1 / 37

CFSRR under the hood

CFSRR under the hood. This is put together with help from many people in the Environmental Modeling Center together with many groups that work closely with EMC on model and data assimilation development efforts. GFS AM.

selah
Download Presentation

CFSRR under the hood

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CFSRR under the hood This is put together with help from many people in the Environmental Modeling Center together with many groups that work closely with EMC on model and data assimilation development efforts

  2. GFS AM • Latest version of Global Forecast System (GFS) Atmospheric Model (AM) is being considered for CFSRR. • GFS AM - developed by the staff of Global Climate and Weather Modeling Branch of EMC. • The first reanalysis (NCEP/NCAR – R1) was based on the operational GFS AM of January 1995. • GFS AM has undergone major revisions since the first reanalysis.

  3. GFS AM improvements through reanalysis • Some specific problems found in NCEP/NCAR reanalysis, addressed in later AM changes • -- valley snow • -- wrong snow cover • -- wrong ocean albedo • -- SH paobs mislocated • -- ”pathological” problems in stratosphere • New reanalysis will find problems in GFS that will be addressed and produce improved GFS, improved future reanalysis and improved future CFS • We’ll keep doing it until we get it right (Glenn White)

  4. Comparison between AMs in R1, CFS (opr) and GFS (opr)

  5. Operational CFS GFDL-LW Radiationvs. RRTM-LW Radiation GFDLRRTM Description: - 15 bands 16 bands - trans table look-up 140 cor-k terms - O3,H2O,CO2 O3,H2O,CO2,O2,CH4 CO, 4 CFCs Advantages/ - comp efficient more comp efficient Disadvantages: - no aerosols effect aerosol effect capable - fixed CO2 only varying CO2 capable - fixed sfc emis varying emis capable - random cld ovlp random or max-ran - larger errors, especially improved accuracy at upper stratosphere, at upper stratosphere - simple cloud optical prop advanced cloud optical property property

  6. The current operational GFS AM has Realistic moisture prediction with better depiction of no-rain areas Prognostic Ozone Prognostic cloud condensate Cloud cover only where cloud condensate > 0 Momentum mixing in deep convection Fast and accurate AER RRTM for IR radiation Mountain blocking parameterization Noah land model Sea-ice model Improved treatment of snow, ice, orography Better hurricane track prediction ESMF based modern computer algorithms

  7. Options in GFS AM being considered for next operational model • Enthalpy (CpT) as a prognostic variable in place of Tv • AER RRTM shortwave radiation with maximum-random cloud overlap • IR and Solar radiation called every hour (Until now IR is called every 3 hours) • Use of historical and spatially varying CO2 and volcanic aerosols • Eighth order horizontal diffusion • Modified background vertical diffusion for stratus

  8. NCEP Operational SW Radiationvs. New RRTM SW Radiation NCEPRRTM Description: - 8 uv+vis, 1-nir 5 uv+vis, 9-nir bnds - 38 k-dis terms 112 cor-k terms - O3,H2O,CO2,O2 O3,H2O,CO2,O2,CH4 Advantages: - Comp. Efficient Accu. (use ARM’s data) clr-sky - 10-30 w/m2 reduction cld-sky - adv. scheme Disadvantages: - large errors Comp. slow, 4 times clear-sky - und est slower than opr sw cloudy-sky - over est YuTai Hou of EMC implemented RRTM in the GFS AM

  9. GSI History • The GSI system was initially developed as the next generation global analysis system • Wan-Shu Wu, R. James Purser, David Parrish • Three-Dimensional Variational Analysis with spatially Inhomogeneous Covariances. Mon. Wea. Rev., 130, 2905-2916. • Originally based on SSI analysis system • Replace spectral definition for background errors with grid point representation • Allows for anisotropic, non-homogenous structures • Allows for situation dependent variation in errors

