3416 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

A Linear Stochastic Model of a GCM’s Midlatitude Storm Tracks

YUNQING ZHANG
Program in Atmospheric and Oceanic Sciences, Princeton University, Princeton, New Jersey

ISAAC M. HELD
NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey

(Manuscript received 26 February 1998, in final form 6 October 1998)

ABSTRACT
A linear stochastic model is used to simulate the midlatitude storm tracks produced by an atmospheric GCM.
A series of six perpetual insolation/SST GCM experiments are first performed for each month. These experiments
capture the ‘‘midwinter suppression’’ of the Pacific storm track in a particularly clean way. The stochastic model
is constructed by linearizing the GCM about its January climatology and finding damping and stirring parameters
that best reproduce that model’s eddy statistics. The model is tested by examining its ability to simulate other
GCM integrations when the basic state is changed to the mean flow of those models, while keeping the stirring
and damping unchanged.
The stochastic model shows an impressive ability to simulate a variety of eddy statistics. It captures the
midwinter suppression of the Pacific storm track qualitatively and is also capable of simulating storm track
responses to El Niño. The model results are sensitive to the manner in which the model is stirred. Best results
for eddy variances and fluxes are obtained by stirring the temperature and vorticity at low levels. However, a
better simulation of the spatial structure of the dominant wave train as defined by covariance maps is obtained
by stirring the temperature equation only, and at all levels.

1. Introduction lence (Leith 1971), but because of the complexity of

the inhomogeneous meteorological problem, one tries
Understanding the distribution of eddy variances and
to proceed with much simpler expressions for the stir-
fluxes due to baroclinic eddies remains one of the central
ring and effective damping than is used in the turbulence
concerns of dynamical meteorology. It is clear that lin-
ear theory has much to tell us about the midlatitude literature. Farrell and Ioannou (1993a,b; 1994; 1995)
storm tracks, as indicated, for example, by the distri- have analyzed a variety of idealized flows, while Whita-
bution of the ‘‘local Eady growth rate,’’ presented in ker and Sardeshmukh (1997) have recently obtained
Hoskins and Valdes (1990). Accepting the hypothesis very promising results with a two-layer balanced model.
that the dynamics linearized about the time-mean flow While studying the seasonal cycle of the Northern
is of value, how does one best make use of it to develop Hemisphere storm tracks, particularly the ‘‘midwinter
a theory for storm track structure? Eigenmode analysis suppression’’ of the Pacific storm track (Nakamura
has been a common starting point (e.g., Frederiksen 1992), we have been attracted to the stochastic modeling
1983), but it is not clear how to convert information approach and have found it to be valuable. We first use
about growing modal structures into time-averaged sta- a full primitive equation general circulation model in a
tistics. series of perpetual insolation/perpetual SST experi-
Farrell and Ioannou (1996a,b) have pioneered an al- ments, which simulate the midwinter suppression phe-
ternative approach in which a statistically steady state nomenon in a clear form. We then use a linearized, dry
is modeled directly by assuming that nonlinearity can version of this GCM, linearized about the time-mean
be replaced by two effects on the linear dynamics: flows produced by these GCM experiments, to construct
damping and stochastic stirring. The resulting model our stochastic models.
resembles stochastic theories of homogeneous turbu- In section 2 the perpetual insolation/SST GCM and
the climates it generates are described. The construction
and tuning of the linear stochastic GCM is described in
Corresponding author address: Dr. Isaac M. Held, NOAA/GFDL,
section 3. The linear stochastic model’s performance in
Princeton University, P.O. Box 308, Princeton, NJ 08542. reproducing nonlinear eddy statistics is examined in sec-
E-mail: [email protected] tion 4. In section 5 we concentrate on the model’s sen-

q 1999 American Meteorological Society

 1 OCTOBER 1999
ZHANG AND HELD

FIG. 1. The rms of bandpass-filtered geopotential height at 205 mb from the perpetual GCM. Contour interval: 15 m. Values larger than 60 m are shaded.
3417
 3418
JOURNAL OF THE ATMOSPHERIC SCIENCES

FIG. 2. Time-mean zonal wind at 205 mb from the perpetual GCM. Contour interval: 15 m s 21 .
VOLUME 56
1 OCTOBER 1999 ZHANG AND HELD 3419

FIG. 3. The horizontal structures of the rms of the geopotential height at 350 mb from the linear
model (a) and from the perpetual Jan run (b). Contour interval 10 m.

