implementing a continuous pm 2.5 federal equivalent method ...€¦ · 2009 through december 2009...

Implementing a Continuous PM2.5 Federal

Equivalent Method (FEM) for NAAQS

Comparisons and AQI Reporting in the

Missouri

Ambient Air Monitoring Network

Missouri Department of Natural Resources

Environmental Services Program

Stephen M. Hall

Air Monitoring Section

2010 Air Quality Conference, Raleigh, NC

Benefits of Using a Continuous

PM2.5 FEM

• Reduced redundancy in 24-hr integrated PM2.5

measurements

• Ability to report real time data to AirNow and have

it “match” what is reported to AQS for NAAQS

purposes

• Increases quantity of 24-hr NAAQS data

• Data Quality Indicators for the FEM are reported

consistently to AQS

• Monitor and report diurnal PM2.5 hourly concentrations

•Reduced network operating costs

St. Louis City

(9) PM2.5 Continuous Monitors

7-FEM (FDMS-8500C)

1-SHARP

1-TEOM-SES

El DoradoSprings (FEM)

(Rural &

Regional

Transport Site)

Arnold

West(FEM)

Liberty(FEM) Kansas City

St. Joseph(FEM) Ladue

(FEM)

Troost (FEM)

RG-South(FEM)

MSU

(SHARP)

Springfield

Blair Ave.(TEOM-SES)

Observations•Continuous FEMs do not always perform “exactly” the same

as agency operated PM2.5 FRMs.

•More meta data should be monitored to ensure the best FEM

performance.

•Data Quality Indicators in 40 CFR Part 58 Appendix A were

typically applied to filter based FRM/FEM collocated pairs.

FRM/FEM(continuous) pairs may yield higher CV and Bias

results due to different methods.

•Evaluate the FEM method to understand how these issues

affect the FRM/FEM comparability in your region.

•EPA’s FEM Test was performed with single filter FRMs.

(Filter preservation techniques and sampling time frames

differ between the FEM test campaigns and the way our

agency operates multi-filter sequential FRMs.)

Historic PM2.5 Method Performance in MO

“Winter” - FDMS Mass concentration data more closely

matches filter-based FRM data when ambient

temperatures are colder. Volatile/semi-volatile aerosol is

more likely retained on FRM filter before being weighed.

“Summer” - FDMS mass concentration data is generally

higher than filter based FRM when ambient temperatures

are higher. Volatile/semi-volatile aerosol not as likely to

be retained on FRM filter before being weighed.

FRM/FDMS comparability - Strongly correlated with

ambient temperature over all “seasons” in our region.

(Apparently, this was not true for the data fit for the FEM

test, but it is true for our state over all seasons.)

Historic Missouri Solution

State Method (proposed ARM): Corrects FDMS-8500B

mass concentration data in real time-based on

temperature to account for what the FRM is not

measuring. (Using Tim Hanley’s algorithm-EPA-OAQPS)

FDMS-8500C FEM: Corrects mass concentration data

using a power algorithm which is not temperature

dependent.

As a consequence, modeling past data using the FEM

algorithm was beneficial to understanding the likely effect

of switching to the FEM method.

FDMS-8500B (June 2008 to June 2009)

Data Set Slope and Intercept, and Limits

-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Arnold West (FMDS corrected by

FEM power algorithm (310 sample

pairs) (St. Louis)

Slope1

Intercept2

Correlation (r)

Statistics for this test site: 0.979 -1.219 0.98063

Upper: 1.100 2.000

Lower: 0.900 -1.915 0.94976

PASS PASS PASS1Multiplicative bias

2Additive bias

Test Results (Pass/Fail):

Regression statistics

Class III

Limits for


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Arnold West (FMDS corrected by

ARM algorithm (310 sample pairs) (St. Louis)

Slope1

Intercept2

Correlation (r)


Upper: 1.100 1.669

Lower: 0.900 -2.000 0.94940


2Additive bias



Class III

Limits for

Precision & Accuracy Model

CV (%) (Eqn 11)

9.75

Bias (%) (Eqn 3)

16.84

Signed Bias (%)

-16.84

Relative Percent Differences

-50.000

-30.000

-10.000

10.000

30.000

50.000

%D

CV (%) (Eqn 11)

7.17

ARM Algorithm: applied to Arnold

West data June 2008 through June

2009

Bias (%) (Eqn 3)

8.16

Signed Bias (%)

+/-8.16


-50.000

-30.000

-10.000

10.000

30.000

50.000

%D

FEM Algorithm: applied to raw ARM

test data-Arnold West June 2008

through June 2009

Precision and Bias

Statistics 40 CFR Part 58

Appendix A. Section 4.3.1

24hr-Avg.