  10. Basic Analysis Problem • 3DVAR minimization of objective function J = Jb + Jo + Jc  Jb = background  Jo = observations  Jc = constraints • where, • xa= analysis state (control vector) which are you solving for • xb= background state • B = background error covariance • yo= observations • R= forward model operator • O = “observation” error  includes instrument and representativeness errors

  11. Analysis (control) vector • xa = stream function, velocity potential, surface pressure, virtual temperature, pseudo-relative humidity, ozone mixing ratio, and cloud condensate mixing ratio • SSI uses • vorticity & divergence for the wind field • specific humidity for moisture • Ozone and cloud condensate are analyzed univariately • Moisture analysis may be univariate or multivariate

  12. Constraint terms, Jc • Moisture constraint • Parameters  and  are tuned such that the analysis maintains negative and super-saturated moisture penalties comparable to guess • SSI formulates constraint in terms of vapor pressure • Dynamic constraint • SSI has divergence tendency constraint – compensate for weaknesses in assumed linear balance described through background error • GSI has several dynamical constraints. Global_GSI runs with “strong” constraint

  13. Background error estimation • “NMC” method • Use 48 & 24 hour forecasts verifying at same time as a proxy for estimating the background error • Originally developed for SSI because • background error represented in spectral space • not clear at that time (1992) how to derive B-1 from innovation statistics as done with OI • Use NMC method to estimate statistical balance between , T, , and ps.

  14. Background error estimation • For SSI, • complete spectrum of correlation estimated, along with latitudinally dependent variances • physical space correlations are isotropic, homogeneous • For GSI, • only correlation length and variance estimated, but both can be functions of position • physical scale correlations may be anisotropic, non-homogeneous • CURRENTLY use isotropic correlations

  15. Assimilated data types • All data types currently assimilated by SSI may also be assimilated by GSI • Sondes, ship reports, surface stations, aircraft data, profilers, etc • Cloud drift and water vapor winds • TOVS, ATOVS, AQUA, and GOES sounder brightness temperatures • SBUV ozone profiles and total ozone • SSM/I and QuikScat surface winds • SSM/I and TMI rain rates

  16. Assimilated data types • GSI additionally has capability to assimilate • COSMIC GPS radio occultation local refracitivity or bending angle • In-situ SST observations • Doppler radar radial velocities • Brightness temperature from SSM/I, AMSRE, SSU, etc.

  17. Tb Assimilation • Community Radiative Transfer Model (CRTM) used in GSI • What CRTM does? • Compute satellite radiances (Forward model) • Compute radiance responses to the perturbations of the state variables (Tangent-linear model) • Compute Adjoint (Adjoint model) • Compute Jacobians(K-matrix model)

  18. Forward ModelRadiances • Convert analysis variables to T, q, Ps, u, v, ozone • Interpolate T profiles, q profiles, ozone profiles, u1,v1, Ps and other surface quantities to observation location • Reduce u1 and v1 to 10m values • Calculate estimate of radiance using radiative transfer model (and surface emissivity model) • Tangent linear of calculation – inner iteration • Currently simulation does not include clouds • Apply bias correction • Compare observation to estimate

  19. Radiative Transfer Model • CRTM is a fast radiative transfer function (and tangent linear, adjoint and Jacobian) (LBL codes much too slow) • Reflected and emitted radiation from surface (emissivity, temperature, polarization, etc.) • Atmospheric transmittances dependent on moisture,temperature, ozone, clouds, aerosols, CO2, methane, ... • Cosmic background radiation (important for microwave) • View geometry (local zenith angle, view angle (polarization)) • Instrument characteristics (spectral response functions, etc.) • Scattering from clouds, precipitation and aerosols

  20. What’s needed to assimilate radiance? • CRTM model, CRTM coeff files • satinfo which contains the usage flag, obs error and Instrument/observation characteristics (microwave/IR, etc.) • Radiance obs bias correction is needed.

