PM2.5 speciation / Source Apportionment. Urs Baltensperger Paul Scherrer Institut, Villigen, Switzerland

Transcription

1 PM2.5 speciation / Source Apportionment Urs Baltensperger Paul Scherrer Institut, Villigen, Switzerland Workshop Measurements and Modelling of PM2.5 in Europe Bilthoven, April, 29

2 The traditional approach for source apportionment: use of specific tracers R. Subramanian et al., 25

3 Applying partitioning theory to primary emissions results in much smaller primary fraction than classical OC/EC ratios suggest, because emission factors are not constant, but decrease with increasing dilution, due to evaporation Robinson et al., Science 27 Donahue et al., Environ. Sci. Technol. 26

4 Our approach: Combination of Carbon-14 analysis PMF (Positive Matrix Factorization) of organic aerosol mass spectra PMF of elemental spectra from rotating drum impactors

5 Fossil vs. non-fossil carbonaceous aerosol: Source apportionment by 14 C/ 12 C Anthropogenic Anthropogene Emissionen Emissions Biogene Biogenic Emissionen Emissions 14 C zerfallen 14 decayed C rezent modern Flux Fluss F = d m /dt a a Flux Fluss F = d m /dt ff f = 1.8 M = f M = 1 in reference year 195 b b Carbonaceous Kohlenstoffhaltiges Aerosolpartikel aerosol f M = F a F b + F b Important: discriminate OC and EC (EC mainly fossil, OC mainly recent) Fraction of modern: f M = ( 14 C/ 12 C Sample ) / ( 14 C/ 12 C) Current Biomass

6 14 C results from different sites P A U L S C H E R R E R I N S T I T U T OCnonfossil Göteborg/summer ECbiomass OCbiomass 2% 7% ECfossil 15% OCnonfossil Zurich/summer OCbiomass 8% ECbiomass 1% ECfossil 17% OCnonfossil OCbio 19% Mexico-City T/Spring OCbiomass 8% ECbiomass 1% ECfossil 34% OCbio 46% OCfossil 3% OCbio 48% OCfossil 26% OCfossil 38% OCnonfossil 43% OCbiomass 32% San Pietro Capofiume/summer ECbiomass 5% ECfossil 21% Average TC: 4. µg/m 3 San Pietro 4.8 µg/m 3 Zurich 3. µg/m 3 Gothenburg 24 µg/m 3 Mexico-City Stations Zurich and Gothenburg represent urban background OCbio 11% OCfossil 31%

7 Caveat: there is a grey zone between OC and EC in our C-14 measurements EC OC Andreae and Gelencser ACP 26

8 Non-refractory aerosol mass spectra with the Aerodyne Aerosol Mass Spectrometer (AMS) Matt Thyson (Lexington, Massachusetts)

9 Output of AMS: particle size and mass spectra (high time res.) Aerosol Sampling Real-time measurement AMS dm/dlogd a (µg m -3 ) Aerosol Mass Concentration (µg m -3 ) Nitrate Sulphate Ammonium Hydrocarbons Organics Oxygenated Etc Aerodynamic Diameter (nm) m/z (Daltons) Sp Mass Size distribution Chemical composition

10 Key Organic Mass Fragments Hydrocarbon : HOA e - C n H m ----> C n H m + 27, 29, 41, 43, 55, 57, 69, 71, Oxygenated : OOA e - C n H m O y ----> HCO+ CO + 2 H 3 C 2 O + HCO + 2 C n H + m , 57,... Note: ToF-AMS allows for unambigous fragment identification

11 AMS summer data for Zurich how can we retrieve source contributions from this? Zurich, Summer 19 µg m -3 15% 8% 8% 69% Org NO3 SO4 NH4 OM: 13 µg/m Zurich, Summer 8 29 Urban ics (%) 69 91

12 Positive Matrix Factorization (PMF) of full OM spectrum for source identification and attribution Ulbrich et al., ACPD 28

13 Results of PMF for Zurich and Pittsburgh Zurich average mass contribution to total organics [%] % 6% 13% 12% 1% 1% 7% 27% 15% 8% 2% 11% 7% 6% 7% 1% 1% 7% 19% 22% 22% 87% 62% 6% 5% 44% 4% number of factors 15 OOA OOA, type I OOA, type II HOA charbroiling wood burning minor (unknwon) minor (food cooking) 2 Six sources found for Zurich, Three sources for Pittsburgh Lanz et al., ACP (27) Ulbrich et al., ACPD (28)

14 Check of spectral similarity of retrieved spectra with measured spectra from sources Lanz et al., ACP (27)

15 There are two different types of OOA OOA, type I [µg m -3 ] AMS particle-sulphate [µg m -3 ] temperature [ C] dat OOA, type II [µg m -3 ] AMS particle-nitrate [µg m -3 ] Lanz et al., ACP (27)

