The GEOS–5 met fields: What’s new and what’s different ...acmg.seas.harvard.edu/presentations/2006/bmy_group_mtg_dec_6_2006.pdfÆInstallation of libraries (HDF, HDF–EOS, ESMF,

The GEOS–5 met fields:What’s new and what’s different

Plus a quick look at some recentGEOS–Chem developments

Bob YantoscaSoftware Engineer

Atmospheric Chemistry Modeling GroupHarvard University

Jacob Group MeetingWednesday, December 6, 2006

Table of Contents

1. A little housekeeping and quick look at recent GEOS–Chem updates

2. GEOS–5 met fields (in gory detail w/ lots of pictures)6-hour instantaneous fields (a.k.a “I6” fields)

3-hour time-averaged fields (a.k.a “A3” fields)

6-hour time averaged fields (a.k.a “A6” fields)

3. Summary and GEOS–5 schedule

Nov 29 2006 Bob Y – Jacob Group Meeting – Page 2

A little housekeeping

+

A quick look at recent updatesto GEOS–Chem


A little housekeeping …We are here to help you!

Introducing Philippe Le SagerPhilippe started working with us in August 2006

Bob and Philippe have many years of experience in working on big software projectsPlease take advantage of our experience! Ask questions!

Please help us help you!Try to be proactive in debugging!

If you can isolate any errors to a particular section of GEOS–Chem, that really helps us out.

Also please be aware that there are many things we are working on simultaneouslyGMI, CTM-componentization, ESMF, Adjoint, GEOS–5, etc.


Division of LaborBob Y. will primarily be involved with software/system issues & short term stuff

Rewriting GEOS–Chem to facilitate implementation of Earth System Model Framework (ESMF)

Interfacing with NASA (GMAO, GMI, SIVO)

Installation of libraries (HDF, HDF–EOS, ESMF, netCDF) on different platforms

Porting of code to different platforms

Met fields (including GEOS–5)

Philippe will primarily be involved with long-term, open-ended, scientific projectsChemistry up to dynamic tropopause (now completed!)

New chemical solver (implementing Yevgeniy’s work)

Adjoint (working with Adrian Sandu) ???

Both Bob & Philippe will work on (for the time being):Minor maintenance bug fixes

Various user support issues

Version control


Some recent updates to the standard codeCurrent version in development: v7–04–11

Improvements and additions to chemistry Ability to do chemistry up to the location of the actual tropopause (Philippe, Brendan)

Added 2 tracers (SOG4, SOA4) to track the SOA prod from isoprene + OH reaction (Daven Henze)

HO2 uptake by aerosols is now turned off in the SMVGEAR mechanism (Bastien Sauvage)

Improvements and additions to emissionsOption to use GFED2 biomass emissions for gas phase + aerosol species

Option to use David Streets' regional emissions for China and SE Asia

Option to use EDGAR global NOx, CO, SO2 emissions

Option to use BRAVO NOx, CO, SO2 emissions over Northern Mexico

New module for IPCC future emissions scenarios (primarily for GCAP)

Near-land lightning NOx emissions module -- currently for GEOS-4 only (Lee Murray, Rynda Hudman)

Updated Hg simulationMore recent Hg simulation updates still need to be standardized

Technical updatesSupport for Sun 4100 platform (with AMD/Opteron chipset)

Removal of obsolete GEOS-1 and GEOS-STRAT met fields

Removal of support for obsolete LINUX_IFC and LINUX_EFC compilers

Various minor bug fixes & updates for the GCAP simulation

Other minor bug fixes and improvements


Things currently in the pipelineYet to be added to GEOS–Chem:

Implementation of a cleaner driver for the emissions code (from Brendan Field)

New, faster chemical solver (Philippe & Yevgeniy)

Updated dust emissions (from Duncan Fairlie)

Bug fix for SOA – save concentrations to restart file for future use (from Havala Taylor, Hong Liao)

Local rescaling for lightning emissions (from Bastien Sauvage)

Updated emission factors for “default” biomass burning (from Bastien Sauvage)

Hydrogen simulation (from Lyatt Jaegle et al at U. Washington)

COS simulation (Parv Suntharalingam)

CO—CO2 simulation (Monika Kopacz)

LINOZ – linear ozone chemistry (from Dylan Jones et al at U. Toronto)

Problems currently being fixed in GEOS–Chem:Need to apply post-2002 overhead O3 columns for photolysis

Existing TOMS O3 columns end in 2002

Improvements & fixes to near-land lightning formulation (GEOS–4 only)To fix an underprediction of the 4x5 global lightning totals

These were detected in the recent 1-year benchmark simulations


GEOS–Chem User’s Meeting: April 2007The 3rd GEOS-Chem User's Meeting will be held at Harvard on April 11-13, 2007 (Wed-Fri, ending at noon Friday). Please mark your calendars.

