Forecasting extreme rainfall Iand getting people to listen to watches or warnings Can be difficul t!! Wes Junker Wes Because…….
Dec 18, 2015
Forecasting extreme rainfall
Iand getting people to listen to watches or warnings
Can be difficult!!
Wes Junker
Wes
Because…….
Climatology of Heavy Rain Events in the United States fromHourly Precipitation Observations
HAROLD E. BROOKS AND DAVID J. STENSRUD
From
Frequency changes with seasons
5 inch an hour rainfall rates lasting an hour are very, very rare but……do happen
Inch an hour rainfall frequency.
THE AREA OBSERVED DECAYS LOGARITHMICALLY AS THRESHOLDS INCREASE
THE ACCURACY OF FORECASTS ALSO DECAYS LOGARITHMICALLY EXCEPT FOR THE VERY HIGHEST THRESHOLDS
Because higher amounts or thresholds are relatively rare, it is difficult getting a big enough sample to
calibrate forecasts using traditional statistical techniques. Verification of 24 hour QPF for various
thresholds
25 50 2575 100mm 50 75mm
From Charles E. Konrad II (2001)
The Most Extreme Precipitation Events over the Eastern United States from 1950 to1996: Considerations of Scale
500,000 km2 100,000 km2 2,500 km2
DJF MAM JJA SON
Precipitation with mesoscale convective systems is very difficult
to predict
Inside the red line, the probability of 1 mm is 100% but the probability of 3 inches is only a little above 10% in the blue area
Extreme rainfall is a moving target
• May occur in association with a synoptic scale event– In winter, 1” to 2 inches of rain over a basin
may cause flooding
• In summer, flash flooding may occur due to a mesoscale system or even occur from a single storm that remains quasistationary and has intense rainfall rates
Forecasting extreme rainfall events
• You need to understand which combination of ingredients that can lead to – High rainfall rates– For an extended period
• Most extreme rainfall events are associated with convection.– You need moisture, instability and lift.
• or by strong lifting due to orography
A various time ranges
• The methods of trying to predict extreme rainfall differ– At shorter time ranges
• Mesoscale analysis, satellite and radar imagery are often the best tools for pinpointing where an event is most likely and are used to monitor ongoing convective trends
– At longer time ranges, numerical models and ensemble forecast systems provide information about potential extreme events.
• THERE IS A CLEAR ASSOCIATION BETWEEN SHORT-WAVE TROUGHS AND CONVECTION
• THE VERTICAL MOTION ASSOCIATED WITH SYNOPTIC SCALE LIFT DOES NOT TYPICALLY ALLOW PARCELS TO REACH THE LEVEL OF FREE CONVECTION (LFC)
• HOWEVER, LARGE SCALE LIFT– STEEPENS LAPSE RATE– PROMOTES MOISTURE TRANSPORT– WEAKENS CAP– AFFECTS VERTICAL SHEAR (more important for severe
weather forecasting)
START BY LOOKING AT SYNOPTIC SCALE (THE BIG
PICTURE)
SHORT RANGE (0-3 HR) FORECASTS
• RELY PRIMARILY ON CURRENT OBSERVATIONS AND TRENDS– NEXRAD AND SATELLITE IMAGERY ARE GREAT
TOOLS PROVIDING INFORMATION ON THE , SIZE AND INTENSITY AND MOVEMENT OF PRECIPITATION SYSTEMS
– HAVE TO KNOW LIMITATIONS OF OBSERVING SYSTEMS
– STILL HAVE TO ANTICIPATE NON-LINEAR CHANGES
• NEW CELLS FORMING UPSTREAM
THE AMOUNT OF RAINFALL THAT FALLS OVER AN AREA DEPENDS ON
• SIZE and SHAPE OF THE RAINFALL AREA
• THE INTENSITY OF THE RAINFALL WITHIN IT
• HOW FAST THESE AREAS MOVE
• HOW FAST NEW RAIN BEARING CLOUDS ARE FORMING UPSTREAM (PROPAGATION)
Sound Easy?
