aerosols and adjoint modeling · 2009-02-06 · daven henze aerosols and adjoint modeling . aerosol...

University of Colorado, Boulder

WRF-Chem Data Assimilation Workshop, NCAR

Daven Henze

Aerosols and

Adjoint Modeling

Aerosol (PM2.5) Data Assimilation

• PM2.5 is key component of - air pollution - climate forcing

• Current models suffer from large uncertainties - sources - mechanisms

• Observations widespread - surface networks: IMPROVE, CASTNet, STN... - remote: chemical precursors (SO2, NO2,

NH3,...) - remote: optical properties (MODIS to APS)

Aerosol adjoint modeling

Areas of development:

• Aerosol microphysics

• Heterogeneous chemistry and thermodynamics

• Radiative transfer

NO2 (g) HNO3(g) NO3-(s)

Aerosol Microphysics

Aerosol dynamics equation (mass-based formulation):

Condensation / Evaporation

Coagulation Nucleation Loss (deposition)

Developed adjoints for several model formulations of aerosol microphysics

Aerosol Microphysics

Box model tests of inverse modeling capabilities:

IC for ideal multi-component mixture with explicit growth

IC, coagulation kernal and growth parameter

Henze et al. (2004) Sandu et al. (2005)

PM2.5 in a global model: heterogeneous chemistry and thermodynamics

SO2 SO2 + OH

HNO3 NO2 + OH

SO2 + O3 , H2O2

N2O5 + H2O

NOx

NH3

SO42-

NO3-

NH4+

gas-phase thermodynamics

aerosol chemistry

Emissions

H2SO4

wet and dry loss

• Forward model v6-02-05 (Park et al., 2004) • GEOS assimilated meteorology • Global (4°x5°), 30 vertical levels • HOx - NOx - HC gas-phase chemistry • Inorganic aerosol (secondary) • Carbonaceous aerosol (primary) • built manually & automatically (TAMC, KPP)

• Additional developers - K. Singh (GEOS-4), P. Eller, A Sandu (KPP), K. Bowman (TES) - M. Kopacz (CO) - C. Shim (2°x2.5°)

GEOS-Chem Adjoint

Calculates sensitivity of PM2.5 concentrations w.r.t precursor emissions, initial conditions and rxn rate coefs

GEOS-Chem adjoint: Inorganic PM2.5

(thermo only)

Adjoint model validation (Henze et al., 2007)

GEOS-Chem adjoint: Inorganic PM2.5

(thermo only)

Adjoint model validation (Henze et al., 2007)

Constraining U.S. NH3 emissions (Henze et al., 2008)

Radiative Flux (forward calculation)

Chemical Transport Model

Mie Code

Aerosol concentrations

optical properties

Radiative Flux (forward calculation)

Chemical Transport Model

Mie Code

Aerosol concentrations

optical properties

Radiative Transfer Code

TOA upward SW fluxes

Radiative Flux Sensitivity

LIDORT Spurr, 2002

Jacobian calculation

GEOS-Chem Adj Henze et al., 2007

Mie Derivative

[SIA]*, [BC]*

Grainger et al., 2004

Radiative forcing sensitivity by sector, species and location

BC fossil fuel BC bio fuel

NH3 anthropogenic SO2 shipping

Yearly average sensitivities for several sectors:

Aerosol adjoint modeling

Areas of development:

• Aerosol microphysics

• Heterogeneous chemistry and thermodynamics

• Radiative transfer

PM2.5 4D-Var with GOCART: Dubovik et al., 2008

Aerosol adjoint modeling

Areas for development:

• Aerosol microphysics

• Heterogeneous chemistry and thermodynamics

• Radiative transfer

Aerosol adjoint modeling

Areas for development:

• Aerosol microphysics

• Heterogeneous chemistry and thermodynamics

• Radiative transfer

• Meteorology

WRF-Chem

Future Directions

Areas for development:

• Integration of microphysics with chemistry adjoints

• Implementation with WRF-Chem

• Effects of aerosols on photolysis

• Organic aerosol

The end

AQ Attainment

Emissions

Consider metric of PM2.5 air quality,

July

Nonattainment

Effectiveness = adjoint sens =

Effectiveness

GEOS-Chem adjoint: Inorganic PM2.5

Aerosol Microphysics

Box model tests of inverse modeling capabilities:

IC for real multi-component mixture with explicit growth

IC, coagulation kernal and growth parameter

Henze 2005 (M-Thesis) Sandu et al. (2005)

Forward sensitivity

The change in all , or some , is calculated separately for one .

Adjoint sensitivity

Adjoint sensitivities propagate from one , or , to all .

Ideal for calculating the sensitivity of a few things with respect to many things.

Testing the Adjoint Model: Gradient Check

Check gradient using finite difference calculation

Component-wise analysis affords domain wide points-of-comparison

model response (radiative forcing)

control parameter (emissions)

adjoint sensitivity

Testing the Adjoint: single processes, 1 week

(thermo only)

(chem only)

Validating Radiative Forcing Sensitivity

Finite difference

Adjo

int

Anth BC Emis Anth SOx Emis

- Truncation of phase moments bias - Lookup table for Mie calculation variance

Validating Radiative Forcing Sensitivity

ignore

Phase function coefficients for SIA(Dwet)

Dmax

Dmin

Validating Radiative Forcing Sensitivity

Mie results for extinction at discrete mode diameters:

AQ Attainment

Nonattainment Emissions Effectiveness

Januar

y

Effectiveness of emissions controls have critical spatial, seasonal, sector and chemical variability

(NH4+)2SO4

2- (mw = 132) 2NH4+NO3

- (mw = 2x80) West et al. 1999; Vayenas et al., 2005; Pinder et al., 2007