presentation about regcm-roms coupled model in regcm workshop 2012

23
Toward fully coupled RegCM-ROMS Atmosphere-Ocean Model: Preliminary Results Ufuk Utku Turuncoglu ICTP (International Center for Theoretical Physics) Earth System Physics Section

Upload: bartimuf7097

Post on 04-Sep-2015

12 views

Category:

Documents


0 download

DESCRIPTION

Explains the main design of the RegCM-ROMS coupled modelling system

TRANSCRIPT

  • Toward fully coupled RegCM-ROMS Atmosphere-Ocean Model: Preliminary Results Ufuk Utku Turuncoglu ICTP (International Center for Theoretical Physics) Earth System Physics Section

  • Outline Motivation Surface Flux Parameterizations in RegCM RegCM-ROMS Coupled Model Design Preliminary Results

    Caspian Sea Additional Notes

    Simple Benchmarking Pros & Cons

    Future Work

    - Outline

  • - Motivation (1/2)

    Earth System Includes several sub-components (ATM, OCN, BIO, ICE, LAND etc.) Each of them has its own complexity Interaction between these components are also very complex and

    include non-linear processes Each sub-component have different temporal and spatial scales Each of these components with different and hierarchical

    complexities can be merged as an Earth System Models (ESMs) Facts Models are getting more complex with the evolution of the computing

    resources and old fashioned coupling strategies are not valid anymore Designing a monolithic ESM model to represent all these sub-

    components is not possible There is no any standardized interface for the models (different

    horizontal and vertical grids, variable units, configuration files etc.) The coupling libraries (i.e. ESMF, OASIS, MCT) and new

    technologies (XML, RDF etc.) can help to reduce the complexity of the models itself and coupling processes between them

  • - Motivation (2/2)

    Atmosphere-Ocean Interaction? Investigating role of water bodies (ocean, sea and inland waters) over

    the regional climate or investigating effects of climate change over water bodies

    Study interaction and feedback mechanisms between atmosphere and water bodies

    The impact of marine aerosol (sea-salt, sea spray etc.) on climate system and atmospheric chemistry

    Validation of the simplified flux algorithms (i.e. Zeng Ocean) created special for water bodies in the regional climate model or designing more realistic and lightweight new sub-grid parameterization methods

    Motivation Investigating role of CAS in regional climate by representing

    atmosphere-ocean interaction more realistically More accurate estimation of water budget of the CAS to reconstruct

    the historical sea level fluctuations. Representation of evaporation flux over the sea is very important in this aspect.

  • - Surface Flux Parameterizations for Water Bodies (1/2)

    RegCM 4.1 Options to represent fluxes over ocean and lake BATS It uses Monin-Obukhov similarity relations.

    It has some limitations such as no special treatment of convective and very stable conditions and constant roughness length

    Zeng Ocean (Zeng et al., 1998) Unlike BATS, it describes all stability conditions. It calculates sensible and latent heat and momentum fluxes in the air-sea interface.

    It allows better representation of air-sea interaction It uses prescribed sea surface temperature (i.e. OISST) as input.

    One-dimensional Lake Model (Hostetler et al., 1993) Heat, momentum and moisture fluxes are calculated based on lower level atmospheric conditions. Heat is transferred vertically between lake model layers by eddy and convective mixing.

    It includes simple ice sub-model No horizontal exchange across the grid points It may require tuning to get realistic results.

  • - Surface Flux Parameterizations for Water Bodies (2/2)

    RegCM 4.3 In addition of the available parameterizations the new version of

    RegCM enables to use three-dimensional regional ocean model (ROMS) in coupled mode (RegCM+ROMS).

    The coupled model may provide better representation of oceans, lakes and inland waters (If the ROMS model is tuned well).

    Supports It can be used together with Zeng Ocean and one-dimensional Lake

    Model. ESMF online re-gridding capability and conservative type interpolation

    is used for flux variables (bilinear type interpolation for others) Works with ROMS 3.5 tagged version and also ROMS-Ice branch. Different number of processor can be assigned to each gridded

    components (coupler component uses all processors) Limitations Coupled code only interacts with BATS not CLM Current version only supports Gregorian calendar The ocean model still needs forcing files but RegCM updates them

  • - Design (1/2)

    General structure (1)

    Ocean / Lake

    Psurf, T2m, Q2m, Rain Swrad, Lwrad, U10m, V10m SST Hice

    ATM Surface atmospheric conditions, heat fluxes and rain. The atmospheric variables is used to calculate fluxes in the ROMS.

    OCN SST and Ice Thickness (if the ice model exist). The SST is used to calculate sensible heat flux, evaporation, ground temperature etc. Ice Thickness is used to change land type and updates the evaporation flux over ice

    The interaction or coupling time step (in second) can be defined using RegCM namelist file

  • - Design (2/2)

    General structure (2)

    Gridded Components (RegCM and ROMS) + Coupler Component

  • - Results (1/X)

    Caspian Sea

    Importance of Caspian Sea It locates in a very sensitive climate region (extreme weather cond.) It is the major source of oil and natural gas Large fishing industry (90% of the Worlds caviar) Facts The Caspian Sea Level (CSL) has fluctuated dramatically in the past. The CSL changes also have great importance over costal areas and it

    causes to change in the coastline and flooding of these regions It is possible that future changes in global and regional climates might

    produce large changes in the Caspian Sea characteristics Needs To study hydrodynamic properties (i.e. main circulation and thermal

    structure) of the Caspian Sea To understand feedback mechanism and complex interaction between

    sea and regional climate

  • - Results (2/X)

