GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO. 7, PAGES 1207-1200, APRIL 1,2001
Mountain
torques and atmospheric oscillations
FranqoisLoft Laboratoire de M6tdorologieDynarnique, Universit6 Pierre et Marie Curie, Paris, France
Andrew W. Robertson
and Michael
Ghil
Department of Atmospheric Sciencesand IGPP, University of California at Los Angeles, USA
Abstract.
Theoretical work and general circulation model [Hide et at., 1980; Hide and Dickey,1991]. Changesin M (GCM) experiments suggest that the midlatitudejet stream's arise either through torques exerted at the lower bound-
interaction with large-scaletopography can drive intraseasonal oscillationsin large-scaleatmosphericcirculation patterns. In support of this theory, we present new observational evidence that mountain-induced torques play a key role in 15-30-day oscillations of the Northern Hemisphere circulation's dominant patterns. The affected patterns in-
ary by small-scale turbulent friction or by surface-pressure differences
across mountains.
On time
scales shorter
than
a season, the relative role of these two factors in affect-
ing M is subject to debate. In the 30-60-day band, Mchangesare driven about equally by the mountain torque TM and by frictional torque TF; changesin TF accompany
clude the Arctic Oscillation(AO) and the Pacific-North- the MJO throughtropicalsurface-windanomalies[Madden, American(PNA) pattern. Positivetorquesboth accelerate 1987; Madden and $peth, 1995; Weickmannet at., 1997]. and anticipate the midlatitude westerly winds at these periodicities. Moreover, torque anomalies anticipate the onsets of weather regimes over the Pacific, as well as the break-ups of hemispheric-scaleregimes.
At periodicitiesbelow 15 days, M-changes are primarily related to mountain torque changesthat accompanysynoptic weather systems as they crossthe ROckiesor the Himalayas
Introduction
jor extratropicaloscillations occur[Branstator,1987;Kush-
[Iskenderianand $alstein,1998]. The intermediate 15-30-day band, however, where ma-
Observational studies show that,
above a broad-band
background,midlatitudelow-frequency variability (LFV) is characterizedby intermittent weatherregimes[Chengand Waltace,1993; Kimoto and Ghil, 1993], and by intraseasonaloscillations[Branstator,1987;Kushnir, 1987; Ghil et al., 1991]. The latter are knownto be modulatedby tropical convection at the 30-60-day time scale of the Madden-
nir, 1987; Ghit and Robertson, 2000, and further references
therein],has not beenexaminedin sufficientdetail. This is surprising,sincein the Northern Hemisphere(NH), significant cross-spectralpeaks between Ta4 and barotropic zonal wind occur at periods above 15 days, while Ta4 also affects
a blockingindex at thesetime scales[Metz, 1985]. Using a much longer and physically more self-consistentdataset
than wasavailableto Metz in 1985,we demonstratethat (i)
JulianOscillation(MJO) [Maddenand Julian,1994;Higgins the 15-30-day band is preciselythe one where the mountain and Mo, 1997],and the 2-6-yeartime scaleof the E1 Nifiotorqueexhibitsits mostsignificantspectralpeaks;(ii) these SouthernOscillation(ENSO) [Rasmusson and Mo, 1993]. spectral peaks in TM are linked to large-scaleatmospheric On the other hand,theoreticalmodelstudies[Charneyand flow patterns; and (iii) changesin TM anticipatethose in DeVore, 1979; Pedlosky, 1981; Legras and Ghil, 1985; Jin
and Ghil, 1990]suggestthat the midlatitudejet stream'sin-
the flow patterns.
teraction with large-scaletopography can drive midlatitude intraseasonal
oscillations in both zonal and non-zonal winds. Results This theory of oscillatory topographic instability is supported by numerical experiments using quasi-geostrophic The 40-yearNCEP/NCAR reanalysisdataset[Kalnayet models with full-sphere geometry and realistic topography al., 1996]is ideal for our purpose.It is a dynamicallycom[Stronget al., 1995]aswell asby GCM simulations[Marcus plete set of meteorological fields for 1958-1997, constructed et at., 1996]. Ghit and Robertson [2000]discuss the topo- with NCEP's current data assimilation system. To assessits graphically driven oscillations' properties acrossa full hier- accuracy we have computed the global angular momentum archy of models. So far, direct observational support for budgetfor 1958-1997 and verifiedthat the tendencydM/dt oscillatory topographic instability has been fairly limited is very similar to the independent estimate of Madden and
[Metz, 1985]. The theory,if correct,impliesthat mountain $peth[1995],basedon ECMWF data. In addition,the total torque Ta4+ T• doesfollow dM/dt very closely(cf. AppendixA). Over the full 40-year,the correlationcoefficient r betweendM/dt and the total torque is r- 0.87.
