Statistical air quality predictions for public health surveillance: evaluation and generation of county level metrics of PM2.5for the environmental public health tracking network
© Vaidyanathan et al; licensee BioMed Central Ltd. 2013
Received: 9 January 2013
Accepted: 24 February 2013
Published: 14 March 2013
The Centers for Disease Control and Prevention (CDC) developed county level metrics for the Environmental Public Health Tracking Network (Tracking Network) to characterize potential population exposure to airborne particles with an aerodynamic diameter of 2.5 μm or less (PM2.5). These metrics are based on Federal Reference Method (FRM) air monitor data in the Environmental Protection Agency (EPA) Air Quality System (AQS); however, monitor data are limited in space and time. In order to understand air quality in all areas and on days without monitor data, the CDC collaborated with the EPA in the development of hierarchical Bayesian (HB) based predictions of PM2.5 concentrations. This paper describes the generation and evaluation of HB-based county level estimates of PM2.5.
We used three geo-imputation approaches to convert grid-level predictions to county level estimates. We used Pearson (r) and Kendall Tau-B (τ) correlation coefficients to assess the consistency of the relationship, and examined the direct differences (by county) between HB-based estimates and AQS-based concentrations at the daily level. We further compared the annual averages using Tukey mean-difference plots.
During the year 2005, fewer than 20% of the counties in the conterminous United States (U.S.) had PM2.5 monitoring and 32% of the conterminous U.S. population resided in counties with no AQS monitors. County level estimates resulting from population-weighted centroid containment approach were correlated more strongly with monitor-based concentrations (r = 0.9; τ = 0.8) than were estimates from other geo-imputation approaches. The median daily difference was −0.2 μg/m3 with an interquartile range (IQR) of 1.9 μg/m3 and the median relative daily difference was −2.2% with an IQR of 17.2%. Under-prediction was more prevalent at higher concentrations and for counties in the western U.S.
While the relationship between county level HB-based estimates and AQS-based concentrations is generally good, there are clear variations in the strength of this relationship for different regions of the U.S. and at various concentrations of PM2.5. This evaluation suggests that population-weighted county centroid containment method is an appropriate geo-imputation approach, and using the HB-based PM2.5 estimates to augment gaps in AQS data provides a more spatially and temporally consistent basis for calculating the metrics deployed on the Tracking Network.
KeywordsParticulate matter Tracking Network Hierarchical Bayesian Air quality system Geo-imputation
Numerous studies have identified a relationship between fine particulate air pollution and its impact on human health . Particles with an aerodynamic diameter of 2.5 μm or less (PM2.5) are small enough to invade air pathways in the body, and have been known to cause adverse health effects . Several epidemiologic and human clinical studies have examined the cardiovascular and respiratory health effects of both acute and long term exposures to PM2.5[3–5]. The Medicare Air Pollution Study (MCAPS), a multi-city study in the United States (U.S.), reported a short-term increase in hospital admission rates associated with elevated ambient PM2.5 concentrations, for health outcomes such as ischemic heart disease, heart failure, chronic obstructive pulmonary disease and respiratory infection . The MCAPS study also concluded that the cardiovascular risks, estimated at the county level, tended to be higher in the eastern U.S. Similarly, the extended follow-up of the Harvard Six Cities study showed that cardiovascular and lung cancer mortality were positively associated with long-term ambient concentrations of PM2.5.
To quantify the health impacts of PM2.5, and to track population exposure to PM2.5, accurate and timely data collected on an ongoing basis are needed at the sub-state level. The Pew Environmental Health Commission report released in 2000 found that the state of environmental public health systems were fragmented and not robust enough to respond to environmental threats . Based on the recommendations of the Pew Commission, the U.S. Congress funded the Centers for Disease Control and Prevention (CDC) to establish a National Environmental Public Health Tracking Program. The cornerstone of this program is the National Environmental Public Health Tracking Network (Tracking Network) which provides nationally consistent data and metrics (indicators and measures) to monitor relationships among hazards, exposures, and health effects . The CDC, U.S. Environmental Protection Agency (EPA), and state local health departments funded by the CDC have been collaborating in the development of air quality metrics for PM2.5 for integration into the Tracking Network (http://ephtracking.cdc.gov/showAirData.action). In July 2009 during the initial launch of the Tracking Network, only Federal Reference Method (FRM) Air Quality System (AQS) monitor data were incorporated into the Network to provide county level air quality metrics.
