Potter T.D., Colman B.R. (co-chief editors). The handbook of weather, climate, and water: dynamics, climate physical meteorology, weather systems, and measurements
Подождите немного. Документ загружается.
tasks take on the order of 1 to 3 h. Forecasts for this time frame will have to be made
without the benefit of new model guidance, although output from a previous model
run may still be applicable. Another reason NWP is of limited use in the short-term
is that the initial conditions of numerical models do not yet include ongoing preci-
pitation processes. This is an area of research and development and progress is being
made, but, presently, there is a so-called spin-up period, typically 3 to 6 h, before a
modern numerical model can produce realistic precipitation. Perhaps the most
important reason that NWP is of limited use for short-term forecasting is that the
atmospheric phenomena of interest are often not explicitly resolved by current NWP
models. Predictions of whether an individual thunderstorm will produce large hail or
not and where such a storm is likely to track in the next hour are examples of short-
term forecasting for which current NWP offers little or no guidance, although
research efforts are ongoing in this area.
The short-term forecast process begins much like the longer-term forecast
process, the forecaster must understand the current state of the atmospher e. Analysis
is even more important for short-term, small-scale forecasting. Objective and manual
analyses are useful, and particular emphasis must be placed on careful examination
of individual observations. Because specific observations may be smoothed by
objective analysis schemes, or even eliminated by quality control algorithms,
manual analysis is usually considered a routine part of short-term forecasting, parti-
cularly during the warm season when gradients are weak and important features may
be subtle.
Most short-term forecasting deals with phenomena on the mesoscale. However,
as part of the analysis step, forecasters must consider scale interaction. Although it is
debatable whether planetary-scale considerations are important for shor t-term fore-
casting, there is no doubt that synoptic-scale conditions are critical to the evolution
of mesoscale phenomena. The location and evolution of synoptic-scale features such
as frontal boundaries, troughs and ridges aloft, regions of warm and cold tempera-
ture advection, high- and low-level jets, and the distribution of moisture in the
horizontal and vertical is necessary knowledge for any forecaster attempting to
predict mesoscale phenomena.
Because short-term forecasting does not have the benefit of primit ive equation
model output, it is absol utely critical that the forecaster understand the current
atmospheric processes and phenomena. A prediction of condit ions 1 to 3 h into
the future will generally fail if the current conditions are misdiagnosed. The fore-
caster must decide whether conditions are evolving linearly or via nonlinear
processes. An example of a linear forecast would be extrapolation. A line of rain
showers is moving east at 10 mph an d is expected to continue east with no change in
speed, direction, or intensity. Again, modern graphic=image workstations allow
tracking and extrapolation of features on radar and satellite imagery to make even
this simple extrapolation forecast more accurate. However, if the forecaster fails to
note that this line of rain showers is moving into an environment of greater instability
and favorable vertical wind shear, then the no nlinear transformation of the line of
showers into a severe squall line could be completely missed in the short-term
forecast.
686 MODERN WEATHER FORECASTING
Conceptual models play a very important role in short-term forecasting, perhaps
even more so than in longer-ter m situations because of the lack of NWP output. A
good example of applying a conceptual model to a short-term forecas t problem can
be illustrated for the prediction of supercell thunderstorms. Once thunderstorms have
started to develop over an area, the short-term forecast for that area can be very
different depending on the character of the convection. Ordinary thunderstorms will
have a life cycle of only an hour or so, they will generally not produce large hail or
tornadoes, and they will tend to move in a straight line. In contrast, supercell
thunderstorm life cycles can be many hours, they are likely to produce severe
weather, including large hail and possible tornadoes, and their movement will
likely be to the right or left of the movement of the ordinary cells. Clearly, the
short-term forecast will be quite different if supercell thunderstorms are expected
as opposed to ordinary thunderstorms.
Considerable research has shown that supercell thunderstorms develop in regions
with at least 20 m=s of vertical wind shear in the 0 to 6 km depth above ground, and
with adequate instability to support deep convection. The basic conceptual model of
supercells is that the combination of buoyancy and shear allow a thunderstorm
structure to become organized such that the updraft and downdraft do not interfere
with each other, as is typically the case in ordinary thunderstorms. The result is a
long-lived storm with very strong updraf ts and downdrafts, and pressure perturba-
tions that lead to rotation within the storm and propagation rather than just advection
of the thunderstorm.
A forecaster will first examine the environment to determine if supercell thunder-
storms are possible. Output from NWP can provide guidance about buoyancy and
shear. However, detailed analysis of direct and remotely sensed observations is
required to identify pertinent features such as old frontal zones, outflow boundaries,
gradients in surface characteristics, and diabatic heating gradients due to cloud cover
that may modulate the shear and instability on the mesoscale. Upon completion of
this analysis the forecaster will know whether supercell thunderstorms are possible
and also if some areas are more favorable than others for their development. The
second step in this process is to use Doppler radar data to carefully examine the
three-dimensional structure of thunderstorms as they develop. The forecaster will
look for features such as tilted reflectivity cores, weak-echo regions, bounded weak-
echo regions, rotation in the velocity data in the region of the storm updraft, and
anomalous storm movement. These are structures that have been documented in
research results and are the basis of current conceptual models of supercells.