This will be an opportunity for all of us to meet and hear about several areas of model development and research, including:

Big-ticket GEOS–Chem development items:capability to run with MPI

standardization of the GEOS–Chem model adjoint

implementation of Earth System Modeling Framework (ESMF) software structure

nested and high-resolution capabilities for GEOS–4 and GEOS–5

interface with GEOS–5 met fields

Several major scientific updates and research applications by GEOS–Chem users

Hotel and registration informationWebsite: http://www.as.harvard.edu/chemistry/trop/geos/geos_meeting_2007.html

Contact Brenda Mathieu to register: [email protected] to cover travel to the G-C mtg are available! Talk to Daniel for more info.

NOTE: Boston Marathon is on the following Monday! Reserve your hotel early!!!


http://www.as.harvard.edu/chemistry/trop/geos/geos_meeting_2007.html

mailto:[email protected]

GEOS–5 met fields


Met field introductionGEOS–Chem does not generate its own meteorology

It reads “assimilated” data fields from disk

Assimilated data is “real world” data that is fed thru a GCMThe GCM is not allowed to run free, but is nudged back to “the truth” on a regular basis

Various observations are ingested into the assimilation system. GEOS–4 uses:Conventional Final Dump - NCEP

Wentz SSM/I

Operational Sea Ice

Interactive TOVS retrievals

QuikSCAT (satellite)

SBUV Ozone

Reynolds SST - Centered Average

MODIS NESDIS Winds

In basic terms, the assimilation system can be thought of as “interpolating” real-world onto a regular grid. Every grid point will have a value, even if there is no “real-world” data at that location.


Met field introductionNASA Global Modeling and Assimilation Office (GMAO) provides assimilated data for the satellite and modeling communities

GMAO is about to release their 5th-generation assimilated data product (not surprisingly called GEOS–5)

GEOS = “Goddard Earth Observing System”

Other GMAO data products have included:GEOS–1

GEOS–STRAT

GEOS–2 (used internally at NASA – never released!)

GEOS–3

GEOS–4


GMAO Met Field Comparison Chart

Quick Look at GMAO met fields

GEOS–1 GEOS–STRAT

GEOS–3 GEOS–4 GEOS–5

# of levels 20 sigma 26 sigma 48 sigma 55 hybrid 72 hybrid

Native grid 2 x 2.5 2 x 2.5 1 x 1 1 x 1.25 0.5 x 0.666

Native File Format

Binary Binary HDF4–EOS HDF4–EOS HDF4—EOS(HDF5– EOS after April 2007)

Regridded to 2 x 2.54 x 5

2 x 2.54 x 5

1 x 1 nested2 x 2.54 x 5

2 x 2.54 x 5

2 x 2.54 x 5

Temporal Coverage

1985 – 1995 1996 – 1998 2000 – 2002 1985 –present (ending in Feb 2007)

2004 –onward + 30 yr reanalysis

Past Present Fut.


Pure Sigma Grid: Used prior to GEOS–4

σ = [ P(IJL) – PTOP ] / [ Psurface(IJ) – PTOP ]

Where I=lon, J=lat, L=alt

Pros: Sigma coordinates are terrain-following. While the surface pressure varies at each (IJ) location, the sigma coordinate for each (IJ) is constant for a given level L.

Cons: The signature of the mountains are evident even very high up in the atmosphere. This leads to very noisy winds and poor STE.


Hybrid Grid: Used in GEOS–4, GEOS–5 Pressure at grid edges:

Pe(IJL) = A(L) + [ B(L) * Psurf(IJ) ]

Pressure at grid centers:Pc(IJL) = [ Pe(IJL) + Pe(IJL+1) ] / 2

Near the surface, the hybrid grid has a “sigma-like” terrain-following coordinate.

Near the model top, the hybrid grid has fixed-pressure levels. The pressure at each level is the same for all latitudes & longitudes.

Pros: Results in much smoother winds near the top of the model, better STE.


Comparison of vertical grids Vertical levels up to 4 km – PBL

GEOS–3 GEOS–4 GEOS–5

GEOS–5 has much finer resolution in the PBL!