In order to better understand air quality for areas and days without monitor data, the CDC collaborated with the EPA on the development of a hierarchical Bayesian (HB) model to predict daily PM2.5 concentrations for use in the Tracking Network. The HB model integrates AQS monitor data with results from the EPA’s Community Multiscale Air Quality (CMAQ) model to generate predicted PM2.5 concentrations for a 36-km grid (individual predictions for 36-km × 36-km grid cells) across the conterminous U.S. and for a 12-km grid across the eastern half of the U.S. (http://www.cmaq-model.org). The statistical approach incorporates prior knowledge of the model parameters in the hierarchical Bayesian model, which results in improved estimation of the PM2.5 concentrations in areas (also covered by the CMAQ grid) and at times (days) that are not monitored . The model also quantifies the prediction error associated with the predicted daily concentrations for each grid cell. In general, the HB model utilizes monitor data, and bias-adjusted CMAQ model output for non-monitored areas and days. Background documents for the HB model can be found on the Tracking Network and at the EPA webpage (http://www.epa.gov/heasd/sources/projects/CDC/index.html).
creating a spatial relationship between grid cells and counties so that daily county level estimates can be generated from HB predictions;
evaluating whether HB predictions at the 12- or 36-km resolution PM2.5 should be used for the calculation of daily county level estimates;
comparing the resultant daily county level HB estimates with AQS county level monitor data;
comparing county level annual averages of PM2.5 based on HB estimates with those based on AQS monitor data.
n k = number of census blocks in county k;
X Bj,k = X coordinate of the centroid of census block j contained within county k;
Y Bj,k = Y coordinate of the centroid of census block j contained within county k;
P Bj,k = population of census block j contained within county k;
Our second approach was again based on centroid containment and relates all grid cell centroids (geometric) to the county into which they fall . We established a relationship between each given county boundary polygon and all the grid cell geometric centroids it contains, and then transferred HB predictions to that county (Figure 2B). For counties that did not contain a grid cell geometric centroid, which were very few, we related the nearest one. Since many counties contain more than one grid cell centroid, we selected the maximum predicted concentration value for each day from all the grid cells with centroids in each given county to create daily county level estimates of PM2.5. This is consistent with the EPA approach of using the maximum concentration among multiple monitors in a county.
The third approach was based on geometric intersection of the county boundary and grid cell polygons . We performed an intersect-overlay analysis to identify geometric intersections between grid cells and counties and related each county to grid cells that either fell within or intersected with the county boundary (Figure 2C). After establishing the appropriate many-to-many relationship between counties and grid cells, we selected the maximum HB prediction for each day from all the grid cells related to each given county to create daily county level estimates of PM2.5 from HB predictions.
Evaluation of county level estimates of PM2.5
the denominator is adjusted accordingly in the event of ties .
We carried out our data analyses using the Statistical Analysis System (SAS® Version 9.2) and Environmental Systems Research Institute’s GIS software (ESRI, ArcGIS® Version 9.3). This study was determined to be research not involving human subjects by the CDC National Center for Environmental Health (NCEH) Office of Science. This study did not require further review by the CDC institutional review board.
For 2005, 587 (19%) of counties in the conterminous U.S. had PM2.5 monitors that operated year-round. Most of these PM2.5 monitors only operated every third day while some operated every sixth day. A few monitors operated on an every-day schedule. HB predictions available at the 36-km grid resolution were available for 11266 grid cells covering the entire conterminous U.S. It should be noted that CMAQ estimates dominated the 36-km HB predictions in the western areas where few monitors are located. The 12-km HB predictions were available for the eastern U.S. with 66960 grid cells, out of which 66123 grid cells overlapped and were aligned with the 36-km grid.
