Saltzman B. (editor) Anomalous Atmospheric Flows and Blocking

Подождите немного. Документ загружается.

354

STEFAN0

TIBALDI

stationary part of the waves.

can then summarize the message contained

in the diagrams of Fig.

by saying that the envelope orography contributes

to largely (but not completely) restore the observed spectral distribution of

kinetic energy among the planetary-scale wavenumbers, with this effect

somehow reducing in amplitude with height and being more evident for the

stationary part of the eddy flow. The somehow vertically evanescent nature

of the “envelope” effect on the kinetic energy spectra is in some contrast with

the widely accepted view that orographic mechanical forcing

adominating

effect in the upper troposphere and in the stratosphere, but less important,

compared

thermal forcing, in explaining the stationary features of the

lower troposphere (e.g., Manabe and Terpstra, 1974). Figure

1 1

contains the

isotropic amplitude spectra for both orographies and shows that, in going

from a mean to an envelope orography, the spectral amplitude has been

‘

FIG.

Amplitude spectra of mean and

J2fib

envelope orographies

a function

the

isotropic two-dimensional wavenumber (Legendre polynomial index

n).

the

axis the

natural logarithm of

I),

on the

axis the natural logarithm of the amplitude

surface

geopotential (m2

se~-~).

ENVELOPE OROGRAPHY AND MAINTENANCE

355

increased almost identically for almost all wavenumbers up to WN

16,

with a

somewhat larger proportional increase for

and for wavenumbers

larger than

16.

This, in

linear framework, hardly justifies the response

implied by the diagrams of Fig.

10,

where we showed an increased

response and a largely decreased

response.2

picture has therefore

emerged of a “selective” model response (in wavenumber space and location

in the vertical) to a generally increased orographic forcing, with a large,

necessarily indirect, impact on the zonal mean mass and wind fields.

At this stage it

interesting to try to diagnose whether this behavior of the

eddies and of the zonal flow in the envelope ensemble

forecasts is the

consequence of a rapid (linear) response of the model atmosphere’s eddy

flow to an enhanced large-scale eddy orographic forcing, eventually modify-

ing the zonal flow via nonlinear wave-mean flow interactions, or the zonal

flow is affected directly by the small scales of the envelope orography and it,

in turn, affects the response of the quasi-stationary planetary waves. The

results of an experiment designed to shed light on precisely this question is

discussed in Section

THE

EXPERIMENTS

WITH THE

BLENDED

OROGRAPHIES

In order to address the question raised at the end of the previous section, it

was decided to construct two spectrally ‘blended’ orographies in the follow-

ing way: the envelope increment, that is, the proportion of the subgrid

terrain height variance that is added to the gridbox mean orography, was

decomposed in spherical harmonics, using the same base functions used by

the ECMWF spectral model. Two orographies were then resynthesized,

blending the spherical harmonic coefficients in the total wavenumber band

from the mean (envelope) orography with the coefficients in the wave-

number band

to truncation from the envelope (mean) orography. We will

call the first

these two LSMO, for large-scale mean orography (implying

that the small scales come from the envelope orography), and the second

LSEO,

for large-scale envelope orography (implying that the small scales

come from the mean orography). Now we have four orographies at our

disposal: MEA, LSMO,

LSEO,

and ENV.

We should point out that although the spectra of Fig.

(zonal wavenumber spectra) and

Fig.

(isotropic wavenumber spectra) are not directly comparable, since we

are

concentrating

here on the planetary scales, the argument is little affected. The correspondent “m” orography

spectrum (not shown) would show a relative change similar to the one shown by Fig.

1 1

in going

from a mean to an envelope orography. The “n”

spectra

were preferred in this case

because

their easier interpretation in terms of the orography “blending” experiments described in

Section

356

STEFAN0

TIBALDI

the behavior of the forecasting model, measured looking in particular at

the

response, using, for example,

LSEO,

similar

its behavior

using

ENV

and different from the one obtained using

MEA,

while, con-

versely, the behavior using

LSMO

similar to the behavior using

MEA,

will infer that the “envelope” effect on the ultralong wavescomes from direct

forcing by the correspondent wavenumbers and that the mean zonal flow is

then influenced via wave- mean flow interaction.

