Transcript nested grid predictive model

Mapping and predicting the weather
The Storm of the Century, a winter storm which struck the entire
eastern half of the US on March 12, 1993.
Mapping and predicting the
weather: a history
• Meteorology involves geography and
cartography
• Prediction is synonymous with mapping:
predicting weather is best done spatially, on
maps.
• To arrive at where we are today with weather
forecasting required innovations in how
atmospheric data was collected,
communicated, mapped, and analyzed.
Weather proverbs and folklore
were the first means for
predicting and understanding
the weather.
http://www.americanfolklore.net/folktales/rain-lore.html
http://en.wikipedia.org/wiki/Weather_lore
Red sky at morning, sailor take warning.
Red sky at night, sailor’s delight.
Visibility of stars can indicate
increasing amounts of
moisture in the troposphere,
the lowermost layer of the
atmosphere where weather
takes place. This has been
known to many cultures and
civilizations.
Andean cultures used the brightness of constellations to predict the weather. Bright stars
indicated more moisture in the troposphere (El Nino conditions) which could several months
later lead to dry conditions (La Nina conditions). This knowledge allowed Andeans to adjust
potato planting dates in anticipation of oscillating climatic conditions.
Weather proverbs and folklore also have a long
history in the prediction and comprehension of the
weather, although their accuracy can be debated.
• The 189-year-old publication
“The Farmers Almanac” claims
80% to 85% accuracy for the
forecasts by its reclusive
prognosticator, Caleb
Weatherbee.
• Weatherbee prepares the
forecasts two years in advance
using a secret formula based
on sunspots, the position of the
planets, and tides.
• Does not meet scientific
standards of weather
prediction.
Weather logs were the first formalization of what would eventually become
weather maps. Prior to late 1700's, weather data was listed in weather logs
rather than spatially analyzed with maps. Integrating weather logs among places
and meteorologists was thwarted by a systematic way to collect and distribute
data quickly.
First weather maps published
in 1811 by Brandes.
Because it took a lot of time
to assemble and plot the
data, he could not predict the
weather; only hindcast it.
Brandes hindcasted
conditions in Europe for each
day of the year in 1783. He
was the first to contour and
plot isobars, the lines on an
air pressure map. He plotted
air pressure values observed
across the continent and then
drew in the contours by hand.
The invention of the
telegraph aided in the rapid
dissemination of weather
data. Now weather
observations and data could
move faster from place to
place. Warnings of severe
weather could be
communicated along
telegraph wires. However,
knowledge of how large
weather systems travelled
was poorly developed.
Sketches from Morse’s notebooks showing early design of the telegraph
(1840’s)
Development of a rapid network of telegraph stations eventually evolved to warn Atlantic
cities of storms developing to the west, in the Midwest and South, shortly after 1846.
Weather gauge ( instrumental) information could be telegraphed to Washington D.C. where
maps were constructed. By 1860 had 45 stations reaching as far west as St. Louis
Signal Corps telegraph office network, (War Department, post-Civil War)
The network of telegraph stations increased under the US Weather
Bureau administration, helped by the completion of the transcontinental
railroad.
US Weather Bureau forecast headquarters, Washington
DC, early 1900’s
Production and standardization of the first national weather maps in 1880’s
Weather Bureau transferred to Dept of Agriculture and network of local and regional
forecast stations constructed. In other words, the office in Washington DC was
decentralized. Telegraph was still main means of communicating weather data. However,
weather over ocean was an unknown….
Institutional evolution of weather
prediction
• US Weather Bureau - US War Department (1870)
• US Weather Bureau - US Department of Agriculture
(1891)
• US Weather Bureau - US Department of Commerce
(1940)
• The US Weather Bureau was renamed the National
Weather Service (1970) and placed under the newly
created National Ocean and Atmospheric
Administration (NOAA) within the US Department of
Commerce
Development of wireless radio
technology, early 1900’s. Wireless
radio communications was pioneered
during WWI and used to transmit
weather conditions and forecasts to
and from ships at sea. Airplanes are
still particularly critical today for making
precise measurements of hurricane
conditions and positions.
View of hurricane eye from airplane.
But what was going on up high in the atmosphere? Initial upper
atmospheric measurements initially made with kites…..note tornado in
background (and the likely presence of lightning). Not a safe work
environment.
