Variability in the altitude of fast upper tropospheric

Variability in the altitude of fast upper tropospheric winds over the

Northern Hemisphere during winter

Courtenay Strong1,2 and Robert E. Davis1

Received 14 July 2005; revised 3 November 2005; accepted 13 February 2006; published 31 May 2006.

[1] The surface of maximum wind (SMW) is used as a frame for examining spatial andtemporal variability in the vertical position of fast upper tropospheric winds over theNorthern Hemisphere during winter. At a given observation time in a gridded data set, theSMW is defined as the surface passing through the fastest analyzed wind above each gridnode, with a vertical search domain restricted to the upper troposphere and anytropospheric jet streams extending into the lower stratosphere. Documenting how SMWpressure varies spatially and temporally guides the operational and climatological analysisof tropospheric jet streams and enables more informed interpretation of isobaric windspeed signals (i.e., isobaric speed variability can be caused by both jet core pressurechanges and jet core speed changes). Using six-hourly Northern Hemisphere (NH) NCEP-NCAR reanalysis data from 1958 to 2004, the mean three-dimensional structure of the NHSMW is mapped for winter, and vertical cross sections are used to show the position of theSMW relative to the mean isotachs, the mean lapse-rate tropopause, and the meandynamic tropopause. The Arctic Oscillation (AO) and El Nino/Southern Oscillation(ENSO) are identified as the leading contributors to SMW pressure variability over theNH using principal components analysis. During the AO positive phase, mean SMWpressures are decreased at high latitudes and increased at middle and subtropical latitudes.The SMW response to the AO is elliptical with the centers of pressure change displacedmore equatorward over Asia and the Atlantic. During the warm phase of ENSO,eastward expansion and strengthening of the Pacific jet stream is associated with SMWpressures up to 27 hPa lower near 30�N. The relatively low SMW pressures of the eastPacific SMW are flanked, during the warm phase of ENSO, by slower upper troposphericwinds and higher SMW pressures near the equator and the Gulf of Alaska.

Citation: Strong, C., and R. E. Davis (2006), Variability in the altitude of fast upper tropospheric winds over the Northern

Hemisphere during winter, J. Geophys. Res., 111, D10106, doi:10.1029/2005JD006497.

1. Introduction

1.1. Perspective

[2] Much effort has been aimed at documenting andunderstanding variability in the horizontal position of con-centrated bands of fast upper tropospheric winds known asjet streams. Knowing the jet stream’s location has provenvaluable to aviation [Reiter, 1958] and short- and long-termforecasting [Riehl et al., 1954]. Jet stream variability alsoinfluences stratosphere-troposphere exchange [Hoerling etal., 1993; Langford, 1999; Wimmers et al., 2003], theintensity and track of synoptic-scale storms [Nakamura,1992; Christoph et al., 1997], the development of severeconvective storms [Uccellini and Johnson, 1979; Kloth andDavies-Jones, 1980], the occurrence of heavy rain [Smithand Younkin, 1972] and snow [Uccellini and Kocin, 1987],

and the development of tornados [Fawbush et al., 1951;Skaggs, 1967]. Jet streams also produce gravity wavesassociated with clear air turbulence [Knox, 1997] and polarstratospheric clouds [Hitchman et al., 2003; Buss et al.,2004].[3] In addition to having immediate practical relevance,

jet stream position variability can be an important indicatorof climate change. Extratropical jet streams are closely tiedto the polar front [Palmen, 1948; Palmen and Newton,1948] and strengthen and expand equatorward during win-ter [Namias and Clapp, 1949]. On interannual or longertime scales, changes in the thermal structure of the atmo-sphere can exert an influence on the shape and intensity offast tropospheric flows. The linkage between temperatureand the geometry and location of the fast mid- to upper-tropospheric circulation has been demonstrated throughstudy of the circumpolar vortex [Angell and Korshover,1977; Davis and Benkovic, 1992, 1994; Burnett, 1993].Vortex variations at 700, 500 and 300 hPa collectivelyaccount for almost two-thirds of midlatitude tropospherictemperature variability [Frauenfeld and Davis, 2003].[4] Because of the jet stream’s close relationship to

tropospheric and stratospheric thermal fields, it has been

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D10106, doi:10.1029/2005JD006497, 2006

1Department of Environmental Sciences, University of Virginia,Charlottesville, Virginia, USA.

2Now at Department of Soil, Environmental, and AtmosphericSciences, University of Missouri-Columbia, Columbia, Missouri, USA.

