Tue 4/12/2016Radiation parameterization:

- Overview of WRF radiation schemes, tests and implications

Representation of clouds and precipitation:- Microphysics: Paper summaries (3-4 more: Keith, Xia, Pat, Hans)

Reminders/announcements:• WRF V3.8 was released on 8 April, see:

http://www2.mmm.ucar.edu/wrf/users/wrfv3.8/updates-3.8.html• Updates to KF, Tiedtke CP, Several PBL, Thompson, NSSL MP• MP experiment assignment (due next week)• Final presentations: 28 April, 1-4 pm (final exam period)

• Schedule optional meetings with me if feedback or assistance is needed • Be sure to emphasize the analysis aspect! • Handout/assignment provides additional guidance for content and evaluation• Extra credit option: YouTube presentation of your project!

A new paradigm for improved interpretation of scientific journal papers? Let’s try this (for extra credit)

Outline for radiation parameterization section

Radiative transfer- Review of radiation basics- Atmospheric radiation- Model representation strategies- An example of physics interactions (MP-RA)- Overview of WRF radiation schemes- Cloud-radiation interactions: Thompson/RRTMG

ra_lw_physics=1 (RRTM)Rapid Radiative Transfer Model (RRTM) scheme – from AER Inc. (Mlawer et al.

1997, 2003)

• Spectral scheme (based on line-by-line (LBL) transfer model) – 16 LW bands

• Look-up tables to draw on accurate LBL calculations (absorption as function of pressure and temperature)

• Interacts with explicit clouds

• Ozone/CO2 from climatology in WRF-ARW

• Namelist default (what we’ve been running, unless you changed it)

• Accounts for water vapor, CO2, O3, N2O, CH4, halocarbons (CFCs)

• Validated for wide range of conditions, seasons, locations

• Designed for versatile applications, including climate and mesoscale models; used in ECHAM5 GCM

• Serious bug fixed in V3.2 (Cavallo) – major cold bias in upper stratosphere

ra_lw_physics=3 (CAM)Community Atmosphere Model (CAM) radiation scheme,

see Collins et al. (2004)

• Requires additional namelist variables (see next slide)

• More sophisticated scheme from climate model, useful for long WRF runs (on order weeks or more)

• Use with CAM SW scheme

• Accounts for CO2 variation over time

ra_lw_physics=4 (RRTMG)RRTMG scheme:

Similar to RRTM: Look-up tables, K-distribution method

More sophisticated cloud treatment than RRTM; RRTMG handles cloud fractions whereas RRTM is 1/0

More interactions with WRF-chem, uses optical depth

Well suited for climate applications

Coupled with Thompson microphysics (more on this soon)

ra_lw_physics=5 (Goddard)Updated Goddard scheme, added 2011 (Chou and

Suarez):

Fewer bands than RRTM schemes (10), also uses look-up tables

Handles cloud fractions? Designed to handle aerosols

Also well-suited for climate applications

ra_lw_physics=7 (UCLA)New UCLA scheme (Fu, Liou, and Gu: FLG)

Added 2012 (V3.4), based on Gu et al. 2011, JGR

Another k-distribution scheme

12 bands, look-up table

Cloud fraction 0/1

Designed for aerosol and trace gas interactions

Lots of work on cirrus problem (scattering by ice crystals, with account of shape and size properties)

Would be nice to couple with microphysics

ra_lw_physics=99 (GFDL)GFDL longwave scheme (“semi-supported”)• Used in Eta/NMM; compared by Tarasova et al.

• Should only be used with Ferrier microphysics (?)

• Spectral scheme from global model

• Also uses lookup tables

• Interacts with explicit clouds

• Ozone/CO2 from climatology

ra_sw_physics=1 (Dudhia)MM5 shortwave (Dudhia)• Simple downward calculation, a wide-band method

• Clear-sky scattering, tunable, and can include aerosols

• Water vapor absorption, but not ozone(!)