sitivity to the basic-state changes associated with the scheme dependent only on relative humidity. Climate
seasonal cycle and interannual variability. The results statistics from a seasonal integration of this model have
are summarized and discussed in section 6. been documented by Alexander and Scott (1996) and
are available at http://www.cdc.noaa.gov/gfdl/.
For our study, the model was forced with insolation,
2. Perpetual insolation/SST GCM
sea surface temperature, sea ice, and ozone concentra-
Nakamura (1992) has drawn attention to the fact that tions that are fixed at values corresponding to the middle
the Pacific storm track attains its maximum strength in of each individual month. For each month from October
spring and autumn although the jet stream and the ver- to March, an integration of 1200 days was completed
tical wind shear over the Pacific are at maximum and the statistics of the last 1000 days are analyzed in
strength in midwinter. The midwinter suppression of the this study.
Pacific storm track is present in the seasonal GCM de- Figure 1 depicts the rms of the bandpass filtered (2–
veloped by the Climate Dynamics Group at the Geo- 6-day period) geopotential height field at 205 mb from
physical Fluid Dynamics Laboratory (GFDL) and in the perpetual insolation model and Fig. 2 shows the
other models (Christoph et al. 1997). Rather than study time-mean zonal wind at the same level. Comparing the
this seasonally varying model, we have performed a storm tracks in this model to those in observations, such
series of GCM integrations with fixed insolation and as in Nakamura’s (1992) analysis, we find that mid-
fixed SSTs so as to isolate this phenomenon in a simpler, winter suppression is even more pronounced in this per-
statistically stationary framework. petual insolation GCM, with a reduction of 50% in var-
The GCM employed is that developed by the Climate iance in midwinter, as opposed to about 30% in obser-
Dynamics Group at the GFDL. It has a global domain vations. Comparison with the seasonal GCM referred
with realistic orography and land–sea contrast, with to above (figures not shown) indicates that the perpetual
rhomboidal 30 truncation and 14 unequally spaced sig- model also has a stronger suppression than the more
ma levels. The physical packages in the model have realistic seasonal model. Therefore, this perpetual in-
been left unaltered. These include moist convective ad- solation format evidently captures this phenomenon in
justment, gravity wave drag, a ground hydrology that a somewhat cleaner form.
includes formation of continental snow cover and a In discussions of storm track structure based on geo-
bucket for water storage, and a simple cloud prediction potential variance, it is necessary to filter out low-fre-
3420 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

structure in the seasonal cycle. Because of the small

changes of SST from, say, November to January this
should not be surprising. We presume that the Pacific
storm track is primarily responding to changing inso-
lation over Eurasia.
Our goal is to determine if the storm track variation
can be understood more directly as a response to chang-
es in the time-mean flow. Having a theory for eddy
fluxes given the mean flow, one could then hope to
construct a theory for how the mean flow itself evolves
due to the changing insolation.

3. Linear stochastic dry GCM

a. Constructing the model
Our starting point is the dynamic core of a 14-level
primitive equation model with all forcing/dissipation
terms removed. The model is discretized vertically in
the same manner as the full GCM and is horizontally
truncated at T30.
We start by linearizing this dynamical core around a
certain climate generated by the GCM. Rayleigh friction
and Newtonian cooling are then added to crudely model
the boundary layer processes in the GCM. Letting V9
and T9 represent horizontal velocity and temperature
perturbations, respectively, these damping terms have
the form
FIG. 4. The distribution of zonally averaged transient eddy heat
dV9
fluxes from the linear stochastic model (a) and the R30 perpetual Jan 5 · · · 2 a f (s)V9 (1)
GCM (b). Units: m s21 K. Vertical axis: s coordinates. dt
dT9
5 · · · 2 b f (s)T9 (2)
quency variability, as in Fig. 1 and in Nakamura (1992). dt
The filtering is a potential source of confusion, as there
 s 2 0.7
could be shifts in the frequency of the eddies with sea-  , 0.7 # s # 1.0
son. With this in mind we have examined other measures f (s) 5 1.0 2 0.7 (3)
of eddy activity that are more strongly dominated by 
0, otherwise,
high frequencies and do not require a filter to isolate
the storm tracks. In particular, we find that the midwinter where a represents boundary layer friction and b rep-
suppression is present in the meridional velocity vari- resents boundary layer thermal damping. Both are con-
ance field without filtering [see Zhang (1997) for these stants and their values are determined when the model
results]. We continue to use filtered geopotential here is tuned. A biharmonic damping is also added to absorb
for easier comparison with previous work. the enstrophy of small scales. This damping coefficient
When we compare these GCM simulations with the is chosen such that the e-folding time for the smallest
observations, several differences are also apparent. The wave is 0.5 days, which is somewhat stronger damping
seasonal cycle as a whole is shifted about one month than that used in the full GCM. There is no explicit
earlier since the heat capacity of the atmosphere creates topography in the model; however, it is implicitly pres-
roughly this much delay in the response to seasonally ent through the field of mean surface pressure. In this
varying forcing. In Fig. 1, the maximum variance occurs s-coordinate model one can show that topography af-
in October and March, rather than in November and fects the linear dynamics only through its effect on the
April as in observations. The seasonal cycle of the At- distribution of mean surface pressure. Finally, because
lantic storm track is not simulated well, as the obser- our model uses s as a vertical coordinate, the basic-
vations have a better-defined maximum in midwinter. It state variables have steep gradients in the Tibetan Pla-
is as if this perpetual model is on the verge of sup- teau region where the s levels are strongly sloped.
pressing eddies in midwinter in the Atlantic as well. Small-scale numerical instabilities result from the steep
Additional experiments in which we fix SST at their gradient. While we do not understand them fully, we
January value but allow the insolation to change show are able to eliminate these small-scale numerical insta-
that the insolation change accounts for nearly all of this bilities by increasing the strength of the lower-level tem-
1 OCTOBER 1999 ZHANG AND HELD 3421

