!"#$%&"'()*+%&",#$-.("&/(01"-*()/2$&$.
#+"#$%&
Satellite-Based PM2.5 Datasets
Melanie Follette-Cook, Pawan Gupta, Carl Malings, Sarah Strode
NASA Air Quality Remote Sensing Training for EPA, March 21-23, 2023
Learning Objectives
By the end of this presentation, you will be able to:
! Give examples of applications for surface PM2.5 estimates
! List several ways satellite observations can be used to estimate surface PM2.5
! Locate relevant PM2.5 estimates
NASA’s Applied Remote Sensing Training Program 2
Examples of Applications Using
Surface PM2.5 Estimates
Satellites Provide a “God’s Eye” View of the Earth
Spatial coverage is the primary advantage of satellite data.
https://openaq.org/
MODIS Terra AOD Sep 2020
Surface Monitors
Satellite
https://earthobservatory.nasa.gov/global-
AOD maps/MODAL2_M_AER_OD
https://www.purpleair.com/
NASA’s Applied Remote Sensing Training Program 4
Health Studies of Exposure WHO Interim Targets
Annual
Average
PM2.5
(µg/m3)
Age-standardized rates of death attributable to PM2.5 in 2019
https://www.stateofglobalair.org/
NASA’s Applied Remote Sensing Training Program 5
UN Sustainable Development Goals (SDGs)
Transforming Our World: The 2030 Agenda for Sustainable Development
Goal 3 – Good Health and Well Being
! Target 3.9; Indicator 3.9.1
! Mortality rate attributed to household
and ambient air pollution (annual
mean levels of air pollution [PM2.5])
Goal 11 – Sustainable Cities and
Communities
! Target 11.6; Indicator 11.6.2
! Annual mean levels of fine particulate
matter (e.g., PM2.5 and PM10) in cities
(population weighted)
Text adapted from “Transforming our world: the 2030 Agenda for Sustainable Development”
NASA’s Applied Remote Sensing Training Program 6
PM2.5 Estimation: Popular Methods
PM2.5
Two-Variable Y=mX + c
Method
AOT
Multivariable m
PM2.5 = !0 + " # $ + % (!n#Mn)
Difficulty Level
Method n=1
Artificial
Intelligence
Model Surface Area Concentration
MSC Estimated PM2.5 = x Satellite AOD
Model AOD
NASA’s Applied Remote Sensing Training Program 7
Satellite-Based Estimates of Surface PM2.5 and
Chemical Composition - Van Donkelaar et al.
(2021)
Van Donkelaar et al. (2021)
Eight retrievals of AOD from four different instruments
Instrument MODIS: Terra/Aqua MISR SeaWiFS
Retrieval Algorithm Deep Blue Dark Target MAIAC MISR Deep Blue
Horizontal
10 km 10 km 1 km 17.6 km 13.5 km
Resolution
Van Donkelaar et al., 2021, https://pubs.acs.org/doi/pdf/10.1021/acs.est.1c05309
NASA’s Applied Remote Sensing Training Program 9
Van Donkelaar et al. (2021)
AOD ! PM2.5
Calculate Monthly Mean
AOD from
&
AOD
Model
(8 retrievals)
Simulations
Calculate PM2.5
from AOD using
Calibrate all using surface observations model AOD-to-PM2.5
of AOD from the AERONET network relationship
NASA’s Applied Remote Sensing Training Program 10
Van Donkelaar et al. (2021)
Geographic Weighted Regression (GWR)
GWR corrects the satellite estimate using the relationship between
PM2.5 from ground monitors and variables such as model aerosol
composition, elevation data, and land use indicators.
PM2.5 Uncertainty Estimates
Uncertainty is estimated by using 1) the range of AOD values going
into the best estimate and simulated PM2.5/AOD relationships, and
2) the predictor coefficients of the GWR.
