ATM
401/501 (Spring 2013): Daily Class
Summary
Thu 11 Apr
NOTE:
Exam #2 will be on Tuesday, 16 April.
The exam will cover quasi-geostrophic theory.
The in-class severe weather and QPF exercise
will begin on Thursday, 18 April.
- We resumed and
concluded our discussion of severe convective storm forecasting, following the
class slides linked below
- Slides 35–97
were discussed, covering:
- dry slots and cold fronts aloft
- CAPE/shear phase space and storm
mode/structure
- RKW theory, describing
buoyancy/shear interactions near cold pools
- typical large-scale environments
associated with derechos
- microbursts
- classic supercell
structure
- splitting supercells
- case study excerpts from Doswell and Bosart (2001)
Tue 9 Apr
- The focus of
class was to begin discussion of severe convective storm forecasting, following
the class slides (Brief Overview of Severe Storm
Forecasting)
- In the most
rudimentary sense, moist convection requires lift, instability, moisture, and
boundaries (LIMB)
- Slides 1–34
were discussed, illustrating:
- the basics of surface fronts and drylines
- low-level jets
- vertically coupled front/jet
circulations
- capping inversions and elevated
mixed layers
- characteristic sounding types
associated with convective storms
Thu 4 Apr
- We continued
with our discussion of the Q-vector
form of the QG ω-equation following the class notes posted here and here (email Jaymes if you need the ID and password).
- For
frictionless, adiabatic, QG flow, the Q-vector
form of the ω-equation states that vertical motion is forced by the divergence
of the Q-vector.
- In the
simplest sense, in order for a Q-vector
to be nonzero, we must have (1) a horizontal temperature gradient, and (2)
horizontal non-uniformities in the geostrophic wind field. These horizontal non-uniformities in the
geostrophic wind are expressed as relative vorticity and/or horizontal
deformation.
- Note that the Q-vector destroys thermal wind balance
(see eqs. 13–16 in the notes). In doing so, a secondary circulation arises.
This 3-D secondary circulation can be regarded as having vertical
branches (i.e., vertical motions) and horizontal branches (i.e., ageostrophic
motions). Both the vertical and
horizontal branches are linked by mass continuity. As a whole, this secondary circulation acts to
restore thermal wind balance, to be discussed shortly.
- Where a
component of the Q-vector is directed
towards warm (cold) air, frontogenesis (frontolysis) by the geostrophic wind is
occurring. Frontogenesis (frontolysis)
is associated with a thermally direct (indirect) circulation. Note that the rising (sinking) branches of
these cross-frontal circulations are found where Q-vectors are convergent (divergent).
- According to
the scaled horizontal momentum equations (Holton eqs.
2.24 and 2.25), horizontal accelerations (e.g., at jet entrance and exit
regions) imply the existence of ageostrophic flow (i.e., departure from
geostrophic balance). Consequently,
thermal wind balance is destroyed in these regions.
- Consider an
idealized, straight, E-W jet streak atop a uniform N-S temperature gradient. At the jet entrance region, the horizontal
temperature gradient is too large for the vertical shear. Re-establishing thermal wind balance in this
region requires (1) decreasing the horizontal temperature gradient and (2)
increasing the vertical shear. A
secondary circulation accomplishes both!
- Note that the Q-vector is directed towards warm air in
the jet entrance region, implying geostrophic frontogenesis and a thermally
direct secondary circulation. The vertical branches of this circulation
involve rising warm air and sinking cold air, thereby serving to weaken the
horizontal temperature gradient, which satisfies requirement (1) above. The horizontal
branches of this secondary circulation involve a southerly ageostrophic
wind at jet level and a northerly ageostrophic wind in the low levels. Owing to the Coriolis force, this
ageostrophic flow yields a westerly
acceleration at jet level and an easterly
acceleration at low levels.
Therefore, the vertical wind shear is increased, which satisfies requirement
(2) above. In this manner, the secondary
circulation acts to restore thermal wind balance at the jet entrance
region. A secondary circulation must
also exist at the jet exit region, but will be thermally indirect.
- The foregoing
interpretation of jet circulations is consistent with the four-quadrant model
of idealized jet streaks. Thus, the
circulations associated with jet streaks are predicted by QG theory (and the QG
ω-equation).
Tue 2 Apr
- We began class
by interpreting QG vertical motion from both Eulerian (i.e., fixed-location) and Lagrangian (i.e., flow-following) viewpoints.