If,

however, the contrary is

true

(LSEO

behaves like

MEA

and

LSMO

like ENV), we will infer that the

smaller scales of the orography have a direct (beneficial) effect on the evolu-

tion ofthe mean zonal flow and that this improved zonal flow affects, in turn,

the response of the planetary waves.

These experiments had to

canied out, for practical reasons, with the

ECMWF

T63

spectral model (Simmons and Jarraud, 1984) rather than with

the gridpoint model. There are, however, a number

experiments showing

that the impact

the envelope orography on the

ECMWF

T63 spectral

model is very similar to the one we have documented in the previous section

on the gridpoint model (e.g., Jarraud and Simmons, 1984). Another differ-

ence that should be mentioned here and that is also reported and explained

in Jarraud and Simmons

(

1984) is that the envelope orography used with the

spectral model is of the type

J2ph

and not

2j4,

as it was

in the case for the earlier experiments using the gridpoint model. This differ-

ence has its basis in

ECMWFs

operational needs (we should point out here

that the T63 spectral model using a

421h

envelope orography replaced in

operations at

ECMWF

the N48 gridpoint model with mean orography in

April 1983) and was canied over to the experiments described here only for

reasons of convenience.

All

available experimentation with the T63

ECMWF

spectral model seems to indicate that whatever strong signal is

emerging from this set

experiments can

transported to the gridpoint

model environment, and vice versa.

Figure

shows the Northern Hemispheric tropospheric anomaly corre-

lation coefficient of geopotential height for all waves (a, b, and c) and for the

wavenumber band

1-3 only (d, e, and f) for the

case

chosen for this

experiment. The initial conditions were 30 January 1982 at

GMT

and all

forecasts were conducted up to

days. These particular initial conditions

were chosen

for

a variety of practical reasons, the main ones being that the

MEA

and

ENV

integrations were already available and that this was the case

study that had shown possibly the greatest sensitivity to the change in

orog-

raphy, among those

used

test the impact

the envelope orography on the

spectral model before their joint operational implementation in April 1983.

Let

first concentrate on panels a, b, d, and e. The important conclusion

that emerges is that, of the two alternatives described earlier in this section,

the second

far the most plausible one, since

LSEO

behaves almost

ENV

MEAN

FIG.

12. Mean tropospheric anomaly correlation coefficient of geopotential height

function of forecast time. (a-c), Full fields; (d-f), wavenumbers

1-3

only.

ENV,

Envelope

orography model; MEA, mean orography model;

LSEO,

large-scale envelope orography model;

LSMO, large-scale mean orography model. All experiments were performed using the T63

ECMWF spectral model. The initial

data

for

all

experiments was 30 January 1982

GMT.

358

STEFAN0

TlBALDl

identically to MEA, and LSMO to ENV. This shows, therefore, that the

envelope effect is due to the smaller scales of the orography. Before conclud-

ing from this that the smaller scales of the orography exert their beneficial

impact by directly affecting the zonal flow (and it, therefore, in turn affecting

the response of the planetary-scale waves), we should mention that the

behavior of the two integrations that use the blended orographies (LSMO

and

LSEO)

evidenced

neatly by the objective scores of Fig.

is not

confined to this particular indicator (Northern Hemisphere tropospheric

geopotential height correlation coefficient). Mean cross-sections of zonal

wind error for the last

days of the

10-day

integrations show that, for the

lower tropospheric belt

50”

-60”N,

LSEO

and MEA have a mean zonal

wind in excess ofthe observed ofalmost

m sec-I. For

LSMO

and ENV this

error is reduced to around

m sec-’.

Panels c and

of Fig.

show the results from further 10-day integrations

using yet two different blended orographies. They have been produced using

exactly the same procedure

LSEO

and LSMO but with the wavenumber

cut off at WN

instead of

Orograph

LSEOT9

(LSMOT9) is, there-

and all other waves from the mean

(J2ph

envelope) orography. This was

done in order to isolate more efficiently the spectral zone where the envelope

increment is directly felt and the spectral zone where we are measuring the

impact

(WN

-3).

