Some recent developments in the ECMWF model

1
Some recent developments in the ECMWF model

• Mariano Hortal ECMWF
• Thanks to A. Beljars (physics),
• E. Holm (humidity analysis)

2
Noise in forecasts
H12 from 26-09-2002 at 00z
H12 Z10 from 28-12-2002 at 12z
3
Linear least square fit interpolation
x
4
SETTLS with LLSI at both departure and arrival in
the vertical trajectory computation
H12 from 26-09-2002 at 00z
H12 Z10 from 28-12-2002 at 12z
5
Recent developments in the ECMWF physics

• Radiation (aerosol climatology)
• Convection and clouds
• Clouds and boundary layer
• Land surface
• Simplified physics for linear and adjoint
  applications
• Orography (MAP reanalysis, turbulent orographic
  form drag)

6
Radiation

• 26R3
• Radiation on a separate grid to save costs
  (instead of 1 out of 4 points). In T511 model
  radiation is done on T255 grid.
• New aerosol climatology
• Post-processing of PAR and UV-B
• Under development
• RRTM short wave
• McICA Monte Carlo Independent Column
  Approximation to represent cloud overlap and
  inhomogeneous clouds by using different samples
  of the clouds in the different computational
  intervals (140 g-points in 16 spectral intervals)

7
Convection and clouds

• 26R3
• Clean-up of code and improved numerics leading to
  better representation of ice fallout
• New cloud base/top algorithm based on entraining
  plume
• Convection from any layer in lowest 300 hPa
• Revised initiation of convection with perturbed
  parcels (in T and q) starting from mixed layer
  properties
• Reduced water load in updrafts through more
  efficient microphysics
• Increased entrainment

8
Convection and clouds
new
old
9
Convection and clouds
10
Clouds and boundary layer

• A statistical cloud scheme based on variance of
  total water is under development
• Moist boundary layer mixing scheme is nearly
  finished (better stratocumulus)

11
Land surface

• Fully implicit tile coupling with less noisy
  results for the tiles with small fraction

• Tiles
• Water
• Ice
• Wet skin
• Low vegetation
• Exposed snow
• High vegetation
• Snow under vegetation
• Bare soil

12
Land surface

• An Extended Kalman Filter (EKF) has been
  developed for soil moisture analysis (as part of
  the EU project ELDAS).
• EKF can assimilate SYNOP-T/RH, Meteosat heating
  rates, and microwave brightness temperatures

Single column simulation for MUREX (France), 1.
Control with no data assimilation, 2. EKF with
microwave Tb 3. EKF with SYNOP T/RH, 4. EKF
with surface heating rates
13
Physics in relation to data assimilation

• Linear and adjoint of radiation code has been
  developed and is currently under test
• Simplified cloud and convection schemes have been
  developed for linear and adjoint applications
• Experiments are under way to evaluate
  assimilation of microwave rain products and
  brightness T in rainy areas via 1DVAR of TCWV
  which is assimilated in 4DVAR
• TRMM precipitation radar is used for verification

14
Physics in relation to data assimilation
TRMM-PR
first guess
assim. of TMI Tb
assim. of TMI-rain rate
15
Orography MAP reanalysis
TCWV from GPS 21-10-1999

• Reanalysis with all the additional MAP data is
  available

TCWV from MAP reanalysis, T511
TCWV from operations 1999, T319
16
New scheme for turbulent orographic form drag

• Alternative to effective roughness length concept
• Drag is distributed in vertical and implemented
  on model levels (Brown and Wood, 2003)
• Scales between 5 km and 10 m are represented by
  this scheme
• Universal orographic spectrum is assumed to
  account for scales smaller than 5 km
• Standard deviation of orography at scales between
  about 10 to 2 km is used to drive the scheme
  (from 1 km data base)

Comparison of orographic drag and turbulent
surface drag (from vegetation) from new scheme
with fine scale model results. Expressed as drag
coefficient versus terrain slope.
17
Nonlinearities in the humidity analysis

• Humidity is bounded from below (gt0) and
  restricted close to saturation by condensation.
• Analysis increments behave asymmetrically at
  different levels of relative humidity.
• A new humidity analysis accounts for this through
  nonlinear flow-dependent change of variable,

18
Some humidity analysis results

• With a better background error description,
  better use is made of humidity observations.
• An example is given by HIRS-12 humidity
  sensitive radiances.
• The new humidity analysis (bottom) has removed
  unrealistic outliers in the background error
  description.
• This results in better humidity forecasts.