Thu 3/31/2016 Representation of clouds and precipitation: - WRF scheme options + some background - Sensitivity to CCN: - Convective storm case study - Stratus dissipation problem Reminders/announcements: • MP paper summary assignment • Upcoming: MP experiments • Tuesday 4/5: Class will not meet, but recorded lecture will be posted on www page
75
Embed
2016 L21 MEA716 3 31 post - Nc State Universitygary/mea716/2016_L21_MEA716_3_31_post.pdf · Shutting off melting in WSM6 MP made little difference ... Keith made a nice observation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Thu 3/31/2016Representation of clouds and precipitation:
- WRF scheme options + some background- Sensitivity to CCN:
- Convective storm case study- Stratus dissipation problem
Reminders/announcements:• MP paper summary assignment• Upcoming: MP experiments• Tuesday 4/5: Class will not meet, but recorded lecture will
be posted on www page
MP papers: Will do sign-up sheet
Re-Cap from TuesdayCase study of Buffalo early-season lake-effect snow event
Shutting off melting in WSM6 MP made little difference
Shutting off evaporation and sublimation resulted in warming, no snow accumulation
Keith made a nice observation that if the NOAH LSM was still in the older mode, then the lack of snow could be due to T above freezing… so went and checked
Great example of model physics interactions!
Little difference in T between NOMELT, control
Up to 1.5 C warmer in NOEVAP than control
No snow in NOEVAP
• similar amount of precipitation, too warm to be snow
Similar snowfall amounts in NOEVAP, control
noevap
control
nomelt
2-meter T at Buffalo
Model snowfall rate at Buffalo
NOAH LSM, then and now
WRFV2.2 NOAH LSM:
WRFV371 NOAH LSM:
A partial listing of microphysical processes
Deposition/sublimation of graupel Melting of hail
Deposition/sublimation of cloud ice Homogeneous freezing of cloud water
Deposition/sublimation of snow Homogeneous freezing of rain water
Deposition on ice nuclei (nucleation) Accretion of cloud water to graupel
Deposition/sublimation of hail Accretion of cloud water to snow (rime)
Condensation/evaporation of cloud water Coalescence of cloud water to rain
Condensation/evaporation of rain water Cloud water to rain (autoconversion)
Freezing of cloud water (immersion, contact) Accretion of cloud ice by snow
Freezing of rain water (immersion, contact) Accretion of snow by snow (aggregation)
Melting of cloud ice Accretion of cloud ice by graupel
Melting of snow Accretion of rain water to graupel
Melting of graupel Accretion of snow by graupel
Breakup of rain drops Splintering of freezing droplets
Breakup of snow crystals Accretion of snow by rain
WRF MP Options (first 11 of 22) (V3.7.1)
mp_physics Characteristics Design
1: Kessler Bulk, single moment, 3 class Simplest possible, use for test case
2: Lin Bulk, single moment, 6 class Classic, ahead of its time. WSM6 updates
3: WSM3 Bulk, single moment, 3 class, simple ice All ice below freezing, all warm above
4: WSM5 Bulk, single moment, 5 class Probably okay for coarse grid spacing
5: Ferrier Bulk, single moment, NAM scheme Clever total ice advection for efficiency
6: WSM6 Bulk, single moment, 6 class, can add hail An updated Lin et al., standard/default
7: Goddard Bulk, single moment, 6 class, hail option Can set switch to have hail vs. graupel
8: Thompson Bulk, 6-class, double moment for ice, rain Computationally efficient part double moment
9: Milbrandt-Yau Bulk, 7-class, double moment for all Expensive. Double moment, fixed alpha
10: Morrison Bulk, 6-class, double moment for cloud ice, cloud water, rain, and snow. Can add hail
Designed for arctic clouds, good radiation int.
11: CAM 5.1 Bulk, 5-class, double moment, NEW in WRF Some aspects not fully implemented in WRF
14: WDM5 Bulk, single moment, 3 class Perhaps useful in special cases
16: WDM6 Bulk, single moment, 6 class, can add hail Double moment for CCN, cloud, rain
17: NSSL 2/4 Bulk, NSSL double moment 4 ice All NSSL schemes allow setting intercept
18: NSSL CCN Bulk, double moment, predicted CCN For double moment, can set shape param.
