

---

Mapping the seamless hourly surface visibility in China: a real-time retrieval framework using a machine-learning-based stacked ensemble model

npj Climate and Atmospheric Science, Год журнала: 2024, Номер 7(1)

Опубликована: Март 13, 2024

Abstract Surface visibility (SV), a key indicator of atmospheric transparency, is used widely in the fields environmental monitoring, transportation, and aviation. However, sparse distribution limited number SV monitoring sites make it difficult to fulfill urgent need for spatiotemporally seamless fine-scale monitoring. Here, we developed operational real-time retrieval (RT-SVR) framework China that incorporates information from multiple data sources, including Chinese Land Data Assimilation System meteorological data, situ observations, other ancillary data. Seamless hourly with 6.25-km spatial resolution are available real time via RT-SVR model, which was built using two-layer stacked ensemble approach combines machine learning algorithms deep module. Sample-based cross-validation model on approximately 41.3 million pairs revealed strong robustness high accuracy, Pearson correlation coefficient (R) value 0.95 root mean square error (RMSE) 3.17 km. An additional hindcast-validation experiment, performed continuous observations obtained over one year (approximately 20.8 pairs), demonstrated powerful generalization capabilities albeit slight degradation performance ( R = 0.85, RMSE 5.28 km). The update capability enable tracking generation, development, dissipation various low-SV events (e.g., fog, haze, dust storms) China. might also prove useful quantitative aerosol-related parameters PM 2.5 , 10 aerosol optical depth).

Язык: Английский

---

Provincial equity and enhanced health are key drivers for China's 2060 carbon neutrality

Journal of Cleaner Production, Год журнала: 2024, Номер 473, С. 143531 - 143531

Опубликована: Авг. 30, 2024

Язык: Английский

---

Long-Term (2015–2024) Daily PM2.5 Estimation in China by Using XGBoost Combining Empirical Orthogonal Function Decomposition

Remote Sensing, Год журнала: 2025, Номер 17(9), С. 1632 - 1632

Опубликована: Май 4, 2025

Fine particulate matter (PM2.5) has garnered significant scientific and public health concern owing to its capacity for deep penetration into the human respiratory system, presenting risks. Despite implementation of strict environmental policies in China over past decade reduce PM2.5 levels, long-term concerns remain a serious issue. Our study aims provide high-quality, seamless daily dataset covering years 2015 2024. A two-step estimation model is established based on machine learning algorithm spatio-temporal decomposition method. First, we utilize XGBoost (EXtreme Gradient Boosting) address gaps MAIAC (Multi-Angle Implementation Atmospheric Correction) AOD (Aerosol Optical Depth), with R2/RMSE (coefficient determination/Root Mean Square Error) 0.67/0.2678 compared AERONET Robotic Network) AOD. Then, novel approach by integrating EOF (Empirical Orthogonal Function) introduced estimation. The integration allows incorporation entire meteorological field information model, significantly enhancing accuracy: spatial CV (cross-validation)-R2 improved from 0.8340 0.8935, CV-RMSE reduced 13.8177 11.0668. Leveraging newly produced dataset, analyze variations across decomposition, particularly noting that levels eastern anthropogenic intensive regions continuously declined 2020, fluctuated steadily during 2020–2024. This research underscores critical need sustained effective air quality management strategies China.

Язык: Английский

---

Coordinated estimates of 4-day 500 m global land surface energy balance components

Remote Sensing of Environment, Год журнала: 2025, Номер 326, С. 114795 - 114795

Опубликована: Май 7, 2025

Язык: Английский

---

A novel framework for quantitative attribution of particulate matter pollution mitigation to natural and socioeconomic drivers

The Science of The Total Environment, Год журнала: 2024, Номер 926, С. 171910 - 171910

Опубликована: Март 24, 2024

Язык: Английский

---

Data Integration for ML-CNPM_2.5: A Public Sample Dataset Based on Machine Learning Models and Remote Sensing Technology Applied for Estimating Ground-level PM_2.5 in China

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2024, Номер 62, С. 1 - 15

Опубликована: Янв. 1, 2024

Ambient fine particulate matter (PM _2.5 ) has significant adverse effects on human health, thereby urgent hunger for accurate monitoring of ground-level PM , especially its space distribution. Since satellites can observe the Earth a large spatial scale, remote sensing technology be applied to estimate concentrations at national level. Based it and machine learning (ML) methods, numerous studies mapped high-accuracy, wholesale continuous . However, different models data in these made their results incomparable, more samples were needed provided. Here, large-column long-term sample dataset (ML-CNPM ML-based was constructed with 5,076,608 records 24 features from 2014 2023 China. Multiple approaches used guarantee quantity quality dataset. Due comprehensiveness objectivity, ML-CNPM train validate models, further improving accuracy estimating. Using eight basic also as baseline judging other derivative models. These daily full-coverage most performed well, 10-fold cross-validation RMSE 16.94-11.21μg/m ³ R xmlns:xlink="http://www.w3.org/1999/xlink">2 0.71-0.89, which is consistent previous effectively capture trends period suffered high pollution. Overall, our construct, validate, compare various estimation, helping develop new algorithms higher robustness.