NASA’s Applied Remote Sensing Training Program 11
Washington Univ of St. Louis - Atmospheric Composition Analysis Group
https://sites.wustl.edu/acag/datasets/surface-pm2-5/
! Global and Regional PM2.5
(V5.GL.03) (1998-2021)
" Annual and Monthly
Means at 0.01# x 0.01#
" Annual and Monthly
Means at 0.1# x 0.1#
" Annual and Monthly Mean
Uncertainty at 0.01# x 0.01#
! North American Regional
Estimates (with composition)
(V4.NA.03)(2000-2016)
" Annual and Monthly
Means at 0.01# x 0.01#
NASA’s Applied Remote Sensing Training Program 12
North American Regional Estimates (V4.NA.03)
https://pubs.acs.org/doi/pdf/10.1021/acs.est.8b06392
“We estimate ground-level fine particulate matter (PM2.5) total and compositional
mass concentrations over North America by combining Aerosol Optical Depth
(AOD) retrievals from the NASA MODIS, MISR, and SeaWIFS instruments with the
GEOS-Chem chemical transport model, and subsequently calibrated to regional
ground-based observations of both total and compositional mass using
Geographically Weighted Regression (GWR) as detailed in the below reference.”
! https://sites.wustl.edu/acag/datasets/surface-pm2-5/#V4.NA.03
! 2000-2017, 0.01# $ 0.01#
NASA’s Applied Remote Sensing Training Program 13
North American Regional Estimates (V4.NA.03)
https://pubs.acs.org/doi/pdf/10.1021/acs.est.8b06392
NASA’s Applied Remote Sensing Training Program Donkelaar et al., 2019 14
Reanalysis-Based Estimates of Surface PM2.5
and Chemical Composition – MERRA-2
Modern-Era Retrospective analysis for Research and
Applications, Version 2 (MERRA-2)
https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/
! NASA’s Global Model and Assimilation
Office (GMAO) produces estimates of
surface PM2.5 over the period of 1980
to the present day.
! The model system assimilates
meteorological data as well as some
atmospheric constituents (e.g., ozone,
AOD).
! Spatial Resolution: 0.5x0.625 deg.
NASA’s Applied Remote Sensing Training Program 16
MERRA-2 Aerosol Observations
! Aerosol assimilation is described in detail in
Randles et al. 2017 and https://gmao.gsfc.nasa.gov/pubs/docs/Randles887.pdf.
! In MERRA-2, AOD at 550 nm is assimilated.
! Some Notes:
" No information on vertical structure or composition
" Daylight observations only
" Subject to meteorological conditions (e.g., clouds) and viewing geometry (e.g., sun glint)
" When there are no observations, MERRA-2 draws towards the GEOS/GOCART simulation.
Table 2 from Randles et al. 2017
NASA’s Applied Remote Sensing Training Program 17
MERRA-2 Aerosol Observations
! When using MERRA-2 products, one must take care to consider the changing
observing system over time.
Figure 3 from Randles et al. 2017
NASA’s Applied Remote Sensing Training Program 18
Machine Learning-Based Estimates
of Surface PM2.5
Machine Learning Ensemble-based PM2.5 over CONUS (2000-2016)
https://sedac.ciesin.columbia.edu/data/set/aqdh-pm2-5-concentrations-
contiguous-us-1-km-2000-2016
! Daily and Annual, 1 km2
! Available in RDS and GeoTIFF
! Also Available at Zip Code Level
! https://sedac.ciesin.colum
bia.edu/data/set/aqdh-
pm2-5-o3-no2-
concentrations-zipcode-
contiguous-us-2000-2016
NASA’s Applied Remote Sensing Training Program 20
Machine Learning Ensemble-Based PM2.5 over CONUS (2000-2016)
Di et al., 2019, Environmental International
https://www.sciencedirect.com/science/article/pii/S0160412019300650
! Meteorology: Relationship Between
! NARR (North American Monitored and Predicted
Regional Reanalysis) PM2.5 at Annual Level
! Satellite Observations:
! MODIS MAIAC
! MERRA-2 Speciation
! CTM:
! GEOS-Chem, CMAQ
! Land-Use Terms:
! Coverage Type, Road
Density, Restaurant
Density, Elevation, NDVI
! Additional downscaling to 100 m
! Includes uncertainty estimates
NASA’s Applied Remote Sensing Training Program 21
Global Deep Ensemble Machine Learning (2000-2019)
Yu et al., 2023, Lancet, https://www.thelancet.com/action/showPdf?pii=S2542-5196%2823%2900008-6
! Deep Ensemble Machine Learning
(DEML) Figure 6, Yu et al., 2023
! Inputs:
" Station PM2.5 (Daily Mean)
" GEOS-Chem
" ERA5 Reanalysis
" MODIS Land Cover
" Population Data Figure 2, Yu et al., 2023
! Daily, Global, PM2.5 Estimates
! 0.1# x 0.1#
! Data not available yet
NASA’s Applied Remote Sensing Training Program 22
Questions?