Comparison of vertical grids

GEOS–3 GEOS–4GEOS–5

Vertical levels – up to 30 km


GMAO met fields come in 3 flavorsInstantaneous 6-hr fields (what we call “I6”)

Pressures, land types, etc.

3-hr time-averaged fields (what we call “A3”)Surface parameters

6-hr time-averaged fields (what we call “A6”)3-D parameters (winds, humidities, temperature, mass fluxes, etc.)


GEOS–5 data processingI wrote new data processing software to extract & regrid the GEOS–5 data

Uses HDF–EOS reading code (written by me)

Uses S-J Lin’s MAP_A2A regridding code This is area preserving mapping (e.g. quantities per unit area are conserved)

MAP_A2A was also used for GEOS–3 and GEOS–4 regridding

With the regridding code, I have created 1 day of sample data files for GEOS–Chemat both 2 x 2.5 and 4 x 5 grids

However, the code is flexible enough to also save out 0.5 x 0.666, 1 x 1, 1 x 1.25 grids too

I will now show “quick-look” comparisons between GEOS–5 and GEOS–4 dataWe will look at the 2 x 2.5 data.

Date and time: 2004/09/20 at 0000 GMT (A6, I6 fields) or 0130 GMT (A3 fields)

CAVEAT: This is not a scientifically rigorous comparison.We will not have a good idea of how GEOS–Chem performs with GEOS–5 meteorology until we obtain 1 year of GEOS–5 met data and run a 1-yr benchmark

However, glaring problems should stand out in the plots, so they are a good first-check of the data.


PS (surface pressure) – I6 field

GEOS-5 – GEOS-4 abs diff GEOS-5 – GEOS-4 % diff


LWI (land/water flags) – I6 field


NOTE: Both GEOS-4 and GEOS-5 call Antarctica as “land”. Technically this is true as there is land below the ice.

In GEOS-Chem we use LWI plus the surface albedo to correctly diagnose places where ice/snow exist on the surface.


SLP (sea-level pressure) – I6 field



TO3 and TTO3 (Ozone cols) – I6 fields

TTO3: Total trop ozone column

NOTE: the sample data did not have realistic values for TTO3, ignore this

TO3: Total ozone column


ALBEDO (visible albedo) – A3 field


NOTE: The sample data did not have the regular ALBEDO field so I constructed it from the direct & diffuse albedo fields just for comparison purposes.

This may account for the difference.


CLDTOT (2d cloud fraction) – A3 field



EVAP (evapotranspiration flux) – A3 field

mm/day kg/m2/s

GEOS–5 EVAP is the total turbulent flux of water vapor at the surface, including fluxes from transpiration, sublimation, and surface condensation.

The turbulent flux of cold condensate (fog) is assumed to be zero.

GEOS–3 EVAP GEOS–5 EVAP

NOTE: EVAP was contained in the GEOS–4 met data but we did not archive it for GEOS–Chem. We did, however, archive EVAP as part of the “Extra” GEOS–3 fields.

EVAP is not currently used in GEOS–Chem but it could be useful for certain simulations (e.g. Hg, CO2).


GWETTOP (topsoil wetness) – A3 field



HFLUX (sensible heat flux) – A3 field



LAI (leaf area index) – A3 field

% m2/m2



LWGNET (long wave @ sfc) – A3 field


NOTE: LWGNET in GEOS–5 seems to be a “net”radiation whereas RADLWG in GEOS–4 was a total incident radiation.


SWGNET (shortwave @ sfc) – A3 field


NOTE: SWGNET in GEOS–5 is a net radiation wheras RADSWG in GEOS–4 was a total incident radiation.


PARDF (diffuse PAR) – A3 field



PARDR (direct PAR) – A3 field



PBLH (PBL height) – A3 field



Precip fields at ground – A3 fieldsGEOS-4 PREACC GEOS-5 PRECTOT

GEOS-4 PRECON GEOS-5 PRECCONmm/day

mm/day

kg/m2/s

kg/m2/s

NOTE Different Units!!!

