using in-network precision data as a basis for cross-network comparisons warren h. white, nicole p....
TRANSCRIPT
Using In-Network Precision Data as a Basis for Cross-Network Comparisons
Warren H. White, Nicole P. Hyslop, and Charles E. McDade
AAAR Specialty Conference: Particulate Matter, Supersites Program & Related Studies
Atlanta, GA 11 February 2005
Module A (PM2.5, Teflon Filter)
Mesa Verde National Park, COProctor Maple Research Facility, VTOlympic National Park, WASaint Marks National Wildlife Refuge, FLSac and Fox Tribe, KSTrapper Creek (Denali National Park), AK Module B (PM2.5, Nylon Filter)
Big Bend National Park, TXBlue Mounds State Park, MNFrostburg Reservoir, MDGates of the Mountains Wilderness, MTLassen Volcanic National Park, CAMammoth Cave National Park, KY
Module C (PM2.5, Quartz Filter)
Everglades National Park, FLHercules Glades Wilderness, MOHoover Wilderness, CAMedicine Lake National Wildlife Refuge, MTSaguaro National Park (Western Section), AZSeney National Wildlife Refuge, MI Module D (PM10, Teflon Filter)
Houston, TX (STN urban site)Jarbridge Wilderness, NVJoshua Tree National Park, CAQuabbin Reservoir, MASwanquarter National Wildlife Refuge, NCWind Cave National Park, SD
collocated samplers in IMPROVE networkstarting summer 2003
STN/IMPROVE Monitoring Intercomparison Sites: Oct. 2001 – Sept. 2003
Mt. Rainier NPS
Phoenix
TontoNational Monument
Haines Point, NPS
Wash. DC
Official or designated STN site, host to IMPROVE sampler
Official IMPROVE site, host to STN sampler
Anderson RAAS 401 STN Samplers
Met One SASS STN Samplers
URG MASS STN Samplers
SeattleBeacon Hill
USDA FS Dolly SodsWildersness
Operated According to Each Network’s Protocols
STN/IMPROVE Monitoring Intercomparison Sites: Oct. 2001 – Sept. 2003
Mt. Rainier NPS
Phoenix
TontoNational Monument
TontoNational Monument
Haines Point, NPS
Wash. DC
Official or designated STN site, host to IMPROVE sampler
Official IMPROVE site, host to STN sampler
Anderson RAAS 401 STN Samplers
Met One SASS STN Samplers
URG MASS STN Samplers
SeattleBeacon Hill
USDA FS Dolly SodsWildersness
Operated According to Each Network’s Protocolscollocated monitoring by Speciation Network
uncertainty reporting started summer 2003STN/IMPROVE Monitoring Intercomparison
Sites: Oct. 2001 – Sept. 2003
Mt. Rainier NPS
Phoenix
TontoNational Monument
Haines Point, NPS
Wash. DC
Official or designated STN site, host to IMPROVE sampler
Official IMPROVE site, host to STN sampler
Anderson RAAS 401 STN Samplers
Met One SASS STN Samplers
URG MASS STN Samplers
SeattleBeacon Hill
USDA FS Dolly SodsWildersness
Operated According to Each Network’s Protocols
STN/IMPROVE Monitoring Intercomparison Sites: Oct. 2001 – Sept. 2003
Mt. Rainier NPS
Phoenix
TontoNational Monument
TontoNational Monument
Haines Point, NPS
Wash. DC
Official or designated STN site, host to IMPROVE sampler
Official IMPROVE site, host to STN sampler
Anderson RAAS 401 STN Samplers
Met One SASS STN Samplers
URG MASS STN Samplers
SeattleBeacon Hill
USDA FS Dolly SodsWildersness
Operated According to Each Network’s Protocols
STN data provided by Shelly Eberly, EPA OAQPS
obligatory introductory slide
• When should both parties be happy with the results of a comparison?
• First impressions – how do the collocated and routine IMPROVE results compare?
• First impressions – how do the STN and IMPROVE results compare?
