GEOS–CHEM: The Agony and The Ecstasy Bob Yantosca Software Engineer Atmospheric Chemistry Modeling Group Harvard University Group Meeting / Telecon 06

GEOS–CHEM: The Agony and The Ecstasy

Bob Yantosca

Software Engineer

Atmospheric Chemistry Modeling Group

Harvard University

Group Meeting / Telecon

06 October 2004

06 October 2004GEOS-CHEM: The Agony and the Ecstacy2

Topics

1. Transition to GEOS–4 Met Fields

2. A Summary of Recent Scientific Upgrades to GEOS–CHEM

3. GEOS–CHEM v7–01–02: New User Interface!

4. Future Directions

Appendix: ICARTT NRT Simulations


Part 1

Transition to GEOS–4



GEOS–4 Overview As of October 31, 2002, GEOS–4 is now the “operational”

data product generated by NASA GMAO

GEOS–4 is very different than GEOS–3 in many ways: GEOS–4 uses a different GCM than before (NCAR fvCCM) GEOS–4 is a hybrid grid (55 vertical levels) GEOS–4 has a 1 x 1.25 horizontal grid (not 1 x 1!) Several GEOS–4 quantities have different units than GEOS–3 Some GEOS–4 fields did not exist in GEOS–3 Et cetera …

GEOS–CHEM had to evolve to deal with GEOS–4!!


Transition to GEOS–4 Comparison of GEOS vertical layers in the PBL

GEOS–3 has 8 levels up to ~850 hPa, but GEOS–4 only has 4

850 hPa



Differences from GEOS–3 Some GEOS–4 quantities can be very different than

GEOS–3, for example: Cloud optical depth & cloud mass fluxes Surface wetness, snow cover, roughness height Precipitation fields

How does this impact a GEOS–CHEM simulation? Cloud optical depth affects J-Values Cloud mass fluxes affects cloud convection Surface wetness & snow cover affects dust mobilization Roughness height affects dry deposition Precipitation affects rainout & washout


Transition to GEOS–4Slide from

J. Logan & S. Wu

????

GEOS–4 July mean Cld Frac GEOS–3 July mean Cld Frac

GEOS–4 July mean OPTD GEOS–3 July mean OPTD

Cloud frac does not show up as cloud

opt depth!


Transition to GEOS–4GEOS–3 Column OPTD, July 2001 GEOS–4 Column OPTD, July 2003

Slide from

J. Logan & S. Wu

Larger O1D over Asian Subcontient!



0 4

GWET = surface wetnessranges from 0 (dry) to 1 (totally wet)

GWET has important implications for dust mobilization. Where it’s wet you

don’t have dust, but mud – this suppresses dust mobilization!

GEOS–4 is much wetter in Africa, SE

Asia & Americas!

-0.33 0 0.78 [unitless]

GEOS–3 GWET at 0 GMT 2002/07/01 GEOS–4 GWET at 0 GMT 2002/07/01

Abs Diff GEOS–4 – GEOS–3



Roughness height (Z0) is a function of land type and is used for dry deposition computations.

The differences in roughness height indicate that GEOS–4 uses a much different land-surface model than GEOS–3. (Don’t know if it’s better!)

GEOS–3 Z0 at 0 GMT 2002/07/01 GEOS–4 Z0 at 0 GMT 2002/07/01

Abs Diff GEOS–4 – GEOS–3

-1.96 0 1.27 [m]

0.0 1.30 2.60 [m]



Differences in the WINDS Synoptic values (GEOS–3) Averaged values (GEOS–4)

U and V winds, Specific Humidity, Temperature, Vis. Albedo

GEOS–4 winds required brand-new transport code We added a totally new version of TPCORE (cf. S-J Lin) We also installed the LLNL P-Fixer (cf. Phil Cameron-Smith)

TPCORE mass-flux diags were rewritten for GEOS–4 Brendan Field recently did this; will be added in v7–01–02

Nested grid does not yet work with GEOS–4 New TPCORE has yet to be modified for boundary conditions Also, GEOS–4 horizontal grid is 1 x 1.25, not 1 x 1



Differences in CLOUD CONVECTION GEOS–4 cloud convection fields are totally different

GEOS–3 style CLDMAS and DTRAIN are no longer provided

GEOS–4 uses the NCAR convection scheme (cf. P Rasch) Hack (shallow) convection: HKETA, HKBETA Zhang/McFarlane (deep) convection: ZMMU, ZMMD, ZMEU Scheme includes both updrafts & downdrafts

