Annual particulate matter (PM2.5) concentrations in India averaged 50.6 micrograms per cubic meter of air (µg/m³) in 2024. While annual PM2.5 levels have fallen roughly 30 percent since 2018, they remain more than 10 times above the World Health Organization's recommended limit of five µg/m³.

In this study, Beijing, the capital of China, is selected as the study area. Hourly mean concentrations of six regulatory air pollutants including O3 (μg/m3), SO2 (μg/m3), NO2 (μg/m3), PM2.5 (μg/m3), PM10 (μg/m3), and CO (mg/m3) were collected from 35 air quality monitoring stations labeled by 1 to 35 from 01/01/2017 to 05/30/2018. The data was provided by the Ministry of Environmental Protection (MEP) of China. Hourly averaged meteorological data in the same period were first accessed from The National Oceanic and Atmospheric Administration (NOAA), then processed by the Weather Research and Forecasting (WRF) model to produce grid meteorological data (21×31 points) with a grid spacing of 5 km. Meteorological parameters including temperature, air pressure, relative humidity, wind speed, and wind direction are selected as the main meteorological features due to their close relationships with the change of ozone concentrations.

Noise pollution levels coming from airplanes in South Korea dropped a little more than one point from 2018 to 2022. This drop may partially be attributed to fewer flights being operated during the height of the COVID-19 pandemic in early 2020 into 2021.

Description

This dataset contains remote sensing data from the ESA Copernicus missions Sentinel-2 and Sentinel-5P (tropsopheric NO2 column density) in the 2018-2020 timespan. The satellite measurements each cover ~3100 locations in Europe and ~100 on the US Westcoast, each with a size of 1.2x1.2km. The locations are selected such that each measurement is centered at the location of an air quality measurement station on the ground (from the European Environment Agency or the US Environmental Protection Agency, measuring NO2). This makes it possible to analyze spatiotemporally aligned remote sensing and ground-based measurements.

The 13 Sentinel-2 bands are upsampled (bilinear) to 10m resolution and cropped to 120x120 pixel. For some locations multiple Sentinel-2 images are available. The images are stored as binary numpy .npy files organized into directories based on their locations.

The Sentinel-5P data was pre-processed by mapping the measurements from consecutive satellite overpasses onto a common rectangular grid of 0.05×0.05◦(∼5×5km) across Europe. To harmonize the Sentinel-2 (10m to 60m, upscaled to 10m) and Sentinel-5P (5×3.5km, rescaled to 5×5km) imaging resolutions, the Sentinel-5P data is linearly interpolated to 10m resolution and cropped to 120×120 pixel around the locations of interest. Additionally, all measurements with a QA flag (qa_value) below 75 were discarded, following ESA recommendations. The Sentinel-5P data are stored as .netcdf file, organized by location. For each location, three such files are available, containing averaged Sentinel-5P measurements at different temporal frequencies (2018-2020, quarterly, monthly).

The

samples_{frequency}_{area}.csv

files provide a list of observations with the corresponding file paths to a (cloud-free) Sentinel-2 image, the Sentinel-5P measurement, and the average NO2 concentration measurement by the EEA or EPA ground station. These files can be used for easy data-loading.

Content

The data is organized into the following files:

• README.md - this file
• sentinel-2-eea.tar.gz [33.1GB]
• sentinel-5p-eea.tar.gz [80.1GB]
• samples_2018_2020_eea.csv
• samples_quarterly_eea.csv
• samples_monthly_eea.csv
• sentinel-2-epa.tar.gz [0.15GB]
• sentinel-5p-epa.tar.gz [1.8GB]
• samples_2018_2020_epa.csv
• samples_quarterly_epa.csv
• samples_monthly_epa.csv

Acknowledgement

If you use this data set, please cite our publication:

*Scheibenreif, L., Mommert, M., Borth, D., "*Estimation of Air Pollution with Remote Sensing Data: Revealing Greenhouse Gas Emissions from Space*", Tackling Climate Change with Machine Learning workshop at ICML 2021.*

Please refer to this publication for additional information on the data set.

This data set contains modified Copernicus Sentinel data acquired in 2018-2020, processed by ESA.