Once identified, appropriate forecasts and warnings can be issued for these storms.
A similar forecast process is used for virtually all short-term forecasting that
attempts to predict the evolution of mesoscale phenomena. A precipitation band
in a cool-season situation can be caused by many different conditions. A short-
term forecast of the behavior of the band can only be made if the forecaster under-
stands the structure and processes responsible for its existence. The short-term
forecast will be very different depending upon whether the band is due to conditional
symmetric instability, or an internal gravity wave, or postfrontal lake effect, or
orographic lift, or low-level convergence.
4 FORECASTING 0 TO 12 h 687
Ensemble techniques may eventually play a role in the 0- to 12-h forecast proces s.
Often, the differences between environmental conditions that can support certain
mesoscale processes and those that cannot are quite subtle. An ensemble of models
may be able to offer forecasters insight into which situations are borderline and
which are not. Agai n, the value of the ensemble technique is the quantification of
uncertainty.
688 MODERN WEATHER FORECASTING
SECTION 5
MEASUREMENTS
Contributing Editor: Thomas J. Lockhart
CHAPTER 33
ATMOSPHERIC MEASUREMENTS
THOMAS J. LOCKHAR T
The technology applied to sensing of atmospheric conditions and content improves
constantly and at an ever-increasing rate as newly available technologies are applied
to old, familiar questions. The increasing capability and decreasing price of personal
computers is both a case in point and a cause of expanding technological applica-
tions. The best venues for describing and discussing advanced instrumentation are
the technical conferences and peer-reviewed journals. The thrust of this section will
be toward common applications and common concerns with respect to the repre-
sentativeness and uncertainty of measurements. The contributors to the measure-
ments section have practical experience dealing with such questions. The goal of this
chapter is to describe the problems, compromises, and pitfalls with measurement
programs of many kinds. This section will also include some examples of program
failures, which seldom reach the readers of technical journals.
In centuries past, instrument design, siting, and data application began with
individuals. Curious individuals used measurements to document their observations
and to help find explanations for what was happening. This led to a need to duplicate
instruments and introduced a commercial opportunity. As communication technol-
ogy developed, the ability to assemble synchronous observations became possible.
Synoptic meteorology generating forecasts and warnings followed. National weather
services were born and skills in measurement technology with the necessary set of
specifications became a specialty.
It is a fundamental principle that the application of data rests on an understanding
of the measurement process and the size of the uncertainty or error bars. It is
therefore incumbent on the organization taking the data to retain expertise in
these disciplines. Those with such expertise usually are thinking about improve-
ments and investigating new technology. This has led to the various national
Handbook of Weather, Climate, and Water: Dynamics, Climate, Physical Meteorology, Weather Systems,
and Measurements, Edited by Thomas D. Potter and Bradley R. Colman.
ISBN 0-471-21490-6 # 2003 John Wiley & Sons, Inc.
691
laboratories and research organizations formed to pursue technological and applica-
tion progress.
The atomic age introduced a need for radioactive fallout predictions. In recent
decades (since the 1960s) federal regulations have driven many of the applications of
special measurement programs in the atmospheric surface layer. First, in response to
the U.S. Department of Defense (DOD) Army chemical warfare research programs, a
need for better understanding of dispersion or diffusion proces ses was recognized.
The Nuclear Regulator y Commission (NRC) required accurate measurements and a
quality program to assure compliance with the requirements stated in the NRC
Safety Guide 23. A group of volunteers working within the auspices of the American
Nuclear Society (ANS) composed a standard (ANS 2.5) that defined the details
necessary to meet the ANS requirements for measurement accuracy and the methods
for qua lity assurance.
The U.S. Environmental Protection Agency (EPA) provided similar guidelines in
response to the needs of the 1967 Clean Air Act. The EPA requi rements were quite
similar to those of the NRC. Clearly, a vacuum of measurement standards existed.
No one could cite consensus standards to justify claims of data compliance. These
two lists of requirements, with the force of the federal government behind them,
brought the measurement community to action. It was immediately clear that stan-
dard definitions and test methods were necessary to determine if the requirements
were being met by the measurement systems used.
The American Meteorological Society’s (AMS) Committee o n Atmospheric
Measurements (CAM) considered the need to write standard methods to determine
the various characteristics that were being speci fied. After due consideration and
conversations between the AMS executive secretary and his counterpart in the
American Society for Testing and Materials (ASTM), it was decided to form a
subcommittee within ASTM D22, Sampling and Analysis of Atmospheres. Subcom-
mittee D22.11, Meteorology, was formed in 1972. The work of volunteer members
of D22.11 provided some of the early standards, but the process is open ended.
