CONVECTIVE PARAMETERIZATION

1
CONVECTIVE PARAMETERIZATION

• For the Lesson
• Precipitation Processes
• December 1998

2
What is Convective Parameterization?

• Cumulus or convective parameterization schemes
  are procedures that attempt to account for the
  collective influence of small-scale convective
  processes on large-scale model variables
• All NWP models with grid spacing larger than that
  of individual thunderstorms or storm clusters
  need to parameterize the effect that convection
  has on larger-scale model variables in each grid
  box

3
Why is ConvectiveParameterization Important?

• Convective storms can significantly influence
  vertical stability and large-scale flow patterns
  by
• Redistributing heat, moisture, and momentum
• Producing cloud cover that affects surface
  temperatures
• Are there other reasons?

4
Discussion Questions

• What are some real-life weather scenarios that
  would be seriously impacted if no attempt were
  made to account for convective processes within
  NWP models?
• How might the model fields differ if they werent
  accounted for?

5
Formulation of Convective Parameterizations

• No matter how they are formulated, all convective
  parameterization schemes must answer these key
  questions
• How does the large-scale weather pattern control
  the initiation, location, and intensity of
  convection?
• How does convection modify the environment?
• What are the properties of parameterized clouds?

6
Convection Initiationand Intensity

• Schemes can initiate convection by considering
  the
• Presence of some convective instability at a grid
  point (perturbed parcels may reach LFC)
• Existence of low-level and/or vertically-integrate
  d mass/moisture convergence that exceeds some
  threshold at a grid point
• Rate of destabilization by the environment at a
  grid point

7
Convection Initiationand Intensity Continued

• Schemes can make the intensity of the convection
• Proportional to the moisture or mass convergence
  or flux
• Sufficient to offset the large-scale
  destabilization rate
• Sufficient to eliminate the CAPE (this is
  constrained by the available moisture)

8
ConvectiveFeedback

• In the real atmosphere, convection modifies the
  large-scale thermodynamics via
• Detrainment (creates large-scale evaporative
  cooling and moistening)
• Subsidence in the ambient environment (creates
  large-scale warming and drying)

9
ConvectiveFeedback Continued

• When a model changes the vertical temperature and
  moisture profiles as a result of convective
  processes, it is referred to as convective
  feedback
• The issue for convective parameterization schemes
  used in any given model is how they determine the
  new vertical distribution of heating, cooling,
  moistening and/or drying caused once convection
  is triggered

10
Two Approaches to Convective Feedback

• Adjustment Schemes
• Either nudge the vertical profile toward an
  empirical reference profile
• Make the profile a function of the difference
  between the moist adiabat inside the cloud and
  the moist adiabat representative of the ambient
  environment
• Mass Flux Schemes
• DO attempt to explicitly model convective
  feedback processes at each grid point

11
Properties ofParameterized Clouds

• If the model includes clouds, it determines their
  properties by using
• The moist adiabat from cloud base (older
  approach)
• OR
• A one-dimensional cloud model (of varying
  complexity in different models)

12
How Does This HelpUse NWP?

• Knowing which approach to the questions an NWP
  model has taken as a result of its convective
  parameterization scheme helps you to
• Understand some of the inherent strengths and
  weaknesses of the resulting convective
  precipitation forecasts
• Realize that the same scheme used in two
  different models will likely produce different
  results due to the way the scheme interacts with
  the other components of each individual model

13
Comparing Schemes
14
NWP Models and Schemes
15
Discussion Questions

• Combining information from the two tables, which
  of these operational models includes convective
  downdraft processes?
• When running or accessing a local mesoscale
  model, does it account for convective downdrafts?
• What implication does no knowledge of outflow
  boundaries have on NWP convective initiation
  forecasts?
• Even if the model produces outflow boundaries,
  why might they have little impact on subsequent
  convection initiation in the model?

