Post on 09-Jun-2020
Thu 3/17/2016Briefly discuss exam resultsRemaining hypothesis presentations – TuesdayModel capabilities:
• Resolution terminology• Domains and nesting• Moving nests• Digital Filter Initialization (DFI)
Begin cloud/precipitation microphysics (MP) section
Reminders/announcements:• Take-home portion of exam due
Semester OutlineModel Physics:
1.) Land-Surface Models (LSM)2.) Turbulence parameterization & the planetary boundary layer (PBL)3.) Convective parameterization (CP)4.) Cloud and precipitation microphysics (MP)5.) Parameterization of radiation
Project:1.) Topic selection, case identification2.) Hypothesis development3.) Control simulation, hypothesis presentation4.) Experiments and final presentation
Technical:1.) Running SCM2.) Running WPS, WRF, postprocessing for real-data cases3.) Model experiments: Terrain and physics modifications4.) Analysis and diagnosis of model output
DoneDoingNot yet
On the Term “Resolution”
Grasso (2000), and many others:• The terms “resolution” and “grid
spacing” are not equivalent
• Generally filter 2x and 3x waves, so at very least, “resolution” is 4x
• Most estimates are between 5 and 10 x (e.g., Walters comment)
• Safe terminology: Refer to “grid length” or “grid spacing” rather than “resolution”
WRF Resolution (Skamarock and Klemp 2008)
• Model initial conditions typically have little kinetic energy in mesoscale
• Requires 6-12 hours for development of mesoscale energy spectrum
• Skamarock and Klemp (2008): WRF “resolves” features ~6 to 8 x (with 5th order advection scheme)
On WRF Resolution (Skamarock and Klemp)
Some important considerations:1) Resolution and grid spacing. What phenomena must be resolved,
and what grid length will accomplish that?2) Domain size and “residence time”. What is the prevailing wind speed,
and how long must air reside in a given domain to benefit from the resolution?
3) One versus two-way nesting (not yet discussed: “feedback” in NL)4) Sources of initial, boundary condition data
Worksheet
Analysis and Prediction of High-Impact Weather Events
Circulations in eyewall
TC flow can be divided into… Primary: azimuthal
Maintains inertial stability of the warm-core
Secondary: radial and vertical Supplies thermodynamic energy in
the form of turbulent fluxes & releases latent heat aloft
Ooyama ’82: TC a “mesoscale power plant with a synoptic-scale support system”
Megan Gentry (now Mallard)
Analysis and Prediction of High-Impact Weather Events
Eye diameter ~ 37 km
4x partially resolved x 9 km
10x fully resolved x 4 km
Scale of a hurricane
• What grid spacing is necessary to “resolve” a hurricane vortex?
• Even coarse grids can resolve basic aspects of vortex
RMW ~ 50 km, diameter ~ 100 km
4x partially resolved x 25 km
10x fully resolved x 10 km
Walters 2000 & comments by Grasso, The Differentiation between Grid Spacing and Resolution and Their Application to Numerical Modeling
Analysis and Prediction of High-Impact Weather Events
Convection in eyewall• Observations: Updrafts and downdrafts in eyewall
are much smaller in scale than vortex itself
Core Diameter (km)
90% of updrafts ~ 4 km or less
4x partially resolved, x = 1 km
10x fully resolved, x = 0.4 km
Eastin 2005
Analysis and Prediction of High-Impact Weather Events
Eyewall processes• Vortex Rossby waves
(VRWs) break in eyewall, forming pools of high PV
• Warm, low-momentum air in eye mixes out into eyewall, additional heat source for convection (Persing and Montgomery 2003)
Schubert et al. 1999
Analysis and Prediction of High-Impact Weather Events
Experimental setup• Hurricane Ivan (2004) simulated at 27, 9, 8, 6, 4, 3, 2, 1 km
• See Gentry and Lackmann (2010), MWR for details
Analysis and Prediction of High-Impact Weather Events
Vortex-tracking nest
Model simulated
radar
Aside about moving nests…
Nested domains on the fly…
“Easy way” to spawn additional nests, but
terrain resolution won’t match grid length
Real-time WRF forecast of Hurricane Earl, initialized 00 UTC 2 September 2010
Real-time WRF forecast of Hurricane Earl, initialized 00 UTC 2 Sept 2010 (with vortex-tracking moving nest feature)
Analysis and Prediction of High-Impact Weather Events