FIG. 5. The distribution of transient eddy heat fluxes at s 5 0.865 from the linear stochastic
model (a) and the R30 perpetual Jan GCM (b). Units: m s21 K.

perature and velocity damping in the region over the has freedom in choosing the spatial structure of the var-
Tibetan Plateau. In this region the low-level damping iance of the noise and the relative strength of the forcing
coefficients a and b are both increased by a constant of different model variables. Although we have exper-
amount, 1.25 day21 . imented with several alternatives, we focus here pri-
An overall uniform linear damping and white noise marily on a model in which we include noise only in
forcing are then added and the final linear stochastic the lower troposphere (s . 0.76) and between 208 and
model has the form 608N. We force only the temperature and vorticity equa-
tions.
dX
5 (L 2 g I )X 1 e. (4) Rather than try to obtain the eddy statistics by direct
dt matrix inversion, given the large matrices involved we
Here X is the vector of all model variables, and L is have chosen to integrate the stochastic model forward
the linear operator of the model including the dissipation in time with the same semi-implicit time step as in the
and damping terms described above; D 5 2g I is the full GCM:
additional linear damping meant to represent the non-
1 2
X i11 2 X i21 X i11 1 X i21
linear scrambling and I is the identity matrix. The forc- 5 L E Xi 1 L I 2gX i21 1 e i .
ing matrix e is white in time and in all three spatial 2d t 2
coordinates, but even with this simplification one still (5)
3422 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

the 1000-day-averaged climate of the perpetual January

GCM. The linear model is run for 500 days and the last
400 days are used to construct the eddy statistics.
Although we have not systematically searched the
entire parameter space, we have found a reasonably
good fit to the geopotential variance and poleward eddy
heat and momentum fluxes with a 5 0.83 day21 , b 5
1.67 day21 , g 5 0.063 day21 , e V 5 8.1 3 10211 s22 ,
and e T 5 4.1 3 1026 K s21 . The sensitivity of the model-
produced second-moment statistics to these parameters
will be discussed later in the paper.
Figure 3 shows the storm tracks at 350 mb from the
perpetual January GCM and from the linear model.
Since the linear stochastic model is meant to model the
high-frequency eddies in the perpetual GCM, the non-
linear time series are bandpass filtered when the eddy
statistics are calculated. When tuned in this way, the
stochastic model provides relatively little low-frequency
variability and requires no time filter. The linear model
performs very well in simulating the two oceanic storm
tracks. The position and relative magnitude of the Pa-
cific and the Atlantic storm tracks and the northeastward
tilt of the Atlantic storm track are all captured by this
linear model.
Eddy heat and momentum fluxes are of central im-
portance in the maintenance of climate and its low-
frequency anomalies. How well the linear model can
reproduce these flux terms will serve as an important
FIG. 6. The distribution of zonally averaged transient eddy mo-
mentum fluxes for the linear stochastic model (a) and the R30 per- indicator of the model’s performance. It is found that
petual Jan GCM (b). Units: m 2 s22 . Vertical axis: s coordinates. with this set of parameters the eddy heat and momentum
fluxes are also reproduced quite well. As in the calcu-
lation of geopotential height variances, the GCM eddy
Here L E and L I represent the terms treated explicitly and fields have been bandpass (2–6 day) filtered to retain
implicitly, respectively. A time filter is used each time the information on baroclinic eddies.
step to relax X i to the mean of X i11 1 X i21 , just as in Figure 4 depicts the vertical structure of the zonally
the GCM. The noise is given by e 5 e 0 R, where R is averaged heat flux from the linear model and the per-
a random number having a Gaussin distribution with petual January GCM. We see that in both models the
unit variance and e 0 is a constant if dt is modified, the maximum heat flux is located near 408N and the vertical
noise should be multiplied by (dt)21/2 to maintain the axis of the flux tilts slightly poleward. In the linear model,
same variance, assuming that dt is shorter than any dy- the heat flux is trapped closer to the surface than in the
namically relevant timescales. The time step dt is 1728 perpetual January GCM. In Fig. 5 we show the horizontal
s (50 steps per day) in all of our experiments. The structure of eddy heat flux at s 5 0.865. It can be seen
strengths of the forcing in the temperature and vorticity that the overall distribution and the magnitude of heat
equations are denoted by e 0 5 e T and e 0 5 e V , respec- fluxes in the Pacific storm track are produced very well
tively. The same values are used at all points in the by the linear model, though the heat flux in the Atlantic
horizontal and vertical at which the forcing is not set storm track is somewhat smaller than that in the GCM.
explicitly to zero. We return to this point in section 3d.
In Fig. 6 we show the zonally averaged meridional
momentum flux u9y 9 from the linear model (Fig. 6a)
b. Tuning the model
and the perpetual January GCM (Fig. 6b). We see that
Our model has been explicitly tuned; that is, the pa- in the GCM, most of the momentum fluxes are con-
rameters a, b, g, and (e T , e V ) were adjusted until various centrated near the tropopause. The dominant feature is
second-moment statistics produced by the model match the strong northward momentum flux around 308N
those generated by the original nonlinear GCM when flanked by two regions where the momentum is trans-
its time-mean flow serves as the linear model’s basic ported southward. Another region of southward mo-
state. In this tuning process, we choose the perpetual mentum transport can be observed around 508N at the
January GCM as the target nonlinear system. Accord- surface levels. The overall distribution and the strong
ingly, the basic state of the linear stochastic model is northward momentum flux around 308N are captured by
1 OCTOBER 1999 ZHANG AND HELD 3423