Panel

shows that, at least up to day

an envelope

orography that is “envelope” only in the very short scales still has the same

effect on the ultralong planetary scales, and this effect shows clearly even in

the earliest part of the forecast

(days

3),

when an influence of the smaller

scales of motion through progressive up-the-spectrum nonlinear interac-

tions could not yet

felt by the planetary, ultralong scales since the time

elapsed is much shorter than the eddy turnover time (Lorenz,

1969;

kith,

1971).

We conclude therefore that the smaller scales

the orography have an

effect

the mean zonal flow by reducing its westerly strength in the lower

tropospheric midlatitudes. Simple linear theory of stationary Rossby waves

(e.g., Held,

1983)

could then suggest that shorter stationary wavelengths are

favored by the weakening

the mean (westerly) zonal wind since

d(p/

U),

where

is the stationary Rossby wavenumber, pis the latitudinal varia-

tion of the Coriolis parameter, and

is the strength of the mean zonal wind.

Despite the crudity of this argument, this is quite consistent with the decrease

of WN

response and the corresponding increase of

response brought

about by the introduction of the envelope orography,

shown in Fig.

10.

The next section will be devoted to the description of some further diag-

nostic work aiming to investigate the physical mechanism by which the

fore, resynthesized with

0-9

from the

2,uh

envelope (mean) orography

ENVELOPE

OROGRAPHY

AND MAINTENANCE

359

(smaller scales of the) orography can influence the zonal flow on compara-

tively shorter time scales.

MOUNTAIN TORQUE

AND

ZONAL FLOW

We have

far concluded that there must be a direct and effective link

between the specification of orography and the strength of the tropospheric

zonal flow. Let us remember that the mountain torque is, together with the

frictional torque, the only possible source or sink of total angular momen-

tum for the global atmosphere:

where

RrZ

cos

is the total angular momentum for unit mass,

is the total mass of the

atmosphere,

is the total surface of the earth, and

and

are the local

frictional and mountain zonal torques per unit area.

The mountain torque would therefore seem as a very good candidate to

hold the responsibility for an efficient, dynamical coupling between the

atmosphere and the underlying solid earth. Figure

shows the mean ana-

lyzed mountain torque for the month of January

1981

a function of

latitude (integrated over longitude) for the mean orography data assimila-

tion cycle (a) and for the envelope orography data assimilation cycle (b),

while Fig.

shows the same for the ensembles of the day

forecasts. The

numerical computation was performed using the

N48

gridpoint model

(analysis or forecast) field of surface pressure

and surface terrain height

and refers, therefore, to the

2ph

envelope orography. The mean zonal torque

per unit area

computed over a (thin) latitude belt of width

was com-

puted using the relationship

-1

cos

8p.-

p.i(hi+l

2nr2

-I

cos

where

and

are, respectively, latitude and longitude,

is the radius of the

earth, the latitudinal integral is taken over one gridbox

(A8

1.875"),

and

the longitudinal integral has been replaced by a sum over a longitudinal

index

1,2,

192.

This method of computing the mountain

torque has well known problems associated with truncation errors in regions

of very steep topography (see Manabe and Terpstra,

1974;

Wahr and Oort,

360

.50c

.“OC

JP’

.20c

.ill

-.om

-.I@

-.m

-.3m

~.UOc

-.Sffi

STEFAN0

TIBALDI

MER

I I

I,,

,--.-

S2 73

-10

-yI

-60

-8Q

-53

LRT

TUDE

ENV

.3?

JL‘

-10

-20

-XI

-yo

-90

LRT

iUCE

FIG.

13.

Analyzed mountain torque

per

unit

area

averaged for the January

198

period

(ensemble of

analyses)

a function

latitude. (a) MEA, Mean orography

used

in the

assimilating model; (b)

ENV,

envelope orography

(2,~~)

used

in the assimilating model.

ENVELOPE OROGRAPHY AND MAINTENANCE

L-

-+,

,-,

.-i-

LRTITUDE

FIG.

14. Analyzed mountain torque per unit area averaged for the January 1981 period

(ensemble of all day

forecasts)

a function of latitude. (a) MEA, Mean orography model;

(b)

ENV, envelope orography

(2&)

model.