torques drive a significant fraction of atmospheric variability within an intraseasonal frequency band and might have predictive value. The link between mountain torques and changesin global atmospheric angular momentum M is well established Copyright2001 by theAmericanGeophysical Union. Papernumber2000GL011829. 0094-8276/01/2000GL011829505.00
Our results are presented for the NH, north of 20øN. The
powerspectrumof NH mountaintorque (Figure 1), constructedusingthe multi-tapermethod(MTM seeAppendix B) [Thomson,1982;Derringeret at., 1995;Mann and Lees, 1996],exhibitsfive significantpeaksin the 15-30-dayrange above an almost white background. The MTM spectrum drops sharply for periods longer than 30 days, indicating
1207
1208
LOTT ET AL.: MOUNTAIN TORQUES AND ATMOSPHERIC
25
,
OSCILLATIONS
variations in NH TM are a key driver of midlatitude changes
20
•
15
•
in M, weintegratedriM/dr between20øN and 90øN overthe 15-30-day band. The angular momentum tendency's variations are very close in amplitude and phase to those in NH mountain torque; their correlation at zero lag is r - 0.7. The surface pressure changesassociated with the hemi-
10
sphericEOFs I and 2 are essentiallymeridional(Figures 2a,b), while the patterns that produce a large TM ex-
5
0 0.01
hibit strong zonal gradients of surfacepressureand 700-hPa geopotential height across the Rockies and Himalayas re-
' 0.03
0.05
0.07
0.09
Cycles/day
spectively(not shown). Hence a strongmountaintorque signal cannot be a passiveby-product of the flow changesassociated with PCs I and 2. This observation is entirely consistent with the correlations between TM on the one hand, and PCs I and 2, on the other, being almost zero at zero
Figure 1. MTM spectraof NH TM basedon time seriesof 3day averages,using8 tapersand a resolutionof 1.3710-3cy/day. lag (Figure3). The light (heavy) dashedline givesthe 95% (99%) significance The Pacific sectoffal level.
PC-3's
correlation
with
the moun-
tain torque is also highly significant and showsthe torque to lead the PC. Thus, mountain forcing anticipates changes that the MJO cannot affect the NH mountain torque in a major way. To capture the dominant spatial patterns of atmospheric
in the dominantpattern of Pacific-sector variability (Figure 2c). Thesechanges,in turn, affectthe intensityof the
jet over the northeastern Pacific and thus the angular moflow variability, we take 3-day-mean geopotential height mentum M. In contrast to the hemispheric PCs, the third Pacific EOF is itself associatedwith a large pressurediffermaps of the 700-hPa pressuresurface and compute empirical orthogonalfunctions(EOFs) overthe NH, as well as over ence acrossthe Rockies and, therefore, a mountain torque. the Pacific-North-American(PAC) sector (120øE-60øW, This is consistent with the smaller phase lag between the 20øN-90øN). The first two NH EOFs have a large zonally torque and PAC PC-3, compared to the hemisphericmodes
symmetriccomponent.The first EOF (Figure2a) describes (Figure3).
changesin the midlatitude zonal-wind speed and mass distribution that are associatedprimarily with the subtropical
To check the potential significance of our findings for extended-range prediction, we have isolated the dominant patterns of atmospheric LFV using an analysis of weather
jet and the seasonalcycle. The secondNH EOF (Figure the NH 2b) describesmodulationsin the strengthof the polar vor- regimes[Ghil et al., 1991]. Our analysisreproduces tex and resembles the lower-tropospheric manifestation of
and PA Cregimes found in previousstudiesduring the winter
eastern Pacific, describes an anomalous extension or contraction of the jet stream in this sector. We project the individual 700-hPa height maps onto the
variations in the 15-30-day band contribute to the break-up of both of these regimes. Over the PA C sector,we find that unfiltered torque anomalies anticipate the onset of the two most significant Pacific regimes- which resemble opposite
the AO [Thomsonand Wallace,1998].The third PAC EOF months[Kimoto and Ghil, 1993; Chengand Wallace,1993; regimes, (Figure2c) corresponds to an east-westdipolecenteredover $mythet al., 1999]. The first andthird hemispheric by frequency-of-occurrence, resemble contrasting phases of the Rockiesthat resemblesthe PNA pattern [Wallace and the AO, i.e. of NH EOF 2 (Figure 2b). NH mountain torque Gutzler,1981];its primary centerof action,overthe north-
three EOFs in Figure 2 to obtain the correspondingprinci-
pal components (PCs). Eachof their spectra(not shown) polarities of PA C EOF 3- by a few days.