Responsible Author

Linus Scheibenreif

University of St. Gallen, Institute of Computer Science

Chair Artificial Intelligence and Machine Learning

linus.scheibenreif ( at ) unisg.ch

Air pollution produced by fossil fuel combustion is a major hazard to human health, and was responsible for an estimated 8.7 million premature deaths worldwide in 2018. Death rates was especially high in Eastern Asia, where an average of 30.7 percent of adult deaths were attributable to fine particulate matter (PM2.5) generated by the burning of fossil fuels such as coal. In China and India there were an estimated 3.9 million and 2.5 million deaths attributable to fossil-fuel related PM2.5. Regions with high levels of air pollution from fossil fuel combustion have higher mortality rates than in regions such as Africa and South America.

SUMMARYPredicted concentration of PM2.5 air pollution: modelled emissions from all sectors (2018).DATA SOURCENational Atmospheric Emissions Inventory. Available here.COPYRIGHT NOTICENational Atmospheric Emissions Inventory. Licensed under the Open Government Licence v3.0. This data is based on and includes information from: Ordnance Survey (GB) data © Crown copyright and database right 2021; Royal Mail (GB) data © Royal Mail copyright and database right 2021; National Statistics (GB) data © Crown copyright and database right 2021. CaBA HEALTH & WELLBEING EVIDENCE BASEThis dataset forms part of the wider CaBA Health and Wellbeing Evidence Base.

Annual levels of PM10 suspended particulate pollutants from Mulhouse agglomeration modelling for 2018. The data comes from the fine-resolution modelling chain (ADMS-URBAN model). All data provided are in μg/m³ (microgram per cubic meter). The statistics produced are comparable to the limit values for health protection: Decree 2010-1250 and Directive 2008/50/EC. The modeling is carried out in accordance with the recommendations and recommendations of the Central Laboratory for Air Quality Monitoring (LCSQA).

China has a vast territory, and different regions have different air quality conditions. The database selects the air quality of 264 major cities in China as the research object. From the Ministry of Ecology and Environment, PRC (http://www.mee.gov.cn/), China's air quality on-line monitoring and analysis platform https://www.aqistudy.cn/historydata/ for the 264 cities in 2018-2021 per hour AQI and CO, NO2 and O3, PM10 and PM2.5, The hourly monitoring concentration data of six pollutant items of SO2 (except for CO which is mg/m3 and the other units are ug/m3) are averaged over the 24 1-hour data every day to obtain the daily pollutant item concentration index and daily air quality index. All seven kinds of daily data sets of 264 cities from 2018 to 2021 were used as data samples for the study.

This dataset contains data contributed by EPA/ORD/NERL/CED researchers to the manuscript " Assessment and economic valuation of air pollution impacts on human health over Europe and the United States as calculated by a multi-model ensemble in the framework of AQMEII3" led by Dr. Ulas Im of Aarhus University in Denmark. This dataset is associated with the following publication: Im, U., J. Brandt, C. Geels, K. Hansen, J. Christensen, M. Andersen, E. Solazzo, I. Kioutsioukis, U. Alyuz, A. Balzarini, R. Baro, R. Bellasio, R. Bianconi, J. Bieser, A. Colette, G. Curci, A. Farrow, J. Flemming, A. Fraser, P. Jimenez-Guerrero, N. Kitwiroon, C. Liang, U. Nopmongcol, G. Pirovano, L. Pozzoli, M. Prank, R. Rose, R. Sokhi, P. Tuccella, A. Unal, M. Garcia Vivanco, J. West, G. Yarwood, C. Hogrefe, and S. Galmarini. Assessment and economic valuation of air pollution impacts on human health over Europe and the United States as calculated by a multi-model ensemble in the framework of AQMEII3. Atmospheric Chemistry and Physics. Copernicus Publications, Katlenburg-Lindau, GERMANY, 18: 5967-5989, (2018).

This statistic displays the public concerns about air pollution in the United Kingdom (UK) in 2018. Of those surveyed, 80 percent showed concern about air pollution, whilst just two percent claimed they were not concerned at all.

This includes supporting datasets and the replication files for "Fine-particulate air pollution and behaviorally inclusive mortality impacts of China’s winter heating policy, 2013-2018", by Alberto Salvo, Qu Tang, Jing Yang, Peng Yin, and Maigeng Zhou.

Context

Air Pollution is a major environmental concern that demands immediate attention. A data-centric approach can help better understand the problem, identify highly impacted areas and target solutions appropriately. The data collected under the NATIONAL AIR QUALITY MONITORING PROGRAMME (N.A.M.P.) is available in the form of a PDF, a nightmare for Data Scientists. Fortunately, it has been converted into a usable CSV format, and every record has been cross-verified to ensure data integrity.

