HR: 0800h
AN: A11B-0873    [Abstracts]
TI: Inverse Modeling of Aerosols Using the Adjoint of GEOS-CHEM
AU: * Henze, D K
EM: daven@caltech.edu
AF: Caltech, 1200 E California Blvd, Pasadena, CA 91125 United States
AU: Seinfeld, J H
EM: seinfeld@caltech.edu
AF: Caltech, 1200 E California Blvd, Pasadena, CA 91125 United States
AB: The relationship between gas phase emissions and resulting aerosol concentrations is well established, yet our ability to analyze precursor emissions inventories based upon ambient particulate measurements has been limited until now due to difficulties that aerosols pose when formulating an inverse model. Overcoming these difficulties using a combination of manual code processing and automatic differentiation, we present an inverse chemical transport model (GEOS-CHEM) created using the adjoint method which includes all main physical process (advection, wet and dry deposition, turbulent mixing, deep convection), tropospheric chemistry, and, for the first time, aerosol heterogeneous chemistry and thermodynamics. Using simulated observations, we force the adjoint model with a cost function that captures the discrepancy between simulated and observed aerosol mass concentrations. Using the adjoint model, we calculate the sensitivity of this cost function with respect to all precursor emissions in an efficient manner, allowing us to establish a global picture of the importance of various tracers (NO_x, HO_x, SO_x) and their emissions on simulated aerosol SO₄, NH₄ and NO₃. Black carbon (hydrophobic and hydrophilic), dust and sea salt species are also included in the analysis. The validity of the computed gradients are assessed by comparison to finite difference calculations and intercomparison (discrete vs. continuous) of various adjoint techniques. Feasibility of using such gradients for large scale optimization (data assimilation) is explored.
DE: 0305 Aerosols and particles (0345, 4801, 4906)
DE: 0345 Pollution: urban and regional (0305, 0478, 4251)
DE: 0368 Troposphere: constituent transport and chemistry
DE: 3315 Data assimilation
DE: 3355 Regional modeling
SC: Atmospheric Sciences [A]
MN: Fall Meeting 2005