Monitoring Insights

Relative accuracy in EPA CAMD's Power Sector Emissions
Data

May 10th, 2022

40 CFR part 75 requires regular quality
assurance (QA) testing to ensure that
continuous emission monitoring systems
(CEMS) are providing accurate, consistent,
and reliable data.

There are several QA tests that power plant operators
are required to perform on the CEMS. One of the
required tests is a relative accuracy test audit, or RATA—
a test used to determine the relative accuracy of sulfur
dioxide (S02), nitrogen oxides (N0X), and carbon dioxide
(C02) or oxygen (02) concentration measurements, and
volumetric flow measurements of the flue gas (known as
stack gas flow). This monitoring insights analysis reviews
RATA results between 2017 and 2021 to assess the
performance of the part 75 monitoring systems.

Relative accuracy test audits are used to
quality assure continuous emission
monitoring systems

What is relative accuracy?

As defined in 40 CFR part 72.2, relative accuracy is "a
statistic designed to provide a measure of the
systematic and random errors associated with data from
continuous emission monitoring systems". Relative
accuracy represents the difference between the
measurements of a CEMS installed at a facility to data
collected concurrently with an EPA emission test
method. The relative accuracy is expressed as a percent
difference (% RA) between a facility's CEMS values and
the applicable reference method test values.

oEPA


-------
Monitoring Insights

Relative accuracy is based on independent
measurements

How is relative accuracy determined?

Relative accuracy is determined by comparing emissions data recorded by a
CEMS to data collected concurrently with an EPA reference method test.1 The
EPA reference methods generally involve a qualified team of stack testers2
using a portable CEMS to measure emissions at a facility.

The relative accuracy is calculated as follows:

D _ | Mean ref. method value—Mean CEMS value\ + confidence coefficient 1 nn

KA —			-—-	-	 X 1UU

Mean ref. method value

1	Refer to section 6.5.10 Reference Methods of appendix A to part 75 for
more information about reference method tests.

2	A qualified individual who meets certain competency standards (e.g.,
ASTM D7036), as certified by an Air Emission Testing Body (AETB), must be
onsite during the RATA.

What are acceptable RA
values?

To pass a RATA, the primary
performance specification requires a
relative accuracy of less than or equal to
10.0% for pollutant gas monitors (e.g.,
S02, NOx, C02, 02) and stack gas flow.
However, if the relative accuracy is less
than or equal to 7.5%, a facility may
qualify to conduct RATAs less frequently.

Facilities with low pollutant
concentrations (or emission rate for NOx
Ib/mmBtu) can pass a RATA based on
the primary performance specifications
expressed as a percentage, or an
alternate performance specifications
expressed as a pollutant concentration
(or emission rate for NOx Ib/mmBtu).

oEPA

2


-------
Monitoring Insights

Relative accuracy s improving or stable for all parameters

Distribution of relative accuracy by year and parameter

C02 concentration	NOx rate

02 concentration

10--

O -¦

2017 2018 2019 2020 2021

The red line indicates the
passing value of 10% and the
blue line indicates 7.5%, the
value to qualify for a reduced
RATA frequency

2017 2018 2019 2020 2021

v>EPA

2017 2018 2019 2020 2021
Year

For information about how to read these figures, refer to page 9


-------
Monitoring Insights

Results and Conclusions

Relative accuracy values are lower than
required

The relative accuracies for all parameters are
concentrated below 7.5%, meaning that most
units qualify for reduced RATA frequency.

A lower RA indicates a smaller difference between the
measured value and the reference method. For
volumetric flow and C02, the median of the data is
concentrated around 2%. ForS02, the median of the
data is concentrated closer to 3%. 02data was
concentrated around 3% in 2017 but has recently been
closer to 2%. Note that 02 data is more variable because
it has the fewest number of RATA tests. For NOx rate, the
data is concentrated around 3.5%. The median RAfor
NOx rate is likely higher because two monitors are
required to measure the value (NOx concentration and
diluent concentration).

oEPA

Relative accuracy values are constant or
improving

For parameters C02, S02, N0X rate, and stack gas flow,
the median RA and spread of data is relatively
consistent from 2015 through 2020.

For 02, the median RA trends down, indicating an
improvement, and the data spread (i.e., interquartile
range) declines starting in 2017. The key finding is that
relatively accuracy values are improving—getting smaller
and more uniform.

4


-------
Monitoring Insights

How does relative accuracy affect the data?

If the RATA results indicate the CEMS is reading lower
than the reference test method, a bias adjustment factor
(BAF) may be applied to subsequent emission
measurements until the next RATA. The BAF is designed
to ensure there is not an underreporting of emissions. If
a BAF is required, each subsequent emission
measurement is multiplied by the BAF.

