Satellite sulfur dioxide (SO2) measurements from the Ozone Monitoring Instrument (OMI) satellite sensor, averaged over a period of several years, were compared with emissions inventories for major US sources. Low- and high- spatial frequency filtration was applied to OMI data to reduce the noise and bias to enhance and reveal weak SO2 signals that are otherwise not readily apparent. Averaging a large number of individual observations enables the study of SO2 spatial distributions near larger SO2 emissions sources with an effective resolution superior to that of an individual OMI observation and even to obtain rough estimates of the emissions level from those sources. It is demonstrated that individual sources (or multiple sources within 50 km) with annual SO2 emissions greater than about 70 kT y−1 produce a statistically significant signal in 3-year averaged OMI data. A correlation of 0.93 was found between OMI SO2 integrated around the source and the annual SO2 emission rate for the sources greater than 70 kT y−1. OMI SO2 data also indicate a 40% decline in SO2 values over the largest US coal power plants between 2005–2007 and 2008–2010, a value that is consistent with the reported 46% reduction in annual emissions due to the implementation of new SO2 pollution control measures over this period.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 OMI provides the best horizontal resolution (13 × 24 km2 footprint at nadir) among instruments in its class. However, even with this resolution, most anthropogenic sources produce elevated SO2 levels that are detectable only within the co-located space of just one or two pixels. The same limitation applies to the standard archived OMI level 2G grid with resolution of 0.125° by 0.125°. This study employs a different analysis technique in which a large number of individual observations are used in an attempt to quantify the SO2 spatial distributions near larger SO2 emissions sources.
2. Data and Analysis
2.1. SO2 Emissions Sources
 The top 100 largest US individual sources (according to the U.S. national emissions inventory for 2005: see http://www.epa.gov/ttn/chief/eiinformation.html) were examined in this study. The majority of these sites are coal-burning power plants. As will be demonstrated later, only ∼40 of the largest sources, those with SO2 emissions levels greater than ∼60 kT y−1, produce a statistically significant signal in the OMI data. The inventory data for these sources, unlike most emissions data, were based on direct stack measurements using Continuous Emissions Monitoring Systems as mandated by Title IV of the 1990 U.S. Clean Air Act Amendments (Public Law 101-549) (e.g., http://www.epa.gov/air/caa/title4.html). For the comparison with satellite data, it was assumed that emissions from these sources are constant throughout the year.
2.2. OMI SO2 Data Product
 The Dutch-Finnish-built Ozone Monitoring Instrument (OMI) is a nadir-viewing, UV-visible spectrometer [Levelt et al., 2006] that has been observing aerosols and trace gases, including SO2, from the NASA EOS Aura satellite platform since 2004 [Schoeberl et al., 2006]. This study focuses on anthropogenic pollution sources that emit SO2 to the planetary boundary layer (PBL), and therefore a data product specifically designed to represent boundary-layer SO2 was used [Krotkov et al., 2006]. Values are given as total column SO2 retrievals optimized for the PBL in Dobson Units (DU), where 1 DU is equal to 2.69·1026 molec·km−2. OMI SO2 data for the period 2005–2010 were analyzed. The OMI measures 60 cross-track positions (pixels), and the pixel size varies depending on the track position from 13 × 24 km2 at nadir to about 28 × 150 km2 at the outermost swath angle. Data from the first and last 10 track positions were excluded from the analysis to limit the across-track pixel width to about 40 km. Beginning in 2007, some track positions were affected by field-of–view blockage and scattered light (so called, “row anomaly”, see http://www.knmi.nl/omi/research/product/rowanomaly-background.php). The affected pixels were excluded from the analysis. Only clear sky data, defined as having a cloud radiance fraction (across each pixel) less than 20%, were used. To exclude cases of transient volcanic SO2, the range of analyzed values were limited to a maximum of 5 DU. Furthermore, results presented in this study are based on May–August data only. While results for the rest of the year are similar, they exhibit larger uncertainties due to various factors such as larger solar zenith angles, variable surface albedo (snow), higher ozone optical depth, etc.
2.3. Spatial Smoothing and Local Bias Correction
Figure 1 shows the mean column SO2 values for the 2005–2007 period plotted as a function of a distance between the OMI pixel center and the location of two large emissions sources: the largest US SO2 source (Bowen power plant in Georgia, estimated at 170 kT y−1) and the 20th-largest source (Belews Creek power plant in North Carolina, 88 kT y−1). As Figure 1 demonstrates, OMI data show elevated SO2 values near the emissions sources but they became insignificant beyond about 50 km, even for the largest single source used in this study. This may explain why many sources are not typically seen by OMI: SO2 from such sources affects only 1–2 pixels and the noise level is high. However, Figure 1 also suggests that it is possible to obtain a statistically significant signal by averaging a large number of individual pixels centered within a several km radius from the source. It was found that averaging over 3 years of data typically produces a statistically significant (at the 95% confidence level) mean value if the source annual emission is greater than ∼70 kT y−1, although some of the sources with the annual emissions of ∼60 kT y−1 also produce significant mean values.