Ambient temp.

low (Nov-Feb)

Other Midwest Testing - March 2007


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

SlopeIntercept, ug/m

3

Comparability of Candidate and FRM Methods*

y = 1.015x - 1.3868

0

5

10

15

20

25

30

0 5 10 15 20 25 30

FRM concentration, ug/m3

Candidate m

ethod concentration,

ug/m

3

Slope1

Intercept2

Correlation (r)


Upper: 1.100 1.652

Lower: 0.900 -2.000 0.93539


2Additive bias



Class III

Limits for

• Concurrent with FEM test Campaign C (different site)

•FDMS-8500B corrected using FEM power algorithm

• Avg. ambient 24-hr temperature for month: 12.9 °C.

• Only one day when 24-hr avg. ambient temperature was below 0 °C.

FDMS-8500C FEM Burn-inArnold West (St. Louis) (FMDS-8500C

FEM) 124 valid pairs (Aug.to Dec. 2009)

Arnold West (St. Louis) (FMDS-8500C

ARM) (Aug.to Dec. 2009)


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Slope1

Intercept2

Correlation (r)


Upper: 1.100 2.000

Lower: 0.900 -0.519 0.95000

FAIL PASS PASSTest Results (Pass/Fail):


Class III

Limits for


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Slope1 Intercept2 Correlation (r)

Statistics for this test site: 0.988 0.122 0.98850

Upper: 1.100 2.000

Lower: 0.900 -2.000 0.95000

PASS PASS PASSTest Results (Pass/Fail):


Class III

Limits for

Precision & Accuracy FRM/FEM-Burn-in

CV (%) (Eqn 11)

8.68


-50.000

-30.000

-10.000

10.000

30.000

50.000

%D

Bias (%) (Eqn 3)

13.26

Signed Bias (%)

-13.26

FEM Method FDMS-8500C –October

2009 through December 2009 (N=68

pairs) Arnold West St. Louis

CV (%) (Eqn 11)

6.74

Bias (%) (Eqn 3)

8.88

Signed Bias (%)

+/-8.88

ARM Method- FDMS-8500C -

October 2009 through December

2009 (N=68 pairs) Arnold West St.

Louis


-50.000

-30.000

-10.000

10.000

30.000

50.000

%D

Precision and Bias

Statistics 40 CFR Part 58

Appendix A. Section 4.3.1

24hr-Avg.

Ambient temp.

low (Nov-Dec)

What is Going On?

Temp (24-hr deg. C)

FRM-FDMS-FEM (ug/m^3)

25.9

24.2

22.5

22.0

21.2

20.4

19.0

17.1

15.7

14.1

12.9

11.4

10.4

9.5

7.8

5.9

2.7

1.0

-1.0

-2.7

-7.9

6

5

4

3

2

1

0

-1

-2

-3

0

5.2

Difference (FRM-FDMS(FEM) vs Ambient Temperature )ArnoldWest 8-1-09 through 12-31-09

FEM method corrects raw data (independent of temperature)

FRM Filter Preservation Temperature: 4 °C

FDMS Reference Purge Filter Temp: 4 °C

Coincidence? Probably not…

Regional Method Corrects FDMS-8500C for

Temperature Effect

Temp (24-hr deg. C)

FRM-FDMS-TH (ug/m^3)

25.9

24.2

22.5

22.0

21.2

20.4

19.0

17.1

15.7

14.1

12.9

11.4

10.4

9.5

7.8

5.9

2.7

1.0

-1.0

-2.7

-7.9

6

5

4

3

2

1

0

-1

-2

-3

0

5.2

Difference (FRM-FDMS(ARM Method) vs Ambient Temperature )ArnoldWest 8-1-09 through 12-31-09

