صفحه 1:
Snow to Liquid-Ratio:
Climatology and Forecast
Methodologies
Martin A. Baxter
Cooperative Institute for
Precipitation Systems
Saint Louis University.
Dept. of Earth and
Atmospheric Sciences
LSX WFO Winter Weather Workshop
7*November:2005
صفحه 2:
Forecasting Winter-Precipitation isa Two-
Step 12006
00 0 current dynamic and منصحموقمصوموط
forcings of the storm must be assessed.
-Numerical model forecasts must be studied, especially
the model quantitative precipitation forecast (QPF)..
* Second, the evolution of the hydrometeors from
. their origin to the surface must be 000:
1 -This evolution will be determined by the vertical eee
» of temperature and moisture.
-This profile.will elucidate the type-of precipitation-
rain, snow, freezing rain, ice pellets, 0۲ 27
combination.
» -If the precipitation is را ۱ ی
to liquid equivalent ratio must be determined to forecast
the actual snow amount.
صفحه 3:
ی ی وش
forecasters?
“After forecasting liquid equivalent (QPF), the'snow-
liquid equivalent ratio must be-estimated. ;
*Significant variations in snow to-liquid equivalent
ratio can occur even within’a single.storm system
*A more clear understanding of the processes that
act to vary snow density will enable the forecaster to
employ a more‘scientific process Oriented method
toward forecasting snowfall, versus.;commonly used
empirical techniques.
* A challénge exists to determine thé extent of
interaction between the dynamical forcing and the
microphysical processes that determine snow
density (i.e.,, how efficient is the forcing in producing
snowfall from a given amount of liquid equivalent’).
EA OE a SEI
2 Pape el Se 8 Aa ا 2 4
صفحه 4:
NAS
(Kyle and Wesley
1996)
“Utilizes surface
temperatures to
estimate snowfall
from: liquid
equivalent
*Is only marginally
effective, as it does
not account for
صفحه 5:
Description of Dataset and 006
+A 30 year (1971-2000) climatology of snow,
to liquid ratios was compiled using NWS
Cooperative Observer Summary of the Day
1۰
۱۵ greater than 2”-and liquid: °
-equivalentsgreater than 0.11” were
- included, as this was the standard for
۱۱۹۹ “له (2003).
_*Estimated.events were discarded.
*A station must have recorded at least 15
observations over the 30 year 9 6 ا
included.
صفحه 6:
ation distribution of the 30 year climatology
صفحه 7:
Average Snow to Liquid Ratios 1971-2000
صفحه 8:
Average Snow to Liquid Ratios (1971-2000) for October & November
صفحه 9:
Average Snow to Liquid Ratios (1971-2000) for December, January, February
صفحه 10:
=
2
ا
م
3
=
&
3
8
5
5
1
۳
2
5
=
=
2
3
3
1
?
2
صفحه 11:
Average SLR for each NWS County Warning
St. Louis, MO
‘Avg SLR:
Standard Dev:
75th Percentile:
‘0th Percentile:
25th Percentile:
http://jwww.eas:-slu,edu/CIPS/Research/
snowliquidrat.html
صفحه 12:
Histogram for the Entire Dataset of SLR
Mean:
- 13.53
(Short
dashed)
واه
Percentile:
9:26 -
Median:
- 12.14
(Long
dashed)
طا5 7 6 30 32 30 28 2428 22 20 18 18 14 1012 8 6 4 2
صفحه 13:
Sample SLR Climatological Distributions
°3674
Observations
4 تدعت ۰۸
12 ۷2۵ x
distribution
when
compared to
116
histogram for
the entire US
۰916010
high values
۲۵۵ 5-5
St. Louis, MO
Avg SLR: 12.0
Standard Dev: 6.0
۰
0-2 46 8 1042 14 16 18.2072 24 25 25 30 260
sir
صفحه 14:
oe
ar
0
en
هه ae oe
3
53
of
صفحه 15:
۱
feature higher snow to
liquid ratios; as they are
colder and contain less
moisture.
*This leads to.growth
by deposition.
*Storm tracks that are
warmer or contain more
Gulf moisture feature
lower snow to liquid
10
فد ال يا
southeastern Wisconsin
with various storm tracks
(Adapted from Harms,
1970 [ ۰۲5 16600 ما growth
۱
صفحه 16:
bullet SS ل« لين
555 solid colum:
eombinatio: dendrite
of needles 7
hollow colum:
crystal with
broad
branches
combination shea
of bullets th
(Pruppather and Klett 198
صفحه 17:
al habit depends.on temperature and degree of satur.
“GRAUPEL
ree DROPLET REGION
>
a
a
a
2
a
4
5
g
3
5
6
w
3
5
c
o
<
1
§
5
NEARLY EQUILIBRIUM REGION
Lice sarusarion 1€E SATURATION
0 5 “15 2000-25 -35 2-40 °C)
TEMPERATURE
(Magono and Lee, 1966)
صفحه 18:
*Supercooled water
droplets impact the
crystal as it falls
*If riming occurs late;
crystals retain original.
form.