Язык: Английский

---

Retrieving ground-level PM_2.5 concentrations in China (2013–2021) with a numerical-model-informed testbed to mitigate sample-imbalance-induced biases

Earth system science data, Год журнала: 2024, Номер 16(8), С. 3781 - 3793

Опубликована: Авг. 27, 2024

Abstract. Ground-level PM2.5 data derived from satellites with machine learning are crucial for health and climate assessments. However, uncertainties persist due to the absence of spatially covered observations. To address this, we propose a novel testbed using nontraditional numerical simulations evaluate estimation across entire spatial domain. The emulates general machine-learning approach by training model grids corresponding ground monitoring sites subsequently testing its predictive accuracy other locations. Our enables comprehensive evaluation various methods' performance in estimating domain first time. Unexpected results shown application China, larger absolute biases found densely populated regions abundant observations all benchmark models higher baseline concentration, though relative error (approximately 20 %) is smaller compared that rural areas (over 50 %). imbalance samples, mostly urban high emissions, main reason, leading significant overestimation lack monitors downwind where transported varying vertical profiles. proposed also provides an efficient strategy optimizing structure or samples enhance satellite-retrieval performance. Integration spatiotemporal features, especially conventional neural network (CNN)-based deep-learning approaches like residual (ResNet) model, has successfully mitigated (by 5–30 µg m−3) exposure 3 million people ⋅ area over 9 years (2013–2021) traditional approach. Furthermore, incorporation 600 strategically positioned identified through essential achieving more balanced distribution thereby ensuring precise facilitating assessment associated impacts China. In addition presenting retrieved surface concentrations China 2013 2021, this study dataset physical modeling which can serve data-driven methodologies, such as learning, community (Li et al., 2024a; https://doi.org/10.5281/zenodo.11122294).

Язык: Английский

---

Land Cover Disaggregated Fire Occurrence and Particulate Matter2.5 Relationship in the Mekong Region: A Comprehensive Study

ISPRS International Journal of Geo-Information, Год журнала: 2024, Номер 13(6), С. 206 - 206

Опубликована: Июнь 17, 2024

Air pollution has become an increasing concern in the Mekong region due to seasonal vegetative burning triggered by related anthropogenic activities and climate change. While assumption of a correlation between agriculture air is common postulation, little evidence exists on association fire incidents concentrations. The current study explores relationship satellite-derived occurrence, land surface characteristics, particulate matter 2.5 (PM2.5) concentrations for five Lower countries, namely Cambodia, Laos, Myanmar, Thailand, Vietnam, effort gain new insights into distributions quality. Publicly available daily active hotspots from VIIRS satellite instrument, annual cover products MODIS satellite, mean monthly ground-level PM2.5 estimates V5.GL.04 database were analyzed two relational assessments; first, distribution counts radiative power (FRP) temporally spatially secondly, correlations counts, cumulative FRP per country type. results suggest statistically significant positive FRP, each country, which differ based cover. strongest incidences was found case Myanmar. For all countries combined, fires detected forests displayed highest with estimates. This demonstrates use product provides important temporal spatial as baseline information prevention mitigation strategies region.

Язык: Английский

---

The Reconstruction of FY-4A and FY-4B Cloudless Top-of-Atmosphere Radiation and Full-Coverage Particulate Matter Products Reveals the Influence of Meteorological Factors in Pollution Events

Remote Sensing, Год журнала: 2024, Номер 16(18), С. 3363 - 3363

Опубликована: Сен. 10, 2024

By utilizing top-of-atmosphere radiation (TOAR) data from China’s new generation of geostationary satellites (FY-4A and FY-4B) along with interpretable machine learning models, near-surface particulate matter concentrations in China were estimated, achieving hourly temporal resolution, 4 km spatial 100% coverage. First, the cloudless TOAR matched modeled solar products ERA5 dataset to construct estimate a fully covered under assumed clear-sky conditions, which increased coverage 20–30% 100%. Subsequently, this was applied matter. The analysis demonstrated that (R2 = 0.83) performed better than original 0.76). Additionally, using feature importance scores SHAP values, impact meteorological factors air mass trajectories on increase PM10 PM2.5 during dust events investigated. haze indicated main driving changes included pressure, temperature, boundary layer height. concentration obtained exhibit high spatiotemporal resolution. Combined data-driven learning, they can effectively reveal influencing China.

Язык: Английский

---