The Challenges of Accurate Snowfall Forecasts: Implications for Observing Strategies and Future Rese

1
The Challenges of Accurate Snowfall Forecasts
Implications for Observing Strategies and Future
Research Efforts
Dr. David Schultz CIMMS and NOAA/NSSL Norman,
Oklahoma
2
Forecasting snowfall is a mesoscale challenge
cloaked in a synoptic-scale culture. Dr. Louis
Uccellini, Director, NOAA/NCEP 3 October 2002,
Midatlantic Winter Storms Conference
3
OBJECTIVES

• Discuss the theory and some supporting
  observations for the importance of snow
  microphysics in determining snowfall.
• Discuss the density of new snowfall and attempts
  to predict it.
• Present research advances required to
  improve snowfall forecasting.

4
Methodologies for Forecasting Snowfall

• Climatology
• Heavy snow is favored 2.5 to the left of the
  track of the 500-mb vorticity maximum (Goree and
  Younkin 1966).
• Personal experience, pattern recognition This
  looks like a 6-inch snowstorm.
• Rules of Thumb
• Average 24-h snowfall in inches is 1/2 of the
  maximum indicated 200-mb warm advection in C
  (Cook 1980).
• Maximum potential snowfall is twice the average
  mixing ratio at 700 mb (Garcia 1994).
• For the problems with rules of thumb, see
  Schultz et al. (2002), Comments on An
  operational ingredients-based methodology for
  forecasting midlatitude winter season
  precipitation.

5
Methodologies for Forecasting Snowfall

• Mesoscale effects
• Conditional symmetric instability (e.g.,
  Schultz and Schumacher 1999)
• Mesoscale banding (Novak et al. 2002)
• Cloud microphysics
• Is this the last frontier?

6
TOP-DOWN APPROACH

• Dan Baumgardt (NWS WFO La Crosse, WI) has been
  advocating the top-down approach to
  forecasting.
• Starts at the top of the environmental sounding
  and traces a hydrometeor trajectory down to the
  surface
• Considers three levels in the sounding
• ice-producing level
• warm layer
• cold surface layer

7
Steps in Producing Snow

• 1. Is it cold enough to activate ice nuclei?
• Function of temperature and type of substrate
• 2. Is the ice crystal growing by deposition?
• Function of temperature and supersaturation
• 3. Is the snow collecting supercooled liquid
  water as it falls through the cloud (riming)?
• Function of temperature, supersaturation, and
  vertical motion
• 4. Are the snow crystals aggregating?
• Function of temperature, crystal shape, and
  amount of turbulence
• 5. Is the phase changing?

8
1. THEORY Will Ice Be Produced in the Cloud?

• Is it cold enough to activate ice nuclei?
• Ice nuclei are a subset of cloud condensation
  nuclei (CCN) that act as a surface for ice growth
  to initiate.
• Some ice nuclei have crystal structures similar
  to ice.
• Only 1 in 108 airborne particles nucleates ice at
  20C.
• Every 4C drop in temperature increases the
  number of ice nuclei by tenfold.
• Ice nuclei activate at different temperatures.
• Ice 0C
• Silver iodide 4C
• Kaolinite 9C
• Vermiculite 15C
• Pseudomonas syringae (bacteria from decaying
  leaves) 2C

9
1. OBSERVATIONS Will Ice Be Produced in the
Cloud?
Oklahoma City soundings for snow/rain/freezing rai
n/ice pellet cases
(Michael Schichtel 1988,OU M.S. thesis)
10
1. OBSERVATIONS Will Ice Be Produced in the
Cloud?
snowno-snow cut-off temperature advocated by
Wetzel and Martin (2001)
cloud-top temperature (C) vs cloud-top pressure
(hPa) from 64 soundings during snowfall events
at Albany, Minneapolis, and Denver (Schultz et
al. 2002).
11
OBSERVATIONS SEEDERFEEDER PROCESS