Comparison of Geo-imputation methods
Correlation between county level HB estimates of PM2.5and AQS-based PM2.5concentrations
Population-weighted county centroid containment
Grid cell centroid containment
Daily county level comparison
Daily differences between county level HB estimates of PM2.5and AQS-based PM2.5concentrations by census regions and divisions
AQS-based PM2.5concentration ranges
Median absolute difference(μg/m3) *
Median relative difference(%)α
Median absolute difference(μg/m3)*
Median relative difference(%)α
Median absolute difference(μg/m3)*
Median relative difference(%)α
Median absolute difference(μg/m3)*
Median relative difference(%) α
East North Central
(−0.2 – 0.8)
(−3.9 – 16.0)
(−0.8 – 0.7)
(−7.2 – 6.7)
(−2.4 – 0.2)
(−11 – 1.0)
(−6.5 – -1.5)
(−14.5 – -3.6)
West North Central
(−0.3 – 0.6)
(−6.2 – 13.9)
(−1.2 – 0.4)
(−11.0 – 3.6)
(−2.7 – -0.2)
(−13 – -1.1)
(−4.9 – -1.6)
(−12.9 – -3.9)
(−0.2 – 1)
(−3.7 – 20.1)
(−1 – 0.9)
(−8.9 – 8.3)
(−2.9 – 0.4)
(−13.6 – 1.8)
(−12.3 – -2.3)
(−28.3 – -6.1)
(−0.5 – 0.6)
(−9.7 – 14.1)
(−1.5 – 0.3)
(−14.6 – 3.1)
(−3.4 – -0.1)
(−16.6 – -0.6)
(−7.9 – -2.3)
(−19.6 – -6.1)
East South Central
(−0.2 – 1)
(−3.1 – 17.4)
(−0.7 – 0.8)
(−6.3 – 7.2)
(−2.0 – 0.3)
(−9.8 – 1.6)
(−6.6 – -1.3)
(−15.4 – -3.2)
(−0.4 – 0.7)
(−7.4 – 12.1)
(−0.8 – 0.7)
(−7.7 – 7.0)
(−1.9 – 0.4)
(−9.5 – 2.4)
(−5.5 – -1.1)
(−14.4 – -2.9)
West South Central
(−0.4 – 0.7)
(−5.8 – 12)
(−1.1 – 0.5)
(−10.1 – 4.3)
(−2.4 – 0.1)
(−12.1 – 0.6)
(−9.4 – -2.1)
(−24.1 – -5.3)
(−0.5 – 0.7)
(−9.5 – 17.7)
(−2.7 – -0.3)
(−25.8 – -3.5)
(−8.5 – -1.9)
(−42.4 – -9.9)
(−25.2 – -8.2)
(−55.7 – -16.2)
(−0.5 – 0.8)
(−9.6 – 17.6)
(−1.7 – 0.5)
(−15.8 – 4.5)
(−5.9 – -0.8)
(−28.9 – -4.0)
(−20.7 – -5.8)
(−43.1 – -12.5)
The HB estimates comported well with AQS data when AQS-based concentrations were less than 35 μg/m3; the median AD and median RD ranged from −4.4 to +0.3 μg/m3 and from −23.3 to +6.8%, respectively. For concentrations less than or equal to 35 μg/m3, the median AD for the Midwest, Northeast, and South ranged from −1.3 to +0.2, -1.6 to +0.3, and −1.0 to +0.3 μg/m3, respectively; the West had a median AD range of −4.4 to +0.1 μg/m3, notably wider and more indicative of bias than the ranges observed for other census regions. When the prevailing AQS-based concentrations were greater than 35 μg/m3, county level HB-based estimates under predicted AQS-based concentrations across all census regions and divisions. For concentrations greater than 35 μg/m3, the median AD ranged between −15.2 and −3.2 μg/m3, and the median RD ranged between −35.3% and −7.5%. Counties in the Mountain and Pacific census divisions in the western U.S. showed the highest magnitude of under prediction.
Comparison of annual averages
Spatio-temporal gaps in monitoring can result in uncertainty in the ascertainment of population-level exposures, especially when fluctuations in PM2.5 concentrations tend to occur at frequencies not detectable given existing monitor sampling schedules. Monitors that operate for regulatory purposes are not usually sited very close to sources, where high concentrations of PM2.5 can be observed. Rather they are sited in places to measure levels of PM2.5 that represent average concentration levels over large areas. Lack of PM2.5 concentration measurements over continuous spatial and temporal scales limits our ability to link air quality levels with health effects data. Thus, modeled predictions may provide a suitable alternative for use in public health surveillance.