  21. Scan bias (slowly evolve) Satellite Bias Correction in GSI Observation bias Air mass bias (dependent on atmospheric state) Constant Zenith angle Cloud liquid water Square of T lapse rate T lapse rate etc. is the coefficients of predictors

  22. 2 BC files are required to run gsi • Coefficients for predictor part of BC (biascr.gdas.*) • Updated within inner loop of analysis • Slowly evolving angle dependent part of BC (satang.gdas.*) • Updated in separate job step following analysis using Tb innovation file • Method: Start from guess values and run a training period. Experiment was conducted to see if this procedure can lead to convergence to the operational values (‘truth’) after the training period.

  23. Coupling of GFS to MOM3 (MOM4) In the operational CFS, AM and OM are coupled daily with AM and OM running sequentially In the new CFS, the coupling is MPI-level (developed by Dmitry Shenin) – AM, OM and the coupler run simultaneously Coupling frequency is flexible up to the OM time step Same AM code can run in coupled or standalone mode Coupler details for MOM4 will be presented later in this meeting

  24. CFS Reanalysis and Reforecast Scripts AM and OM Post post.sh Start here Copy IC files copy.sh 9 (or 48) hr Coupled Model Forecast (first guess) New GFS + MOM4 with Sea Ice MPI-level Coupling fcst.sh Verify vrfy.sh CFSRR website Prep step Hurricane relocation Data preparation prep.sh GODAS Global Ocean Data Assimilation oanl.sh Archive data arch.sh Retrospective Forecast? Time 00Z ? GDAS Global Atmospheric Data Assimilation GSI anal.sh GLDAS Global Land Data Assimi- lation lanl.sh Run Retrospective Forecast fcst.sh

  25. Sea-Ice in new CFS: thermodynamics • Winton (2000) 3-layer thermodynamic model plus ice thickness distribution • 2-layer of sea-ice and 1-layer of snow • Fully implicit time-stepping scheme, allowing longer time steps • 5 categories of sea-ice Winton, M. 2000. A reformulated three-layer sea ice model. J. Atmos. Ocean. Tech., 17(4), 525-531

  26. Sea-Ice in new CFS: dynamics • Hunke and Dukowicz (1997) elastic-viscous-plastic (EVP) ice dynamics model • Improved numerical method for Hibler’s viscous-plastic (VP) model • Computionally more efficient than Hibler’s VP model, suitable for fully coupled models Hunke, E. C. and J. K. Dukowicz, 1997. An elastic-viscous-plastic model for sea ice dynamics. J. Phys. Oceanogr., 27, 1849-1867

  27. Coupler Time Step Δc Ocean Time Step Δo GFS (LAND) Time Step Δa Sea-Ice Time Step Δi Atmosphere grid Sea-ice is one component of the new CFS Fast loop:Δa= Δc= Δi Slow loop: Δo Tsfc Sea-Ice Fluxes X-grid

  28. What’s MOM4 • The Modular Ocean Model (MOM4) is a numerical representation of the ocean’s hydrostatic primitive equations developed by GFDL. It is designed primarily as a tool for studying the ocean climate system.

  29. Release & Version of MOM4 • Jan, 2004: MOM4p0a • Mar, 2004: MOM4p0b • Aug, 2004: MOM4p0c • Dec, 2004: MOM4cDec10 • Jun, 2005: MOM4d • Final version: MOM4p0 this is what we are using now

  30. Main Feature • Coded in Fortran90, units are KMS • Re-engineered Parallel infrastructure (MPI) • All I/O is handled via NetCDF, will support IEEE format output in the future. • Only contains one cpp-preprocessor option (ifdefs), (MOM3 has 300-400 ifdefs). • Upside: clean up the logic in the separate modules, code more readable, short compiling time. • Downside: increase code maintenance. System structure becomes complicated (~80 directories). • 2D horizontal domain decomposition. No memory windows or slabs. Increase computation efficiency.

