GEOS-Chem Adjoint

From Geos-chem
Revision as of 02:16, 9 April 2009 by Kumaresh (Talk | contribs) (Modes)

Jump to: navigation, search


Historical Overview

Original work on the adjoint of GEOS-Chem began in 2003, focusing on the adjoint of the offline aerosol simulation. By 2005, the adjoint was expanded to include a tagged CO simulation and a full chemistry simulation as well as observational operators for MOPITT (CO) and IMPROVE network (aerosols). The adjoint code has been constructed in a hybrid fashion using a combination of automatic differentiation software (TAMC, KPP) and manual coding of both discrete and continuous adjoints. Current development aims to standardize the implementation of the adjoint model and make it more user friendly, with the eventual goal of making the adjoint publicly available as part of the standard GEOS-Chem code base.


  • Caltech: John Seinfeld
  • Carleton University: Amir Hakami
  • Columbia / NASA GISS / CU Boulder : Daven Henze
  • Dalhousie: Randall Martin, Chulkyu Lee
  • Harvard: Monika Kopacz, Lin Zhang, Daniel Jacob
  • JPL: Kevin Bowman, Qinbin Li, Xun Jiang, Changsub Shim, Yang Chen, Sunita Verma
  • U. of Toronto: Dylan Jones, Mark Parrington, Michael Seymour, Thomas Walker
  • Virginia Tech: Adrian Sandu, Kumaresh Singh, Paul Eller

Current Activities

  • streamlining GEOS4 adjoint transport among the three code branches, some bug fixes to be implemented

Get Involved

Support and development of the GEOS-Chem adjoint has clearly become a multifaceted undertaking. In attempt to focus such developments, we've created a GC-adjoint mailing list. If you wish to become involved, you should add yourself to this list by sending an email to with the word subscribe in the subject or body.

Kevin Bowman organizes bimonthly conference calls concerning current adjoint model activities.

Adjoint Model Branches

There are currently three main branches of the GEOS-Chem adjoint model.

  • GEOS-Chem v6-02-05, full chemistry with online aerosols
  • GEOS-Chem v6, tagged CO (and similar tagged Ox)
  • GEOS-Chem v7, full chemistry

GEOS-Chem v6-02-05, full chemistry with online aerosols


  • GEOS-3 and GEOS-4 met fields
  • 2x2.5, 4x5 model resolution
  • Observational Operators
    • IMPROVE PM2.5 (NO3, SO4, OC, BC)
    • CASTNet (NH4+)
    • GOME / SCIAMACHY NO2 column
      • using KNMI retrieval (Henze)
      • using Dalhousie retrieval (Shim)
    • TES NH3
  • Control parameters
    • Emissions scaling factors (linear or log)
      • SOx, NH3, primary BC/OC: anthropogenic, natural, bioburn, biomass, volcanoes
    • Initial Conditions scaling factors (linear)
  • Adjoint sensitivities
    • all control parameters
    • w.r.t Reaction Rate Parameters
    • w.r.t all emissions
    • to AQ attainment metrics
    • to spatiotemporally averaged species concentrations (e.g., arctic O3)
  • Inverse Hessian approximation
  • Code manual
  • Code diagrams

Primary Developer

Daven Henze


Daven Henze (Columbia University), Chulkyu Lee (Dalhousie), Dylan Jones, Mark Parrington, Thomas Walker (U. Toronto), Changsub Shim (JPL)


  • Henze, D. K., A. Hakami and J. H. Seinfeld (2007), Development of the adjoint of GEOS-Chem, Atmos. Chem. Phys., 7, 2413-2433.
  • Henze, D. K., J. H. Seinfeld and D. T. Shindell (2008), Inverse modeling and mapping U.S. air quality influences of inorganic PM2.5 precursor emissions with the adjoint of GEOS-Chem, Atmos. Chem. Phys. Discuss., 8, 15033-15099.

GEOS-Chem v6-02-05 (builds on dkh version), tagged CO


Primary Developer

Monika Kopacz


Monika Kopacz (Harvard), Guergana Guerova (U. Wollongong), Dylan Jones (U. Toronto), Lin Zhang (Harvard, Ox branch)

Publications and presentations

GEOS-Chem v7, full chemistry

A full chemistry adjoint model for GEOS-Chem v7, geos-4. The package includes the following modes and features.


  • Forward GEOS-Chem with SMVGEAR chemistry
  • Forward GEOS-Chem with KPP chemistry
  • Adjoint Finite Difference module to test old and new process adjoints
  • Adjoint Sensitivity Analysis module to calculate sensitivities with respect to tracer and emission species
  • Variational Data Assimilation code modules with diagonal and non-diagonal background error covariance matrix with real data (TES observations) to perform
    • 3-D Var Data Assimilation
    • 4-D Var Data Assimilation


  • GEOS-4 met fields
  • 4x5 model resolution
  • Observational Operators
    • TES O3
  • Control parameters
    • Emissions scaling factors
    • Initial Conditions
  • Adjoint sensitivities w.r.t.
    • all control parameters
    • Reaction Rate Parameters
  • Project Page

Primary Developer

Kumaresh Singh


Dylan Jones (U. Toronto), Changsub Shim (JPL)