362

STEFAN0 TlBALDl

1984), but was employed for convenience and in the light of the compara-

tively high resolution ofthe ECMWFgridpoint model. These problems seem

to become excessive only in the region

the Andes and only in the case of

the much steeper envelope orography. Since we will concentrate here on the

Northern Hemisphere, we are confident that the technique employed is

adequate.

Figure 13 shows the comparison between the torques produced by the

mean and by the envelope orography in the ensemble of 3

initial conditions

for the January 198

set of experiments. Figure

shows the same for the

ensemble of the

day

forecasts. In the Northern Hemisphere midlatitudes

(30" -60"N) the higher negative torque exerted by mountains (mostly the

Rockies) is, using the envelope orography, such

to slow down the lower

tropospheric westerlies (approximately between

100

and

500

mbar) and is,

therefore, consistent with the observed reduction

the model's excessive

westerlies shown in Fig.

The increase of the positive torque taking place in

the Northern Hemisphere subtropical and tropical belt

(Oo

-30"N) is also

quite consistent with the exaggerated easterlies of the mean orography

model, also reduced by the introduction of the envelope orography.

Figure 15 shows the time evolution during the January 198

10-day inte-

grations of the ensemble mean latitudinal integral of the mountain torque

between approximately 30" and 60"N. The mean and envelope orography

MEA

ENV

-301

I I

Forecast

time

(days)

FIG.

15.

Mountain torque

per

unit area averaged for the

forecasts

January

1981

and

over the latitudinal belt

28"-60"N

a function of forecast time. MEA, Mean orography

model;

ENV,

envelope

(24,)

orography model; OBS, appropriate observed winter values (e.g.,

Oort

and Bowman,

1974).

ENVELOPE OROGRAPHY AND MAINTENANCE

363

ensembles are compared with estimates of the real torque derived from

observations (0013 and Bowman, 1974). It is quite clearly shown that the

ensemble of mean orography forecasts underestimates the mountain torque,

while the

2ph

envelope orography overestimates it for the first 6 days. Both

ensembles tend toward more reasonable (and closer) values toward the end

of the forecast period, suggesting that a new dynamical balance has been

attained and that the surface pressure pattern has adjusted to give again a

reasonable total torque. We know, however, from Fig. 6, that these new

balances are attained for different values of the mean zonal wind for that

latitude belt.

If we add to the results of this diagnosis the fact that the smaller scales of the

orography (and of the pressure field) are likely to dominate the generation of

mountain torque, we can conclude that the changes brought about in the

Northern Hemisphere midlatitudinal mountain torque are quite consistent

with the reduced model errors in the zonal mean of zonal wind, indicating

this as

plausible physical mechanism to account for an important part of

the observed impact of the envelope orography.

6. THE EFFECTS

THE

ENVELOPE OROGRAPHY

THE

TROPICAL

REGIONS

Figures

and 6 show that the effects of the envelope orography on the

Southern Hemisphere’s midlatitudes are, as it was reasonable to expect,

smaller in amplitude but essentially of the same nature of those experienced

by the Northern Hemisphere. The impact, however, is largest on the day 7

upper tropospheric and lower stratospheric tropical temperature

SE.

The

zonal-wind SE is also strongly affected in this portion of the model’s atmo-

sphere but, unlike for temperature for which the benefit has vanished by day

10,

the impact seems to last well into the whole forecast range explored. It

becomes, therefore, natural to investigate the impact of the envelope orog-

raphy or the mean meridional circulation and, in particular, on the Hadley

cell.

Figure 16 shows the mean meridional circulation for the two January

198

ensembles at days 1,4,7, and 10. The most conspicuous difference is a

considerable reduction of the spin-up time of the model’s tropical atmo-

sphere. The mean orography ensemble needs

10 days to reach the value of

140 ton sec-’ for the total Hadley cell mass transport, while the envelope

ensemble shows an “overshoot” value of 180 ton sec-’ at day

followed by a

lower value of 130 ton sec-’, then settles down at values around 160 ton

sec-’ during the second

days of the forecast. The existence of a spin-up

phase that is an integral part of the model’s systematic error is a well known