FIG. 7. The distribution of transient eddy momentum fluxes at s 5 0.26 for the linear stochastic
model (a) and the R30 perpetual Jan GCM (b). Units: m 2 s22 .

the linear model fairly well, though the weak southward ulate the baroclinic eddies in the GCM, unfiltered data
momentum transport around 508N is overestimated by are used to calculate the lag-covariances. Data from the
the linear model. nonlinear GCM are still bandpass filtered as above.
Figure 7 depicts the meridional momentum flux at s Figures 8 and 9 show the lag-correlation maps of the
5 0.26 for the two models. The pattern of the momen- geopotential height at 350 mb from the GCM and the
tum flux is not as well simulated as the pattern of the linear model for lags 22, 0, 12 days. The reference
heat flux. The momentum fluxes are shifted eastward point is located at (408N, 1808). Although wave prop-
somewhat in the Pacific as compared with the GCM. agation is clearly present in the linear model, the wave
The fluxes are also too zonally symmetric in the Atlan- train is not as persistent and part of the Pacific wave
tic. train is refracted too strongly into the Tropics. This re-
sult is improved slightly when a bandpass filter is ap-
plied to the data from the stochastic model as well (not
c. The propagation of eddies in the linear model
shown)
Lag-correlation maps of bandpass filtered data are To further study the wave propagation properties in
useful for illustrating the statistical structure and evo- the linear model we have conducted experiments in
lution of high-frequency disturbances (Wallace et al. which we force the temperature and vorticity equation
1988). Since the linear model has been designed to sim- separately. These experiments indicate that temperature
3424 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

FIG. 8. One-point correlations of bandpass filtered geopotential height at 350 mb from the R30
perpetual Jan GCM. Reference point is (408N, 1808).

forcing produces a more persistent, zonally propagating found parameters for which the dominant wave train
wave train and vorticity forcing is more responsible for and the eddy fluxes of heat and momentum are all well-
the propagation to the Tropics. This difference appears simulated simultaneously.
to be a consequence of the fact that white noise tem-
perature forcing excites smaller horizontal scales than
d. Sensitivity studies
white vorticity forcing, and these smaller scales do not
propagate as efficiently into the Tropics. We have ex- Our primary goal in tuning the model was to repro-
perimented with the model parameters to determine how duce the second-moment statistics of the perpetual
best to mimic the GCM’s wave train. In Fig. 10 we GCM. In particular, we have concentrated on the hor-
perturb the temperature equation everywhere over the izontal and vertical distribution of geopotential height
globe, at all levels, with noise that is once again white variance, heat fluxes, and momentum fluxes. In this sec-
in the vertical as well as in the horizontal. In this case, tion we show how these characteristics change when
the stochastic model produces significant low-frequency the model parameters deviate from the control values
variability, so a bandpass filter is applied to the 350-mb described above. When a parameter is changed in the
geopotential height. Under these conditions the model experiments described in this section, we always re-
produces a wave train that resembles fairly closely that adjust the strength of the stirring so that the magnitude
in the GCM, but the fluxes generated by this model are of the geopotential height variance at the center of the
less satisfactory than those shown earlier in this section. Pacific storm track at 350 mb matches that in the GCM.
Within our self-imposed constraints of spatial white This is equivalent to examining the relative magnitudes
noise forcing vorticity and temperature, we have not of the fluxes and variances relative to the 350-mb geo-
1 OCTOBER 1999 ZHANG AND HELD 3425

FIG. 9. One-point correlations of geopotential height at 350 mb from the linear stochastic
model. Reference point is (408N, 1808).

