10 dec 2007 precipitation uncertainty due to variations in particle parameters within a simple...
TRANSCRIPT
10 Dec 2007
Precipitation Uncertainty due to
Variations in Particle Parameters within a Simple Microphysics
Scheme
Dr. Matt GilmoreDAS/UIUC
(Photo Credit: NCAR/NSF)
Presented 10 Dec 2007
10 Dec 2007
Collaborators
• Jerry Straka - OU School of Meteorology
• Erik Rasmussen - OU/NSSL
Special Thanks toBob Wilhelmson, Adam Houston, Leigh Orf, Ted Mansell, Lou Wicker
Sue van den Heever
10 Dec 2007
Purpose
• Test precip. sensitivity to hail, snow, and rain particle characteristics in a single-moment bulk ice microphysics package (used in WRF)
• Can we trust precipitation forecasts from such models? Tornado forecasts?
• Motivate the use of more sophisticated microphysics
10 Dec 2007
National Hail Research Experiment
1972-1974, 1976, Grover, CO
(Photo Credit: NCAR/NSF)
• Examined the influence of cloud seeding on hailfall
(Photo Credit: NCAR/NSF)
Former division that worked on NHRE became MMM.
10 Dec 2007
“... the specific sources of error must be identified in current microphysical parameterizations, and physically based improvements to the model physics must be developed, particularly for ice formation...” (MMM/NCAR Science Plan, Oct. 2000)
Prediction of Precipitating Weather Systems - Cloud Microphysics and Precipitation
(Photo Credit: NCAR/NSF)
The Costner’s, April 1975
10 Dec 2007
Top-5 Severe Weather Hazardsthat Caused Crop and Property Damage
(1995 - 2000)
Average Cost PerYear(Billions of U.S. dollars)
Flood
Tornad
o
Hurrica
ne Hail
Drough
t$0.0
$1.0
$2.0
$3.027% 25%
11% 9% 8%
Total=$11.2 Billionper Year
(NWS Natural Hazard Statistics)
10 Dec 2007
Hail Vulnerability• Tobacco, tea, and soybeans ->
numerous 5 mm hail,(Changnon 1971, 1977, 1999)
• Other crops, farm animals, & property -> larger hail(Largest from supercells;
Changnon 2001)
• Could we someday forecast hail occurrence & its characteristics?
(Photo Credit: NCAR/NSF)
10 Dec 2007
Supercell Flooding Potential
• Hypothesized that supercells might contribute to climatology of extreme rainfall (100-year recurrence interval). (Smith et al. 2001, J. Hydrometeor.)
e.g., dense gauge network, max rainfall rates:Orlando, FL: 330 mm h–1 (26 Mar. 1992)Dallas, TX: 231 mm h–1 (5-6 May 1995)
Would cloud model flooding predictions help?
10 Dec 2007
Hail Climatology
Changnon and Changnon (1999)
Changnon (1996)
# Hail Days/year (1901-1994)
Mean Hail Diameter (cm)
10 Dec 2007
Hail Forecasting• Climatology
• Synoptic Pattern Based– See review articles by Longley and Thompson (1965); Johns and Doswell (1992).
• Sounding Based– Edwards and Thompson (1998) demonstrated thermodynamic parameters to be practically useless in hail severity forecasting: VIL, CAPE, maximum parcel level, EL, convective cloud depth, wet-bulb zero, freezing level. Suggested that forecasts of storm rotation be made. Progress?
• Reliable NWP cloud model forecasts? (We show …not yet)
10 Dec 2007
Numerical Experiments
• Straka Atmospheric 3-D Cloud Model (SAM). Similar to the N-COMMAS model.
• 6th order Crowley advection on scalars. 2nd order flux scheme on momentum.
• Prognostic equations for u, v, w, , p, TKE, qv, qc, qr. Optional ice physics.