A FEW IDEAS TO HELP DETERMINE HOW BIG AN AREA OF RAINFALL TO FORECAST
• THE SIZE IS DEPENDENT ON HOW MUCH MOISTURE IS PRESENT AND ON THE STRENGTH OF THE MOISTURE TRANSPORT – IS DEPENDENT ON BOTH THE ABSOLUTE (PWS, MIXING RATIOS),
RELATIVE MOISTURE (RH) AND WIND SPEED AND DIRECTION
• SIZE IS DEPENDENT ON THE SCALE OF THE FORCING– PATTERN RECOGNITION OFTEN PROVIDES CLUES BUT YOU
NEED TO UNDERSTAND WHY THE PATTERN IS FAVORABLE FOR HEAVY RAINFALL.
• MODEL GUIDANCE PROVIDES A DECENT FIRST GUESS, ESPECIALLY OF COOL SEASON STRATOFORM EVENTS– NOT SO GOOD WITH RAINFALL FROM CONVECTION
• An argument for using pattern recognition• And ensemble guidance to get a better feel about a systems potential.
Schematic representing the affect the shape and movement of a system has on the rainfall at a particular point. The shaded colors on the system represent the
radar echoes.
RA
INF
AL
L R
AT
E
RA
INF
AL
L R
AT
E
RA
INF
AL
L R
AT
E
TIME TIME TIMETIME
From Doswell et al., 1996 (Weather & Forecasting, 11, 560-581)
RA
INF
AL
L R
AT
E
You live at the blue dot Adopted from Doswell et al. 1996
Where new cells form in relation to the initial convection is important, to
• Determining the mode of development– Whether the system will look like this
– Determining the speed and direction the system will move
Or this
Or this
The mode of development and propagation is influenced by a
number of factors.
• OUTFLOW– EVAPORATIONAL COOLING RELATED TO THE
ENVIRONMENTAL HUMIDITY– GUST FRONT SPEED RELATED TO TEMPERATURE DEFICIT
BETWEEN OUTFLOW AND AIR AROUND IT.
• WHERE THE STRONGEST LOW-LEVEL CONVERGENCE, MOISTURE AND INSTABILITY IS LOCATED
• SYSTEM RELATIVE WINDS– DETERMINES WHERE LOW LEVEL CONVERGENCE WILL
BE LOCATED.
• INTERACTION OF UPDRAFT WITH ENVIRONMENTAL WIND
Rainfall rates
• Vertical moisture transport into the system which is dependent – on the amount of moisture that is available. Precipitable water,
dewpoints, winds, moisture flux– The magnitude of the vertical motions
• What produces the strongest vertical motion and rainfall rates? – Convection– High CAPE
• The proportion of the condensate that is fed into the cloud that reaches the ground (the precipitation efficiency)– Which is dependent on the shear– Relative humidity– An even shape of the sounding
What is precipitation efficiency
• The ratio of moisture that falls beneath the cloud to how much moisture is being fed into it.
• High precipitation efficiency is more likely when– Mean relative humidity is high
• Cloud bases are low
– Don’t want a lot of shear– More likely with warm rain processes– And airmass with maritime origins
• Because of having a more larger hygroscopic nucleii (more large salt particles.
• On radar look for high reflectivities below the zero wetbulb.
Madison County flash flood event at 1800 UTC
1800 UTC IAD sounding
Low centroid of max echo returns <55 DBZ
CAPE is low but positive
Relative humidity is high
PW is <2.00 inches
Not much shear, weak mean winds compared to low level winds
Low centroid mean it’s not hail but rain…..very heavy rainfall
A high precipitation efficiency event
MOVEMENT OF THE SYSTEM
• SLOW MOVING SYSTEMS ARE USUALLY THE HEAVIEST RAINFALL PRODUCERS
• AT SHORTER TIME RANGES-EXTRAPOLATION BASED ON RADAR AND SATELLITE PROVIDES PRIMARY GUIDANCE
• AT LONGER RANGES, MODELS CAN PROVIDE A FIRST GUESS, BUT– YOU STILL NEED TO TAKE INTO ACCOUNT MODEL
CHARACTERISTICS AND BIASES.
