Saltzman B. (editor) Anomalous Atmospheric Flows and Blocking
Подождите немного. Документ загружается.
374
STEFAN0 TIBALDI
Simmons,
A.,
and Jarraud, M. (1984). The design and performance of the new ECMWF
Operational Model Proceedings of the ECMWF 1983 Seminars on ‘Numerical Methods
for Weather Prediction,’ Vol. 2, pp.
1
13- 164. Available from ECMWF.
Sumi, A., and Kanamitsu, M. (1984).
A
study of systematic errors in a numerical weather
prediction model. Part
I:
General
aspects
ofthe systematic errorsand their relation with the
transient eddies.
J
Meteorol.
SOC.
Jpn.
62,234-250.
Tibaldi,
S.
(
1982). The production of a high resolution global orography and associated clima-
tological surface
fields
for operational
use
at ECMWF.
Riv.
Meteorol. Aeronaut.
42,285-
308.
Tibaldi,
S.
(1984). On the relationship between the systematic error of the ECMWF forecast
model and orographic forcing. Predictability of fluid motions
(La
Jolla Institute, 1983).
AfPConJ
Proc.
106,397-418.
Tiedtke, M. (1983). Winter and summer simulationswith the ECMWF model. Proceedings of
the ECMWF Workshop on ‘intercomparisons of large-scale models used for extended
range forecasts,’ pp. 263-314. Available from ECMWF.
Wahr,
J.
M., and
Oort,
A.
H.
(1984). Friction and mountain torque estimates from global
atmospheric data.
J.
Almos.
Sci.
41, 190-299.
Wallace,
J. M., and Woessner,
J.
K.
(
198
1
).An analysis offorecast errors in the NMC hemisphe-
ric primitive equation model.
Mon.
Wearher
Rev.
109,2444-2449.
Wallace, J. M., Tibaldi,
S.,
and Simmons,
A.
(1983). Reduction of systematic forecast error in
the ECMWF model through the introduction ofan envelope orography.
Q.
J.
R.
Meteorol.
Williamson, D.
L.,
and Temperton,
C.
(1981). Normal mode initialisation for a multi-level
SOC.
109,683-717.
grid-point model. Part 11: Nonlinear
aspects.
Mon.
Weuther
Rev.
109,
744-757.
NUMERICAL FORECASTS
OF
TROPOSPHERIC AND STRATOSPHERIC
EVENTS DURING THE WINTER OF
1979:
SENSITIVITY TO THE MODEL’S
HORIZONTAL RESOLUTION AND
VERTICAL EXTENT
CARLOS
R.
MECHOSO, MAX
J.
SUAREZ,* KOJI
YAMAZAKI,?
AKIO
KITOH,?
AND
AKIO
ARAKAWA
Department
of
Atmospheric Sciences
University
of
California,
Los
Angeles
Los
Angeles, California
90024
1.
INTRODUCTION
In this paper we report the results of several medium-range (10-day) exper-
imental forecasts with the University of California,
Los
Angeles (UCLA)
general circulation model (GCM). Our primary interest in these experi-
ments is in the impact of increased horizontal resolution and vertical extent
on the accuracy of the forecasts. It is a pervasive idea in numerical weather
prediction that increased horizontal resolution generally results in more
accurate forecasts (e.g., Bengtsson, 198
1).
On the other hand, the impact of
the artificial upper boundary is a recognized but relatively unexplored prob-
lem. Here we note that having the model top at zero pressure is, in practice,
equivalent to having the model top at some small finite nonzero pressure as a
result of inevitable finite-difference errors (Lindzen
et
al.,
1968).
Lambert (1980) investigated the sensitivity
of
short-range forecasts to
increased vertical resolution and the addition of stratospheric levels. He
compared forecasts with versions of the Canadian Atmospheric Environ-
*
Present address: Laboratory for Atmospheres, National Aeronautics and Space Adminis-
t
Present address: Meteorological Research Institute, Yatabe-machi, Tsukuba-gun, Ibaraki-
tration, Goddard Space Flight Center, Greenbelt, Maryland 2077
1.
ken 305, Japan.
375
Copyright
0
1986
by Academic Press,
Inc.
All
rights
of
reproduction in any
form
reserved.
ADVANCES
IN
GEOPHYSICS,
VOLUME
29
376
CARLOS
R.