16
The Aviation Medium-Range Forecast Model (AVN/MRF)

• AVN/MRF uses the Grell-Pan (GP) convective
  parameterization scheme
• Convection initiation within a column considers
• Time rate of change in stability as primary
  convective trigger
• Presence of positive buoyancy (must have some)
• Cap strength

17
The AVN/MRF Model Continued

• Properties of the scheme
• Modifies column buoyancy toward equilibrium (done
  as function of vertical motion at cloud base)
• Evaporation efficiency a function of wind shear
  strength (over ocean only)
• No direct mixing between cloudy air and
  environmental air
• All cloud water converted to rain and leaves the
  cloud
• Shallow clouds (lt 250 hPa) have no downdrafts and
  detrain moisture more easily than deeper clouds

18
The Eta Model (32-km)

• Uses the Betts-Miller-Janjic (BMJ) convective
  parameterization scheme
• BMJ scheme simultaneously nudges temperature and
  moisture profiles at a grid point toward a
  reference profile (acts to adjust model
  atmosphere to a post-convective environment)
• Post-convective profile adjustment happens ONLY
  if precipitation occurs (otherwise BMJ scheme
  does nothing or simulates limited vertical
  mixing)
• BMJ scheme wont trigger convection if cloud
  layer too dry (regardless of amount of CAPE)

19
The Shallow Convection BMJ Scheme in the Eta

• Shallow portion of BMJ scheme triggers if
• Cloud depth (resulting from lifting the most
  unstable parcel)
• gt 10 hPa deep
• lt 200 hPa deep
• Covers at least two model layers

20
The Shallow Convection BMJ Scheme in Eta Cont.

• Shallow schemes role - to prepare pre-convective
  environment via vertical mixing by transporting
  moisture upward
• Mimics process of condensation near cloud base
  (warming and drying) and evaporation near cloud
  top (cooling and moistening) so net change in
  sounding from shallow convection results in no
  precipitation

21
The Deep Convection BMJ Scheme in the Eta Model

• BMJ deep convective parameterization scheme
  identifies the most unstable parcel in the lowest
  130 hPa at each grid point

22
The Deep Convection BMJ Scheme in the Eta Model
Cont.

• Then calculates cloud depth. If gt 200 hPa deep,
  scheme modifies
• Temperature profile to be 90 of slope of moist
  adiabat through cloud base
• Moisture profile using a procedure that considers
  the distance a parcel needs to be lifted to reach
  saturation and the "cloud efficiency" (CE) factor
  (a measure of the convective columns ability to
  transport enthalpy upward, while at the same time
  producing as little precipitation as possible)

23
The Deep Convection BMJ Scheme in the Eta Model
Cont.

• BMJ scheme ensures if rain is produced, net
  latent heat release is in balance with the net
  moisture change due to condensation
• Intensity of convection produced by BMJ scheme
  very moisture dependent
• More moist the column, more intense the convection

24
An Example of the BMJ Scheme in the Eta

• It is often easy to recognize where the BMJ deep
  conv. param. scheme has been active by the
  well-defined reference profile. Example shows Eta
  model soundings before and after scheme has been
  active.

25
Discussion Question

• What impact will the recent change in the shallow
  cloud depth threshold (from 290 hPa to 200 hPa)
  that is responsible for triggering the deep
  convection scheme have on Eta model precipitation
  forecasts?

26
Eta Model with Kain-Fritsch (KF) Parameterization

• An experimental Eta model output using KF
  parameterization is available on the Web

27
The Eta Model with KF Parameterization Continued

• KF scheme is a mass flux scheme similar to Grell
  and GP schemes running in the RUC-2 and AVN/MRF
  notable differences
• CIN evaluated by amount of negative area, not
  just pressure depth of the cap
• Large-scale destabilization not required to
  trigger convection, only CAPE
• Updraft and downdraft formulations more
  sophisticated
• Intensity of convection based on instantaneous
  CAPE, rather than time rate of change of CAPE

28
The Nested Grid Model (NGM)

• Uses a modified Kuo scheme
• Convection triggered at a grid point when
• Moisture convergence in the lowest six layers
  reaches a certain threshold
• A parcel in the lowest four layers can achieve
  buoyancy if lifted
• Total moisture convergence in the column below
  cloud base is positive

29
The NGMContinued

• Modifies by adjustment process, including
  converting 80 of moisture in the column to
  precipitation (with a corresponding latent heat
  release)
• Precipitation allowed to fall and evaporate, but
  lower layers only need to reach 48 RH before
  precipitation falls to next layer