 Based on this finding, we use this pixel-averaging approach to analyze the long-term mean spatial SO2 distribution near the source. For this, a geographical grid is established around the source and the average of all OMI pixels centered within a several km radius from each grid point is calculated. This approach is illustrated in Figure 2a. For Figure 2a, a 60 km × 60 km grid with a 2 km step was centered over the 10th-largest source, the John E. Amos power plant in West Virginia. An average of all SO2 data falling within a 12 km radius of the center of a grid cell was assigned to that grid point, where the pixel center is used as the location of the measurement. The resultant distribution of SO2 values, presented in Figure 2a, reveals the highest mean SO2 values occurred at grid points located within a small area around the source. Thus, this procedure provides a detailed "subpixel-resolution” spatial distribution of long-term mean SO2 value in the vicinity of the source. The choice of averaging radius determines the degree of smoothing: averaging with a large radius reduces the noise, but it also reduces the spatial resolution.
 Systematic errors in retrieved SO2 resulting from imperfect instrument calibration as well as from, for example, forward model simplifications, were substantially reduced in the present OMI algorithm by empirical corrections [Yang et al., 2007]. Nonetheless, some large-scale biases remain. It can be expected that, other factors being the same (e.g., average solar illumination angles, wind speed, cloud cover, surface albedo), sources with similar emission strengths should produce similar mean observed SO2 values. As Figure 2b shows, this is not always the case and residual biases are comparable with the mean SO2 values from the sources. Since these biases appear as large-scale patterns, they can be removed with a spatial high-pass filter. To accomplish this, SO2 values within a 300 km radius were averaged and then this mean value was subtracted. It was also found that these local biases are somewhat different from year to year so that the local bias correction was calculated for each year. Mean SO2 values around similar sources corrected in this manner are very similar as illustrated by Figure 2b.
 This combination of spatial smoothing and local bias correction can be used to produce high-resolution, long-term mean SO2 maps. Figure 3a presents such maps for the eastern US, where the majority of large SO2 sources are located (indicated by the black dots). Areas of high SO2 values are centered over these major emissions sources. The maps in Figure 3 were generated for two 3-year intervals, 2005–2007 and 2008–2010. A 24 km averaging-radius was used to smooth the data for this plot. A substantial decline in OMI SO2 values at major sources between the two time intervals is evident from Figure 3. This reduction is attributed to the installation of additional flue-gas desulfurization units (or “scrubbers”) at many US power plants over this period (e.g., http://www.epa.gov/airmarkets/images/CoalControls.pdf) to meet stricter emissions limits introduced by the Clean Air Interstate Rule.
 OMI data can be further used to evaluate the reduction in the measured SO2 values and then compared to the actual reported reduction in emissions levels. The sum of SO2 values from the top 40 US emissions sources was calculated (this corresponds to sources with annual emissions greater than 60 kT y−1) as a function of distance from the source for 2005–2007 and 2008–2010 and is shown in Figure 3b. For this plot, the mean SO2 value was calculated for each source as a function of distance from the source and then these mean values were added up to form the sum. The ratio between the two sums is about 0.6, indicating a 40% reduction in OMI mean SO2 values. The actual measured reduction for these same sources over this period based on emission reported to the U.S. Environmental Protection Agency (see http://camddataandmaps.epa.gov/gdm/index.cfm?fuseaction=emissions.wizard) was 46%.
3. Emissions Inventories and OMI SO2 Values
 Assuming comparable SO2 sources produce similar long-term mean OMI SO2 values, OMI SO2 can be related to emissions levels from individual sources and furthermore, provide an estimate of annual emissions. In order to quantify the total amount of SO2 near a source, a two-dimensional Gaussian function = a · f(x, y) was fit to OMI SO2 measurements within a time window and radius, where
and x and y refer to the co-ordinates of the OMI pixel center. Note that actual OMI measurements, not smoothed data described in section 2 were used for the fit. The elliptical shape of the SO2 distribution near the source is determined by parameters σx, σy, and ρ. The parameters μx and μy were included since the position of the emissions source may be different from the position of the fit maximum due to for example, prevailing winds or if the source is comprised of two closely located power plants. Since f(x, y)dxdy = 1, the parameter a represents the total observed number of SO2 molecules near the source. If is in DU, i.e., in 2.69·1026 molec km−2, and σx, σy are in km, then a is in 2.69·1026 molec. For SO2 emissions sources less than 140 kT y−1, the fitting was done using OMI pixels centered within a 40 km radius from a source and local bias was removed. For sources larger than 140 kT y−1 that affect SO2 values at greater distances, the radius was 60 km. Note that function (1) has a single maximum and therefore describes an SO2 distribution near a single source. If two or three sources are located in a close proximity (within 50 km for very large sources), they were counted as a single source with the total emissions equal to the sum of emissions from these sources. Otherwise, instances where the secondary source was located within the fitting radius were excluded from the analysis.