Copyright 2006 by the American Geophysical Union.0148-0227/06/2005JD006497$09.00

D10106 1 of 15

observed to vary in conjunction with climate phenomenathat are linked to temperature variability. The Arctic Oscil-lation (AO), for example, is the meridional seesaw ofatmospheric mass between the polar region and midlatitudes[Lorenz, 1951; Kutzbach, 1970] manifested as the leadingprincipal component of wintertime sea level pressure[Thompson and Wallace, 1998]. For zonally averaged dataover the Northern Hemisphere (NH), the positive phase ofthe AO is associated with a column of tropospheric warm-ing near 45�N, cooling in the upper troposphere andstratosphere over the polar region, and slower (faster) zonalwind speeds near 30�N (60�N) from the surface into thestratosphere [Thompson and Wallace, 2000].[5] Planetary circulation is also influenced by the El

Nino/Southern Oscillation (ENSO) – the coupled variationof sea surface temperature and sea level pressure over thetropical Pacific [Bjerknes, 1969]. The ENSO warm phasecorresponds to anomalously low pressure and high surfacewater temperatures over the eastern Pacific. Associatedtropical phenomena include weakening or reversal of theeasterly trade winds [Ichiye and Petersen, 1963], increasesin tropical tropospheric temperature [Horel and Wallace,1981; Yulaeva and Wallace, 1994; Sobel et al., 2002], watervapor mass, and precipitation intensity [Soden, 2000]. AnENSO-related intensification and equatorward displacementof the mid- to upper-tropospheric circulation over the NHeastern Pacific has been detected in upper tropospheric windspeed [Arkin, 1982; Rasmusson and Mo, 1993; Yang et al.,2002], middle and upper tropospheric geopotential heighttopography [Horel and Wallace, 1981; Hastenrath, 2003],storm tracks [Chen and Van den Dool, 1999], second orderstatistics indicative of storm track regimes [Straus andShukla, 1997], and the circumpolar vortex [Angell andKorshover, 1985; Frauenfeld and Davis, 2000; Angell,2001]. The influence of ENSO on extratropical circulationis also manifested as variations in the equatorial zonalWalker cell, the tropical meridional Hadley cell, the extra-tropical meridional Ferrel cell, and the midlatitude zonal cell[Wang, 2002].[6] The vertical position of fast upper tropospheric wind

features has also received some attention. Initial interest injet stream altitude produced cross sections of zonallyaveraged summer and winter wind speed data from alllatitudes [Mintz, 1954], physical explanation of the jetstream’s location [Staff Members, 1947], and detailed map-ping of the layer of maximum wind’s altitude relative to jetstream axes [Reiter, 1958]. The U.S. Navy began opera-tionally mapping the altitude and vertically averaged speedof the three-dimensional layer of maximum wind (LMW)beginning at the end of 1958 for aviation purposes [Reiter,1961].[7] Little is known, however, about climatological vari-

ability in the pressure of jet streams and other uppertropospheric wind maxima. Significant interannual variabil-ity in the pressure of tropospheric jet streams is an intrin-sically interesting and important phenomenon in theatmosphere. From a practical perspective, mapping howthe pressures of upper tropospheric wind maxima varyspatially guides operational and climatological analyses oftropospheric jet streams, and documenting how the pres-sures of upper tropospheric wind maxima vary temporallyenables more informed interpretation of isobaric wind speed

signals. Climate studies considering upper troposphericwind speed variability often utilize an isobaric surfacepresumed to lie close to the jet stream level. Some inaccu-racy may result when data obtained from a single isobaricsurface are used to evaluate climate trends in jet streams orfast upper tropospheric winds in general [Strong and Davis,2005]. At any one time, no isobaric surface can capture a jetstream whose altitude varies spatially in the hemisphere,and, at any one location, no isobaric surface can capture ajet stream whose vertical position varies temporally.[8] In prior research, we identified statistically significant

temporal trends in wind maxima pressure (as large as30 hPa decade�1) over the Northern Hemisphere tropicsfor summers 1958–2004, indicating decadal jet streamdescent consistent with observed changes in the temperaturestructure of the troposphere and lower stratosphere [Strongand Davis, 2006]. Here, we document spatial and temporalvariability in the vertical location of wind maxima close tothe tropopause during winter, including those associatedwith the tropopause jet streams. In the Introductionsection 1.2, we give theoretical consideration to the rela-tionship between horizontal temperature gradients and thevertical location of wind speed maxima. Section 2 describesthe data and the surface of maximum wind (SMW) [Strongand Davis, 2005] used to track the pressure of upper tropo-spheric wind maxima. Spatial variability of the SMW isconsidered in section 3.1 and principal contributors to jointspace time variability of the extratropical and tropical SMWare considered in section 3.2.