New GEOS–4 met fields required new convection code Convection code had to be stripped out of MATCH model We also had to re-install the wet scavenging in cloud updrafts Almost all of this work was done by Shiliang Wu



Differences between GEOS–4 and GEOS–4 To complicate matters, there were 2 versions of GEOS–4

Version 3: (GMAO: 1.3 r7) was made from 2002 thru Jan 2004 Version 4: (GMAO: 1.4 r2) was introduced in Jan 2004

GMAO says V4 contains improvements over V3: New Land Model Improvements to Skin Temperature analysis NOAA-17 satellite assimilation MODIS cloud track winds assimilation over the poles Relative Humidity fix near surface Diag wind output stream for stratospheric transport customers

http://gmao.gsfc.nasa.gov/operations/recent_mods.php



Differences between GEOS–4 and GEOS–4 An annoyance: the averaging periods for certain fields

were different in GEOS–4 V4 than in GEOS–4 V3 In GEOS–4 v4, the averaging periods for “A6” fields are the

same as in the GEOS–1, GEOS–STRAT, and GEOS–3

In GEOS–4 v3, the averaging periods for the “A6” fields are shifted 3 hours later than V4.

“A6” fields = 6-hour average fields. In GEOS–4, these include winds, temperature, specific humidity.

IF YOU THINK IT’S CONFUSING, YOU’RE NOT ALONE!



Z = “Zulu” = abbreviation for GMT

From the fvDAS file spec on GMAO’s website

GEOS-1

GEOS-S

GEOS-3

GEOS-4 ver. 4

GEOS-4 ver. 3

(now obsolete)

A-6 fields

A-3 fields

A-6 fields

A-3 fields

Times of day when you need to read 6-hr avg met fields

Start of day Start of next day



Differences between GEOS–4 and GEOS–4 File formats between GMAO “raw” data files vary

GEOS–4 V3: most are HDF–EOS; some are HDF GEOS–4 V4: all files are HDF–EOS

Problems To process the GEOS–4 V3 files, I had to write new F90 code that

could read the HDF data files. When GEOS–4 V4 was introduced, I didn’t need to use the code

to read the HDF reader code anymore. Some of the file and variable names in GEOS–4 V4 were different

from V3, which required me to modify the scripts & code that I used to process the data.

THIS WAS VERY TIME CONSUMING !!!!!



GEOS–CHEM v6–01–05 was 1st GEOS–4 compatible version This version contained the following features:

Release Date: 09 April 2004 Could only support GEOS–4 V3 met fields Introduced support for compiling on Altix platform Contained new GEOS–4 transport code & LLNL P-fixer Contained new convection code for GEOS–4 met fields GEOS–4 Lightning emissions are now scaled to 6 Tg N/yr Bug fix: ACET emissions are the same independent of grid Other fixes here and there to enable GEOS–4 met fields

A 1-yr benchmark was done with v6–01–05 (year 2003) See: www-as.harvard.edu/chemistry/trop/geos/geos_1yr.html



Philosophy of 1–year benchmarks: Conduct simulations that are as similar as possible with two

sets of met. data - from different GCMs.

Met. data: GEOS–3 run for 2001, GEOS–4 (V3) run for 2003 (same year would have been ideal)

Chemical mechanism and reaction rates: Identical

Emissions: Identical—anthropogenic, biomass burning Scaled to the same annual total—isoprene, NOx from lightning

SYNOZ (stratospheric source of ozone) Spun up for each model (494 Tg/yr in GEOS–4)

Slide from

J. Logan


Transition to GEOS–4 OH (x 106) MCF lifetime (yr)GEOS-STRATrvm, V. 4.26 1.21 5.61 Old Chem.Sept. 96-Aug. 97

GEOS-3V. 5.07.08 1.08 6.53 New chem.2001

GEOS-4 V. 6.01.05 1.17 5.55 New chem2003

Ozone budget GEOS-3 GEOS-4 GEOS-STRAT

Prod. ozone (Tg/y) 4383 5087 4924Loss ozone (Tg/y) 3830 4540 4377

Slide fromJ. Logan

OH changes

by approx 10% from

GEOS–3 to GEOS–4!!

O3 P&L are 16–18%larger in GEOS-4



Why did OH change so much?