10 Dec 2007
Model Set-up
• 91x91x22 km domain
• dx,dy = 1000 m, dz = 500 m
• 1°C perturbation bubble.
• Horizontally homogeneous environment
10 Dec 2007
Experimental Design• Idealized environments from Weisman and Klemp (1984).
HodographsT , Td Profile
CAPE=2200 J kg-1
10 Dec 2007
Experimental Design• Microphysics scheme:
– Three-class bulk ice scheme (ice, snow, and hail/ graupel). [Lin et al. (1983), “LFO”] Includes warm rain scheme for cloud and rain. Referred to herein as “3-ICE”
– Scheme described in Gilmore et al. (2004; MWR, August issue)
– Similar schemes still being used (Purdue-Lin and WSM-6)
10 Dec 2007
“3-ICE” Process Rates
• 6 species• Single moment
PRECIPITATION ON GROUND
RAINHAIL/GRAUPEL
WETOR DRYGROWTH
SNOW
CLOUDICE
CLOUDWATER
WATER VAPOR
qhcns, qhacs,
qhdpv
qvevr
qsdpv
qrmlh
qiint
qsacw, qsfw
qrmls
qrcnw, qracw, qh
acw qs
aci, qsfi,
qscni
qhfzr,
, qsacr
qhaci,
qvsbsqvsbh
qrshh
qwmliqwcdv
qvevw qvsbi
qidpv
qifzw
qraci,
qsacw
qiacr
qiacr,
qracs
qraci
qsacrqhacr,
10 Dec 2007
Many Limitations
• Species are constrained to a predefined distribution function which experiences bulk/average fallout.
• Scheme doesn’t predict number concentration or other moments.
• Inconsistencies in diameter’s used for different processes (see McFarquhar & Black 2004; Potter 1991)
• Efficiencies are assumed constant.
10 Dec 2007
Limitations (continued)
• Only a single graupel/hail (qh) species.
• Pick parameters such as the particle intercepts, densities, and fall velocity coefficients a priori
(There are many more “gotchas”… but we will focus on the two above)
10 Dec 2007
NHRE-Motivated Sensitivity Testing
Relative Frequency
104 106 108
Intercept (dm-3 mm-1 )
.06
.12
Knight et al. (1982)
10 Dec 2007
Intercept Observations
Hail/graupel intercept (noh) (dm-3
mm-1 )
References
103 – 105 (hail) Federer and Waldvogel (1975), Dennis et al. (1971), and Spahn (1976)
102 – 106 Cheng et al. (1985)
104 – 108
graupel/hailKnight et al. (1982)
10 Dec 2007
Particle Density Observations
Density (kg m-3)
Hail: 700 to 900Graupel: 50 to
890
Hail: 400
Reference
Pruppacher and Klett (1978)
• D. Zrnić, personal communication
10 Dec 2007
Species Size Distributions
Standard LFOparamsno
1
104
106
102
108
Particle Diameter (mm)
0 10 20 30 40
rainsnowN2ρ93N ρ94N ρ95N ρ94N ρ45N ρ46N ρ47N ρ48N ρ4
a) Number Conc. per mm bin
n(D) {count dm
–3 mm
–1}
10 Dec 2007
Species Fallspeeds
• Larger and more dense fall faster
10 155
Snow
Water Content (g m–3)
Rain
N5ρ4
3N ρ9
4N ρ9
5N ρ9
6N ρ4 7N ρ48N ρ4
4N ρ4
5
10
15
20
25
Vt (m s–1)
c) Mass-weighted mean fall speed
rainsnowN2ρ93N ρ94N ρ95N ρ94N ρ45N ρ46N ρ47N ρ48N ρ4
10 Dec 2007 4
x3x
10 Dec 2007 2 h accumulation of qr & qh
10 Dec 2007
N3ρ9
PeasDimesGolfballs
10 Dec 2007
N4ρ9
PeasDimes
10 Dec 2007
N5ρ4
Peas
10 Dec 2007
Hail counts over a 100 m2 area
10 Dec 2007
Midlevel Storm Structure and Surface Gust Front Evolution
t=30t=60 t=90t=120
10 Dec 2007
Midlevel Storm Structure and Surface Gust Front Evolution
N8
t=30t=60 t=90t=120Warmer!