30
The Navy Operational Global Atmospheric
Prediction System (NOGAPS 4)

• Uses Relaxed Arakawa-Schubert (RAS) convective
  parameterization scheme
• This version of AS scheme relaxes state toward
  equilibrium each time invoked instead of
  requiring end state to be balanced
• Other noteworthy difference in RAS from AS
  relates to handling of detrainment
• Precipitation assumed to fall to ground without
  re-evaporation into lower layers

31
European Centre for Medium-Range Weather
Forecasts (ECMWF) Global Model

• Uses mass flux convective parameterization as
  part of prognostic cloud scheme initially
  developed by Tiedtke
• Tiedtke approach uses one-dimensional model to
  predict population of cloud types allowing for
• Shallow convection
• Deep convection (including anvil cirrus)
• Elevated convection

32
The ECMWF Global Model Continued

• Cumulus scale downdrafts included in Tiedke
  scheme
• One strength of ECMWF global model -
  parameterization of precipitation processes
  handled same way for convective clouds as for all
  other clouds, including those resulting from
  large-scale ascent

33
The Rapid Update Cycle RUC-2

• Uses a version of Grell convective
  parameterization scheme
• Scheme was updated from RUC-1 scheme to fix
• Downdraft detrainment
• Calculation of cloud top
• Minimum cloud depth
• Capping criteria

34
The Rapid Update Cycle RUC-2 Continued

• Produces somewhat larger amounts of precipitation
  and more coherent rainfall patterns in convective
  areas than RUC-1
• Because Grell scheme includes downdrafts, RUC-2s
  convective precipitation patterns may appear more
  detailed than those in the Eta model (which uses
  BMJ, no downdrafts)

35
MesoscaleModels

• So it is important to know which scheme choice
  is operational in any mesoscale model accessed
  (as well as other physical parameterizations,
  PBL, etc.)
• Does it modify by adjustment or mass flux?
• How well do the various schemes interact?

36
MesoscaleModels Continued

• Mesoscale models (including 32-km Eta) can
  predict precipitation resulting from storms
  associated with some topographically induced
  boundaries, slantwise convection, etc.
• In mesoscale models, distinction between
  "grid-scale precipitation" and "convective
  precipitation" begins to disappear
• In storm-scale models (lt 2 km), all precipitation
  can be calculated "explicitly no convective
  parameterization is necessary (although
  microphysical processes are still parameterized)

37
MM5 Performance with Various Schemes

• 1997 study compared Anthes-Kuo (AK), Grell, BM,
  and KF schemes in 6 heavy precipitation events
  (both warm cold season) some key results
  regarding precipitation forecast skill
• Skill higher for cold season events than for warm
  season
• Skill better for rainfall volume than areal
  coverage or peak amount
• 12-km grid superior to 36-km (especially for
  heavy precipitation amounts)

38
MM5 Performance with Various Schemes Cont.

• KF and Grell predicted total precipitation volume
  and storm life-cycles well, but over-predicted
  light precipitation
• BM did good job of predicting areal extent of
  light precipitation and maximum rain rates, but
  tended to over predict areas of moderate to heavy
  rainfall in warm season
• AK had the most difficulty predicting warm season
  events

39
MM5 Performance with Various Schemes Cont.

• All 4 schemes had difficulty predicting high
  based convection
• Overall, KF consistently performed best of those
  evaluated
• Partition of rainfall into subgrid scale (that
  precipitation produced by the convective
  parameterization CP scheme) and grid-scale
  precipitation was more sensitive to the
  particular CP scheme chosen than to model grid
  size or convective environment

40
Discussion Questions

• Why might the forecast skill of MM5 precipitation
  be better in the cold season than warm?
• What is probably the main reason that KF and
  Grell were better at predicting storm
  life-cycles?
• What might contribute to all of the schemes
  having difficulty with high-based convection?

41
The Future

• In the next 1-2 years, the operational NCEP suite
  will begin to assimilate precipitation data into
  the models initial conditions. They will use
  radar, rain gauge, and satellite data to
  initialize model clouds and precipitation during
  the data assimilation stage.
• This is expected to improve the initial
  specifications of humidity, vertical motion, and
  instability in the models and should lead to
  better numerical forecasts of convection and its
  associated precipitation