- We can
undertake a simple thought experiment:
consider an idealized midlatitude shortwave trough in adiabatic, stably
stratified, QG flow. The flow is also equivalent
barotropic (i.e., the height contours and isotherms at a given level are
parallel), with temperature decreasing in the same direction as geopotential
height. Note that equivalent barotropic flow implies
that horizontal temperature advection is zero everywhere. Furthermore, we will assume that in the upper
troposphere, air parcels overtake the trough axis (i.e., flow through it). In the lower troposphere, air parcels are overtaken
by the trough axis. The Coriolis
parameter f is constant everywhere.
Eulerian perspective: Consider an observer atop a very tall tower
located downstream of the trough axis.
As the trough axis approaches, the observer notes a decrease in the
temperature. According to the QG
thermodynamic equation (Holton Eq. 6.13), adiabatic cooling via upward vertical motion is the only process that
can account for this local cooling, given the conditions stated above. Therefore, upward (downward) vertical motion
is expected ahead of (behind) the trough axis.
Lagrangian perspective: Consider an observer in the upper
troposphere, initially at the trough axis, who is then free to move with the flow. As the observer proceeds downstream following
a particular air parcel, the observer notes a decrease in the relative
vorticity of the parcel. According to
the QG vorticity equation (Holton Eq. 6.18), divergence (of the ageostrophic wind) is the only process that can
account for this vorticity tendency following the flow, given the conditions
stated above. Therefore, in the upper
troposphere, divergence (convergence) is expected ahead of (behind) the trough
axis.
Meanwhile, in the lower troposphere,
consider another observer, initially downstream of the trough axis, who is also
free to move with the flow. At this
level, the observer is overtaken by the trough axis, and observes an increase
in the relative vorticity as the trough approaches. Again, according to the QG vorticity
equation, convergence is the only
process that can account for this Lagrangian vorticity tendency. Therefore, in the lower troposphere,
convergence (divergence) is expected ahead of (behind) the trough axis.
Note that this distribution of
divergence and convergence implies upward (downward) vertical motion ahead of
(behind) the trough axis. Furthermore, these
vertical motions are wholly consistent with the Eulerian perspective discussed
above!
- We proceeded
with a derivation of the Q-vector
form of the QG ω-equation. The Q-vector was first derived by Hoskins et al. (1978). See
also Sanders and Hoskins (1990). Note
a key conceptual duality: the Q-vector does to the ω-equation what
QGPV does to the height tendency equation.
These two equations, formulated in terms of the Q-vector and QGPV (respectively), are each Galilean invariant and
contain single forcing functions (i.e., are not prone to cancellation between
terms).
Thu 28 Mar
- The main objective
of class was to review several forms of the QG ω-equation
- For adiabatic,
frictionless, quasi-geostrophic flow, the traditional form of the QG ω-equation
includes two RHS terms (or forcing functions):
Term A: differential geostrophic absolute vorticity
advection by the geostrophic wind
Term B: the horizontal Laplacian of temperature advection
by the geostrophic wind
This equation is discussed in Tom Galarneau’s overview.
- Both terms in
the traditional ω-equation can be can be qualitatively evaluated on synoptic
charts. However, there are notable
drawbacks to this form of the equation:
(1)
Terms A and B are not independent.
Note that temperature and geostrophic wind can be written in terms of
geopotential (or geopotential height), so that an instantaneous 3-D
distribution of geopotential prescribes both terms (A and B). In other words, if one term is known, then so
is the other.
(2)
Terms A and B (the forcing functions) often oppose each other. In these situations, qualitative
assessment of the total forcing for ω is difficult.
(3)
Assessment of term A requires information from multiple isobaric levels
(i.e., we must evaluate the vertical derivative of geostrophic absolute
vorticity advection).
(4)
Terms A and B are not Galilean invariant, as demonstrated in the class notes.
Note, however, that the sum of terms A and B (i.e., the total forcing) is
Galilean invariant.
- We therefore
seek to recast the QG ω-equation in a manner that avoids these pitfalls.
- The Sutcliffe-Trenberth
form of the ω-equation is once such approach, as derived in the class notes and
discussed by Tom Galarneau (see links above).
This equation states that (to a good approximation) ω is forced by a
single forcing function: the advection of geostrophic relative vorticity by the
thermal wind. See Trenberth (1978) for the original presentation of this
result. As with classic Sutcliffe
Development Theory, note again the significance of the thermal wind in this
equation.
