-------
ACCOUNTABILITY: DECREASES DUE TO THE ACID RAIN PROGRAM

SO2 is both a local and regional pollutant, so intra-urban differences in ambient
concentrations are expected. If local sources are close to monitors, they may obscure long-term
regional trends. Continuous SO2 data for 1993-2005 are available from the EPA's Air Quality
System (AQS) for five sites in the Detroit area. National S02 emission trend estimates are
available from 1993-2002,1 and power generating facility emissions are available from
1995-2005. In addition to S02, sulfate aerosol, visibility extinction from sulfate aerosol, and
acid deposition should be impacted by the Acid Rain Program. To understand the multi-media
effect of S02 regulations, 1993-2005 data for ambient sulfate aerosol concentration, sulfur
deposition, and light extinction due to sulfate aerosol were obtained from the Ann Arbor,
Michigan, CASTNET site.

Between 1993 and 2002, national SO2 emissions reductions were gradual. Emissions
from power generating facilities in Michigan and regionally showed a large decrease in
concentrations from 1998-2001. Specific dates and locations of local SO2 controls in the Detroit
area are not known; regional controls may impact concentrations in the Detroit area.

Overall, all sites showed a decrease in ambient SO2 concentrations from 1993 to 2005
(Figure 9). Three-year averages were used for most of this analysis to reduce year-to-year
variability. A large decrease (about 30%) in year-to-year SO2 concentrations is evident between
1994 and 1995, corresponding to the largest decrease in year-to-year emissions nationally (28%).
Changes noted include a

•	14% decrease in Michigan SO2 emissions from power generation (1995-1997 to
2003-2005);

•	26% region-wide decrease in SO2 emissions from power generation (1995-1997 to 2003-
2005);

•	26% decrease in average SO2 concentrations in Detroit (1993-1995 to 2003-2005);

•	24% decrease in sulfate concentrations in Ann Arbor (1991-1993 to 2003-2005);

•	7% decrease in sulfate concentrations in Allen Park (2001-2003 to 2003-2005);

•	26% decrease in total sulfur deposition in Ann Arbor (1991-1993 to 2003-2005); and

•	17%) decrease in light extinction due to sulfate (1991-1993 to 2003-2005).

Additional details on this analysis are available in Appendices B and C.

1 National Emission Inventory; .

10

-------
Legend notes: (1) Excluding 1996-1997 (incomplete data)

(2)	Ann Arbor CASTNET data (representative of other Michigan sites)

(3)	STN data

(4)	utility emissions from

http://www.epa.gov/airmarkets/emissions/prelimarp/index.html.

Axis note: Light extinction calculated from b=(3)ft(RH)[S042"], where RH is relative
humidity.

Figure 9. Annual total SO2 emissions in Michigan and the regional area (black
and blue lines); three-year averaged concentrations of sulfur species in Michigan
(final year of three-year average is indicated on the x-axis).

ACCOUNTABILITY: INITIAL EVALUATION OF THE NOx SIP CALL

N0X is both a locally and regionally emitted and distributed pollutant. Intra-urban
differences in N0X concentrations are likely, and mobile sources typically produce the most N0X
in urban areas. N0X from power generation (the target of the N0X SIP Call) constitutes about
30% of total NOx emissions in the Detroit area (U.S. Environmental Protection Agency, 2004),
so changes in NOx emissions from other sources (such as mobile sources) could confound how
NOx concentration trends (or lack of trends) are interpreted. In addition, NOx data are only
available from two sites in the Detroit area for 2002-2005, and regulations in Michigan to reduce
NOx were not implemented until 2004. Decreases in NOx concentrations in Detroit because of
these regulations are probably not large enough to be noticeable with such a short data record.
Ozone concentrations are expected to decrease corresponding to a decrease in NOx
concentrations. Nitrate and PM2.5 mass are not expected to change as a result of this regulation
because the regulation is only in effect during summer months; when nitrate formation is
minimal and the nitrate contribution to PM2.5 mass is very small.

11

-------
When summer-only yearly box whisker plots were examined for the two Detroit NOx
measurement sites, East 7 Mile and Linwood, no consistent change in concentrations across sites
after 2004 was seen. At East 7 Mile, NOx concentrations decreased in 2004, followed by an
increase in 2005 (Figure 10). This change in ambient NOx concentrations was not observed at
the Linwood site, even though it is closer to NOx point sources. Differences between the NOx
concentrations at these two sites are likely due to differences in the proximity of both mobile and
point NOx sources to the monitors. Concentrations were segregated by hour to examine rush-
hour (i.e., mobile source-dominated) versus non-rush hour concentrations, nighttime hour
(lowest mobile source contribution) concentrations, and daytime hour concentrations, but no
consistent trend was evident. Because mobile source activity is lower on weekends but power
generation activity generally is not, ambient NOx concentrations were also segregated by day of
week and hour to determine whether examining periods when mobile source emissions are low
could reveal trends from power generation sources. Again, no consistent trend was observed at
sites from which data were available.