1.2. Theory

[9] The geostrophic wind speed’s rate of change withrespect to pressure can be expressed as (details inAppendix A)

@Vg

@ ln p¼ �Rd

f~rTv

�� cos b ¼ Rd

f

@Tv@n

ð1Þ

where Rd is the dry air gas constant, f is the Coriolisparameter, Tv is virtual temperature, the del operator isapplied on an isobaric surface, b is the angle between thegeopotential gradient (� ~rF) and the virtual temperaturegradient (� ~rTv), and n is oriented normal to ~Vg and definedpositive to the right of the flow direction. Equation (1)indicates that wind speed will increase with height (@Vg/@ lnp will be negative) when higher temperatures on an isobaricsurface are collocated with the higher values of geopotential(i.e., b < p/2). The temperature gradient in the b < p/2 caseis associated with stronger gradients of geopotential in theoverlying layer in accordance with the hypsometricrelationship, supporting an increase in the overlyinggeostrophic wind speed. In the b � p/2 case, lowertemperatures are collocated with higher values of geopo-tential, indicating weaker overlying geopotential gradientsand geostrophic wind speeds.[10] Since k ~rTvkcos b and @Tv/@n are equivalent, the

more compact notation @Tv/@n will be used in subsequentsections of this manuscript. In this section, however, thebehavior of equation (1) in the atmosphere will be consid-ered using the k ~rTvkcos b form to stress that equation (1)may be forced to zero (marking the local wind maxima) byeither the angle between the geopotential and temperature

D10106 STRONG AND DAVIS: FAST UPPER TROPOSPHERIC WIND ALTITUDE

2 of 15

D10106

gradients or the magnitude of the temperature gradientitself. As an example, Figure 1a shows a cross section ofvirtual temperature isopleths and isotachs along 47.5�E on1 January 2000 from the NCEP-NCAR Reanalysis [Kalnayet al., 1996]. The virtual temperature gradients associatedwith the polar front integrate vertically to produce theisobaric geopotential gradients supporting the polar frontjet stream (PFJ: �250 hPa, 45.7�N). Likewise, the isobaricvirtual temperature gradients associated with the subtropicalfront integrate vertically to produce the isobaric geopoten-tial gradients contributing to the subtropical jet stream (STJ:�200 hPa, 22.5�N). The STJ also receives energy fromtropical heating and the conservation of angular momentum

in the upper, poleward branch of the Hadley circulation[Krishnamurti, 1961a, 1961b].[11] Figures 1b–1d show profiles of the terms from

equation (1) evaluated on the dashed vertical line passingthrough the PFJ core in Figure 1a. The magnitude of theisobaric virtual temperature ascendent (k ~rTvk) was nonzeroand fluctuated around 6 10�6 K m�1 below the PFJ core(Figure 1b; Figure 1a shows the virtual temperature iso-pleths associated with the @Tv/@y component of k ~rTvk).The geopotential gradients were generally aligned with thevirtual temperature gradients up to 250 hPa (b < p/2,Figure 1b), indicating an increase in geopotential gradientmagnitude with height (positive k ~rTvkcos b, Figure 1c),

Figure 1. (a) Meridional cross section showing virtual temperature isopleths (labels in K) and isotachs(grayscale shading every 7 m/s up to 56 m/s) along 47.5�E on 1 January 2000. Bold lines show theposition of the polar front, the subtropical front, and the Reanalysis tropopause. The dashed vertical lineshows the position of the profiles in Figures 1b–1d (47.5�N, 47.5�E). (b) The magnitude of the isobaricvirtual temperature ascendent (k ~rTvk), and the angle b between ~rTv and ~rF. (c) From equation (1), thechange of the geostrophic wind speed with respect to �ln p so that a positive value indicates speedincreasing with height. (d) Reanalysis wind speeds. The shaded box across Figures 1b–1d shows thelayer where the wind maximum likely occurred.


3 of 15

D10106

and wind speed increasing with height (Figure 1d). At somepressure between 250 and 200 hPa, the virtual temperaturegradient was nonzero but orthogonal to the geopotentialgradient (b = p/2, Figure 1b), the first derivative of windspeed with respect to pressure went to zero (Figure 1c), andthe wind speed reached a local maximum (Figure 1d).Above 200 hPa, the virtual temperature gradients opposedthe geopotential gradients (b > p/2, Figure 1b), and windspeed decreased with height (Figure 1d).[12] The ageostrophy prevalent in the vicinity of jet

streams [Newton, 1959; Shapiro and Kennedy, 1981] vio-lates the assumption of geostrophic balance, so equation (1)will not always provide a reliable estimate of the truechange of wind speed with pressure. Nonetheless, thegeostrophic thermal wind relationship indicates how tem-perature influences the vertical distribution of geopotentialgradients and is thus a useful theoretical model for inves-tigating jet core pressure variability. The pressure variabilityof wind maxima is determined in this research withoutassuming geostrophic balance (i.e., the wind field will beused rather than the fields of geopotential and temperature).Equation (1) will then be used to link the observed windmaxima pressure variability to atmospheric temperaturevariability. A change in jet core pressure represents a changein the position of the local wind speed maximum on thepressure axis and is therefore associated with nearbychanges in the first derivative @Vg/@ ln p.