• Ice crystals from a mid to high layer of clouds
  fall into a lower layer of supercooled liquid
  water clouds, sparking ice nucleation
• Distance between clouds is less than about 5000
  feet (1.5 km)

12
OBSERVATIONS SEEDERFEEDER PROCESS
(Hentz)
13
2. THEORY How does ice grow in cloud?

• Growth by deposition (vapor condenses directly
  onto ice crystal as ice) BergeronFindeisen
  process
• Function of supersaturation with respect to ice
  (temperature) and pressure

14
2. THEORY How Does Ice Grow in Cloud?
maximum depositional growth rate (dendrites)
(Dennis Lamb, Penn State)
15
2. OBSERVATIONS How Does Ice Grow in Cloud?
After 30 mins., dendrites grow to 10 times the
mass of the next largest ice crystal.
Fukuta and Takahashi (1999)
16
2. OBSERVATIONS How does ice grow in cloud?
(Mahoney)
17
2. OBSERVATIONS How does ice grow in cloud?
(Mahoney)
18
2. OBSERVATIONS How Does Ice Grow in Cloud?

• Waldstreicher (2001)
• http//www.erh.noaa.gov/er/hq/ssd/snowmicro/
• Following Auer and White (1982)
• Intersection of temps of 12 to 18C and omega
  at least 10 microbars s-1 in RH75
• 4 winters in northeast PA and central NY, 55
  synoptic-scale snow events that met warning
  criteria, 75 synoptic-scale snow events that met
  advisory level, examined Eta/Mesoeta output.
• 76 of warning-level events showed this
  intersection, whereas only 9 of advisory-level
  events met this criteria

19
OBSERVATIONS How does ice grow in cloud?
2 ft of snow during rush hour
20
NSHARP utility at the Storm Prediction Center
21
AWIPS utility to estimate the residence time of
ice crystals in the dendritic-growth region
(minutes)
(Dan Baumgardt, NWS)
22
3. THEORY How does ice grow in cloud?

• Growth by accretion ice crystal collects
  supercooled liquid water drops (riming to produce
  graupel)
• Solid evidence of saturation at 1 to 5C

(David Babb)
23
Growth by accretion will eventually
dominate ice-crystal growth
Fukuta and Takahashi (1999)
24
3. THEORY How does ice grow in cloud?

• HallettMossop (1974) secondary ice production
  mechanism
• Rime will splinter at 5 to 10C as it freezes,
  thus producing more ice nuclei
• These rime splinters can get lifted in the
  updraft again, thus acting to sweep out more of
  the supercooled liquid water.
• Increased precipitation efficiency

25
Convective snow environments

• Deeper circulation (likely to reach toward colder
  temps and produce ice nuclei, acts as a seeder to
  supercooled liquid water regime)
• Strong vertical motions, heavy precipitation
• Greater possibility of riming
• Look for elevated CAPE (Trapp et al. 2001)
• Thundersnow

26
4. THEORY How does ice grow in cloud?

• Growth by aggregation joining of multiple ice
  crystals to form a snowflake
• Most important at 0 to 5C as surface of ice
  becomes sticky, with a secondary maximum around
  15C due to interlocking dendrites

27
4. OBSERVATIONS How does ice grow in cloud?
Enhancing growth by turbulence
IMPROVE II NOAA/ETL S-band Radar 1314 December
2001
Reflectivity
(Houze et al.)
28
IMPROVE II NOAA/ETL S-band Radar 1314 December
2001

• aggregation /or riming enhanced by the turbulent
  overturning

bright band

• turbulence likely overwhelmed by fall speeds of
  rain

Upward Radial Velocity
(Houze et al.)
29
5. Hydrometeor-altering environments

• Warm layers snow-rain, sleet, freezing rain
• Wet-bulb temperature and dry layers rain- snow
  (e.g., Kain et al. 2000)