It is not easy to change mean OH by 10% when the chemistry is the same, water vapor is similar, the ozone column is the same.

One likely candidate is the cloud optical depth, as these are very different in the tropics.

Slide fromJ. Logan


Too much ozone at high latitudes in winter

GEOS-STRAT, GEOS-3, and GEOS-4

Slide from J. Logan



GEOS–4 Summary GEOS–4 V4 is now the stable, operational GMAO data set

We have GEOS–4 V4 data for 2004 to date GEOS–4 V3 is now obsolete; 2003 data will be replaced

GMAO is reanalyzing GEOS–4 data for years prior to 2004 This is the CERES reanalysis product Soon we will have GEOS–4 V4 data from 2001 thru 2004

GEOS–CHEM v6 & higher are now GEOS–4 compatible

GEOS–CHEM OH & Ozone are higher with GEOS–4 Likely culprit: lower GEOS–4 cloud optical depths


Part 2

A Summary of Recent Scientific Upgrades

to GEOS–CHEM


Scientific Upgrades

Scientific Upgrades to GEOS–CHEM The transition to GEOS–4 took longer than planned

For all of the reasons mentioned previously, plus… Spin-up simulation w/ offline Tagged Ox was necessary in order

to generate initial conditions w/ Ox in steady-state (for Synoz)

In the meantime, several additional scientific upgrades had become mature, but were languishing in the pipeline Black Carbon and Organic Carbon Aerosols (R. Park) Secondary Organic Aerosols (H. Liao, CALTECH) Desert Dust Aerosols (D. Fairlie) Sea Salt Aerosols and SO2 Emissions from Ships (B. Alexander)

These upgrades were fast-tracked to be ready for ICARTT


Scientific Upgrades

Black Carbon & Organic Carbon Aerosols (cf. Rokjin Park): Adds 4 more transported tracers to previous fullchem sim

Hydrophilic and Hydrophobic Black Carbon Hydrophilic and Hydrophobic Organic Carbon

Emissions: Anthropogenic: From Bond et al 2004 (default) or

Cooke et al 1999 with seasonality imposed by R. Park Biomass: From GEOS–CHEM (default) or Bond et al 2004

Chemistry: A fraction of hydrophobic BC & OC hydrophilic BC & OC Hydrophobic BC & OC dry deposit Hydrophilic BC & OC dry deposit and wet deposit BC & OC aerosols also affect J-values and het chem rates


Scientific Upgrades

Black Carbon & Organic Carbon Aerosols (cf. Rokjin Park): Emission totals with GEOS–4 2003 (Tg C/yr) are below 20% of total BC emitted is Hydrophilic 50% of total OC emitted is Hydrophilic

Emissions Anthro Biofuel Biomass Biogenic Total

H-phobic BC 2.51 1.31 1.83 - 5.65

H-philic BC 0.62 0.33 0.46 - 1.41

Total BC 3.13 1.64 2.29 - 7.06

H-phobic OC 1.22 3.24 9.35 6.02 19.83

H-philic OC 1.22 3.24 9.35 6.02 19.83

Total OC 2.44 6.48 18.70 12.04 39.66


Scientific Upgrades

Secondary Organic Aerosols (cf. Hong Liao, Caltech): Adds 9 more transported tracers

ALPH: a-pinene, b-pinene, sabinene, carene, terpenoid ketones LIMO: Limonene ALCO: myrcene, terpenoid alcohols, ocimene SOG1: lump of gas products of ALPH+LIMO+TERP HC ox. SOG2: gas product of ALCO oxidation SOG3: gas product of SESQ oxidation SOA1: lump of aerosol products of ALPH+LIMO+TERP) HC ox. SOA2: aerosol product of ALCO oxidation SOA3: aerosol product of SESQ oxidation

You can use these in both the FULLCHEM or OFFLINE AEROSOL simulations!


