Transcript Lesson 22

AOSC 200
Lesson 21
WEATHER FORECASTING
• FOLKLORE
–
–
–
–
Red sky at night, shepherd’s delight,
Red sky in morning, shepherd’s warning
When spiders’ webs in air do fly
The spell will soon be very dry
• PERSISTENCE
– The weather tomorrow will be the same as the weather today (two
times out of three)
• CLIMATOLOGY
– This takes persistence one step further
– The average weather say for a particular month is the same each year *
• ‘COLD in December – HOT in July’
– English saying:
• In July the Sun is hot,
Is it raining? No it’s not.
Fig. 13-1, p. 375
Climatology Forecast of a White Christmas
TREND AND ANALOG
• We know that persistence forecasts will eventually be wrong
because weather does change.
• A trend forecast assumes that the weather-causing patterns
are themselves unchanging in speed, size, intensity, and
direction of movement (this is called steady-state).
– For instance: we know that an approaching cyclone will bring rain
(weather does change) but assume that the amount of rain or its speed
will not change during all the path the cyclone will travel.
• The analog forecast also acknowledges that weather changes,
but unlike the trend method, it assumes that weather patterns
can evolve with time.
– The main assumption is that weather repeats itself.
– Therefore, this method “searches” for past weather patterns that are
similar (analog) to the current situation.
– In this sense, the future weather patterns “should” be similar to those
that happened in the past.
Trend forecast based
on the assumption that
a mid-latitude cyclone
moves up the East
coast unchanged.
Fig. 13-3, p. 378
The D-Day Forecast: June 1944
– Suitable weather for the invasion:
•
•
•
•
•
•
•
•
Initial invasion around sunrise
Initial invasion at low tide
Nearly clear skies
At least 3 miles of visibility
Close to full Moon
Relatively light winds
Non-stormy seas
Good conditions persisting for at least 36 hours, preferably for 4
days
– Three meteorology groups worked independently:
• Analog forecast
• Bergen Schools: air masses, cyclones and upper level patterns
• Waves forecast
Weather patterns leading up to D-day
The D-Day Forecast: June 1944
– First question: What are the odds, month-by-month, that the
weather required for the invasion would actually occur?
• May: 24-to-1
• June: 13-to-1
• July: 33-to-1
– However, the weather changed from a placid and calm May
to a very stormy June. A winter-like pattern not seen in the
Atlantic in June in past forty years!
– At the beginning the invasion was planned for June 5th but
postponed to the 6th due to the weather forecast. This
decision turned out to be correct!
NUMERICAL WEATHER
PREDICTION
• Step One: Weather Observations
Surface observations, Rocket and balloon
observations, Satellite observations
• Step Two: Data Assimilation
• Model grid and grid points
• Measurements do not cover all of the globe and
are not at set grid points
• The input data need to be interpolated,
smoothed and filtered. This process is called
Data Assimilation
Data Assimilation
Water vapor image
NUMERICAL WEATHER
PREDICTION
• Step Three: Forecast Model Integration
• The measured data (initial conditions) and the “primitive
equations” of the atmosphere are used to forecast what the
status of the atmosphere will be in the future. In order to get a
“good” (accurate and precise) forecast enormous
computational resources are needed
• Step Four: Tweaking and Broadcasting
– Current forecasts do not sample the atmosphere on a grid
size that picks local events or resolve small scale
phenomena
• Local forecasters use local knowledge and experience to tweak
the final forecast for the public
Fig. 13.9
Concept of a stretched-grid model
Fig. 13.13
Richardson’s Model Grid
Numerical Weather Prediction Models
•
•
•
•
Short-Range Forecast Models
US government uses two models for this purpose
ETA model – Run four times per day
Rapid Update Cycle (RUC) model– Run every three
days
• Forecast out to 48 hours
• Medium-Range Forecast Models
– Spectral-models
– Medium range Forecast (MRF) model
• Forecast out to 10 days
Numerical Weather Prediction Models
• Why Do Forecasts Still Go Wrong Today?
• Imperfect data
• Models cannot solve small scale phenomena:
parameterization*
• Chaos: The atmosphere could react very differently to
slightly different initial conditions (non-linear
system) – butterfly flapping its wings.
• Is there any solution?
• Ensemble forecast
• Vary initial conditions*
• Use different models
ETA 48 hr Prediction – 0Z Nov 20 2004
MRF 48 hr Prediction- 0Z Nov 20, 2004
Forecasting
• Let’s consider a car that travels at constant speed v
from point B towards point C
• We can use the equation
x = x0 + vt
(1)
to determine its location (the distance x) at a given
time t. x0 is the distance from point A to point B at
t=0
Forecasting
•  INITIAL CONDITION
• This equation comes from a MODEL or
idealization of reality.
• If for any reason x0 is NOT well known, or
there is an “error” in determining the exact
location of B, then the equation will give us
a different distance to point C
Forecasting
If we now ask the driver “to go straight” but we don’t give
him/her any point of reference (there is no road, trees or
anything to use as a reference), the final path could be not as
straight as the driver might think
Numerical integration takes one small step at a time to move
forward
Ensemble Forecast