Bi-plane with weather instruments, 1934. Planes were expensive to operate
for weather data collection, and safety concerns certainly remained.
Balloons could be launched with people
and weather instruments on board, or
better, safer, and less expensive….
..by tracking unmanned balloons,
meteorologists could gain an idea of how
wind speeds and directions change moving
up through the troposphere.
Image at right shows an airplane
dropping a hurricane warning
notice to a crew of sponge divers
of Florida. Today, airplanes drop
dropsondes, which record the
weather conditions in its fall to the
surface. These data are
transmitted back to the forecast
office. Without dropsondes our
knowledge and prediction of
hurricane track and intensity would
be greatly diminished.
A radiosonde is a balloon released from the surface with an attached weather recording and
transmitting instrument. A dropsonde is a weather recording and transmitting instrument
dropped from a plane. The instrument shown above is what records and transmits weather info
on a radiosonde. It is attached underneath the balloon.
Radiosondes, dropsondes, and later, satellites, led to the mapping of the upper
atmosphere. Shown below is the location of the polar jet stream. Colors
indicate wind speed. Altitude of the polar jet stream ranges from 10-15 miles.
This information could not be collected without these technological innovations.
Development of polar front theory(1922) , which described mid-latitude cyclones, a
major weather maker. Pioneering work by Bjerknes (left) contributed to our
conceptual and mathematical understanding of weather prediction.
Midlatitude cyclone (left and above) encompass low
pressure center and attached warm and cold fronts. Cold
front is shown in blue. Warm front in red. Entire system
rotates counterclockwise
Integration of upper-level conditions, surface conditions, and polar front theory are
essential to forecasting the weather. The low pressure on the surface is a midlatitude cyclone. Note how the winds in it are coupled to upper level flow.
Skew-t plot
for Tallahassee
Red line shows
air temp and blue
line shows dew
point temp
y-axis: temp
y-axis: pressure level
At what pressure level
are clouds likely to
form ?
Invention of the skew-t plot from existing graphical methods (late 1940’s)
Finite difference technique of
forecasting (Richardson, 1922)
Grid cells used by Richardson to map his prediction of air motion
determine weather patterns. Richardson’s work pioneered a method
for mathematically and cartographically predicting the weather.
3D block of the atmosphere showing
complexity of flow
To understand the finite difference
technique, you have to understand
some basic principles about weather
prediction.
At its simplest, weather prediction is predicated
upon anticipating the motion of the atmosphere.
Region of
sinking air: clear
conditions and
high pressure
Region of rising
air: cloudy
conditions and
low pressure
Various air flow geometries of atmospheric
motion are associated with certain weather
conditions.
Rising air is associated with cloudy conditions
and low pressure. Sinking air is associated with
clear conditions and high pressure.
If you can predict the relative motion of the
atmosphere, moving up or down, a few days in
advance, you can predict the likelihood of
cloudy or clear conditions.
Based on these descriptions, what is the
general motion of air in Montana and Idaho?
(Rising). What is the general motion of air over
the mid Atlantic states? (Sinking)
Steps in the finite difference technique:
1. Collect data from
weather stations. In this
case we will collect air
temperature.
2. Contour data so that
temperature an be
estimated for all points
on the map
3. Draw a grid over the entire forecast area and assign a discrete value of
temperature to the center of each grid cell. The value of temperature is estimated
from the contour map.
4. Repeat this same process of contouring, gridding, and assigning
a value to a point at the center of each grid cell for other weather variables,
like air pressure.
5. Apply equations of fluid motion and
thermodynamics to predict the motion
of air to and from grid cells based
on their temperature, air pressure, and
other weather variables (dew point,
wind speed and direction)
These calculations will reveal general
movement of air and thus the regions
where there are rising (potentially
cloudy) and sinking (potentially clear)
atmospheric motions
Richardson’s method is the basis for 3dimensional numerical weather
prediction models used today,
After WWII, surplus radar donated for weather forecasting
Early radar image of thunderstorms in Florida
One of the first radar images of hurricane structure, 1946
This was state of the art radar-based hurricane tracking in 1960
Now the biggest hurdle to making weather maps and forecasts
was computational. The calculations were too tedious to do by hand. ENIACone of the first general purpose computers (1940) used to make forecasts.
Vacuum-tube technology (eventually replaced by transistor
and integrated circuits----the “chip”) was used in ENIAC
and early computers. Unreliable and prone to overheat.