exhibits significant peaks in the 15-30-day band. To quantify the correspondence between these peaks and those in
the NH TM (Figure 1), we focuson the 15-30-dayrangeby band-pass filteringall seriesbetween15 and 35 days(seeAppendixB). The amplitudesof the four resultingseries(for the torque and the three PCs respectively)are quite substantial: after subtracting the seasonalcycles they account for typically 25% of the varianceof the respectiveunfiltered time series. The NH TM and the PCs show highly significant
Summary Our observationalresultsindicate that large-scalechanges in the extratropical atmosphere with periods of 15-30 days are often anticipated by mountain-torque changes. For the NH EOFs i and 2 we find strong evidence that the moun-
tain torque actively drivesthese changesbecause:(i) the torque leadsthe PCs in phasequadrature;and (ii) both
lag correlationsin all three cases(Figure 3). The correla- EOFs patterns are zonally symmetric to a large degreeand tions
between
the unfiltered
time
series with
the
seasonal
cycleremoved(not shown)are about half as largeas in the
thus very different from flow patterns that would produce a large mountain torque per se. For the Pacific-sector EOF 3,
filtered
we find TM to lead its PC by about 7r/3. This phaserela-
data.
For NH PCs i and 2, the correlations are nearly anti--
tionship is consistentwith PA C PC 3 being associatedwith symmetricwith respectto lag, and lffear-zeroat zero lag. changesin the jet intensity over the northeastern Pacific and This phase quadrature agreeswith the dominance of TM in thus atmospheric angular momentum M. The changesinduced by the torque affect the onset and Eq. (1) for our intraseasonal frequencyband. Positivevalues of TM, i.e. an eastward acceleration of the atmosphere, lead break of important patterns of NH LFV, in particular the positive coefficients of NH EOFs I and 2. Thus stronger AO and the PNA pattern. These observational findings are than usual midlatitude westerly winds follow larger Tx• by entirely consistentwith the theory of oscillatory topographic up to about 10 days. To corroborate that these 15-30-day instability[Legrasand Ghil, 1985;Jin and Ghil, 1990].The
LOTT ET AL.- MOUNTAIN
TORQUES AND ATMOSPHERIC
OSCILLATIONS
1209 ß
Figure 2. EOFs of ?00-hPageopotential heightsevaluatedover1958-1997usingmapsof three-daymeans.a) EOF 1 for NH; b) EOF 2 for NH; and c) EOF 3 for PAC. Negativecontoursare dashed.
theory's predictions have been verified now across a full hi-
forecasts. Gravity-wave stressesare difficult to estimate ac-
erarchyof models[Strongeta/., 1995;Marcus et al., 1996; curately [Loft and Miller, 1997]. They were found to deGhil and Robertson,2000] and providethe most plausible grade the angular momentum balance and were therefore mechanismso far to explain our observationalfindings.
Appendix budget
A: Angular momentum
The budget of angular momentum M is given by dM dt
= T• + T•, ,
(A1)
where TM is the mountain torque and Tr is the friction
torque (seefor instance[Maddenand $•t•, 1995]). Daily zonal-windaveragesu, at the 19 reanalysispressure(p) levels, together with daily averagesof surface pressure were used to compute M. TM consists of an explicit pressure term which involves the zonal gradient of the mountain height h and a gravity-wave stressTo. The latter and the boundary-layer stress TB needed for Tr are estimated usingtheir parameterizedvaluestaken from NCEP's 6-hour
excluded from our calculations. We analyze three-day averages of all quantities and focus on Ta• evaluated using only pressureand topographic height, both of which are directly measured quantities. The 40-year grand mean was subtracted to eliminate systematic biases due to inaccuracies in the parameterized values of TO and TB.
