DATA ABSTRACT AIM OF STUDY

Investigates spatio-temporal variation

Satellite Data: MODIS AOD and FMF products and decadal trends in atmospheric Analyze spatio-temporal and decadal
(Terra and Aqua). aerosols over Peninsular Malaysia. trends of aerosols.
Focuses on Aerosol Optical Depth (AOD)
Ground Data: AERONET AOD, PM10 Identify aerosol sources, types, and their
trends and their impact on solar radiation.
concentrations, solar radiation recordings. relationship with solar radiation.
Explores aerosol types, sources, and
HYSPLIT Model: Used for back-trajectory seasonal transport pathways using Support understanding of regional air
analysis of aerosol transport pathways. satellite and ground-based data. pollution and its climate impact.

SIGNIFIICANCE OBJECTVIES
Provides detailed insights into aerosol dynamics Study spatial and temporal evolution of AOD
and their regional impact. using MODIS data.
Highlights the role of aerosols in air quality and Compare MODIS data with AERONET ground-
solar radiation attenuation. based observations.
Classify aerosol types and identify their
Supports policy-making for environmental
sources.
management and pollution control in Southeast Investigating aerosol properties Quantify aerosol effects on solar radiation
Asia. attenuation.
in Peninsular Malaysia
via the synergy of satellite
RECOMMENDATION PROBLEM STATEMENT
remote sensing and Limited aerosol studies in Malaysia despite
Implement policies to reduce emissions in urban and industrial
areas. ground-based measurements significant pollution from local and transboundary
Enhance ground-based monitoring networks, especially sources.
AERONET stations. 2013 Challenges in remote sensing of aerosols in
Conduct further studies on aerosol-climate interactions using Southeast Asia due to variability and cloud cover.
improved datasets. Kasturi Devi Kanniah a, Hui Qi Lim a , Lack of comprehensive long-term datasets for
Promote low-carbon strategies to mitigate air pollution.
Dimitris G. Kaskaoutis b , Arthur P. Cracknell aerosol impact analysis.

Kanniah, K. D., Lim, H. Q., Kaskaoutis, D. G., & Cracknell, A. P.

(2014). Investigating aerosol properties in Peninsular Malaysia
via the synergy of satellite remote sensing and ground-based
FINDINGS measurements. Atmospheric Research, 138, 223–239. METHODS
https://doi.org/10.1016/j.atmosres.2013.11.018
Data Sources: MODIS satellite data (Terra and Aqua), AERONET,
Aerosols have strong seasonal and regional variability in Peninsular PM10 measurements, and solar radiation records.
Malaysia. Trend Analysis: Decadal trend analysis using statistical regression.
Dry season aerosols originate from biomass burning in Indonesia. Source Tracking: HYSPLIT model for seasonal back-trajectory
Urban and industrial aerosols are dominant in densely populated analysis.
Scatter Analysis: AOD vs. FMF scatterplots for aerosol type
areas.
identification.
Significant reduction in solar radiation linked to high aerosol loading.
Solar Radiation Correlation: Examined AOD effects on solar
radiation.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Terra MODIS AOD values were higher in the morning High aerosol loading during dry seasons due to biomass
than Aqua MODIS in the afternoon. burning in Indonesia.
Sparse aerosol data in Malaysia due to limited ground-based observations.
Seasonal variation in aerosol types: biomass burning Wet seasons show lower AOD due to rain and humid
maritime air masses.
Challenges in retrieving satellite data in cloudy regions. aerosols dominated during the dry season; coarse-
Terra AOD showed a neutral-to-declining trend; Aqua AOD
mode particles (maritime influence) during the wet
exhibited a slight increase.
Sensor-specific biases (e.g., Terra MODIS calibration drift). season. PM10 concentrations increased over many locations.
Correlation analysis showed aerosols significantly Aerosols reduce solar radiation by ~0.21 MJ/m² on average.
attenuate solar radiation.
 DATA ABSTRACT AIM OF STUDY
Ground-Based Data: Examines the estimation of PM10 levels in Malaysia using
PM10 concentrations from 45 air quality satellite-based aerosol optical depth (AOD) data. Develop and validate empirical models to
Incorporates meteorological parameters (temperature,
monitoring stations.
relative humidity, and atmospheric stability) in models.
estimate PM10 concentrations using AOD
Satellite Data: Compares two modeling techniques: Multiple Linear and meteorological variables.
MODIS AOD at 550 nm, surface Regression (MLR) and Artificial Neural Network (ANN). Explore the potential for satellite data to
temperature, relative humidity, and Demonstrates ANN's superior accuracy in estimating PM10
levels. enhance air quality monitoring in Malaysia.
atmospheric stability index (k-index).

SIGNIFIICANCE OBJECTVIES
Demonstrates the potential of remote sensing and Use AOD and meteorological data to estimate
machine learning for air quality management in PM10 levels over Malaysia.
developing regions. Test the effectiveness of statistical (MLR) and
Provides a scalable methodology for PM10 machine learning (ANN) methods.
estimation in areas with limited ground-based
Estimating Particulate Matter using
Evaluate the spatial and seasonal distribution of
monitoring. satellite based aerosol PM10 across Malaysia.
optical depth and meteorological variables
in Malaysia
RECOMMENDATION 2017 PROBLEM STATEMENT
Expand satellite monitoring and ground-based validation Limited ground-based PM10 monitoring stations in
networks in Malaysia. Nurul Amalin Fatihah Kamarul Zamana , Kasturi Devi Malaysia hinder regional-scale pollution analysis.
Adopt ANN models for regional air quality assessments. Kanniaha,c , Existing models using AOD alone show limited accuracy in
Strengthen policies targeting biomass burning and urban- PM10 prediction.
Dimitris G. Kaskaoutisb
industrial emissions. Need for integrating meteorological factors and advanced
modeling techniques for improved estimations.
Kamarul Zaman, N. A. F., Kanniah, K. D., & Kaskaoutis, D. G. (2017).
Estimating particulate matter using satellite-based aerosol optical depth and meteorological variables in Malaysia.
Atmospheric Research, 193, 142–162. https://doi.org/10.1016/j.atmosres.2017.04.019

FINDINGS METHODS
Data Sources:
PM10 data from ground-based monitoring stations.
ANN provided a robust framework for integrating AOD and MODIS AOD and meteorological data (temperature, relative humidity,
meteorological variables to estimate PM10. atmospheric stability index).
PM10 estimates were most accurate when all variables were Statistical Models:
MLR using stepwise regression to identify significant predictors.
included, especially during the dry season. ANN with a Multi-Layer Perceptron (MLP) architecture to handle non-linear
Models revealed hotspots of PM10 pollution linked to urbanization relationships.
and seasonal biomass burning. Validation:
Validation of models against PM10 measurements from 16 stations not used
in training.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Extensive cloud cover and data gaps in satellite observations over Malaysia. MODIS AOD is a key predictor of PM10, but its accuracy ANN model outperformed MLR with higher accuracy (R² = 0.71,
improves with meteorological variables. RMSE = 11.61 µg/m³ vs. R² = 0.66, RMSE = 12.39 µg/m³).
Challenges in correlating columnar AOD with surface PM10 due to varying ANN effectively captured the non-linear interactions Meteorological parameters significantly improved PM10 predictions
aerosol sources and meteorology. compared to AOD alone.
between variables, unlike MLR.
Seasonal variation observed: higher PM10 levels during the dry
Limited availability of ground-based data for model validation.
Spatial distribution highlighted urban-industrial areas as
season due to biomass burning aerosols from Indonesia.
hotspots for high PM10 levels.
 DATA ABSTRACT AIM OF STUDY
Ground PM2.5 data from the Department of The study focuses on estimating PM2.5 concentrations in
Environment, Malaysia. Malaysia using satellite-derived aerosol optical depth (AOD)
To develop an empirical model for estimating
AOD data from the Himawari-8 satellite. and pollutant gases.
It highlights the importance of remote sensing in areas PM2.5 levels across Malaysia using satellite
Pollutant gas data from the Sentinel-5P
satellite.
lacking ground monitoring data. data.

SIGNIFIICANCE OBJECTVIES

This research contributes to the understanding To utilize AOD from the Himawari-8 satellite and
of air quality in Malaysia and demonstrates the gases (CO, O3, NO2, SO2) from Sentinel-5P.
utility of satellite data in environmental To assess the accuracy of the Support Vector
monitoring Regression (SVR) model in estimating PM2.5.
FINE PARTICULATE MATTERS MAPPING IN
THE MARITIME REGION OF MALAYSIA
USING AEROSOLS AND POLLUTANT GASES DERIVED
FROM SATELLITE REMOTE SENSING
RECOMMENDATION 2023 PROBLEM STATEMENT
Future studies should address the issue of
Nurul Amalin Fatihah Kamarul Zaman1 , Kasturi Devi Kanniah Malaysia faces challenges in monitoring air
missing AOD data to enhance model 1, 2*, Dimitris G. Kaskaoutis3, 4 , Nurul Asyiqin Mohamad quality due to a limited number of ground-
accuracy and reliability Fadzil1 and Mohd Talib Latif based PM2.5 monitoring stations.

Kamarul Zaman, N. A. F., Kanniah, K. D., Kaskaoutis, D. G., Mohamad Fadzil, N. A., & Latif, M. T. (2023).
Fine particulate matters mapping in the maritime region of Malaysia using aerosols and pollutant gases derived from satellite remote sensing.
IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium.
Kamarul Zaman, N. A. F., Kanniah, K. D., & Kaskaoutis, D. G. (2017).
Estimating particulate matter using satellite-based aerosol optical depth and meteorological variables in Malaysia.
Atmospheric Research, 193, 142–162. https://doi.org/10.1016/j.atmosres.2017.04.0
1109/IGARSS52108.2023.10281576

FINDINGS METHODS
The model successfully identified AOD, The study employed the Support Vector Regression (SVR) technique to
analyze the relationship between PM2.5 and satellite-derived parameters.
CO, O3, NO2, and SO2 as significant predictors
of PM2.5 concentrations. Data was collected from ground monitoring stations and satellite
observations over a two-year period (May 2018 - April 2020).

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The study faced challenges related to missing AOD data, The model effectively estimated PM2.5 The SVR model demonstrated acceptable accuracy
which could affect the seamless mapping of PM2.5. concentrations even in the absence of ground with R² = 0.67, RMSE = 13.36 μg m⁻³, and NSE =
data, indicating its potential for broader 0.645.
applications.
 DATA ABSTRACT AIM OF STUDY
Meteorological and air quality data from WRF- The study addresses serious air pollution in Malaysia,
CMAQ modeling. - Emission inventories from particularly during the southwest monsoon, which adversely
various sources for accurate assessments.
TThe study evaluates air pollution in
affects human health.
It proposes using satellite data and robust statistical Southeast Asia due to local emissions and
techniques to increase the availability of PM2.5 data. transboundary air pollution (TAP), quantifying
their health impacts using air quality modeling

SIGNIFIICANCE OBJECTVIES
To integrate atmospheric gases data from Sentinel
TThe study provides critical data and insights for
5p and AOD from Himawari-8 with ground-based
regulatory actions to improve air quality and PM2.5 data.
public health in Southeast Asia. To provide PM2.5 estimates for the entire region of
Malaysia at a spatial resolution of 5 km and daily
intervals.
Source emission contributions to particulate matter and ozone,
and their health impacts in Southeast Asia
2024

RECOMMENDATION PROBLEM STATEMENT

Formulation of comprehensive emission
restriction strategies. - Improved There is inadequate understanding of the
regional cooperation for effective TAP contributions of local and transboundary
management. sources to air quality in Southeast Asia, despite
significant health impacts.

FINDINGS METHODS
Local emissions are the primary cause of Air quality modeling with the WRF-CMAQ system. - Species tagging method for
source apportionment. - Integrated Source Apportionment Method (ISAM) for
mortality due to PM2.5. - Industrial emissions estimating health effects.
are predominant in health impacts, particularly
in urban areas.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

TThe study highlighted seasonal variations in Approximately 900,000 premature mortalities
Limitations in understanding TAP contributions. PM2.5 and O3 concentrations, with biogenic annually due to air pollution. - 77% due to local
- Marginal effectiveness of existing collaborative measures among emissions, 23% attributed to TAP. - PM2.5 major
emissions contributing significantly to O3 levels
Southeast Asian nations. contributor, especially from industrial and residential
and local sources dominating PM2.5 pollution
sources
 DATA ABSTRACT AIM OF STUDY
Ground measurements from 65 air-quality The study estimates PM2.5 concentrations across Malaysia
stations. using multi-satellite data and machine learning (ML) models
Satellite data from Himawari-8, Suomi-NPP,
To develop simplified and computationally
(Random Forest, Support Vector Regression, and XGBoost).
and Sentinel-5P. Random Forest showed the best performance with a high efficient ML models for estimating PM2.5
Global PM2.5 product from Washington correlation (R2 = 0.64) and RMSE of 12.17 μg/m³. concentrations in Malaysia.
University.

SIGNIFIICANCE OBJECTVIES
Provides high-resolution PM2.5 maps for
To provide high spatial resolution estimates of
Malaysia.​ PM2.5 concentrations.​
Helps in formulating air quality improvement To cover remote areas without measurement
strategies. networks.
Useful for epidemiological studies and public To integrate multi-satellite data for seamless
health assessments. seasonal PM2.5 mapping.

Improving the quantification of fine particulates (PM2.5)

concentrations in Malaysia
using simplified and computationally
RECOMMENDATION efficient models PROBLEM STATEMENT
Nurul Amalin Fatihah Kamarul Zaman a,b , Kasturi Devi Kanniah a,c,* , Dimitris G. Kaskaout
High spatio-temporal variability of aerosols
Use Geostationary Environment Monitoring Spectrometer
(GEMS) for better AOD data. and pollutants makes air pollution
Expand the air-pollution monitoring network with low-cost assessment challenging.
sensors.
Existing models are complex and
Include vertical profiles of air pollutants in future models. computationally intensive.

FINDINGS METHODS
RF models provided satisfactory PM2.5
Multi-satellite data integration (Himawari-8, Suomi-NPP, Sentinel-5P).
estimates. Machine learning models: Random Forest (RF), Support Vector Regression
(SVR), and XGBoost.
Multi-satellite data integration improved Six sub-models for different locations and se
data coverage.​ Data normalization and gap filling techniques

PM2.5 levels were higher in urban/industrial

areas and during the dry season.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

RF outperformed XGBoost and SVR in most Random Forest showed the best performance
with R2 = 0.64 and RMSE = 12.17 μg/m³.
Insufficient data during wet and inter-monsoon seasons.​ cases.​
Models for urban/industrial and suburban sites,
Cloud contamination affecting satellite data accuracy. CO, AOD, and O3 were the most influential
Limited ground measurement data in remote areas and different seasons were developed
predictors. Lower performance in wet and inter-monsoon
Seasonal models showed higher PM2.5 seasons due to fewer data
levels in the dry season.
 DATA ABSTRACT AIM OF STUDY
Urban air quality has been deteriorating due to
Hourly data from 2005 to 2015 (11 years). anthropogenic sources and meteorological conditions.
Parameters: CO, O₃, NO₂, SO₂, PM₁₀, wind Study aims to determine seasonal variation in air pollutants
speed, and wind direction. (CO, O₃, NO₂, SO₂, PM₁₀) in urban Malaysia. To investigate long-term variations in
Standards referenced: Malaysia Ambient Air
Analyzed 11 years of data (2005–2015) across four stations concentrations of major air pollutants in an
using statistical software Openair.
Quality Standard (MAAQS) and National Key findings: O₃ and PM₁₀ exceed standards during monsoon urban Malaysian environment and their
seasons; shipping emissions and biomass burning are
Ambient Air Quality Standard (NAAQS). significant contributors. seasonal variations.

SIGNIFIICANCE OBJECTVIES
Helps understand the role of local and
Identify the trend and spatial variability of
seasonal factors in urban air pollution. pollutants (CO, O₃, NO₂, SO₂, PM₁₀).
Supports policymakers in crafting region- Assess pollutant levels against air quality
specific air quality management strategies. standards (MAAQS, NAAQS).
Provides long-term data for addressing Explore the effects of meteorological conditions on
health and environmental concerns. pollutant levels.
Variation of major air pollutants in diferent
seasonal conditions in an urban environment in Malaysia

RECOMMENDATION PROBLEM STATEMENT

Improve emission inventories, especially for vehicular and Urban air pollution is worsening globally
industrial sources. and regionally.
Strengthen regional policies to mitigate transboundary haze. Local emissions and seasonal
Promote cleaner fuels and stricter industrial emission meteorological factors are insufficiently
standards.
studied in urban Malaysia.
Enhance public awareness of monsoon-related pollution
Seasonal monsoon effects on pollutants
risks.
need further exploration.

FINDINGS METHODS
O₃ and PM₁₀ consistently exceed air quality standards,
influenced by meteorology and local sources. Study Locations: Klang, Petaling Jaya, Shah Alam, Cheras (four air quality
Industrial and vehicular emissions dominate CO, NO₂, and monitoring stations).
SO₂ levels. Data Collection: Hourly data from 2005 to 2015 on air pollutants and
Biomass burning contributes to severe haze episodes, meteorological variables.
Data Analysis: Statistical tools (Openair software) for temporal, spatial, and
affecting PM₁₀.
exceedance trends.
Meteorological factors (wind speed/direction) significantly LOESS smoothing for trend visualization.
affect pollutant distribution. Polar plots for wind direction and pollutant interaction.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Local sources (industrial, vehicular, biomass burning)
CO: Highest levels at industrial areas, related to vehicle and biomass combustion.
Significant seasonal variability makes pollution control heavily influence pollutant trends. O₃: Frequently exceeds standards during northeast monsoon, linked to NOₓ precursors.

complex. Monsoon seasons amplify pollutant levels through NO₂: Highest in industrial sites; related to vehicle and industrial emissions.
SO₂: Elevated near shipping ports and power plants; reduced over time due to cleaner
Lack of specific emission inventories limits precise source transport mechanisms. fuels.
Biomass burning in dry seasons is a critical source of PM₁₀ PM₁₀: Peaks during haze episodes, especially during the southwest monsoon (biomass
identification. burning).
and CO.
High concentrations during monsoon seasons challenge Regional variations in SO₂ and NO₂ reflect the influence of
Seasonal and spatial patterns linked to monsoon winds and anthropogenic sources.

mitigation strategies. industrial and traffic emissions.

 DATA ABSTRACT AIM OF STUDY
Himawari-8 AOD data (accessible via JAXA).
Ground-based pollution data (available upon The study focuses on estimating PM2.5 concentrations in Malaysia using
To estimate PM2.5 concentrations across
machine learning techniques.
request). It highlights the importance of remote sensing data, particularly from Malaysia using machine learning models.
Meteorological data including wind speed, Himawari-8, in air quality monitoring.

direction, temperature, and humidity.

SIGNIFIICANCE OBJECTVIES
The study contributes to the development of To evaluate the effectiveness of Support Vector Regression
models for systematic PM2.5 monitoring in SVR) and Random Forest (RF) models.
Malaysia. To analyze the influence of various meteorological and pollution
parameters on PM2.5 levels.
It highlights the potential of satellite technology in
assessing air quality over large areas.

Evaluation of Machine Learning Models for

Estimating PM2.5 Concentrations
RECOMMENDATION across Malaysia PROBLEM STATEMENT
Establish more air quality monitoring stations across Limited air quality monitoring stations in Malaysia
Malaysia. (only 65) are insufficient to cover the entire territory.
Encourage further studies using remote sensing data to There is a need for systematic monitoring of PM2.5,
enhance understanding of atmospheric particulates. especially in remote areas.

FINDINGS METHODS
Remote sensing data can effectively supplement ground-
Utilized hourly AOD data from Himawari-8 and ground-based pollution data.
based measurements for PM2.5 estimation. Developed and evaluated two machine learning models: SVR and RF.
The study emphasizes the need for more air quality Input variables included AOD, SO2, NO2, CO, O3, wind speed (WS), wind
monitoring stations to improve data coverage. direction (WD), temperature (TEMP), and relative humidity (RH)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Variable importance analysis indicated key predictors for The study found satisfactory accuracy in estimating PM2.5 concentrations despite
Challenges in representing PM2.5 pollution episodes due to challenges.
PM2.5 estimations. PM2.5 levels showed a mean of 21.9 μg/m³, with episodic high values due to regional
limited ground measurement stations. The models demonstrated the ability to represent spatial haze.
Biases in data representation and the need for more and seasonal variations in PM2.5 distributions.
comprehensive monitoring.
 DATA ABSTRACT AIM OF STUDY
MODIS AOD550 data (3 km resolution).
Ground-based PM10 data from 29 monitoring The study addresses air pollution as a significant risk factor affecting human
To estimate PM10 concentration over Malaysia using
health and the environment.
stations. Focuses on mapping PM10 spatial distributions in Malaysian cities due to remote sensing data.
Meteorological data (temperature, humidity, local and trans-boundary sources.

atmospheric stability) from the Department of

Environment (DOE).
AOD data from AERONET for validation.

SIGNIFIICANCE OBJECTVIES
The study contributes to understanding urban air To develop an empirical model for estimating PM10 concentration.
quality in Malaysia and highlights the potential of To validate satellite-derived AOD data against ground-based
remote sensing for environmental monitoring. measurements.

Satellite data for upscalling urban air pollution in

Malaysia
RECOMMENDATION PROBLEM STATEMENT
Future studies should include additional meteorological Limited PM10 monitoring stations in Malaysia (only
parameters (e.g., wind speed) for better PM10 estimation. 74) to cover a vast area (330,290 km²).
Caution in using MODIS AOD550 data due to its surface noise Need for effective methods to assess air quality
using remote sensing.

FINDINGS METHODS
Urban areas experience higher PM10 concentrations due to
Utilized MODIS AOD550 data at 3 km spatial resolution.
local and trans-boundary pollution sources. Ground-based PM10 data from 29 stations for model development and 16
Meteorological parameters improved PM10 estimation stations for validation.
results. Employed Artificial Neural Network (ANN) for PM10 estimation.
Incorporated meteorological data (temperature, humidity, atmospheric
stability).

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Spatial distribution of PM10 showed higher concentrations MODIS AOD550 correlated well with AERONET AOD (R² = 0.58, RMSE = 0.13).
High levels of surface noise in MODIS AOD550 data. ANN model yielded moderate accuracy for PM10 estimation (R² = 0.41, RMSE = 12.99
in urbanized areas (Klang, Shah Alam, Petaling Jaya). µg/m³).
Limited accuracy of the model when validated with The model's accuracy was moderate when validated
independent data. against additional stations.
 DATA ABSTRACT AIM OF STUDY
The analysis is based on data from various
monitoring stations, including ground-based This study provides an overview of atmospheric aerosol research in
The primary aim is to summarize existing research on
Malaysia, focusing on aerosols from anthropogenic sources and their
sensors, AERONET stations, and satellite impacts on health and climate. It emphasizes the need for more aerosol properties in Malaysia while identifying
observations, providing a comprehensive view of comprehensive monitoring and research to understand aerosol behavior and
critical knowledge gaps that hinder effective air
effects better.
aerosol behavior in Malaysia. quality management.

SIGNIFIICANCE OBJECTVIES
Understanding the dynamics of atmospheric aerosols The objectives include assessing the optical, physical, and chemical
characteristics of aerosols, monitoring air quality in urban areas, and
is crucial for formulating effective air quality policies
recommending further studies to fill identified gaps
and mitigating their adverse health effects in
Malaysia.

Overview of atmospheric aerosol studies in

Malaysia: Known and unknown
RECOMMENDATION PROBLEM STATEMENT
The authors recommend enhancing monitoring capabilities, The study highlights the lack of widespread aerosol
increasing research on aerosol sources and health impacts, and monitoring networks and field studies in Malaysia, which
improving public awareness strategies. impedes effective assessment and understanding of
aerosol impacts on air quality and public health.