10 Dec 2007
Getting microphysics “right” is important for simulating
the right downdraft
Z ~ 500 m
Z = 0 m
10 Dec 2007
N8
Low-level Storm Structure and Surface Gust Frontt=60 min (Us=50)
10 Dec 2007
Temporally and Horizontally-AveragedVertical Profiles
Average Water Content (g m-3 ).10 .15 .20.05 .250 .30 .05 .10
Hail/Graupel
Rain
Snow Cloud ice
CloudWater
10
8
6
4
2
12
14
.035
Z(km)
10 Dec 2007
Temporally and Horizontally-AveragedVertical Profiles (cont’d)
Z(km)
N3ρ94N ρ94N ρ45N ρ95N ρ46N ρ47N ρ48N ρ4
LFO warm rain
10
8
6
4
2
12
14
.10 .15 .20.05 .25 .30
Total Species
( Average Water Content g m-3 )
10 Dec 2007
Thus, N8r4 case has larger massof qh but spread over larger areaSmaller fall velocities
10 Dec 2007
Second Sensitivity Test
• When the intercepts for the following species were increased individually by 2 orders of magnitude, then total precipitation fall decreased by...– Rain (N6 to N8): 3 Tg – Snow (N5 to N7): 4 Tg– Large Ice (N4 to N6): 12 Tg
• Greatest sensitivity to qh
10 Dec 2007
Third Sensitivity Test
• Change the ground-relative (GR) motion but not the vertical wind shear...
• Result: point rainfall and hailfall doubles although total system mass preserved
New GR-Motion: subtract 10 m s–1 from u windspeeds
10 Dec 2007
Results Summary• Mass fallout changes by a factor of 4 when the “large ice” distribution is varied within its observational limits. Weaker shear shows differences of a factor of 3. Us=20 m/s cases (not shown) had factor of 2.
• Thus, sensitive to the vertical wind shear
• Model precipitation is more sensitive to changes in the “large ice” distribution than the snow or rain distribution.
• Precipitation depth is also a function of GR-storm motion.
10 Dec 2007
Conclusion
• While simple bulk microphysics schemes such as LFO can be “tuned” and may have value in a research mode, they are probably not well suited for precipitation forecasts.
10 Dec 2007
Future Work: Improved schemes
Multiple Moments for 3 or 4 ice categories• e.g., Ferrier (1994),
Siefert and Beheng (2001)
More ice categories (necessary - McCumber et al.)
More liquid and ice categories, more moments, particle density prediction,
Straka and Gilmore (2008)
10 Dec 2007
Two Moment, 5-ice, 3-
rain scheme
• Straka and Gilmore (2008)
– Will this reduce the precipitation uncertainty?
10 Dec 2007
(Photo Credit: NCAR/NSF)
Joe and Jill Newham, April 1975
Thanks for your attention!
10 Dec 2007
10 Dec 2007
Rates t = 30 min
10 Dec 2007
Supplemental Figures
Rates t = 30 min
10 Dec 2007
Rates t = 90 min
10 Dec 2007
Supplemental Figures
Rates t = 90 min
10 Dec 2007
Conventional versus Dual-Polarization Radar
Light Rain
Wet Snow
Heavy Rain
Large Drops
Moderate Rain
Vertical Ice
Horizontal Ice
Dry Snow
Hail
Graupel/Small Hail
Rain/Hail
Conventional Radar Reflectivity
Hydrometeor Classification
Cimmaron, OK, az=148.2°,2316 UTC 6 June 1996. (Adapted from Zrnic et al. 2001).