452 Precipitation
Prediction of Clouds and
Precipitation often Poor
• Most precipitation is inherently mesoscale
and many models lack sufficient resolution.
• H20 vapor is poorly observed on both the
mesoscale and synoptic scale
• Microphysics problems, with other physics
problems (e.g., PBL) adding to the
difficulty.
Model Microphysics
Model Microphysics
• Many schemes use bulk parameterizations
that only prediction mixing ratio of varies
cloud/precipitation species (like cloud
water). Called single moment schemes
• Newer schemes are double moment,
predicting mixing ratio and number of each
precipitation/cloud type.
• Both single and double moment schemes
assume size distributions, often some kind
of exponential.
Droplet Distribution
Bin Schemes are the most
sophisticated
• Predicts the details of the size
distribution…but far too expensive.
• Some new schemes offer variable riming:
coating of ice crystals by water that freezes.
Precipitation Forecast Skill
• Only modest improvement during the past
30 years
• Skill declines with intensity
• Higher in winter when precipitation is more
synoptic and stratiform
• Less skillful in summer when more
convective
• Humans add less skill than for most other
parameters
Prob. Of Precip.– Cool Season
(0000/1200 UTC Cycles Combined)
0.7
Brier Score Improvement over Climate
0.6
Guid POPS 24 hr
Local POPS 24 hr
Guid POPS 48 hr
Local POPS 48 hr
0.5
0.4
0.3
0.2
0.1
0
1966
1969
1972
1975
1978
1981
1984
Year
1987
1990
1993
1996
1999
2002
Mechanisms producing saturation, clouds and
precipitation
• Dynamical forcing as expressed by QG
dynamics: broad frontal scale clouds.
• Topographic: upslope precip, mountain
waves clouds.
• Convection: produced by destablization of
the atmosphere.
Total precipitation is the product of all three
Annual
Precipitation
SW Olympic Slopes-Hoh Rain Forest: 150-170 inches yr-1
Sequim: roughly 15 inches per year.
.
Orographic
precipitation
enhancement
Annual
Precipitation
Orographic Enhancement of Precipitation
12
Convective
precipitation
Precipitation Rate
Precipitation Rate (m m /hr)
(mm/hr)
Surface
frontal
rainband
Post-frontal
convection
Upper-level
frontal band
10
Upslope site
(elev. 1067m)
Pre-frontal
precipitation
8
Valley site
Upslope site at 1067m
6
Valley site
4
2
0
-600
-400
-200
0
Distance (km
)
Distance
(km)
200
400
600
Rainshadows Shift with
Approaching Flow Directions
and Depend on Stability
• As the flow approaching a barrier changes
direction, so does the orientation of the rain
shadow.
• Less rainshadow during warm frontal period, more
post-frontal.
Rainshadow
Rainshadow in westerly flow
Coastal frictional convergence
• Heavier precipitation at coast as winds slow
Flow over terrain can be highly 3D and complex
Convective Precipitation
• Historically, subject approaches dominated
• Models lacked sufficient resolution to
model explicitly (2-4 km) but gave
guidance where convection might break out.
• Dependence on cumulus parameterizations,
which have serious issues.
Atmospheric Rivers
• Meridional flow of moisture is often limited
to relatively narrow currents of moisture
and usually warm temperatures.
Most West Coast heavy precip events are associated with
“atmospheric rivers”, a.k.a. the “Pineapple Express”
A relatively narrow current of warm, moist air from the
subtropics…often starting near or just north of Hawaii.
Associated with
extraordinarily narrow
filaments of moisture
Precipitable Water
From Mike Warner
A Recent Devastating Pineapple Express:
November 6-7, 2006
November 6-9,
2006
Dark Green: about 20 inches
We know quite a bit about
atmospheric rivers and heavy NW
precipitation events, although there
are still gaps in our knowledge
Synoptic Set-Up for Top Fifty
Events at Forks
Courtesy of Michael Warner
Precipitable Water
500 mb height
SLP
850 mb Temp
Extreme Precipitation Events
• The current of warm,
moist air associated with
atmospheric rivers are
found in the warm sector,
parallel, near, and in front
of the cold front.
• Thus, atmospheric rivers
are closely associated
with the jet core and the
region of large
baroclinicity.
Orographic Enhancement
• Upslope flow
greatly increases
precipitation rates
on terrain.
• Thus, wind speed
and angle of attach
can greatly modify
the extreme nature
of the precipitation.