• Computationally inexpensive, cloud reflection & absorption

• This is namelist default (what we’ve been using, unless you changed it)

• Often produces less SWDN at surface relative to other schemes

ra_sw_physics=2 (Goddard)

Goddard shortwave (Chou and Suarez 1999, as mentioned in Tarasova et al. (2006) study)

• Spectral method from Goddard GCM

• Interacts with grid-scale clouds

• Does include ozone absorption, but uses climatological distributions

• Better account of water vapor absorption, oxygen absorption line in SW, aerosols (as discussed)

• Seems to consistently yield greater SWRAD than Dudhiascheme… consider this for convective initiation?

ra_sw_physics=3 (CAM)

CAM shortwave scheme, corresponds to LW scheme (use together)

• Includes more advanced capabilities, such as chemistry and aerosol interactions

• Perhaps best for long-term simulations, has solar constant variations evidently built in (11-year cycle plus longer variations)

• Uses IPCC A2 scenario for future CO2 as discussed earlier

• Can specify better ozone climatology (o3input = 2) using namelist

ra_sw_physics=4 (RRTMG)RRTMG shortwave scheme, corresponds to LW scheme

(use together)

• Also spectral, with 14 bands

• Handles cloud fractions, better cloud interactions with Thompson MP scheme

• Chemistry interactions, such as aerosol optical depth

• Added climatological aerosol values; soon will allow to utilize analyzed aerosols

• Better suited for climate applications than RRTM

• Lower clear-sky SCM SWDOWN value – aerosol?

• Add aer_opt = 1 for climatological aerosol

ra_sw_physics=5 (Goddard 2)

Updated Goddard SW scheme, corresponds to LW scheme (use together)

• Also spectral, with 11 SW bands

• Handles cloud fractions

• Evidently adds several minor absorption bands, important collectively

• As before, k-distribution method in SW

• Accounts for cloud properties, including effective particle size

ra_sw_physics=7 (UCLA/FLG)

UCLA FLG scheme (Fu, Liou, Gu)

Has aerosol capabilities

6 shortwave bands, k-distribution method

Aerosol capability there, but not implemented yet in WRF-ARW?

Again, specialty cirrus and radiation-ice particle interactions

SCM clear-sky test has highest SWDOWN of all schemes tested (too large)

ra_sw_physics=99 (GFDL/Eta)

GFDL shortwave (“semi-supported”)• Used in Eta/NMM model

• Can only be used with Ferrier microphysics

• Ozone effects

• Interacts with clouds

• Compared with Goddard by Tarasova et al. (2006) as discussed previously

SWDOWN comparison for 6/21/2011 SCM

RRTMG: 1018.45 Wm-2 RRTM: 1058.99 Wm-2

CAM: 1067.32 Wm-2 ETA: 1078.06 Wm-2

SWDOWN comparison, 6/18/2014 SCM V3.5.1

RRTMG: 1022.0 W/m2, with climo aerosol, 990 W/m2

RRTM: 1111.2 W/m2

CAM: 1043.54 W/m2 UCLA: 1141.3 W/m2 peak

Dudhia

CAM

Goddard

• SWDN 100-200 W/m2 lower for much of Dudhia run• Dudhia appears to have too small clear-sky values…

RRTMG

Dudhia

FLG

Goddard

• SWDN 100-200 W/m2 lower for much of Dudhia run• Dudhia appears to have too small clear-sky values…

RRTMG

Difference field, Dudhia minus RRTMG (SWRD)

Difference field, Dudhia minus CAM (SWRD)

Difference field, Dudhia minus Goddard (SWRD)

Difference field, Dudhia minus FLG (SWRD)

Difference field, CAM minus Goddard (SWRD)

Difference field, Dudhia minus Goddard (2-m T)

What are the implications?

Default SW scheme (Dudhia) generally results in lower values of SWRD relative to other schemes

UCLA FLG scheme highest

What are some other fields that could be affected?

How much difference does 50-100 W m-2 make over a period of a few hours?