potential height variance at the center of the Pacific Newtonian cooling. In Fig. 12a we reduce b to 0.83
storm track. day21 and e V , e T to (6.1 3 10211 s22 , 3.0 3 1026 K s21 ).
We first examine the model sensitivity to the lower- The heat flux is doubled in both storm tracks. We can
level mechanical damping and Newtonian cooling. We no longer simultaneously generate the correct magnitude
find that the lower-level heat flux is most sensitive to of the upper-level eddy variance and the low-level heat
the change of these parameters. Figure 11 shows the flux. When b is increased to 2.5 day21 and e V , e T to
heat flux at s 5 0.865 with a 5 0.42 (a), 0.83 (b), and (9.4 3 10211 s22 , 4.7 3 1026 K s21 ) in (c), the heat flux
1.25 (c) day21 [the results described above correspond consistently decreases in both oceans. The relative mag-
to (b)]. The strength of the stirring e V , e T is adjusted to nitude of the heat fluxes over the two oceans does not
(5.5 3 10211 s22 , 2.8 3 1026 K s21 ) in (a) and (1.0 3 change appreciably. In short, while mechanical damping
10210 s22 , 5.2 3 1026 K s21 ) in (c). (The ratio of e V to and Newtonian cooling both affect the strength of the
e T is held fixed in this readjustment.) All the other pa- lower-level heat fluxes in the storm tracks, for fixed
rameters are kept the same as in (b). When the lower- upper-level eddy amplitudes in the Pacific, the mechan-
level mechanical damping is weakened (decreasing the ical damping can further control the relative magnitudes
strength of the stirring to keep the upper-tropospheric of the heat fluxes in the two oceanic storm tracks. The
geopotential variances in the Pacific at the right level), upper-level momentum fluxes are not very sensitive to
it is primarily the heat flux in the Atlantic storm track these two parameters in the range we have covered,
that increases. given that we adjust the stirring level in each case to
The model responds differently to a change in the maintain the same upper-troposphere variances. Though
3426 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

FIG. 10. One-point correlations of geopotential height at 350 mb from the linear stochastic model
when only the temperature equation is forced everywhere. Reference point is (408N, 1808).

stronger damping naturally brings weakened lower-level track is hardly present when only a local region up-
eddy statistics, the vertical profiles of fluxes and vari- stream of the Pacific Ocean is perturbed.
ances do not change drastically in these experiments. It is interesting to compare this result with that in
We have also examined aspects of the model sensi- which we perturb only the temperature equation be-
tivity to the horizontal distribution of the random forc- tween 208 and 608N, 308 and 908E but at all levels in
ing, continuing to assume that it is spatially uncorre- the vertical, shown in Fig. 14. We expect a persistent
lated. In one experiment, shown in Fig. 13, we perturb wave train propagating more directly from the Pacific
the temperature and vorticity field between 208 and 608N to the Atlantic Ocean in this version of the model, as
and below s 5 0.76, just as in our standard case, but already seen in Fig. 10. Consistently, this version pro-
only in a longitudinally confined region over Asia be- duces a well-defined Atlantic storm track, even without
tween 308 and 908E. The other parameters assume the stirring in the Atlantic sector. It is interesting that this
same values as in the standard integration. Compared kind of change in the structure of the stirring can have
to the standard run in which temperature and vorticity such a dramatic impact on the upstream seeding of the
are forced everywhere in the extratropics, this experi- storm tracks.
ment produces a well-defined Pacific storm track, while To explore this sensitivity further we have repeated
the Atlantic storm track is very weak. Since part of the this experiment with several other distributions of forc-
Pacific wave train is strongly refracted into the Tropics ing, including vorticity forcing at all levels, vorticity
in this version of the stochastic model, the Atlantic forcing at low levels only, and temperature forcing at
storm track is not seeded very efficiently by eddies com- low levels only. In all cases, the forcing is once again
ing from the Pacific, and therefore the Atlantic storm localized in the same central Asian region. It is found
1 OCTOBER 1999 ZHANG AND HELD 3427

FIG. 11. The distribution of transient eddy heat fluxes at s 5 0.865 from the linear stochastic model:
a 5 (a) 0.42, (b) 0.83, and (c) 1.25 day21. Units: m s21 K.

that temperature forcing produces a stronger Atlantic But in fitting the structure of the Pacific wave train and
storm track, for a given strength of the Pacific storm its ability to seed the Atlantic, the inclusion of some
track, than vorticity forcing with the same vertical struc- upper-tropospheric stirring is useful in creating distur-
ture. Further, perturbing at all vertical levels rather than bances that propagate more zonally. As stated earlier,
only at low levels also generates a stronger Atlantic spatially white temperature forcing, as contrasted with
storm track. spatially white vorticity forcing, also seems to favor
We do not understand these results fully. When fitting smaller scales that do not radiate as strongly into the
the variances and fluxes, we find it to be beneficial to Tropics and therefore propagate more easily from the
localize the stirring at low levels. Conservation of wave Pacific into the Atlantic.
action causes eddy energy and geopotential variance to The overall damping g also has an effect on the rel-
increase as eddies propagate upward into a region of ative magnitudes of the two storm tracks, but this sen-
stronger westerlies, a fact of central importance to the sitivity is itself sensitive to the structure of the forcing.
structure of the singular vectors that define those struc- When g is increased in our control linear model, with
tures which can grow most rapidly over finite time in- stirring only at low levels, the Atlantic storm track gets
tervals (see Buizza and Palmer 1995). The larger scales somewhat weaker with respect to the Pacific storm track,
propagate upward more easily and these larger scales but this change is not very large. (We do not display
appear to be needed to obtain the correct amplitudes for this result here.) This is consistent with the picture that
the upper-tropospheric eddy fluxes. If one forces only a large part of the Pacific wave train is refracted into
at upper levels, the solution is noisier, less dominated the Tropics instead of propagating into the Atlantic
by particular waves that have been selectively amplified. Ocean so that g cannot easily regulate the relative
3428 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