Scientific Upgrades

Desert Dust Aerosols (cf. Duncan Fairlie): Adds 4 more transported tracers to G–C fullchem sim

Dust bins w/ Reff = 0.7, 1.4, 2.4, and 4.5 m (NOTE: there are 4 transport dust bins, but 7 FAST–J bins)

Emissions: Source function from either the DEAD model (C. Zender) or

from GOCART (P. Ginoux) can be used. DEAD is the default. DEAD scheme is more detailed than the GOCART scheme, and

also takes as input more physical fields

Chemistry: Dust aerosols both dry settle and dry deposit Dust aerosols also affect J-values and het chem rates


Scientific Upgrades

Desert Dust Aerosols (cf. Duncan Fairlie): GEOS–4 dust emissions are 3 – 4 X lower than GEOS–3

This may be due to the higher surface wetness in GEOS–4 Duncan is currently looking into this…

Dust Reff(m)

GEOS–3 2001 (Tg/yr)

GEOS–4 2003 (Tg/yr)

0.7 117.6 35.2

1.4 243.3 72.8

2.4 310.0 92.8

4.5 290.4 86.9


Scientific Upgrades

Sea Salt Aerosols (cf. Becky Alexander): Adds 2 more transported tracers to G–C fullchem sim

Accumulation Mode: Reff = 0.1 – 2.5 m• NOTE: Accumulation mode seasalt should REALLY be 0.1 – 0.5

m. This will be the default in v7–01–02. Coarse Mode: Reff = 2.5 – 10.0 m

Emissions: Uses parameterization from Monahan et al (1996) as described

by Gong et al (1997) (Ask Becky to explain!)

Chemistry: Sea salt aerosols wet settle, wet deposit, and dry deposit Sea salt aerosols also affect J-values and het chem rates


Scientific UpgradesAccum Sea Salt, Jan 2003

Accum Sea Salt, Jul 2003

Coarse Sea Salt, Jan 2003

Coarse Sea Salt, Jul 2003

EMISSIONS

0.0 0.5 1.0 1.5 [Tg]


Scientific Upgrades

Month Accumulation (0.1 – 2.5 mm)

Coarse (2.5 – 10 mm)

Jan 387 257

Feb 346 229

Mar 367 244

Apr 363 241

May 384 255

Jun 344 228

Jul 416 276

Aug 367 243

Sep 346 230

Oct 344 228

Nov 353 234

Dec 376 250

Total 4394 2913

Emission totals for

Accumulation Mode and

Coarse Mode Sea Salt

Aerosols in GEOS–CHEM using GEOS–4 winds for 2003

Units: Tg/yr


Scientific Upgrades

Ship SO2 emissions (cf. Becky Alexander): Doesn’t add any extra tracers to G–C fullchem simulation

Ship SO2 exhaust is added to anthropogenic SO2 emissions

Emissions From Corbett et al. Original emissions are on a 2 x 2 degree grid Regridded to 2 x 2.5 and 4 x 5 GEOS grids by Becky Alexander

Global Total 4.2 Tg S/yr is emitted from ship exhaust


Scientific UpgradesAnnual Avg Distribution of SO2 Emissions from Ship Exhaust

Major shipping lanes are readily visible in this plot!

0 6.6 [Mg S]2.2 4.4

Rotterdam, NL

Annual Total = 4.2 Tg S /yr

Great Lakes


Scientific Upgrades

Summary: GEOS–CHEM v6–02–05 contains the following mature

scientific upgrades over v6–01–05 Black carbon & organic carbon aerosols Mineral dust aerosols Sea salt aerosols Ship SO2 exhaust is added to anthropogenic SO2 emissions

GEOS–CHEM v6–03–02 (internal release) contains Secondary organic aerosols

GEOS–CHEM v6–02–05, v6–03–02: & higher can use: GEOS–4 V4 met data GEOS–4 V3 met data


Part 3

GEOS–CHEM v7–01–02:New User Interface!


GEOS–CHEM v7–01–02

Logistical shortcomings in existing GEOS–CHEM versions: GEOS–CHEM user interface is too confusing!

Switches have to be set in 6 or 7 files to specify output options It is easy to make mistakes It is hard to add new functionality There is LOTS OF HISTORICAL BAGGAGE in the input files There is a steep learning curve for GEOS–CHEM users

Timeseries diagnostics need rewriting Tracer #’s were not consistent between ND49, ND50, ND51 ND48 station timeseries diagnostic output is hard to understand Timeseries diagnostics were not parallelized



Existing GEOS–CHEM user input files: input.geos: specifies start & end dates; turn operations on/off input.ctm: schedule diagnostics and dates for output inptr.ctm: specifies tracer names & molecular weights tracer.dat: specifies which tracer constituents are emitted prodloss.dat: specifies prod/loss families for diagnostic output diag.dat: specifies which tracers you want to print out timeseries.dat: specifies options for ND49, ND50, ND51