2. Data and Methods

2.1. Data

[13] We use wind speed on seven isobaric surfaces (500,400, 300, 250, 200, 150, and 100 hPa) resolved every sixhours on a 2.5� grid from the NCEP-NCAR Reanalysis, andReanalysis lapse-rate tropopause pressure data at the sametemporal and horizontal spatial resolution. In addition tousing the Reanalysis lapse-rate tropopause, we also identifya ‘‘dynamic tropopause’’ based on the rapid increase inpotential vorticity between the troposphere and stratosphere[Danielson, 1968; Danielson and Hipskind, 1980; Hoerlinget al., 1991; Morgan and Nielsen-Gammon, 1998]. Tolocate the dynamic tropopause, we use Reanalysis dataand an expression for Ertel’s potential vorticity in isobariccoordinates [Bluestein, 1992]

Pq ¼ �g zþ f þ Rd

sp@v

@p

@T

@x� @u

@p

@T

@y

� �� p

@q@p

; ð2Þ

where relative vorticity z = @v/@x � @u/@y, q is potentialtemperature, and s is a static-stability parameter

s ¼ �RT

p

@ ln q@p

: ð3Þ

Various values of Pq have been suggested for locating thedynamic tropopause ranging from 1 to 3.5 PVU [Hoerling etal., 1991; Spaete et al., 1994], where 1 PVU is defined as10�6 m2 s�1 K kg�1. In section 3, we calculate and show thewinter mean pressure of the 1, 2, 3, and 3.5 PVU isopleths.[14] Artificial trends or discontinuities may exist in the

Reanalysis because of variations in the observing systemincluding changes in the density of rawinsonde data and the

use of satellite data after 1978 [Kistler et al., 2001].Kanamitsu et al. [1997] found that satellite data impactedthe Northern Hemisphere troposphere Reanalysis less thanit impacted data sparse areas such as the stratosphere andeastern oceanic areas of the Southern Hemisphere. Thepotential for perceiving an artifact of changes to the Re-analysis observing system as climate variability is reducedhere because conclusions are not based solely on temporalvariability in the Reanalysis. Rather, results obtained fromthe Reanalysis are contextualized through correlation withexisting, well-studied climate phenomena and indices.[15] Climate indices or time series employed here include

the Arctic Oscillation Index (AOI), the Multivariate ElNino/Southern Oscillation Index (MEI), and the mean NHsurface temperature (TNH). AOI values were obtained fromthe Climate Prediction Center, and were developed byprojecting the monthly mean 1000 hPa geopotential heightanomalies onto the first principal component from theprincipal component analysis (PCA) of the monthly mean1000 hPa geopotential height field north of 20�N. MEIvalues were obtained from the Climate Diagnostic Center,and capture variability in sea-level pressure, zonal andmeridional components of the surface wind, sea surfacetemperature, surface air temperature, and total cloudinessfraction of the sky over the tropical Pacific [Wolter andTimlin, 1993, 1998]. Monthly mean TNH data were obtainedfrom the NASA Goddard Institute for Space Studies, andwere calculated by combining meteorological station mea-surements over land with sea surface temperatures obtainedprimarily from satellite measurements [Reynolds and Smith,1994; Smith et al., 1996]. We averaged the December–February monthly values for each predictor time series forwinters 1958–2004 to match the temporal structure of thesurface of maximum wind winter seasonal means.

2.2. Surface of Maximum Wind

[16] We illustrate the algorithm for locating the uppertropospheric SMW using data from 1200 UTC on 9 January2004 along 175�E (Figure 2). Data in the upper troposphereand lower stratosphere along the meridian are first evaluatedfor SMW candidacy, and then the fastest wind speed fromthe candidates above each node is selected as a point on theSMW. From the column of data above a node (above a tickmark on the abscissa in Figure 2), each wind speed from500 to 100 hPa is a SMW candidate unless the datum isabove the measurement level closest to the lapse-ratetropopause and outside of a 25.7 m/s (50-knot) contour thatcaptures at least one tropospheric datum. In other words, thesearch domain for the SMW is restricted to the uppertroposphere and any tropospheric jet streams that extendinto the lower stratosphere.[17] For the example in Figure 2, points that are not SMW

candidates are left blank, points that are SMW candidatesare shown with open circles, and the SMW is shown withsolid circles. No SMW candidates appear outside 500 to100 hPa, and all points from 500 hPa to the lapse-ratetropopause are SMW candidates. The point at 40�N and200 hPa is a SMW candidate because it is bounded by a25.7 m/s wind speed contour that captures at least onetropospheric datum. The point at 87.5�N and 100 hPa is not aSMW candidate because its 25.7 m/s contour does notcapture at least one tropospheric datum. Properties associated