FIG. 12. The distribution of transient eddy heat fluxes at s 5 0.865 from the linear stochastic model:
b 5 (a) 0.83, (b) 1.67, and (c) 2.5 day21. Units: m s21 K.

strength of the two storm tracks. In the case where only retuning the model when the basic state is altered. In
temperature forcing is present at all levels and the wave this section we examine the sensitivity of the storm
train tends to propagate more directly into the Atlantic tracks to the basic-state changes associated with sea-
Ocean, the relative magnitude of the two storm tracks sonal and interannual variations.
is more sensitive to the change of g, as shown in Fig.
15, with weaker g producing a relatively strong Atlantic a. The midwinter suppression of the Pacific storm
storm track. track in the linear stochastic model
Having tuned the linear model with the perpetual Jan-
4. Seasonal and interannual variation of storm uary data, we now change the basic state of the linear
tracks in the linear stochastic model model to the time-mean flow of the other five perpetual
In the previous sections we have shown that the sto- runs without changing any other model parameters. All
chastic model is capable of reproducing many features the procedures to run the stochastic model and to obtain
of the climatological storm tracks of the perpetual Jan- the statistics are unaltered. All parameters in the linear
uary GCM, such as the horizontal and vertical distri- model are kept fixed, including the amplitude of the
bution of variances and momentum and heat fluxes. We white noise forcing. Figure 16 shows the rms of geo-
have also shown that these eddy statistics are sensitive potential height at 205 mb obtained from the linear mod-
to the lower-level boundary layer dampings. The power el using the six time-mean flows as the basic state. Com-
of a stochastic model of this kind will be determined paring these figures with their counterparts from the
by its ability to predict storm track variations without perpetual GCM runs in Fig. 1, we see that the seasonal
1 OCTOBER 1999 ZHANG AND HELD 3429

FIG. 13. The distribution of rms of the geopotential height at 350 mb produced by the linear stochastic
model when only the region between 208 and 608N, 308 and 908E is perturbed (vorticity and temperature
forcing at low levels). Units are meters.

variation of the intensity of the Pacific storm track is nostics Group. The ‘‘ENSO storm track’’ is first cal-
largely captured by the linear model. Most interestingly, culated for each 1982/83 DJF period and then averaged
the midwinter suppression of the Pacific storm track is over four integrations.
qualitatively produced by the linear model, though the In Fig. 17 we present the rms of the bandpass filtered
degree of suppression is less than in the nonlinear mod- DJF geopotential height at 205 mb for the control storm
el, and one of the two maxima is found in November track (a) and the ENSO storm track (b). The difference
instead of in October. Significant differences are found between (b) and (a) is shown in (c). We see that in the
in February, with the relative strength of the Atlantic winter of the El Niño years, while the Atlantic storm
and Pacific storm tracks poorly reproduced in the linear track shifts southeastward, the Pacific storm track shifts
model. The linear model also captures the eastward shift southward, with a modest intensity increase in the mod-
of the Pacific storm track from fall to midwinter. The el. In Fig. 18 (M. Ting 1998, personal communication)
result is robust with respect to changes in model pa- we show the regression of monthly mean bandpass
rameters. In particular, when the model is tuned to obtain transient eddy kinetic energy against the El Niño index
a more zonally propagating wave train (Fig. 10), this for the period from November through March of 1979–
predicted seasonal cycle is qualitatively unchanged. 94 using NCEP/NCAR twice-daily reanalysis wind
data. The El Niño index used for this purpose is the
time series of the first EOF of Pacific SST. This ENSO
b. Modeling the interannual variation of storm tracks
composite is not quantitatively comparable to the GCM
during El Niño episodes
simulations of 1982/83, but it illustrates that the GCM
Having shown that the model does fairly well in sim- responds in a fairly realistic way to the imposition of
ulating the seasonal variation of the storm tracks, we ENSO boundary conditions, especially over the Pa-
now check if it is able to capture the basic features of cific.
storm track variations associated with ENSO. Figure 19 depicts the control storm track (a), the
The ‘‘control climate’’ we choose for this purpose is ENSO storm track (b), and the difference between the
the 30-yr-mean DJF (Dec–Feb) from a GFDL 14-level two reproduced by the linear stochastic model. Com-
seasonal R30 GCM. The ‘‘control storm track’’ is also paring Fig. 19a with Fig. 17a, we see that the overall
defined during this DJF period and averaged for 30 distribution of the two storm tracks in the El Niño win-
years. The ‘‘ENSO climate’’ is the average of four in- ters are reproduced yet the magnitude is somewhat un-
tegrations for the 1982/83 DJF from a series of ENSO derestimated. The southward shift of the storm tracks
experiments carried out by the GFDL’s Climate Diag- in the El Niño winters is captured fairly well by the