Tim Hanley's method- corrects based on ambient temperature

Raw FDMS-8500C

Temp (24-hr deg. C)

FRM-FDMS-Raw (ug/m^3)

25.9

24.2

22.5

22.0

21.2

20.4

19.0

17.1

15.7

14.1

12.9

11.4

10.4

9.5

7.8

5.9

2.7

1.0

-1.0

-2.7

-7.9

6

5

4

3

2

1

0

-1

-2

-3

0

5.2

Difference (FRM-FDMS(Raw Data) vs Ambient Temperature )ArnoldWest 8-1-09 through 12-31-09

No Data Corrections

Troost (Kansas City) FRM/FEM-Burn in

FEM Method FDMS-8500C -October

2009 through January 2010 (N=53

pairs)


-50.000

-30.000

-10.000

10.000

30.000

50.000

%D

ARM Method- FDMS-8500C -

October 2009 through January 2010

(N=53 pairs)

Precision and Bias Statistics

40 CFR Part 58 Appendix A.

Section 4.3.1

CV (%) (Eqn 11)

9.32

Bias (%) (Eqn 3)

17.38

Signed Bias (%)

-17.38


-50.000

-30.000

-10.000

10.000

30.000

50.000

%D

24hr-Avg.

Ambient temp.

low (Nov-Jan)

Bias (%) (Eqn 3)

7.36

Signed Bias (%)

+/-7.36

CV (%) (Eqn 11)

6.58

Western MO “Left Shift” (Nitrate?)


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Slope1

Intercept2

Correlation (r)


Upper: 1.100 2.000

Lower: 0.900 -0.129 0.95000

FAIL FAIL PASSTest Results (Pass/Fail):


Class III

Limits for


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Slope1

Intercept2

Correlation (r)

Statistics for this test site: 0.958 0.405 0.99632

Upper: 1.100 2.000

Lower: 0.900 -1.535 0.95000

PASS PASS PASSTest Results (Pass/Fail):


Class III

Limits for

Troost (Kansas City) (FDMS-8500C FEM)

53 valid pairs (Oct. 2009 to Jan. 2010)

Troost (Kansas City) (FDMS-8500C ARM)

53 valid pairs (Oct. 2009 to Jan. 2010)

Troost (Kansas City)


y = 0.9576x + 0.4052

0

5

10

15

20

25

30

35

40

45

0 10 20 30 40 50


Candidate m


ug/m

3


y = 0.8764x - 0.2831

0

5

10

15

20

25

30

35

40

45

0 10 20 30 40 50


Candidate m


ug/m

3

Troost (Kansas City) (FDMS-8500C FEM)

(Oct. 2009 to Jan. 2010)

Troost (Kansas City) (FDMS-8500C ARM)

(Oct. 2009 to Jan. 2010)

Temp 24-hr-Deg C

FRM-FDMS-FEM (ug/m^3)

18.2

16.6

15.3

13.8

12.4

11.3

10.6

9.2

8.1

7.4

5.5

4.4

3.0

2.1

0.2

-1.1

-3.1

-4.6

-7.2

-9.1

-12.7

-16.9

6

5

4

3

2

1

0

-1

-2

-3

0

5.3

Valid Data Pairs 9-25-2009 through 1-31-2010

Difference (FRM-FDMS(FEM) vs. Ambient Temperature )

Troost (Kansas City)-FEM method corrects raw data (independent of temperature)

Temp 24-hr-Deg C

FRM-FDMS-TH (ug/m^3)

18.2

16.6

15.3

13.8

12.4

11.3

10.6

9.2

8.1

7.4

5.5

4.4

3.0

2.1

0.2

-1.1

-3.1

-4.6

-7.2

-9.1

-12.7

-16.9

6

5

4

3

2

1

0

-1

-2

0

5.1

Valid Data Pairs 9-25-2009 through 1-31-2010

Difference (FRM-FDMS(ARM) vs. Ambient Temperature )

Troost (Kansas City)-Tim Hanley's method- corrects based on ambient temperature

Regional P&A Statistics Summary (For burn-in period, using 40 CFR Part 58 Appendix A. Section 4.3.1 and 4.3.2)