4 of 15

D10106

with the SMW will be denoted by a tilde so that SMWpressure is eP and SMW wind speed is eV .[18] A dynamic tropopause based on isentropic potential

vorticity [Hoerling et al., 1991, and references therein]could alternatively be used in applying the SMW method.The Reanalysis lapse-rate tropopause is used here becauseit is also viable in the tropics. The algorithm presented inthis section could likewise be adapted to locate the surfaceof maximum geostrophic wind using the fields of temper-ature and geopotential to interpolate the pressure whereequation (1) reaches zero near the analyzed geostrophicwind maximum.[19] It is the SMW’s flexible upper bound that distin-

guishes it from the operationally defined layer of maximumwind (LMW) [Reiter, 1958] and the maximum wind layer(MWL) available in the Reanalysis. The MWL has a fixedvertical domain between 500 and 70 hPa, and is shown forcomparison in Figure 2. In this example, differences be-tween the SMWand MWL in the tropics and through the jetcore near 30�N result from the MWL’s use of cubic splinesfor interpolation. Differences in higher latitudes are moresubstantial and result from the SMW algorithm’s exclusionof stratospheric wind data outside a tropospheric jet stream(25.7 m/s contour). In particular, the winter polar night jetstream is excluded by the SMW algorithm. In the summer,the MWL tends to capture stratospheric jets excluded by theSMW over the tropical lapse-rate tropopause and the SMWand MWL have better agreement in mid- to high-latitudes(not shown).

2.3. Statistical Methods

[20] Principal components analysis (PCA) is used toidentify uncorrelated patterns of simultaneous anomaliesthat account for variability in winter mean eP. A separatePCA is performed in the tropics (equatorward of 20�N) and

the extratropics (poleward of 20�N). For each PCA, thevariables are Reanalysis grid nodes and the cases are theseasonal mean SMW pressures at each node over 47 years.Since jet stream level thicknesses may be more than 10%larger in the tropics than the polar region, the correlationmatrix is used as input into the PCA (rather than thevariance-covariance matrix). To ensure proper representa-tion in the correlation matrix, data are standardized andweighted by the square root of the cosine of latitude.Singular value decomposition is used to obtain the eigen-values and unit-length eigenvectors of the area-weightedcorrelation matrix. The time series produced by projectingthe weighted, standardized data onto the mth eigenvectorwill be referred to as ‘‘principal component m’’ (PC m), andthe contoured map of the mth eigenvector’s elements will bereferred to as the PC m ‘‘loading pattern.’’ For display andcomparison with existing climate indices, each principalcomponent time series is standardized by subtracting itsmean and dividing by its standard deviation. The varianceaccounted for by PC m is eigenvalue m divided by thesummation of all the correlation matrix eigenvalues.[21] Correlations are measured with least squares linear

regression and reported p statistics are based on degrees offreedom (df) adjusted to account for Pearson lag-1 autocor-relation in the regression residuals. Where the results ofcorrelation tests are mapped (Figures 5a and 8a), thefraction of the map area with significant results is verifiedto be field significant at a = 0.05 using Monte Carlosimulations [Livezey and Chen, 1983] confirming there isless than a five percent probability that the amount of areawith significant local results occurred by chance given thefield’s spatial degrees of freedom.

3. Results

[22] Section 3.1 is focused on the spatial variability of thewinter SMW. In section 3.2, PCA is used to uncover the AOand ENSO as principal contributors to SMW joint space-time variability. The relationship between the SMWand AOis explored in section 3.3, and the relationship between theSMW and ENSO is explored in section 3.4.

3.1. SMW Spatial Variability

[23] Figures 3a and 3b show the 47-year mean winter ePand eV for all locations over the NH. SMW pressuresgenerally decrease from the pole toward the position ofthe polar front jet (PFJ) in the midlatitudes. The winterSMW maintains a broad shelf of high pressure in the Arcticbecause of the expanded polar high, and is steeply slopednear the fast jet stream cores in the midlatitude westernPacific, off the east coast of the United States and east of theMediterranean Sea (Figure 3a). Just equatorward of thesteep SMW pressure decreases, high values of eV associatedwith the presence of jet cores are found in the west Pacific,eastern United States, and to the south of the MediterraneanSea (Figure 3b). The depression in the SMW topography(area of relatively high pressure) in the equatorial westernPacific (Figure 3a) is associated with a corridor of relativelylow eV (Figure 3b) and will be discussed in the context ofENSO in section 3.4.[24] The cross section in Figure 3c shows how the

47-year mean SMW relates to the 47-year mean lapse-rate

Figure 2. Wind speed (contours), lapse-rate tropopausepressure, the maximum wind level (MWL), and surface ofmaximum wind (SMW) along 175�E at 1200 UT on9 January 2004. Open circles show data that are candidatesfor being SMW pressures, and solid circles show the SMW.The gray areas show wind speeds greater than or equal to25.7 m/s (50 knots).