CMAQ model output as well as output from any model that relies on CMAQ output has a spatial unit, the grid cell that differs from the spatial unit of health and demographic data, which are often available at geo-political resolutions, such as county, census tract, etc. Geo-imputation procedures are necessary to convert grid-level data to county level estimates, which are needed to generate metrics for environmental public health surveillance through the CDC Tracking Network and for linkage to spatially comparable health data. The three geo-imputation methods mentioned in this paper are routinely used in public health and other allied fields. In our study, a population-weighted county centroid containment approach performed best among the methods considered for translating grid-level HB predictions to county level estimates. Also, a population-weighted county centroid denotes a spatial mean of the underlying population distribution within each county and estimates of PM2.5 generated using this method are intended to represent the ambient concentration levels to which most of the population are potentially exposed.
In the context of linking with health data, the spatial scale of modeled predictions was a very important consideration in developing county level PM2.5 estimates. For 2005, HB predictions of PM2.5 were available at a 12-km resolution for the eastern U.S., whereas HB predictions of PM2.5 were available at a 36-km resolution for the entire conterminous U.S. County level estimates of PM2.5 derived from 36-km predictions correlate more strongly with AQS-based PM2.5 concentrations than do estimates derived from 12-km predictions. The predominant reason for the difference in performance of 12- and 36-km HB predictions was the underlying CMAQ estimates. The input needed to generate the 12-km and 36-km CMAQ estimates were processed with different assumptions and, for certain inputs, resolving to a finer spatial scale could add uncertainty to the final model output . We developed county level estimates of PM2.5 from 36-km HB predictions since our primary goals were to have the HB-based metrics approach the values of the AQS-based metrics, and generate metrics for the entire conterminous U.S.
The strength and consistency of the relationship between daily HB- and AQS-based PM2.5 county level estimates are acceptable at concentrations below the daily NAAQS and, at these concentrations, differences between HB- and AQS-based PM2.5 estimates are reasonable for most census regions. For 2005, less than 2% of the measurements available from AQS monitoring reflected concentrations greater than 35 μg/m3. At these higher concentrations, HB-based county level estimates are more likely to under predict AQS concentrations, with the largest differences observed for the western region of the U.S. Some of these differences can be explained based on model features. The HB model fuses monitor data when available with CMAQ estimates and, for most locations, days with only CMAQ estimates outnumber days with both CMAQ and AQS data. CMAQ estimates are primarily used for predicting background concentrations and do not adequately capture spikes in air quality levels as a result of exceptional events . While the bias in the CMAQ estimates is adjusted using a global (national-level) regression approach, and the AQS data measurement error is accounted for in the HB model , the daily HB predictions can be different from the coincident AQS measurements when CMAQ estimates greatly differ from the AQS data. Additionally, there are relatively fewer monitor-based observations available for the western U.S. and CMAQ estimates under predict AQS concentrations in the western U.S., especially when the terrain is mountainous . Hence, HB estimates rely heavily on CMAQ in the western U.S. and we see larger absolute and relative differences between county level HB and AQS estimates with increasing PM2.5 concentrations (Table 2). Users of HB-based PM2.5 estimates should be aware of the limitations of these data as well as the benefits of having data over continuous spatio-temporal scales.
Annual county level metrics of PM2.5, such as annual averages, provide a useful summary of prevailing concentration levels. However, averages created from AQS-based PM2.5 concentration measurements are limited to counties with monitors and therefore do not provide a complete picture of prevailing air quality throughout the conterminous U.S. Moreover, PM2.5 annual metrics derived from AQS data based on a sampling frequency that is predominantly every third day can be taken to accurately characterize general conditions only under the assumption that days included in the sample fairly represent the full calendar. Given that the HB-based estimates are available for every day of the year, annual averages incorporating these estimates can be interpreted without any assumptions concerning days without data.
The benefits of employing HB predictions should be considered in light of the added uncertainty which they introduce. As noted, the annual county level HB-based annual averages can understate or overstate the air quality problem in specific areas compared to averages based on AQS concentrations. At higher concentrations, especially near the annual NAAQS—15 μg/m3, and in the western U.S., HB-based annual averages are more likely to fall below monitor-based measurements. Notably, combining HB-based estimates with AQS-based concentrations results in annual averages that comport well with annual averages created using AQS data exclusively; however, a few counties have lower annual averages when compared with the annual averages obtained exclusively from AQS-based concentrations.
