Advancd Regional Prediction System (ARPS)

1
Advancd Regional Prediction System (ARPS)
Ming Xue mxue_at_ou.eduSchool of Meteorology
and Center for Analysis and Prediction of
StormsUniversity of Oklahoma
2
Model Dynamics, Equations and Numerical
FormualtionsSee PDF file
3
Initial Condition
4
ARPS Components
5
Ways to Initialize ARPS

• Idealized, single sounding
• Interpolation from GFS, Eta, RUC, etc
• ADAS

6
ARPS Data Analysis System (ADAS)

• Manages the real time ingest, QC, objective
  analysis of observations
• Doppler radar data (NIDS, base Level II from n
  systems, VAD)
• MDCRS commercial aircraft wind and temperature
  reports
• Wind profilers
• RAOBS (conventional, CLASS, dropsondes)
• Mobile and fixed mesonets
• SAO and METAR observations
• GOES satellite visible and IR data for cloud
  analysis
• NCEP gridded model output
• Based on Bratseth successive correction method
• Handles retrieved radar data (from SDVR et al)
• Had its root in FSL LAPS. Data format is about
  the only one left though.

7
Braseth Analysis Scheme

• ADAS use the Bratseth analysis scheme which is a
  successive correction scheme
• The scheme theoretically converges to optimal
  interpolation (O/I), but without explicit
  inversion of large matrices
• Multi-pass strategy used where more detailed data
  can be introduced after a few iterations using
  broad-scale data.
• Like OI, the Bratseth method accounts for the
  relative error between the background and each
  observation source, and is relatively insensitive
  to large variations in data density.
• Vertical correction in terms of z or q

8
Formulation of Bratseth Scheme
9
Formulation (Continued )
10
ARPS Data Analysis System (ADAS)

• User specifies background error covariances and
  structure functions. Codes to calculate
  background error statistics being developed.
• Performed on ARPS native (terrain-following) grid
• 3-D cloud analysis and diabatic initialization
  package using GOES, Doppler radar and surface
  data.
• Water vapor, cloud, rain, ice and temperature
  fields are affected by the cloud analysis
• Used to initialize realtime high-res (kms)
  forecasts at CAPS since 1996
• Linked closely with ARPS data assimilation system
  (via, e.g., intermittent assimilation,
  incremental analysis update method)

11
Incremental Analysis Update Cycles
(from Brewster 2003)
12
(No Transcript)
13
ADAS analysis Total u
ADAS
Background
73x73x43 grid, dx12km
14
ADAS Analysis Total q
Background
ADAS
15
Example of Initial Condition with cloud analysis
on a 3km Grid
16
Application to fine-scale analysis at Kennedy
Space Center (Case et al 2002 Wea. Forecasting)
17
(No Transcript)
18
(No Transcript)
19
Boundary Conditions

• Lateral Boundary Conditions
• Rigid, zero-gradient, periodic
• Open/radiative LBC (only applied to normal
  velocity)
• Externally (can be from the same model) forced
• Davies-type relaxation zone, arbitrary width
• w not forced
• variables (e.g., water) not found in exbc are
  excluded from relaxation zero gradient is
  usually applied
• Ensure same terrain at nesting boundaries
• Carpenter (1982) radiation BC with external
  forcing(?)
• Carpenter, K. M., 1982 Note on radiation
  conditions for the lateral boundaries of
  limited-area numerical models. Quart. J. Roy.
  Meteor. Soc., 108, 717-719.

20
Vertical Boundary Condition

• Radiation top BC based on cosine Fourier
  transform (Klemp and Durran 1983)
• periodicity requirement at the top relaxed
• Still based on linearized equations difficult
  to apply to large domain
• Upper boundary sponge/absorbing layer
• relaxation to coarse grid/external model solution
  in the layer
• or relaxation to the mean state
• Rigid, zero-gradient and periodic top-bottom BC
• Semi-slip lower BC

21
Stratiform Clouds and Precipitation

• Microphysics parameterization for grid-scale
  precipitation
• Can be used together with cumulus
  parameterization schemes
• Option to allow condensation at subsatuation
  (lt100 RH)
• helps retaining clouds in IC for large grid
  spacing
• improves surface temperature forecast at
  low-resolution by introducing clouds earlier
• Sedimentation term treated implicitly or using
  time splitting

22
Model Physics

• SGS Turbulence
• Smagorinsky-Lilly, 1.5-order TKE, Germano dynamic
  closure
• Fully three dimensional formulation, including
  map factor
• Simplified 1-D option available for efficiency
  purpose
• Cumulus Parameterization
• Kuo scheme
• Old and New versions of Kain-Fritsch cumulus
  parameterizations
• Eta Betts-Miller-Jancic scheme
• Microphysics
• Kessler warm rain
• Lin-Tao ice microphysics
• Schultz NEM grid-scale microphysics