Computer punch card---eventually replaced by magnetic tape,
then floppy discs, zip discs, compact discs, and dvd’s.
IBM 7090 (1965) used in weather forecasting
The “chip” was developed in late 50’s and 60’s
Refinement of integrated silicon-chip based computing, last
half of 20th century.
Launch of first weather satellite,
TIROS 1.
Launch of the first weather satellite
Real time satellite tracking
Further development of quantitative
models of weather prediction
• New models possible because of faster
computers, better data integration and
communication
• Predictive model types
– Global model
– Nested grid model
– Ensemble forecast
A global predictive model applies the finite difference technique over
a large area, with very large grid cell sizes. This allows
meteorologists to predict the weather by looking “upstream” to see
what kinds of conditions are coming their way. Remember, weather
systems tend to flow from west to east. One of the weaknesses of
this model is that it cannot make local weather predictions: the grid
cells are too large. It is also very computationally intensive.
A 24-hour global forecast system (GFS) prediction of potential precipitation. The map shows the
predicted weather for 8 pm Thursday night Eastern Standard Time (which is midnight, or 0Z Friday
on the Prime Meridian). I downloaded this map Thursday morning, March 13th. As this is a 24 hour
prediction, the model was run Wednesday night, 8 pm Eastern Standard Time. (Also see this link:
http://www.wunderground.com/modelmaps/maps.asp?model=GFS&domain=TA)
A nested grid predictive model also uses the finite difference
technique, but it combines the large scale global grid with a smaller
grid with much finer grid cells. This allows forecasters to see and
predict the big picture, while also accounting for local conditions. The
smaller grid cells in the nested grid allow more detail for local
forecasts.
A 24-hour nested grid predictive model (NGM) showing areas of potential precipitation. This is the
prediction for 8 am Friday Eastern Standard Time (which is noon, or 12Z Thursday on the Prime
Meridian). I downloaded this map Thursday morning, March 13th. As this is a 24 hour prediction,
the model was run Thursday morning 8 am Eastern Standard Time.
Hurricane forecasters look at the way models agree and disagree with each other. Each of the
lines below is predicted track from different computer models (NGM and GFS models are two of
them). By looking at how the models agree and disagree a measure of how much certainty in the
track can be estimated. If models disagree to a large extent, there is a greater sensitivity to local
conditions . Small differences in initial conditions input into the models can lead to vastly different
forecasts The result is a much harder to predict hurricane.
Which hurricane is more sensitive to the local conditions surrounding it? In other words, which
storm, when weather variables are fed into the forecast models, is more dependent upon small
differences in them? The white envelope indicates the potential 1-3 day area in which the eye of the
hurricane may pass.
(The atmospheric and oceanic conditions with Ivan (left) are much more difficult to predict based on
the shape of the white cone. Very localized variations in air temperature, ocean temperature, or
winds around the storm could cause Ivan to shift tracks over a very wide area. With Noel, any local
variability in air and ocean temperatures or winds are unlikely to have any effect on the outcome of
the storm track. You could enter a range of initial conditions in the models for Noel (right) and the
models would likely agree with each other.
Predicting hurricane intensity is more difficult than predicting its track. This is because the
conditions that determine intensity exist on much smaller scales and are difficulty to measure within
and around the hurricane. These parameters are a challenge to integrate into track forecast
models.
Hurricane Charley, 2004
The detection of “hot tower” thunderstorms
in hurricanes are thought to herald an
increase in strengthening (an example of
remote sensing)
NEXRAD Doppler radar coverage for the
US, another geographic-meteorologic
innovation
Outside of one of the first Doppler radar units, late 1970’s
Inside of one of the first Doppler radar units, late 1970’s
Doppler radar coverage out
of the Tallahassee National
Weather Service Office
Interface for selecting Doppler radar
reflectivity and velocity from across the US:
http://www.rap.ucar.edu/weather/radar/
Doppler radar can estimate
two parameters:
Reflectivity - a measure of
the size of the particles in the
air, which can not only tell
you the type of precipitation
(rain, sleet, snow, freezing
rain) but also its relative
intensity.
Rotational velocity - the
speed and direction of the air
relative to the radar station
Tallahassee National
Weather Service Doppler
To understand how Doppler
radar works….
…..you have to understand the
Doppler effect.