5 of 15

D10106

tropopause and isotachs using data from 135�E as anexample. The seasonal mean SMW at 135�E is below thetropopause at the equator, undulates in the tropical latitudes,and extends laterally poleward through the lapse-rate tro-popause in the vicinity of the western Pacific jet’s 70 m/smean isotach (200 hPa, 35–45�N, Figure 3c). The range ofwinter eP at 135�E is minimized close to the jet core at32.5�N, reflecting the stability of this feature’s seasonalmean pressure.[25] The SMW also has a structural relationship to the

dynamic tropopause. Figure 3d shows various isosurfacesused to represent the dynamic tropopause ranging from 1to 3.5 PVU with gray shading over intervening pressures.The SMW’s relationship to the Reanalysis lapse ratetropopause is similar to the SMW’s relationship to the3.5-PVU dynamic tropopause recommended for the extra-tropics by Hoerling et al. [1991] (compare Figures 3c and3d). Specifically, the SMW resides below the 3.5-PVUdynamic tropopause in the tropics, extends through it intothe lower stratosphere at the 70-m/s wind speed maximum

near 35�N, and resides generally below it over highlatitudes.[26] The tendency for a nearly vertical tropopause near jet

streams is well known [e.g., Platzman, 1949; Defant andTaba, 1957; Morgan and Nielsen-Gammon, 1998], and thedynamic tropopause and lapse-rate tropopause are bothsteeply sloped through the jet core near 35�N. The SMW,by contrast, is most steeply sloped poleward of the fastestwind speeds (Figure 3c, also seen in the contour packing inFigure 3a just inside the spiral-shaped corridor of fast windsin Figure 3b).

3.2. Contributors to SMW Temporal Variability

[27] In the extratropics, the first principal component ofwinter mean SMW pressure (extratropical eP PC 1,Figure 4a) accounts for 28% of the extratropical variance.The eP PC 1 loading pattern has oppositely-signed rings inthe polar and midlatitudes, suggesting an AO-related phe-nomenon, and produces scores correlated with the AOI(Figure 4b). The AO has its strongest positive and negative

Figure 3. (a) Mean surface of maximum wind pressure (eP) and (b) mean surface maximum wind speed(eV ) for winters 1958–2004. The lines in Figures 3a and 3b mark the meridian used for the cross sectionsin Figures 3c and 3d. (c) Cross section showing eP (bold dashed line) with mean winter tropopausepressure (bold solid line) and isotachs (dotted lines at 10 m/s interval) along 135�E for winters 1958–2004. Dashed lines bound the 10th to 90th percentile of eP and the gray area bounds the 10th to 90thpercentile of seasonal mean tropopause pressure. (d) Same as Figure 3c, except the tropopause is replacedwith the dynamic tropopause as shown by isopleths of 1, 2, 3, and 3.5 potential vorticity units (PVU) withshading between 1 and 3.5 PVU.


6 of 15

D10106

loadings (Figure 4a: labeled +AO and �AO, respectively)where the extratropical eP PC 1 loadings are relatively weak.The AO and eP loading centers of action have a latitudinaloffset because changes in the AO reflect temperaturevariability (related to pressure variability) whereas changesin eP reflect temperature gradient variability. In the tropics,the first principal component of winter mean SMW pressure(tropical eP PC 1, Figure 4c) accounts for 14% of the tropicalvariance. The loading pattern suggests an ENSO-like phe-nomenon with a pair of oppositely signed loading centerswhere convection shifts from Indonesia toward the centralPacific during El Nino, and the loading pattern producesscores correlated with the MEI (Figure 4d).[28] A PCA was also performed using 47 years of winter

mean SMW pressure at each node over the entire NH asinput (not shown). An AO-like phenomenon was again theleading contributor to SMW variability (NH eP PC 1 [20%of the NH variance], score versus AOI: p < 0.01, r2 = 0.64,

df = 43, mean square error [MSE] = 0.36) followed by anENSO-like phenomenon (NH eP PC 2 [9% of the hemi-spheric variance], score versus MEI: p < 0.01, r2 = 0.29, df =29, MSE = 0.71). The leading principal component’scorrelation with the AOI does not markedly depend onthe inclusion or exclusion of the tropics during the PCA.However, the leading principal component of the tropical-only PCA is better correlated with the MEI than the secondprincipal component of the hemispheric analysis, perhapsbecause manifestations of ENSO outside the tropics canvary among events [Rasmusson and Wallace, 1983]. Thethird principal component of the hemispheric PCA issignificantly correlated with mean NH surface temperature(NH eP PC 3 [8% of the hemispheric variance], score versusTNH: p < 0.01, r2 = 0.36, df = 34, MSE = 0.64). Attentionhere will be restricted to the AO (section 3.3) and ENSO(section 3.4) as the two leading contributors to SMW spacetime variability.