FINDINGS METHODS
The study concludes that aerosols have considerable effects on
The research utilizes ground-based monitoring techniques, remote sensing data,
air quality, climate, and human health, especially in densely and satellite-based observations to analyze aerosol properties and trends
populated urban areas.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The authors conduct a thorough review of existing Findings reveal significant seasonal variations in aerosol concentrations and their
Key challenges include limited monitoring infrastructure, associations with health impacts, particularly during haze events.
literature and data, evaluating the strengths and
insufficient data on fine particulate matter (PM2.5), and weaknesses of current aerosol research in Malaysia
difficulties in assessing transboundary aerosol contributions.
 DATA ABSTRACT AIM OF STUDY
The analysis utilized aerosol optical depth (AOD)
data from the Himawari-8 satellite, tropospheric This study explores the significant impact of the COVID-19 lockdown on
The primary aim is to evaluate the extent of
atmospheric conditions in Southeast Asia, focusing on reductions in aerosols
NO2 levels from the Aura-OMI satellite, and and air pollutants. The findings reveal marked decreases in harmful reduction in anthropogenic emissions and
ground-based air quality monitoring data from substances, highlighting the potential for improved air quality during
consequent improvements in air quality due to
restrictions.
numerous stations across Malaysia, providing a governmental measures implemented during the
comprehensive assessment of air pollution trends. COVID-19 pandemic.

SIGNIFIICANCE OBJECTVIES
This research highlights the potential for significant The study seeks to quantify changes in aerosol optical depth (AOD) and
air quality improvements through reduced human concentrations of key pollutants (NO2, PM10, PM2.5, SO2, and CO) in
urban areas during lockdown periods compared to historical data from
activity and offers insights into how such findings can
2018 and 2019.
inform future environmental policies aimed at
combating air pollution

COVID-19's impact on the

atmospheric environment in
RECOMMENDATION PROBLEM STATEMENT
the Southeast Asia region
The study suggests that post-COVID-19 policies should be The COVID-19 pandemic led to widespread lockdowns,
formulated to maintain and enhance air quality improvements significantly altering human activities and industrial
observed during lockdowns, emphasizing the need for stricter operations, which prompted an investigation into its
regulations on emissions. effects on air pollution levels in Southeast Asia.

FINDINGS METHODS
The findings underscore the significant improvement in air
The researchers employed satellite observations from the Himawari-8 satellite for
quality during the COVID-19 lockdown, with notable AOD data and used tropospheric NO2 data from the Aura-OMI satellite.
reductions in major health-related air pollutants linked to Additionally, ground-based measurements from 65 monitoring stations across
decreased industrial and vehicular emissions. Malaysia were analyzed to assess air quality changes.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Statistical comparisons were made between the pollution The study reported a substantial reduction in air pollutants during the lockdown phase.
One of the challenges included accounting for the variability in Specifically, PM10 concentrations decreased by 26-31%, PM2.5 by 23-32%, NO2 by 27-
levels during the lockdown and the same periods in 34%, SO2 by 9-20%, and CO by 25-31% in urban areas of Malaysia.
pollution due to seasonal biomass burning, which could confound previous years (2018 and 2019). The results indicated a
the results. Additionally, the influence of meteorological factors consistent trend of reduced pollutant levels across various
on air quality was not fully quantified. urban centers.
 DATA ABSTRACT AIM OF STUDY
PM2.5 data (April 2018 - March 2020)
Malaysia faces significant air pollution with insufficient PM2.5
from 65 monitoring stations, AOD from data. This study uses Himawari-8 satellite AOD and Sentinel
To explore the potential of remotely sensed
Himawari-8, and gas data from Sentinel 5p gas data to map PM2.5 via machine learning algorithms, aerosol and gas data for mapping PM2.5
achieving high accuracy (R²=0.7) and aiding healthcare across Malaysia.
5p. analysis.

SIGNIFIICANCE OBJECTVIES
The study’s findings help address air 1. Enhance PM2.5 data availability.
quality issues, informing strategies to 2. Utilize satellite data for regional PM2.5 mapping.
mitigate health risks associated with air 3. Analyze the spatial and temporal distribution of
PM2.5.
pollution.

EXPLORING THE LINK BETWEEN GROUND BASED

RECOMMENDATION PROBLEM STATEMENT
PM2.5 AND REMOTEDLY
Integrate satellite-derived data for SENSED AEROSOLS Limited ground-based PM2.5 data hampers
continuous PM2.5 monitoring across AND GASES DATA effective air quality management in
Malaysia to inform public health policies. TO MAP FINE PARTICULATE MATTERS Malaysia.
IN MALAYSIA USING MACHINE LEARNING ALGORITHMS

FINDINGS METHODS
PM2.5 concentrations are influenced Machine learning algorithms: Random Forest and
predominantly by AOD and atmospheric Support Vector Regression.
gases, essential for health impact Satellite data: AOD from Himawari-8 and gas data
from Sentinel 5p
assessments.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Limited availability of ground-based data. Data indicated that atmospheric gases High accuracy in modeling PM2.5 concentrations was
Challenges in integrating satellite data due to cloud cover. and AOD have a stronger influence on achieved, revealing significant spatial and temporal
patterns.
PM2.5 levels than meteorological
factors.
 DATA ABSTRACT AIM OF STUDY
The study investigates the increase in air pollutants (PM10,
Historical air quality data from ground-based and satellite PM2.5, O3, CO) due to biomass burning, especially during the
observations.
dry season in Southeast Asia (SEA). To analyze the impact of biomass burning
It highlights the gradual rise in O3 concentrations linked to
Emission inventories related to biomass burning and
biomass burning emissions and the expected increase in PM2.5 on air quality in Southeast Asia.
industrial activities. due to industrial development and biomass burning. To improve the understanding of air
Meteorological data to assess the influence of weather on The importance of integrating observations and models for
air pollutant dispersion. better understanding air pollutant sources and movements is pollutant emissions and their effects on
emphasized. health and the environment.

SIGNIFIICANCE OBJECTVIES
The findings underscore the urgent need for improved air
1. To assess the trends of air pollutants in SEA.
quality management strategies in Southeast Asia.
The study contributes to the understanding of the relationship 2. To evaluate the effectiveness of inverse modeling
between biomass burning and air pollution, informing in reducing uncertainty in emission inventories.
policymakers and researchers.
3. To integrate observational data with modeling
techniques for better air quality predictions.

Air Pollution Modeling in Southeast Asia

RECOMMENDATION PROBLEM STATEMENT
An Overview
Air pollution in Southeast Asia is
Further emphasis on the development and application of inverse
modeling techniques in air quality assessments. exacerbated by biomass burning, leading to
Increased investment in monitoring infrastructure to gather more increased health risks and environmental
accurate data on emissions. degradation.
Implementation of policies to mitigate biomass burning and its
associated air quality impacts.
There is a need for improved modeling
techniques to accurately assess and predict
air quality impacts

FINDINGS METHODS
Biomass burning is a significant contributor to air
Utilization of in situ, airplane, and satellite
pollution in Southeast Asia, particularly affecting O3
measurements for monitoring air pollutants.
and PM2.5 levels.
Application of inverse modeling techniques to
The integration of observational data into models
integrate observational data with existing models.
enhances the reliability of air quality predictions. Analysis of historical data and trends in air quality
and emissions.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The study analyzed the correlation Significant increases in O3 and PM2.5
Challenges in obtaining accurate and comprehensive data on
emissions from biomass burning. between biomass burning activities and concentrations were observed, particularly during
Variability in air pollutant concentrations due to diverse air pollutant concentrations. peak biomass burning periods.
sources and meteorological conditions. It assessed the effectiveness of different Inverse modeling effectively reduced simulation
Limitations in existing models to fully capture the complexity emission inventories on modeled air errors and improved emission inventory accuracy.
of air pollution dynamics.
quality outcomes.
 DATA ABSTRACT AIM OF STUDY
The study presents a deep learning-based method for
Ground-level PM2.5 concentration data from 67 estimating surface PM2.5 concentrations using atmospheric To develop an accurate and interpretable
monitoring stations across Thailand. gas data from the TROPOMI satellite.
model for estimating ground-level PM2.5
Atmospheric gas concentration data from the TROPOMI It addresses the challenge of limited ground-level PM2.5
satellite, including NO2, SO2, O3, CO, and HCHO. monitoring stations in Thailand and the need for improved concentrations across Thailand using
spatial coverage.
Meteorological data from ERA5-Land, including satellite-derived atmospheric gas data.
The model demonstrates strong performance in capturing
temperature, humidity, and wind speed. To assess the influence of fire emissions
spatiotemporal variations in PM2.5 concentrations, particularly
Fire radiative power data to assess the impact of biomass during periods of high pollution linked to biomass burning.
burning on air quality.
on air quality in Thailand.

SIGNIFIICANCE OBJECTVIES
The study contributes to the understanding of air quality 1. To utilize the TROPOMI satellite data to estimate
dynamics in Thailand and highlights the importance of
PM2.5 concentrations without relying on aerosol
integrating satellite data with machine learning for effective air
optical properties.
quality management.
It provides a framework for future research on air pollution 2. To evaluate the model's performance against existing
estimation and management strategies in Southeast Asia. machine learning algorithms.
3. To analyze the impact of seasonal variations and fire
emissions on PM2.5 levels.

RECOMMENDATION PROBLEM STATEMENT

Estimation of surface PM2.5 concentrations from atmospheric gas species retrieved
from TROPOMI using deep learning: Impacts of fire on
Ground-level PM2.5 monitoring is limited in
Enhance the monitoring network to include more ground
stations, particularly in regions with high pollution levels. air pollution over Thailand Thailand, hindering effective air quality
Continue to refine and validate the deep learning model with management.
additional data to improve its accuracy and reliability. Existing models often underestimate PM2.5
Implement policies to mitigate the impacts of biomass burning on
air quality.
concentrations, especially during high
pollution events linked to biomass burning.

FINDINGS METHODS
Implementation of the Attentive Interpretable Tabular
The model successfully estimated PM2.5 concentrations Learning neural network (TabNet) to analyze
without relying on aerosol optical depth, demonstrating atmospheric gas species (NO2, SO2, O3, CO, HCHO)
its potential for remote air quality monitoring. from TROPOMI.
Fire emissions, particularly from agricultural burning, Data collection from 67 ground monitoring stations
significantly influenced PM2.5 levels, especially in across Thailand for model training and validation.
northern Thailand. Use of meteorological data and geographical features
as additional predictors in the model.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Limited availability of high-quality data, particularly in the The study conducted a detailed evaluation of model The TabNet model achieved an R² score of 0.873, indicating
southern region of Thailand, which affected model training performance across different regions in Thailand, strong predictive performance.
highlighting regional variations in PM2.5 The model outperformed traditional machine learning
and validation.
concentrations. algorithms (e.g., SVM, Random Forest, XGBoost) in estimating
Challenges in accurately capturing the spatial distribution of PM2.5 concentrations, particularly during high pollution
It assessed the correlation between fire radiative power
PM2.5 due to the uneven distribution of monitoring stations. events.
(FRP) and PM2.5 levels, demonstrating the impact of
The complexity of atmospheric processes and the influence of biomass burning on air quality.
The analysis revealed that carbon monoxide (CO) was the
various factors on PM2.5 concentrations posed difficulties in most significant predictor of PM2.5 levels.
model interpretation.
 DATA ABSTRACT AIM OF STUDY
Topic: Analysis of air pollution over Hanoi, Vietnam.
MERRA-2 BC: 3-hourly Black Carbon concentration Key Focus: Utilized satellite datasets (UVAI, MERRA-2 To analyze the current status of air pollution
(ug/m³) for 5 years (2012–2016). reanalysis data, MODIS fires) and weather information to
in Hanoi and assess the contribution of
UVAI: Ultraviolet Aerosol Index for biomass burning analyze air pollution trends and rice residue burning
aerosols. emissions. biomass burning (including rice residue) and
Findings: Forest biomass burning impacts were detected, but
Rainfall Data: CHIRPS daily precipitation data. other sources using advanced satellite
emissions from rice residue burning were difficult to capture
MODIS Active Fires: Satellite-based fire detections. datasets and meteorological information.
due to cloud cover.
Cloud Cover: MODIS cloud fraction data

SIGNIFIICANCE OBJECTVIES
1. Characterize pollution trends over Hanoi using satellite
Scientific: Highlights the limitations of current satellite systems data (UVAI, MERRA-2 BC).
in monitoring air pollution under cloud-prone conditions. 2. Identify sources contributing to air pollution during
Policy: Provides a foundation for better air quality management
different months (e.g., rice residue burning, forest fires).
in Southeast Asia, especially in high-risk regions like Hanoi.
3. Examine the influence of weather variables (rainfall,
wind patterns) on pollutant levels.
4. Determine the potential and limitations of satellite
datasets for monitoring rice residue burning emissions.

RECOMMENDATION PROBLEM STATEMENT

Enhanced Monitoring: Analysis of air pollution over Hanoi, Vietnam using multi-satellite and Air pollution in Hanoi is a major environmental issue caused by
Deploy higher-resolution satellite systems to improve spatial and temporal data quality. MERRA reanalysis datasets
Establish a network of ground-based monitoring stations to validate satellite data and diverse sources, including biomass burning and urban emissions.
capture surface-level pollutants. Satellite-based monitoring systems face challenges in accurately
Improved Data Integration:
capturing air pollution events due to:
Combine multiple datasets (e.g., PM2.5 and BC) to better characterize emissions from rice
residue burning and urban sources. High cloud cover in the Red River Delta.
Policy and Action: Limitations in detecting small-scale and episodic burning events.
Implement stricter controls on biomass burning and industrial emissions.
Lack of high-resolution datasets and validation from ground
Raise awareness among local communities about the health impacts of residue burning.
monitoring systems.

FINDINGS METHODS
Forest biomass burning during March-April significantly impacts air
Analysis Techniques:
quality in Hanoi.
Time-series analysis of pollutant concentrations over 5 years.
Rice residue burning in June and October does not show notable Rainfall adjustment using regression models to isolate BC trends.
pollution peaks, likely due to cloud cover, wet deposition, and low Monthly and diurnal variability of BC levels.
smoke plume heights. Active fire trends and spatial distribution analysis.
Urban and industrial emissions dominate during the dry season Wind pattern analysis to track pollutant transport from external sources.
Metrics:
(December-January). Peak UVAI values (positive for absorbing aerosols).
Pollutant transport from Laos, NW Vietnam, and Southern China Diurnal and monthly BC variations.
contributes significantly to Hanoi’s pollution levels. Number of active fire detections and seasonal fire trends.

ISSUES & DIFFICULTIES ANALYSIS Major Pollution Sources: RESULTS

Cloud Cover: UVAI and BC Data: March-April: Forest biomass burning in NW Vietnam and Laos impacts Hanoi.
Hanoi’s persistent cloudiness limited satellite-based observations, especially during UVAI effectively captured absorbing aerosols during forest burning seasons but failed during rice December-January: Industrial emissions and pollutant transport from Southern China.
biomass burning months (e.g., June). residue burning months due to cloud interference. June-October: Minimal increase in BC or UVAI during rice residue burning months due to rainfall
Coarse Spatial Resolution: BC data showed higher concentrations in the dry season, likely from industrial and urban emissions. and cloud cover.
MERRA-2 and UVAI datasets lacked the granularity needed to capture small-scale rice Weather Influence: Trends:
Rainfall significantly influenced pollutant levels, with BC concentrations dropping during high BC levels were significantly higher during nighttime (peaks at 9:30 PM–12:30 AM).
residue burning events.
rainfall months. Severe pollution episodes observed in December 2013 and January 2014.
Limited Ground Data: Key Observations:
Wind patterns explained pollutant transport from Northern Vietnam, Laos, and Southern China.
Lack of ground-based air pollution monitoring stations restricted validation of satellite Episodic Events: Hanoi experienced peak cloud cover during June, coinciding with rice residue burning, which
data. Severe pollution episodes were observed during certain dry months (e.g., December 2013), linked hindered satellite detection.
Rainfall Effects: to long-range pollutant transport and regional burning. Rainfall reduced BC levels through wet deposition.
Heavy rainfall diluted BC levels, complicating analysis of emissions during certain months. Synoptic wind patterns transported pollutants into Hanoi from regional sources.
 DATA ABSTRACT AIM OF STUDY
Focus: The study develops a systematic methodology to
Collection: Indoor and outdoor air samples using active air quantify and characterize airborne microplastics (AMPs) in To provide a rigorous and standardized
samplers. both indoor and outdoor environments.
methodology for analyzing airborne
Analysis: Spiked samples with synthetic polymers (HDPE, Techniques: Utilizes sonication, oxidative digestion with
PP) and real air samples. hydrogen peroxide, and density separation with zinc chloride. microplastics, ensuring reproducibility and
Findings: The methods are efficient, reproducible within 2
Techniques: Microscopy, µ-Raman spectroscopy, ATR- comparability in future research.
days, and cause minimal damage to particles, making them
FTIR, SEM-EDX for particle characterization.
suitable for environmental samples.

SIGNIFIICANCE OBJECTVIES
1. Develop an efficient and systematic approach for
Provides a comprehensive protocol for AMP analysis, collecting, treating, and analyzing AMPs.
addressing current gaps in methodology. 2. Test and validate the methodology on spiked samples
Offers a foundation for future research into AMPs'
to ensure effectiveness.
environmental behavior and health implications.
3. Characterize AMPs based on their physical and
polymeric properties.
4. Reduce the time and cost of analyzing airborne
microplastics while improving accuracy.

RECOMMENDATION PROBLEM STATEMENT

Tracking Microplastics in the Air: Cutting-edge Methods fo
r Indoor and Outdoor Environments
Adopt standardized methods for AMP analysis to enable data comparability across studies. The lack of standardized protocols for analyzing airborne microplastics
Invest in cost-effective and accessible analytical tools for broader application.
has limited the comparability and reliability of studies. This gap hinders
Enhance quality control measures to prevent contamination during sampling and
processing. our understanding of AMPs' transport, environmental dynamics, and
impacts on human health.

FINDINGS METHODS
Sampling:
Indoor and outdoor air sampled using portable active air samplers.
Quartz filter papers (pore size: 2.2 µm) used for particle collection.
AMPs are present in both indoor and outdoor air. Sample Pretreatment:
Fibrous microplastics dominate, likely originating from textiles and Sonication: Used to dislodge particles from filter paper.
Oxidative Digestion: Organic matter removed using 15% H2O2 at 70°C for 1 hour.
environmental plastic erosion. Density Separation: Zinc chloride solution used to separate synthetic particles from inorganic impurities.
The proposed methodology is time-efficient and reliable for AMP Analysis:
Visual examination using microscopy.
analysis.
Spectroscopic techniques (µ-Raman, ATR-FTIR) for polymeric analysis.
SEM-EDX for surface and elemental analysis.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Spiked Sample Testing:
The methodology is robust, efficient, and suitable for both indoor and outdoor air samples.
Recovery rate of synthetic particles was >95%.
Contamination risks during sampling and sample preparation. Key characteristics (size, shape, polymer type) of microplastics were successfully identified and Minimal weight loss (<10%) during oxidative digestion, indicating no damage to particles.
quantified.
High cost and technical expertise required for advanced spectroscopic and SEM analyses. Real Sample Application:
Techniques used ensure minimal contamination and high recovery rates. Microplastics detected in all indoor (4.34 ± 1.93 items/m³) and outdoor (0.93 ± 0.32 items/m³)
Challenges in distinguishing synthetic fibers from natural ones.
samples.
SEM analysis revealed morphological differences like cracked edges, pits, and smooth surfaces.
 DATA ABSTRACT AIM OF STUDY
Focus: The study investigates the influence of secondary
PM2.5 and PM0.1 Concentrations:
Collected from urban and transportation sites in Hanoi during October–
inorganic aerosol (SIA) on PM2.5 and PM0.1 concentrations To determine the contribution of SIA to
during air pollution episodes in Hanoi, Vietnam.
December 2020. PM2.5 and PM0.1 during air pollution
Meteorological Data: Key Findings:
Wind speed, direction, relative humidity, temperature, and radiation from SIA contributed 29.0% to PM2.5 and 14.1% to PM0.1 during episodes in Hanoi and understand the
local monitoring stations. episodes.
SIA Analysis:
roles of meteorological factors and long-
Long-range transport and meteorological factors such as wind
Ion concentrations measured via ion chromatography.
speed and relative humidity significantly influenced PM levels. range pollutant transport.
Trajectory Models:
HYSPLIT and CWT for long-range pollutant transport assessment.

SIGNIFIICANCE OBJECTVIES
1. Quantify the contribution of SIA to PM2.5 and PM0.1
Provides critical insights into the composition and behavior of concentrations during air pollution episodes and non-
PM during pollution episodes, aiding air quality management episode periods.
and policy development in Vietnam.
2. Analyze the effects of meteorological conditions on SIA
Highlights the importance of addressing both local and
and particulate matter levels.
transboundary pollution sources.
3. Investigate the impact of long-range transport on
PM2.5 and SIA levels using trajectory analysis.

RECOMMENDATION PROBLEM STATEMENT

Influence of Secondary Inorganic Aerosol on the Concentrations of PM2.5 and PM0.1
Policy Interventions:
during Air Pollution Episodes in Hanoi, Vietnam
Reduce emissions of SIA precursors (SO₂, NOₓ, NH₃) from transportation, industrial, and Air pollution, particularly PM2.5, poses significant health risks in
agricultural sources.
Hanoi, often exceeding national and WHO standards.
Enhanced Monitoring:
Establish comprehensive monitoring networks for PM2.5, PM0.1, and chemical Limited research exists on the chemical composition of PM during
components. high-pollution episodes in Vietnam, especially regarding SIA.
Research Expansion:
Further studies on PM chemical composition in other regions of Vietnam to develop
tailored air quality policies.

FINDINGS METHODS
Sampling:
PM2.5 and PM0.1 samples collected at two sites (HUST: urban, and CEM: transportation) in Hanoi during dry
winter (October–December 2020).
SIA is a significant contributor to PM2.5 during pollution episodes in Monitoring using quartz fiber filters with cyclone samplers.
Hanoi. Analysis:
Ion chromatography to measure water-soluble ions (SO₄²⁻, NO₃⁻, NH₄⁺).
Long-range transport from China and regional meteorology play
Mass and composition analysis of particulate matter.
critical roles in PM2.5 levels. Data Tools:
PM0.1 sources are predominantly local, influenced by traffic and HYSPLIT and concentration-weighted trajectory (CWT) models for long-range transport analysis.
Pearson correlation and polar plots for meteorological impact assessment.
urban activities.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

SIA Contributions:
Episodes: SIA contributed 29.0% to PM2.5 and 14.1% to PM0.1 concentrations.
Non-episodes: SIA contributions were slightly lower at 25.6% (PM2.5) and 10.6% (PM0.1).
Local vs. Regional Sources:
Ion Analysis:
Difficulty in isolating SIA sources due to overlapping contributions from local and PM0.1 was influenced more by local sources (e.g., transportation).
SO₄²⁻, NO₃⁻, and NH₄⁺ were the dominant ions in SIA.
transported pollutants. PM2.5 had regional sources, as indicated by similar concentrations across sites. NO₃⁻ showed the highest percentage increase during episodes.
SIA Trends:
Limited monitoring data for PM0.1 in Hanoi, restricting comparisons with global datasets. Meteorological Effects:
SIA levels were consistent with PM peaks, emphasizing their role in pollution episodes. Low wind speeds (<2 m/s) led to higher pollutant concentrations.
Challenges in analyzing meteorological effects on PM0.1 due to their local nature.
Relative humidity and temperature influenced PM0.1 more significantly than PM2.5.
Long-range Transport:
Pollutants from South China and the Gobi Desert contributed to PM levels in Hanoi during
episodes.
 DATA ABSTRACT AIM OF STUDY
The study focuses on validating the GEMS (Geostationary
Environment Monitoring Spectrometer) Aerosol Optical Depth
GEMS AOD product data collected from the GEMS sensor onboard the
GEO-KOMPSAT-2B satellite. (AOD) product against AERONET (Aerosol Robotic Network) The primary aim is to evaluate the accuracy
AERONET ground-based measurements of AOD at 550 nm, which served ground-based measurements. and reliability of GEMS AOD data in
as a reference for validation. It emphasizes the significance of monitoring air quality in
Historical data on biomass burning activities and air quality trends in the mainland Southeast Asia, particularly due to the influence of monitoring aerosol levels over mainland
region. agricultural biomass burning during the dry season. Southeast Asia, particularly in the context of
The findings indicate a strong correlation between satellite and
ground-based data, highlighting the potential of GEMS for air seasonal variations and pollution sources.
quality assessments.

