BAYESIAN AND ADJOINT INVERSE MODEL ANALYSES OF PM SOURCES IN THE UNITED STATES USING OBSERVATIONS FROM SURFACE, AIRCRAFT, AND SATELLITE PLATFORMS Daniel J. Jacob, Harvard (P.I.) and John H. Seinfeld, Caltech (Co-I.) Publications acknowledging support from this EPA-STAR grant: Park, R.J., D.J. Jacob, and J.A. Logan, Fire and biofuel contributions to annual mean aerosol concentrations in the United States, AE (in press) Henze, D.K. and J.H. Seinfeld, Global secondary organic aerosol from isoprene oxidation, GRL, 33, L09812, Heald, C.L., D.J. Jacob, et al., Concentrations and sources of organic carbon aerosol in the free troposphere over North America, JGR 111, D23S47, Fairlie, T.D., D.J. Jacob, and R.J. Park, The impact of transpacific transport on mineral dust in the United States, AE, 41, , Henze, D.K., A. Hakami, and J.H. Seinfeld, Development of the adjoint of GEOS-Chem, Atmos. Chem. Phys., 7, , Drury, E.E, D.J. Jacob, et al., Improved algorithm for MODIS satellite retrievals of aerosol optical depths over land, to be submitted, 2007

QUANTIFYING THE BIOMASS BURNING SOURCE OF PM2.5 FROM CORRELATION OF TOTAL CARBON (TC) WITH NON-SOIL K + AT IMPROVE SITES Non-soil K + : West: summer max, low background East: winter and summer maxes, high background Background ns-K + (traces industrial biofuel) Park et al. [2007] use TC vs. ns-K + linear regression together w/satellite data to quantify TC from open fires d[TC]/d[ns-K + ]

FIRE AND BIOFUEL CONTRIBUTIONS TO TC PM2.5 Annual mean C concentrations, Mean values for W and E at top of plot Park et al. [2007]

IMPLICATIONS FOR AQ STANDARDS AND EMISSION INVENTORIES PM2.5,  g m-3 (x1.8 for TC) WestEast Summer wildfires Other fires Residential biofuel Industrial biofuel Total fire+biofuel TC (this work) Observed TC Observed total PM Tg C y -1 Top-down This work Open fires Biofuel Park et al. [2003] Open fires Biofuel (40% industrial) (80% industrial) Bottom-up Bond et al. [2004] Open fires Biofuel NEI99 Biofuel (<10% industrial) 0.23 (<10% industrial) Source attribution at IMPROVE sites (annual means) Emissions for contiguous U.S. (climatological estimate for fires) Park et al. [2007]

Henze and Seinfeld [2006] SOA FROM ISOPRENE Mechanistic treatment of SOA formation from isoprene oxidation in GEOS-Chem: semi-volatile products partitioning fitted to data of Kroll et al. [2006] w/o isoprene w/ isoprene Biogenic SOA: Isoprene accounts for half of the global biogenic SOA source Surface Free troposphere

WATER-SOLUBLE OC (WSOC) AEROSOL OVER NORTHEASTERN U.S. Observations GEOS-Chem Total Fires Fossil fuel Biogenic SO x WSOC (R. Weber) Model OC attribution low-altitude WSOC aircraft measurements are consistent with IMPROVE data Observed free tropospheric concentrations are lower than in ACE-Asia; model is only 25% too low in the mean for both boundary layer and free trop. …but is incapable of reproducing observed variability in boundary layer (r 2 = 0.11) or free troposphere (r 2 = 0.05) Observed free tropospheric WSOC has significant correlation (r 2 ~ 0.3) with methanol combined with sulfate, nitrate, or toluene; suggests cloud production mechanism and combined bio-anthro source – DICARBONYLS? Heald et al. [2006] NOAA ITCT-2K4 aircraft campaign in Jul-Aug 2004 out of Portsmouth, NH