Figure 4. (a) Loading pattern of the first principal component (PC 1) of winter mean surface ofmaximum wind pressure (eP) in the extratropics (unshaded area poleward of 20�N). Contour interval is0.01 with negative values dashed. The +AO and –AO labels relate to the loading pattern of the ArcticOscillation and are referenced in the text. (b) Time series of the extratropical eP PC 1 score and the ArcticOscillation Index. (c) Loading pattern of tropical (unshaded area from 0–20�N) eP PC 1 contoured at 0.01with negative values dashed. (d) Time series of the tropical eP PC 1 score and the multivariate El Nino/Southern Oscillation Index (MEI). The effective degrees of freedom are denoted ‘‘df’’ and the meansquare error is denoted ‘‘MSE.’’


7 of 15

D10106

3.3. SMW and the AO

[29] SMW pressure tends to seesaw with the AO, and themap of the eP versus AOI correlation is field significant ata = 0.05 (Figure 5a). During the positive phase of the AO,SMW pressures are generally decreased over the highlatitudes (negative eP versus AOI correlation) and increasedover the midlatitudes or subtropics (positive eP versus AOIcorrelation). Figure 5a has been divided into five sectorsbased on the pattern of the SMW versus AO relationshipand the nature of the underlying tropospheric temperaturechanges associated with the AO (shown later). The patternof eP versus AOI correlation is elliptic with a longer axisoriented through the Atlantic and Asian sectors (Figure 5a).The ellipticity of the SMW response to the AO is furtherillustrated in Figure 5b where we show the SMW pressureunder positive AO minus the SMW pressure under negativeAO for a sample meridian from each of the five sectors. TheAtlantic and Asian zones of AO-related SMW pressurechanges (�55�N and 20�N, respectively) are displacedequatorward from the corresponding belts of SMW pressurechange in the East Pacific and European sectors (Figure 5b).The SMW relationship to the AO in the Central Pacificsector is relatively weak, with only isolated significantdifferences poleward of 20�N.[30] The ellipticity of the SMW response and the presence

or absence of accompanying upper tropospheric wind speedchanges can be understood by examining AO-relatedchanges in the thermal structure of the atmosphere. Thisis illustrated in Figures 6 and 7, beginning with the EastPacific sector and moving east through the Asian sector (thetemperature changes in the Central Pacific sector areweaker, but spatially similar to, those in the Atlantic sectorand will not be shown separately). Some comments on the

organization of Figures 6 and 7 will facilitate their discus-sion. Each column shows data from one of the four sectorsin a meridional cross section collocated with an arrow at theequator in Figure 5a. The three panels in each column show,from the bottom to the top, the AO-related change intemperature (DT , bottom panel), the AO-related change inthe �Rdf

�1@Tv=@n term from equation (1) (middle panel),the AO-related change in wind speed (DV , upper panel),and the mean SMW during the positive phase of the AO andduring the negative phase of the AO (also upper panel).[31] In the East Pacific at 120�W (Figure 6, left column),

the temperature changes associated with the AO are largelystratospheric, consisting of a high latitude cooling andmidlatitude warming (Figure 6c). The intervening increasein @Tv/@n near the tropopause at 70�N (Figure 6b) raises thealtitude where @Tv/@n goes to zero and is associateddecreased values of eP (Figure 6a). Likewise, decreased@Tv/@n near the tropopause at 30�N (Figure 6b) is associ-ated with increased SMW pressures (Figure 6a). Althoughthe pressure of the SMW varies with the AO in the EastPacific sector, there is little change in wind speed at theSMW (Figure 6a) because weak changes in @Tv/@n throughthe depth of the underlying troposphere (Figure 6b) verti-cally integrate to cause only weak changes in upper tropo-spheric geopotential gradients. AO-related wind speedchanges do occur over the East Pacific, but are predomi-nantly stratospheric with faster speeds above the column ofpositive D@Tv=@n and slower speeds above the column ofnegative DTv=@n (Figures 6a and 6b).[32] The Atlantic sector (Figure 6, right column) also

exhibits high latitude cooling north of warming, but here thetemperature changes appear in the stratosphere and tropo-sphere (Figure 6f). The deep temperature changes over the

Figure 5. (a) Pearson correlation coefficient (r) for winter (December–February) mean surface ofmaximum wind pressure (eP) versus the Arctic Oscillation Index (AOI) for 1958–2004. Contours are at0.3 with negative values dashed. Results locally significant at a = 0.05 account for 53% of the map area(including, but not limited to, all of the contoured area), and the map of results is field significant at a =0.05. The hemisphere is divided into sectors (e.g. East Pacific) discussed in the text. (b) For each of thesectors in Figure 5a, the mean winter surface of maximum wind pressure (eP) for positive ArcticOscillation Index (AOI) minus eP for negative AOI. The arrows in Figure 5a indicate where the samplemeridians are located, and all results outside of the gray region are individually significant at a = 0.05.