SIGNIFIICANCE OBJECTVIES
1. To validate the GEMS AOD data by comparing it with
The study provides valuable insights into air quality
management strategies in Southeast Asia, emphasizing the role AERONET measurements from multiple sites.
of satellite data in understanding pollution dynamics. 2. To analyze seasonal variations in AOD and identify
It supports public health initiatives by identifying pollution peak periods associated with biomass burning.
sources and their seasonal variations, which can inform policy 3. To investigate the sources of aerosols and their impact
decisions. on air quality in the region.

RECOMMENDATION PROBLEM STATEMENT

Validation and analysis of GEMS Aerosol Optical Depth product 2
Enhance the processing and calibration of satellite data to improve accuracy during the against AERONET over mainland Southeast Asia
Air pollution, especially from aerosols, poses significant health risks
rainy season.
Increase the number of ground-based monitoring stations to provide complementary data and environmental challenges in Southeast Asia.
for satellite observations. There is a critical need for reliable satellite data to monitor and
Develop and implement policies aimed at managing and reducing biomass burning manage air quality effectively, particularly in regions affected by
practices to mitigate air quality issues.
seasonal agricultural practices.

FINDINGS METHODS
The study involved collecting GEMS AOD data at 550 nm wavelength from twelve AERONET sites over the
The GEMS AOD product is effective for monitoring aerosols during period from November 2021 to April 2024.
dry periods, particularly in relation to biomass burning. A statistical analysis was conducted to compare GEMS AOD with AERONET observations, focusing on
Significant peaks in AOD were observed in April, coinciding with the correlation coefficients and regression analysis.
Seasonal analysis was performed to assess variations in AOD and the influence of aerosol load density.
intensification of agricultural burning practices.
The rainy season presented challenges for monitoring, indicating a
need for improved data collection methods during this period.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

A strong overall correlation (R² = 0.75) was found between GEMS and AERONET AOD data from
8,569 paired observations.
The study utilized statistical methods to evaluate the performance of GEMS AOD, including
GEMS AOD tended to underestimate actual values, particularly for low AOD values (<0.5), with
The study faced challenges in obtaining consistent AOD data during the rainy season due regression analysis to determine the relationship between satellite and ground-based
a linear regression slope of 0.56.
to increased cloud coverage and lower aerosol concentrations. measurements.
Seasonal analysis revealed higher correlations during the dry season (R² values of 0.63 in 2022,
Seasonal trends were analyzed to understand the impact of agricultural practices, particularly
Attributing specific sources of aerosols was complicated by overlapping emissions from 0.79 in 2023, and 0.72 in 2024) compared to the rainy season (R² values of 0.11 in 2022 and 0.13
biomass burning, on aerosol concentrations.
various activities, including urban transportation and industrial sources. in 2023).
 DATA ABSTRACT AIM OF STUDY
The study presents an overview of the 7-SEAS/BASELInE
campaigns, which focused on understanding air quality and
The study utilized data from the 7-SEAS/BASELInE campaigns conducted
between 2013 and 2015, which included: aerosol-cloud interactions in Southeast Asia. The primary aim is to enhance the
AERONET and MPLNET data for assessing aerosol properties. It highlights the integration of satellite remote sensing and understanding of how biomass-burning
Satellite data from MODIS for monitoring aerosol optical thickness and ground-based measurements to assess the impact of biomass-
smoke plume distribution. burning aerosols on regional air quality. aerosols affect air quality and climate in
Ground-based measurements to complement satellite observations and
enhance understanding of local air quality dynamics.
The research emphasizes the collaborative efforts of Southeast Asia, particularly through their
international scientists and the significance of the findings for
future studies. interactions with clouds.

SIGNIFIICANCE OBJECTVIES
1. Characterization of Aerosols: To characterize the chemical
The findings provide a critical baseline for understanding the composition, microphysical properties, and radiative effects of
interactions between aerosols and clouds, contributing to the biomass-burning aerosols in the region.
broader knowledge of climate and air quality dynamics. 2. Assessment of Air Quality: To evaluate how biomass-burning
The study serves as a foundation for future research initiatives activities contribute to regional air quality and meteorological
aimed at addressing air quality challenges in Southeast Asia and conditions, particularly during peak burning seasons.
3. Interaction Evaluation: To investigate how biomass-burning
beyond.
aerosols interact with seasonal stratocumulus clouds, affecting
cloud formation and properties.

RECOMMENDATION PROBLEM STATEMENT

Satellite-Surface Perspectives of Air Quality and Aerosol-Cloud Effects on
the Environment: An Overview of 7-SEAS/BASELInE
Future research should focus on in-situ profiling of the planetary boundary layer (PBL) to The increasing concern over air quality due to both natural and
validate remote sensing retrievals and improve model initialization. anthropogenic emissions necessitates a deeper understanding of
There is a need to fully utilize high-frequency satellite observations to capture the diurnal
cycle of PBL height and its effects on biomass-burning aerosols.
aerosol-cloud interactions.
There is a critical need to quantify the effects of these interactions
on climate and air quality, especially in regions like Southeast Asia
that are significantly impacted by biomass burning.

FINDINGS METHODS
The research confirmed that biomass-burning smoke significantly Instrument Deployment: Utilization of AERONET (AErosol RObotic NETwork) sun-sky spectroradiometers
impacts air quality, particularly during the dry-winter monsoon and irradiance radiometers to collect data on aerosol properties.
Airborne Platforms: Employment of airborne measurement platforms to provide flexibility and mobility in
season.
data collection, allowing for real-time monitoring of aerosol properties.
The study highlighted the importance of local meteorological Satellite Data Integration: Use of satellite data (e.g., MODIS) to retrieve aerosol optical thickness and assess
conditions in determining the extent and impact of aerosol pollution the spatial extent of smoke plumes from biomass burning.

on air quality.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The study identified extensive aerosol plumes associated with biomass burning, particularly
during the dry season (March-April).
The analysis revealed distinct meteorological patterns driven by the winter northeast and summer
Significant findings include the retrieval of aerosol optical thickness (AOT) above clouds, which
Modeling Challenges: Difficulty in accurately representing the complex processes of southwest monsoons, which affect aerosol transport and dispersion.
enhances understanding of aerosol distribution and its effects on air quality.
aerosol-cloud-radiation interactions in atmospheric models. The study examined how local circulation patterns during peak burning seasons enhance the uplift
The results indicate that biomass-burning smoke is prevalent in the boundary layer during the
of pollutants, contributing to regional air quality issues.
Variability in Measurements: Challenges associated with the variability of aerosol dry-winter monsoon season, with local meteorological conditions influencing pollutant uplift.
properties and their interactions with changing meteorological conditions, complicating
data interpretation.
 DATA ABSTRACT AIM OF STUDY
The study investigates the variation in the global chemical
PM10 and PM2.5 mass concentration data collected during field studies.
Chemical composition data obtained from EDX analysis of individual
composition of PM2.5, focusing on emissions from biomass To characterize the chemical composition
burning, traffic, and cooking sources.
particles.
It highlights the significant contributions of these sources to air of PM2.5 and identify the contributions
Meteorological data, including humidity and temperature, to contextualize
pollution levels and source contributions.
quality issues, particularly during haze episodes. from various pollution sources, particularly
The findings underscore the need for targeted interventions to
improve air quality. biomass burning and urban emissions.

SIGNIFIICANCE OBJECTVIES
1. Analyze the individual particle composition of PM2.5 from
The study provides critical data for understanding the sources
different sources.
and impacts of PM2.5 on air quality and public health. 2. Assess the contributions of biomass burning, traffic, and cooking
Findings can inform policy and regulatory measures aimed at to overall PM2.5 levels.
improving air quality and protecting public health 3. Compare pollution sources in different geographical regions to
understand their impact on air quality.

RECOMMENDATION PROBLEM STATEMENT

Aerosol Pollutants during Agricultural Biomass Burning:
A Case Study in Ba Vi Region in Hanoi, Vietnam
Implement better management practices for agricultural burning to reduce emissions. High levels of PM2.5 pose serious risks to air quality and public
Enhance monitoring and regulation of urban emissions to mitigate their impact on air health, particularly from local emissions such as biomass burning
quality.
Promote public awareness and policy measures aimed at reducing pollution from identified
and urban traffic, necessitating a detailed analysis of their sources
sources. and impacts.

FINDINGS METHODS
Emissions from smoldering rice straw burning were found to Field emission scanning electron microscope (SEM) was employed for detailed particle morphology analysis.
significantly affect local air quality, contributing to high PM2.5 Energy dispersive X-ray (EDX) spectroscopy was used to determine the chemical composition of individual
particles.
levels.
Light attenuation measurements were conducted on quartz filters to assess aerosol absorption
Traffic and cooking sources also played a notable role in local characteristics.
pollution, highlighting the need for comprehensive source
management.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The study found high PM10 mass concentrations, particularly from rice straw burning, with
values reaching up to 167 µg m–3 during haze episodes.
Cluster analysis was utilized to categorize particles based on their chemical composition and
Total carbon (TC) was a significant contributor to PM10 mass, with organic carbon (OC)
Variability in smoke plume characteristics required repeated sampling to obtain morphology, revealing distinct pollution source signatures.
dominating PM10 mass by up to 42% during high smoke days.
representative data. Light attenuation data provided insights into the absorption characteristics of aerosols, indicating
The OC/EC (elemental carbon) ratio approached 20, indicating a predominance of smoldering
the impact of different sources on regional haze.
Isolating specific pollution sources in densely populated areas posed challenges, emissions.
complicating the attribution of PM2.5 levels to individual sources.
 DATA ABSTRACT AIM OF STUDY
The study investigates PM2.5 levels in Ho Chi Minh City (HCMC)
PM2.5 and PM10 Samples: Collected from various urban sites, reflecting
different levels of pollution exposure.
over a year (March 2017 - March 2018). To comprehensively analyze PM2.5
It focuses on urban and residential areas to assess air quality and
Meteorological Data: Included parameters such as wind speed, solar
identify sources of pollution. concentrations and their correlation with
radiation, relative humidity, and temperature, collected from local weather
stations to correlate with PM2.5 levels
The findings aim to inform local authorities for better pollution meteorological parameters in HCMC.
control measures.

SIGNIFIICANCE OBJECTVIES
1. Assess PM2.5 Concentrations: Measure and compare PM2.5
The study provides critical data for local air quality management levels across different urban and residential sites.
efforts in HCMC, aiming to align with national and WHO air 2. Analyze Spatial Variation: Determine how PM2.5 levels differ
quality standards. based on location and urban density.
It highlights the need for targeted interventions to mitigate air 3. Investigate Meteorological Effects: Examine how factors like
pollution and protect public health. temperature, humidity, and wind speed influence PM2.5
concentrations.
4. Identify Emission Sources: Explore potential sources contributing
to PM2.5 pollution in the studied areas
.

RECOMMENDATION PROBLEM STATEMENT

Current Status of Fine Particulate Matter (PM2.5) in
Vietnam’s Most Populous City, Ho Chi Minh City
Monitoring Strategies: Establish continuous air quality monitoring systems to track PM2.5 HCMC experiences high PM2.5 levels, which pose significant health
levels in real-time. risks and frequently exceed national air quality standards,
Urban Planning: Implement urban planning measures aimed at reducing traffic congestion
and promoting cleaner transportation options.
necessitating urgent investigation and intervention.

FINDINGS METHODS
Traffic and Urban Activities: High PM2.5 levels were closely linked Sampling Locations: Five sites were selected: two roadside (RO1, RO2), two residential (RE1, RE2), and one
to traffic density and various urban activities, such as construction urban background (UB).
Sampling Frequency: Air samples were collected twice weekly over 24-hour periods.
and commercial operations.
Statistical Analysis: Employed bivariate correlation and multiple linear regression to analyze the relationship
Meteorological Influence: Weather conditions, including wind between PM2.5 levels and meteorological data.
patterns and temperature fluctuations, significantly impacted PM2.5
concentrations.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Concentration Levels: PM2.5 concentrations were highest at RE2 (41.1 µg/m³) and lowest at RE1
(31.4 µg/m³).
Regression Analysis: The model showed an adjusted R² value of 0.31, indicating that
Correlation Findings: Strong correlations were found between PM2.5 levels and meteorological
Weekend Effect: Variations in PM2.5 levels on weekends may have distorted the annual meteorological factors explain about 31% of the variance in PM2.5 levels.
factors, particularly wind speed and solar radiation.
mean values, complicating data interpretation. Key Influencers: Wind speed and solar radiation were identified as significant factors affecting
PM2.5 concentrations
Source Isolation Challenges: The complexity of urban environments made it difficult to
pinpoint specific emission sources contributing to PM2.5 levels.
 DATA ABSTRACT AIM OF STUDY
The study explores the correlation between aerosol optical
AOD data from MODIS (2001-2018).
PM2.5 monitoring data from U.S. Embassy and other local studies.
depth (AOD) and PM2.5 concentrations in Kuwait and Iraq, To develop a model that accurately
emphasizing the public health risks associated with PM2.5
Meteorological data from ERA5, including temperature, humidity, and
exposure. represents the spatial distribution of
wind speed, to support the analysis.
It addresses the scarcity of historical PM2.5 data in many PM2.5 concentrations using AOD data as a
regions, which hampers effective air quality management and
health assessments. proxy.

SIGNIFIICANCE OBJECTVIES
The findings contribute valuable insights into air quality 1. To evaluate temporal and spatial variations in PM2.5
management and public health policy, particularly in regions where concentrations.
direct PM2.5 monitoring is lacking 2. To leverage AOD data to estimate PM2.5 levels in areas lacking
direct monitoring.

RECOMMENDATION PROBLEM STATEMENT

Estimation of ambient PM2.5 in Iraq and Kuwait
from 2001 to 2018 using machine learning and remote sensing
It is recommended to enhance PM2.5 monitoring networks and data collection efforts in The absence of comprehensive historical PM2.5 data in many countries
regions with limited historical data to improve air quality assessments and public health complicates the assessment of air quality and its health impacts
interventions.

FINDINGS METHODS
A strong correlation was found between AOD measurements and AOD data was sourced from the Moderate Resolution Imaging Spectroradiometer (MODIS) for the years
PM2.5 concentrations, indicating that AOD can serve as a reliable 2001-2018.
PM2.5 concentrations were obtained from various monitoring stations and studies.
proxy for estimating PM2.5 levels.
Meteorological data was sourced from the European Centre for Medium-Range Weather Forecasts (ERA5) to
The study underscored the significant public health risks associated support the analysis.
with PM2.5 exposure, particularly in urban areas.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

A statistically significant relationship between AOD and PM2.5 concentrations was established
for both Kuwait and Iraq.
The study employed regression analysis and random forest modeling techniques to compare the
Daily PM2.5 concentration estimates were generated, revealing patterns of air quality over the
The primary challenge was the limited availability of historical PM2.5 data, which effectiveness of AOD in predicting PM2.5 levels.
study period.
restricted comprehensive analysis. The models demonstrated varying degrees of accuracy, with random forest providing better
predictive capabilities.
Variability in pollution sources and meteorological conditions across different regions
posed additional complexities.
 DATA ABSTRACT AIM OF STUDY
The study investigates the levels and sources of airborne
The study utilized mass concentrations of airborne particles, elemental
concentrations from chemical analyses, and environmental parameters
particulate matter (PM) in school environments, focusing on both To evaluate the concentration,
indoor and outdoor air quality.
including temperature, humidity, CO, and CO2 levels to draw conclusions
It aims to understand the chemical composition of PM and its composition, and sources of airborne
about air quality in schools.
potential health impacts on children, providing a comprehensive particulate matter in schools to inform air
analysis of air quality in educational settings.
quality management strategies.

SIGNIFIICANCE OBJECTVIES
The study underscores the critical need for attention to air 1. To measure and analyze different sizes of particulate matter:
quality in schools, highlighting its direct implications for PM0.1, PM0.5, PM1, PM2.5, and PM10.
children's health and educational outcomes. 2. To assess the relationship between particulate matter
concentrations and environmental factors such as carbon
monoxide (CO), carbon dioxide (CO2), temperature, and humidity.

RECOMMENDATION PROBLEM STATEMENT

Assessment of Air Quality in School Environments in Hanoi, Vietnam: A Focus on
Mass-Size Distribution and Elemental Composition of
Schools should implement improved ventilation systems and regular air quality The presence of high levels of airborne particulate matter in schools
assessments to mitigate the risks associated with airborne particulate matter. Indoor-Outdoor Ultrafine/Fine/Coarse Particles
poses a risk to children's health and can negatively impact their
Establishing a routine monitoring program for airborne particles is essential to maintain a
safe learning environment.
cognitive development and learning abilities.

FINDINGS METHODS
Elevated concentrations of specific elements, such as heavy metals, Aerosol sampling was conducted using five-stage cascade impactors to collect particulate matter.
were detected, suggesting local pollution sources. Chemical analysis of the collected samples was performed using Scanning Electron Microscopy with Energy
Dispersive X-ray Spectroscopy (SEM-EDS) and Inductively Coupled Plasma Optical Emission Spectroscopy
Significant correlations were found between particulate matter
(ICP-OES) and Mass Spectrometry (ICP-MS).
levels and environmental factors, indicating the influence of Statistical analyses, including t-tests and Spearman correlations, were employed to evaluate the data.
external conditions on air quality.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Data quality assurance measures included weighing field and laboratory blank filters to ensure A total of 23 elements were identified in the airborne particles, with significant variations in their
Variability in elemental concentrations posed challenges in establishing consistent accuracy. concentrations across different sampling locations.
Statistical significance was determined at a 0.05 level, indicating reliable results for the Average concentrations of CO and CO2 were monitored, alongside comfort parameters such as
patterns. temperature and humidity, revealing concerning levels in certain areas
correlations observed.
Detection limits for certain elements resulted in incomplete data, affecting the overall
analysis.
 DATA ABSTRACT AIM OF STUDY
The data comprised PM 10 mass concentrations, metal content
measurements, and seasonal variations collected throughout the year 2018
The study investigates inhalable particulate matter (PM) in Bangkok, The primary aim is to evaluate the mass and
highlighting its health risks linked to respiratory and cardiac
diseases. PM concentrations were measured throughout 2018, size-resolved metal content of aerosols in
covering cool, hot, and rainy seasons, with a focus on metal content urban Bangkok, assessing their potential
and size distribution of aerosols
health impacts due to high exposure levels

SIGNIFIICANCE OBJECTVIES
This research underscores the urgent need for public health
interventions in cities with elevated PM levels, particularly 1. To measure PM 10 concentrations and analyze their metal
regarding the health impacts of metal exposure content.
2. To examine the aerodynamic size and mass distribution of
aerosols.
3. To identify the sources of metals found in PM

RECOMMENDATION PROBLEM STATEMENT

Aerosol mass and size‑resolved metal content in urban Bangkok, Thailand

The study recommends ongoing monitoring and regulation of PM emissions in urban areas to Urban areas like Bangkok experience high PM concentrations, which
mitigate health risks associated with high metal exposure are associated with serious health issues, necessitating a
comprehensive study of their composition and sources

FINDINGS METHODS
Smaller particles were found to contain higher concentrations of PM 10 sampling was conducted using a Sven Leckel LVS3 PM 10 sampler.
metals, raising concerns about their deposition in the lungs and Long-term measurements were taken from a rooftop and a 5th-floor inlet at the Chulabhorn Research
associated health risks Institute [2].
Particle number concentration was measured using a portable condensation particle counter (CPC)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Principal component analysis (PCA) revealed that the metals' sources included crustal minerals and PM 10 concentrations ranged from 30 to 100 μg m −3, with the highest levels observed during
The study encountered challenges such as filter overload due to high mass deposition, traffic emissions, indicating potential health risks from metals like nickel, arsenic, cadmium, and lead the cool season. The largest metal fractions in PM 10 were calcium, iron, and magnesium [

which sometimes led to premature termination of measurements

 DATA ABSTRACT AIM OF STUDY
The study investigates the causes and sources of PM2.5 pollution
Sampling Data: 88 PM2.5 samples (44 quartz and 44 Teflon).
Meteorological Data: Wind speed, direction, temperature, and humidity. during winter haze in industrial regions of northwest China. To analyze the sources, chemical composition,
Pollutants Monitored: PM2.5, NO2, SO2, CO, O3 concentrations. Sampling and analysis of PM2.5 were conducted in January 2016, and factors contributing to PM2.5 pollution
revealing secondary particulate matter, coal combustion, fugitive
dust, industrial emissions, and vehicular exhaust as major during severe winter haze episodes in
contributors. Results emphasized the significant role of unfavorable
meteorological conditions and secondary particle formation in
industrial areas of the Kuitun-Dushanzi-Wusu
exacerbating pollution. (K-D-W) region in Xinjiang, China.

SIGNIFIICANCE OBJECTVIES
he study provides actionable insights into the sources and drivers of Characterize PM2.5 composition under varying pollution levels.
PM2.5 pollution, aiding policymakers in designing effective Investigate meteorological influences and gas pollutants
strategies for air quality improvement in NW China. contributing to PM2.5 formation.
Use models like Positive Matrix Factorization (PMF) to identify
PM2.5 sources.
Pinpoint potential pollution source areas with the MeteoInfo model.

RECOMMENDATION PROBLEM STATEMENT

Composition Characteristics and Potential Regions of PM2.5 during
1. Strengthen regional emission controls, especially for coal and industrial sources. Winter Haze Pollution in Typical Industrial Areas, NW-China Rapid industrialization and coal dependency in NW China have
2. Promote clean energy alternatives and stricter vehicular emission standards. worsened PM2.5 pollution, especially in winter. The limited
3. Implement joint regional air quality management policies.
understanding of pollution causes and sources hinders effective air
quality management in industrial zones like K-D-W.

FINDINGS METHODS
1. PM2.5 pollution in K-D-W is significantly influenced by secondary Sampling: PM2.5 samples collected from three locations (Kuitun, Dushanzi, Wusu) using quartz and Teflon filters
aerosol formation and meteorological factors. over 15 days.
Chemical Analysis: Ion chromatography, thermal/optical analysis, and mass spectrometry to analyze PM2.5
2. Emissions from coal combustion and industrial activities were
components (e.g., SO4²⁻, NO3⁻, NH4⁺, organic/elemental carbon).
critical contributors. Statistical Models: Positive Matrix Factorization (PMF) for source apportionment.
3. Pollution was compounded by unfavorable topography and weather Trajectory Analysis: MeteoInfo model for potential source area mapping.

conditions.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

SO4²⁻ was the dominant component, increasing with pollution severity. High PM2.5 Levels: Average daily concentration of ~210 µg/m³, exceeding national air quality
Severe weather conditions (low temperatures, high humidity) aggravated pollutant Secondary particulate matter contributed significantly to PM2.5 levels, indicating poor atmospheric standards.
dispersion and enhanced chemical reactions under high humidity. Chemical Composition: Dominated by secondary inorganic aerosols (SO4²⁻, NH4⁺, NO3⁻) and
dispersion and transformation. carbonaceous compounds.
Fossil fuel combustion and industrial activities were major sources.
Limited regional studies on PM2.5 during specific pollution episodes. Source Apportionment: Secondary particulates (47%), coal combustion (19%), fugitive dust
(14%), industrial (10%), and vehicular emissions (10%).
Meteorological Impact: Low wind speed and high relative humidity promoted pollutant
accumulation and secondary particle formation.
 DATA ABSTRACT AIM OF STUDY
The study investigates volatile organic compound (VOCs) emissions
VOC concentration data from mass spectrometry.
Meteorological inputs from the WRF model. in Shaoxing, an industrial city in Zhejiang Province, during the ozone To analyze VOCs' characteristics and sources
Emission inventories for industrial, mobile, and solvent-related sources. (O3) pollution season. It measures VOC concentrations, identifies in Shaoxing during the O3 pollution season
key components, and calculates their Ozone Formation Potential
(OFP). The findings highlight industrial processes and solvent usage and evaluate their contribution to secondary
as dominant local sources and underscore the significant
contribution of regional transport from neighboring provinces to
pollutants like O3 and PM2.5.
VOC levels.

SIGNIFIICANCE OBJECTVIES
The study provides valuable insights into VOC emissions and their Measure VOC concentrations and identify their chemical
role in O3 pollution in Shaoxing, enabling evidence-based composition.
policymaking for air quality improvement in industrial cities. Calculate OFP for individual VOC species to assess their reactivity.
Apportion VOC sources, distinguishing between local and regional
contributions.
Recommend mitigation strategies to control VOC emissions and
reduce pollution.