The sound of a horn from a moving
train will sound different for an
observer in front of the train
compared to someone the train has
already passed.
This is the Doppler effect applied to
sound. The wavelengths of sound
are compressed for the person out
in front of the moving train.
Doppler radar bounces
microwave radiation off the
atmosphere and the clouds
within it.
By looking at how the return
wavelengths have changed,
Doppler can detect three aspects
of the weather:
1.
The type of precipitation,
(rain, sleet, snow, freezing
rain) because their
differences in size, shape,
and composition will bounce
back a different signal.
2.
Potential intensity of
preciptitation. For example,
light to heavy rain will have a
different signal.
3.
The rotational movement of
the wind in terms of its
direction to (inbound) or
away (outbound) from the
radar station and at what
speed.
Doppler reflectivities showing
precipitation type (top) and rainfall
intensities (bottom).
These dBZ values equate to
approximate rainfall rates
indicated in the table right.
These are hourly rainfall
rates only and are not the
actual amounts of rain a
location receives. The total
amount of rain received
varies with intensity changes
in a storm as well as the
storm's motion over the
ground.
Doppler rotational velocities
With Doppler radar, the rotational
component of winds can be detected.
This is particularly useful for detecting
tornadoes. The pattern of a tornado in
Doppler is called a hook echo.
Doppler can detect winds that are moving
toward (inbound) the radar and away from
(outbound) the radar. It can also provide
the windspeeds.
Hook echo shown in white box. A hook echo is indicative of tornadic
circulation. When a hook echo is detected a tornado warning is announced
to the public.
In a hook echo, adjacent pixels will have sharply different directions, typically
with one inbound and one outbound. This permits the detection of rotation in a
tornado. Dopper is also useful for detecting the rotational component in the
eye of a hurricane.
Doppler is not perfect. Some parts of the
US are not covered completely by Doppler
radar. However, Doppler radar coverage
for most of the lower 48 overlaps.
The area right around the Doppler is the
cone of silence. Weather phenomena in
this area may be hard to detect.
At greater distances from the radar, the
beam may overshoot severe weather and
miss potentially dangerous conditions
closer to the ground.
Development of Internet 1960’s - today
• Emergence of the Internet in the
early 1990’s revolutionized the
dissemination and analysis of
weather data.
• Weather stations could be
connected in real time, allowing
instantaneous data feeds, and
rapid dissemination of forecasts
and warnings.
National Weather Service offices, including Tallahassee
Map of the Internet, 2007
NWS data and forecasting network
Surface observation map
Each symbol is called a
station model and
communicates information
about wind direction/speed,
temp, air pressure, dew point,
and cloud cover
Post-1990 to present
• Faster computer processor speeds, larger hard drive
storage capacities, and rapid data integration through
the Internet encouraged the development of:
– AWIPS (Advanced Weather Interactive Processing
System)
– Cartographic animation
– Global climate models
– Mesoscale forecasting models
– Integration of dynamical and statistical models.
– Privatization of weather forecasting
AWIPS
Integrated weather monitoring
and forecasting system
comprised of:
Network of Doppler radar
Automated weather
observation stations
Data stream of satellite
imagery and numerical
forecast models
Animation of atmospheric
conditions used to visualize and
predict tropical storm tracks and
intensity
Animations of global weather
patterns.
Global climate models (GCM’s) are being used to predict global
and hemispheric changes in temperature and precipitation decades
into the future.
Mesoscale forecasting models
Mesoscale
numerical models
can forecast the
weather at smaller
spatial
and temporal
scales.
http://www.mmm.u
car.edu/mm5/
In a statistical model, the past is used
to forecast the future.
For example, with hurricanes, the
records of when and where hurricanes
formed in the past provides the raw data
about how strong they became as they
tracked across the ocean.
This historic data can be used to assign
probabilities that a present-day tropical
storm, developing near the same
location at the same time of year, would
have characteristics (wind speeds, air
pressure in the eye) of the older tropical
storms.
Statistical models are combined with
dynamical (grid-based models) to make
a joint forecast of hurricane track and
intensity.
Private forecasting
Weather “Wars”
Sometimes called TV radar wars or Doppler wars, are a
kind of sensationalist journalism primarily concerning
weather news. The "war" is typified by competing local
TV news stations engaging in technological oneupmanship to increase viewership.
http://en.wikipedia.org/wiki/Weather_wars