These are now all rolled into a single file! The new input.geos

Some “sneak peeks” of the new input.geos file follow…



GEOS-CHEM SIMULATION v7-01-02

------------------------+-------------------------------------------

%%% SIMULATION MENU %%% :

Start YYYYMMDD, HHMMSS : 20020915 000000

End YYYYMMDD, HHMMSS : 20020916 000000

Run directory : /users/ctm/bmy/T/run.v7-01-02.2x25/

Input restart file : restart.YYYYMMDDhh

Make new restart file? : T

Output restart file(s) : restart.YYYYMMDDhh

Root data directory : /data/ctm/GEOS_2x2.5/

-- GEOS-1 subdir : GEOS_1/YYYY/MM/

-- GEOS-STRAT subdir : GEOS_STRAT/YYYY/MM/

-- GEOS-3 subdir : GEOS_3/YYYY/MM/

-- GEOS-4 subdir : GEOS_4_v4/YYYY/MM/

Temporary directory : /users/ctm/bmy/TEMP/T1/

Unzip met fields? : F

Wait for met fields? : F

Global offsets I0, J0 : 0 0



%%% TRACER MENU %%% :

Type of simulation : 3

Number of Tracers : 41

Tracer Entries -------> : TR# Name g/mole Tracer Members; ()= em

Tracer #1 : 1 NOx 46.0 NO2 (NO) NO3 HNO2

Tracer #2 : 2 Ox 48.0 O3 NO2 2NO3

Tracer #3 : 3 PAN 121.0

Tracer #4 : 4 CO 28.0 (CO)

Tracer #5 : 5 ALK4 12.0 (4C)

Tracer #6 : 6 ISOP 12.0 (5C)

Tracer #7 : 7 HNO3 63.0

Tracer #8 : 8 H2O2 34.0

Tracer #9 : 9 ACET 12.0 (3C)

Tracer #10 : 10 MEK 12.0 (4C)

Tracer #11 : 11 ALD2 12.0 (2C)

Tracer #12 : 12 RCHO 58.0

Tracer #13 : 13 MVK 70.0

Etc . . .



%%% TRANSPORT MENU %%% :

Turn on Transport : T

-- Use Flux Correction?: F

-- Fill Negative Values: T

-- IORD, JORD, KORD : 3 3 7

Transport Timestep [min]: 30

Use strat O3/NOy BC's : T

------------------------+--------------------------------------------

%%% CONVECTION MENU %%% :

Turn on Cloud Conv? : T

Turn on PBL Mixing? : T

Convect Timestep (min) : 30

------------------------+--------------------------------------------

%%% DEPOSITION MENU %%% :

Turn on Dry Deposition? : T

Turn on Wet Deposition? : T



%%% EMISSIONS MENU %%% :

Turn on emissions? : T

Emiss timestep (min) : 60

Include anthro emiss? : T

-- scale 1985 to year : 1995

Include biofuel emiss? : T

Include biogenic emiss? : T

-- Scale ISOP to MONOT?: T

Include biomass emiss? : T

-- Seasonal biomass? : T

-- Scaled to TOMSAI? : F

Individual NOx sources :---

-- Use aircraft NOx? : T

-- Use lightning NOx : T

-- Use soil NOx : T

Use ship SO2 emissions? : T



Other improvements in GEOS–CHEM v7–01–02 Timeseries diagnostics were rewritten for consistency All logical switches were grouped into a single module All GEOS–CHEM directory paths grouped into a single module WETDEP code now parallelizes on the Altix platform Convection code now parallelizes on the Altix platform AOD’s can be saved from both fullchem & offline aerosol runs Tracer array is now allocatable; this will save memory Family prod/loss diagnostics (ND65) were cleaned up Mean OH diagnostic (ND23) were cleaned up Mass-flux diagnostics implemented into GEOS–4 TPCORE

v7– 01–02 should make everyone’s lives MUCH easier!



Thanks to GEOS–CHEM v7–01–02 Beta Testers! Prasad Kasibhatla Randall Martin Aaron von Donkelaar Becky Alexander Colette Heald Brendan Field (for mass-flux diagnostics!) and others!