GEOS-4 is in mm/day

GEOS-5 is in kg/m2/s


SNOMAS (snow depth, H2O equiv) – A3 field



SNODP (geom. packing of snow) – A3 field



T2M (T @ 2m, proxy for surf. air temp) – A3 field



TSKIN (skin (aka ground) temp) – A3 field



USTAR (friction velocity) – A3 field



U10M (U-wind at 10m) – A3 field



V10M (V-wind @ 10m) – A3 field



Z0M (roughness height) – A3 field



U and V winds – A6 fieldsGEOS–4 U-wind GEOS–5 U-wind

GEOS–4 V-wind GEOS–5 V-windZonal Means


Specific Humidity and Temperature – A6 fieldsGEOS–4 specific humidity (Q) GEOS–5 specific humidity (QV)

NOTE: GEOS-5 QV field is in kg/kg so the scale is 1000X smaller than the GEOS-4 value, which was in g/kg.GEOS–4 Temperature GEOS–5 TemperatureZonal Means


Optical depth: sum of ice + water paths (A6)GEOS–4 optical depth GEOS–5 ice optical depth (TAUCLI)

Ice optical depth scale is 1/10th of the other plots!!!

GEOS–5 water optical depth (TAUCLW) GEOS–5 optical depth (sum of TAUCLI + TAUCLW)

No cloud optical depth higher than ~5km

TAUCLI and TAUCLW are summed on the high-res grid and then the sum is regridded to 2 x 2.5.

Cloud OD > 5km is restored

Zonal Means


Optical depth: sum of ice + water paths (A6)GEOS–4 column optical depth GEOS–5 column TAUCLI

GEOS–5 column TAUCLW GEOS–5 column OPTDEPTH(sum of TAUCLI + TAUCLW)


Cloud fraction and optical depth – A6 fieldsGEOS–4 3D cloud fraction GEOS–5 3D cloud fraction

GEOS–4 OPTDEPTH GEOS–5 OPTDEPTH

Cloud fraction doesn’t translate to cloud optical depth at altitudes > 5

km in GEOS–4

Zonal Means


CLDMAS and DTRAIN – A6 fields GEOS–3 CLDMAS GEOS–5 CMFMC

GEOS–3 DTRAIN GEOS–5 DTRAINZonal Means


MOISTQ: sum of liquid, ice, vapor tendencies

Units have been converted to g/kg/day for comparison to the GEOS-4 fields.

GEOS–5 DQIDTMST (ice tendency) GEOS–5 DQLDTMST (liquid tendency)

GEOS–5 DQVDTMST (vapor tendency) GEOS–5 equivalent MOISTQ= sum of all three tendencies

Zonal Means


MOISTQ: GEOS–5 vs. GEOS–4 GEOS–4 MOISTQ GEOS–5 MOISTQ

Comparison of MOISTQ fields

MOISTQ is provided in GEOS–4

MOISTQ has been computed from DQIDTMST, DQVDTMST, DQLDTMST in GEOS–5

Zonal Means


DQRCON, DQRLSC – A6 fieldsGEOS–5 DQRCON GEOS–5 DQRLSC

DQRCON = layer production rate of preciptating condensate from

convective processes.

DQRLSC = layer production rate of precipitating condensate from

large-scale processes

Includes both liquid & frozen precipitating condensate but not

the cloud condensates.

DQRCON, DQRLSC are only in GEOS–5. They were not included in GEOS–4. They could help us improve the GEOS–Chem rainout parameterization.

Zonal Means


GEOS–5 met fields: summaryWe looked GEOS–5 vs. GEOS–4 data at ~ 0 GMT on 2004/09/20

For EVAP, CLDMAS, DTRAIN we had to compare to 2002/09/20 in GEOS–3

Some GEOS–5 fields have significant differences than in GEOS–4 Root cause: different cloud & convection package in GEOS–5 ?

Different land model or landtype database in GEOS–5?

GEOS–5 optical depth looks better than GEOS–4 The cloud optical depth at 5 km altitude and higher is present in GEOS–5, where this was “mysteriously” absent in GEOS–4.

Changes in GMAO met fields from one version to another often require modifications in GEOS–Chem on our part

There is always something a little bit different

Fields are in different files data extraction codes have to be rewritten

Units are different, must be converted, etc.


GEOS–5 scheduleReprocessing of “historical” GEOS–5 data:

May 1 2004 – March 31, 2005 (done by Dec 2006)

March 1, 2005 – March 31, 2006 (done by January 2007)

Feb 15, 2006 – forward (will catch up to real-time in late December, 2006)

There will be 1.5 months of real-time overlap between GEOS–4 and GEOS–5. GEOS–4 will end in mid-February, 2007.

In late April, 2007, GEOS–5 met fields will change from HDF4–EOS to HDF5–EOS file format

HDF5–EOS is incompatible with HDF4–EOS. Chief advantage is that HDF5–EOS allows you to store more than 2GB per file.