FIG. 14. The distribution of rms of the geopotential height at 350 mb produced by the linear stochastic
model when only the region between 208 and 608N, 308 and 908E is perturbed (temperature forcing
only, at all levels). Units are meters.
3430 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

FIG. 15. The horizontal structures of the rms of the geopotential height at 350 mb from the perpetual
Jan run (a) and from the linear model when temperature fields at all levels are perturbed and g 5 (b)
0.083, (c) and 0.125, and (d) 0.25 day21. Units are meters.

linear model. However, the modest strengthening of the 5. Conclusions and discussion
Pacific storm track during ENSO is not present in the
linear model. In fact, the model predicts a slight weak- The goal of this work is to relate the nonlinear baro-
ening of the stormtrack. clinic eddy statistics to the climatological mean flow.
 1 OCTOBER 1999
ZHANG AND HELD

FIG. 16. The distribution of rms of the geopotential height at 205 mb produced by the linear stochastic model. Units are meters.
3431
3432 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

FIG. 17. The rms of the bandpass-filtered geopotential height at 205 mb during DJF: (a) from the
14-level seasonal GCM, (b) from the 1982/83 ENSO experiment, and (c) the difference between (b)
and (a). Units are meters.

We have first constructed a linear stochastic model by in this way, the Pacific wave train is refracted too strong-
adding linear damping and white noise forcing terms to ly into the Tropics. If we try to correct this deficiency
a linearized primitive equation model. With the param- by stirring at upper levels, as well as by forcing the
eters chosen appropriately, the linear stochastic model temperature equation preferentially, which produces a
can reproduce with a fair degree of accuracy the eddy more zonally elongated wave train, we lose fidelity in
geopotential variance as well as eddy heat and momen- our simulation of the fluxes.
tum fluxes of the perpetual January GCM, when the The modeled heat fluxes are sensitive to the lower-
perpetual January climate is used as the linear model’s level friction and Newtonian cooling coefficients, while
basic state. Constraining our subjective tuning to spa- the momentum fluxes respond only weakly to these
tially white forcing, the best result is obtained using changes in low-level damping if the stirring is adjusted
forcing in both the temperature and vorticity that is lo- to maintain the correct upper-level eddy variance in the
calized in the lower troposphere. However, when tuned Pacific storm track.
1 OCTOBER 1999 ZHANG AND HELD 3433

FIG. 18. Regression map of monthly mean bandpass transient eddy kinetic energy against the El Niño index based on the
data during the period Nov through Mar of 1979–94. The original wind data were taken from NCEP–NCAR twice-daily
reanalysis. The El Niño index is the time series of the first EOF of Pacific SST.

To examine the model’s predictive power we predict es in eddy amplitudes quantitatively with fixed stirring
the storm track seasonal variation from the seasonal amplitudes.
changes of the time-mean flow. This effort has been It is interesting that we are able to produce realistic
generally encouraging. The seasonal change of the Pa- heat fluxes and momentum fluxes simultaneously, in
cific storm track intensity and its location are reproduced contrast to the related study of Whitaker and Sardesh-
to a certain degree by the linear stochastic model. The mukh (1997) using a two-layer model. The freedom we
midwinter suppression is present qualitatively, but the have given ourselves in perturbing both the temperature
suppression is weaker than in the GCM. Since the am- and vorticity field when tuning the model could account
plitude of the eddy stirring should increase as eddy am- for this improved result. If we think of the forcing of
plitudes increase, models with fixed stirring potentially potential vorticity, modifying the ratio of the vorticity
miss a positive feedback from changes in stirring am- to the temperature forcing is, in part, a crude way of
plitude. So we should not be surprised by the muted modifying the spatial spectrum of the most dynamically
character of the linear model’s suppression. relevant part of the forcing. On the other hand, it may
This result suggests that the midwinter suppression be that we are not obtaining these flux patterns for en-
can be attributed at least partly to the variation in struc- tirely correct reasons, since the shape of the model’s
ture of the background flow, as one progresses through dominant wave trains in the Pacific is distorted.
the winter season. One should now be able to analyze While this stochastic approach is promising, we clear-
the linear model dynamics to establish a better theo- ly need theoretical guidance on how best to stir and to
retical understanding of the phenomenon. We can im- damp the linear model. The stirring and damping must
mediately deduce from this result that the direct effects eventually be tied to the eddy statistics themselves, as
of latent heat release on the eddies is not of dominant in turbulent closure theories.
importance in this seasonal cycle, for our linear model Rather than linearizing about the time-mean flow, as-
does not include any effects of moisture on the eddies. suming a form for the damping, and then tuning the
However, we cannot totally exclude the possibility that damping parameters, one can also try to optimize the
latent heating effects enhance the suppression some- fit of the stochastic model to the GCM or observations
what. more directly, as described by DelSole (1996) in his two
As for interannual variations, the model has the ca- layer quasigeostrophic experiments. One should be able
pacity to reproduce the shift of storm tracks associated to fit the eddy statistics more accurately with this ap-
with ENSO, although the change of storm track mag- proach since the choice of the time-mean flow as the
nitude is not captured with a fixed level of stirring. Once flow about which to linearize and the choice of the form
again, we should not expect to be able to explain chang- of the damping are both somewhat arbitrary. The dis-
3434 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 56