19: NSSL single 7 Bulk, single moment, 7 class Predicts graupel
21: NSSL single 6 Bulk, single moment, 6 class, set intercepts Can set many namelist parameters
28: Thompson AA Bulk, Thompson aerosol aware Can utilize climatological aerosol values
30: HUJI Bin, fast version Hebrew University of Jerusalem, Israel (HUJI)
32: HUJI all Bin, full version (expensive) Even fast version ~100x slower than M-Y?
95: Ferrier old Bulk, old Eta version, still in NAM? Not sure why you would use this over 5
WRF Microphysics
WRF Microphysics
• Cloud processes and latent heating
• Several levels of sophistication from warm rain to graupel schemes
• Many options now… several are new– Kessler warm rain– WRF Single-Moment 3-class, 5-class, 6-class– Ferrier (Eta) scheme– Thompson et al. graupel scheme– Lin et al. graupel scheme– Double-moment schemes…
• 3-class microphysics with ice, no graupel or hail
• Only ice processes below 0C, includes ice sedimentation
• Cloud ice only below freezing
• Liquid only above freezing
• Very simple melting layer
• Bulk, single moment
WSM5: mp_physics = 4WSM 5-class scheme• Also from Hong, Dudhia and Chen (2004)
• Replaces previous NCEP5 scheme
• 5-class microphysics with ice, still no graupel or hail
• Includes supercooled water and snow melt
• Ice sedimentation
• Similar to WSM3, except more completewith Bergeron process, realistic mixed-phase
Adequate for coarse grid spacing, parameterized convection, stable atmosphere
mp_physics=5Ferrier (current NAM) scheme• Designed for computational efficiency
– Advection of total condensate– Cloud water, rain, & ice (cloud ice, snow/graupel)
from storage arrays – assumes fractions of water & ice within the column are fixed during advection
• Has mixed-phase processes to -30C
• May be challenged for proper advection of hydrometeors
• Introduces ice at too warm a temperature?• Tunable parameters, very efficient
Ferrier
Qi/Qs/Qg
Qv
Qc
Qr
Ferrier scheme (cont.)• Ferrier Scheme is in operational NAM model
• Four classes of hydrometeors: • Suspended cloud liquid water droplets • Rain • Large ice (includes snow, graupel, sleet, etc.) • Small ice (generally suspended cloud ice, evaporates quickly in air
subsaturated with respect to ice)
• Only has full prediction equations, including advection, for total condensate (sum of all four hydrometeor classes)
• Does separate (diagnose), but not track with advection, total cloud ice (large ice + small ice), rain, and cloud water
WSM6: mp_physics = 6WSM 6-class scheme: Hong and Lim (2006 JKMS)
• 6-class microphysics with graupel, ice [#] as WSM3, WSM5
• Essentially similar to Purdue Lin, but updated
• Modified processes for more accurate vertical heating profile
• Easy to identify manipulate
• Recommended over Lin scheme
• Still single moment
• Now hail_opt NL switch: 1 to add hail
WSM6 (Hong and Lim 2006)
Goddard: mp_physics = 7Goddard scheme• Reference Tao et al. (1989 MWR - Note)
• 6-class microphysics with graupel
• Similar to Lin, but more hail options – use – gsfcgce_hail: 0 for graupel, 1 for hail – gsfcgce_2ice: 0 for all, 1 for ice and snow, 2 for ice + graupel (?)
• Still single moment
• I have limited experience with it…
• Proven “reasonably accurate” for tropical events, midlatitude convection
Thompson: mp_physics = 8Thompson graupel scheme:
– Reference: Thompson et al. (2004, MWR, updated Thompson et al. 2008)
– Newer version of Reisner2 scheme (MM5)
– 6-class microphysics with graupel
– Double moment, 5 class (now rain also)
– Sophisticated scheme, but computationally affordable
– Two implementations in WRF, 1st had problems, was only 2x Mom. for ice
4) PI size distribution assuming Gamma distribution, but shape parameter mu_s=0 (reduces to exponential) currently
7) The Liu and Daum autoconverion is quite sensitive on Nt_c. For mixed-phase cloud and marine environment, Nt_c of 10 or 20 (E6) is suggested. Default value is 10E.6. Change accordingly for your use.