30
Summary of Top-Down Microphysics Approach for
Snow

• Need ice nuclei (cold temps to activate or
  seederfeeder)
• Need growth mechanism
• Deposition (vertical motion at 15C)
• Riming (supercooled liquid water at 1 to 5C)
• Aggregation (near 0C and/or turbulent)
• Embedded convection (CAPE)
• Diabatic effects (advection small)

31
Even if you were able to predict the liquid
equivalent perfectly

• . . . youd still have to know the snow density.
• Usually this is assumed to be 10 inches of snow
  to 1 inch of liquid water (snow ratio).
• This will vary, however, depending on ice-crystal
  habit (function of RH and T), degree of riming,
  surface compaction due to weight and wind.
• Need to consider crystal shape when formed and
  the compaction of crystals on the ground.

32
isometric crystals
isometric crystals
Apparent crystal density for a single ice crystal
4550 s after seeding (Fukuta 1969)
columns
dendrites
Apparent ice crystal density at a growth time of
10 minutes (Takahashi et al. 1991)
Density can vary by a factor of 29, depending on
crystal shape
33
Apparent density will decrease, then stabilize as
crystal grows
(Fukuta and Takahashi 1999)
34
Density will decrease as snowflakes increase in
size, but it is not a simple relationship.
(Rogers 1974)
35
Factors Affecting Snow Ratio
Snow ratio versus liquid equivalent for snowfall
from five stations in western Canada
(Courtesy of Gabor Fricska and Alex Cannon)
36
Factors Affecting Snow Ratio

• Simple measures like lower-tropospheric
  temperature rarely work, except in very special
  cases.

37
NWS snow-density vs temp. table
Function of surface temperature only!
Developed as a guide for QC of observations Not
intended as substitute for obs or as a forecast
method
38
Roebber et al. (2003)Improving Snowfall
Forecasting by Diagnosing Snow Density, Wea.
Forecasting.

• GOAL To do better than the 101 ratio.
• PROBLEM Science on what controls the snow ratio
  is unknown.
• Dataset constructed of 1650 snowfall events at 28
  radiosonde stations in the U.S. 2 inches snow
  (0.11 inch liquid) with wind
• Snow densities binned
• heavy 11 91
• average 91 151
• light 151

39
(Roebber, Bruening, Schultz and Cortinas)
40
(No Transcript)
41
Properties of Snow Ratio

• A principal component analysis isolates factors
  influencing snow ratio
• Month (solar radiation)
• Temperature profile (lowmid, midupper)
• RH profile (lowmid, mid, upper)
• External compaction (wind speed, liquid
  equivalent)
• Compaction of snowfall once on the ground was the
  most crucial parameter to predict snow ratio
  (wind speed and liquid equivalent).

42
How are we doing now?

• For diagnosing snow ratio class (heavy,
  average, light) in a test sample
• 101 rule 45.0 correct
• climo 41.7 correct
• NWS table 51.7 correct

43
How are we doing now?

• For diagnosing snow ratio class (heavy,
  average, light) in a test sample
• 101 rule 45.0 correct
• climo 41.7 correct
• NWS table 51.7 correct
•  Ensemble of neural networks that are fed
  sounding parameters, surface windspeed, and
  liquid-equivalent amount
• 60.4 correct

44
How are we doing now?

• For diagnosing snow ratio class (heavy,
  average, light) in a test sample
• 101 rule 45.0 correct
• climo 41.7 correct
• NWS table 51.7 correct
•  Ensemble of neural networks that are fed
  sounding parameters, surface windspeed, and
  liquid-equivalent amount
• 60.4 correct
• Heidke skill score improves 184 between NWS
  table (0.120) and neural network (0.341)

45
The Fall Velocity of Snow and Why It Matters

• These sensitivities to snow fall speed will
  impact where snows will fall in numerical models
  with small horizontal grid spacing.