Tue 26 Mar
- The main
objective of class was to introduce quasi-geostrophic potential vorticity
(QGPV) and its applications. QGPV is
derived here (as posted on the class website); see also Holton, 4th
edition, sections 6.3.2, 6.3.3, and 6.3.4.
- To motivate
development of QGPV, we begin by reexamining the traditional form of the QG
height tendency equation (Holton eq. 6.23, which also appears in the derivation
linked above).
- Note that in
the traditional form of the QG height tendency equation, height tendency is
forced by (1) geostrophic absolute vorticity advection by the geostrophic wind,
and (2) differential thickness (or temperature) advection by the geostrophic
wind. These forcing functions appear as
separate terms on the RHS of the equation.
- At a vorticity
maximum or minimum, geostrophic absolute vorticity advection must vanish. Thus, geostrophic absolute vorticity
advection cannot amplify troughs or ridges, but can only move them along. The amplification mechanism for troughs and
ridges is differential thickness (temperature) advection.
- Conceptually,
the two terms in the traditional form of the QG height tendency equation are
straightforward. However, in practice,
there may be significant cancellation between them. A cursory inspection of synoptic charts will
often identify regions where such cancellation is likely occurring.
- QGPV
alleviates this cancellation problem in the height tendency equation. After deriving QGPV, observe that the height
tendency equation may be rewritten in terms of QGPV. See Tom Galarneau’s
discussion here (his section 2).
- According to the
QGPV form of the QG height tendency equation, height tendency is forced solely by the advection of QGPV by the
geostrophic wind. Thus, recasting the
height tendency equation in terms of QGPV eliminates the cancellation problem,
since a single forcing function appears.
- Note that QGPV
is simply the sum of geostrophic absolute vorticity and stability. QGPV is a linearized form of Ertel PV. For adiabatic conditions, QGPV is conserved
following the geostrophic flow on isobaric surfaces (Holton, p. 160).
Thu 7 Mar and
Thu 14 Mar:
NOTE:
Project #1 (macroclimatology) is due on 2 Apr. Over the next several days, please email a
brief synopsis of your project (~1 paragraph) to both Lance and Jaymes.
- During these
class periods, we completed the derivation of the Sutcliffe Development
Equation and discussed its application/interpretation
- The Sutcliffe
Development Equation is strictly a diagnostic expression for 1000-hPa
convergence. However, given the
relationship between convergence, vorticity tendency, and pressure tendency, we
can use this diagnostic expression to make prognostic inferences.
- Note that in
all three terms, the thermal wind serves as the advecting
flow. This underscores the fundamental
importance of thermal wind balance in understanding midlatitude dynamics.
- Term A may be
regarded as a steering term. This term
cannot contribute to surface development, and therefore can only move systems
along. Note again that the thermal wind
effectively serves as a steering current for surface cyclones and anticyclones.
- Term B
contributes to both development and steering.
Note that assessing term B requires knowledge of the thickness field
alone.
- Term C
contributes to both development and steering, but is generally small compared
to terms A and B.
- We also
discussed Laplacian operators and how they may be interpreted:
- The Laplacian
of a scalar is simply the divergence of the gradient of that scalar. Recall that the gradient is a vector field.
- Many fields
(e.g., geostrophic relative vorticity) can be written in terms of Laplacians of other fields (e.g., geostrophic relative
vorticity is proportional to the Laplacian of geopotential height). Using geostrophic relative vorticity as an
example, this relationship essentially states that the height field directly
prescribes the geostrophic relative vorticity field. In other words, if we have a height field,
and wish to know the geostrophic relative vorticity field, then there is no
need to compute the geostrophic wind as an intermediate step.
- Remembering
that gradients are vectors (having direction and magnitude), we can qualitatively
evaluate the divergence of gradients to understand Laplacian operators.
- In the
simplest sense, Laplacian operators provide a measure of the distortion of the
flow. In the case of geostrophic
relative vorticity, we know that straight, uniformly spaced isohypses yield
zero geostrophic relative vorticity. In
such a case, the Laplacian of geopotential height is zero.
Tue 5 Mar:
NOTE:
Exam #1 will be on Tuesday, 12 March.
The exam will cover forecast verification and ensemble forecasting. The material presented today will not be on
Exam #1.