GLOBAL GLYOXAL BUDGET IN GEOS-Chem Including reactive uptake by aq. aerosols + clouds with  =2x10 -3 [Liggio et al., 2005] Global SOA formation of 18 Tg yr -1 (12 in clear sky + 6 in cloud); compare to 16 Tg yr -1 from terpenes/isoprene by semivolatile mechanism May Fu, Harvard, in prep. 10  = 1.5 h

GLOBAL METHYLGLYOXAL BUDGET IN GEOS-Chem Including reactive uptake by aerosols and clouds with  =2x10 -3 Global SOA formation of 38 Tg yr -1 (21 in clear sky + 17 in cloud); compare to 16 Tg yr -1 from terpenes/isoprene by semivolatile mechanism May Fu, Harvard, in prep  = 0.9 h

TESTING DUST EMISSION SCHEMES WITH IMPROVE DATA GOCART [Ginoux et al., 2004], DEAD [Zender et al.,2003] and combined schemes implemented in GEOS-Chem global simulation IMPROVE dust (annual mean, 2001) GEOS-Chem w/GOCART GEOS-Chem w/DEAD GEOS-Chem w/combined  g m -3 GOCART has two low wind threshold DEAD has insufficient areal restriction combination of two (GOCART dust areas + DEAD wind threshold) largely removes biases Fairlie et al. [2007] bias = 0.17 RMS=0.43 bias = 0.41 RMS=0.53 bias = RMS=0.26

ASIAN CONTRIBUTION TO U.S. DUST (2001)  g m -3 Model simulation of spring 2001 Asian outflow (TRACE-P aircraft data) Annual time series; Asian dust is combination of events + seasonal background Lava Beds Great Smoky Mountains observed Asian (model) Spring (MAM) is max dust season in most of U.S., and source is mainly from Asia Fairlie et al. [2007]

ADJOINT MODEL APPROACH TO CONSTRAIN PM NITRATE SOURCES IMPROVE (Jan 2002) GEOS-Chem prior simulation (4 o x5 o ) Nitrate Sulfate Henze et al. [2007] Use LSQ cost function J to fit [NO 3 - ] and [SO 4 2- ] as function of model parameters

SENSITIVITY OF COST FUNCTION TO MODEL PARAMETERS NO x emissions N 2 O 5 reactive uptake by aerosols Anthropogenic NH 3 emissions Initial conditions Henze et al. [2007]

NH 4 +, observed NH 4 +, model (prior) NH 4 +, model (posterior) Henze et al. [2007] OPTIMIZATION OF NH 3 EMISSIONS USING NO 3 - AND SO 4 2- DATA Anthropogenic NH 3 Natural NH 3 Prior [molec x / cm 2 ] Prior GEIA 0.11 Tg N mo -1 Posterior 0.10 Tg N mo -1 Other recent national estimates are in range Tg N mo -1 [Gilliland et al., 2003, 2006; Pinder et al., 2003, 2006; Park et al., 2004] but don’t constrain spatial variability Jan 2002

IMPROVING MODIS SATELLITE RETRIEVALS OF AEROSOL OPTICAL DEPTHS OVER LAND SURFACE AEROSOL 0.47  m 0.65  m 2.13  m Interpretation of TOA reflectance in terms of AOD requires assumptions on surface reflectance, aerosol optical properties Use TOA reflectance at 2.13  m (transparent atmosphere) to derive surface reflectance MODIS operational algorithm relies on general assumptions for 0.47/2.13 and 0.65/2.13 surface reflectance ratios and aerosol optical properties; we improve by deriving those locally using GEOS-Chem model information The resulting AOD retrieval (or model reflectance) is suited for quantitative interpretation including inverse analyses MODIS measures top-of-atmosphere (TOA) reflectance in several wavelength channels Drury et al. [2007]

MODIS vs. AERONET 0.47  m AODs (Jul-Aug 2004) MODIS (this work) MODIS (collection 4) AERONET MODIS (collection 5) Drury et al. [2007]