16 The situation in winter is different: Concentrations mainly driven by meteorology high correlation between the time patterns from the different sources Traditional factor analysis fails Org NO3 SO4 NH4 32% 15% 17% 36% Zurich, Winter 44 µg m -3 OM: 16 µg/m 3

17 The use of a Multilinear Engine (ME-2) instead of PMF a. HOA factor: spectral similarity to references and evolution R 2 Diesel/starting profile (xx; ) Fuel (xx; ) Lubricating oil (xx; ) a-value m/z 2 b. Wood burning factor: spectral similarity to references R 2 1. R norm. intensity 4 6 a =.8 ( ) a =.6 ( ) a =. ( ) 8 1 c. OOA factor: spectral similarity to references a-value = : profile fixed a-value = 1: intensities can evolve from to 2% Additional constraints by radiocarbon analysis: a-value cannot be higher than.8 (otherwise HOA overestimated (fossil SOA negative).6.5 Ambient wood burning (xx; ) Chestnut burning (xx; ) Levoglucosan (xy; ) a-value Secondary Pittsburgh (xx; ) Secondary Vancouver (xy; ) Fulvic acid (xz; ) a-value.8 Lanz et al., ES&T 28

18 ME-2 result for OA in Zurich during winter Percent contribution [%] Date measured organics [µg m -3 ] From C-14 analysis: 69% of OOA non-fossil Sources? Percent contribution [%] overall average: 3% 7% 13% 57% 4% 2 55% 4 38% 6 52% 35%..6.8 a-values 8 HOA wood burning OOA 1 Lanz et al., ES&T 28

19 Temporal evolution of the three components c. HOA [µg m -3 ] ( ) d. wood burning [µg m -3 ] ( ) dat dat NO x [ppb] R 2 =.7 CO [ppm] R 2 =.78 K [µg m -3 ] R 2 = Comparison of sum of these 3 components with orginal data yields slope of 1. and R 2 of e. OOA [µg m -3 ] ( ) AMS-NH 4 [µg m -3 ] R 2 = No split into OOA1 & OOA Date Lanz et al., ES&T 28

20 RDI aerosol sampling RDI: Rotating Drum Impactor RECTANGULAR JET INLET A Size fractionation in 3 classes based on rectangular jet impaction Rotating substrate flexible time resolution using a step motor DECREASING PARTICLE SIZE B RECTANGULAR JET RECTANGULAR JET PM1 Inlet: 16.6 L/min (1 m 3 /h) 1h steps! 3 size fractions: A: µm B: µm C: µm C Modified design of the original impactor by Lundgren (1967) PUMP

21 RDI sample handling Substrate: Mylar (thickness 1.8 µm, greased with APIEZON L) RDI impactor wheels RDI aerosol sampling Sample holder for SR-XRF analysis Aerosol covered Mylar films

22 Accessible elements with Synchrotron X-ray fluorescence spectrometry (SR-XRF) 5 2 kev primary monochromatic beam Detected Z range ( via K-Lines / L-Lines

23 P A U L S C H E R R E R I N S T I T U T Elemental time series Size fractionation + High time resolution + Low elemental detection limit + Broad range of elements = New insight into short-term dynamics of atmospheric pollution by trace elements Bukowiecki et al., ES&T, µm µm.1-1 µm Zürich (Switzerland), Winter 23/ Bromine (ng m -3 ) Sulfur (ng m -3 ) Iron (ng m -3 )

24 Sector contributions to PM1 in Zurich in winter

25 Conclusions PMF is a powerful means for source apportionment (needs to be used with care!) Can be applied to organics and inorganics The combination with C-14 analysis adds much additional strength Wood burning OA much higher than expected SOA often dominating, also high in winter We cannot (yet) discriminate between SOA from different sources

26 Is this of any help for the air quality authorities? Yes, e.g. Swiss project: temporal evolution of fossil/non-fossil carbonaceous aerosol at ~1 sites over the next 5 years Not yet for AMS: too expensive for routine applications; may change in near future with introduction of new Mini-AMS

27 Conclusions on PM source apportionment / short + We know how to do it - It s expensive

28 Thank you for your attention Acknowledgments: S. Szidat, Univ. Bern C. Hüglin, Empa A. Prévôt, group leader at PSI R. Alfarra, R. Chirico, P. DeCarlo, M. Furger, M. Heringa, V. Lanz, C. Mohr, N. Perron, A. Richard, R. Richter, G. Wehrle, E. Weingartner,... Funding: - BAFU (EPA Switzerland) - Competence Centre for Energy and Mobility - Competence Centre for Environment and Sustainability - EC projects ACCENT, EUCAARI, EUROCHAMP, POLYSOA - ESF project INTROP