- The focus of
class was to discuss the fundamentals of Sutcliffe Development Theory
- Driven by
wartime necessity, Sutcliffe sought to find a physical basis for forecasting
the intensification/weakening and movement of surface cyclones and anticyclones
- Aircraft
limitations of the era meant that upper-tropospheric measurements were not
available to forecasters
- Further,
Sutcliffe sought to work with spatial derivatives only, which could be
qualitatively evaluated on synoptic charts for a single time (i.e., no time
derivatives)
- Given these
constraints, Sutcliffe appealed to both dynamical and thermodynamical aspects
of the problem, and made key simplifications to the governing equations
- Sutcliffe can
be credited for inventing QG theory
- Derivation of
the Sutcliffe Development Equation is available on the class website, and will
be completed in class on Thursday
Thu 28 Feb:
- Concluded our
discussion of ensemble forecasting topics
- Skilled use of
ensemble guidance requires linking “signals” in the ensemble with inferred synoptic
and mesoscale features
- For example, high MSLP spread
north of a mean E-W trough suggests high uncertainty in the progression of a
warm front
- Often,
ensemble mean and spread charts are best used in conjunction with spaghetti
charts:
- Where spread is high, spaghetti
charts will help identify whether members tend to cluster around several
outcomes, or are uniformly dispersed
- Visualizing
ensemble QPF with mean and spread charts is not always useful, especially when
the ensemble spread approaches the ensemble mean (often in cases of convective
precipitation).
- Consider QPF decile charts instead
- In addition to
simple mean and spread charts, normalized spread is often available
- For every grid point and lead time,
a running mean (e.g., 30 days) of ensemble spread is calculated
- The spread for the current
forecast is then divided by the running mean, thereby normalizing the spread to
account for lead time and geography
- Normalized spread above (below) 1
indicates more (less) uncertainty than the recent average
- See the NCEP ensemble training page and the MetEd COMET module, both discussed in class, for a good
review of ensemble forecasting
Tue 26 Feb:
- Continued with
discussion of ensemble forecasting
- Plotting
ensemble mean with ensemble spread highlights the areas of largest
forecast uncertainty
- Large spread
in the trough access is mostly related to amplitude uncertainty
- Large spread
in between troughs and ridges is related to phase uncertainty
- Decile plots
show where "X" percent (e.g. 60%) of ensemble members are forecasting
at least "Y" inches/mm (e.g. 0.25") of precipitation
- Probability
plots illustrate what percentage of ensemble members are forecasting a certain
precipitation threshold
- Plume diagrams
display individual ensemble member forecasts of a particular variable (at a
given location) as a function of forecast time
- Precipitation
type maps indicate the probability of frozen vs. liquid precipitation
- Note that we
should not overlook the individual ensemble members (especially in
the case of large uncertainty)
Thu 21 Feb:
- Continued with
discussion of ensemble forecasting
- A
well-constructed ensemble will occupy a large portion of the actual phase space
(i.e., the ensemble will be suitably dispersive)
- Despite having
a coarser resolution, the ensemble mean Briar score (forecast skill) exceeds
the control run Briar score at all forecast times (on average)
- Ensemble mean
offers approximately one additional day of forecast skill
- Features
associated with large uncertainty (low predictability) will not be well
represented in the ensemble mean. Instead,
the ensemble mean retains the synoptic and global features common to many
members, often at lead times of ~6 days.
- Exercise
caution when using an ensemble forecast whose members tend to cluster. This may indicate an underdispersive
ensemble.
- Exercise
caution when the ensemble spread exceeds the ensemble mean (especially for
precipitation forecasts)
Tue 19 Feb:
- The focus of
class was on ensemble forecasting concepts
- Discussed
probability density functions (PDFs) and their relevance to ensemble
forecasting
- A PDF describes the likelihood of
observing a certain value (or range of values) for a given field/variable
- You can create PDFs for many
variables (temp, dewpoint, precip, wind speed, etc.)
- Note that some variables are not
normally distributed in the real world (e.g., precipitation amount)
- A PDF should include the entire range
of plausible values (i.e., span the phase space)
- An underdispersive PDF has an
artificially small phase space
- An overdispersive PDF has an
artificially large phase space
- What are the implied
consequences if one solution (or an ensemble of solutions)
cannot capture the entire phase space
(atmospheric extremes)?
-Discussed
probabilistic forecasting vs. deterministic forecasting
- Deterministic forecasts rely on
a single model solution
- Ensemble forecasts are derived from
multiple model solutions
- Ensemble forecasting allows you
to perform statistical analyses of PDFs for a given field (mean, spread,
range, etc.)