The required frequency for performing a RATA is
dependent on the results of the previous RATA. Section
2.3 in appendix B to part 75 defines the standard RATA
frequency as once every two QA operating quarters or
semi-annually; however, a unit may qualify for a reduced
frequency of once every four QA operating quarters or
annually if the relative accuracy is less than or equal to
7.5% for pollutant monitors and stack gas flow.

What happens if a RATA is failed,
performed incorrectly, or not completed?

If a RATA is failed, performed incorrectly, or not
completed by the required deadline, the facility must
deem the CEMS measurements as invalid and use
substitute data methods described in 40 CFR 75.30-37
to calculate emissions until the CEMS passes a
subsequent RATA. More information about validity of
CEMS data is available in the analysis of percent monitor
data availability, and more information about substitute
data is available in the analysis on substitute data.

A failed or missing RATA can also lead to a field audit of
the CEMS by state and/or federal regulators.

oEPA

5


-------
Monitoring Insights

For more information about the data or this analysis

EPA's part 75 monitoring and reporting
program

¦	40 CFR part 75—Continuous Emission Monitoring

¦	Plain English Guide to Part 75 (PDF)

¦	EPA CAMP power sector programs—progress reports

Power Sector Emissions Data

¦	CAMD's Power Sector Emission Data

¦	CAMD's Power Sector Emissions Data Guide (PDF)

Contact information

Stacey Zintgraff

EPA's Clean Air Markets Division

202-564-2204

z i nteraff .sta cev@ e oa. gov

oEPA

6


-------
Monitoring Insights

Analytical methodology

This analysis was completed in R. If you would like to review the code or

source data, contact Stacev Zinteraff.

The steps in this analysis include:

1.	Compiled all RATA results from 2017 through 2021, including relative
accuracy, parameter (e.g., pollutant), year, and test result

2.	Filter out parameters not included in this analysis, tests that passed using
alternate performance specifications (qualifying units only), and aborted tests
Note that failed RATAs are included in this analysis.

3.	Calculate quartiles and median for each year and parameter.

4.	Create box and whisker plots. Outliers, including failed tests, were excluded
from the plot on page 3, refer to appendix B on page 9 for a box and whisker
plot including outliers.

By the numbers

Test counts in this analysis
2021

¦	C02: 578 RAs (median 2.1% RA)

¦	S02: 324 RAs (median 3.5% RA)

¦	02:19 RAs (median 2.3% RA)

¦	N0X rate: 1,901 RAs (median 3.1% RA)

¦	Flow: 1,456 RAs (median 2.2% RA)

2017

¦	C02: 757 RAs (median 1.9% RA)

¦	S02: 458 RAs (median 3.0% RA)

¦	02: 23 RAs (median 2.9% RA)

¦	N0X: 1,985 RAs (median 3.2% RA)

¦	Flow: 1,905 RAs (median 2.3% RA)

v>EPA

7


-------
Monitoring Insights

Appendix A: How to read a box plot

Box plot key

• Outliers/Maximum

1.5 times the upper IQR

75th percentile
Median

25th percentile
Minimum

The RATA results for all reporters are displayed using a "box plot"
also known as a "box-and-whisker plot".

A box plot is a method to depict groups of numerical data in quartiles. It illustrates the

distribution, central tendency, and variability.

In the example box plot on the left:

¦	The blue box represents the middle half of all values—also known as the interquartile range
(IQR)—those that fall between the 25th and 75th percentile.

¦	The grey horizontal line represents the median value (i.e., the 50th percentile value).

¦	The green vertical line, or top whisker, represents the values between the 75th percentile
and 1.5 times the upper interquartile range.

¦	The red vertical line, or bottom whisker, represents the values between the minimum and
the 25th percentile.

¦	The black dot represents outliers or values outside 1.5 times the IQR. The highest dot
represents the maximum value.

Each box plot provides visual representations of both the magnitude and variability of

values for all reporters in a given year in a single chart.

oEPA


-------
Monitoring Insights

Appendix B: Box and Whisker Plot including outliers

Distribution of relative accuracy by year and parameter

C02 concentration

NOx rate

02 concentration

25-
20-
15-
10-

„ 5-

1Z2

l

f-



>

o

n o

=5

O

o
<

jro
a>
CT

o-

i— —r

T

I	I	I	I

2017 2018 2019 2020 2021

2017 2010 2019 2020 2021
Year

v>EPA

S E

S02 concentration

stack gas flow























•
t





















•















	J

•

: . • :



...









	1 __T I

	1 _























2017 201S 2019 2020 2021


-------