In summary, we characterized the difference between HB-based estimates and AQS-based concentrations with the intent that the results can guide public health professionals and researchers on the appropriate use of the county level estimates of PM2.5. Our analysis of daily differences between AQS-based concentrations and HB-based estimates of PM2.5 indicate that the differences can vary across census regions and divisions, and that generally HB-based county level estimates tend to under predict at higher concentrations. This needs to be considered when using daily county level HB-based estimates to identify exceedances of the daily NAAQS in different parts of the country or to conduct studies that assess health outcomes related to short-term PM2.5 exposures. The annual averages developed by combining HB- and AQS-based PM2.5 data show less variation with AQS-based annual averages. Given the need to correctly characterize air quality levels and minimize the discrepancy with county level annual averages created from AQS data that are commonly used, we suggest that the county level annual averages of PM2.5 for the Tracking Network be calculated using AQS data when and where they are available and using HB-based estimates for days and locations without such data.
Poor air quality is a worldwide problem. The global burden of disease report (2010) identifies air pollution as a major contributor to premature mortality . Measurements from monitors are limited in space and time, and modeled data can be an alternative to ascertain population level exposures. Our evaluation of HB predictions and AQS measurements explores the utility of modeled data from a public health perspective. Using the HB-based predictions to augment gaps in AQS data provides a more spatially and temporally consistent basis for calculating the metrics deployed on the Tracking Network. Further, “data fusion” techniques, combining monitor and modeled data, are being used in many countries to produce grid-level air quality predictions . The evaluation framework presented in this paper is robust and can be extended to areas outside the U.S.
Converting grid-level predictions to estimates by geo-political units, such as counties or county equivalents, is needed to link health and population information with air pollutant data. This manuscript suggests that assigning modeled predictions to counties using a population-weighted centroid containment method is an effective approach for making the translation between a fixed grid and county-specific geography. Counties or similar administrative units exist in several European counties, for example, and the geo-imputation methods suggested in the paper can be adopted seamlessly to obtain pollutant concentrations to which most of the population is potentially exposed.
Air quality system
Centers for disease control and prevention
Community multi-scale air quality
US environmental protection agency
Environmental systems research institute
Federal reference method
Medicare air pollution study
National ambient air quality standards
Statistical analysis system
The authors would like to thank CDC Tracking program’s air workgroup for its helpful comments on the project. We would also like to appreciate the help provided by Dr. James A. Mulholland, Dr. Armistead G. Russell, Mr. Eric Hall, Ms. Ellen Baldridge, Mr. Steve Anderson, and Mr. Thomas Talbot.
- Samet J: The perspective of the national research council’s committee on research priorities for airborne particulate matter. J Toxicol Environ Health. 2005, 68 (Suppl 13): 1063-1067.View Article
- Rom WN, Samet JM: Small particles with big effects. Am J Respir Crit Care Med. 2006, 173: 365-366. 10.1164/rccm.2601003.PubMedView Article
- Dockery DW, Pope CA: Acute respiratory effects of particulate air pollution. Annu Rev Public Health. 1994, 15: 107-132. 10.1146/annurev.pu.15.050194.000543.PubMedView Article
- Dockery DW, Pope CA, Xu X: An association between air pollution and mortality in six U.S. cities. N Engl J Med. 1993, 329: 1753-1759. 10.1056/NEJM199312093292401.PubMedView Article
- Pope CA, Burnett RT, Thun MJ: Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. J Am Med Assoc. 2002, 287: 1132-1141. 10.1001/jama.287.9.1132.View Article