Content

Column	Description
State	Name of the state
City	Name of the city
Location	Location in the city where the recording was taken
{Pollutant} Monitoring days / year	Number of days that pollutant's recordings were taken that year.
{Pollutant} Minimum	Minimum value of that pollutant that year
{Pollutant} Maximum	Maximum value of that pollutant that year
{Pollutant} Annual Avg	Average value of that pollutant that year

Acknowledgements

NAMP data - Source Image Credits - Veeterzy on Unsplash

AQI: Florida: Orlando-Kissimmee-Sanford: Ozone data was reported at 58.000 Index in 23 Mar 2025. This stayed constant from the previous number of 58.000 Index for 22 Mar 2025. AQI: Florida: Orlando-Kissimmee-Sanford: Ozone data is updated daily, averaging 38.000 Index from Jan 1980 (Median) to 23 Mar 2025, with 16444 observations. The data reached an all-time high of 151.000 Index in 18 Apr 2018 and a record low of 16.000 Index in 25 Aug 2020. AQI: Florida: Orlando-Kissimmee-Sanford: Ozone data remains active status in CEIC and is reported by United States Environmental Protection Agency. The data is categorized under Global Database’s United States – Table US.ESG.E001: Air Quality Index and Air Pollutants. [COVID-19-IMPACT]

Annual hourly air quality and meteorological data by monitoring site for the 2018 calendar year. For more information on air quality, including live air data, please visit https://www.qld.gov.au/environment/pollution/monitoring/air.

Data resolution: One-hour average values (one-hour sum for rainfall)
Data row timestamp: Start of averaging period
Missing data/not monitored: Blank cell
Calm conditions: No hourly average wind direction is reported when the hourly average wind speed is zero
Barometric pressure: Values are at monitoring station elevation, not corrected to mean sea level
Daily zero/span response check: Automated instrument zero/span response checks are conducted daily between midnight and 1am at Queensland Government sites (can differ at industry sites). Where this takes place an ambient hourly value cannot be reported.
Sampling height: Four metres above ground (unless otherwise indicated)

Regionalisation of national emission data on the basis of EU reporting obligations (data status 2020).In the Federal States Air Pollution Inventory (BLI), the Federal Environment Agency assigns the national emission data from the Austrian Air Pollution Inventory to the individual Federal States. The report shows the evolution of greenhouse gases and other selected air pollutants (nitrogen oxides, sulphur dioxide, ammonia and non-methane volatile hydrocarbons) for the years 1990 to 2018. For the particulate matter fractions PM10 and PM2.5, the Federal States' Air Pollution Inventory (BLI) contains the emission data for the years 2000 to 2018. The federal states' specific analysis is continuously supplemented by new surveys and detailed analyses of emission data and influencing factors.The federal states' air pollutant inventory is compiled annually by the Federal Environment Agency in cooperation with the offices of the state governments.

Annual levels of PM10 suspended particulate pollutants from the Metz agglomeration modelling for 2018. The data comes from the fine-resolution modelling chain (ADMS-URBAN model). All data provided are in μg/m³ (microgram per cubic meter). The statistics produced are comparable to the limit values for health protection: Decree 2010-1250 and Directive 2008/50/EC. The modeling is carried out in accordance with the recommendations and recommendations of the Central Laboratory for Air Quality Monitoring (LCSQA).

Annual levels of NO2 nitrogen dioxide pollutants from the Colmar agglomeration modelling for 2018. The data comes from the fine-resolution modelling chain (ADMS-URBAN model). All data provided are in μg/m³ (microgram per cubic meter). The statistics produced are comparable to the limit values for health protection: Decree 2010-1250 and Directive 2008/50/EC. The modeling is carried out in accordance with the recommendations and recommendations of the Central Laboratory for Air Quality Monitoring (LCSQA).

