1. Pinning down the global organic aerosol budget Atmospheric Chemistry Gordon Research Conference July 28, 2011 Colette L. Heald With thanks to : many individuals for invaluable field and satellite measurements noted along the way…

2. OA can dominate aerosol loading Likely to become larger part of the pie with clean-up of SO 2 Should we regulate OA? How do we regulate SOA? What is the anthropogenic lever on OA? What are climate feedbacks on OA? What are ‘natural’ OA loadings that define sensitivity of indirect effect? How do organics affect CCN? Policy Questions Air Quality Climate Organics

3. THE PICTURE CIRCA 2005-2007: A LARGE MISSING SOURCE OF ORGANIC AEROSOL? Room for everyone’s favourite source/process in models that are woefully inadequate. Could be up to 10x more OA in atmosphere than sulfate! Models drastically underestimate SOA from 4 campaigns [Volkamer et al., 2006] ACE-Asia (2001): 3 groups measured high OA off Asia. GEOS-Chem simulation factor of 10-100 too low [Heald et al., 2005] Goldstein and Galbally [2007] suggest that SOA source may be anywhere from 140-910 TgC/yr. Obs (Maria et al., 2003) GEOS-Chem

4. THE MODELING CHALLENGE: BOTTOM-UP VS. TOP-DOWN Too many (underconstrained) processes… Not enough globally representative obs…. Until now? SV-POA & IVOCs [Robinson et al., 2007] Aqueous SOA formation [Lim et al., 2005; Carlton et al, 2006; Sorooshian et al., 2007; Volkamer et al., 2007] New precursors [Kroll et al., 2005; Lim and Ziemann, 2009; Volkamer et al., 2009] OA has MORE SOURCES and is MORE DYNAMIC than understood in 2005. SOA Yields = f(RH, NOx, acidity..) [Iinua et al., 2004; Ng et al., 2007; Surratt et al., 2007] What about sinks?? Dynamic volatility [Donahue et al., 2005] Global (MODIS, MISR), but not species-specific First OA profile in 2000 Largely filter samples until early 2000s AMS proliferation: NH now reasonably well characterized…coverage of the rest of the world still sparse PBAP [Jaenicke et al., 2005; Heald and Spracklen, 2009; Burrows et al., 2009] OH Recycling [Lelieveld et al., 2008]

5. MY OBJECTIVE: USE OBSERVATIONS TO SET A TARGET FOR GLOBAL MODELING OF OA (and maybe get some insight into key processes along the way…)

6. IF ONLY AEROSOL IN THE ATMOSPHERE WAS OA, WHAT LOADING IS IMPLIED BY SATELLITE AOD? Calculate the “hypothetical” AOD implied by a constant 1  g/sm 3 profile over the land, and see how we need to scale this locally to make up ENTIRE AOD reported by MISR. Inverted OA loading is 3.5 TgC over land. Assume a 6 days lifetime = 215 TgC/yr  extrapolate to include outflow ~430 TgC/yr. (middle of Goldstein & Galbally range) Inverted total MISR AOD: Surface OA concentrations With contributions from: David Ridley, Sonia Kreidenweis, Easan Drury, and MISR team

7. A MORE REALISTIC POSSIBILITY: REMOVE CONTRIBUTIONS FROM DUST, BC, INORGANICS (assuming all the negative bias in the model is ONLY OA) If remove N. Africa & the Middle (dust), estimate that ~170 TgC/yr source is required to close the MISR-GEOS- Chem* discrepancy. DJFJJA MISR GEOS-Chem* MISR- GEOS-Chem* *excluding OA

8. UNCERTAINTIES ATTRIBUTED TO VERTICAL DISTRIBUTION Uniform vertical profile perhaps not very realistic… If the same mass is distributed with exponential drop off (atmospheric scale height assumed), the AOD increases by 15%. OA burden implied by AOD would be 15% lower if distributed exponentially. This reduces required source to ~150 TgC/yr Note: with this profile surface concentrations would be ~ twice as high. Will think more about this next… 1212 OA [  g/m 3 ]

9. This is more than THREE TIMES what is currently included in global models…. BUT at the low end of Goldstein & Gallbally [2007] range. HAVE WE REDUCED THE UNCERTAINTY ON THE OA BUDGET? 910 47 Existing GEOS- Chem sources 140 Our satellite top-down estimate 150 Bottom-up estimate [Goldstein and Galbally, 2007] All units in TgCyr -1 Satellite-based estimate [Heald et al., 2010] AMS surface-based optimization [Spracklen et al., 2011] 94 24 POA (fixed) SOA (optimized) Also in relatively good agreement with Spracklen et al. [2011] estimate.

