Empirical Model Database (BETA version)

A Repository of Air Quality Estimates From Empirical Models


What is the Empirical Model Database?

Air pollution researchers in the U.S. and worldwide have developed national and continental-scale models for estimating concentrations of several air pollutants. The purpose of the modeling is to estimate concentrations at locations without measurements, and to do so with excellent spatial precision.

Researchers whose results are shared here have elected to make the model estimates publicly available in the hope that they might be useful to others. They and we welcome feedback, comparison, and corroboration that can help improve the methods and estimates. Estimates provided here are most appropriate for understanding averages, broad patterns, and national trends. The underlying models can also produce estimates at individual locations, which can be appropriately used as inputs into analyses that consider a large number of locations, such as studies of the health effects of air pollution (which was the motivating application for many of these models). Researchers generally do not use estimates of the types given here to draw conclusions about air quality at one location or a small number of specific locations; as such, we do not recommend that type of use.

Models shared here are generally derived from publicly available concentration measurements from regulatory monitors. (In the U.S. these measurements are collected by EPA, state, local, and tribal agencies and made publicly available via the U.S. EPA Air Quality System). The models also use information about land use (for example, locations of major and minor roads; elevation; and whether an area is urban or rural) and in some cases satellite-derived estimates of air pollution.

The tables below summarize peer-reviewed large-scale empirical models, and provide links to access model concentration estimates if they have been made publicly available. In some cases, there is more than one model available for a given pollutant and year. Those models have been developed by different research groups, using different methods, or are applied at different spatial locations. We expect that users will familiarize themselves with corresponding publications, and encourage them to contact researchers (contact information provided below) for questions about the data provided, or to request customized data sets via formal collaboration.


Thank you for your interest. If you have suggestions for improvement or have model estimates you would like to share here, please contact us (Matthew Bechle).


Subscribe to the mailing list to be updated when we add new models and data sets.


Nitrogen Dioxide (NO2) Concentration Estimates

Pollutant Years Temporal Unit Location Type Geographic Coverage Model Type Citation Contact Data
NO2 (ppb) 1996-2012 Annual;
3-yr mean
0.1°×0.1° Global RS-derived estimates Geddes et al., 2015
DOI, PubMed
J.A. Geddes Link
NO2 (ppb) 1990-2012 Annual 25-km grid*; Tract Centers* Contiguous U.S. LUR/UK with and without RS Young et al.,2016
DOI, PubMed
M.T. Young
NO2 (ppb) 2006 Annual Block-face* Canada LUR with RS Crouse et al., 2015
DOI, PubMed
D.L. Crouse
NO2 (ppb) 2000-2010 Annual;
Monthly
Block Centers*;
Block Group;
Tract
Contiguous U.S. LUR with RS Bechle et al.,2015
DOI, PubMed
J.D. Marshall Download
NO2 (ppb) 2006-2011 Annual;
Monthly
Mesh Block Centers* Australia LUR with RS Knibbs et al.,2014
DOI, PubMed
L.D. Knibbs Available by request
NO2 (ppb) 2005-2010 Daily 1-km grid New England LUR with RS Lee & Koutrakis, 2014
DOI, PubMed
H.J. Lee
NO2 (µg/m3) 2005-2007 Annual 100-m grid Western Europe LUR with RS Vienneau et al.,2013
DOI, PubMed
D. Vienneau Link

BME = Bayesian maximum entropy; LUR = land use regression; RS = remote sensing; UK = universal kriging; Models noted with * provide point-specific concentration estimates


Particulate Matter (PM) Concentration Estimates

Pollutant Years Temporal Unit Location Type Geographic Coverage Model Type Citation Contact Data
PM2.5 (µg/m3) 1998-2014 Annual 0.01°×0.01° Global RS-derived estimates van Donkelaar et al., 2016
DOI, PubMed
A. van Donkelaar Link
PM2.5 (µg/m3) 1990; 1995;
2000; 2005;
2010; 2013
Annual 0.1°×0.1° Global Fused model with RS Brauer et al., 2016
DOI, PubMed
M. Brauer Link
PM2.5 (µg/m3) 2003-2011 Daily 1-km grid Southeast U.S. LUR with RS Lee et al., 2015
DOI, PubMed
PM2.5 (µg/m3) 2003-2011 Daily 1-km grid Northeast U.S. LUR with RS Kloog et al., 2014
DOI
I. Kloog
PM2.5 (µg/m3) 1999-2009 Annual 1.404°×0.784°* Contiguous U.S. LUR/BME Reyes & Serre,2014
DOI, PubMed
M.L. Serre Maps
PM10 (µg/m3) 2006 Annual 100-m grid Western Europe LUR with RS Vienneau et al.,2013
DOI, PubMed
D. Vienneau Link
PM2.5 (µg/m3) 1999-2011 Annual 25-km grid*; Tract Centers* Contiguous U.S. LUR/UK Sampson et al.,2016
DOI, PubMed
P.D. Sampson
PM2.5 (µg/m3) 1999-2008 Monthly Tract Centers* Contiguous U.S. LUR/BME with & without RS Beckerman et al.,2013
DOI, PubMed
B.S. Beckerman Link
PM2.5 (µg/m3) 2001-2008 Daily Tracts Contiguous U.S. Fusion model U.S. EPA, 2012 D. Holland Link
PM2.5 (µg/m3) 2006 Annual Block-face* Canada LUR with RS Hystad et al., 2011
DOI, PubMed
P. Hystad

BME = Bayesian maximum entropy; LUR = land use regression; RS = remote sensing; UK = universal kriging; Models noted with * provide point-specific concentration estimates


Other Pollutant Concentration Estimates

Pollutant Years Temporal Unit Location Type Geographic Coverage Model Type Citation Contact Data
O3 (ppb) 2001-2009 Long-term avg. Postal code* Canada Fusion model;
avg. 8-hr daily max during warm season
Crouse et al.,2015
DOI, PubMed
D.L. Crouse
O3 (ppb) 2001-2008 Daily Tracts Contiguous U.S. Fusion model U.S. EPA, 2012 D. Holland Link
Benzene (µg/m3) 2006 Annual Block-face* Canada LUR Hystad et al., 2011
DOI, PubMed
P. Hystad
Ethylbenzene (µg/m3) 2006 Annual Block-face* Canada LUR with RS Hystad et al., 2011
DOI, PubMed
P. Hystad
1,3-butadiene (µg/m3) 2006 Annual Block-face* Canada LUR with RS Hystad et al., 2011
DOI, PubMed
P. Hystad

BME = Bayesian maximum entropy; LUR = land use regression; RS = remote sensing; UK = universal kriging; Models noted with * provide point-specific concentration estimates