If rimihg occurs early,
the droplet can provide a
nucleus for a new
crystal.
1
صفحه 19:
Different crystal types will have different amounts of
air in between them at the surface (More air =
higher'SLR) (Less.ai lower;SLR)
Stellar’ / Dendritic 91۵۱1
صفحه 20:
air Space in ع1
the cfystal itself’
Saul معط 5806+ :
PE Sots ا كر
صفحه 21:
Temperature Scanning Electron Microscope Photog:
Air space is reducedias a Air space is reducedieven
snowpack settles more,as snow melts
صفحه 22:
Billings, MT-ACARS Sounding
1 SLR Observed (crystals like thesdbjbbrecht 2004)
Descent sauncing trom elings/Logan, MT (BIL)
[lasting 27 mn, ana covert vautical mies (Arcraft #671
والخر مره تم
صفحه 23:
hysically-Based Method Using Glimatol
_*SLR is'determined largely by the. vertical
temperature-profile
*Thus; the 30-year average SLR is likely
associated with an average vertical
temperature profilé
*An SLR value that is higher-or lower than the
30-year average is presumably associated with
an anomalous vertical temperature profile that,
15 colder or warmer, respectively
*For this study, 850 mb anomalies were used,
with modifications made based upon the
temperature profile below this level and the
surface ات
صفحه 24:
Why must we use Climatology?
*No correlations between atmospheric
variables (temperature and humidity aloft and
at the*surface) and SLR have been established
*Climatology provides an “initial guess” that
can be'refined’by ete ge the details of the
ومناهناتو « ١
Where is the maximum vertical motion?
«What crystal types will form and how will
they evolve?
*Will'significant riming occur?
*How will the surface تست نت0 دوع the
fallen snow?
*This method shows-how a forecaster.can .
2 یت هت ualel the 1205 that.affect دا as
صفحه 25:
Novemb:
SD &‘LacrosseyWI marked (SLReon-bottomy# of reports
صفحه 26:
rams for 1200 UTE 22 November 1996 +1200 UTG-23. Nove
صفحه 27:
Column and Spatial Forts
een
ar
7162...
ماو
صفحه 28:
صفحه 29:
850 ۱۱۱ Anomalies’00 UTC 23’ November 199
4
eyed
cooler
تصقطا
average
*18 SLR
vs. 13-14
Avg SLR
for Fall
صفحه 30:
50 mb:Temperature Anomalies 12 UfTC.23 November 1996
0
colder than
ات
اكت نالفل
due to
یج
ground
temps
هس
warm a: سس سر
تت اضرا
صفحه 31:
ow do I find the 850.mb Temp Anomalie
*Rich Grumm’s (SOO, State College) ensemble &
anomaly page at Penn State University
*http://eyéwall:met.psu-edu/ensembles/java/
صفحه 32:
‘Methods for Forecasting SLR + Neural Net
+ Neural.Network ~ Paul و ةا (UW-MW), Be
Schultz (NSSL)
toe ا ا اا
(temp, RH, etc.) associated with oe values for
many/cases
+ ۱6 وگ ره Network is then able to Sir *
based upon the nonlinear relationships ۱
derived from the training data
¢“nonlinear” - an exponential relationship for
example - changes in a ‘given variable are‘not
associated with changes of equal proportion in
another variable
و امم *Crude treatment of vertical
likelihood alee ontemcl men act scout) نگ
exact number aiven.
صفحه 33:
۱ SLR
1. Solar radiation (month)
Low-to midlevel temperature (>>850 aie
Mid- to upper-level temperature (875-400 mb)
. Low- to midlevel relative humidity (> 850 mb)
ا لل ل ملفل ا
Upper-level relative humidity (700-500 mb)
External compaction
ی( یر
+ RH information useful, but ار رت رت
0 ی ی is to sort out the non-
linearity.
*. Behaves like a human brain’- takes\in new data
and matches it to’patterns it has “experience”
صفحه 34:
Verification. of the Roebber Algorithm
تا
network vs.
surface chart
% Snowfall
forecast error
(cm)
15.2 cm = 6”
4.0 بطم ۳
37 5
2004-05
ی
Using the 37 cases trom 2004.05, Horizontal
50th. 75th anc 90th percentile errors (numbers indicate the snowfall forecast
enor, in em)
صفحه 35:
SLR - Cobb Met ی ی
*Cobb Method - Dan Cobb (SOO, WFO CAR)
*Similar to a top-down approach to forecasting
*Uses vertical:motion information:
*Keys in on the Snow Production, Zone (SPZ),
where températures are -12 to-18 °C and the
Bergeron process is maximized
*Accounts for temporal evolution of SLR
» *Code is not complex anda ان product 126
“been created
*Beneficial for forecasting responsibilities at
both national and local/regional levels
sonal restate ۱ te teteite tie
صفحه 36:
The Cobb Method
1. Fiid max UVV in a cloudy layeri(RH with: —
respect to water >'75%)
2. Calculate.a weighting factor.to be applied to
all layers that méet criteria
* Although the concept is;physically sound, the
determination of the formulation of the
weighting factor is highly empirical (and very
important).