Because the large mobile source contribution to NOx may confound any changes in
Detroit urban NOx concentrations due to the NOx SIP Call, wind direction analysis was also
performed to isolate NOx point sources in the ambient data record. Data were divided by wind
direction into three groups: (1) 180-225 degrees, expected to be dominated by point sources,
(2) winds from the Detroit area—mobile-dominated, and (3) winds from Canada—no emissions
information available. Concentrations were significantly higher at East 7 Mile and Linwood
when the wind was from 180-225 degrees, supporting the hypothesis that large NOx point
sources in this direction impact ambient concentrations. However, no significant year-to-year
change in concentrations from this sector was evident at either site. Data were divided by hours
to further isolate the point source-versus-mobile source contribution, but no consistent change
across years was seen with morning hourly data only or nighttime hourly data only.

Ratio analysis was also conducted using ratios of NOx and mobile source-dominated
pollutants. If the mobile source species (benzene and total hydrocarbons) do not change with
time, a change in their ratios to NOx could indicate a change in the point source contribution.
However, no consistent year-to-year change was seen in these ratios at either site (e.g., see
Figure 9). Additional details on this analysis are available in Appendices B and C.

12

-------
(a) E 7 Mile, 24-hr avg NOx

100

b) E 7 Mile, TNMOC:

2002

2003

2004

2005

2002

2003

2004

2005

Figure 10. Notched box plots of concentrations of (a) NOx (ppb) and (b) the total
nonmethane organic compounds (TNMOC):NOx ratio at the East 7 Mile site;
summer of 2002-2005 data used.

ACCOUNTABILITY: HOW DO PEAK EMISSIONS IMPACT OZONE?

An accountability analysis to examine the connection among weather, peak emissions,
and resulting air quality (e.g., electric generating unit [EGU] S02 peak emissions and PM2.5;
EGU NOx peak emissions and ozone) was performed for 2005. The year 2005 was selected
because air quality was worse in 2005 than in 2004, and ambient NOx plus daily emissions data
were available for 2005. Peak emissions were defined as emissions from facilities that operated
for less than 1,000 hours during the 2005 ozone season, i.e., facilities that were only operated
when demand was highest. However, as seen in Figure 11, several large power plants in the
Detroit area (including Monroe and J H Campbell) are not considered peak emitters and typically
account for almost all of the NOx emissions; therefore, peak emissions were not a large fraction
of total NOx emissions in this area. Additional details on this analysis are available in
Appendix D.

13

-------
120



450

(/>

400

£

O

350

V)

300

C

o

250

"3>

200

"g

150

u

X

100

O
2

50



0

I Fteak NOx Bniss ion s
Base NC& Emissions

A

a

\yV

Apr

May

Jun

Jul

Aug

Sep

Duly temperature is daily max temperature for the Detroit MSA (source: AQS)

AQI is based an daily max 8-hour 03 value for Detroit MSA (source: AirNowTech)

Output is total daily output for the Monroe Power Plant (source: EPA Clean Air Markets; h&pJtfripub.epQ.yuv4/drrfndex.cfm?fij9eaction=emissiQn5. wizard)
Emissions are from all -facilities in Michigan

Peak NOx emissions are from facilities that operated for <1000 hours during the 2005 03 season

(source: EPA Clean Air Markets; tdip:Atfpuixeps.go\^gdm^nd&i.cfm?fus^cik3fj=cmtssians.wizard)

Base NOx emissions are from facilities that operated far >1000 hours during the 2005 Q3 season

(source: EPA Clean Air Markets; htipVAzfpuiiepa.gDu/gdm/inciw.cfm?fuseactkm=*mis3ian&-wizard)

Figure 11. Daily ozone Air Quality Index (AQI), maximum daily temperature,
EGU output (1000 MWh), and peak and base NOx emissions during the 2005
ozone season (April through September). Peak emissions are from facilities that
operated for less than 1,000 hours during the 2005 ozone season.

TRENDS IN ANNUAL AVERAGES OF MOBILE SOURCE RELATED SPECIES

To explore whether the ambient data indicate a change in source contributions over time,
annual average concentrations of mobile source related species OC, EC, 1,3-butadiene, and
benzene) were examined at Allen Park and Southwest High School. Data were investigated
using both a t-test, which is used to detect a significant difference between two years of data, and
an f-test, which is used to detect whether the slope of average concentrations by year is
significantly different than zero. In both cases, a p-value of less than 0.05, corresponding to a
95% confidence level, was considered significant. Benzene and 1,3-butadiene did not have
significant trends at either site (Figure 12). Only 1,3-butadiene at Southwest High School had a
significant change from 2004 to 2005 (p-value <0.01). OC and EC had significant trends at
Allen Park; however, this may be misleading because 2002 was the only year that was
significantly different than other years. Additional years of data are needed to confirm this trend.
Weekday versus weekend and seasonal differences were also examined for these pollutants.
Benzene and 1,3-butadiene had no significant differences by day of week or season at either site.