0
0.001
0.002
0.003
0 0.001 0.002 0.003
Se, ug/m3
VAL/MDL > 3
0
0.001
0.002
0.003
0 0.001 0.002 0.003
IMPROVE routine
IMP
RO
VE
col
loca
ted
0
0.001
0.002
0.003
0 0.001 0.002 0.003
PMRF
SAFO
OLYM
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
IMPROVE
ST
N
Se, ug/m3
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
MASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
Se, ug/m3
0
0.001
0.002
0.003
0 0.001 0.002 0.003
IMPROVE routine
IMP
RO
VE
col
loca
ted
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
IMPROVE
ST
N
Se, ug/m3
Se, ug/m3
0
0.001
0.002
0.003
0 0.001 0.002 0.003
PMRF
SAFO
OLYM
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
MASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
0
0.001
0.002
0.003
0 0.001 0.002 0.003
Se, ug/m3
VAL/MDL > 3
0
0.001
0.002
0.003
0 0.001 0.002 0.003
PMRF
SAFO
OLYM
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
MASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
V, ug/m3
0
0.005
0.01
0.015
0.02
0.025
0 0.005 0.01 0.015 0.02 0.025
IMPROVE
ST
NMASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
V, ug/m3
0
0.0025
0.005
0 0.0025 0.005
IMPROVE routine
IMP
RO
VE
col
loca
ted
0
0.001
0.002
0.003
0 0.001 0.002 0.003
PMRF
SAFO
OLYM
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
MASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
V, ug/m3
0
0.005
0.01
0.015
0.02
0.025
0 0.005 0.01 0.015 0.02 0.025
IMPROVE
ST
N
V, ug/m3
0
0.0025
0.005
0 0.0025 0.005
IMPROVE routine
IMP
RO
VE
col
loca
ted
0
0.001
0.002
0.003
0 0.001 0.002 0.003
PMRF
SAFO
OLYM
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
MASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
Fe, ug/m3
0
0.25
0.5
0.75
0 0.25 0.5 0.75 1 1.25 1.5
IMPROVE
ST
N
Fe, ug/m3
0
0.025
0.05
0.075
0.1
0 0.025 0.05 0.075 0.1
IMPROVE routine
IMP
RO
VE
col
loca
ted
0
0.001
0.002
0.003
0 0.001 0.002 0.003
PMRF
SAFO
OLYM
0
0.002
0.004
0.006
0.008
0 0.002 0.004 0.006 0.008
MASS/URG PUSO
RASS/AndersonDOSO, WASH
SASS/MetOne PHOE, TONT
Fe, ug/m3
0
0.025
0.05
0.075
0.1
0 0.025 0.05 0.075 0.1
IMPROVE routine
IMP
RO
VE
col
loca
ted
0
0.25
0.5
0.75
0 0.25 0.5 0.75 1 1.25 1.5
IMPROVE
ST
N
Fe, ug/m3
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100
(col
lo-r
outin
e)/s
qrt(
2),
ug/m
3
-75%
-50%
-25%
0%
25%
50%
75%
0.01 0.1 1 10 100
(col
lo-r
out)
sqrt
(2)/
(col
lo+
rout
)
-75%
-50%
-25%
0%
25%
50%
75%
0.01 0.1 1 10 100
-75%
-50%
-25%
0%
25%
50%
75%
0.01 0.1 1 10 100
(collocated+routine)/2, ug/m3
-75%
-50%
-25%
0%
25%
50%
75%
0.01 0.1 1 10 100
-75%
-50%
-25%
0%
25%
50%
75%
0.01 0.1 1 10 100
(collocated+routine)/2, ug/m3
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100
(col
lo-r
outin
e)/s
qrt(
2),
ug/m
3
-75%
-50%
-25%
0%
25%
50%
75%
0.01 0.1 1 10 100
(col
lo-r
out)
sqrt
(2)/
(col
lo+
rout
)
IMPROVE SO4=
relative error
arithmetic error
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100
(col
lo-r
outin
e)/s
qrt(
2),
ug/m
3
-15
-10
-5
0
5
10
15
0.01 0.1 1 10 100
obse
rved
diff
. / e
xpec
ted
diff.
IMPROVE SO4=
-15
-10
-5
0
5
10
15
0.01 0.1 1 10 100
(collocated+routine)/2, ug/m3
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100
(collo
-routin
e)/
sqrt
(2),
ug/m
3
observed expected
-5
-4
-3
-2
-1
0
1
2
3
4
5
0% 25% 50% 75% 100%cumulative percentage of observations
ob
serv
ed
/ e
xpec
ted
-15
-10
-5
0
5
10
15
0.01 0.1 1 10 100
-15
-10
-5
0
5
10
15
0.01 0.1 1 10 100
obse
rved
diff
. / e
xpec
ted
diff.