• AT ALL TIME RANGES, YOU MUST ANTICIPATE WHEN NEW ACTIVITY MAY FORM UPSTREAM (Propagation effects)
A slow moving or backbuilding MCS is more likely when
• When weak 850-300 mean winds are present.• When the low level jet is upstream and of the MCS
location and the low level jet is strong compared to the 850-300 mb mean wind.
• When you expect the MCS to be near the upper level ridge.
• When veering winds with height dominate speed shear
• When the strongest moisture transport and low-level convergence is located upstream from the MCS.
• When the airmass is unstable upstream but stable downstream
• When the mean RH is high.
Models often have problems handling the development and motion of MCSs
– Because of lack of data• This is especially true when dealing with
– The vertical and horizontal distribution of temperature and moisture
» Unfortunately, the vertical distribution of moisture and temperature governs where the airmass is unstable or not.
– Because certain physical processes occur below the scale that the NCEP operational models can resolve
• The following physical processes are therefore parameterized (and are handled in rather crude ways)
– Convection (in mesoscale models), within cloud microphysical processes, radiation, boundary layer processes
USE MODELS TO IDENTIFY SYNOPTIC AND MESOSCALE PATTERNS THAT ARE
FAVORABLE TO HEAVY RAINS
• CAN USE THE SURFACE, 850- AND 500-MB PATTERNS TO IDENTIFY MADDOX ET AL. OR OTHER TYPES OF HEAVY RAINFALL EVENTS – ALSO NEED TO LOOK CLOSELY AT MOISTURE, MOISTURE
TRANSPORT AND INSTABILITY
• MODELS OFTEN PROVIDE DECENT FORECASTS OF LOW-LEVEL WIND AND MOISTURE FIELDS– 850 MOISTURE TRANSPORT (MOISTURE FLUX)– PWS
• OUTPUT CAN BE USED TO ASSESS FORCING AND TO FORECAST THE LOCATION OF BOUNDARIES.
• Ensembles can provide information about the probability of the pattern actually occurring.
Use models to identify patterns associated with extreme rainfall.
Composite for East coast synoptic type
Borrowed from Rich Grumm
Lets look at a classic synoptic heavy rainfall event in the east
SREF has synoptic pattern right
Correctly predicts strong southern wind anomalies, PW anomalies of greater than 2
250wind & v-anom
LL
Bottom panels 850 winds and normalized anomalies
SREF again does better at recognizing pattern than forecasting QPF
SREF probability of 2.00 inches
Observed
Probability of 2.00 inches was 20% and in the wrong place
Stronger than normal low level jet….over 3 SD v-wind component at 850, Some uncertainty about placement of moisture plume (PW across threat area)
SREF trending towards high MF anomalies…greater than 5 SD. NAM (right) forecasting greater than 5 SD.
For extreme events is there a better way than trying to deterministically forecast a extreme event or forecast
the probability of an event at a single point
s fn
Forecast Actual Event
30 miles Reporting station
A perfectly predicted 125 mm area having a position error A perfectly predicted 125 mm area having a position error may be a terrible forecast. Or is it?may be a terrible forecast. Or is it?
Imagine a non-hydrostatic model forecasting 125 mm (5 inches) or more of rain (magenta). It verifies as the light blue area
Or is it? How Good Are these hypothetical ensemble forecasts?
How do you statistically forecast the probability of such a small scale event?
Your point probability forecasts of such an event will always be small
Ensemble members
observed
In this case, the probability of 125 mm (5 inches) would be zero based on the raw ensemble members but, all ten forecast a major event. Within the circle, 100% of the members forecast 5 inches
Even with 100% of the ensemble members forecasting an event,
what does it mean• Do the model members have an individual
bias for each member
• If you use probabilities based on ensembles, are they well calibrated.
• is there a resolution problem that may bias the members? – For the GEFS and SREF when dealing with
weak summertime forcing, probably yes
Forecasting extreme rainfall requires
• Looking at the synoptic, mesoscale and storm scale environments
• For any one location they are rare and often of small scale making it difficult to deterministically predict the exact location of the rainfall maximum
• We need to develop better ways to convey the potential for extreme rainfall