FIG. 19. The rms of the geopotential height at 205 mb from the linear stochastic model linearized
around (a) the time-mean DJF flow of the 14-level seasonal GCM, (b) the time-mean DJF flow of the
ENSO experiment 1982/83, and (c) the difference between (b) and (a). Units are meters.

advantage of this approach is that one loses contact with YZ was supported in part by NOAA Grant NA67RJ0120
the underlying dynamical operator. A combination of to Princeton University.
the two approaches should help us better understand the
connections between storm tracks and the mean flow, REFERENCES
as well as to appreciate the limitation of the linear sto-
Alexander, M. A., and J. D. Scott, 1996: Atlas of Climatology and
chastic framework. Variability in the GFDL R30S14 GCM. U.S. Government Print-
ing Office, 1996-774-842, 121 pp. [Available from NOAA-CI-
Acknowledgments. We thank Peter Phillipps for in- RES Climate Diagnostics Center, 325 Broadway, Boulder, CO
80303.]
tegrating the perpetual insolation GCM, and Paul Kush- Buizza, R., and T. N. Palmer, 1995: The singular-vector structure of
ner, Zuojun Zhang, Isidoro Orlanski, and Steve Garner the atmospheric global circulation. J. Atmos. Sci., 52, 1434–
for assistance in interpreting the linear model results. 1456.
1 OCTOBER 1999 ZHANG AND HELD 3435

Christoph, M., U. Ulbrich, and P. Speth, 1997: Midwinter suppression Frederiksen, J. S., 1983: Disturbances and eddy fluxes in Northern
of Northern Hemisphere storm track activity in the real atmo- Hemisphere flows: Instability of three-dimensional January and
sphere and in GCM experiments. J. Atmos. Sci., 54, 1589–1599. July flows. J. Atmos. Sci., 40, 836–855.
DelSole, T., 1996: Can quasigeostrophic turbulence be modeled sto- Hoskins, B. J., and P. J. Valdes, 1990: On the existence of storm-
chastically? J. Atmos. Sci., 53, 1617–1633. tracks. J. Atmos. Sci., 47, 1854–1864.
Farrell, B., and P. Ioannou, 1993a: Stochastic forcing of perturbation Leith, C. E., 1971: Atmospheric predictability and two-dimensional
variance in unbounded shear and deformation flows. J. Atmos. turbulence. J. Atmos. Sci., 28, 145–161.
Sci., 50, 200–211. Nakamura, H., 1992: Midwinter suppression of baroclinic wave ac-
, and , 1993b: Stochastic dynamics of baroclinic waves. J. tivity in the Pacific. J. Atmos. Sci., 49, 1629–1642.
Atmos. Sci., 50, 4044–4057. Wallace, J. M., G. H. Lim, and M. L. Blackmon, 1988: Relationship
, and , 1994: A theory for the statistical equilibrium energy between cyclone tracks, anticyclones and baroclinic waveguides.
spectrum and heat flux produced by transient baroclinic waves. J. Atmos. Sci., 45, 439–462.
J. Atmos. Sci., 51, 2685–2698. Whitaker, J. S., and P. D. Sardeshmukh, 1998: A linear theory of
, and , 1995: Stochastic dynamics of the midlatitude at- extratropical synoptic eddy statistics. J. Atmos. Sci., 55, 237–
mospheric jet. J. Atmos. Sci., 52, 1642–1656. 258.
, and , 1996a: Generalized stability theory. Part I: Auton- Zhang, Y., 1997: On the mechanisms of the mid-winter suppression
omous operators. J. Atmos. Sci., 53, 2025–2040. of the Pacific stormtrack. Ph.D. thesis, Princeton University,
, and , 1996b: Generalized stability theory. Part II: Non- 152 pp. [Available from NOAA/GFDL, Princeton University,
autonomous operators. J. Atmos. Sci., 53, 2041–2053. Princeton, NJ 08542-0308.]