MECHOSO
ET
AL.
ment Service model which have
5,
1
1,
and 15 levels. The first two versions
have the top level at approximately
100
mbar, while the last version has the
top at
10
mbar and is identical to the
1
l-level version below
100
mbar. He
found that the differences between these forecasts were generally very small
up
to
48
hr. Derome
(
198
1
)
analyzed the averaged fields for an ensemble of
seven I0-day forecasts with a 15-level version of the European Centre for
Medium Range Weather Forecasts (ECMWF) model. He found that errors
in geopotential height increase with time at all levels while having maximum
amplitudes at the uppermost levels. He also detected downward propagation
of the geopotential error growth throughout the entire vertical domain in
about
2
days. Simmons and Striifing (1983) examined the sensitivity of
tropospheric forecasts
to
stratospheric resolution using different versions of
the ECMWF model. They compared forecasts with versions which have
14,
16, and 18 levels and top levels at
50,
25,
and
10
mbar, respectively. They
found that the forecasts for
500
mbar up to
10
days are largely insensitive to
changes in the position of the top level. This conclusion was based on skill
scores, which are integrated quantities. However, when they looked for
impact on specific phenomena, they found some sensitivity of the plane-
tary-wave structures at high latitudes.
In Mechoso
et
al.
(1
982) we studied the impact of an upper boundary in
the lower stratosphere on tropospheric forecasts using two versions of the
UCLA GCM which have 9 and 15 layers and tops at 51.8 and
1
mbar,
respectively. We took a 15-day
period
from a long-term integration with the
1
5-layer version as the “control” and performed a “forecast” for this period
with the 9-layer version. We found that significant “errors” in the geopoten-
tial field rapidly appeared in the uppermost model layers, and later propa-
gated to all other layers. In particular, the ultralong waves in the forecast
showed a strong tendency to quickly become equivalent barotropic due to
reflection at the lowered upper boundary. Such a behavior of the ultralong
waves progressively affected shorter waves. At 500 mbar, significant errors in
the ultralong waves appeared within
5
days, while errors in the total field
became large after
10
days. From this experiment we expect that a lowered
position of the upper boundary can contaminate actual tropospheric fore-
casts after a few days, and that this contamination will become more appar-
ent as models improve and the period
of
useful prediction is lengthened.
Our strategy in this paper consists
of
performing 10-day actual forecasts
with four versions
of
the UCLA GCM that differ in horizontal resolution
and/or vertical extent, and of analyzing in detail the accuracy of these fore-
casts. We have selected four cases from the Northern Hemisphere winter of
1979. Two
of
these cases are from January 1979, when there was a minor
stratospheric warming; and the other two cases are from February 1979,
when there was a major stratospheric warming.
NUMERICAL FORECASTS
OF
EVENTS
377
We begin in Section
2
by describing some selected observed features of the
tropospheric and stratospheric circulations during January and February of
1979. The UCLA GCM is briefly described in Section
3.
Tropospheric and
stratospheric forecasts are presented in Sections
4
and
5,
respectively. In
Section
6
we discuss the influence of the upper boundary
on
tropospheric
forecasts.
2.
SELECTED
FEATURES
OF
THE
ATMOSPHERIC CIRCULATION
DURING
THE
NORTHERN HEMISPHERE WINTER
OF
1979
In this section we describe some selected observed features of the tropo-
spheric and stratospheric circulations of the Northern Hemisphere during
the period
12
January
to
27 February 1979. The data are identical to those
used by Mechoso
et
al.
(1985). They consist
of
the First GARP Global
Experiment (FGGE) Level 111-b analysis by ECMWF below
10
mbar, and
United States National Meteorological Center (NMC) global height and
temperature fields above that level. Winds above 10 mbar and at latitudes
higher than
20"
were obtained by using the geostrophic relations, while for
90
80
70N
w
0
60
2
50
UO
Q
30
20N
10
J
'12
16
20 2U 28 32
36
UO
1111
118
52 56
DAY
FIG.
1.
Longitudinally averaged geopotential field Z(meters) at 500
mbar
(a) between 90"W
and
0"
and
(b)
between 135"E and 135"W duringthe period
12
January 1979 to 27 February
1979.
378
CARLOS
R.
MECHOSO
ETAL
lower latitudes they were obtained by linear interpolation between 20"N and
20"s.