Summary: RadiationRadiation scheme namelist default is questionable for some

applications (e.g., climate-type or longer runs, high-latitude runs, runs with high model top, stratosphere)

Major advances are starting to happen with aerosol and cloud interactions, analyzed trace gases/aerosols

Some new capabilities require namelist modifications, see README.namelist

Radiation choice may matter for many applications, tends to garner less attention

Worth exploring choices more fully than we will have time to do here

WRF Radiation Options

• These choices impact:– Surface air temperature– Stability– Convective precipitation– PBL depth– Cloud cover– Soil moisture– Etc.

radtRadiation time-step recommendation• Radiation too expensive to call every step

• Frequency should resolve cloud-cover changes with time

• radt = 1 minute per km grid size is about right (e.g. radt=10 for dx=10 km)

• If radt very long, consider swint_opt = 1 (to interpolate zenith angle between calls)

Cloud-radiation interactionsCloud-radiation interactions are critical:

Climate: Cirrus warms, stratus cools

Cloud optical properties and lifetimes are crucial to forecast

How can radiation schemes be designed to optimally represent clouds?

- Grid-scale clouds?

- Subgrid-scale clouds?

See worksheet, will discuss on Thursday

Cloud-radiation interactionsLongwave:- Are sub-grid clouds included? Slingo 1987- Partial cloud cover: Overlap strategy

Shortwave:- Key variable is cloud optical thickness,

- Absorption of SW by cloud water cannot be neglected at some wavelengths

- But, we have cloud/hydrometeor information from MP scheme, right? But it isn’t used

32

cld water patheffectivedrop radius

Cloud-radiation interactionsIf we alter the cloud-droplet number concentration, say from

maritime to continental, will the radiation scheme know if the clouds are different?

No.

Unless… we are using Thompson and RRTMG

Let’s test this to make sure it works the way we think it will

Cloud-radiation interactionsExperiment:

Run 1: RRTM/Thompson with default nt_cRun 2: RRTMG/Thompson with default nt_cRun 3: RRTM/Thompson with 1000 nt_cRun 4: RRTMG/Thompson with 1000 nt_c

What do you expect will happen with these 4 runs?

Run 1 minus Run 3(RRTM 100 – RRTM 1000)

Run 2 minus Run 4(RRTMG 100 – RRTMG 1000)

Cloud-radiation interactionsSummary: nt_c = 100 versus nt_c = 1000 per cm3

The cloud properties were NOT communicated from Thompson to the RRTM radiation scheme:- Changes in radiation were due to changes in cloud water

distribution (coverage)

The cloud properties WERE communicated from Thompson to the RRTMG radiation scheme- Changes in SWDN were due to both cloud cover changes and also

to changes in cloud properties

Simulated Electrification of a Small Thunderstorm with Two-Moment Bulk Microphysics

Edward (Ted) R. Mansell, Conrad L. Ziegler, Eric C. BruningJournal of the Atmospheric Sciences, 2010

• Implemented new microphysics scheme including six hydrometeor classes

• Simulated thunderstorm charge structure and lightning

Key topics:

Microphysics Scheme• Double-moment (mass and number

concentration predicted)• Also predicts bulk concentration

of CCN and densities of graupeland hail

• Six hydrometeor species• Droplets, rain, ice crystals,

snow, graupel, and hail• Can select to include graupel

and hail or only one (in this study, only graupel used)

• Largely based off of Ziegler (1985)

Key Concepts:• Greater diversity in fall speed of ice

hydrometeors (graupel/hail) through use of rime mass density

• Distribution function in terms of particle volume

• Tunable shape parameters for graupel/hail

• Full (3D) gradient of supersaturation• Separate mass- and number-

weighted average terminal fall speeds for all hydrometeor species

Electrification Processes• Parameterized “noninductive charge separation in

rebounding graupel-ice collisions”• Attachment/drift motion of small ions explicit• Based on laboratory results• Lightning explicitly represented

• Includes tunable “propagation threshold factor”

Aerosol Effects on Intensity of LandfallingHurricanes as Seen from Simulations with theWRF Model with Spectral BinMicrophysics

(HUJI BIN scheme – Fast Version)Khain et al. 2009Journal of the Atmospheric Sciences

Motivation:• Aerosols substantially affect cloud microphysics• Microphysics scheme of the current operational TC forecast model is insensitive to aerosols

• Bulk microphysics schemes (especially single moment) have limitations• Fast‐SBM keeps the main advantages of SBM, requires less than 20% of the computer time of the full SBM

Experimental design:• Hurricane Katrina (August 2005)• 9km for outer domain and 3km for the finest grid• GFS Reanalysis data (100km), SST of Gulf of Mexico was not updated• SBM is applied at the finest grid and Thompson scheme for the outer grid• Two simulations: over the sea N0 equals to 100cm‐3