WDM5: mp_physics = 14WRF Double Moment 5-class:
– Reference: Lim and Hong (2010)
– 4 classes, partial double moment
– Predicts CCN [#], and number concentration for rain, cloud
– Good in same situations as WSM5?
– Ice species (cloud ice, snow) all follow WSM5
– Can set background CCN [#]?
WDM6: mp_physics = 16WRF Double Moment 6-class:
– Reference: Hong et al. (2010) and Lim and Hong (2010, MWR)
– 5 classes, double moment for cloud, rain, CCN [#]
– Eliminates biases in WSM6 in published case
– Journal papers compare this to WSM6 for squall line, monsoon
– Utilizes hail_opt, can set CCN?
NSSL: mp_physics = 17, 18NSSL double moment schemes:
– Include both graupel and hail; double moment for 6 classes
– Option 18 predicts CCN, and graupel volume?!
– These have been added to WRF recently (V3.4)
– Additional namelist options available, can set shape parameter, intercept, and density in namelist
– Designed for dx < 2km, maybe idealized best
NSSL: mp_physics = 19,21NSSL single moment schemes:
– Include both graupel and hail
– These have been added to WRF recently (V3.4+) – I’ve very little experience with them
– Additional namelist options available
NSSL: mp_physics = 19,21NSSL schemes:
21, NSSL 1-moment, (6-class), very similar to Gilmore et al. 2004Can set intercepts and particle densities in physics namelist, e.g., nssl_cnorFor NSSL 1-moment schemes, intercept and particle densities can be set for snow, graupel, hail, and rain. For the 1- and 2-moment schemes, the shape parameters for graupel and hail can be set:
Aerosol aware, discussed in Thompson and Eidhammer (2014, JAS)
Can utilize CCN climatology (or eventually analysis) for initial conditions
Nate Farrington thesis results – for stratus dissipation
Also excellent interactions with RRTMG radiation scheme (missing for other choices)
Stratus Dissipation: Bottom‐Up Erosion
Morning Afternoon
LCL
LCL
Solar radiation warms PBLLCL raised in response
Stratus Dissipation: Bottom‐Up Erosion
Morning Afternoon
LCLLCL
Larger number concentrationLess solar warming of PBLIncreased cloud lifetime
Stratus Dissipation: Bottom‐Up Erosion
Can a numerical model such as WRF accurately represent this type of cloud‐radiation interaction? (Twomey, cloud‐lifetime effects)
If cloud‐droplet number concentration increases, should see reduced solar flux at surface for a given amount of cloud (at a certain time of day)
Testable hypothesis: Plot surface solar flux as function of cloud liquid water content for different experiments
WRF Single‐Column Model Tests
Thompson, RRTMG, nt_c set to 10000
Thompson, RRTMG, nt_c set to 1000
Thompson, RRTMG, nt_c set to 100 (default)
58 W/m2 230 W/m2185 W/m2
GOES 15 Visible satellite17 UTC 18 July 2013
YSU (default and aerosol) + MYJ (default) cloud water with topography, hour 29 (17Z)
17 UTC 18 July 20135x10‐6 kg/kg isosurfaces
YSU (default and aerosol) + MYJ (default) cloud water with topography, hour 29 (17Z)
17 UTC 18 July 20135x10‐6 kg/kg isosurfaces
YSU (default and aerosol) + MYJ (default) cloud water with topography, hour 29 (17Z)
17 UTC 18 July 20135x10‐6 kg/kg isosurfaces
YSU (default and aerosol) + MYJ (default) cloud water with topography, hour 29 (17Z)
17 UTC 18 July 20135x10‐6 kg/kg isosurfaces
HUJI fast: mp_physics = 30New in WRF, BIN scheme from Hebrew
University of Jerusalem
Fast version (still 8,977 lines of code)
Tried running on 64 processors (winter storm case); crashed after ~ 10 hours run time (due to adaptive time step?)