Fukuta and Takahashi (1999)
46
850-hPa wind dir.
Overprediction bias 140 (solid lines)
Underprediction bias
Colle et al. (1999)
47
Idealized MM5 2-D Simulation
(Courtesy of Brian Colle)
IPEX IOP 3
48
What do we need to do to forecast snow
better?Observations

• Larger quantity and in real time (daily to
  every 1-minute)
• Cooperative Weather Observers upgrade
• Weather Support to Deicing Decision Making
  (WSDDM)
• Better quality
• Take measurements! Dont rely on simplistic
  tables or constant snow ratios.
• Nolan Doeskens snow measurement video
• Observations of crystal types
• Can dual-polarimetric radars be of use?
• Can satellite IR data be used to estimate
  cloud-top temperature for identifying activation
  of ice nuclei?

49
The Promise of Polarimetric Radar

• Hydrometeor discrimination
• real-time algorithm exists for discriminating
  rain, nonaggregated ice crystals, aggregated dry
  snow, and aggregated wet snow
• discrimination among the habits of
  nonaggregated ice crystals is also possible
• Quantitative analysis
• If the snow is heavily aggregated, then
  reliable quantitative measurements of liquid
  equivalent, snow density, or snowfall rate are
  difficult at this time.
• If snow is nonaggregated or moderately
  aggregated, then robust estimates of ice water
  content can be made.
• Multiparameter (dual-pol, dual-wavelength)
  radar measurements provide the best promise for
  snow quantification.
• (Courtesy of Alexander Ryzhkov)

50
(Jay Hanna, NESDIS)
http//www.ssd.noaa.gov/PS/PCPN/ice-images.html
51
What do we need to do to forecast snow
better?Research

• Better understanding of precipitation
  processes, esp. in orography
• Climatologies of snowstorm soundings (Eric
  Ware, OU)
• Relationship between sounding structure and
  crystal type
• Relationship between crystal type and density
• Lack of understanding of cloud microphysical and
  aerosol processes
• Lack of understanding of electrical effects on
  microphysics
• Idealized microphysical simulations?

52
What do we need to do to forecast snow
better?Numerical Weather Prediction

• Improved microphysical parameterizations
• Models are very sensitive to cloud
  microphysical
• parameterizations, especially at high
  resolution (
• km) (Brian Colle and collaborators).
• Parameterization to predict snow depth
  explicitly
• Recognition that one parameterization does
  not fit all.
• Statistical prediction techniques
• Roebber et al. (2003) neural net will be
  tested by 11 groups this winter (NWS offices,
  HPC, TV station, Canadian Weather Centres)

53
Cloud Microphysics . . . The Ultimate Limitation?

• Steady progress on the synoptic and mesoscale
  dynamics of snowfall forecasting
• Microphysics, by contrast, has not been advancing
  as quickly, but there is an increasing
  recognition of its importance.
• Unobserved in-cloud quantities will ultimately
  limit our ability to forecast snowfall (e.g.,
  microphysics, electrical charges, vertical
  motion).
• Forecasters, researchers, and the public need to
  recognize these limitations, otherwise
  disappointment in snowfall forecasts will
  continue.

54
Acknowledgments
Dan Baumgardt (NWS, La Crosse, Wisconsin) Harold
Brooks (NSSL) John Cortinas (NSSL/CIMMS) Norihiko
Fukuta (University of Utah) Jay Hanna
(NOAA/NESDIS) Robert Houze (University of
Washington) Jack Kain (NSSL/CIMMS) David
Kingsmill (Desert Research Institute) David Novak
(NWS, Eastern Region SSD) Paul Roebber and Sara
Bruening (University of WisconsinMilwaukee) Al
exander Ryzhkov (NSSL/CIMMS) Jeff Waldstreicher
(NWS, Eastern Region HQ) Eric Ware (University
of Oklahoma) Melanie Wetzel (Desert Research
Institute)