• FRM/FEM Collocated Precision CV: <10%

PEP Bias - Audit Conc. 6.41µg/m3

Nov. 3, 2009, Avg. 24-hr temp: 10.3 °C

FRM/PEP: -17.6% (FRM Conc. 5.28µg/m3)

FEM/PEP: -14.2% (FEM Conc. 5.5µg/m3)

ARM/PEP: -6.4% (ARM Conc. 6.0µg/m3)

Raw/PEP: +12.3% (Raw Conc. 7.2µg/m3)

• FRM/FEM Bias (Absolute Method): Between

10% and 20%

Important Meta Data to Monitor

for Best FEM Performance

• Dryer Efficiency - Monitor Dew Points! (not discussed in FDMS manual)

• Dryer Purge Line Vacuum - MUST MONITOR!

(maintain 80% to70% of ambient pressure, about 20 inHg vacuum or better)

• Noise (Frequency) (<0.10 per manual, but <0.030

is better, 0.004 to 0.007 is typical of our monitors)

• Flow Rates (not usually a problem)

• Filter Loading ( <100% per manual, <60% is

better, or once per month whichever is sooner)

Purge Line Vacuum Effect on Dryer Efficiency

FDMS-RG South-Sample Dew Point as a function of rank ordering the External

Dew point

-35

-30

-25

-20

-15

-10

-5

0

5

10

0 200 400 600 800 1000 1200 1400 1600

Data point (hourly value)

Dew Point Deg C.

Samp Dew (PRC-099)

Extrn Dew (PRC-114)

After new pump

12/22/09

Vacuum at -22 inHg

Nov. & early

Dec. Vacuum

at -18 inHg

Sample Dew Point vs. Time -Ladue FDMS-8500C

"Age" Effect on Nafion Dryer (12/29/06 to 10/2/09)

-25

-20

-15

-10

-5

0

5

10

1/0 9/8 5/18 1/24 10/3 6/11

Date/hour

Sample Dew Point Deg C

Samp Dew (PRC-099)

FDMS-8500C

upgrade: about

5/17/07

May

2008

New Nafion Dryer

installed 8/26/09

May 2009

January 2008January

2009

Dryer “Age Effect” on FDMS Reference Channel

Observation

Individual Value (ug/m^3)

1601441281129680644832161

2

0

-2

-4

-6

-8

_X=-2.80

UCL=0.03

LCL=-5.63

144

111

1

1

11

1

1

11

1

Average 24 hr. Reference channel mass concentration (ug/m^3)

Dryer Replaced at Reference Line 144 (8/26/09)

Ladue TEOM-FDMS-8500C Reference Channel

Period 4/1/09 to 9/17/09

FRM Comparability - Dryer Efficiency


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3


-8

-6

-4

-2

0

2

4

6

8

0.8 0.9 1.0 1.1 1.2

Slope

Intercept, ug/m

3

Less Efficient Dryer

(typically yields greater

additive bias)

Efficient Dryer

(normal additive bias)

General Network Issues• Presently maintain FRM/FEM collocation at design value site in Kansas City MSA for quality assurance requirements and FEM performance monitoring.

• Report Regional Method (not an approved ARM) to AQS

under Parameter Code 88502 for the FRM/FEM collocated

site. (To support a potential modification of methods by user

request if cold weather bias is a problem for NAAQS

attainment decisions)

• Report PM2.5 Volatile Channel (reference) as AQS

Parameter Code 88503

• Report PM2.5 Total Atmospheric (uncorrected) as AQS

Parameter Code 88500 (meets Class III equivalency in

colder months)

Conclusions• The FEM in MO is comparable to the FRM averaged over all

seasons, but its use may be less desirable if concentrations areclose to attainment due to cold weather bias.

• There are several TEOM-FDMS methods that are comparable to the

FRM but only the FEM is approved for NAAQS comparisons.

• Options:

� Start new ARM test with FDMS-8500C (Best solution but is too

costly).

� Propose a Modification of Method by User: Substitute raw FDMS

data (already reported to AQS under method code 88500) at the

applicable temperature threshold.

• Monitor dryer performance and/or replace dryer annually.

� Continue with FEM “as is” and accept the bias.

implementing a continuous pm 2.5 federal equivalent method ...€¦ · 2009 through december 2009...

Documents