Wall clock per time step ~ 45 seconds versus 0.14 seconds (Kessler, Ferrier), 0.3 WDM5
How to chose a MP scheme?What are you simulating? What MP processes
could affect it, and what scheme best represents those processes?
Is computational cost worth it?
More complex MP schemes may not be compatible with other parameterization schemes (e.g., radiation) which have been developed and tuned to work with simpler schemes (- Morrison workshop presentation)
Performance for winter storm run
0.1
1
10
100
Kessler Fast Bin HUJI Milbrandt‐Yau WDM5 WSM6 (hail opt)
Damage exceeded $2B USD (Durkee et al. 2012)
3-day precipitation exceeded 250 mm over substantial area
Q2 radar estimate vs. 18-km WRF domain, 72-h total
Q2 radar estimate vs. 6-km WRF domain, 72-h total
Q2 radar estimate vs. 2-km WRF domain, 72-h total
Nate’s MP Assignment: Thompson with RRTMG
Exp - Ctrl
Experiment:
• Tennessee flood event
• CDNC from 100 to 1000 cm-3
Results• Precipitation displaced
north and west of ctrl
• Synoptic precipitation field displaced northward
Ctrl (CDNC 100 cm-1) Exp (CDNC 1000 cm-1)
Results: Albedo (SWDOWN)
Results• The density of the entire cloud shield increased in Exp
• Much less sw radiation reached the surface in Exp• Much more nucleation took place, increasing total droplet SA
Increasing the CDNC of the Thompson MP scheme from 100 cm-3 to 1000 cm-3 was shown to: Displace (in some instances enhance) precipitation downwind Increased the amount of time (and therefore distance) each cloud took
to produce precipitation With more time, more lift, increased LW Path, downstream precip
maxima were observed
Increase cloud nucleation/density (albedo) Much less SWDOWN was observed in Exp versus Ctrl Difference field showed couplets of upwind decrease (increase) and
downwind increase (decrease) in cloud albedo (SWDOWN)
NOTE: Due to my own inability to correctly plot vertical latent heating profiles, I
could not test the latent heating portion of my hypothesis
Conclusions
THE EFFECT OF CHANGING THEINITIAL CLOUD DROPLETCONCENTRATION ON PRECIPITATIONTim GlotfeltyMEA 793, spring 2011
INTRODUCTION
Aerosols impact the environment: One aspect is aerosol indirect effect Several processes are lumped into “indirect effect”,
most well known is the impact of increased aerosols on cloud condensation nucleus distribution
Here, indirect effect of aerosols on precipitation was modeled by changing initial CCN concentration from 100 cm-3 to 1000 cm-3 in WDM6 microphysics scheme for the Tennessee Flooding Case May 1st-4th 2010
HYPOTHESIS
Increasing number of CCN in should result in an overall decrease in precipitation during the event.
More CCN will distribute water vapor over a larger number of smaller drops.
This will decrease ability of drops to grow large enough to fall as rain, thus resulting in lower accumulated rainfall.
PRECIPITATION COMPARISON
00Z May 4th
Control Increased CCN
mm
PRECIPITATION DIFFERENCE
The precipitation appears to have shifted slightly from its original location (As evidenced by the couplet banding).
There are large areas of decreases and smaller areas of increases of roughly the same amount (100-150 mm)
SUMMARY AND CONCLUSION
Increased CCN led to a decrease in the maximum amount of accumulated precipitation by 22 mm.
The increased CCN also resulted in some hot spots of increased precipitation, shifts in cell location
This most likely results from the predicted CCN concentration. The indirect effect is not a linear process. CCN are beneficial
to precipitation until the concentration becomes too large. So in the areas where precipitation increased the increased CCN were most likely in the regime where they helped increase rain formation and the areas that saw decreases were probably in the other. [Also, consider strength of ascent]
Implies that more CCN could cause flooding damage from an event like this to be more localized.
Which MP scheme is best for you?
Discussion
5 minutes: small group5 minutes: discuss in classIdentify group spokesperson
Considerations:
• What phenomena/processes must you forecast/represent?• What grid length are you using?• How much computational expense are you willing to