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
cumulative distribution of observed / expected
-5
-4
-3
-2
-1
0
1
2
3
4
5
0% 25% 50% 75% 100%cumulative percentage of observations
IMPROVE SO4=
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
/
exp
ect
ed
-5
-4
-3
-2
-1
0
1
2
3
4
5
0% 25% 50% 75% 100%
obse
rved
/
expe
cted
-5
-4
-3
-2
-1
0
1
2
3
4
5
0% 25% 50% 75% 100%cumulative percentage of observations
unit norm
al distrib
ution*
* corresponding to normal error, unbiased with known precision
IMPROVE SO4=
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
/
exp
ect
ed
0
7.5
15
22.5
0 7.5 15 22.5
ST
N
0
7.5
15
22.5
0 7.5 15 22.5
RASS/Anderson DOSO, WASH
SASS/MetOne PHOE, TONT
MASS/URG PUSO
0
7.5
15
22.5
0 7.5 15 22.5
ST
NRASS/Anderson DOSO, WASH
SASS/MetOne PHOE, TONT
MASS/URG PUSO
0
7.5
15
22.5
0 7.5 15 22.5
IMPROVE
0
7.5
15
22.5
0 7.5 15 22.5routine IMPROVE
collo
cate
d IM
PR
OV
E
MACA
LAVO
BIBE
GAMO
0
7.5
15
22.5
0 7.5 15 22.5
MACA
LAVO
BIBE
GAMO
0
7.5
15
22.5
0 7.5 15 22.5
collo
cate
d IM
PR
OV
E
0
7.5
15
22.5
0 7.5 15 22.5routine IMPROVE
IMPROVE-IMPROVE STN-IMPROVE
SO4=
-6.0
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%cumulative percentage of observations
obse
rved
/
expe
cted
STN
- IMPROVE
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%
Se
V
Fe
Are observed STN-IMPROVE differences accounted for by the two networks’ reported uncertainties?
Yes!
no!
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
/
exp
ect
ed
PM2.5
PM10
IMPROVE:collocated - routine
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
PM2.5
PM10
How about collocated-routine differences within IMPROVE?
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%cumulative percentage of observations
obse
rved
/
expe
cted
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
EC
OC
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%cumulative percentage of observations
obse
rved
/
expe
cted
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
SO4
NO3
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
diff
ere
nce
/ e
xpe
cte
d d
iffe
ren
ce V, 351/372 detects (94%)
Ni, 299/372 detects (80%)
As, 243/372 detects (65%)
Se, 339/372 detects (91%)
normal error,zero bias,accurateuncertainty
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
V, 351/372 detects (94%)
Ni, 299/372 detects (80%)
As, 243/372 detects (65%)
Se, 339/372 detects (91%)
normal error,zero bias,accurateuncertainty
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
diff
ere
nce
/ e
xpe
cte
d d
iffe
ren
ceVAL/MDL > 3
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
V, 330/372 detects (89%)
Ni, 200/372 detects (54%)
As, 53/372 detects (14%)
Se, 241/372 detects (65%)
normal error,zero bias,accurateuncertainty
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
diff
ere
nce
/ e
xpe
cte
d d
iffe
ren
ce
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%
Ca, 372/372 detects (100%)
K, 372/372 detects (100%)
Ti, 356/372 detects (96%)
Fe, 372/372 detects (100%)
normal error,zero bias,accurateuncertainty
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
diff
ere
nce
/ e
xpe
cte
d d
iffe
ren
ce
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%
Al, 154/372 detects (41%)
Si, 338/372 detects (91%)
Mn, 362/372 detects (97%)
Fe, 372/372 detects (100%)
normal error,zero bias,accurateuncertainty
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
diff
ere
nce
/ e
xpe
cte
d d
iffe
ren
ce Al, 132/372 detects (35%)
Si, 338/372 detects (91%)
Mn, 343/372 detects (92%)
Fe, 372/372 detects (100%)
normal error,zero bias,accurateuncertainty
VAL/MDL > 3
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
2.5% 50.0% 97.5%
Al, 132/372 detects (35%)
Si, 338/372 detects (91%)
Mn, 343/372 detects (92%)
Fe, 372/372 detects (100%)
normal error,zero bias,accurateuncertainty
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%cumulative percentage of observations
ob
serv
ed
diff
ere
nce
/ e
xpe
cte
d d
iffe
ren
ce
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
ob
serv
ed
/
exp
ect
ed
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
2.5% 50.0% 97.5%
IMPROVE:collocated - routine
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
2.5% 50.0% 97.5%
H, 372/372 detects (100%)
S, 372/372 detects (100%)
Br, 372/372 detects (100%)
Pb, 368/372 detects (99%)
normal error,zero bias,accurateuncertainty
What I just said
• When should both parties be happy with the results of a comparison? When observed differences are consistent with reported uncertainties.
• First impressions – how do the collocated and routine IMPROVE results compare? We are generally ‘happy’, except with the ‘crustal’ elements and S.
• First impressions – how do the STN and IMPROVE results compare? See C.E. McDade et al., this room, just 1 hour from now.