• MAR:  N0 equals to 100cm‐3 (typical of maritime atmosphere)• MAR_CON: N0 equals to 1500cm‐3 (typical of continents under not very polluted conditions)

• TC in MAR_CON is weaker, when TC reached maximum intensity the difference was ~15hPa

• Aerosol concentration becomes similar to that over the land at distances of a few hundred km

• Aerosols can penetrate clouds in the eyewall

• Higher aerosol concentration can foster the lightening formation

• Decrease in lightening in the TC center and its increase at the TC periphery can serve as a precursor of TC weakening

Red dots: Zones of lightening. Green: TC eye

TC weakening mechanism:

Improvements:• High resolution needs to resolve all clouds at cloud periphery and convective clouds in the TC eyewall

• Take into account the effects of sea spray on cloud microphysics• Size of the finest grid should be increased to include additional intensification of rainband

72h simulation ~ 10 days (8 processors)!

Increasing of w and mass updraft also increases

Extra convective heating lowers the surface pressure, decreasing pressure gradient

Compensating downdrafts increase between these zones

Compensating downdrafts caused by convection at the periphery also damp the convection in the eye

Gilmore, M.,Straka, J., Rasmussen, E.

Presented by Patrick HawbeckerMEA 716Spring 2016

▪ 1 thermal profile (fromWeisman & Klemp1984) w/ several vertical shear variations

and Lr is evaporation rates

▪ 3D, nonhydrostatic cloudmodel (StrakaAtmospheric Model –SAM)

▪ Compared 3 MP schemes▪ “Kessler” – adjusted to work in SAM▪ Lr = “Lin–Farley–Orville” w/out ice▪ Li = “Lin–Farley–Orville” w/ice

▪ Main difference between Kessler

▪ Ice processes à stronger updraftsANDdowndrafts / cold pools

▪ Larger stratiform region and overallprecipitation totals with ice processes

▪ Kessler and Lr produce ~similar results, but Li produce more precipitation and updraft / downdraft strengths

▪ Supercell longevity was promoted with weakshear and weak cold pool (Lr)

▪ Interesting note: they admit their solution maybe resolution dependent

a) K20 d) K30 g) K40

TotalRM=36.53

J)K50

TotalRM=36.40

Max. Total Max.R 0=50.69 RM=18.45 R0=50.01HO= 3.68 HM= . 9 HO= 5.07

Total Max.RM=34.35 R0=48.69HM= 0.72 HO=6.26

Total Max. TotalRM::47.30 R0::48.02HM= 1.08 HO= 5.96

I0

' I

I Ikm 50RM::52.25HM"" 1 42

Max. Total Max. Total Max. Total Max.R0=32.95 RM=14.33 R0=30.10 RM "' 25.96 R0=33.60 RM=31.51 R0=33.83b) Lr20 e) Lr30 h) Lr40 k) Lr50

Max. Total Max. Total Max. Total Max.R0::50.41 RM=17.65 R0=36.33 RM =26.01 R0=38.46 RM=30.60 R0::40.89

WSM6 Performance Study

Objective‐ Assess and compare the 

performance of WSM6, a more complex WSMMPs, with WSM3 and WSM5

WSM6 ‐ Includes graupel as an additional 

predictive variable ‐ Water vapor, cloud water/ice, 

snow, rain, and graupel

Graupel “Communications”

• Freezing of rain droplets

• Accretion of cloud ice, cloud water, snow, and rain

• Depositional growth

• Sublimation decay• Melting

Experimental Design

Case Study – Heavy Rainfall EventJuly 15, 2001, local maximum of 371.5 mm

Design – 45‐km grid spaced domain, 2 nestedCP: KF | PBL: YSU | LSM: Noah | Radiation: RRTM 

23 Vertical Layers

Precipitation45‐km Grid Spacing

• Negligible differences across the schemes in the spatial distribution and intensity of precipitation 

5‐km Grid Spacing• Overall distribution of simulated 

precipitation is not changed across WSM3, WSM5, and WSM6

• Maximum amount of precipitation is greater with increasing complexity of scheme

Vertical Distribution of Water Species