We begm with the circulation at 500 mbar and consider
two
sectors of the
Northern Hemisphere: the Atlantic Ocean sector (9OOW to
O'),
and the
Pacific Ocean sector (135"E to 135"W). Figure
1
shows that there were
episodes with persistent ridges at
500
mbar between about
40"
and
70"N
in
both sectors. For the Atlantic Ocean sector, the major episode was roughly
during
20
-
3
1
January. For the Pacific Ocean sector, it was after
20
Febru-
ary. These episodes were accompanied by splitting
of
the westerly flow into
two branches separated by easterly flow, as shown in Fig. 2. They were also
accompanied by strong zonally averaged northward eddy heat flux at high
latitudes in the lower stratosphere,
as
shown in Fig.
3.
This is revealing since
the zonally averaged northward eddy heat flux
is
proportional to the vertical
component
of
the Eliassen -Palm flux vector, which represents upward en-
ergy flux in the quasi-geostrophic framework. Therefore, both major epi-
sodes were associated with intense tropospheric forcing on the stratosphere.
Figure 3 also shows that zonal wavenumber
1
was the principal contributor
to the forcing during the January maximum, while zonal wavenumber
2
was
dominant during the February maximum.
The stratospheric events
of
the Northern Hemisphere winter of
i
979 have
90
80
70N
60
50
(I
30
20N
10
;z
(10
-1
'12
16
20
2V
28
32
36
(10
UU
118
52 56
DClY
FIG.
2. As
in
Fig.
1,
except
for
zonal
wind
u
(m
sec-I).
0
01
NO2
OE
D
I
Oh
n
09
NOL
08
06
0
01
NO2
OE
Oh
-
09
NOL
08
06
0s
2
0s
2
6Lf
SIN3A3
dO
SISV33XOd 1V3IX3AflN
FIG.
3.
Zonal mean northward eddy heat
flux
[v*T*]
(K
m
sec-')
at
50
mbar
for (a)
all
waves,
(b)
wavenumber
1,
and
(c)
wavenumber 2 during
the
period
12 January 1979
to
27 February
1979.
been described in several studies (e.g., Quiroz,
1979;
Palmer,
1981;
McIn-
tyre,
1982).
At
10
mbar the typical wintertime circulation consisting
of
a
circumpolar vortex changed during a minor warming in mid-January to a
pattern consisting
of
a cyclonic vortex and an Aleutian high. This pattern
persisted with slight variations in phase until it changed again during a major
warming in February to another pattern consisting
of
an anticyclonic vortex
over the polar region and two well-separated cyclonic vortices. The major
warming reversed the direction
of
the zonal flow in deep stratospheric layers.
380
CARLOS
R.
MECHOSO
ETAL
3.
DESCRIPTION
OF
THE
MODEL
The UCLA
GCM
is described mainly in Suarez
et
al.
(1983). Briefly, the
model is based on the primitive equations in a modified a-coordinate system
for which the lower boundary, the top of the planetary boundary layer
(PBL), and isobaric surfaces above a prescribed pressure level
(100
mbar),
including the model top, are coordinate surfaces. The primary prognostic
variables of the model are the surface pressure, the horizontal velocity, and
the potential temperature. In addition, the model predicts the water vapor
and ozone mixing ratios and the PBL depth. Formulation of the diabatic
processes include parameterizations of cumulus convection (Arakawa and
Schubert, 1974) and of boundary-layer processes (Suarez
ef
al.,
1983) and a
radiation calculation (Katayama,
1972;
Schlesinger, 1976) with both sea-
sonal and diurnal changes.
In
this study we use the 9- and 15-layer versions of the model with low and
high resolutions(4" oflatitude by
5"
oflongitudeand2.4" oflatitude by
3"
of
longitude, respectively). The 15-layer version has the top at
1
mbar and
6
layers above
5
1.8 mbar (with interfaces at 26.8, 13.9,
7.20,
3.73, and
I
.93
mbar), resulting in a stratospheric resolution of about
5
km. The 9-layer
version has the top at 5
1.8
mbar and is identical to the 15-layer version below
that level except that ozone is prescribed.
For
the primary prognostic variables, the initial conditions of the forecasts
were obtained from the dataset described in Section 2. The surface pressure
was obtained from the observed sea level pressure and geopotential fields.