Progress in Precipitation Prediction in Terrain
NGM,
80 km,
1995
NGM, 1995
2001: Eta Model, 22 km
Move to roughly 12 km by 2005
36-km
12-km
NWS WRF-NMM (12-km)
2007-2008
4-km MM5
Real-time
Smaller Scale Terrain Modulates
Precipitation
10-km
Small-Scale Spatial Gradients in Climatological Precipitation on the Olym
Alison M. Anders, Gerard H. Roe, Dale R. Durran, and Justin R. Minder
Journal of Hydrometeorology
Volume 8, Issue 5 (October 2007) pp. 1068–1081
Annual Climatologies of MM5 4km domain
Verification of Small-Scale
Orographic Effects
Convective Precipitation
• Need at least 2-4 km resolution to get most
convection even half right.
• If grid spacing is more coarse must have cumulus
parameterization
• Convection is least skillful precipitation in
virtually all models.
• High resolution models can potentially give a
heads on type of precipitation a day ahead (squall
line, supercell, etc.)
Terminology Review
• Explicit convection: model has enough
resolution to do it itself (less than 4 km grid
spacing)
• Parameterized convection: model
parameterization uses large scale variables
to determine precipitation and other impacts
of convection. Example: Kain-Fritsch,
Grell, Kuo.
Real-time WRF 4 km BAMEX Forecast
Valid 6/10/03 12Z
4 km BAMEX forecast
36 h Reflectivity
4 km BAMEX forecast
12 h Reflectivity
Composite
NEXRAD Radar
Real-time 12 h WRF Reflectivity Forecast
Valid 6/10/03 12Z
4 km BAMEX
forecast
10 km BAMEX
forecast
22 km CONUS
forecast
Composite
NEXRAD Radar
Hurricane Isabel Reflectivity at Landfall
18 Sep 2003 1700 Z
Radar Composite
41 h forecast from 4 km WRF
Predicting Clouds
• Zero-order approach: use relative humidity
• 70% 1000-500 mb RH often corresponds
with thicker, middle-level clouds that have a
serious impact on radiation.
• 700 and 850 mb RH is also used by some.
Direct Use of Model Clouds
• All modern models predict clouds,
specifically mixing ratios of cloud liquid
water and cloud ice.
• The quality varies…and keep in mind there
are serious deficiencies of even the best
microphysical schemes.
• And problems with other physical
parameterizations: boundary layer turbulence
or radiation can also mess up model clouds.
Direct Use of Model Clouds
• Most problematic: stratus, stratocumulus,
and FOG.
• Remember, some models have spin-up
issues: precipitation and clouds improve
during the first 3-9 hours. Particularly true
of UW WRF which is now cold-started (no
clouds at initialization!)
Cloud/Precip Forecasting
Strategy
• Very short-term (0-3 hr): temporal
extrapolation (informed by human
judgement) is HARD to beat. Nowcasting.
– Use radar and satellite animation. Models
generally not that useful.
– New data assimilation/modeling systems: e.g.,
EnKF, RUC/RR/HRRR are becoming viable
and eventually will take this on successfully.
Strategy
• Short-term (2-6 hr): Satellite extrapolation
becomes central. Models becomes more
dominant at the end. Particularly HRRR
• Daily (6-24 h): Satellite extrapolation and
model, weighing model more at longer
range.
• Remember model spin up issues.
http://ruc.noaa.gov/hrrr/
Strategy
• A wise forecaster ALWAYS evaluates
model’s initial moisture fields—often a
failure mode and a good indicator of
potential model failures.
• How? Compare initial model fields and RH
to satellite imagery.
850
700
Strategy
• Know regional precip/cloud climatology
– Diurnal and geographic features tend to be very
repeatable, particularly during the warm season
– Example: front range convection
Human Forecaster Issues
• Precipitation is a parameter where in
general forecasters add the least skill
compared to objective guidance.
• Psychological issues for high and low
precip probabilities
Human Versus
Objective Skill for
Precipitation
Forecasts: NWS
Offices Around the
US
Brier Scores for Precipitation for all stations for the entire
study period.
Brier Score for all stations, 1 August 2003 – 1 August
2004. 3-day smoothing is performed on the data.
Precipitation
Brier Score for all stations, 1 August 2003 – 1 August
2004, sorted by geographic region.
Reliability diagrams for period 1 (a), period 2 (b),
period 3 (c) and period 4 (d).
Rain Psychology
More than a day out….Human Forecasters
Tend to Overpredict 10-30%
--exaggerating the threat of rain when it is not
likely.
“Its probably not going to rain…but I will
throw in 10-30% anyway to cover myself”
And underpredict rain when it is fairly definite.
“Well it looks like it will rain…but I am unsure…so I will
knock down the probabilities to 60-80% to cover myself”
Humans Help with
Heavier Precipitation
The End
Bias scores for the (a) 1.33and (b) 4-km model
simulations for 1400 UTC 13
Dec 2001 through 0800
UTC 14 Dec 2001.
Puget Sound Convergence Zone