What can affect AOD-PM(2.5) association?

Although satellite remote sensing has advanced significantly in recent years, there are inherent weaknesses in the use of this technology. The association between satellite-based aerosol optical depth (AODS) and air pollution monitored on the ground can be influenced by a number of factors. In their article, Paciorek and Liu (2009) highlighted the weaknesses of AODS to predict the spatial distribution of fine particulate matter ≤ 2.5 μm in aerodynamic diameter (PM2.5). It is a timely article given the increasing importance of indirect methods, including satellite data, to estimate air quality because of scarce and ad hoc spatial–temporal coverage of air pollution monitored by federal regulatory methods. It is important that the robustness of these methods is evaluated, and Paciorek and Liu’s article is such an attempt. However, they failed to address the role of five major factors that can influence the AODS–PM2.5 association. These factors include decomposition of AODS by aerosol types, mismatch in spatial–temporal resolution, collocation and integration of AODS and PM2.5 data, and control for spatial–temporal structure in the statistical model. Consequently, the weaknesses in Paciorek and Liu’s study lead me to question their findings.

The columnar measurement of AODS consists of aerosols generated by anthropogenic (human) sources (AODSh), such as emissions from industries and vehicles, and natural sources (AODSn), such as water vapor or dust in the air. AODSn that constitutes a large fraction of AODS is influenced by moving large air masses and observes a strong spatial and temporal structure. The concentration of PM2.5, however, can vary significantly within short distances. Therefore, there is a significant mismatch in the magnitude and extent of spatial and temporal variability of AODSn and AODSh; without an adequate control for AODSn, it is difficult to develop a reliable PM2.5 predictive model using AODS (Kumar et al. 2008).

Paciorek and Liu (2009) recognized that the spatial–temporal resolutions of AODS and PM2.5 they used were different, but they did not address how the mismatch in the spatial–temporal resolutions of these data can influence their association. The spatial resolutions of MISR (multiangle imaging spectroradiometer), MODIS (moderate resolution imaging spectroradiometer), and GEOS (geostationary operational environmental satellite) AOD were 17.6 km, 10 km, and 4 km, respectively, and PM2.5 data were point measurements aggregated across 24 hr. A recent study suggests the strength of the AODS–PM2.5 association diminishes with the increase in time interval used for their aggregation (Kumar et al. 2007). It would have been useful for Paciorek and Liu (2009) to document the implications of the spatial–temporal resolutions and aggregation of AOD and PM2.5 (data they used) on their findings.

AODS retrieval and PM2.5 are not available on the same days: AODS retrieval is not possible on cloudy days, and PM2.5 data are recorded every third or sixth day. It seems that Paciorek and Liu (2009) averaged all AODS at 4-km pixel (i.e., 16 km2 area; monthly and yearly) and all PM2.5 (in the pixels where a monitoring station was situated). This could have resulted in a weak association between AODS and PM2.5, because there were systematic temporal gaps in both AODS and PM2.5 data sets. A reasonable approach to address this problem is to aggregate AODS-PM2.5 data for those days only when both AODS and PM2.5 are available.

Paciorek and Liu’s method for aggregating 17.6-km and 10-km AODS to a 4-km pixel seems problematic. First, a radiative transfer model is used to retrieve AODS (Remer et al. 2006) which removes pixels with the upper 50% and lower 20% of the reflectance values. This removal can be systematic. For example, pixels with high reflectivity (such as buildings and roads) are more likely to be removed than the vegetated pixels (i.e., pixels under vegetation canopy). Thus, the centroid of a 10-km AODS pixel may not represent the AODS value for the entire 10-km area. Second, AODS registers a strong spatial–temporal autocorrelation. Thus, time–space kriging that utilizes large number of data points is appropriate for AODS aggregation rather than a single AODS value to avoid an area specific bias.

The robustness of AODS retrieval is evaluated by its comparison with the AOD recorded by sunphotometers at AERONET sites (AODA) (NASA 2007). The spatial resolution at which AODS is retrieved and the spatial–temporal intervals within which these data are aggregated may directly influence its comparison with the AODA . This, in turn, can influence the association between AODS and PM2.5. Recent literature suggests that 1-km and 5-km AODS observe a significantly better association with PM2.5 monitored on the ground than the 10-km AODS (Kumar et al. 2007; Li et al. 2005). Therefore, the optimal spatial resolution of AODS retrieval and the optimal spatial and temporal intervals for aggregating these data are critically important for developing time–space resolved estimates of air quality with the aid of AODS.

Because meteorologic conditions are largely influenced by the prevailing air masses and do not vary significantly within thousands of miles for a short period of time, the AODSn component of AODS is likely to have a strong spatial–temporal structure. PM2.5 that constitutes particulate mass associated with anthropogenic factors, however, varies significantly within short distances from emission sources. Therefore, to develop a PM2.5 predictive model it is important that only AODSh is used instead of AODSn. If such data are not available, an alternative is to indirectly control for AODSn and its associated spatial–temporal structure. Otherwise the predicted PM2.5 surface is likely to have an unrealistic spatial trend, as reported by Paciorek and Liu (2009), as well as unrealistic temporal trends.