1 Collocated STN-CSN and IMPROVE carbon measurements WHW, UCD
1/22/08
Slide 2
2 IMPROVE analytical upgrade
Slide 3
3 EXPECTIONS & WORKING ASSUMPTIONS analysis : Analytical
methods agree for TC = OC + EC: TC CSN = TC IMP i.e., carbon itself
is measured unambiguously. The relationship between the splits is
linear and homogeneous: OCOC measured OC measured = (1-f)OC + gEC
ECEC measured EC measured = fOC + (1-g)EC IMPROVEs f and g may
differ between old and new TOR analyses.
Slide 4
4 EXPECTATIONS & WORKING ASSUMPTIONS sampling : The
collected sample may be affected by filter artifacts (FA) and
sampling artifacts (SA): C = (1-B SA )[C]V + A FA,B SA > 0 and A
FA > 0 The filter and sampling artifacts for EC are negligible
in both networks: EC = [EC]V (B SA = 0 and A FA = 0) These
assumptions lead us to expect an unadjusted value (reported by CSN)
of: TC/V = OC/V + EC/V = (1-B SA )[OC] + A FA /V + [EC] = [TC] B SA
[OC] + A FA /V
Slide 5
5 EC Each point is the median from all observations on days -3,
0, +3 by the indicated sampler. Out-of-range points are always
plotted at the appropriate boundary.
Slide 6
6 TC Note the switch from IMPROVE / CSN to CSN / IMPROVE. (Both
plots show the larger measurement on top.) inverse of previous
ratio
Slide 7
7 TC Neither ratio nor difference is quite right for a
relationship of the form [CSN] = a + b[IMPROVE]. difference, not
ratio
Slide 8
8 For EC, the systematic differences between CSN and IMPROVE,
and between old and new IMPROVE, seem to be multiplicative.
Slide 9
9 For TC, the difference between CSN and IMPROVE appears to
include an additive offset in addition to a multiplicative factor.
There is no obvious difference between old and new IMPROVE.
Slide 10
10 2005-6 For EC, the difference between CSN and IMPROVE shows
little dependence on the CSN sampler, suggesting that it is mainly
analytical.
Slide 11
11 2005-6 For TC, the difference between CSN and IMPROVE
clearly does vary with the CSN sampler.
Slide 12
12 2005-6 Site-specific differences between CSN and IMPROVE are
evident only at Phoenix, where the MetOne - IMPROVE TC difference
tends to be higher than it is elsewhere.
Slide 13
13 Phoenix, 2004-6 A collocated IMPROVE monitor has operated at
Phoenix since March 2004. These plots compare the two collocations
on days with observations from both. Nothing out of the ordinary is
evident.
Slide 14
14 IMPROVE analytical upgrade
Slide 15
15 RECALL OUR WORKING ASSUMPTION : TC/V = [TC] B SA [OC] + A FA
/V (i) The MetOne face velocity (6.7 L/min through a 47 mm filter)
is much lower than the IMPROVE face velocity (~22.8 L/min through a
25 mm filter). (ii) Reported IMPROVE concentrations are corrected
for the filter artifact. It will simplify our interpretation if we
accordingly neglect (i) the MetOne sampling artifact and (ii) the
IMPROVE filter artifact. Then [TC] IMP = [TC] B IMP [OC] and[TC]
CSN = [TC] + A CSN /V MetOne. Solving for [TC] in both expressions
and equating the two solutions yields [TC] CSN = [TC] IMP + B IMP
[OC] + A CSN /V MetOne Estimate OC: ~ [TC] IMP + B IMP *[OC] IMP +
A CSN /V MetOne = [EC] IMP + (1+B IMP *)[OC] IMP + A CSN /V
MetOne
Slide 16
16 EXPECTATION : [TC] CSN = [EC] IMP + (1+B IMP *)[OC] IMP + A
CSN /V MetOne OLS REGRESSION: [TC] CSN = (1+b EC )[EC] IMP + (1+b
OC )[OC] IMP + a 1 ++ a 12 + e 2005-6 observations at 7 MetOne
sites (excluding Phoenix): b EC = 0.008 (+/-0.05) no sampling
artifact for IMPROVE EC b OC = 0.22 (+/-0.03) ~ 20% sampling loss
for IMPROVE OC rms(e) = 0.9 ug/m 3 (r 2 = 0.986, n = 779) a mm next
slide
Slide 17
17 MetOne artifacts
Slide 18
18 OLS REGRESSION FOR EC: [EC] CSN = (1-g)[EC] IMP + f[OC] IMP
+ a 1 ++ a 12 + e 2005-6 observations at 7 MetOne sites (excluding
Phoenix): g = 0.40 (+/-0.02) f = 0.03 (+/-0.01) rms(e) = 0.3 ug/m 3
(r 2 = 0.942, n = 779) a mm : mixed signs, marginal
significance
Slide 19
19 REVIEW OF EVIDENCE FOR DIFFERENCES: Observed differences
Changed at 2004 2005 TOR transition Vary with CSN sampler Suggest a
seasonally varying additive artifact in CSN OC (relative to
IMPROVE) Suggest a multiplicative negative artifact in IMPROVE OC
(relative to CSN)
Slide 20
20
Slide 21
21 The 779 MetOne observations from 2005-6 at seven sites can
be linearly transformed into IMPROVE values with rms errors of EC:
0.4 ug/m 3 (27% of mean value) TC: 0.8 ug/m 3 (16% of mean
value)