File format switch will coincide with “MERRA”, which is the long-term reanalysis with GEOS–5 (1970’s – present)

This will require further modification to our data processing code (swap out HDF4–EOS modules and insert HDF5–EOS modules.

This should be a minor software change, as we do already have HDF5–EOS libraries and software installed here at Harvard.


Extra Slides


GEOS–1: 1st generation met dataGEOS–1 parameters

Assimilation GCM used: NASA/GSFC GEOS–1 GCM

Temporal coverage: 1985 – 1995

Native horizontal grid: 2o lat x 2.5o lon

Native vertical grid: 20 pure-sigma layers, model top at 10 hPa

Native file format: NASA PHOENIX format (binary Fortran unformatted)

We regridded to: 4o lat x 5o lon, 20 pure-sigma layers

Pros: •Long temporal coverage (10 years)

Cons: •Very low model top (10 hPa)•Strat-trop exchange was too fast by 2-3X


GEOS–STRAT: continuation of GEOS–1 GEOS–STRAT parameters


Temporal coverage: 1996 – 1997 (some 1998)


Native vertical grid: 46 pure-sigma layers, model top at 0.1 hPa

Native file format: NASA PHOENIX format (binary Fortran unformatted)

We regridded to: 2o lat x 2.5o lon, 26 pure-sigma layers4o lat x 5o lon, 26 pure-sigma layers

Pros: •Higher model top, better stratospheric representation

Cons: •Vertical grid did not correspond to GEOS–1•STE still too fast by 2-3X•Surface data had a 6-hr temporal resolution, we had to interpolate to a 3-hr temporal resolution ourselves


GEOS–2: A research dataset GEOS–2 parameters


Temporal coverage: Data was never released outside of NASA

Native horizontal grid 2o lat x 2.5o lon

Native vertical grid 70 pure-sigma layers, model top at 0.01 hPa

Native file format NASA PHOENIX format (binary Fortran unformatted)

We regridded to: N/A

Pros: N/A

Cons: N/A


GEOS–3: 1st high-resolution met data GEOS–3 parameters


Temporal coverage: 2000 – 2002 (with some non-std data for 1998)

Native horizontal grid: 1o lat x 1o lon

Native vertical grid: 48 sigma layers, model top at 0.01 hPa

Native file format: HDF4–EOS gridded files(Output data was on pressure levels, not sigma levels)

We regridded to: 1o x 1o nested grids; 2o x 2.5o; 4o x 5o w/ 48 sigma layers

Pros: •High resolution made 1o x 1o nested sims possible•Higher model top

Cons: •P sigma vertical interpolation had to be done•STE still not realistic•U, V winds were very noisy at the top of the atmosphere


GEOS–4: High-resolution, hybrid grid GEOS–4 parameters

Assimilation GCM used: NCAR fvCCM GCM

Temporal coverage: 1985 – present


Native vertical grid: 55 hybrid P-sigma layers, model top at 0.01 hPa

Native file format: HDF4–EOS gridded files

We regridded to: 2o x 2.5o; 4o x 5o with 55 hybrid layers

Pros: •Long temporal horizon – 20 years of consistent met data•U and V winds are 6-hr averages, not instantaneous •Much better STE than any prior GEOS met fields

Cons: •Convection variables totally different than GEOS–3•GMAO was unhappy w/ the convection results•Cloud OD’s are too low J-values, chemistry are “too hot”


GEOS–5: Ultra-high resolution! GEOS–5 parameters

Assimilation GCM used: NCAR fvCCM GCM + NASA convection modules

Temporal coverage: 2007 – onward (past years will also be re-analyzed)

Native horizontal grid: 0.5o lat x 0.666o lon

Native vertical grid: 72 hybrid P-sigma layers, model top at 0.01 hPa

Native file format: HDF4–EOS gridded files (with SZIP data compression)will switch to HDF5–EOS after April 2007

We will regrid to: 2o x 2.5o; 4o x 5o with 72 hybrid P-sigma layersAlso higher resolution if required

Pros: •High resolution will facilitate comparison to satellites•Convection, optical depth probably better than in GEOS–4

Cons: •1 day of “raw” data files = 3 GB! •Regridded data will also require a lot of storage


Documents

The GEOS–5 met fields: What’s new and what’s different ...acmg.seas.harvard.edu/presentations/2006/bmy_group_mtg_dec_6_2006.pdfÆInstallation of libraries (HDF, HDF–EOS, ESMF,