- Each ensemble member should be equally
likely to verify; however, members from the same parent model tend to cluster
near each other
- When viewing multiple ensemble
solutions, it is important to acknowledge differences in both phase and
amplitude (e.g., as illustrated by a 500 hPa height spaghetti diagram)
- Linear climate
change involves a shift in the mean (PDF shape is retained)
- Non-linear
climate change involves a shift in both the mean and variance (PDF
shape changes)
Thu 14 Feb:
- Discussed
forecast performance of the National Hurricane Center (NHC), Storm Prediction
Center (SPC), and Hydrometeorological Prediction Center (HPC).
- NHC: official storm track errors have improved
markedly since 1970, especially at longer time ranges. Forecasting storm intensity remains a
challenge.
- SPC: many recent severe weather outbreaks have
been anticipated several days in advance, which was rare decades ago. Moderate- and high-risk events are often
highlighted with remarkable skill at days 2 and 3.
- HPC: human forecasters continue to add skill to
NWP precipitation forecasts, as measured by threat and bias scores. Skill tends to be lowest during the summertime,
when precipitation is commonly convective in nature. In 2012, HPC forecasters outperformed MOS in
maximum temperature forecasts, but sometimes struggled with minimum temperature
forecasts.
Tue 12 Feb:
- Discussed the
Ocean Prediction Center, which is responsible for maritime weather forecasting
- Major concern
is the forecasting of extratropical cyclones producing hurricane force winds
- High
frequencies of hurricane force winds overlap with major transoceanic shipping
routes
- Implementation
of QuickSCAT greatly improved the detection of
hurricane force winds, but this instrument is no longer operational
- Many hurricane
force wind events go undetected, especially over the Pacific
- Diabatic
processes contribute substantially to the explosive deepening of oceanic bomb
cyclones
- Intense
extratropical cyclones are more common in the SH, where mean meridional
baroclinicity is pronounced and expansive
Tue 7 Feb:
- The frequency distributions of GFS and ECMWF
ACs for 500 hPa height show that NWP forecast skill has improved remarkably
over the past 15 years. Greatest
improvement has occurred in the SH.
- The same conclusion can be reached by
considering the percentage of excellent (AC > 0.9) and poor (AC < 0.7)
forecasts for both hemispheres.
- Root-mean-square error (RMSE) of wind speed
in the tropics shows that the GFS is systematically fast compared to the ECMWF.
- The ECMWF exhibited superior performance in
its handling of Sandy, especially at forecast ranges beyond 72 hours
Tue 5 Feb:
- GFS anomaly correlation has increased
considerably over the last 20–25 years, but still lags ECMWF (due primarily to
differences in data assimilation)
- Annual mean GFS die-off curves for 500 hPa height AC suggest that useful forecast
skill (AC = 0.6) is typically lost at day-8 in the NH, and day-7.6 in the SH.
- The CDAS (Climate Data Assimilation System) is
frozen; any increase in CDAS forecast
skill is related to improved data assimilation and initialization
Thu 31 Jan:
- Discussed how terrain influences cyclone
movement along the West Coast and the near the Appalachian Mountains.
- Completed presentation on weather and climate
prediction (by A. Thorpe):
- Sources of predictability (and unpredictability) are found across a
broad range of time scales
- Anomaly correlation (AC, or ACC) is a common metric of NWP
forecast skill. Forecasts verifying with
ACs > 0.6 are considered useful.
- Advances in NWP capability have yielded a
steady increase in ACs over the past three decades for medium range (~
3–10-day) forecasts.
- Note especially the shrinking forecast skill
gap between northern and southern hemispheres, largely attributable to improved
satellite data assimilation practices.
- A fairly
comprehensive description of various forecast skill
metrics is available
Tue 29 Jan:
- Clarified the mechanism by which a barrier
jet can form as a result of terrain-blocked flow. This is often observed during Appalachian
cold air damming episodes.
- We used simple arguments to deduce that
downslope (upslope) flow contributes to cyclonic (anticyclonic) vorticity
tendency at the surface. This
relationship explains (1) the development and maintenance of lee troughs and
windward ridges, and (2) the synoptic rule-of-thumb that surface cyclones and
anticyclones near mountain ranges tend to move with higher terrain on their
right (in the NH).
Thu 24 Jan:
- Student-led map discussions will begin on Thu
31 Jan
- The six questions that map discussions should
address are posted on the course website (under Map Discussion). Prof. Bosart explains the motivation behind
this approach in his 2003 paper, Whither
the Weather Analysis and Forecasting Process?, available here.