Part 4

Future Directions


Future Directions

Continue pulling met data for GEOS–4 V4 late-look data 2001, 2003 is now available 1996, 1997 may be available soon

JPL is interested in doing the following … To solve the scalability problem in GEOS–CHEM ?? To code GEOS–CHEM for MPI parallelization so that it can be

used on distributed memory machines (e.g. Linux clusters) People:

• Qinbin Li et al (JPL), • Daven Henze (Caltech), • Hamid Oloso (GSFC), • Bob Y, Jack Yatteau, Daniel Jacob (Harvard)


Future Directions

Creating an adjoint of GEOS–CHEM Monika Kopacz, Parvadha Suntharalingam (Harvard) Dylan Jones (U. Toronto) Daven Henze (Caltech)

Aerosol microphysics modules Peter Adams et al (CMU)

Mercury chemistry Lyatt Jaegle et al (UWA) Noelle Eckley (Harvard)

CO2 chemistry Parvadha Suntharalingam (Harvard)


Future Directions

Interface with CMAQ model Daewon Byun (U. Houston) Joshua Fu (U. Tenn) Rokjin Park (Harvard)

Online data assimilation at GMAO Steven Pawson (GMAO) Daniel Jacob (Harvard)

GCAP: Interface with GISS–GCM winds Loretta Mickley & Shiliang Wu (Harvard) David Rind, Jean Lerner (GISS) Etc.


Future Directions

GEOS–5 is coming! GEOS–5 will become the operational GMAO data set in 2005 GEOS–4 will be turned off sometime next year

We will need to evaluate GEOS–CHEM with GEOS–5 fields Spinup with Synoz Ozone for 10 or 11 years Emissions will have to be tested (e.g. isoprene, acetone) Lightning NOx will have to be re-scaled What if cloud convection / optical depths change again? We will have to do a 1-year benchmark again

INTEX–B: 2006 GEOS–CHEM will have to use GEOS–5 fields for this Visualization website will be set up


Appendix

ICARTT Near-Real-Time Simulations


ICARTT NRT Simulations

Overview Harvard did daily NRT simulations for ICARTT

NRT = Near Real Time fullchem–aerosol coupled runs People: Solene, Carine & Bob. Also thanks to Rynda & Lyatt!

ICARTT model was v6–02–05 w/ minor modifications User starts each model run w/ a script – everything else (model

run & data transfer) happens automatically! Daily flight tracks were obtained for each DC8, P3B, Falcon, etc.

flight and imported into GEOS–CHEM Model output was saved as daily means, 3-D timeseries, daily

columns, and along the mission flight tracks Output data was archived on geos.as.harvard.edu server Output data was also downloaded by Lyatt’s web server at UW


ICARTT NRT Simulations

My part of ICARTT Script & code to obtain the GEOS–4 met data from GMAO

Pre-programmed to start automatically each day!

IDL code to make input files for G–C plane flight diag. Ingests flt tracks from NASA DC-8, NOAA, WP-3B, DLR Falcon

Script to start GEOS–CHEM runs automatically Downloads input and restart files from geos.as.harvard.edu Submits a GEOS–CHEM job to Europa via the queue system Pushes diagnostic output files to geos.as.harvard.edu

Other Stuff EPA emissions prep, regular GAMAP development


Extra Slides …



GEOS–4 vertical grid is hybrid; GEOS–3 is pure-sigma

Pure-sigma formula for computing pressure:

1) P(I,J,L) = (L) * ( Ps(I,J) – PTOP ) + PTOP

Hybrid formula for computing pressure:

2) P(I,J,L) = ( Bp(L) * Ps(I,J) ) + Ap(L)

The pure-sigma formula can be expressed in hybrid terms if: Bp(L) = e(L) and Ap(L) = PTOP for all L

All instances of Eq. 1 in the code had to be replaced by Eq. 2


Transition to GEOS–4GEOS–3 OPTD at 0 GMT 2002/07/01 GEOS–4 OPTD at 0 GMT 2002/07/01

Where arethe clouds?

0 1 2 4 [unitless] 0 1 2 4 [unitless]


Scientific Upgrades0.7 m 1.4 m

2.4 m 4.5 m

GEOS-3 dust in July


Scientific Upgrades0.7 m 1.4 m

2.4 m 4.5 m

GEOS-4 dust in July

Documents

GEOS–CHEM: The Agony and The Ecstasy Bob Yantosca Software Engineer Atmospheric Chemistry Modeling Group Harvard University Group Meeting / Telecon 06