  31. Numerical/dynamic characteristics • Generalized orthogonal horizontal coordinator. Support both standard spherical coordinator and “tripolar” grid (no polar filter required, full Arctic Ocean). • Z-coordinator vertically. Only support B-grid currently. • Partial cell for bottom topography • Non-Boussinesq explicit free surface (Greatbatch et al. 2002). • explicit fresh water flux • McDougall et al. (2003) equation of state with in situ density a function of local potential temperature, salinity and hydrostatic pressure. • Various kind of tracer advection schemes. • KPP vertical mixing (Large et al. 1994) • Chlorophyll based shortwave penetration (Morel and Antoine, 1994) • runoff process

  32. GODAS for the CFSRR • Prototype GODAS to become operational in 2010. • - Based on MOMv4 (1/2o x 1/2o x 40L) and the 3DVAR assimilation • scheme. • - Assimilation data are temperature profiles (XBT, Argo, TAO, TRITON, • PIRATA), synthetic salinity profiles derived from a seasonal T-S relation, • Reynolds’ daily OI SST, Levitas seasonal SSS, TOPEX/Jason-1 Altimetry. • Data window extends from 10-days before to analysis time. • Data window for SST is 1-day. • - The analysis system is quasi-coupled to the CFS in the sense that the first • guess for the assimilation is provided by the CFS. After each analysis cycle • the ocean model is stepped forward as a fully coupled component of the • CFS. • - No external atmospheric forcing fields are needed. No relaxation of surface • temperature or salinity is used.

  33. Comparability of MOM4 and MOM3 versions of GODAS • Differences in how physical parameterizations are implemented. • Assimilation code has been rewritten using FORTRAN 90 structures and MOM4 coding conventions. Comparison of 2 long runs with same resolution (1ox1o), using the same forcing (Reanalysis 2), and the same assimilation data set (T(z), S(z)).

  34. Summary • The new version of GODAS is based on MOM4, is global and has increased resolution (1/2ox1/2o). • When run in a similar configuration, the new GODAS is comparable to the current operational GODAS. • The new GODAS incorporates SST and SSS into the assimilation in place of surface relaxation. • The new GODAS assimilation extends to 2200m, taking better advantage of the Argo data and limiting mid-depth temperature drift. • As part of the CFSRR, the GODAS will not be forced by an external analysis (e.g. R2), but instead will be integrated into the CFS.

  35. Noah LSM 4 soil layers (10, 30, 60, 100 cm) Frozen soil physics included Two snowpack states (SWE, density) Surface fluxes weighted by snow cover fraction Improved seasonal cycle of vegetation cover Spatially varying root depth Runoff and infiltration account for sub-grid variability in precipitation & soil moisture Improved thermal conductivity in soil and snowpack column Higher canopy resistance Improved evaporation treatment over bare soil and snowpack OSU LSM 2 soil layers (10, 190 cm) No frozen soil physics Only one snowpack state (SWE) Surface fluxes not weighted by snow fraction Vegetation fraction never less than 50 percent Spatially constant root depth Runoff & infiltration do not account for subgrid variability of precipitation & soil moisture Poor soil and snow thermal conductivity, especially for thin snowpack GFS and CFS: Land Model UpgradeNoah LSM (new) versus OSU LSM (old): Noah LSM replaced OSU LSM in operational NCEP medium-range Global Forecast System (GFS) in late May 2005

  36. GLDAS versus Global Reanalysis 2 (GR2):Land Treatment • GLDAS: an uncoupled land simulation system driven by observed precipitation analyses (CPC CMAP analyses) • Executed using same grid, land mask, terrain field and Noah LSM as GFS in experimental CFS • Non-precipitation land forcing is from CFS analysis cycle • Executed along side the CFS Reanalysis • GR2: a coupled atmosphere/land assimilation system wherein land component is driven by model predicted precipitation • applies the OSU LSM • nudges soil moisture based on differences between model and CPC CMAP precipitation

More Related