• Repeat: Ivan (2004) simulated using 27, 9, 8, 6, 4, 3, 2, 1 km grid length
• First, let’s visit “no man’s land”, examine 9-km output
• Purpose here: Examine with, without CP scheme on
• Note, with 1-way feedback, nested grids can be examined independently
• 2 runs
• KK -> Kain-Fritsch CP scheme on all domains
• K0 -> no CP in inner domain, KF outer
• Similar nomenclature for BMJ scheme
Analysis and Prediction of High-Impact Weather Events
Azimuthal averages (9 km)• Good way to look at averaged structure of TC in profile
R
R
ZY
X
Analysis and Prediction of High-Impact Weather Events
Updraft velocity (9 km)• Makes significant difference in
updraft velocity
• ~ 0.25 m/s difference with updraft of 1 – 1.5 m/s
• Think about vortex stretching…
KK
K0
Analysis and Prediction of High-Impact Weather Events
Downdraft velocity (9 km)• Downdraft velocities are more
intense with explicit convection
KK
K0
Analysis and Prediction of High-Impact Weather Events
No CP vs. CP• Weakening of vertical motions with CP… expected since CP
handles convection implicitly
• Overall, secondary circulation, spiral bands strongly influenced by whether or not CP used
• Influence of CP scheme is felt by balanced vortex… no clear scale separation between vortex and convective scale on which CPS is working…. what Molinari paper warns about
Analysis and Prediction of High-Impact Weather Events
Implications of using CP• Use of CP weakens vertical velocities in/around eyewall,
affects intensification
• Without CP: Stronger updrafts, greater compensating subsidence in core… more compressional warming, hydrostatic lowering of surface pressure
• From vorticity perspective, weaker vortex stretching with CP
• Intensity differences (between CP vs. no CP with all else equal): ~ 5 to 10 hPa in central pressure
Analysis and Prediction of High-Impact Weather Events
Outline• Hurricane simulations with & without the use of a
convective parameterization scheme (CPS)
• Sensitivity of hurricane structure & intensity to horizontal resolution
Analysis and Prediction of High-Impact Weather Events
• What changes in TC structure and intensity occur when horizontal grid spacing is reduced?
• What physical processes might be responsible for these changes?
30 hPa of deepening
between 8 & 1 km grid spacing
Analysis and Prediction of High-Impact Weather Events
Model-simulated
• Smaller vortex, RMW shrinks by 40%
• Development of spiral bands
• Convection becomes more axisymmetric, with more & smaller maxima in eyewall
8km 4km
2km 1km
Analysis and Prediction of High-Impact Weather Events
Vertical motion
8km 6km
4km 2km
Analysis and Prediction of High-Impact Weather Events
Eyewall convection• “Dumbell” behavior
(seen in coarser) simulations vanishes at small grid length
• Eyewall has interspersed up/downdrafts
850-hPa vertical velocity (w)
Analysis and Prediction of High-Impact Weather Events
Eyewall Updraft• As resolution increases, eyewall shrinks and vertical velocity
increases, even when interpolated to same grid
8 6 4
3 2 1
Analysis and Prediction of High-Impact Weather Events
VRWs and PV mixing• Straight-line PV segments and PV pools appear in 4-
km and finer simulations
Analysis and Prediction of High-Impact Weather Events
As grid length decreases…– Up/downdrafts in eyewall stronger, more numerous, and
smaller in spatial extent
– Eyewall less elliptical, more likely to transition from circular to polygonal eyewall appearance
– Spiral bands develop
– VRWs phase-lock & break, eye/eyewall exchange of entropy & momentum
• Overall, much better representation of eyewallprocesses & intensity with 4-km and finer grid spacing
DFI
• Another important WRF capability (that should probably be used for most simulations): Digital Filter Initialization
• Following is extracted from WRF workshop talks by Peckham et al. and Huang et al. (2008)
• DFI is a way to filter noise and ensure dynamical consistency at model start-up: “Initialization”
• It does not involve interpolation, bogusing, nudging, or data assimilation
DFI
• Puts momentum and mass fields into dynamical balance
• Spins up cloud, precipitation fields at model start (unlike usual clear-sky start)
• Easy to use, a namelist option
• Drawback: Adds some computational expense