RECOMMENDATION PROBLEM STATEMENT

Characteristics and Sources of VOCs Emission during the O3
1. Strengthen emission controls for industries and solvents. Pollution Prone Season Shaoxing, part of the Yangtze River Delta, faces severe O3 pollution
2. Enhance cross-provincial collaboration to manage regional transport of VOCs. in an Industrial City in Eastern China exacerbated by industrial growth. However, limited research exists on
3. Promote cleaner production technologies and enforce stricter vehicle emission standards.
VOC emissions and their role in pollution, particularly during the high-
risk O3 season.

FINDINGS METHODS
1. Industrial activities and solvent usage are the primary local Data Collection: Continuous VOC monitoring using high-sensitivity mass spectrometry from April to September
contributors to VOCs. 2019.
Chemical Analysis: Identification of 98 VOC species, including alkanes, benzene compounds, halogenated
2. Benzene compounds and olefins are critical targets for O3
hydrocarbons, and olefins.
mitigation. Modeling Approaches:
3. Effective pollution control requires cooperation between provinces Ozone Formation Potential (OFP) calculated using Maximum Incremental Reactivity (MIR) values.
Source apportionment using the Community Multiscale Air Quality (CMAQ) model with Integrated Source
to address regional transport. Apportionment Method (ISAM).

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Industrial and solvent-related activities dominate Shaoxing's VOC emissions. Average VOC Concentration: 106.96 ppb, dominated by halogenated hydrocarbons (31.68%),
Complex interactions between local and regional pollution sources. Regional transport significantly influences VOC concentrations due to monsoon-driven airflows. alkanes (30.36%), and benzene compounds (23.32%).
Benzene compounds, including m/p-xylene and toluene, are the most reactive and impactful on O3 OFP Results: Benzene compounds, olefins, and ketones contributed 41.62%, 25.81%, and
Limited understanding of the seasonal dynamics of VOC emissions. 11.48% of the total OFP, respectively.
formation.
Challenges in accurately modeling the transport of VOCs from neighboring regions. Source Contributions:
Local emissions: 64.51% from industrial processes, 25.69% from solvent usage, and 4.53% from
vehicle exhaust.
Regional transport: 46.26% from neighboring provinces, primarily Jiangsu (9.18%) and Fujian
(7.82%).
 DATA ABSTRACT AIM OF STUDY
The study investigates carbonaceous components (organic carbon,
OC, and elemental carbon, EC) in PM2.5 from September 2022 to
Hourly concentrations of OC, EC, and gaseous pollutants from September
2022 to August 2023. August 2023 in Jincheng, an industrial city in China. Annual OC and To analyze the seasonal variation and sources
Meteorological data, including wind speed, humidity, and temperature. EC concentrations were 5.54 µg/m³ and 0.98 µg/m³, respectively, of carbonaceous components in PM2.5 in
SOC estimated using the MRS method based on the OC/EC ratio. with seasonal variations. Biomass and coal combustion contributed
significantly in winter, while vehicle emissions were predominant in Jincheng and provide actionable insights for
summer. The OC/EC ratio was used to estimate secondary organic
carbon (SOC) using the Minimum R-Squared (MRS) method.
air pollution control.
Findings highlight that controlling emissions from combustion and
industrial processes can mitigate pollution.

SIGNIFIICANCE OBJECTVIES
This study provides vital insights into the seasonal dynamics of Characterize the concentrations and variations of OC and EC in
carbonaceous aerosols in industrial cities like Jincheng. The findings different seasons.
can guide policy interventions to address PM2.5 pollution Estimate SOC using the MRS method.
effectively and mitigate its environmental and health impacts. Identify the primary and secondary sources of carbonaceous
components.
Develop recommendations to mitigate carbonaceous aerosol
pollution.

RECOMMENDATION PROBLEM STATEMENT

Pollution Characteristics and Source Analysis of
1. Strengthen controls on biomass and coal combustion, especially in winter. Carbonaceous Components in Jincheng, as a typical industrial city, faces persistent PM2.5 pollution
2. Enhance vehicle emission standards to reduce summer pollution. PM2.5 in a Typical Industrial Cit driven by industrial emissions, coal combustion, and biomass burning.
3. Implement seasonal air quality management strategies, focusing on secondary organic
aerosol control in warmer months.
Despite its environmental significance, studies on carbonaceous PM2.5
components and their sources in Jincheng remain limited, hindering
effective mitigation strategies.

FINDINGS METHODS
1. Biomass and coal combustion emissions are major contributors to Site Description: Sampling was conducted at Jincheng's National Automatic Monitoring Station, located in a
residential-commercial area.
carbonaceous aerosols in winter, while vehicle emissions dominate Data Collection:
in summer. PM2.5 carbonaceous components measured using online OC/EC analyzers.
Hourly data collected for OC, EC, and gaseous pollutants (SO2, NO2, CO, O3).
2. SOC formation is seasonally dependent, driven by photochemical Quality Assurance: Data cleaned for outliers and missing values, yielding 8346 valid data sets.
and multi-phase processes. SOC Estimation: MRS method applied to differentiate primary organic carbon (POC) and SOC using seasonal
OC/EC ratios.
3. The OC/EC ratio is a useful indicator for identifying primary and
Source Analysis: Correlation and regression analyses conducted to identify key sources.
secondary sources.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Concentration Levels:
Annual average OC and EC concentrations: 5.54 µg/m³ and 0.98 µg/m³.
Carbonaceous components accounted for a significant portion of PM2.5 in Jincheng. Seasonal variations: Highest in winter due to combustion emissions; lowest in summer.
Complex interactions between primary and secondary sources made source The OC/EC ratio was influenced by seasonal meteorological factors (e.g., low wind speed in winter SOC Estimation:
promoted pollution accumulation). SOC concentrations peaked in winter (3.84 µg/m³) and autumn, with photochemical reactions
apportionment challenging.
SOC formation mechanisms varied by season, highlighting photochemical oxidation in summer and enhancing SOC formation in summer.
Meteorological conditions like frequent inversions in winter exacerbated pollution. SOC/OC ratio highest in autumn and summer due to secondary processes.
multi-phase reactions in colder months.
Lack of comparable historical data limited the analysis of long-term trends. Source Analysis:
Winter: Biomass and coal combustion dominated.
Summer: Vehicle emissions contributed most significantly.
Secondary processes like photochemical reactions and multi-phase transformations played key
roles.
 DATA ABSTRACT AIM OF STUDY
The study presents a machine learning approach to simulate ground
visibility, particularly during fog and haze events. It utilizes K-
Meteorological Data: Minute-level data on visibility, temperature, relative The primary aim of the study is to develop a robust model for
humidity, wind direction, wind speed, and dew point temperature collected Nearest Neighbors (KNN) and ensemble model algorithms,
integrating data from atmospheric boundary layer observations and predicting ground visibility by leveraging atmospheric
from December 2020 to January 2021.
Pollutant Concentration Data: Second-by-second measurements of conventional ground meteorological measurements. The findings boundary layer data and pollutant concentrations. The
various pollutants (SO₂, NO₂, VOCs, O₃, CO, PM levels) obtained from a indicate that the ensemble model consistently outperforms the KNN research seeks to address the limitations of traditional
drone-based atmospheric boundary layer detection system. algorithm in predicting visibility, especially when atmospheric visibility prediction methods that often rely solely on ground-
boundary layer data and pollutant concentrations are included. This
level meteorological data.
research underscores the importance of these factors in enhancing
visibility simulations, which can significantly improve transportation
safety and reduce economic losses.

SIGNIFIICANCE OBJECTVIES
Comparison of Algorithms: To evaluate the performance of KNN
This research is significant as it contributes to the fields of air
and ensemble model algorithms in simulating ground visibility.
quality and meteorology by providing a framework for improved
Impact Analysis: To assess how atmospheric boundary layer
visibility prediction. Enhanced visibility forecasting can lead to
meteorological data and pollutant concentrations affect visibility
better traffic safety measures, reduced economic losses due to predictions.
weather-related disruptions, and improved public health outcomes. Understanding Visibility Changes: To enhance the understanding of
the dynamics of visibility changes during fog and haze events,
providing insights for better forecasting.

RECOMMENDATION PROBLEM STATEMENT

Simulation of Ground Visibility Based on Atmospheric
1. Further refinement of the algorithms to improve their predictive capabilities. Boundary Layer Data Ground visibility is a critical factor for transportation safety, particularly
2. Incorporation of real-time data to enhance the accuracy and responsiveness of visibility Using K-Nearest Neighbors and Ensemble Model Algorithms during adverse weather conditions like fog. Traditional models often
forecasting systems.
3. Exploration of additional machine learning techniques to further improve model
rely on limited meteorological data, which can lead to inaccurate
performance. predictions. This study addresses the need for more comprehensive
models that incorporate atmospheric boundary layer data and pollutant
concentrations to improve visibility forecasting.

FINDINGS METHODS
1. The study confirmed that both KNN and ensemble models are Data Collection: The study utilized minute-level meteorological data (visibility, temperature, humidity, wind
speed, etc.) collected from December 2020 to January 2021 by the Meteorological Bureau of Donghai County,
effective for simulating ground visibility, with ensemble methods Jiangsu Province. Additionally, second-by-second atmospheric pollutant concentration data (including SO₂, NO₂,
yielding better results. PM levels, etc.) were obtained from a multi-rotor drone equipped with a boundary layer meteorological detection
system.
2. Atmospheric boundary layer conditions and pollutant Model Development: Three distinct visibility simulation schemes were developed using KNN and ensemble
concentrations were identified as critical factors affecting visibility, model algorithms. The models were designed to compare the simulation effects of the two algorithms based on
different input data configurations.
emphasizing the need for their inclusion in predictive models.
Performance Evaluation: The models were evaluated using metrics such as Mean Absolute Error (MAE) and Root
Mean Square Error (RMSE) to determine their accuracy in predicting visibility during fog events.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The ensemble model consistently outperformed the KNN algorithm in both fog events analyzed.
The analysis involved a comparative assessment of the simulation results obtained from the KNN and The integration of atmospheric boundary layer data significantly enhanced model accuracy,
Ensuring a balanced dataset that accurately represented different fog and haze conditions. ensemble models. The study highlighted the importance of using high-resolution data from the particularly in the visibility range of less than 200 meters.
atmospheric boundary layer, which provided a more comprehensive understanding of the factors For the first fog event, the ensemble model achieved a reduction in MAE by 33% and RMSE by
The complexity of integrating multiple data sources, which required careful preprocessing
influencing visibility. 24% compared to the KNN model.
and validation to ensure data quality.
The addition of pollutant concentration data further improved the accuracy of visibility
The need for real-time data integration to enhance the applicability of the models in predictions.
practical scenarios.
 DATA ABSTRACT AIM OF STUDY
The study investigates the chemical characteristics and source
Air quality levels and meteorological data from Yulin Air Quality
profiles of fine particulate matter (PM2.5) between suburban and TTo understand the variability and behavior of PM2.5
Monitoring Platform.
urban areas during clean and polluted periods in Yulin, China. It components under different conditions and assess the
Elemental compositions from chemical analysis of PM2.5 samples​
highlights significant contributors like vehicle exhaust, combustion associated human health risks. The study aims to improve
sources, and crustal dust. Health risks associated with PM2.5 are mechanisms for responding to polluted weather
also analyzed​

SIGNIFIICANCE OBJECTVIES
Analyze the chemical composition and pollution sources of PM2.5 in
TThe findings provide insights into pollution characteristics and
urban and suburban areas.
their health impacts, supporting better policy development for air Assess health risks using models such as PMF, APCS-MLR, and
quality management PSCF.
Identify potential regional pollution sources and their contribution to
air quality​

RECOMMENDATION PROBLEM STATEMENT

LPotential Risks of PM2.5 in Urban and Suburban Environments
1. Enhance inter-regional pollution control collaborations. : A Dual Perspective on Chemical Constituents and Pollution Sources DPM2.5 pollution poses significant health risks and remains a
2. Develop policies targeting industrial emissions and vehicle exhaust. challenge in many Chinese cities. Understanding its characteristics,
3. Establish comprehensive databases for spatiotemporal variabilit
sources, and health impacts in underdeveloped regions like Yulin is
limited, necessitating further investigation​

FINDINGS METHODS
Combustion sources and mixed sources significantly impacted Study Area and Sampling: Data collection occurred between November 2019 and January 2020 across urban and
suburban sites.
health risks. Chemical Analysis: Elements, ions, and carbon components were analyzed using techniques like ICP-MS and ion
Regional and local sources heavily influenced PM2.5 pollution. chromatography.
Models Used: Positive Matrix Factorization (PMF), Absolute Principal Component Scores-Multiple Linear
Urban areas showed greater exposure risks due to higher levels of Regression (APCS-MLR), and Potential Source Function (PSCF) models were applied to identify and apportion
industrial activities​ sources

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Higher pollution levels were recorded in urban areas.
The geo-accumulation index revealed significant anthropogenic influences, identifying harmful species Vehicle exhaust contributed 36.0%, combustion sources 22.0%, and crustal dust 20.3% during
Data collection and analysis highlighted challenges like varying pollution levels across
like As, Pb, and Zn. Combustion sources were the main contributors to carcinogenic and non- pollution periods.
regions, complex source apportionment, and meteorological influences carcinogenic risks Regional pollution primarily originated from local and neighboring cities​
 DATA ABSTRACT AIM OF STUDY
The study investigates the chemical characteristics and source
Air quality levels and meteorological data from Yulin Air Quality
profiles of fine particulate matter (PM2.5) between suburban and To understand the variability and behavior of PM2.5
Monitoring Platform.
urban areas during clean and polluted periods in Yulin, China. It components under different conditions and assess the
Elemental compositions from chemical analysis of PM2.5 samples
highlights significant contributors like vehicle exhaust, combustion associated human health risks. The study aims to improve
sources, and crustal dust. Health risks associated with PM2.5 are mechanisms for responding to polluted weathe
also analyzed​

SIGNIFIICANCE OBJECTVIES
Analyze the chemical composition and pollution sources of PM2.5 in
The findings provide insights into pollution characteristics and their
urban and suburban areas.
health impacts, supporting better policy development for air quality Assess health risks using models such as PMF, APCS-MLR, and
management​ PSCF.
Identify potential regional pollution sources and their contribution to
air quality​

RECOMMENDATION PROBLEM STATEMENT

How Does PM2.5 Impact the Urban Vertical Temperature Structure?
1. Enhance inter-regional pollution control collaborations. A Case Study in Nanjing PM2.5 pollution poses significant health risks and remains a challenge
2. Develop policies targeting industrial emissions and vehicle exhaust. in many Chinese cities. Understanding its characteristics, sources, and
3. Establish comprehensive databases for spatiotemporal variability​
health impacts in underdeveloped regions like Yulin is limited,
necessitating further investigation​

FINDINGS METHODS
Combustion sources and mixed sources significantly impacted Study Area and Sampling: Data collection occurred between November 2019 and January 2020 across urban and
suburban sites.
health risks. Chemical Analysis: Elements, ions, and carbon components were analyzed using techniques like ICP-MS and ion
Regional and local sources heavily influenced PM2.5 pollution. chromatography.
Models Used: Positive Matrix Factorization (PMF), Absolute Principal Component Scores-Multiple Linear
Urban areas showed greater exposure risks due to higher levels of Regression (APCS-MLR), and Potential Source Function (PSCF) models were applied to identify and apportion
industrial activities sources​

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Higher pollution levels were recorded in urban areas.
The geo-accumulation index revealed significant anthropogenic influences, identifying harmful species Vehicle exhaust contributed 36.0%, combustion sources 22.0%, and crustal dust 20.3% during
Data collection and analysis highlighted challenges like varying like As, Pb, and Zn. Combustion sources were the main contributors to carcinogenic and non- pollution periods.
carcinogenic risks Regional pollution primarily originated from local and neighboring cities​
pollution levels across regions, complex source apportionment,
and meteorological influences​
 DATA ABSTRACT AIM OF STUDY
The study analyzes PM2.5 sources in two coastal cities, Ulsan (South
Hourly PM2.5 measurements and chemical composition data for Ulsan
Korea) and Dalian (China), between July 2018 and September 2019. To identify, apportion, and locate the sources of PM2.5
(2018–2019).
Using methods like positive matrix factorization (PMF) and trajectory affecting two industrial coastal cities in East Asia and explore
Daily integrated PM2.5 samples and chemical analyses for Dalian during
the same period​ analysis, the study identifies the major contributors to PM2.5 in how these sources contribute to air quality during different
these cities and the probable source regions. The findings highlight seasons
shared and unique challenges faced by both cities in managing air
quality, emphasizing the need for both localized and cooperative
strategies​

SIGNIFIICANCE OBJECTVIES
Compare PM2.5 concentrations and chemical compositions in Ulsan
The study provides actionable insights into PM2.5 source
and Dalian.
apportionment in two industrial cities, contributing to regional Identify and apportion PM2.5 sources using PMF.
strategies for improving air quality and mitigating health impacts of Investigate the transport and local contributions of PM2.5 using
air pollution trajectory-based models like PSCF and SQTBA.
Develop strategies for PM2.5 reduction through national and
international collaboration​

RECOMMENDATION PROBLEM STATEMENT

Apportioning and Locating PM2.5 Sources Affecting Coastal Cities:
1. Strengthen local air quality management by addressing dominant sources in each city (e.g., Ulsan in South Korea and Dalian in China Despite substantial research on PM2.5, the transport of pollutants
residential burning in Dalian and industrial sources in Ulsan). between neighboring countries like South Korea and China and their
2. Foster international collaboration to manage cross-border PM2.5 transport.
3. Enhance real-time monitoring and source apportionment studies for dynamic air quality
respective contributions to PM2.5 concentrations in coastal cities
management​ remain understudied. Ulsan and Dalian, despite their geographic and
industrial similarities, require a comparative analysis to inform localized
and collaborative air quality policies​

FINDINGS METHODS
Regional and local PM2.5 sources were identified for both cities, Data Collection: Hourly PM2.5 concentrations and constituents were measured in Ulsan using in-situ monitors
and daily integrated samples in Dalian using filter-based methods.
with a notable influence from residential burning in Dalian and Source Apportionment: Positive Matrix Factorization (PMF) was used to resolve source contributions.
industrial emissions in Ulsan. Trajectory Analysis: PSCF and SQTBA models identified potential source regions and transport pathways of
PM2.5​
Seasonal variations emphasized the importance of targeted
interventions during heating periods

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Ulsan's PM2.5 was predominantly influenced by secondary sulfate, secondary nitrate, and
PM2.5 transport from Dalian to Ulsan showed stronger correlations during the heating period. Shared traffic-related sources, which contributed 83% during the heating period.
Differences in sampling methods between the two cities sources like secondary nitrate, sulfate, soil, and oil combustion exhibited regional transport influences, Dalian's major sources included residential burning in addition to sulfate and nitrate,
but local sources had significant contributions in both cities​ contributing 84.3% during the same period.
complicated direct comparisons. Seasonal differences showed higher contributions from residential burning in Dalian and
PM2.5 transport processes are influenced by seasonal secondary sulfate in Ulsan during the heating period

meteorology, requiring separate analyses for heating and

non-heating periods
 DATA ABSTRACT AIM OF STUDY
The study investigates the impact of meteorological factors and pollutant
precursors on the concentrations of PM2.5 and O3 in Harbin, Changchun,
Meteorological Data: Sourced from the National Climate Data Centre The primary aim of the study is to explore and quantify the
(NCDC) and the European Centre for Medium-Range Weather Forecasts and Shenyang, three major cities in Northeast China. The research employs
data-driven machine learning techniques, specifically Random Forest (RF) effects of meteorological conditions and pollutant precursors
Reanalysis 5 (ERA5).
Pollutant Data: Daily averaged concentrations of PM2.5, O3, CO, NO2, and Back Propagation Neural Network (BPNN) models, to analyze the on the concentrations of PM2.5 and O3 in the urban
SO2, and HCHO were collected from various air quality monitoring relationships between these factors from 2015 to 2021. The findings reveal environments of Harbin, Changchun, and Shenyang. This
stations in the three cities. significant seasonal variations in PM2.5 and O3 pollution, with PM2.5 understanding is crucial for developing effective air quality
peaking in winter months and O3 in summer months. The study emphasizes management strategies in these regions.
the need for integrated air quality management strategies that consider
both meteorological and pollutant influences.

SIGNIFIICANCE OBJECTVIES
Identify Key Meteorological Factors: Determine which meteorological
variables (e.g., temperature, humidity, wind speed) significantly influence
This research provides valuable insights into the dynamics of air PM2.5 and O3 levels in the three cities.
quality in Northeast China, contributing to the broader Assess Combined Effects: Evaluate how the interaction between
understanding of how meteorological and pollutant factors interact. meteorological conditions and pollutant levels affects air quality.
The findings can inform policymakers and researchers in developing Model Development: Develop predictive models using machine learning
targeted strategies to improve air quality and public health. techniques to forecast PM2.5 and O3 concentrations based on identified
factors.
Provide Recommendations: Offer actionable insights and recommendations
for policymakers to improve air quality management.

RECOMMENDATION PROBLEM STATEMENT

Impact of Meteorology on PM2.5 and O3 Pollution in
1. Enhanced Monitoring: The study recommends improved monitoring of both meteorological Three Provincial Cities in Northeastern Provinces of China Northeast China, particularly the provinces of Heilongjiang, Jilin, and
and pollutant factors to inform air quality management strategies. Liaoning, faces severe air pollution challenges, primarily due to
2. Policy Development: Policymakers should consider local conditions and seasonal variations
when developing interventions to mitigate air pollution.
industrial activities and high coal consumption. The cities of Harbin,
3. Public Awareness: Increasing public awareness about the sources and effects of air Changchun, and Shenyang are particularly affected by PM2.5 and O3
pollution can help foster community engagement in air quality improvement efforts. pollution, which poses significant health risks to the population.
Understanding the complex interactions between meteorological
conditions and pollutant emissions is essential for developing effective
strategies to mitigate air quality issues.

FINDINGS METHODS
Data Collection: The study utilized a comprehensive dataset that included:
Meteorological Data: Daily averaged temperature (T), relative humidity (RH), daily precipitation rate (PRE), wind
CO and NO2 are critical pollutants that significantly affect PM2.5 speed (WS), planetary boundary layer (PBL) height, and atmospheric pressure (PRS).
Pollutant Data: Daily averaged concentrations of PM2.5, O3, carbon monoxide (CO), nitrogen dioxide (NO2),
levels across all three cities. sulfur dioxide (SO2), and formaldehyde (HCHO).
Meteorological factors, particularly temperature and humidity, play Modeling Techniques:
a vital role in influencing O3 concentrations. Back Propagation Neural Network (BPNN): A machine learning model used to predict PM2.5 and O3
concentrations based on input features.
The impact of pollutants often exceeds that of meteorological Random Forest (RF): An ensemble learning method that improves prediction accuracy by combining multiple
factors, highlighting the need for comprehensive air quality decision trees.
Importance Ranking: The study ranked the importance of various meteorological and pollutant factors using the
management strategies that address both. RF model to identify which had the most significant impact on air quality.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Data Quality and Availability: The study faced challenges in obtaining
high-quality meteorological and pollutant data, particularly for certain Seasonal Variations: PM2.5 pollution was found to be most concentrated in the winter months
The study's analysis demonstrated that both meteorological and pollutant factors significantly (November, December, January), while O3 levels peaked during the summer months (May to
parameters like PBL height. influence PM2.5 and O3 concentrations. The RF model outperformed the BPNN in terms of predictive July).
Complex Interactions: The nonlinear relationships between accuracy, indicating that incorporating both types of factors leads to better model performance. The Model Performance: The R² values for the predicted versus observed PM2.5 concentrations were
meteorological conditions and pollutant levels complicated the modeling contributions of meteorological factors to PM2.5 and O3 were found to be between 16.73%–30.08% 0.47, 0.40, and 0.44 for Harbin, Changchun, and Shenyang, respectively. For O3, the R² values
and 15.31%–25.35%, respectively, while pollutants contributed 69.92%–83.27% and 74.65%–84.69%. were 0.64, 0.38, and 0.73.
process, requiring advanced machine learning techniques to capture Key Factors: The analysis identified CO and HCHO as the most significant factors influencing
these dynamics effectively. PM2.5 levels, while temperature and relative humidity were crucial for O3 concentrations.