 The scatter plot of the total number of SO2 molecules retrieved from the OMI measurements, a, versus daily SO2 emissions strength for each source location is shown in Figure 4. The 2005–2007 period was used in Figure 4 because the emissions were fairly stable during that period. Sources emitting less that 70 kT y−1 typically produced values of a that were not statistically significant and were not shown in the plot. Thus 70 kT y−1 represents the threshold for which this methodology can be applied to the present OMI SO2 data. The correlation coefficient between total molecules and annual emission is 0.93 and this high degree of correlation implies that SO2 emissions can be estimated from OMI data using a linear regression.
 There is also a physical interpretation of this relationship since the slope of the regression line represents the effective SO2 removal time due to advection, deposition and chemical conversion to sulfate aerosols. Its value, 5 hours, is 3–5 times less than the current estimates for the eastern US in summer [Lee et al., 2011]. Possible reasons for this discrepancy include the use of a constant air mass factor (AMF) in deriving vertical columns, neglecting dry deposition, dispersion due to variable wind speeds, and statistical arguments. Moreover, it is unclear if a lifetime, obtained from a model simulation at ∼200 km resolution, is representative within several km of the emission site. A constant AMF is reasonable given that the majority of locations considered were in the eastern US and so to first order, any error in AMF will be common to most locations [Lee et al., 2009]. Clearly, detailed AMFs based on local conditions, including a realistic representation of aerosols (which are likely elevated near large pollution sources), are necessary to establish a quantitative link or to expand the analysis to other regions. Another contributing factor is that satellite retrievals may consistently underestimate the small (compared to the pixel sizes) emission plumes, because they do not usually occupy the entire pixel footprint [Yang et al., 2010]. Furthermore on average advection and deposition may be the prevailing removal mechanisms in the vicinity of the power plants. The power plants in the eastern US are located in an area of typically weak summer winds, days with stronger winds causing a faster dispersion of the SO2 plume still contribute to the overall statistics.
4. Summary and Discussion
 Pollution plumes from individual power plants are not typically detectable in daily OMI SO2 maps using standard analysis techniques. However it is possible to obtain a statistically significant signal from large sources (>70 kT y−1) by averaging OMI SO2 pixels centered in 10–20 km from the source over a period of several years. For such large sources, there is a high correlation (0.93) between the annual emissions from an individual source and mean summertime SO2 integrated over the area around that source.
 A pixel-averaging technique, i.e., averaging a large number of individual OMI pixels, together with a local bias correction (i.e., low- and high- spatial frequency filtration) were used to map the mean distribution around the major sources and to produce a composite SO2 maps for the eastern US. Such maps for 2005–2007 and 2008–2010 periods demonstrate a substantial decline in SO2. Moreover, the mean OMI SO2 values summed near 40 largest emissions sources were 40% lower in 2008–2010 than in 2005–2007, consistent with the measured 46% reduction in emissions due to the implementation of new SO2 pollution control measures at some sources over this period.
 Large errors of individual OMI SO2 retrievals and large pixel size (∼300 km2) are the main reasons why sources with emissions below 70 kT y−1 typically do not produce a statistically significant signal in OMI observations. Smaller sources may be detectable in the future if the algorithm is improved. Ultimately, the proposed technique will work best with frequent high spatial resolution measurements from the future GEO-CAPE UV imaging spectrometer on geostationary satellite [Fishman et al., 2008]. The unprecedented temporal and spatial resolution possible from geostationary vintage point will offer the best possibilities for monitoring emissions and understanding pollution processes.
 We acknowledge the NASA Earth Science Division for funding of OMI SO2 product development and analysis. The Dutch-Finnish-built OMI instrument is part of the NASA EOS Aura satellite payload. The OMI project is managed by KNMI and the Netherlands Agency for Aero-space Programs (NIVR). The US Environmental Protection Agency provided SO2 emissions data. The authors also thank two anonymous reviewers for their thorough and thoughtful comments.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.