14

-------
EC at Allen Park was significantly lower in the spring and winter months compared with the rest
of the year as well as on weekends compared with weekdays. The lower concentrations in the
cooler months is most likely due to less atmospheric mixing, but the lower concentrations on the
weekends implies that EC is dominated by diesel emissions, which are reduced on weekends.
OC was significantly higher in the summer months but had no difference in concentrations on
weekends compared with weekdays. Higher concentrations in the warmer months are most
likely due to increased photochemistry.

4.5

4
3.5

£ 3

o>

— 2.5
O

2001	2002 2003 2004 2005

Figure 12. Annual average concentrations of mobile source-related compounds at
Allen Park and Southwest High School.

EXPLORATORY APPORTIONMENT WITH GASEOUS AIR TOXICS DATA

To identify sources of ambient air toxics to further understand the relationships among
source types and air pollutants, gaseous air toxics data from the SWHS site were explored using
PMF. Samples were collected during 2001 to 2006; 150 samples were suitable for PMF and
20 species were available and had sufficient data above detection. Uncertainty estimates were
developed using duplicate samples from the Environmental Laboratory of the Michigan
Department of Environmental Quality. Because multiple minimum detection limits (MDLs)
were reported for each species, the mode MDL was used to prevent introducing a false signal if
many of the data are below detection. Data below the detection limit were substituted with MDL
divided by 2 and given uncertainties of 5/6 times the MDL. Missing data were replaced with the
median and given an uncertainty of 4 times the median.

-A—

-0—
-B—
+-
¦A—

EC (Allen Park)

OC (Allen Park)
1,3-Butadiene (Allen Park)
Benzene (Allen Park)
1,3-Butadiene (SWHS)
Benzene (SWHS)

15

-------
Three- to eight-factor solutions were explored. The four-factor solution was chosen as
the final solution. Factors included chlorinated compounds, acetonitrile, secondary formation
(including acetaldehyde and formaldehyde), and mobile exhaust (including benzenes, xylenes,
and toluene) (Figure 13). Additional factors could potentially be resolved using additional
species above detection, species that are unique source tracers, and more samples. For example,
no biogenic tracer was available for these analyses, therefore no biogenic factor could be
resolved even though biogenic emissions are expected to be an important contributor to VOCs.

(0

o
'o

-------
Figure 14. Average PMF contribution of factors resolved for SWHS gaseous air
toxics data; all values are normalized to 2001 except the acetonitrile factor which
was zero in 2001; therefore, acetonitrile averages are normalized to 2002.

CONCLUSIONS

Several methods were applied to multipollutant data sets in the Detroit area to better
understand trends over time and relationships to emissions controls. The influence of the largest
regional control affecting PM2.5 concentrations, the Acid Rain Program, was observed in the
ambient sulfur and sulfur-related data. Impacts of the NOx SIP Call were not observed in the
ambient urban NOx data, but data were limited and the sites were dominated by mobile source
emissions. Results from a range of ozone, NOx, PM2.5, and air toxics analyses indicate little
impact from local controls (if any were implemented) and few relationships among pollutants
through mobile source emissions. Trends and source apportionment analyses can be enhanced as
additional data become available.

REFERENCES

Brown S.G. (2006) Accountability—Tracking and evaluating air quality programs, Task 3—
Detroit, Michigan. Technical memorandum prepared for the U.S. Environmental
Protection Agency, Research Triangle Park, NC, by Sonoma Technology, Inc., Petaluma,
CA, STI-905213.05-3040-TM, September.

17

-------
Frank N.H. (2006) Retained nitrate, hydrated sulfates, and carbonaceous mass in federal

reference method PM2.5 for six eastern U.S. cities. J Air Waste Manag Assoc 56 (4), 500-
511.

Hafner H.R. and Roberts P.T. (2006) A technical approach using ambient data to track and
evaluate air quality programs. Draft final report prepared for the U.S. Environmental
Protection Agency, Research Triangle Park, NC, by Sonoma Technology, Inc., Petaluma,
CA, STI-905213.02-2918-DFR, March.

Paatero P. and Tapper U. (1994) Positive matrix factorization: a non-negative factor model with
optimal utilization of error estimates of data values. Environmetries 5, 111-126.

Paatero P. (1997) Least squares formulation of robust non-negative factor analysis.
Chemometries and Intelligent Laboratory Systems 37,23-35.

Rubin J.I., Brown S.G., Wade K.S., and Hafner H.R. (2006) Apportionment of PM2.5 and air

toxics in Detroit, Michigan. Final report prepared for the U.S. Environmental Protection
Agency, Research Triangle Park, NC, by Sonoma Technology, Inc., Petaluma, CA, STI-
906201.06-3103-FR, December.

U.S. Environmental Protection Agency (2004) National Emission Inventory (NEI) Database for
Criteria and Hazardous Air Pollutants. Available on the Internet at
.

Wade K.S., Rubin J.I., McCarthy M.C., and Hafner H.R. (2007) Developing uncertainty

estimates for gaseous air pollutants (HAPs). Technical memorandum prepared for the
U.S. Environmental Protection Agency, Research Triangle Park, NC, by Sonoma
Technology, Inc., Petaluma, CA, STI-905306.02-3230, September 19.

18

-------