Seasonal Variability: The significant seasonal variability in pollution

levels necessitated careful consideration of temporal factors in the
analysis.
 DATA ABSTRACT AIM OF STUDY
The study focuses on the pollution characterization, meteorological effects,
and sources of carbon aerosols in PM2.5 in urban Xiangtan during winter. It
Hourly measurements of OC and EC concentrations from December 2022 The primary aim of the study is to investigate the
to February 2023. highlights the increasing air pollution due to urbanization and
industrialization in China, with a specific emphasis on carbonaceous characteristics and sources of carbonaceous aerosols in PM2.5
Meteorological data, including temperature and relative humidity.
Information on local pollution sources, including traffic and industrial aerosols. The research involved measuring hourly concentrations of in Xiangtan, particularly during the winter months when
emissions. elemental carbon (EC) and organic carbon (OC) from December 2022 to pollution levels are typically higher. The study seeks to
February 2023, revealing significant insights into the sources and variations understand how these aerosols vary over time and how
of these pollutants. meteorological conditions influence their concentrations.

SIGNIFIICANCE OBJECTVIES
Measurement of Carbon Concentrations: To quantify the concentrations of OC
and EC in PM2.5 during the winter months.
This research is significant as it provides valuable insights into the Temporal Variation Analysis: To analyze the temporal variations of OC and EC
air quality challenges faced by Xiangtan and similar urban areas in concentrations and their correlation with meteorological factors.
China. The findings can inform policymakers and public health Source Identification: To identify the primary sources contributing to
officials in developing strategies to improve air quality and protect carbonaceous aerosols in the urban environment of Xiangtan.
public health. Impact Assessment: To assess the impact of local pollution sources and
meteorological conditions on the levels of carbon aerosols.

RECOMMENDATION PROBLEM STATEMENT

Pollution Characterization, Meteorological Effects,
1. Stricter Emission Regulations: Implementing stricter regulations on emissions from and Sources of Carbon Aerosols in Xiangtan, like many urban areas in China, is experiencing severe air
vehicles and industrial sources to reduce carbonaceous aerosol levels. PM2.5 in Urban Xiangtan during Winter quality issues due to rapid industrialization and urbanization. The
2. Public Awareness Campaigns: Enhancing public awareness regarding air quality issues and
promoting the use of cleaner technologies.
increase in PM2.5 levels poses significant health risks to the
3. Further Research: Conducting additional studies on source apportionment and the population. However, there is a lack of comprehensive studies focusing
formation mechanisms of carbon aerosols, particularly focusing on secondary organic on the specific sources and characteristics of carbonaceous aerosols in
carbon.
this region, which hampers effective air quality management and
policy-making.

FINDINGS METHODS
Concentration Levels: The average winter concentrations of OC and EC were found to be 10.75 ± 5.75 μg m–3
and 1.92 ± 0.77 μg m–3, respectively. These values accounted for 16.5% and 3.5% of the total PM2.5, indicating
The study concluded that traffic emissions and industrial activities were a significant presence of carbonaceous aerosols in the urban atmosphere.
OC/EC Ratio: The winter average OC/EC ratio was calculated to be 5.82, suggesting that secondary organic
significant sources of carbonaceous aerosols in Xiangtan. The research carbon (SOC) was a major contributor to the OC levels, accounting for approximately 66% of the total OC.
highlighted the importance of local pollution sources and Diurnal Variations: The study observed clear diurnal variations in OC levels, which did not align with the patterns
meteorological conditions in influencing the concentrations of OC and typically seen in metropolitan areas where vehicle emissions are the primary source. EC concentrations showed
relatively flat diurnal variations.
EC. The findings also indicated that the highest concentrations of OC
and EC were associated with low wind speeds, suggesting localized
pollution effects.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Interference from VOCs: One of the challenges faced during the study
was the potential interference from VOCs in the OC measurements, Seasonal Variations: PM2.5 pollution was found to be most concentrated in the winter months
The analysis revealed that OC and EC concentrations were influenced by meteorological conditions, (November, December, January), while O3 levels peaked during the summer months (May to
which required careful calibration and the use of denuders to minimize such as temperature and relative humidity (RH). The highest concentrations of OC were observed at July).
this effect. lower temperatures (7°C to 17°C) and higher RH (60% to 85%), conditions conducive to the formation Model Performance: The R² values for the predicted versus observed PM2.5 concentrations were
Data Variability: The variability in meteorological conditions and local of SOC. The study also employed potential source contribution function (PSCF) analysis to identify the 0.47, 0.40, and 0.44 for Harbin, Changchun, and Shenyang, respectively. For O3, the R² values
geographical sources of OC and EC. were 0.64, 0.38, and 0.73.
pollution sources posed challenges in establishing consistent patterns in Key Factors: The analysis identified CO and HCHO as the most significant factors influencing
the data. PM2.5 levels, while temperature and relative humidity were crucial for O3 concentrations.
 DATA ABSTRACT AIM OF STUDY
The scoping review summarizes published articles examining how air quality
changed due to community-level pandemic mitigation measures in the
The review analyzed studies that retrieved air quality data from both ground To summarize findings from published research on changes in
monitors and satellite remote sensing instruments, ensuring a comprehensive United States. It aims to connect measurable relationships between policies
that reduce emissions and pollutant concentration during the COVID-19 air pollution levels during the implementation of COVID-19
overview of air pollution across various locations in the U.S.
restrictions period. mitigation practices in the U.S.

SIGNIFIICANCE OBJECTVIES
Investigate how ambient air pollution changed from previous years to the
period of COVID-19 restrictions.
This review highlights the impact of reduced emissions on air Examine geographical variability and temporal factors affecting air pollution
quality during the COVID-19 pandemic, providing valuable insights changes.
for policymakers on how mitigation strategies can effectively Identify contextual factors influencing the relationship between air quality
enhance air quality. changes and COVID-19 mitigation measures.

RECOMMENDATION PROBLEM STATEMENT

Investigation of Black Carbon Wet Deposition to the
Future research should focus on conducting measurement-based studies with robust United States from National Atmospheric Deposition Network Samples How did community-level pandemic mitigation measures affect the
methodologies to address limitations found in current studies, aiming for a more detailed trends in air quality across America, and what contextual factors
understanding of the interactions between mobility patterns, meteorological variables, and air
quality.
influenced these changes?

FINDINGS METHODS
The study employed a scoping review methodology based on the framework proposed by the Joanna Briggs
Institute, consisting of:
Findings indicated significant improvements in air quality during the Identifying a research question and defining search strategies.
Selecting relevant studies from several online databases.
COVID-19 restrictions due to reduced vehicular traffic and industrial Extracting and charting data.
activity, though some pollutants showed mixed results contingent on Summarizing and reporting results
geographic and temporal contexts.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Some limitations included variability in data sources, the potential
inaccuracies due to remote sensing under specific conditions (e.g., cloud The review identified 66 relevant studies examining air quality changes during the pandemic,
Most studies utilized a comparative approach (before-and-after analysis) to assess air quality changes showcasing reductions primarily in nitrogen dioxide and carbon monoxide but varied results for
cover), and the challenge in performing comprehensive assessments due to using data from ground monitors and remote sensing. Various studies applied percent change particulate matter and ozone.
the diverse methodologies employed across studies. methodology to quantify the changes.
 DATA ABSTRACT AIM OF STUDY
The study investigates the effectiveness of portable air cleaners (PACs)
equipped with HEPA filters in reducing indoor particulate matter (PM)
Data were collected using REDCap electronic tools and analyzed with SPSS, To evaluate the real-world effectiveness of PACs in reducing
focusing on PM concentrations in µg m -3 during filter and sham conditions [ concentrations in homes of adults who tested positive for COVID-19. It was
a single-blinded randomized cross-over study conducted between PM concentrations in homes, particularly in the context of
November 2020 and May 2021, sampling 29 homes over two consecutive COVID-19
24-hour periods in both primary and secondary rooms

SIGNIFIICANCE OBJECTVIES
To measure the reductions in PM2.5 and PM10 concentrations during the filter
condition compared to the sham condition
This study contributes to understanding the real-world effects of
PACs on indoor air quality, particularly during the COVID-19
pandemic

RECOMMENDATION PROBLEM STATEMENT

Real-World Effectiveness of Portable Air Cleaners
Further studies should explore the long-term effectiveness of PACs and their performance in in Reducing Home Particulate Matter Concentrations The performance of PACs in naturalistic settings and their impact on
various indoor environments PM concentrations beyond the room containing the PAC is not well
characterized [

FINDINGS METHODS
A randomized cross-over trial design was used, with participants operating PACs in filter and sham conditions for 24
hours each. PM concentrations were continuously measured in both primary and secondary rooms
PACs were effective in reducing PM concentrations, especially in high-
PM environments, with close to 80% effectiveness reported

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Variability in PM contributions from indoor and outdoor sources affected the
effectiveness calculations, particularly in low-PM homes The study found significant reductions in PM concentrations, with PAC effectiveness of 78.8% for
Effectiveness was calculated as the percent difference in mean PM concentrations between sham and PM2.5 and 63.9% for PM10 in primary rooms, and 57.9% and 60.4% in secondary rooms,
filter conditions, highlighting the impact of central air systems on PAC performance respectively
 DATA ABSTRACT AIM OF STUDY
The study estimates the source contributions to 370 nm near-UV light
absorption at Fresno and Bakersfield, California, using source
The study utilized yearlong source apportionment data and MAC values from To quantify the contributions of various sources to near-UV
literature to estimate contributions to near-UV light absorption apportionment results and literature values of mass absorption cross-
sections (MAC) for BrC, BC, and dust. The reconstructed light absorption light absorption in the San Joaquin Valley, focusing on the
showed strong agreement with measurements from an AE33 aethalometer, roles of BrC, BC, and dust.
indicating the effectiveness of the approach used in the study

SIGNIFIICANCE OBJECTVIES
The findings contribute to understanding the dynamics of light- To analyze seasonal variations in light absorption.
absorbing aerosols in the San Joaquin Valley, which is vital for air To assess the impact of different sources on light absorption in urban and
rural settings.
quality management and climate change mitigation.

RECOMMENDATION PROBLEM STATEMENT

Attribution of Source Specific 370 nm UV
Mitigation strategies should prioritize reducing wood combustion in winter and vehicle Light Absorption from Understanding the contributions of various carbonaceous aerosols to
emissions in summer to effectively lower near-UV light absorption Dust, Brown Carbon, and light absorption is crucial for addressing air quality and climate change
Black Carbon at Two Locations in the San Joaquin Valley issues in the San Joaquin Valley.

FINDINGS METHODS
The study concluded that using source-specific MAC values effectively The study employed yearlong source apportionment results and applied literature MAC values to estimate light
models near-UV light absorption, highlighting the importance of absorption contributions from various sources

targeting wood combustion and vehicle emissions for mitigation

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The study found that winter absorption was dominated by wood combustion, while summer
Variability in measurements and uncertainties in MAC values were The results indicated significant seasonal changes in source contributions, with vehicle emissions being absorption was primarily from SOA and vehicle emissions. The correlation between total BrC
challenges faced during the study a major contributor in summer and wood combustion in winter absorption and OC concentrations was noted
 DATA ABSTRACT AIM OF STUDY
The study developed the Toxic-metal Aerosol Real Time Analyzer (TARTA)
for monitoring particulate metals in the atmosphere in near real-time.
Data on metal concentrations were collected from TARTA's measurements in To assess the concentrations of metal elements in PM2.5
various environments, including traffic-related emissions in the Caldecott TARTA is compact, inexpensive, and has high sensitivity and temporal
resolution. It was evaluated in three environments: a tunnel, urban area, and across different environments and evaluate TARTA's
tunnel and agricultural emissions from a rural site
rural site, identifying various trace metals associated with different sources effectiveness in real-time monitoring

SIGNIFIICANCE OBJECTVIES
This study highlights the importance of real-time monitoring of Determine metal concentrations in PM2.5.
atmospheric metals for public health and environmental Examine temporal patterns of metal emissions
assessments

RECOMMENDATION PROBLEM STATEMENT

Detection of Atmospheric Particulate Metals in
Future work should focus on refining TARTA's quantification models and improving its Near Real-Time: Tunnel, Urban and Rural Environments There is a need for effective monitoring of atmospheric particulate
calibration standards metals due to their health impacts and varying concentrations in
different environments [

FINDINGS METHODS
TARTA effectively measures particulate metals, revealing higher TARTA uses spark-induced breakdown spectroscopy (SIBS) for real-time measurement of metals. It was tested in a
concentrations in urban and tunnel environments compared to rural tunnel, urban, and rural settings [

areas

ISSUES & DIFFICULTIES ANALYSIS RESULTS

TARTA detected higher metal concentrations in the tunnel compared to rural areas, with Cu
Calibration discrepancies and environmental conditions affected the The study found significant variations in metal concentrations due to environmental factors, with averaging 24 ng m -3 in the tunnel and 15 ng m -3 in rural settings
reliability of TARTA's measurements TARTA showing moderate correlations with established techniques for some metals [
 DATA ABSTRACT AIM OF STUDY
The study examines PM2.5 total mass and species data from the Chemical
Speciation Network and the Federal Reference Method network to analyze
Data were sourced from the Chemical Speciation Network and the Federal To estimate the PM2.5 mass deficit trends and identify
Reference Method network, focusing on PM2.5 total mass and species mass deficit trends, contributing species, and their relation to temperature
across three New York State sites. A regression analysis was employed to contributing species while exploring their relationship with
measurements
harmonize measurements and assess the mass closure technique, revealing ambient temperature.
an increasing trend in mass deficit over time, particularly linked to organic
matter and temperature

SIGNIFIICANCE OBJECTVIES
The findings underscore the importance of understanding PM2.5 To determine the organic matter to organic carbon (OM/OC) ratio.
mass deficits for climate change projections and air quality To analyze the mass deficit trends over time.
To assess the correlation between mass deficit and temperature
managemen

RECOMMENDATION PROBLEM STATEMENT

Estimating the PM2.5 Mass Deficit Trends, Contributing Species,
Further investigation into nitrate ion associations and additional measurements of organic and Relation to Ambient Temperature at Three New York State Sites The study addresses the uncertainty in climate change projections due
matter precursors are suggested to enhance understanding of mass deficit dynamics to particulate matter and the need to understand the mass deficit in
PM2.5 measurements

FINDINGS METHODS
The study found a strong correlation between mass deficit and The study utilized data from two networks, applying a multiple linear regression model to explore contributing
temperature, particularly at higher temperatures, and noted an increase species to the mass deficit and employing mass closure techniques

in biogenic emissions contributing to the mass deficit [

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The results indicated that organic matter is the primary contributor to the mass deficit, with
Challenges included the loss of volatile organic matter and the evaporation The analysis highlighted fluctuations in the soil coefficient and emphasized the increasing role of significant seasonal variability observed
of ammonium nitrate, which affected mass measurements organic matter in the total mass [
 DATA ABSTRACT AIM OF STUDY
The study investigates the impact of COVID-19 stay-at-home orders on air
pollution levels in the US using EPA monitoring data. 1​It finds that during
Daily-average in-situ air pollution concentrations from EPA monitors. 19​ To quantify the impact of COVID-19 stay-at-home orders on
Meteorology data from ECMWF ERA5 Reanalysis. 20​ these orders, PM2.5 levels were slightly higher than expected, while ozone,
NO2, CO, and PM10 levels were lower. 2​The anomalies ended 5-6 weeks air pollution levels across the US. 4
US public transit mobility data from Google Community Mobility Reports.
after the orders were issued, with ozone, NO2, and CO returning to expected
levels, and PM2.5 and PM10 remaining higher than expected

SIGNIFIICANCE OBJECTVIES
The study provides insights into the effects of human activity on air Evaluate changes in air pollution levels before, during, and after stay-at-home
quality and the impact of large-scale social and economic changes orders. 1​
Analyze the variability in air pollution across different states.
on environmental conditions
Understand the broader implications of human activity on air quality. 5
Page 3
"More broadly, responses to Covid create a unique opportunity to quantify the
effect of human activity on air quality."

RECOMMENDATION PROBLEM STATEMENT

Changes in criteria air pollution levels in the US before, during, and
Further research is needed to quantify the extent to which observed changes are attributable after Covid-19 stay-at-home orders: Evidence from regulatory monitors TThe widespread social and economic changes due to COVID-19 might
to COVID-19-related changes versus other factors. 18 be expected to improve air quality, but national monitoring data
provide inconsistent support for this expectation

FINDINGS METHODS
Modest and temporary reductions in ozone, NO2, CO, and PM10 levels 1. Data acquisition from EPA AirData and ESDR. 7​
during stay-at-home orders. 2. Temporal correction using robust differences and regression analyses. 8​
3. Sensitivity analysis with temporal and weather correction. 9
4. Page 4
5. "2.4. Sensitivity analyses: temporal and weather correction, using regression analyses"

ISSUES & DIFFICULTIES ANALYSIS RESULTS

PM2.5 levels were slightly higher than expected during stay-at-home orders. 2​
Substantial variability in air pollution data. 14​ Temporal correction showed that PM2.5 levels were higher than expected, while other pollutants Ozone, NO2, CO, and PM10 levels were lower than expected. 10​
Complexity in isolating the effects of COVID-19 responses from other were lower. 12​ Concentration anomalies ended 5-6 weeks after the orders were issued.
Sensitivity analysis confirmed these findings, with minor variations. 13
factors. Page 4
"(As a sensitivity analysis, we used 5-rather than 10-year trends; results were similar (Fig.
S3).)"
 ABSTRACT
The study examines the projected increase in fine particulate matter (PM2.5)
concentrations and the degradation of air quality due to rising carbon dioxide levels
and climate change. 1​Using thirteen models from the Coupled Model Intercomparison
Project Phase 6 (CMIP6), the study finds that global average PM2.5 concentrations are
DATA expected to increase due to more abundant natural aerosols, such as dust and
secondary organic aerosols (SOA), driven by intensified West African monsoon and
AIM OF STUDY
enhanced emissions of biogenic volatile organic compounds (BVOCs). 2
CMIP6 model simulations. 8​ To quantify the impact of increasing CO2 on PM2.5-based air
Preindustrial control and 1% per year CO2 experiments. 23​
s43247-023-00688-7.pdf quality and understand the role of natural aerosols in
Data on natural and anthropogenic emissions, including BVOC emissions
Page 1-Current doc degrading air quality under current and future warming
from vegetation.
"Here, we show that thirteen models from the Coupled Model Intercomparison Project
Phase 6 all project an increase in global average concentrations of fine particulate scenarios
matter in response to rising carbon dioxide concentrations, but the range of increase
across models is wide."

SIGNIFIICANCE OBJECTVIES
The study emphasizes the importance of natural aerosols in future Assess the projected changes in PM2.5 concentrations due to rising CO2
air quality and the need to reduce CO2 emissions to mitigate the levels. 1​
Identify the main contributors to the increase in PM2.5. 3​4​
climate penalty on air quality.
Evaluate the inter-model variability in PM2.5 projections.
Highlight the importance of natural aerosols and BVOCs in future air quality

RECOMMENDATION PROBLEM STATEMENT

The projected future degradation in air quality is caused
Further research is needed to improve the understanding of climate-dependent BVOC by more abundant natural aerosols in a warmer world Greenhouse gas-induced warming is expected to increase fine
emissions and their impact on air quality. 16​Earth system models should focus on particulate matter concentrations, but significant uncertainties remain
representing chemistry-climate-biosphere interactions more comprehensively
regarding the magnitude and mechanisms of this response.

FINDINGS METHODS
Rising CO2 levels will lead to increased PM2.5 concentrations, 1. Analysis of thirteen CMIP6 models. 8​
2. Comparison of PM2.5 concentrations in 1% per year CO2 simulations relative to preindustrial control simulations.
primarily due to natural aerosols. 13​
9​
Enhanced BVOC emissions and dust from the Sahara are major 3. Examination of the contributions of dust, SOA, and other natural aerosols to PM2.5 increases. 3​
contributors to the projected increase. 19​ 4. Sensitivity analysis of models with and without climate-dependent BVOC emissions.

The increase in PM2.5 is comparable to the improvements in air

quality expected from strong air quality control measures

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Global annual mean PM2.5 is projected to increase by 0.62 ± 0.26 μg/m³ (10.8 ± 3.9%). 11​
The largest contributors to the increase are organic aerosols (63.7%), dust (26.8%), and sea salt
Significant inter-model variability in PM2.5 projections. The increase in PM2.5 is driven by enhanced emissions of natural aerosols due to warming. 13​ (6.6%). 3​
Uncertainties in the response of BVOC emissions to climate change. 16​ BVOC emissions are strongly correlated with surface temperature increases. 14​ Models with climate-dependent BVOC emissions show a larger increase in PM2.5 (1.7 ± 0.9 μg/m³)
The variability in PM2.5 projections is largely due to differences in the treatment of BVOC compared to those without (0.28 ± 0.27 μg/m³).
Limited understanding of the biochemical basis for CO2-mediated emissions among models Significant regional increases in PM2.5 are projected, particularly in South America, northern and
variations in BVOC emissions southern Africa, and the US. 12

s43247-023-00688-7.pdf
Page 3-Current doc
"Over the US and South America, the relatively large increase in total PM2.5 is due to large increases
in OA PM2.5 at 0.5 ± 0.3 μgm−3 (29.4 ± 19.2%) and 1.6 ± 1.4 μgm−3 (51.9 ± 32.8%), respectively."
 ABSTRACT
The review summarizes the evidence regarding changes in air quality in the United
States due to mitigation measures implemented during the COVID-19 pandemic.
Traffic-related emissions are a significant source of air pollution, and previous
DATA policies restricting traffic have shown to affect air quality levels. The pandemic
provided a unique opportunity to study these changes as state and local
AIM OF STUDY
governments enacted various measures, including stay-at-home orders and travel
The review utilized data from 66 scientific articles that measured various
restrictions. The review identified 66 articles that met the inclusion criteria, The aim of the study is to summarize existing evidence on
air pollutants, including NO2, CO, PM2.5, PM10, O3, and SO2.
revealing that nitrogen dioxide (NO2) and carbon monoxide (CO) levels generally how air quality in the United States changed as a result of
Data sources included ground monitors from the U.S. Environmental
decreased, while changes in fine and coarse particulate matter (PM2.5, PM10), community-level pandemic mitigation measures during
Protection Agency (EPA), satellite data from NASA, and other local air
ozone (O3), and sulfur dioxide (SO2) showed mixed results. The studies
quality monitoring agencies. The studies employed a variety of data
highlighted the nonuniform nature of air quality changes across different regions the COVID-19 pandemic.
collection methods, including ground-based measurements and remote
and within cities.
sensing techniques.

SIGNIFIICANCE OBJECTVIES
The review provides valuable insights into the relationship Exposure data used in studies.
between public health interventions and air quality, which can Methodologies employed in the assessments.
Results of the assessments regarding air quality changes.
inform future policies aimed at reducing air pollution and
Contextual factors that may explain variations in air quality changes.
improving public health outcomes. Understanding these Mediators and other outcomes related to air quality during the pandemic.
dynamics is crucial for developing effective environmental
policies in the context of ongoing and future public health
crises.

RECOMMENDATION PROBLEM STATEMENT

Understanding Air Quality Changes after Implementation of
Longer study periods to capture the full impact of COVID-19 mitigation measures on air Mitigation Measures during a Pandemic The COVID-19 pandemic has led to significant changes in human
quality. : A Scoping Review of Literature in the United States behavior and mobility patterns, which in turn may have impacted air
Comprehensive analyses that consider additional factors affecting air quality, such as
climate change and other environmental hazards.
quality. Understanding these changes is crucial for public health
Improved methodologies that allow for better comparisons across studies. and environmental policy, especially given the observed
nonuniformity in air quality changes across different geographic
regions and urban settings.