DFI
DFI
DFI
DFI
DFI
DFI
DFI
DFI
I often reduce forward and backstop time to 1 hour or even 30 mins –still derive benefit
DDFI
I often reduce forward and backstop time to 1 hour or even 30 mins –still derive benefit
Microphysics Outline
• Basics of microphysics schemes– Why classes matter– Representation of number concentration
• The WRF schemes
• Model simulated radar
• Case-study analysis:– Winter storm– Convective storm– Tropical cyclone– Other…
Microphysics References
Microphysics Outline
• Basics of microphysics schemes– MP scheme responsibilities– Distinguishing characteristics: Classes, Distribution, & Processes– Why classes matter– Representation of number concentration
• The WRF schemes
• Model simulated radar
• Case-study analysis:– Winter storm– Convective storm– Tropical cyclone– Other…
Microphysics – cloud interactions
Water Vapor
evap
orat
ion
OceanLANDev
apor
atio
n
evap
otra
nspi
ratio
n
RunoffPrecipitation
Condensation, Deposition
CLOUDS
Long-Wave radiation
Solar shortwave
Adapted from F. Carr
evaporation
shallow Cu,downdrafts from
deep Cu
Grid-Scale Precipitation• Also known as “explicit” precipitation… why might this be a
misnomer?
– Still “parameterization” of very small-scale processes (cloud physics)! – But, the *motions* needed to supersaturate are represented on grid
• Precipitation produced in model as result of grid-scale ascent, removal of supersaturation
• Scheme that handles grid-scale precipitation dubbed “microphysics” or “cloud microphysics” scheme
Microphysics Background
• As model grid spacing decreases, it becomes easier to saturate a grid box
The role of microphysics schemes in NWP becomes increasingly important at higher resolution
Sources:Jason Millbrandt, McGill, excellent microphysics materialsHong et al. 2004Jimy Dudhia (NCAR)WRF technical description, WRF web pageMilbrandt and Yau papersStensrud parameterization text
Microphysics
Ascent, saturation, and condensation
Grid-scale precipitation: – Requires grid box RH to
exceed a specified threshold
– Excessive moisture eliminated through condensation or deposition
What must MP schemes do?• Account for change in vapor, condensate (fill 3-D arrays for
advection), latent heat release/absorption – compute mixing ratio
• Account for falling precipitation (different fall speeds for different size hydrometeors) – must know size distribution, number concentration
• Determine all phase change processes, compute changes in hydrometeor class (e.g., autoconversion, melting, riming, accretion, aggregation, etc.)
• Provide cloud information to radiation scheme
• Determine amount, type of precipitation reaching surface (interact with Land-Surface Model)
• Interact with CP scheme
What are MP distinguishing characteristics?
1.) Classes: # of different water species (cloud, precipitation, vapor):
Cloud liquid waterCloud iceSnowRainGraupelHailVapor…. Could also have different degrees of rimed snow crystal, different crystal habit, wet versus dry hail, melting (aggregate snow), etc.
2.) Specification of particle size distribution:
BinBulkSingle, double, triple moment
What are MP distinguishing characteristics?
3.) Processes
Specification of classes and size distribution dictates which processes may be represented in a given scheme
Braham and Squires (1974, BAMS)
Ice Processes: Habit as f(T) (Caltech photos)
http://www.caltech.edu/content/snowflake-science
When is it no longer snow?
http://www.inscc.utah.edu/~tgarrett/Snowflakes/WASHARX.html
Graupel vs Hail
http://www.komonews.com/weather/blogs/scott/112648409.html & Witchita NWS
Many WRF schemes include only one or other (mainly graupel)
High-end schemes include both
Worksheet example
Katrina, WSM6 run, Hour 18, Simulated radar, SLP
Hour 18, Cross section, cloud liquid water, cloud ice
Hour 18, cross section: rain water
What is approximate fall velocity of rain?
What about snow?
What is typical vertical velocity in eyewall of a mature hurricane?
Why does this matter?
Hour 60, Cross section, vertical velocity (m/s)
Not a bad updraft for 20 km grid with CP!
Hour 18, cross section: rain water, snowWhy is the snow so far above freezing level?
Hour 18, cross section: rain water, snow, graupel