- Dominici F, Peng RD, Bell ML, Pham L, McDermott A, Zeger SL: Fine particulate air pollution and hospital admission for cardiovascular and respiratory diseases. J Am Med Assoc. 2006, 295 (Suppl 10): 1127-1134.View Article
- Laden F, Schwartz J, Speizer FE, Dockery DW: Reduction in fine particulate air pollution and mortality: extended follow-up of the Harvard six cities study. Am J Respir Crit Care Med. 2006, 173: 667-672.PubMedPubMed CentralView Article
- Pew Charitable Trusts: Public Opinion research on Public Health. 1999, Washington, DC: Mellman Group Inc. and Public Opinion Strategies Inc
- McGeehin MA, Qualters JR, Niskar AS: National environmental public health tracking program: bridging the information gap. Environ Health Perspect. 2004, 112 (Suppl 14): 1409-1413.PubMedPubMed CentralView Article
- National Ambient Air Quality Standards(NAAQS). http://www.epa.gov/air/criteria.html,
- Environmental Public Health Tracking Network- Monitoring location maps. http://ephtracking.cdc.gov/showAirMonitor.action,
- Ambient Monitoring Technology Information Center – Sampling schedule Calendar. http://www.epa.gov/ttnamti1/calendar.html,
- Federal Advisory Committee Act – Ozone, PM, Regional Haze Implementation. http://www.epa.gov/ttn/faca/,
- McMillan N, Holland DM, Morara M, Feng J: Combining different sources of particulate data using Bayesian space-time modeling. Environmetrics. 2009, 10: 1002-
- Hibbert JD, Liese AD, Lawson A, Porter DE, Puett RC, Standiford D, Liu L, Dabelea D: Evaluating geographic imputation approaches for zip code level data: an application to a study of pediatric diabetes. Int J Health Geogr. 2009, 8: 54-10.1186/1476-072X-8-54.PubMedPubMed CentralView Article
- Saporito S, Chavers JM, Nixon LC, McQuiddy MR: From here to there: methods of allocating data between census geography and socially meaningful areas. Soc Sci Res. 2007, 36: 897-920. 10.1016/j.ssresearch.2006.05.004.View Article
- Andersson N, Mitchell S: Epidemiological geomatics in evaluation of mine risk education. Int J Health Geogr. 2006, 5: 1-10.1186/1476-072X-5-1.PubMedPubMed CentralView Article
- Chakraborty J, Armstrong MP: Exploring the use of buffer analysis for the identification of impacted areas in environmental equity assessment. Cartography and Geographic Inf Syst . 1997, 24 (Suppl 3): 145-157.View Article
- Zandborgen PA, Chakraborty J: Improving environmental exposure analysis using cumulative distribution functions and individual geocoding. Int J Health Geogr. 2006, 5: 23-10.1186/1476-072X-5-23.View Article
- Vaidyanathan A, Staley F, Shire J, Muthukumar S, Meyer PA, Brown MJ: Screening for Lead Poisoning: A geospatial approach to determine testing of children in at-risk neighborhoods. J Pediatr. 2009, 154: 409-414. 10.1016/j.jpeds.2008.09.027.PubMedView Article
- Rodgers JL, Nicewander WA: Thirteen ways to look at the correlation coefficient. Am Statistician. 1988, 42 (1): 59-66.View Article
- Noether GE: Why Kendall Tau?. http://www.rsscse-edu.org.uk/tsj/bts/noether/text.html,
- Kendall MG, Gibbons JD: Rank Correlation Methods. 1990, London: Arnold, 5
- SAS Institute Inc: SAS/STAT® 9.2 User’s Guide. 2011, Cary, NC: SAS Institute Inc
- Appel KW, Bhave PV, Gilliand AB, Sarwar G, Roselle SJ: Evaluation of community multiscale air quality (CMAQ) model version 4.5: sensitivities impacting model performance; part II-particulate matter. J Atmospheric Environ. 2008, 42: 6057-6066. 10.1016/j.atmosenv.2008.03.036.View Article
- Isakov V, Irwin J, Ching J: Using CMAQ for exposure modeling and characterizing the subgrid variability for exposure estimates. J Appl Meteor Climatol. 2007, 46: 1354-1371. 10.1175/JAM2538.1.View Article
- U.S. Environmental Protection Agency: CMAQ Model Performance Evaluation. http://www.epa.gov/cair/pdfs/CMAQ_Evaluation.pdf,
- Lim SS, Vos T, Flaxman AD, Danaei G: A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2012, 15;380 (9859): 2224-2260.View Article
- Denby B, Garcia V, Holland DM, Hogrefe C: Integration of air quality modeling and monitoring data for enhanced health exposure assessment. EM: AWMA. 2009, 10: 46-49.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.