FINDINGS METHODS
The findings highlighted a consensus that NO2 and CO levels 1. Identifying a valid research question and defining search strategies.
2. Identifying relevant studies through searches in databases such as PubMed, Web of Science, and Google Scholar.
decreased during the pandemic, while the changes in PM2.5, O3,
3. Selecting studies based on predefined inclusion criteria.
and SO2 were more variable. 4. Extracting and charting data relevant to the research question.
The review noted that reductions in NO2 were linked to decreased 5. Collating, summarizing, discussing, and reporting the results.

vehicle traffic, while increases in O3 were attributed to complex

photochemical reactions influenced by reduced NOx emissions.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

A general decrease in NO2 and CO levels across the U.S.
Mixed results for PM2.5, PM10, O3, and SO2, with some studies reporting increases and others
Many studies had truncated timeframes, often only covering the early The analysis involved examining the methodologies used in the included studies, which varied reporting decreases.
months of the pandemic, which may have missed significant changes in widely. Many studies utilized percent change methodologies to quantify air quality changes, while The changes in air quality were not uniform, varying significantly by geographic location and even
others employed machine learning algorithms, regression analyses, and time-series analyses. within cities.
air quality during later surges of COVID-19. Contextual factors such as mobility data (from sources like Google and Apple) and traffic flow were
The impact of other factors, such as wildfires and seasonal variations, analyzed to understand their relationship with air quality changes.

was often not accounted for, potentially skewing results.

Variability in methodologies and reporting units across studies made
comparisons challenging.
 ABSTRACT
Fine-mode aerosols, despite their small size, significantly impact climate, the
environment, and human health. However, global assessments of their trends remain
DATA limited. This study uses satellite retrieval data and an interpretable deep learning
model to analyze fine-mode aerosol optical depth (fAOD) and fine-mode fraction
AIM OF STUDY
(FMF) globally from 2008 to 2017. Results show a significant increase in land FMF
Satellite Observations: MODIS, AERONET.
(2.34 × 10⁻³/year), while fAOD trends over land and ocean remained mostly To examine global trends in fine-mode aerosols from 2008–
Meteorological Data: ERA5 (temperature, pressure, wind speed,
unchanged. High fAOD and FMF levels were observed in China, India, Southeast Asia, 2017 using satellite data and deep learning, identifying
precipitation).
and Africa. Seasonal variations showed peak FMF in summer and lower levels in spatial, seasonal, and anthropogenic factors influencing these
Air Quality Data: OMI (ozone, NO₂, SO₂), MERRA-2 (dust and sea salt
spring. A decline in fAOD over China suggests that air pollution control measures
aerosols).
have been effective. The study also finds a strong correlation between ozone (O₃) changes.
Land Surface Data: MODIS-derived vegetation indices and land cover
levels and fine-mode aerosol changes, emphasizing the need for integrated pollution
classification. control policies

SIGNIFIICANCE OBJECTVIES
The study provides a comprehensive understanding of fine-mode Investigate the global distribution and trends of fAOD and FMF.
aerosol changes, offering valuable insights for climate research, air Analyze seasonal variations and their underlying causes.
Identify key environmental and anthropogenic factors driving aerosol changes.
quality management, and policymaking. It highlights the need for
Assess the effectiveness of pollution control measures in reducing fine-mode
coordinated regulations on both fine particles and ozone pollution. aerosols.

RECOMMENDATION PROBLEM STATEMENT

Understanding global changes in fine-mode aerosols during 2008–2017
Improve satellite retrieval accuracy for fAOD and FMF. using statistical methods and deep learning approach Fine-mode aerosols have a profound effect on air quality, climate, and
Develop policies addressing both fine particles and ozone pollution. health. However, existing studies lack comprehensive global
Expand ground-based aerosol monitoring networks.
Conduct further research on aerosol-climate interactions.
assessments due to challenges in measuring small-scale atmospheric
particles. Improved methodologies are needed to understand their
trends and influencing factors.

FINDINGS METHODS
and FMF increased significantly, while fAOD remained stable. 1. Satellite Data: Newly developed retrievals of fAOD and FMF.
2. Trend Analysis: Mann-Kendall trend test and Theil-Sen estimator to detect significant trends.
China’s emission control measures effectively reduced fine-mode
3. Deep Learning Model: Attentive interpretable deep learning (TabNet) for feature importance analysis.
aerosols. 4. Validation: Comparison with ground-based AERONET measurements for accuracy.
Seasonal variations align with dust storms, biomass burning, and 5. Deep Neural Networks (DNNs): Used for modeling nonlinear relationships.
6. Tree-Based Machine Learning Methods: Applied for feature importance analysis.
industrial emissions. 7. Attentive Interpretable Deep Learning Model (TabNet): A hybrid approach combining DNN accuracy with tree-
Ozone pollution is linked to changes in fine-mode aerosols. based interpretability, allowing identification of key environmental factors influencing aerosol trends.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Increasing FMF: Significant upward trend in land FMF (2.34 × 10⁻³/year).
Stable fAOD: No major changes over land or ocean.
Inaccuracies in satellite-based fAOD measurements over land. Declining AOD suggests reductions in coarse-mode aerosols, while FMF increases indicate shifts High Aerosol Concentrations: Found in China, India, Southeast Asia, and Africa.
Limited ground-based validation in some regions. toward fine particles. Seasonal Variability: FMF peaks in summer (>0.70) and dips in spring (<0.65).
Biomass burning, industrial emissions, and meteorological factors influence regional variations. Regional Trends: fAOD declined in China due to emission control policies but increased in India and
Difficulty distinguishing anthropogenic vs. natural aerosol sources. Policies targeting PM2.5 reductions may have led to rising ozone pollution. Africa.
Key Factor: Ozone (O₃) levels strongly correlate with fine-mode aerosol trends.
 ABSTRACT
This study evaluates long-term air quality in China using the Air Quality Index (AQI)
and the Air Quality Composite Index (AQCI) through the multinomial logistic
DATA regression method. Two models were developed using different dependent variables
(AQI and AQCI) while maintaining the same controlled variables, including GDP and
AIM OF STUDY
primary pollutants (O₃, PM₂.₅, PM₁₀, NO₂, SO₂, and CO). The models were tested using
Air pollution data (AQI, AQCI) from 31 capital cities in China (2013–2019).
real air quality data from 31 capital cities in China (2013–2019). The results indicate To develop and evaluate a multinomial logistic regression
GDP data from the National Bureau of Statistics of China.
that AQCI has greater significance than AQI for air quality predictions, as shown by model for predicting long-term air quality trends in China,
the ROC curve analysis. considering economic and environmental factors.

SIGNIFIICANCE OBJECTVIES
This study provides policymakers with a data-driven approach for Assess air quality trends in China using AQI and AQCI.
integrating economic and environmental policies in air quality Compare the performance of AQI and AQCI models.
Identify key pollutant contributors to air quality deterioration.
management.
Integrate economic factors, such as GDP, into air quality assessments.

RECOMMENDATION PROBLEM STATEMENT

LONG-TERM AIR QUALITY EVALUATION SYSTEM PREDICTION
Expand monitoring networks to include smaller cities. IN CHINA Traditional air quality prediction models focus on short-term
Implement stricter controls on PM₂.₅ and O₃ emissions. BASED ON MULTINOMIAL LOGISTIC REGRESSION METHOD assessments and often overlook economic influences. A comprehensive,
Develop policies integrating economic growth and pollution control.
long-term predictive model is needed to incorporate both pollution and
economic factors.

FINDINGS METHODS
AQCI is a better predictor of long-term air quality than AQI. 1. Data Collection: Air quality data (AQI, AQCI) from 31 capital cities in China (2013–2019).
2. Multinomial Logistic Regression: Used to classify air quality based on GDP levels and pollution concentrations.
PM₂.₅ and O₃ are the dominant pollutants influencing air quality.
3. Model Validation: ROC curve analysis and statistical evaluation in R-Studio.
Economic variables significantly impact pollution trends.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

AQI and AQCI effectively predict air quality trends.
PM₂.₅ and O₃ were identified as primary pollutants.
Data limitations for smaller cities. Economic growth does not always correlate with pollution control. Higher GDP regions showed varying pollution patterns compared to lower GDP areas.
Difficulty in accounting for external pollution sources (e.g., O₃ levels have increased, highlighting challenges in air pollution management. AQCI model demonstrated better predictive accuracy than AQI.
The ROC curve suggests that AQCI is a more reliable metric for long-term air quality prediction.
transboundary pollution).
Uncertainties in long-term predictions due to policy changes.
 ABSTRACT
This study employs an Artificial Neural Network (ANN) to predict hourly PM₂.₅ and
PM₁₀ concentrations in Chongqing, China. The model takes time, meteorological
DATA parameters, and PM₂.₅ (PM₁₀) concentrations as input and predicts their values for the
next hour. Thirteen training functions were tested, with the Bayesian Regularization
AIM OF STUDY
(trainbr) and Levenberg-Marquardt (trainlm) algorithms demonstrating the best
Hourly PM₂.₅ and PM₁₀ data from Environmental Monitoring of China.
predictive performance. The results indicate that ANN models provide accurate and To develop and evaluate an ANN-based model for predicting
Meteorological data from the China Meteorological Administration (Nov–
reliable short-term air pollution forecasts, aiding in pollution management. hourly PM₂.₅ and PM₁₀ concentrations in Chongqing,
Dec 2022).
improving air pollution forecasting accuracy.

SIGNIFIICANCE OBJECTVIES
This research enhances air quality management by providing a Develop an ANN model for short-term air quality prediction.
reliable, data-driven method for predicting PM₂.₅ and PM₁₀ Compare different ANN training algorithms.
Evaluate model accuracy using statistical metrics.
concentrations, allowing for better pollution control measures.
Provide data-driven insights for air pollution management.

RECOMMENDATION PROBLEM STATEMENT

Prediction of Hourly PM2.5 and PM10 Concentrations in Chongqing
Implement ANN models in real-time air quality monitoring systems. City in China Based on Artificial Neural Network Traditional air quality forecasting methods struggle with nonlinear
Incorporate additional meteorological variables for improved accuracy. pollution patterns influenced by meteorological conditions. ANN
Extend the study to other cities for validation.
models offer a potential solution but require validation for
effectiveness in real-world applications.

FINDINGS METHODS
ANN models provide high-accuracy short-term air pollution 1. Data Collection: Hourly PM₂.₅, PM₁₀, and meteorological data from Chongqing (Nov–Dec 2022).
2. Artificial Neural Network: Multi-layer ANN with various training algorithms (e.g., trainbr, trainlm).
forecasts.
3. Model Evaluation: Metrics include correlation coefficient (R), root mean square error (RMSE), and mean absolute
Meteorological factors significantly impact air quality predictions. error (MAE).
Trainbr is the most reliable training function for PM forecasting.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Best Training Algorithm for PM₂.₅: Trainbr (R = 0.9783, RMSE = 1.2271, MAE = 0.9050).
Best Training Algorithm for PM₁₀: Trainbr (R = 0.9773, RMSE = 1.8270, MAE = 1.4341).
Short-term forecasting is limited by data availability. ANN effectively captures complex pollution patterns. ANN successfully predicts hourly PM₂.₅ and PM₁₀ concentrations.
ANN performance may vary based on seasonal meteorological The Bayesian Regularization algorithm (trainbr) outperforms other training methods.
The model's predictive accuracy is higher for PM₂.₅ than PM₁₀.
conditions.
Computational complexity increases with additional variables.
 ABSTRACT
This study employs an Artificial Neural Network (ANN) to predict hourly PM₂.₅ and
PM₁₀ concentrations in Chongqing, China. The model takes time, meteorological
DATA parameters, and PM₂.₅ (PM₁₀) concentrations as input and predicts their values for the
next hour. Thirteen training functions were tested, with the Bayesian Regularization
AIM OF STUDY
(trainbr) and Levenberg-Marquardt (trainlm) algorithms demonstrating the best
Hourly PM₂.₅ and PM₁₀ data from Environmental Monitoring of China.
predictive performance. The results indicate that ANN models provide accurate and To develop and evaluate an ANN-based model for predicting
Meteorological data from the China Meteorological Administration (Nov–
reliable short-term air pollution forecasts, aiding in pollution management. hourly PM₂.₅ and PM₁₀ concentrations in Chongqing,
Dec 2022).
improving air pollution forecasting accuracy.

SIGNIFIICANCE OBJECTVIES
This research enhances air quality management by providing a Develop an ANN model for short-term air quality prediction.
reliable, data-driven method for predicting PM₂.₅ and PM₁₀ Compare different ANN training algorithms.
Evaluate model accuracy using statistical metrics.
concentrations, allowing for better pollution control measures.
Provide data-driven insights for air pollution management.

RECOMMENDATION PROBLEM STATEMENT

Prediction of Hourly PM2.5 and PM10 Concentrations in Chongqing
Implement ANN models in real-time air quality monitoring systems. City in China Based on Artificial Neural Network Traditional air quality forecasting methods struggle with nonlinear
Incorporate additional meteorological variables for improved accuracy. pollution patterns influenced by meteorological conditions. ANN
Extend the study to other cities for validation.
models offer a potential solution but require validation for
effectiveness in real-world applications.

FINDINGS METHODS
ANN models provide high-accuracy short-term air pollution 1. Data Collection: Hourly PM₂.₅, PM₁₀, and meteorological data from Chongqing (Nov–Dec 2022).
2. Artificial Neural Network: Multi-layer ANN with various training algorithms (e.g., trainbr, trainlm).
forecasts.
3. Model Evaluation: Metrics include correlation coefficient (R), root mean square error (RMSE), and mean absolute
Meteorological factors significantly impact air quality predictions. error (MAE).
Trainbr is the most reliable training function for PM forecasting.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Best Training Algorithm for PM₂.₅: Trainbr (R = 0.9783, RMSE = 1.2271, MAE = 0.9050).
Best Training Algorithm for PM₁₀: Trainbr (R = 0.9773, RMSE = 1.8270, MAE = 1.4341).
Short-term forecasting is limited by data availability. ANN effectively captures complex pollution patterns. ANN successfully predicts hourly PM₂.₅ and PM₁₀ concentrations.
ANN performance may vary based on seasonal meteorological The Bayesian Regularization algorithm (trainbr) outperforms other training methods.
The model's predictive accuracy is higher for PM₂.₅ than PM₁₀.
conditions.
Computational complexity increases with additional variables.
 ABSTRACT
With rapid industrialization, air pollution has become a major global issue threatening
human health. The World Health Organization (WHO) estimates that millions of
DATA deaths each year are linked to air pollution. This study aims to predict Air Quality
Index (AQI) levels in major Chinese cities using deep learning techniques. A dataset
AIM OF STUDY
containing hourly AQI data from 1,615 observation sites across China (2015–2019) is
Source: China Environmental Monitoring Center (2015–2019).
constructed. The study applies multiple machine learning methods, including Back The primary aim is to develop and evaluate deep learning
Variables: PM2.5, PM10, CO2, CO, O3, SO2, NO2.
Propagation Neural Networks (BPNN), Convolutional Neural Networks (CNN), Gated models for air quality prediction in major Chinese cities, using
Coverage: 1,615 monitoring stations across 190 cities in China.
Recurrent Units (GRU), Long Short-Term Memory (LSTM), and Bi-directional Long large-scale air pollution datasets.
Short-Term Memory (BiLSTM), to predict AQI levels. The results indicate that BiLSTM
outperforms all other models in forecasting air pollution levels accurately.

SIGNIFIICANCE OBJECTVIES
Construct an AQI dataset (2015–2019) containing hourly air quality data from
Provides real-time, accurate air quality forecasting for major 1,615 monitoring stations.
Chinese cities. Evaluate and compare different deep learning models (CNN, LSTM, BiLSTM,
Helps government agencies and policymakers implement better etc.) for AQI prediction.
pollution control measures. Identify the best-performing model for accurate air pollution forecasting.
Analyze the correlation between AQI and major pollutants (PM2.5, PM10, CO,
Supports public health initiatives by alerting citizens about
NO2, etc.).
hazardous air conditions. Provide real-time air quality forecasting for public health and policymaking.

RECOMMENDATION PROBLEM STATEMENT

Prediction of Hourly PM2.5 and PM10 Concentrations in Chongqing
Expand the dataset by including real-time pollution data from additional cities. City in China Based on Artificial Neural Network Traditional air pollution prediction models struggle with accuracy due
Integrate meteorological variables such as humidity, wind speed, and precipitation to to the complex, non-linear nature of pollution data. Existing
improve predictions.
Develop hybrid models combining CNN and BiLSTM for enhanced accuracy.
deterministic models based on physical laws often fail to find optimal
Use Transfer Learning to adapt models to different regions with varying pollution levels. parameters, leading to low prediction accuracy. Deep learning methods
offer a promising alternative but require validation with real-world air
quality datasets.

FINDINGS 1. Data Collection

METHODS
2. The dataset includes hourly AQI readings from 1,615 monitoring stations across 190 Chinese cities (2015–2019).
3. Core air pollutants included in the dataset:
4. PM2.5, PM10, CO2, CO, O3, SO2, NO2.
BiLSTM is the most effective model for air pollution forecasting. 5. The dataset is obtained from the China Environmental Monitoring Center.
Historical air quality data from the past 10 hours provides the best 6. Deep Learning Models:
7. Several machine learning techniques were used:
prediction results. 8. Back Propagation Neural Network (BPNN)
9. Convolutional Neural Networks (CNN)
PM2.5 and PM10 levels are the strongest predictors of AQI. 10. Gated Recurrent Units (GRU)
11. Long Short-Term Memory (LSTM)
12. Bi-directional Long Short-Term Memory (BiLSTM)
13. Performance Evaluation Metrics
14. The models were evaluated using:
15. Root Mean Square Error (RMSE)
16. Mean Square Error (MSE)
17. Mean Absolute Error (MAE)
18. R² (coefficient of determination)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Data limitations: Some monitoring stations had missing or inconsistent BiLSTM achieved the best performance overall, with the lowest RMSE and highest R² values.
readings. BPNN and CNN performed well, especially in Shanghai.
GRU and LSTM had slightly lower accuracy than BiLSTM but still performed better than traditional
Computational cost: Training deep learning models required high models.
computational resources. PM2.5 and PM10 had the highest correlation with AQI.
Air pollution levels were higher in winter compared to summer.
External factors: Sudden environmental changes (e.g., wildfires, BiLSTM consistently outperformed other models in all tested cities.
sandstorms) made predictions more challenging. CNN was more effective for short-term predictions, while LSTM-based models excelled in long-
term forecasting.
 ABSTRACT
Droughts are a major environmental concern and significantly impact air quality, yet
their precise causal effects remain unclear. This study investigates the relationship
between drought severity and air pollution levels in China from 2001 to 2018 using a
DATA Vector Autoregression (VAR) model combined with deep learning techniques. The
AIM OF STUDY
study uses the Standardized Precipitation Evapotranspiration Index (SPEI) as a
PM2.5 and PM10 concentrations (2001–2018) from China National measure of drought intensity and examines its impact on PM2.5 and PM10 The primary objective of this study is to quantify the causal
Environmental Monitoring Center. concentrations. Various deep learning models—including Artificial Neural Networks relationship between droughts and air pollution levels in
SPEI (drought index) data from meteorological agencies. (ANN), Deep Neural Networks (DNN), Long Short-Term Memory (LSTM), and Stacked
Meteorological data (temperature, humidity, wind speed). LSTM (SLSTM)—are applied to predict air pollution variations due to drought events.
China, using advanced econometric and deep learning
Results indicate that droughts lead to a 9.08 µg/m³ increase in PM2.5 and a 12.46 techniques.
µg/m³ increase in PM10 within a 95% confidence interval. The study highlights LSTM
as the most accurate predictive model, offering valuable insights for policymakers in
environmental management.

SIGNIFIICANCE OBJECTVIES
Analyze long-term trends (2001–2018) in air pollution and drought
Bridges the gap between climate change (droughts) and air occurrences in China.
pollution research. Examine the causal relationship between drought severity (SPEI) and
Provides a data-driven approach to managing air quality during PM2.5/PM10 concentrations using the VAR model.
extreme weather events. Apply deep learning models to improve air pollution forecasting during
drought periods.
Helps policymakers design targeted environmental
Compare different models (VAR, ANN, DNN, LSTM, SLSTM) for predicting air
interventions. pollution changes.
Provide insights for policymakers on mitigating the impact of drought-induced
air pollution.

RECOMMENDATION PROBLEM STATEMENT

Exploring Causal Effects of Droughts on Air Quality
Expand real-time monitoring networks to improve data accuracy. for Environmental Sustainability in Droughts have become increasingly frequent due to climate change, yet
Integrate drought forecasting models with air pollution prediction systems. China Using VAR Model and their impact on air pollution is not well understood. Traditional
Implement stricter emission controls during drought periods.
Develop early warning systems using LSTM models for pollution forecasting. Deep Learning Techniques statistical models struggle to capture non-linear relationships between
drought and pollution, requiring advanced deep learning methods to
improve predictive accuracy.

FINDINGS . Data Collection

METHODS
1. Time Period: 2001–2018
2. Pollution Data: Hourly and daily PM2.5 and PM10 concentrations from China's National Environmental Monitoring Center (CNEMC).
3. Drought Data: Standardized Precipitation Evapotranspiration Index (SPEI) from climate monitoring sources.

Droughts have a measurable impact on air pollution, particularly 4. Meteorological Data: Wind speed, humidity, temperature from the China Meteorological Administration.
5. Study Regions: Focused on Southwest and Central China, where drought frequency is high.

PM2.5 and PM10 levels. Vector Autoregression (VAR) Model

1. Used to establish causal relationships between drought intensity (SPEI) and air pollution levels.

LSTM-based models outperform traditional methods in predicting 2. Applied Granger Causality Tests to determine whether past drought levels could predict future air pollution changes.
3. Impulse Response Function (IRF) used to measure how long a drought event affects air quality.

pollution trends during droughts. 4. Forecast Error Variance Decomposition (FEVD) examined how much of the pollution variability is explained by droughts.
Deep Learning Models for Prediction

Southwest and Central China are most vulnerable to increased air 1. Artificial Neural Network (ANN): Basic model with multiple layers for pollution forecasting.
2. Deep Neural Network (DNN): An advanced version of ANN with more layers to capture complex relationships.

pollution during dry periods. 3. Long Short-Term Memory (LSTM): Used to model sequential dependencies in pollution data.
4. Stacked LSTM (SLSTM): A deeper LSTM architecture to enhance long-term prediction accuracy.
Performance Evaluation Metrics
1. Root Mean Square Error (RMSE)
2. Mean Absolute Error (MAE)
3. R² (coefficient of determination)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Causal Relationship between Drought and Air Pollution (VAR Model Results)
Droughts increased PM2.5 by 9.08 µg/m³ and PM10 by 12.46 µg/m³ within a 95% confidence interval.
Regions with the highest impact:
Data availability: Some monitoring stations had missing records. Southwest China: Highest increase in PM10.
Central China: More severe PM2.5 accumulation.
Computational complexity: Training deep learning models required high Impulse Response Analysis:
PM levels remained elevated for ~7–10 days after a drought event.
processing power. Variance Decomposition (FEVD) Findings:
Up to 27% of PM2.5 variation can be explained by drought conditions.
Uncertainty in climate variables: Variations in temperature and wind LSTM had the highest accuracy (RMSE = 6.98, R² = 0.91).
SLSTM was slightly less effective but still performed well.
patterns affected pollution transport. ANN and DNN were less effective at capturing long-term dependencies
Drought conditions significantly increase PM pollution levels.
LSTM and SLSTM models were best suited for capturing long-term air pollution trends.
Short-term pollution spikes (1–2 days) were better predicted by DNN models.
Regions with high industrial activity showed a stronger correlation between drought and air pollution.
 ABSTRACT
Accurate prediction of air quality response to emission changes is essential for
developing effective pollution control policies. Traditional methods such as Chemical
Transport Models (CTMs) and Polynomial Response Surface Models (pf-RSMs) are
DATA computationally expensive and require multiple simulations. This study introduces a
AIM OF STUDY
novel Deep Learning-based Response Surface Model (DeepRSM) that integrates deep
Air Quality Data: CMAQ CTM model simulations for PM2.5 and O3 learning techniques with chemical indicators of pollutant formation. By requiring only To develop and validate a deep learning-based response
concentrations. two CTM simulations (baseline and fully-controlled scenarios), DeepRSM significantly surface model (DeepRSM) that can accurately predict the
Emission Data: 40 emission control scenarios covering NOx, SO2, NH3, reduces computational cost while maintaining high prediction accuracy. The results
VOCs. demonstrate that DeepRSM captures the nonlinear response of PM2.5 and O3 to
response of air quality (PM2.5 and O3) to emission changes
Chemical Indicator Data: 18 chemical species used for model training. emission changes and performs better than traditional pf-RSM models. while significantly reducing computational requirements
Meteorological Data: Temperature, wind speed, humidity from WRF
compared to traditional modeling approaches.
Model.