Appendix B' Spectral analysis methods All spectral analysesshown in Figure 1 and mentioned in the text were performed using the SSA-MTM Toolkit
[Derringeret al., 1995]; its latest version,Version4.0, is availableas freewareat http://www. atmos.ucla. edu/tcd/. The band-pass filter used in Figure 3 and elsewhere in the text is based on a Kaiser window and its parameters are
adjustedto minimizeGibbseffects[Hamming,1983; Scavuzzo et al., 1998]. The resultingtransfer functionis very closeto unity for 16-30 days and nearly zero above 44 and below 13 days; its half-power points are at 15 and 35 days. Acknowledgments.
The authors are grateful to asso-
dates on three continents for interesting exchanges on lowfrequency atmospheric variability. Comments from J. O. Dickey, K. Ide, P•. A. Madden, S. L. Marcus, W. Metz, H. L. Swinney, Y. Tian, K. M. Weickmann and an anonymous reviewer helped im-
1.0
prove the presentation. The NCEP/NCAI• l•eanalysis data are providedthrough the N OAA Climate DiagnosticsCenter (h•p ://www.cdc.noaa.gov). This work was supported by NASA's Global Modelingand AnalysisProgram (F.L.), DOE's Officeof Biologicaland EnvironmentalResearch(A.W.I•.), and an NSF SpecialCreativity Award (M.G.). This is publicationno. 5496 of
0.5
0.0
UCLA's
IGPP.
-0.5
References i
-45
-30
-15
0
15
i
I
30
i
i
45
laaldavs)
Figure 3. Lag correlationsbetweenthe NH TM and the PCs in the 15-30-day band.
Solid curve: NH PC-l;
short dashed
curve: NH PC-2; long dashed curve PA C PC-3. Shaded: 99% confidence interval from a Monte Carlo test using correlations at random lags.
Branstator, G. W., A striking example of the atmosphere's leading traveling pattern, J. Atmos. Sci., 44, 2310-2323, 1987. Charney, J. G. and J. G. DeVote, Multiple flow equilibria in the atmosphere and blocking, J. Atmos. Sci., 36, 1205-1216, 1979. Cheng, X. and J. M. Wallace, Cluster analysis of the northern hemisphere wintertime 500-hPa height field: spatial patterns, J. Atmos. Sci., 50, 2674-2696, 1993. Derringer, M.D., Strong, C. M., Weibel, W., Ghil, M. and P. You, Software for singular spectrum analysis of noisy time series,
Eos Trans. A GU 76(œ),12-14-21, 1995.
1210
LOTT ET AL.: MOUNTAIN TORQUES AND ATMOSPHERIC OSCILLATIONS
Ghil, M., Kimoto, M. and J. D. Neelin, Nonlinear dynamics
Mann, M. E. and J. M. Lees, l•obust estimation of background noise and signal detection in climatic time series, Climate Suppl., 29, 46-55, 1991. Change, 33, 409-445, 1996. Ghil, M., and A. W. l•obertson, Solving problems with GCMs: Marcus, S. L., Ghil, M. and J. O. Dickey, The extratropical 40General circulation models and their role in the climate modday oscillation in the UCLA general circulation model. Part II: eling hierarchy. In General Circulation Model Development: Spatial structure. J. Atmos. Sci., 53, 1993-2014, 1996. Past, Present and Future, Edited by D. Randall, pp. 285-325, Metz, W., Wintertime blocking and mountain forcing of the zonally averaged flow: a cross-spectrM time series analysis of obAcademic Press, San Diego, 2000. Hamming, P•. W. Digital Filters, Chapters 7 ge 9, Prentice-Hall, served data, J. Atmos. Sci., •2, 1880-1892, 1985. 1983. Pedlosky, J., l•esonant topographic waves in barotropic and baroHide, 1•. and J.O. Dickey, Earth's variable rotation, Science, 253, clinic flows, J. Atmos. Sci., 38, 2626-2641, 1981. 