23
Model Physics (continued)

• PBL scheme
• Convective PBL mixing parameterization based on
  1.5-order TKE formulation (Xue et al 1996)
• Surface Physics (fluxes)
• Stability-dependent bulk aerodynamic drag for
  surface heat, momentum, and moisture fluxes
• Soil Model
• 2-layer soil model (multiple soil types in 1 grid
  cell API initialization)
• A new multi-layer soil model
• High-resolution surface characteristics data base
  (consistency among surface fields important)
• Radiation
• Full long- and short-wave radiation (NASA code)
  including cloud interactions, cloud shadowing,
  and terrain gradient effects

24
ARPS Physical Processes
25
Subgrid Scale Fluxes(Land surface, surface
layer, PBL and SGS turbulence)

• See PDF file

26
Radiation
27
Radiation Parameterization

• ARPS radiation package came from NASA GSFC.
• Shortwave is based on Chou (1990 1992)
• Longwave based on Chou and Suarez (1994)
• Cloud-radiation interaction described in Tao et
  al (1996)
• Allows checkboard-type staggered calculations to
  save computation
• Recently implemented terrain shading effect
• Terrain slope accounted for
• Cloud fraction diagnosed from RH and qs
• Verifications against OK Mesonet radiation
  measurements show good agreement, in clear sky
  conditions at least

28
Shortwave Radiation

• Solar spectrum is divided into the ultraviolet
  and visible region (lt0.69mm), and the
  near-infrared (IR) region (gt0.69mm),
• UV and visible region includes ozone absorption,
  Rayleigh and cloud scattering. It is further
  divided into 4 bands, with effective ozone
  absorption and scattering coefficients given to
  each band
• The IR region includes absorption due to water
  vapor, cloud, CO2, O3 and scattering due to
  clouds.
• Further division into 7 water vapor absorption
  bands, with k-distribution method used to
  calculate the absorption.
• Liou et al (1988) 4-stream discrete ordinate
  scattering algorithm used for multiscattering in
  cloud layer
• Single scattering albedo from King et al (1990)

29
Longwave Radiation

• IR spectrum divided into 8 bands
• Water vapor transmission function computed used
  k-distribution method
• CO2 and O3 transmission functions computed using
  lookup tables
• Includes aerosol effects
• Absorption due to cloud hydrometeors also
  included. Clouds assumed to be gray and
  nonscattering
• Cloud optical properties
• Scheme 1 Broadband emissivity method of Stevens
  (1978, 1984)
• Scheme 2 follows Fu and Liou (1993), Sui et al
  (1996)

30
Radiation Fluxes Verification
31
Microphysics Schemes

• Kessler warm rain microphysics (qc and qr)
• Lin et al (1983) ice microphysics
• includes rain, cloud water, cloud ice, snow,
  graupel/hail,
• lookup tables for power and exponential functions
• ice-water saturation adjustment procedure of Tao
  et al (1989)
• modifications to hydrometeo fall speeds (Ferrier
  1994 and updated coefficients)
• Shultz (1995) simplified ice scheme (also include
  3 ice categories)

32
ARPS Ice Microphysics Processes 30 processes
33
Accumulated Precipitation from 1977 Del City
Supercell Storms with warmrain and ice
microphysics
34
Simulation of 1977 Del City Supercell Storms with
warmrain and ice microphysics
35
Convective Parameterization
36
Convective Clouds and Precipitation

• At high resolutions (lt 3km), use explicit
  microphysics, hopefully the model can resolve the
  convection well
• Cumulus parameterization schemes
• Kuo scheme
• Old and new Kain-Fritch schemes
• Betts-Miller-Janjic scheme
• New K-F scheme used most

37
Use of Cumulus Scheme

• K-F scheme used most
• New K-F scheme using at 27 and 9km during IHOP
  realtime forecast
• BMJ scheme tends to produce much smoother
  precipitation field
• Cold pool important for propagation of convective
  systems over the plains
• Triggering of spurious propagating precipitation
  pattern observed during IHOP

38
CAPS Real Time Forecast Domain during IHOP_2002
183163
273195
213131
39
June 15, 2002, 9km Grid
NCEP Hourly Precip
9 km Forecast Precip Hourly Rate.
24 hour forecast
40
June 15, 2002 3km grid
3 km Forecast Hourly Precip Rate
NCEP Hourly Precip Analysis
11 hour forecast
41
00-12UTC, June 13, 2002, Hourly Precip