This paper is under review

Abstract:

Kampala, the political and economic capital of Uganda and one of the fastest urbanising cities in sub-Saharan Africa, is experiencing a deteriorating trend in air quality with emissions from multiple diffused local sources like transportation, domestic and outdoor cooking, and industries, and sources outside the city airshed like seasonal open fires in the region. PM2.5 (particulate matter under 2.5um size) is the key pollutant of concern in the city with monthly spatial heterogeneity of 60-100 ug/m3. Outdoor air pollution is distinctly pronounced in the global south cities and lack the necessary capacity and resources to develop integrated air quality management programmes including ambient monitoring, emissions and pollution analysis, source apportionment, and preparation of clean air action plans. This paper presents an integrated assessment of air quality in Kampala drawing from ground measurements (from a hybrid network of stations), satellite observations (from NASA’s MODIS and OMI), global reanalysis fields (from GEOS-chem and CAMS simulations), high resolution (~1km) multi-pollutant emissions inventory for the airshed, WRF-CAMx based PM2.5 pollution analysis, and a qualitative review of institutional and policy environment for air quality management in Kampala. The proposed clean air action plans aim for better air quality in the region using a combination of short-, medium-, and long-term emission control measures for all the dominate sources and institutionalize pollution tracking mechanisms (like emissions and pollution monitoring and reporting) for effective management of air pollution.

This data archive serves as a supplemenary to the journal article and with a short description of the files below:

• File: AQ-Kampala-Analysis-Summary.pptx (Caution: large 60MB)
  A composite presentation including the following
  
  • Grid summaries
  • Snapshots of airshed GIS files, emission activities
  • Summary of meteorology from WRF simulations and historical synoptics
  • Summaries of ambient monitoring data
  • CAMS reanalysis summary
  • Summaries of Emission inventory and WRF-CAMx modelling (annual and monthly)
  • Summaries of PM2.5 Source apportionment (annual and monthly)
• File: grids_kampala.rar
  Grid file (KML and ESRI shapefiles format) for the airshed spanning 0.0N to 0.6N and 32.3E to 32.9E with a spatial resolution of 0.01deg (~1km)
• File: gis_roads_from_opensteetmaps.rar
  ESRI shapefiles of primary roads and all roads, extracted from the openstreetmaps
  Raw data archive @ https://download.geofabrik.de/index.html
• File: gis-scanned2021image-quarries.kml
  KML file of quarries scanned using the imagery on Google Earth platform
• File: population_kampala_2000-2022.csv
  Gridded population data 2000 to 2022
  Raw data archive is from LANDSCAN - https://landscan.ornl.gov
• File: Monitoring-Kampala_USEmbassy_2017-2024.xlsx
  Summary of monitoring data collected at the US Embassy in Kampala
  Raw data archive is @ https://www.airnow.gov/international/us-embassies-and-consulates/#Uganda$Kampala
• File: meteo_wrf_stats.xlsm
  Summary of output of WRF simulations for the Kampala region. Have to activiate macros to summarize the results by month and update the charts. The tool can be used to other cities also by changing the input data.
• File: meteo_precip-era5-reanalysis.csv
  Summary of monthly precipitation date (mm/day) from ERA5 reanalysis fields
  Raw data archive is @ https://psl.noaa.gov/data/atmoswrit/timeseries
• File: TROPOMI_EastAfrica_NO2_Maps.zip
  Images of monthly average TROPOMI NO2 extracts covering East Africa (Uganda and Ethiopia)
  Extracted from Google Earth Engine, using 10% cloud fraction
• File: TROPOMI_EastAfrica_CSVs.zip
  CSV files of gridded monthly average NO2, SO2, HCHO, and Ozone columnar densities
  Extracted from Google Earth Engine, using 10% cloud fraction
  Read the data descriptions and applicability of the data for analysis before using (for example, negative numbers in the SO2 file).
  NO2 - https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_OFFL_L3_NO2
  SO2 - https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_OFFL_L3_SO2
  HCHO - https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_OFFL_L3_HCHO
  O3 - https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_OFFL_L3_O3
• File: composite_emisson_factors_gains.xlsx
  A composite library of emission factors for reference
• File: kampala_gridded_emissions_2018.rar
  Gridded emissions inventory for Kampala - PM25, PM10, SO2, NO, NO2, and CO
  PM25 is speciated into FPRM, BC, and OC (sum all for PM25)
  PM10 is speciated into FPRM, CPRM, BC, OC (sum all for PM10)
  All emissions in tons/year/grid
  Emissions are seggragted into sectors and fuels - included in the filenames
• File: kampala_gridded_modelled_monthavgp25.csv
  Gridded PM2.5 concentrations for 2018, from WRF-CAMx modelling system
  Monthly averages in ug/m3