629-637, 1991. Rasmusson,E. M. and K. C. Mo, Linkages between 200 mb tropHide, 1•., Birch, N. T., Morrison, L. V., Shea, D. J. and A. ical and extratropical circulation anomalies during the 1986A. White, Atmospheric angular momentum fluctuations and 1989 ENSO cycle, J. Climate, 6, 595-616, 1993. Scavuzzo, C. M., Lamfri, M. A., Teitelbaum, H. and F. Lott, changesin the length of day, Nature, 286, 114-117, 1980. Higgins, P•. W. and K. G. Mo, Persistent North Pacific circuA study of the low frequency inertio-gravity waves observed lation anomalies and the tropical intraseasonal oscillation, J. during the Pyrinies Experiment, J. Geophys. Res., 103, 1747Climate, 10, 223-244, 1997. 1758, 1998. Iskenderian, H. and D. A. Salstein, Regional sources of mounSmyth, P., Ide, K. and M. Ghil, Multiple regimes in northern tain torque variability and high frequency fluctuations in atmohemisphere height fields via mixture model clustering, J. Atsphericangularmomentum,Mon. Wea. Rev., 126, 1681-1694, mos. Sci., 56, 3704-3723, 1999. 1998. Strong, C., Jin, F. F. and M. Ghil, Intraseasonal oscillations in Jin, F.-F. and M. Ghil, Intraseasonal oscillations in the extratropa barotropic model with annual cycle, and their predictability, ics: Hopf bifurcation and topographic instabilities, J. Atmos. J. Atmos. Sci., 52, 2627-2642, 1995. Thomson, D. J., Spectrum estimation and harmonic analysis, Sci., J7, 3007-3022, 1990. Kalnay E., M. Kanamitsu, P•. Kistler, W. Collins, D. Deaven, L. Proc. IEEE, 70, 1055-1096, 1982. Gandin, M. Iredell, S. Saha, G. White, J. Woollen, Y. Zhu, M, Thompson, D. W. and J. M. Wallace, The Arctic Oscillation signature in the wintertime geopotential height and temperature Cheillab, W. Ebsuzaki, W. Higgins, J. Janowiak, K. C. Mo, field, Geophys. Res. Left., 25, 1297-1300, 1998. C. l•opelewski, J. Wang, A. Leetma, P•. Reynolds, 1•. Jenne and D. Joseph,The NCEP/NCAP• 40-year reanalysisproject, Wallace, J. M. and D. S. Gutzler, Teleconnectionsin the geopotentiM height field during the Northern-Hemisphere winter, Bull. Amer. Meteor. Soc., 77, 437-470, 1996. Mon. Wea. Rev., 109, 784-812, 1981. Kimoto, M. and M. Ghil, Multiple flow regimes in the Northern Weickmann, K. M., Kiladis, G. N. and P. D. Sardeshmukh, The Hemisphere winter, J. Atmos. Sci., 50, 2625-2673, 1993. dynamics of intraseasonM atmospheric angular momentum osK ushnir, Y., Retrograding wintertime low-frequencydisturbances cillations, J. Atmos. Sci., 5•, 1445-1461, 1997. over the North Pacific Ocean, J. Atmos. Sci., 44, 2727-2742,
andpredictability in the atmospheric sciences, Rev.Geophys.
1987.
Legras, B. and M. Ghil, Persistent anomalies, blocking and variations in atmospheric predictability, J. Atmos. Sci., •2, 433471, 1985.
Loft, F. and M. J. Miller, A new subgrid-scaleorographic drag parameterization: Its formulation and testing, Q. J. R. Meteorol. Soc., 123, 101-127, 1997.
Madden, P•. A., Relationships between changesin the length of day and the 40 to 50 day oscillation in the tropics, J. Geophys. Res., 92, 8391-8399, 1987.
F. Loft, Laboratoire de M•t•orologie Dynamique du CNi•S, Universit• Pierre et Marie Curie, 4 place Jussieu, Boite 99, 75252
Paris, France. (e-mail:
[email protected]) M. Ghil and A. W. l•obertson, Department of Atmospheric Sciencesand Institute of Geophysicsand Planetary Physics, University of California at Los Angeles, 405 Hilgard Avenue, Los An-
geles,California 90095-1565,USA. (e-mail:
[email protected]. edu;
[email protected])
Madden, P•. A. and P. 1•. Julian, Observations of the 40-50 day tropical oscillation-A review, Mon. Wea. Rev., 122, 814-837, 1994.
Madden, 1•. A. and P. Speth, Estimates of atmospheric angular momentum, friction, and mountain torques during 1987-88, J. Atmos. Sci., 52, 3681-3694, 1995.
(ReceivedMay 26, 2000; revisedOctober 18, 2000; acceptedNovember20, 2000.)