ASTM has a requirement of a 5-year review and reapproval process. New standards
are generated as the need, individual willingness to contribute, and expertise come
together. This national program is now contributing to a similar international
program through ISO (International Standards Organization) TC 146 (Technical
Committee 146, Air Quality) SC 5 (Subcommittee 5, Meteorology).
It is generally recognized that approaching perfection in the definitions and
performance of instruments is reducing the least important contributor to measure-
ment uncertainty. Take wind speed as an example. A perfect anemometer will only
suggest what the wind speed would be if the calibration facility were a true analog to
the turbulent environment. Its performance has been tested in the wind tunnel. Its
response to a speed change (sensitivity to inertia in a mechanical desig n) is defined
in terms of its distance constant, defined by a standard test using a speed step from
zero to the wind tunnel speed. The time const ant, 1 7 1=e, at that speed is measured
between the 30 and 74% points (of maximum value) during its recovery to avoid the
stall condition at release. The test is run at two or more speeds , which documents
that the response is constant with distance rather than time.
692 ATMOSPHERIC MEASUREMENTS
It has been recognized that the distance constant is different for a step change
from a higher speed to a lower speed than for a step change from lower to higher
speeds. The ratio of the two distance constants will predict the ‘‘overspeeding’’ of a
cup anemometer. This overspeeding is the inertial effect that causes the anemometer
to speed up faster than it slows down, creating an average that is slightly larger in
turbulent flow than in laminar flow. The word overspeeding has also been used to
describe the difference between the scalar average and the resultant vector average of
samples taken over a period of time.
There is further confusion when the effect of off-axis or nonhorizontal flow is
considered. Cup anemometers will usually turn a lit tle faster when the angle of the
flow is not quite horizontal. This is an aerodynamic or lift-drag effect that can be
simulated in the wind tunnel. When the distance constant model, based on the wind
tunnel test, is used to estimate the underreporting of average speed in turbulent flow,
it fails. The step change in speed does not descri be the speed changes in atmospheric
flow. While the model might predict a 20% error, the measurements in the atmo-
sphere show little or no effect.
There are effects, however, that do bias the measurements. Mounting hardware
and=or supporting structures may influence the speed. Nearby trees or buildings will
alter the flow. While the anemometer may faithfully report the speed where it is
mounted, it may not represent an area as large as the application might require. The
overall subject of representativeness will usually contain much more uncertainty
than the uncertainty in the transfer function (relationship between wind tunnel
speed and measured rate of rotation) of the anemometer. However, there is a general
rule of thumb that suggests one should not ignore those parts of the uncertainty
analysis that can be understood simply because there are other parts that are larger
and more difficult to define.
There are a broad variety of atmospheric variables that can be measured and an
equally broad variety of applications that need these measurements, each with their
own unique requi rements. The contributors to this section will touch on many of
these. It remains the responsibility of the person who uses atmospheric measure-
ments to understand, to whatever degree possible or practical, the process by which
the measurements were made. This knowledge should range from the engineering
specifications of the instrument to the siting biases that might contribute to
representativeness of samp les.
ATMOSPHERIC MEASUREMENTS 693
CHAPTER 34
CHALLENGES OF MEASUREMENTS
THOMAS J. LOCKHAR T
1 ETHICS
If there is one absolute in the ethics of measurement, it is that the data must speak for
themselves. But what can numbers say? If one applies standard statistical processes
to a series of numbers, the answer is reproducible. Of course, there is some control in
selecting a subset of a series of numbers (see ‘‘All That Is Labeled Data Is Not Gold’’
below). Even if a time series is continuous, the beginning point and the ending point
can influence the outcome. There is nothing unethical in presenting a continuous
subset of data and listing their statistical parameters, if the fact that it was a selected
subset is disclosed. When the parameters are used to make a political (or societal)
point, the ice is thinner.
A classic example of mixing true science with ‘‘political’’ science comes from
comments made by Stephen Schneider (1987) , a proponent of the theory that cholro-
fluorocarbons (CFCs) are depleting the ozone layer. He said: ‘‘We have to offer up
scary scenarios, make simplified, dramatic statements, and make little mention of
any doubts we may have. Each of us has to decide what the right balance is between
being effective and being honest.’’ Those of us who subscribe to the ethics of
measurement will have no problem with that decision. Science, including its
subset measurement, requires honesty. There is no type-A or type-B honesty. If
any consideration influences how numbers are gathered or generalized, that consid-
eration must be defendable without subjectivity.
The measurement community often expresses a genuine skepticism when consider-
ing the results of model simulations. The assumptions used for the model simplifica-
tions and for the ‘‘data’’ inputs deserve scrutiny. When simulations go beyond the data
in time, they become a forecast. There is real value in general circulation models, and
Handbook of Weather, Climate, and Water: Dynamics, Climate, Physical Meteorology, Weather Systems,
and Measurements, Edited by Thomas D. Potter and Bradley R. Colman.
ISBN 0-471-21490-6 # 2003 John Wiley & Sons, Inc.
695