The horizontal velocity was then obtained by interpolation of the observed
values. The geopotentials at the interfaces of the model's layers were also
obtained by interpolation from the observed values. This particular interpo-
lation procedure is designed such that the model's pressure gradient force is a
proper interpolation of that on the pressure surfaces. The potential tempera-
ture was then obtained from those geopotentials using the model's hydro-
static equation.
No
further initialization procedures were performed for
these variables.
To
obtain initial conditions for the ozone mixing ratio, the PBL depth,
and the water vapor mixing ratio with some degree of consistency with the
model, we performed an "update cycle," which consists of a 5-day integra-
tion preceding each forecast period.
For
this integration. the initial PBL
depth was
50
mbar everywhere, the water vapor mixing ratio was obtained
from the dataset described in Section 2 below 300 mbar and was set equal to
2.5
X
above that level, and the initial ozone mixing ratio was zonally
uniform and taken from Schlesinger and Mintz
(
1979). During the update
cycle, the primary prognostic variables were updated at regular time inter-
vals. Below
10
mbar, the updating interval was 6
hr
during the FGGE Special
NUMERICAL FORECASTS OF EVENTS
38
1
Observing Period, when data are available with that frequency, and 12 hr
outside of that period. Above 10 mbar, the interval
was
24 hr, since the
NMC
data are available only at 1200 GMT.
4.
TROPOSPHERIC
FORECASTS
4.1.
Forecast
from
16
January
The geopotential field at
500
mbar for 16 January is shown in Fig.
4.
Notable features are two cutoff lows, one at about
40"N
and 25"E and
another at about
30"N
and
30°W,
and the ridges at higher latitudes over
western and eastern Europe. The North American region shows a mature
cyclone over the northeastern part of the continent and a prominent trough
west of the Rocky Mountains. The Pacific Ocean region is dominated by the
Aleutian Low while the Asian region exhibits a high-latitude cyclone and a
middle-latitude trough.
The 5day low-resolution 9- and 15-layer forecasts and the high-resolution
9-layer forecast are shown in Fig.
5.
All three forecasts predict the merging of
0000
GMT
16
JAN
1979
FIG.
4.
Geopotential field (meters) at
500
mbar
for
16
January.
382
CARLOS
R.
MECHOSO ET
AL.
DAY
5
9-LAYER
4x5
FIG.
5.
Geopotential field (meters) at
500
mbar
for
2
1
January (top left), and forecasts from
16
January with the low-resolution 9-layer version (top right), high-resolution 9-layer version
(bottom left), and low-resolution 15-layer version (bottom right).
the ridges over northern Europe, the merging of the two cutoff lows, and the
associated split ofthe westerly jet around
40"
W.
However, they fail to predict
the deepening
of
the low
off
the east coast of North America, and placed the
trough over the continent too far east.
The 5-day means from
2
1
to
26
January of the geopotential field at the
loo-,
300-,
500-, and 700-mbar levels are shown in
Fig.
6.
These figures
illustrate the vertical structure observed during the second half
of
the forecast
period. The high over Greenland (the major episode over the Atlantic Ocean
sector described in Section
2)
and the low over the central North Atlantic
Ocean have little tilt with height. On the other hand, the Aleutian Low has a
NUMERICAL FORECASTS OF EVENTS
383
0000
GMT
21-28
JAN
1979
0000
GMT
21-26
JAN
1979
FIG.
6.
Five-day means From
21
January
to
26
January
of
the geopotential field at
100
(top
left),
300
(top right),
500
(bottom left), and
700
mbar (bottom right).
pronounced westward tilt with height, particularly in the upper troposphere.
The corresponding low-resolution 9-layer forecast, shown in Fig.
7,
success-
fully predicts the equivalent barotropic structure over the North Atlantic
Ocean, although the amplitudes are weaker than those observed. The pre-
dicted Aleutian Low
also
has an equivalent barotropic structure but here it
disagrees with the observed. The low-resolution 15-layer forecast shown in
Fig.
8,
on the other hand, is more accurate in predicting the baroclinic
structure of the Aleutian
Low
although it is less accurate over the North
Atlantic Ocean. The predicted flow at
100
mbar by both versions is generally
stronger than observed at midlatitudes, but weaker than observed over the
polar region.