10. DIGGING IN DEEPER: 17 AIRCRAFT FIELD CAMPAIGNS * All AMS measurements, except ITCT-2K4 (PILS) and ACE-Asia (filters). 2001-2009 Aircraft constraints on the organic aerosol distribution through depth of troposphere in remote, polluted and fire influenced regions. GOAL: investigate vertical profile and compare with one CONSISTENT model. Measurements PIs: Hugh Coe (ITOP, ADRIEX, DABEX, DODO, AMMA, ADIENT, EUCAARI, OP3, VOCALS-UK, TROMPEX), Jose Jimenez, (MILAGRO, IMPEX, ARCTAS), R. Weber (ITCT-2K4), Ann Middlebrook (TexAQS), Lynn Russell (ACE-Asia) GEOS-Chem SOA simulation: 2 product model, monoterpenes/sesquiterpenes +OH/O3/NO3 (Griffin et al, 1999), low-NOx isoprene+OH (Kroll et al., 2006), NOx dependent aromatics +OH(Ng et al., 2007)  latest description Henze et al., 2008

11. HOW MUCH ORGANIC AEROSOL AND WHAT FRACTION OF TOTAL AEROSOL DOES IT CONTRIBUTE? Mean concentrations of OA span <1  g/m 3 to 8  g/m 3 On average 15-70% of non-refractory fine aerosol is OA

12. COMPARISON OF VERTICAL PROFILE General profile: drops off with altitude, with BB plumes aloft. Over remote regions, little structure to profile. Outliers: AMMA, ACE-Asia. Model roughly captures profile. “Reasonable” assumption on profiles made in Heald et al. [2010] looking at satellite

13. OVERALL COMPARISON OF OA SIMULATION (an “update” to Volkamer et al., 2006) Median model overestimate less than a factor of 5. Key difference from Volkamer et al [2006]: Discrepancy is largest close to source. Range = 0.45-4.5

14. IN CASE YOU WERE FEELING OPTIMISTIC… R 2 between gridded observations and GEOS-Chem of OA Model captures less than 40% of the observed variability (except for TexAQS). Model has most skill in pollution-influenced environments.

15. IS THERE A SIGNATURE OF AQUEOUS PRODUCTION OF SOA? Observational signature of aqueous processing is murky (need better chemical constraints). Comparison of model with observations doesn’t provide strong evidence. GEOS-Chem 2006 simulation with simple aqueous SOA [Fu et al., 2008] Sorooshian et al. 2010 Hennigan et al. 2008 Observed trends with RHModel underestimate *possibly* increases with RH Model suggestive of a vertical enhancement from aq sources

16. CAN WE ATTRIBUTE THE MODEL UNDERESTIMATE? Adding ~100 Tg/yr source of ASOA (as suggested by Spracklen et al., 2011) improves comparison in polluted regions, but leads to too much OA aloft and in remote regions. Higher volatility OA? OA sink?

17. FRAGMENTATION LOSS OF ORGANICS FunctionalizationFragmentation (break C-C bonds) OH + Oxidation By OH 5% 95% Simple sensitivity tests in GEOS-Chem: assuming that 5% of reacted organics are fragmented (LOSS) assume that 95% are functionalized (NO CHANGE) Testing 2 types of fragmentation loss: 1.Oxidation of gas-phase organics (faster, k OH =2x10 -11 molecules/cm 3 /s) 2.Heterogeneous oxidation of SOA (slower, k OH =1x10 -12 molecules/cm 3 /s) [Molina et al., 2004; Kwan et al., 2006; Kroll et al., 2007; Chan et al., 2007; Kroll et al., 2009] Increase volatility Decrease volatility

18. IMPACT OF FRAGMENTATION ON SIMULATED SOA BUDGET (GEOS-Chem: 2008 simulation) GASPARTICLE Gas-Phase Fragmentation -47% Heterogeneous Fragmentation = slow leak -16% Fragmentation of gas-phase organics efficiently “prevents” SOA formation. Is most efficient for more volatile organics. Fragmentation from heterogeneous oxidation is much slower but is still a potentially important sink of SOA. Gas-Phase: -47% (likely upper limit) Heterogeneous: -16%

19. IMPACT OF FRAGMENTATION SINK & CHANGING VOLATILITY ON COMPARISONS WITH FIELD DATA Effect of gas-phase fragmentation sink is comparable to increasing volatility away from source (via enthalpy of vaporization). All of these modifications added ~5% to the variability captured by the model in anthropogenic regions. ASOAx30 +heterogeneous fragmentation sink ASOAx30 +gas-phase fragmentation sink ASOAx30 +  H=25 kJ/mol (increase volatility) ~equivalent

20. GEOS-Chem simulation looks decent! Median differences < 1  g/m 3 for 15 of 17 campaigns. IF ADD 100 Tg/yr ASOA AND GAS PHASE FRAGMENTATION LOSS [Heald, Coe, et al., in prep]

21. Target: OA source is likely in the 100-150 TgC/yr range Current global model source is about ½ of this. And largest discrepancies with observations are in anthropogenic source regions. Marine OA and continental PBAP do not appear to be dominant contributors to global fine PM [Lapina et al., 2011; Heald and Spracklen, 2009] Fragmentation may be an important sink of OA (will allow us to add more sources without filling up the atmosphere) PRIORITY: aqueous-phase SOA and constraints on OA deposition