SIGNIFIICANCE OBJECTVIES
Develop a deep learning-based air quality response model (DeepRSM) using
Convolutional Neural Networks (CNNs).
Enables real-time, high-accuracy air quality forecasting for
Compare DeepRSM’s performance with traditional pf-RSM models and CTM
policymakers. simulations.
Reduces computational cost and time compared to pf-RSM and Analyze the nonlinear relationship between emission changes and air
CTM. pollutant concentrations.
Supports proactive pollution control strategies with data-driven Evaluate DeepRSM’s effectiveness across different time periods and spatial
decision-maki domains.
Demonstrate how DeepRSM can be used for real-time air quality
management and policy-making.

RECOMMENDATION PROBLEM STATEMENT

Deep learning for prediction of the air quality response to emission changes Traditional air quality prediction models (e.g., CTMs, pf-RSMs) are
computationally expensive and time-consuming, making them
Integrate DeepRSM into real-time air quality monitoring systems.
Expand the dataset with more diverse emission scenarios. impractical for real-time policy decisions.
Combine DeepRSM with meteorological forecasting models for improved predictions. CTMs require 20+ simulations to capture the effects of emission
Develop user-friendly tools to assist policymakers in air quality management.
changes.
pf-RSM models are limited in handling non-linearity and require a
large number of training simulations.
A more efficient, accurate, and scalable method is needed to predict
PM2.5 and O3 responses to emission changes.

FINDINGS Data Collection

METHODS
Air Pollution Data: PM2.5 and O3 concentrations obtained from Community Multiscale Air Quality (CMAQ) CTM simulations.
Chemical Indicators: 18 key atmospheric chemical species were used to predict pollutant formation (e.g., NOx, VOCs, NH3, SO2).
Simulation Data:

DeepRSM is an efficient alternative to traditional air quality 40 emission control scenarios from CMAQ modeling.
Baseline and fully controlled emission conditions.

response models. 4 modeling domains covering China (CN27), Northern China Plain (NCP), Fen-Wei Plain (FWP), Chuan-Yu Region (CYR).
Meteorological Data: From the Weather Research and Forecasting (WRF) model for different seasons (January, April, July, October).

It achieves near CTM-level accuracy with only two simulations 5.2. Deep Learning Model - DeepRSM (CNN-Based Model)
A deep Convolutional Neural Network (CNN) was used to model the nonlinear relationship between air pollutant levels and emissions.

instead of 20+. CNN layers capture spatial and chemical interactions.

Residual blocks improve predictive accuracy and computational efficiency.

The model generalizes well across different seasons and spatial LeakyReLU activation function ensures better gradient flow.
5.3. Performance Evaluation Metrics

domains. Root Mean Square Error (RMSE)

Mean Absolute Error (MAE)
Coefficient of Determination (R²)
CNN-based deep learning is effective in predicting air quality 5.4. Benchmark Comparisons
pf-RSM (Polynomial Response Surface Model)
responses to emission changes. CTM (Full-scale Chemical Transport Model simulations)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

DeepRSM achieves accuracy close to CTM but requires only two simulations, compared to 20+ for pf-RSM models.

CNN training required extensive computational resources. DeepRSM outperforms pf-RSM in capturing nonlinear effects.
CNN-based DeepRSM generalizes well across different seasons and spatial domains.
Limited availability of high-resolution real-world air pollution data. DeepRSM performs well across all regions with RMSE values consistently lower than pf-RSM.
Transferability: DeepRSM generalizes across different locations without retraining.
DeepRSM required fine-tuning in regions with unique pollution sources. DeepRSM successfully captures the nonlinear response of air quality to emissions.
CNN architecture effectively models complex chemical reactions.
The model's accuracy does not degrade significantly when applied to new spatial domains.
Fine-tuning DeepRSM with additional training data improves accuracy.
 ABSTRACT
High-resolution aerosol optical depth (AOD) retrieval is crucial for monitoring air
pollution and climate change. Traditional retrieval methods using radiative transfer
models require assumptions about surface reflectance, making them less accurate.
DATA This study proposes a deep neural network (DNN)-based AOD retrieval algorithm
AIM OF STUDY
using Landsat-8 top-of-atmosphere (TOA) reflectance data. The model was trained
Landsat-8 TOA Reflectance Data (2013–2020) and validated using 8 years (2013–2020) of Landsat-8 data and AERONET AOD To develop a high-spatial-resolution AOD retrieval algorithm
AERONET AOD Measurements (329 stations) measurements from 329 stations worldwide. A total of 6,390 samples were used for based on deep neural networks (DNNs) using Landsat-8 TOA
ERA5 Reanalysis Meteorological Data model training and validation. The study demonstrates that the DNN model can
Digital Elevation Model (DEM) Data retrieve AOD with high accuracy, achieving a correlation coefficient of 0.83, RMSE of
reflectance data, overcoming the limitations of traditional
0.15, and 61% of retrievals within ±(0.05 + 20% AOD) of AERONET AOD values. radiative transfer-based AOD retrieval methods.

SIGNIFIICANCE OBJECTVIES
Develop a DNN-based AOD retrieval model for high-resolution aerosol
monitoring using Landsat-8 TOA data.
Revolutionizes AOD retrieval using deep learning.
Validate the model using collocated AERONET AOD ground measurements.
Achieves high accuracy while reducing computational cost. Compare the DNN method with traditional radiative transfer models to
Supports high-resolution air pollution and climate research. evaluate accuracy and efficiency.
Assess the model’s adaptability to Sentinel-2 TOA data, which has a different
spectral band configuration.
Utilize cloud computing (Google Earth Engine - GEE) to efficiently process
large volumes of Landsat data.

RECOMMENDATION PROBLEM STATEMENT

A Deep-Neural-Network-Based Aerosol Optical Depth (AOD) Retrieval raditional radiative transfer model-based AOD retrieval methods suffer
from the following challenges:
Expand the training dataset with more AERONET observations. from Landsat-8 Top of Atmosphere Data Require assumptions on surface reflectance, leading to errors in
Adapt model for Sentinel-2 data, which has different spectral bands. urban and heterogeneous land cover areas.
Integrate cloud-detection algorithms to improve accuracy.
Limited spatial resolution (≥1 km) in existing satellite-based AOD
Develop real-time AOD monitoring systems using cloud computing.
products (e.g., MODIS, MISR, Himawari-8).
Computationally expensive and complex, making them unsuitable for
real-time applications.
Thus, an alternative deep learning approach is proposed to directly
learn the relationship between TOA reflectance and AOD without
requiring surface reflectance assumptions.

FINDINGS 5.1. Data Collection

Landsat-8 TOA Reflectance Data (2013–2020)
METHODS
AERONET AOD Data from 329 global sites
Meteorological Data from ERA5 reanalysis (water vapor, ozone concentration)
Digital Elevation Model (DEM) for terrain correction

DNN-based AOD retrieval is a viable alternative to traditional

5.2. Data Processing
Google Earth Engine (GEE) was used for large-scale data collation and pre-processing.
A total of 6,390 collocated Landsat-8 and AERONET AOD samples were used for training and validation.

radiative transfer models. 5.3. Deep Neural Network (DNN) Model Architecture
Input Features (17 variables):
7 Landsat-8 TOA reflectance bands
High spatial resolution (30m) enables detailed urban aerosol 2 TOA brightness temperature (BT) bands
Solar zenith and azimuth angles

monitoring.
Viewing zenith and azimuth angles
Scattering angle
Digital Elevation Model (DEM)

Google Earth Engine (GEE) facilitates large-scale model Total column water vapor & ozone concentration
DNN Structure:
Three hidden layers (256, 512, 512 neurons)

deployment. LeakyReLU activation function

Mini-batch gradient descent (batch size: 256, epochs: 200)
Batch normalization to avoid overfitting
5.4. Model Evaluation
Leave-One-Station-Out Cross-Validation
Performance Metrics:
Correlation coefficient (R)
Root Mean Square Error (RMSE)
Mean Absolute Error (MAE)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

DNN AOD estimates closely match AERONET AOD values.

Some errors in AOD retrieval for regions with high aerosol loadings. RMSE and MAE values indicate high retrieval accuracy.
DNN outperforms traditional AOD retrieval methods (DT, DB, MAIAC).
Cloud contamination affects accuracy despite pre-processing steps. RMSE improvement of 25-30% over DT and DB methods.
DNN model effectively captures complex atmospheric relationships without requiring assumptions on surface reflectance.
Limited availability of high-resolution training data for some regions. Higher accuracy than traditional models when applied to heterogeneous urban regions.
Model generalization is robust across different geographic locations and seasons.
 ABSTRACT
PM2.5 is a critical air pollutant that affects human health and climate. Satellite-
derived aerosol optical depth (AOD) is widely used for estimating PM2.5
concentrations, but existing models have significant gaps in spatiotemporal coverage
DATA due to cloud cover, haze, and nighttime limitations. This study develops a deep
AIM OF STUDY
spatiotemporal neural network (ST-NN) to improve PM2.5 estimation by integrating
PM2.5 Concentrations (2017–2020): CNEMC ground station data. satellite AOD, meteorological data, and geospatial information. The model is trained To develop a deep learning model (ST-NN) capable of
AOD Data: MODIS (3 km), Himawari-8 (0.05°). on four years of data (2017–2020) from central and eastern China and achieves high- accurately estimating hourly PM2.5 concentrations across
Meteorological Data: ERA5 reanalysis. resolution (0.01°) hourly PM2.5 estimates, even under cloudy or hazy conditions.
Geospatial Data: Digital Elevation Model, land cover. Cross-validation results show an R² of 0.8–0.9 and an RMSE of 6–26 µg/m³,
China, overcoming limitations in satellite-based monitoring
demonstrating significant improvement over traditional methods. due to cloud cover, haze, and nighttime gaps.

SIGNIFIICANCE OBJECTVIES
Develop a deep spatiotemporal neural network (ST-NN) for PM2.5 prediction.
Integrate multi-source data (satellite AOD, meteorological variables,
Improves air pollution monitoring in areas with no ground
geospatial information) to improve estimation accuracy.
sensors. Fill observational gaps in satellite-derived AOD to provide full-coverage
Supports policy-making and public health decisions. PM2.5 monitoring.
Enhances real-time PM2.5 forecasting capabilities. Evaluate model performance using cross-validation and compare with
traditional methods.
Quantify the impact of different input variables on PM2.5 prediction using
sensitivity analysis.

RECOMMENDATION PROBLEM STATEMENT

A Deep Learning Approach to Increase the Value of
Ground-based PM2.5 monitoring stations are unevenly distributed,
Expand training data with more ground-based PM2.5 measurements. Satellite Data for PM2.5 Monitoring in China making it difficult to track air pollution in remote areas.
Adapt the model for global applications beyond China. Satellite AOD retrieval is hindered by cloud cover, haze, and
Integrate ST-NN with chemical transport models for hybrid forecasting. nighttime conditions, leading to missing data.
Deploy ST-NN in real-time air quality monitoring systems.
Traditional regression and machine learning models fail to accurately
estimate PM2.5 under challenging conditions.
A deep learning-based approach is needed to address these gaps
and improve spatiotemporal PM2.5 estimates.

FINDINGS 5.1. Data Collection

METHODS
PM2.5 Concentration Data: Hourly PM2.5 measurements from the China National Environmental Monitoring Center (CNEMC).
Satellite AOD Data:
MODIS 3 km aerosol product
Himawari-8 AOD (0.05° × 0.05° resolution)

Deep learning significantly improves PM2.5 estimation accuracy

Meteorological Data: ERA5 reanalysis (temperature, wind speed, humidity, pressure).
Geospatial Data: Digital Elevation Model (DEM), land use, population density.
5.2. Deep Spatiotemporal Neural Network (ST-NN) Model

compared to traditional models. Input Features (15 variables):

AOD values (MODIS, Himawari-8)
Meteorological variables (wind speed, temperature, humidity)
ST-NN successfully fills observational gaps in satellite-based Geospatial features (altitude, land cover, population density)
Time-based features (seasonality, day/night cycles)

monitoring.
ST-NN Architecture:
Convolutional Neural Network (CNN): Extracts spatial features.
Recurrent Neural Network (RNN) - LSTM: Captures temporal dependencies.

It enables real-time, full-coverage PM2.5 mapping across China. Attention Mechanism: Assigns weights to key predictors.
Dropout Regularization: Prevents overfitting.
5.3. Model Training and Validation
Data split: 70% training, 20% validation, 10% testing.
Loss function: Mean Squared Error (MSE).
Optimizer: Adam with a learning rate of 0.001.
Evaluation Metrics:
Coefficient of Determination (R²)
Root Mean Square Error (RMSE)
Mean Absolute Error (MAE)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

ST-NN significantly outperforms traditional models, achieving higher R² and lower RMSE.

Errors increase in areas with complex terrain (mountains, coastlines). It accurately estimates PM2.5 even in areas with limited monitoring data.
ST-NN provides full-coverage hourly PM2.5 estimates, unlike traditional satellite-based methods that fail under cloudy or nighttime
Limited ground truth data affects accuracy in remote regions. conditions.
It matches or exceeds the accuracy of chemical transport models but requires much less computation.
ST-NN requires significant computational resources for training AOD is the most important predictor of PM2.5, followed by meteorological factors (humidity, wind speed).
ST-NN successfully reconstructs PM2.5 levels even when AOD data is missing.
High-resolution predictions (0.01°) enable localized pollution analysis.
 ABSTRACT
Atmospheric chemistry transport models (CTMs) have been widely used for regional
aerosol forecasting but face challenges due to uncertainties in emission rates,
meteorological data, and oversimplified chemical parameterizations. This study
DATA introduces a spatiotemporal deep learning framework named Pollution-Predicting Net
AIM OF STUDY
(PPN) for PM2.5 to enhance the accuracy and efficiency of regional PM2.5
PM2.5 concentration data from China's national monitoring network. concentration forecasting. The encoder-decoder model integrates previous PM2.5 To develop and validate a deep learning-based regional
Meteorological variables from WRF model simulations. observations with numerical weather prediction (NWP) data, applying a weighted loss aerosol forecasting model (PPN) that integrates numerical
Emission data from MEIC inventory. function to improve forecasting during extreme pollution events. The study evaluates
the model over the Beijing-Tianjin-Hebei (BTH) region for 3-day PM2.5 predictions,
weather prediction (NWP) data and previous PM2.5
achieving R² = 0.7 and RMSE = 17.7 µg/m³. The PPN model outperforms the WRF- observations for improved short-term PM2.5 forecasts.
Chem model (a state-of-the-art numerical model with data assimilation), particularly
within the first 24 hours of forecasting.

SIGNIFIICANCE OBJECTVIES
Develop a deep learning-based model (PPN) for short-term (0–72 hours)
PM2.5 prediction.
Reduces computational cost compared to traditional numerical
Incorporate numerical weather prediction (NWP) data and PM2.5
models. observations into an encoder-decoder framework.
Enhances real-time pollution forecasting accuracy. Apply a weighted loss function to improve forecasting performance during
Supports early warning systems for extreme pollution events. extreme pollution events.
Compare model performance with WRF-Chem simulations and assess
improvements.
Analyze the influence of different input variables on model accuracy.

RECOMMENDATION PROBLEM STATEMENT

Regional aerosol forecasts based on Traditional numerical models (e.g., WRF-Chem, CMAQ, GEOS-Chem)
have limitations due to:
Expand training datasets with more meteorological and pollution data. deep learning and numerical weather prediction Uncertainties in emissions and meteorology.
Enhance long-term forecasting capabilities by integrating additional spatial transport Over-simplified chemical parameterizations.
variables. High computational costs (several hours per simulation day).
Develop hybrid models combining PPN with chemical transport models.
Existing machine learning (ML) models struggle with capturing
Deploy real-time PM2.5 forecasting systems for public health applications.
spatiotemporal relationships, limiting their effectiveness in regional
air quality forecasting.
A hybrid deep learning approach is needed to integrate CTM,
meteorology, and ML techniques for more accurate and efficient
PM2.5 predictions.

FINDINGS 5.1. Data Collection

METHODS
PM2.5 Concentration Data: Hourly PM2.5 observations from national air quality monitoring stations in the Beijing-Tianjin-Hebei (BTH) region.
Meteorological Data:
Collected from the Weather Research and Forecasting (WRF) model.
Includes temperature, wind speed, humidity, precipitation, planetary boundary layer height (PBLH), sea level pressure (SLP), and geopotential height at 700 hPa.
PPN outperforms traditional numerical models (WRF-Chem) for Emission Data: From the Multi-resolution Emission Inventory for China (MEIC).
5.2. Model Architecture - Pollution-Predicting Net (PPN)

short-term PM2.5 forecasting.

Encoder-Decoder Framework:
Encoder processes historical PM2.5 observations and meteorological variables.
Decoder predicts future PM2.5 concentrations.

Deep learning can effectively replace computationally expensive Spatiotemporal Feature Extraction:
Uses convolutional layers for spatial relationships.
Uses long short-term memory (LSTM) layers for temporal dependencies.
CTMs for rapid air quality predictions. Weighted Loss Function:
Reduces overestimation of low values and underestimation of high values in extreme pollution cases.

Weighted loss function improves extreme pollution event

5.3. Model Training and Evaluation
Training Period: January 2022.
Training Method:

forecasting. Adam optimizer with learning rate 0.001.

Mean Squared Error (MSE) loss function with weighting adjustments for extreme events.
Performance Metrics:
R² (coefficient of determination).
Root Mean Square Error (RMSE).

ISSUES & DIFFICULTIES ANALYSIS RESULTS

PPN outperforms WRF-Chem, particularly in short-term (0–24 h) forecasts.

Computational resources: Training deep learning models requires RMSE is reduced by ~2 µg/m³ compared to WRF-Chem.
PPN effectively captures high PM2.5 concentrations in southern BTH while showing lower values in northern regions.
significant GPU/TPU processing power. Impact of Weighted Loss Function:
Improves accuracy during extreme pollution events.
Forecast degradation beyond 72 hours: Accuracy declines as forecast Reduces systematic biases (e.g., overestimating low values, underestimating high values).
PPN consistently outperforms WRF-Chem at all forecast lead times.
lead time increases. Greatest improvement is observed within the first 24 hours.
PPN successfully integrates deep learning with numerical weather prediction (NWP).
Inter-regional transport limitations: The model does not explicitly It generalizes well across different pollution levels and meteorological conditions.
Weighted loss function significantly improves extreme pollution forecasts.
capture long-range pollution transport events.
 ABSTRACT
Accurate air pollutant prediction is critical for environmental management and public
health. Traditional deterministic and statistical methods have large prediction errors
and struggle to capture the complex temporal dependencies in air pollution time
DATA series data. This study proposes the Autocorrelation Error-Informer (AE-Informer)
AIM OF STUDY
model, which builds on the Informer deep learning model by incorporating
Air Quality Data (2019–2020): China Environmental Monitoring Centre. autocorrelated error correction. The AE-Informer model is applied to predict hourly To develop and evaluate the AE-Informer deep learning
Meteorological Data: Not explicitly mentioned but assumed from standard concentrations of PM10, PM2.5, NO2, and O3 in Henan Province, China. Experimental model for accurate air pollutant prediction by leveraging
sources. results show that AE-Informer reduces prediction errors by 3% compared to the
original Informer model, and outperforms ARIMA, BiLSTM, GRU, and LSTM models.
autocorrelation errors and a self-attention-based Informer
Additionally, a stacking ensemble model is used to fill missing time series data, framework.
further improving prediction performance.

SIGNIFIICANCE OBJECTVIES
Develop the AE-Informer model, integrating autocorrelation error correction
into the Informer architecture.
Improves air pollution forecasting accuracy for Henan Province.
Predict hourly air pollutant concentrations (PM10, PM2.5, NO2, O3) using
Provides a deep learning-based solution to supplement missing deep learning.
air quality data. Compare AE-Informer’s performance with traditional models (ARIMA,
Supports environmental policymakers in developing effective BiLSTM, GRU, LSTM).
pollution control strategies. Use a stacking ensemble method to supplement missing air quality time
series data.
Evaluate model effectiveness across multiple monitoring stations in Henan
Province.

RECOMMENDATION PROBLEM STATEMENT

Improving air pollutant prediction in Henan Province Traditional methods (e.g., ARIMA, regression models) assume linear
relationships and cannot accurately predict nonlinear pollution
Use transfer learning to apply AE-Informer in other regions beyond Henan Province. , China, by enhancing the concentration trends.
Optimize model hyperparameters to further reduce computational costs. prediction accuracy Deep learning models (RNNs, LSTMs) have limitations in handling
Integrate additional meteorological variables (e.g., wind direction, humidity) for better long-range dependencies and autocorrelation errors in air pollutant
predictions. using autocorrelation errors and an Informer deep learning model time series.
Enhance extreme pollution event forecasting by incorporating real-time satellite data.
Missing data in pollution monitoring records reduces the accuracy of
predictions

FINDINGS 5.1. Data Collection

METHODS
Study Area: Henan Province, China.
Time Period: January 1, 2019 – December 31, 2020.
Air Pollutant Data: Hourly PM10, PM2.5, NO2, O3 concentrations from 60 monitoring stations.
Data Source: China Environmental Monitoring Centre (CEMC).
AE-Informer improves pollutant prediction accuracy by 3% 5.2. Model Framework
5.2.1. Informer Model (Baseline)

compared to the standard Informer model.

An advanced Transformer-based time series forecasting model.
Uses a self-attention mechanism to efficiently capture long-term dependencies.
5.2.2. AE-Informer Model (Proposed Model)

It significantly outperforms traditional time series models like Enhances Informer by incorporating autocorrelation error correction (AE).
Key Improvement: Instead of predicting pollutant concentrations directly, AE-Informer models prediction errors, reducing biases from time series autocorrelation.
5.2.3. Stacking Ensemble Model for Missing Data Imputation
ARIMA and BiLSTM. Combines multiple machine learning models to fill gaps in pollutant concentration records.
Base Learners:

The stacking ensemble method enhances missing data imputation,

XGBoost, LightGBM, GBDT, Random Forest (RF), and Extra Trees (ET).
Final Meta-Model: Weighted combination of the base models.
5.3. Model Training and Validation

reducing RMSE. Training Data: 80%

Testing Data: 20%
Loss Function: Mean Squared Error (MSE)
Optimizer: Adam
Batch Size: 512
Evaluation Metrics:
Coefficient of Determination (R²)
Root Mean Square Error (RMSE)
Mean Absolute Error (MAE)

ISSUES & DIFFICULTIES ANALYSIS RESULTS

AE-Informer achieved the highest R² (0.976), meaning it captured the most variance in pollution trends.

Computational cost: Informer models require large memory and GPU MAE and RMSE were reduced by 3–7% compared to other deep learning models.
AE-Informer outperformed ARIMA by a significant margin (13% lower MAE/RMSE)

resources. Highest accuracy for O3 and PM10 predictions.

Lower errors for PM2.5 and NO2 compared to existing models.
Extreme event forecasting: AE-Informer still struggles with predicting AE-Informer successfully corrects for autocorrelated errors, leading to more accurate time series predictions.
Stacking ensemble learning significantly improves missing data imputation, leading to better forecasts.
sudden pollution spikes. Informer-based models outperform traditional deep learning approaches (LSTM, GRU, CNN) in long-term forecasting.