* The layer with the highest vertical motion ‘will
contribute the most to the, observed snow ratio
٠ It will determine the dominant crystal type
صفحه 37:
Snow Ratio as a Function of Source Layer
Temperature
3. Calculaté a
snow ratio
for each
model layer
based on
temperatur
6
* Curve'is generated via a’cubic spline through 6
data points that are based on observations of SLR
vs. crystal typé by Ivan Dube (MSC) and from my
climatology (this.curve is also “bumped up” to
~account for extreme events)
م a PR 4 Ea am SRE ABR, ie all ORR 4 ig هه
صفحه 38:
The Cobb Method
4. Use'this formula to calculate the weighted
contribution te the snow ratio for each layer:
5 که the 26516115 02 لله 61 6ك ةن مصعم قنتل 6 ٍ
layers to receive the predicted snow ratio
T#25C SRM =10 WWeSen's
THA8C SRT)=45 UW= tenis
صفحه 39:
۱0۹۹0
7-26 6 SR(T)=10 UW=6ems 0.4
C8 Oe tick kor Dorit = P%
T=A5C SR(T)=45 UW=8emls See
O00 tick ker Orr = 8% 45 x 36% =
197
وا 12 - ثالانا 6 - (58)1 1-56
26 ۳۷۲ ها +660
6 x 60% =
3:6
صفحه 40:
The Cobb Method.
Vertical motion max er iterate with SPZy
. High, م2 of high ratio snowfall 0
Vertical motion max ال
* Warmer temps, و
_. leading to riming, lower ratio snowfall
Vertical motion’ max above the SPZ:
* More difficult to discern resulting ratio,
crystals fall’ through many layers ABE و 11
different. types. 94 سای
Vertical motion is required to supply 0
water in lower levels - thus ماع عه تن of ee
- are implicitly included
eee ci tall ل ook aa اک دک وا
صفحه 41:
Verification ofthe Cobb Method
. ¢For the 2001- 2002 winter,season (Qctober-April)
+24 hour observations of SLR were paired with.
model vertical profiles where snowfall was
observed during the 24 hour périod
*Using the S:hourly, 32 km, Regional
تا تن data
*25 mb vertical resolution below 700 mb, 7
mb above
°3 hourly Cobb SLR’s were summed to SS a
+031 هه
Ne Be of error are considerable when doing
_ point based verification
*Bad SLR measurement
**Reanalysis still ای مه 0 اف
atmosphere
صفحه 42:
Mean Absolute Error for 128 Stations
Weighting Method #1
Mean = 6.1
Avg # of Reports: 3
Total of 401 Pairs
Mean = 5.3
for MAE <
15
o 2 4 6 @ m 2 4
Frequency
صفحه 43:
for 128 Stations 80۳0۲ و۱
Weighting Method #2
Mean = 7.9
Avg # of Reports: 3
Total of 404 Pairs
Mean = 6.7
for MAE <
| 15
۱ ۱ i | الا ۳ i سه سر
8
Frequency
صفحه 44:
ification of the Cobb Method - Jan 31 2€
Cobb
1 “oy
شما
True
AE iy
صفحه 45:
* Makes the “top-down method” quantitative via a
decision-tree styled qs
۰ ۷/۵۲80
algorithm by
Jessica Cox (McGill
اتيت 1ف
indicates the
method performs
equal to or better
than the 10:1
approximation 83%
of the time
| Nera Arg
16 oe me dee bacnt a bape oy
صفحه 46:
صفحه 47:
صفحه 48:
صفحه 49:
Snow to Liquid Ratio Overview
°A 30 year (1971-2000) ی ۷25 0ل یت (9
snow to liquid ratio .
** Average snow toliquid ratios are typically اون
than 10:1; more like 13:1 for much.of the country
°Certain storm tracks will exhibit varying snow.to
liquid ratios based upon in-cloud temperatures and
relative huntidities. Time of year and.external ,
compaction also play dominate roles.
. ‘We can attempt to extract out the effects of
“microphysical interactions by determining the type of
erystals that formed*and how these كفك grew and
changed.
. *In most cases it is possible.to eee 161116 5
of higher and lower SLR values & obtain a crude
estimate, through use of low-level temp, anomalies
*Two poten forecasting methods تفت being tested Finis 8 es