    Test Configuration

    Tests - 1995-1997 (first year is spin-up) ATM-STD RegCM ATM-LAK RegCM and 1D-Lake Model ATM-CPL RegCM and ROMS (online coupling) DTCPL = 3 hour OCN-STD ROMS (forced by ATM-STD results, offline coupling) RegCM Initial and Boundary Conditions for RegCM: ERA-Interim Resolution: 50 km (18 sigma layers) Parameterizations: Grell, Zeng Ocean, BATS ROMS (with ICE Model) Initial Condition for ROMS: monthly T,S climatology (Ibrayev et al.,

    2001) and started from rest (no motion). Resolution: 10 km (32 sigma layers) Parametrizations: GLS Mixing, Wet-Dry, river runoff

  • - Results (3/X)

    Model Grids River discharge points (climatological daily data)

    ocean model domain

    ETOPO1 is used for ocean model bathymetry and smoothing is applied to prevent PGE. The minimum depth of sea is set to 5 m.

  • - Results (4/X)

    Sea Surface Temperature

    The observational datasets are GRSST, ARCLAKE and CLIM (CTD measurements)

    Coupled model little bit warmer than standalone mode of ROMS (OCN-STD)

    MAM sea surface temperature in NCAS is improved in the coupled simulation.

    The main spatial pattern is preserved in both simulation (ATM.STD & ATM.CPL)

  • OCN-CPL OCN-STD

    - Results (4/X)

    Sea Surface Temperature

    The GHRSST data is used as observation In both cases correlations are around 0.98 for all sub regions (NCAS,

    CCAS, SCAS) The coupled model warmer than standalone model but it improves the

    results especially in the MAM but not in SON

  • - Results (5/X)

    Sea Surface Salinity

    Go to animation

    Spatial pattern is very similar in both

    run (OCN-STD, OCN-CPL). The salinity is underestimated in the

    NCAS when it is compared with available observational dataset

    It is possible that the OBS has problem in the NCAS. Because the salinity values are consistent with other studies and observations.

  • - Results (6/X)

    Ice Coverage The observational

    dataset is ARCLAKE which covers 1996-2009 period but the model results for 1996-1997.

    In DJF, ROMS seems underestimate the ice coverage in the NCAS.

    The positive bias in OCN-STD does not appear in OCN-CPL simulation in MAM. The results are more realistic than the standalone run.

    JJA also looks better in coupled model results.

  • - Results (7/X)

    Evaporation (1)

    The observational dataset is OAFLUX which relatively low resolution 1 deg.

    The evaporation response are quite different in each ATM model case.

    The OCN-STD has larger evaporation bias in SON and DJF than OCN-CPL in other seasons they are very similar

  • - Results (8/X)

    Evaporation (2)

    Monthly evaporation over CAS ATM-LAK (CC: 0.73) and OCN-STD runs overestimates the evaporation ATM-STD (0.92) run generally has negative bias The coupled simulation (ATM-CPL, 0.96) are the closest to the OAFLUX

  • - Results (9/X)

    Surface Temperature

    ATM-STD - ATM.LAK ATM-LAK ATM.CPL

  • - Results (9/X)

    Precipitation

    ATM-STD ATM.LAK ATM-LAK ATM.CPL

  • - Additional Notes (1/2)

    Benchmarking Basic Performance Estimation for Test Case (DTCPL = 3h, DTOCN=300s, DTATM=150s) ATM-STD: 8.5 month/hour (32 CPU) OCN-STD: 4.5 month/hour (28 CPU) ATM-CPL: 3 month/hour (32+28 = 60 CPU) Extra Overhead for Coupling The ocean model %23 slower then standalone mode The atmosphere model %65 slower than standalone mode The coupled model performs interpolation between grids and data

    exchange each 3 hours. The overhead is combination of computing and data transfer.

    Possible Ways to Increase Performance Increasing coupling time step Load balance between modeling components. Increase OCN CPU !!! Extensive benchmarking is needed !!!

  • - Additional Notes (2/2)

    Pros & Cons Pros The version of ocean component (ROMS) can be changed easily by

    porting patch (it may also requires minor change in the RegCM component code). This is tested before is tested before, 3.4 3.5 3.5_ice.

    The coupled model provides better representation of oceans, lakes and inland waters (If the ROMS model is tuned well).

    New components (like CLM) can be added by modifying coupler interface.

    Cons It needs experience also in ocean modeling, which the configuration of

    realistic application is not straightforward (create ocean model grid, initial, boundary and forcing files).

    Adds additional computational and messaging overhead into the model User needs to tune the ocean model like atmospheric model. This

    must be done in standalone mode first !!!

  • - Future Plans

    More work Support for identical mesh/grid.

    No need to interpolation between grid. The possible candidate to test this configuration is the BAND

    version of the RegCM. The tool to create ROMS grid from RegCM domain file is almost

    ready. It is written in NCL. Add support for boundary smoothing. It is important for the ocean or

    sea centric applications (i.e. Mediterranean Sea, Indian Ocean). Add support for selective interpolation types for each variable. User

    can change from conservative type re-gridding to bilinear or others. Improve interface of air-sea interaction. In the current implementation,

    ocean send SST and ICE thickness to atmospheric model. The aerosol related variables, surface roughness etc. can be added into list.

    Benchmarking is needed to see the actual performance of the model. It also helps to find possible performance bottleneck sources of the current implementation.

    Test case for additional domains such as Mediterranean Sea.

  • The coupled model tutorial is on Wednesday afternoon (14:00-18:00).

    Come and learn the details and usage. It is good time to learn basics of the ROMS ocean

    model as well !!! The paper about coupled model and the preliminary

    results for Caspian Sea is on the way.