Data quality: Missing or inconsistent monitoring station data affects

accuracy.
 ABSTRACT
Monitoring fine particulate matter (PM2.5) is essential for assessing air quality and its
impact on public health. However, the sparse distribution of ground-based monitoring
stations limits the ability to capture pollution levels comprehensively. This study
presents a two-step deep learning approach to estimate daily gap-free surface PM2.5
DATA concentrations across the contiguous United States (CONUS) from 2018 to 2022 at a
4 km spatial resolution.
AIM OF STUDY
Step 1: A Depthwise-Partial Convolutional Neural Network (DW-PCNN) is used to
Ground-Based PM2.5 Data: U.S. EPA Air Quality System (AQS)
fill gaps between PM2.5 monitoring stations using MODIS Aerosol Optical Depth To develop a deep learning framework that integrates
Satellite Data: MODIS AOD
(AOD) data. satellite-derived AOD data, meteorological variables, and
Meteorological Data: ERA5 Reanalysis
Step 2: The imputed PM2.5 grids are integrated with meteorological and anthropogenic factors to produce high-resolution, gap-free
Land-Use Data: Urbanization, road density, vegetation cover
anthropogenic variables into a Deep Convolutional Neural Network (Deep-CNN) to
enhance accuracy.
daily PM2.5 estimates across the U.S., thereby enhancing
Performance: The model achieves a Pearson’s correlation coefficient (R) of 0.92 spatial air quality assessments.
and an Index of Agreement (IOA) of 0.96 in a ten-fold cross-validation.
Additionally, the study introduces a grid-based method for calculating PM2.5
Design Values (DV), improving upon traditional station-based assessments.

SIGNIFIICANCE OBJECTVIES
Develop a deep learning-based PM2.5 estimation model using MODIS AOD,
meteorological, and land-use data.
Enhances spatial resolution of air quality data, benefiting
Reduce spatial gaps in PM2.5 monitoring by leveraging satellite-derived data.
environmental health research. Improve model accuracy over traditional statistical and chemical transport
Addresses limitations of traditional PM2.5 monitoring networks. models (CTMs).
Provides a scalable, AI-driven solution for real-time pollution Provide a comprehensive, high-resolution dataset for air quality assessments
assessments. and policy applications.
Validate the model's performance using a ten-fold cross-validation strategy

RECOMMENDATION PROBLEM STATEMENT

A Deep Learning Framework for Satellite-Derived Surface Ground-based monitoring networks are sparsely distributed, leading
Expand the dataset by integrating additional satellite products (e.g., Sentinel-5P, VIIRS PM2.5 2 Estimation: Enhancing Spatial Analysis in the United States to incomplete pollution assessments, particularly in rural and remote
AOD). regions.
Enhance cloud correction techniques to reduce missing data issues. Satellite-based PM2.5 estimates suffer from missing data due to
Develop real-time air quality monitoring tools using the proposed model. cloud cover and retrieval errors.
Extend the approach globally to assess air pollution in other countries. Traditional chemical transport models (CTMs) have high
computational costs and often struggle with bias and inaccuracies.
A deep learning-based solution is needed to integrate multi-source
data and generate more accurate, high-resolution PM2.5 estimates.

FINDINGS 5.1. Data Collection

METHODS
Study Period: 2018–2022
Study Area: Contiguous United States (CONUS)
Key Data Sources:
PM2.5 Ground Observations: U.S. Environmental Protection Agency (EPA) Air Quality System (AQS)
Deep learning models can effectively downscale satellite data, Aerosol Optical Depth (AOD): MODIS satellite data
Meteorological Data: ERA5 Reanalysis (temperature, wind speed, precipitation, boundary layer height, etc.)

providing gap-free, high-resolution PM2.5 estimates.

Land-Use Data: Urbanization indices, road density, vegetation cover
5.2. Deep Learning Model Architecture
Step 1: Depthwise-Partial Convolutional Neural Network (DW-PCNN)

Grid-based DV calculations provide a more complete assessment Fills missing data in PM2.5 observations using MODIS AOD and auxiliary data.
Step 2: Deep Convolutional Neural Network (Deep-CNN)
Refines and enhances PM2.5 estimates by incorporating additional meteorological and land-use variables.
compared to traditional station-based methods. Loss Function: Mean Squared Error (MSE)
Optimizer: Adam

The model offers a scalable solution for real-time air quality

Evaluation Metrics: Pearson’s R, RMSE, IOA
5.3. Model Validation
Ten-fold cross-validation to evaluate robustness.

monitoring and policy applications. Comparison with traditional models, including statistical regression and CTMs.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Deep learning model significantly outperforms traditional methods in estimating PM2.5.

Uncertainty in AOD-PM2.5 relationships in some regions. Reduction in RMSE (~30% lower) compared to traditional regression and CTMs.
6.2. Grid-Based PM2.5 Design Values (DV) Approach

Cloud cover and extreme weather conditions affect MODIS AOD Traditional methods rely on station-based DVs, leading to incomplete coverage.
This study introduces a continuous, high-resolution grid-based DV calculation.
retrieval. Improved identification of pollution hotspots, benefiting public health assessments
The DW-PCNN effectively reconstructs missing PM2.5 values, reducing biases in satellite-based data.
Computational costs associated with training deep learning models. The Deep-CNN model captures complex non-linear relationships between PM2.5, meteorology, and land use.
High spatial resolution (4 km) improves policy and research applications.
 ABSTRACT
his report presents the outcomes of a project supported by the Texas Commission on
Environmental Quality (TCEQ), focusing on estimating daily surface PM2.5
concentrations at a 4 km spatial resolution across Texas from 2018 to 2022 using an
advanced deep learning (DL) model.
DATA Key Contribution: Development of gap-free, high-resolution PM2.5 estimates,
addressing missing data challenges in satellite imagery.
AIM OF STUDY
Problem Addressed: Satellite-derived Aerosol Optical Depth (AOD) often has
Ground PM2.5 Data: U.S. EPA AQS
gaps and does not always correlate well with surface PM2.5 levels. To develop a deep learning-based PM2.5 estimation model
Satellite Data: MODIS AOD (Terra/Aqua)
Solution: A two-phase deep learning model combining a Deep Convolutional using satellite-derived aerosol data, meteorological
Meteorological Data: ERA5 Reanalysis
Neural Network (DeepCNN) with a Depthwise Partial Convolutional Neural conditions, and land-use information, thereby providing high-
Land-Use Data: Urbanization, vegetation cover, road density
Network (PCNN) for more accurate PM2.5 predictions.
Evaluation: The model was validated against ground-based monitoring stations,
resolution, reliable PM2.5 maps for Texas.
achieving high accuracy with a correlation coefficient (R) of 0.89–0.93 and Index of
Agreement (IOA) of 0.94–0.96.

SIGNIFIICANCE OBJECTVIES
Develop an AI-based model that enhances PM2.5 estimation accuracy using
AOD and meteorological variables.
Enhances air quality monitoring with high-resolution data.
Address missing data problems in satellite observations through deep
Addresses limitations of ground-based monitoring networks. learning techniques.
Provides an AI-powered tool for air quality policymaking. Improve air quality monitoring and exposure assessment in Texas, especially
in areas with limited ground-based stations.
Validate the model's predictions against existing monitoring networks using
statistical performance metrics.
Expand the model’s applicability for broader air quality policy and public
health applications

RECOMMENDATION PROBLEM STATEMENT

Estimating Surface PM2.5 from Traditional air pollution monitoring relies on ground stations, which
Expand the dataset by integrating additional satellite products (e.g., Sentinel-5P, VIIRS Satellite Aerosols Proposal for have limited spatial coverage.
AOD). Grant Activities/PCR Number: 582-23-41321-025 Satellite AOD data suffers from retrieval gaps due to cloud cover,
Improve cloud correction algorithms for better AOD retrieval. instrument limitations, and uncertainties in AOD-PM2.5
Deploy AI-based PM2.5 estimation in real-time air quality monitoring systems. Grant Number: 582-18-81339 Grant Name: University of Houston relationships.
Extend the model beyond Texas for nationwide applications. Existing models, such as chemical transport models (CTMs), are
computationally expensive and have inherent biases.
A more robust, AI-driven solution is needed to provide accurate,
spatially continuous PM2.5 estimates.

FINDINGS 5.1. Data Collection

METHODS
Study Area: Texas, U.S.
Study Period: 2018–2022
Data Sources:
PM2.5 Ground Observations: EPA's Air Quality System (AQS).
AI-driven models can significantly enhance satellite-based PM2.5 Satellite Data: MODIS AOD from Terra/Aqua satellites.
Meteorological Data: ERA5 Reanalysis (temperature, wind speed, humidity, etc.).

monitoring.
Land-Use Data: Urbanization indices, road networks, vegetation cover.
5.2. Deep Learning Model Framework
Step 1: Depthwise Partial Convolutional Neural Network (PCNN)

Deep learning models outperform traditional regression and CTMs. Handles missing data issues in AOD retrievals.
Creates complete, daily PM2.5 grids at 4 km resolution.
Step 2: Deep Convolutional Neural Network (DeepCNN)
Grid-based PM2.5 DV calculations provide more comprehensive air Enhances PM2.5 estimation by incorporating meteorological and land-use factors.
Corrects residual biases from AOD-based estimates.

quality assessments.
Loss Function: Mean Squared Error (MSE)
Optimizer: Adam
Evaluation Metrics: Pearson’s R, RMSE, IOA
5.3. Model Validation
Ten-fold cross-validation to test robustness.
Comparison with traditional models such as multiple linear regression and chemical transport models.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Deep learning model significantly outperformed traditional methods.

Uncertainty in AOD-PM2.5 relationships in some regions. Gap-filling with PCNN reduced satellite retrieval errors.
DeepCNN improved spatial consistency in PM2.5 estimations.

Cloud cover and extreme weather conditions affected MODIS AOD 6.2. Spatial Cross-Validation Results
R values ranged from 0.81–0.93 across different Texas regions.
retrievals. The model accurately captured pollution hotspots near urban centers and industrial zones.
6.3. Design Value (DV) Tool
Computational demands for training deep learning models. A novel grid-based PM2.5 DV calculation was introduced.
Improved assessment of long-term air quality trends compared to traditional station-based approaches.
PCNN successfully reconstructed missing PM2.5 data, overcoming satellite AOD retrieval limitations.
DeepCNN effectively captured complex relationships between PM2.5, meteorology, and land-use.
High-resolution estimates (4 km) improve exposure assessments for public health studies.
 ABSTRACT
This study presents a deep learning-based post-process correction approach for
improving aerosol parameter retrievals from the Sentinel-3 Level-2 Synergy product.
The Sentinel-3 Synergy product provides aerosol retrievals at a high spatial resolution
(300 m) but suffers from low accuracy due to uncertainties in input parameters and
DATA approximations in retrieval algorithms. To address this, the researchers employ a
machine-learning-based correction model, which significantly improves the accuracy
AIM OF STUDY
of aerosol optical depth (AOD) retrievals while maintaining the high spatial
Satellite Data: Sentinel-3 Level-2 Synergy aerosol product.
resolution. To develop a deep-learning-based post-process correction
Ground-Based Observations: AERONET aerosol measurements.
The correction is based on a feed-forward neural network trained on collocated model that enhances the accuracy of aerosol retrievals from
Sentinel-3 and AERONET aerosol data. Sentinel-3 while maintaining its high spatial resolution.
Compared to the official Synergy product, the post-process-corrected model
achieves a 9% improvement in R², an 8% reduction in RMSE, and a 20% decrease
in bias.
The study also compares post-process correction with a fully learned machine
learning approach, showing that the hybrid approach performs better in
generalization to different aerosol conditions.

SIGNIFIICANCE OBJECTVIES
Improve the accuracy of Sentinel-3 Synergy aerosol retrievals using a
machine-learning-based correction model.
Improves the accuracy of Sentinel-3 aerosol retrievals without
Compare the post-process correction approach with fully learned regression
requiring full algorithm reprocessing. models.
Enhances spatial resolution of air quality data, enabling Enhance AOD estimates by leveraging Sentinel-3 and AERONET collocated
localized pollution assessments. data.
Supports climate research and atmospheric modeling by Ensure the corrected model generalizes well across various geographical
providing more reliable aerosol estimates. locations.
Assess the feasibility of using deep learning to correct satellite retrieval errors
without requiring reprocessing of the retrieval algorithms.

RECOMMENDATION PROBLEM STATEMENT

Deep-learning-based post-process correction of the Conventional satellite aerosol retrievals suffer from low accuracy
Expand training datasets by incorporating additional aerosol data sources. aerosol parameters in the high-resolution Sentinel-3 Level-2 Synergy product due to approximations in radiative transfer models and uncertainties
Develop real-time correction pipelines for operational air quality monitoring. in auxiliary parameters.
Improve handling of cloud-contaminated pixels in Sentinel-3 retrievals. Reprocessing satellite retrieval algorithms is computationally
Apply the approach to other satellite datasets such as MODIS and VIIRS expensive and often not feasible for real-time applications.
A hybrid approach is needed that combines physics-based retrieval
algorithms with machine-learning-based error corrections.
The Sentinel-3 Synergy aerosol product has high spatial resolution
(300 m) but poor accuracy, necessitating an effective correction
strategy.

FINDINGS METHODS
5.1. Data Collection
Satellite Data: Sentinel-3 Level-2 Synergy aerosol product.
Ground-Truth Data: AERONET (Aerosol Robotic Network) ground-based measurements.
Study Regions: Five selected regions with diverse aerosol conditions.
Post-process correction is an effective and computationally efficient 5.2. Deep Learning Model Framework
Post-Process Correction Model:
approach for improving satellite aerosol retrievals. A feed-forward neural network trained to correct errors in Sentinel-3 Synergy AOD estimates.
The model learns the approximation error between Sentinel-3 retrievals and AERONET ground-truth data.

Hybrid models outperform purely data-driven approaches by Fully Learned Machine Learning Model:
A separate regression model trained to predict AOD directly from Sentinel-3 Level-1 data.

incorporating physics-based information. 5.3. Model Training and Evaluation

Training Data: Sentinel-3 Synergy and AERONET data from selected regions.
Validation Approach: Models tested on independent datasets outside training regions.
The corrected Sentinel-3 Synergy product enables high-resolution Loss Function: Mean Squared Error (MSE).
Optimization Algorithm: Adam.
aerosol monitoring, benefiting both air quality studies and climate Performance Metrics:
R² (Coefficient of Determination)

research. RMSE (Root Mean Square Error)

Bias Reduction

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The post-process correction approach significantly improved AOD accuracy compared to the original Sentinel-3 Synergy product.

Limited AERONET data coverage affected model training in some Bias was reduced by 20%, and RMSE was 8% smaller in the corrected model.
The post-process model outperformed the fully learned model, especially in high AOD conditions.

regions. 6.2. Generalization to Different Regions

The post-process correction approach generalized well to new geographical locations, unlike the fully learned model, which struggled
Machine learning models require extensive computational resources. with out-of-sample predictions.
6.3. Improvement in High-Resolution Aerosol Monitoring
Cloud contamination in Sentinel-3 data introduced uncertainties in some The corrected Sentinel-3 data enabled more accurate fine-scale aerosol mapping, potentially allowing for street-level pollution
monitoring.
retrievals.
The post-process correction model effectively reduces approximation errors while maintaining the original physics-based retrieval
algorithm.
The hybrid approach provides a balance between computational efficiency and accuracy.
The fully learned model, while useful, lacks the ability to generalize well across diverse atmospheric conditions.
 ABSTRACT
This study explores the use of high-spatial-resolution satellite imagery and deep
learning models to generate hundred-meter-scale air quality (AQ) maps for urban
areas.
DATA Traditional air quality models rely on kilometer-scale data, which lacks the fine AIM OF STUDY
resolution needed for precise urban pollution mapping.
Satellite Imagery: High-resolution (2.5 m) commercial satellite images. The study develops a deep neural network (DNN) model trained on very high- To develop and validate a deep learning-based air quality
Air Quality Data: PM2.5 and NO2 ground station measurements. resolution (2.5 m) satellite imagery to estimate PM2.5 and NO2 concentrations.
estimation model that utilizes high-resolution satellite
Land-Use Data: Road networks, vegetation cover, industrial zones. Study Areas: London, Vancouver (BC), Los Angeles, and New York City.
The model achieves an RMSE < 2 µg/m³, highlighting green areas, roads, and imagery for fine-scale air pollution mapping in urban
urban density as key contributors to AQ variations. environments.
Results indicate strong potential for applying this method globally, particularly in
developing urban areas lacking ground monitoring stations.

SIGNIFIICANCE OBJECTVIES
Develop a deep learning model that estimates PM2.5 and NO2 levels based
Improves AQ mapping resolution from kilometers to 100 m solely on satellite imagery.
scale. Generate high-resolution (100–200 m scale) AQ maps for urban areas.
Provides a cost-effective alternative to ground-based Assess how different urban features (e.g., roads, green spaces) influence air
pollution.
monitoring stations.
Validate model predictions against ground monitoring station data and land-
Enables urban planners to design cleaner cities using high- use regression models.
resolution pollution data. Demonstrate the model’s applicability to other urban environments globally.

RECOMMENDATION PROBLEM STATEMENT

Mapping and Understanding Patterns of Air Quality Ground-based AQ monitoring networks are sparse, limiting real-time
Expand training datasets by integrating additional satellite data sources (e.g., Sentinel-2, Using Satellite Data and Machine Learning pollution assessments.
MODIS). Traditional satellite-derived AQ models operate at kilometer-scale
Develop real-time AQ monitoring applications using the trained model. resolution, missing local pollution hotspots.
Refine model calibration for extreme pollution events. Urban environments contain complex pollution patterns influenced
Apply this method to developing cities with poor AQ monitoring networks. by roads, vegetation, and industrial activity, requiring higher spatial
resolution modeling.
A deep learning-based solution that integrates high-resolution
imagery can provide better AQ estimates at finer spatial scales.

FINDINGS METHODS
5.1. Data Collection
Satellite Data: Very high-resolution (2.5 m) satellite images from commercial providers.
Ground-Based AQ Data: PM2.5 and NO2 measurements from government monitoring stations.
Land-Use Regression (LUR) Model Outputs: Used for additional validation.
Deep learning improves urban AQ estimation at a fine spatial scale. Study Cities: London, Vancouver, Los Angeles, and New York City.
5.2. Deep Learning Model Architecture
Road density and green spaces are critical variables in pollution Base Model: Modified VGG16 Convolutional Neural Network (CNN).
Input Data: Satellite imagery patches (40 × 40 pixels for PM2.5, 80 × 80 pixels for NO2).

modeling. Output Data: Estimated PM2.5 and NO2 concentrations at 100–200 m resolution.
Training Strategy:

The method can be applied globally, particularly in cities lacking AQ Stage 1: Model trained on London, Vancouver, and Los Angeles data.
Stage 2: Model tested on New York City (unseen during training) to evaluate generalization.
Loss Function: Mean Squared Error (MSE).
monitoring infrastructure. Optimizer: Adam.
Validation Metrics: Root Mean Square Error (RMSE), Normalized RMSE (NRMSE), and residuals.
5.3. Model Validation
Cross-validation with LUR-based estimates.
Comparison with annual ground monitoring data.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

Deep learning-based AQ maps significantly outperformed LUR models, especially in areas with limited ground monitoring.

Satellite images require preprocessing to remove cloud cover artifacts. The model successfully identified localized pollution sources, including road networks and green spaces.
6.2. Key Findings on Urban Features & Pollution

Model bias exists in extreme pollution events, requiring additional High AQ correlation with road density (higher NO2 near major highways).
Green areas reduce PM2.5 levels, demonstrating the role of urban vegetation in pollution control.
calibration. Industrial zones show significant pollution spikes, particularly in LA and NYC
The CNN model effectively captures spatial pollution patterns based on urban surface features.
Computational costs are high, limiting real-time deployment potential High-resolution satellite data enables detailed AQ mapping, previously unavailable from coarser-resolution satellite-based models.
Model generalization is strong, suggesting scalability to other cities.
 ABSTRACT
This study presents a novel deep learning framework for downscaling atmospheric
pollutant data to improve the spatial resolution of PM2.5 and NO2 estimates.
The method downscales coarse-resolution atmospheric data (40 km) to a fine
DATA resolution (1 km × 1 km). AIM OF STUDY
The framework integrates static and dynamic datasets, including elevation, land-
Ground Monitoring Data: EMEP stations (NO2, PM2.5). use data, and in situ measurements, to enhance prediction accuracy. To develop and evaluate a deep-learning-based data fusion
Satellite Data: Sentinel-5P (NO2), MODIS (AOD). A custom loss function is designed to handle ground station sparsity by prioritizing
approach that enhances the spatial resolution of atmospheric
Reanalysis Data: ERA5 (meteorological variables). regions with available station data while approximating model-based outputs
Model Data: CAMS (pollutant concentration estimates). elsewhere. pollutant data for better air quality monitoring and policy
Land Cover Data: CORINE Land Cover (CLC). The proposed method outperforms both traditional physics-based modeling applications.
Elevation Data: Digital Elevation Model (DEM). approaches and other deep-learning-based models, with improvements of up to
19% for NO2 and 13% for PM2.5.

SIGNIFIICANCE OBJECTVIES
Develop a deep learning model to improve the resolution of NO2 and PM2.5
mproves the spatial resolution of air quality estimates from 40 pollution data.
km to 1 km. Integrate multiple data sources, including satellite-based aerosol retrievals,
Provides a cost-effective alternative to high-resolution CTM elevation maps, and land-use information.
Design a custom loss function to address ground station measurement
simulations.
sparsity.
Enhances urban pollution assessments, supporting air quality Compare the proposed model’s performance against standard modeling-
management and policy decisions. based and deep-learning-based approaches.
Ensure model generalization across different land use types and geographical
regions.

RECOMMENDATION PROBLEM STATEMENT

AI for Air Quality: Leveraging Data Fusion for Traditional air quality models (e.g., CAMS, CTMs) operate at coarse
Extend the approach to other pollutants (e.g., O3, SO2). Deep Downscaling of Atmospheric Pollutants resolutions (40 km), which is insufficient for local-scale pollution
Increase the temporal resolution of meteorological inputs for better NO2 estimation. assessment.
Optimize model efficiency for real-time applications. Ground monitoring stations are sparse, making it difficult to train
Apply the model globally, particularly in developing regions with limited AQ monitoring. data-hungry deep learning models.
Satellite-based AOD retrievals are often incomplete, requiring robust
fusion techniques for accurate downscaling.
A novel deep learning approach is needed to fuse multiple data
sources and provide high-resolution air quality estimates.

FINDINGS Study Area: Northern Italy (500 × 500 km region).

METHODS
Data Sources:
Ground Station Data (EMEP): Daily PM2.5 and NO2 concentrations.
CAMS (Copernicus Atmospheric Monitoring Service): Coarse-resolution pollutant estimates (10 km).
CORINE Land Cover (CLC): High-resolution land-use data (100 m).
Multi-source deep learning models can significantly improve Sentinel-5P: NO2 tropospheric column densities (7 km).
Digital Elevation Model (DEM): 25 m resolution elevation data.

pollution mapping resolution. ERA5 Reanalysis: Meteorological variables (wind direction, temperature).
MODIS: Aerosol Optical Depth (AOD) at 1 km resolution.

Land-use data plays a crucial role in pollutant distribution modeling.

5.2. Deep Learning Model Architecture
Base Model: EfficientNet-B0 CNN architecture.
Custom Sampling Process: Balances land-use representation in training batches.
The proposed model generalizes well across different geographic Loss Function:
Heavily penalizes errors at ground station locations.

regions.
Uses CAMS as a fallback when ground station data is unavailable.
Training Process:
64 × 64 tiles are extracted from preprocessed data.
Data fusion enhances the accuracy of pollutant estimates at finer Ground station data is used as a priority target for loss minimization.
Optimization: Adam optimizer, RMSE-based loss.

spatial scales. 5.3. Model Validation

Comparison with Existing Methods:
Baseline models: CAMS, bilinear interpolation, standard CNN models.
Validation Metrics: RMSE, MAE, Mean Bias Error (MBE).
Cross-validation across different land-use types.

ISSUES & DIFFICULTIES ANALYSIS RESULTS

The proposed deep learning framework significantly outperformed traditional methods, particularly for NO2 downscaling.

Computational cost: Requires significant resources for training. Land-use data and DEM information contributed to accuracy improvements, demonstrating the effectiveness of multi-source data
fusion.

Uncertainty in AOD-PM2.5 relationships affects estimation quality. 6.2. Improvements Across Land-Use Types
The model performed well across different land-use categories, except for one industrial region where performance was similar to
Sparse ground station data limits direct supervision during training. CAMS.
Highest improvement observed in urban areas (up to 19% for NO2, 13% for PM2.5).
Deep learning successfully enhances pollutant resolution from 40 km to 1 km.
The custom sampling process prevents model biases toward specific land-use types.
The use of multiple temporal snapshots of ERA5 & CAMS improves prediction accuracy.
The model outperforms both standard physics-based and deep-learning-based approaches.