8 of 15

D10106

Figure 6. The symbol D indicates mean value for the positive phase of the Arctic Oscillation Index(AOI) minus mean value for the negative phase of the AOI for 1958–2004, and each panel shows the47-year mean tropopause with a thin gray line. Figures 6a–6c are for 120�W, and Figures 6d–6f are for60�W. (a, d) DV contoured gray at 1.5 m/s with negative values dashed. The AOI positive phase surfaceof maximum wind pressure (eP, solid black line) and AOI negative phase eP (dashed black line) withcircles showing differences significant in a t test at a = 0.05. (b, e) D[�Rdf

�1@Tv=@n] contoured at1.5 m/s with negative values dashed. (c, f) DT contoured at 0.5 K with negative values dashed.


9 of 15

D10106

Atlantic sector support a vertically tilted column of in-creased @Tv/@n from the surface at 45�N into the strato-sphere (Figure 6e) and decreased SMW pressures near55�N. There are also speed increases on the SMW near55�N (Figure 6d) because of the deep underlying changes in

@Tv/@n (Figure 6e). Warming of the column at 40�N duringthe positive phase of the AO (Figure 6f) weakens @Tv/@nin the equatorward column at 30�N (Figure 6e) anddecreases the wind speeds at the overlying SMW(Figure 6d). Near-zero values of D@Tv=@n close to the

Figure 7. Same as Figure 6, but for (a–c) 0�E and (d–f) 110�E.


10 of 15

D10106

SMW (Figure 6e), however, are associated with little or nochange in the mean SMW pressure (Figure 6d). Side by sidecomparison of the @Tv/@n changes in the East Pacific andAtlantic sectors (Figure 6b versus Figure 6e) reveals that thecenters of maximum positive and negative @Tv/@n changeare approximately 15� closer to the equator in the Atlanticsector, consistent with the relative positions of the SMWpressure changes. The equatorward displacement in theAtlantic sector contributes to the ellipticity of the SMWversus AOI relationship shown in Figure 5.[33] In the European sector (Figure 7, left column), the

positive phase of the AO is associated with troposphericwarming poleward of 40�N and tropospheric cooling equa-torward of 40�N (Figure 7c). The warming poleward of40�N is associated with a strengthening of @Tv/@n in acolumn from 60�N at the surface up into the polar strato-sphere (Figure 7b), SMW pressures decrease and windspeeds are generally faster (Figure 7a). Between the col-umns of tropospheric warming and cooling, @Tv/@n isweakened (40�N, Figure 7b) during the positive phase ofthe AO, and the overlying SMW is slower and lower(Figure 7a).[34] The Asian sector features high latitude stratospheric

cooling and a shallow tropospheric warming near 65�Nduring the positive phase of the AO (Figure 7f). Tropo-spheric virtual temperature gradients are large in this regionduring winter and the shallow tropospheric warmingamounts to little change in @Tv/@n and little change in windspeed between the positive and negative phases of the AO.The stratospheric cooling, however, increases @Tv/@n nearthe SMW and raises the altitude at which @Tv/@n goes tozero, supporting a broad zone of SMW pressure decreasebetween 50 and 70�N (Figure 7d). Side by side comparisonof the Asian and European sector @Tv/@n changes(Figures 7b and 7e) shows that the location of the largestnear-tropopause increase in @Tv/@n is closer to the pole overEurope and closer to the equator over the Asian sector,

contributing to the ellipticity of the SMW versus AOrelationship.[35] Farther south in the Asian sector, the positive phase of

the AO is also associated with higher temperatures near thesteep slope in the tropopause (Figure 7f, 25–45�N). Al-though the associated, equatorward changes in @Tv/@n arelarge (10–35�N, Figure 7e), the SMW pressure differencesare small (10–35�N, Figure 7d) because the pressure of themaximum change in @Tv/@n is very close to the SMWpressure at this latitude (indicating generalized slowing ofwind speed without a relocation of the speed maximum). Forcomparison, consider the pressure offset between the SMWand the center of @Tv/@n change near 65�N in Figures 7d and7e (or any other SMW pressure change shown in themanuscript). The comparison illustrates two points: (1) sig-nificant SMW pressure differences occur when there is apressure offset between the maximum @Tv/@n change and theSMW, and (2) SMW pressure changes are always associatedwith nearby changes in @Tv/@n, but changes in @Tv/@n nearthe SMW are not always associated with SMW pressurechanges. Both points follow from the general relationshipbetween the position of a function’s maximum (in thisapplication, the pressure of the fastest wind speed) and thefunction’s first derivative (in this application, equation (1)).

3.4. SMW and ENSO

[36] SMWpressure is significantly correlatedwith theMEIat nodes amounting to more than one-third of the NH surfacearea and more than one-half the tropics equatorward of 20�N(Figure 8a), and the map of the eP versus MEI correlation isfield significant at a = 0.05. At 30�N, the warm phase ofENSO is associated with SMW pressure decreases fromChina across the Pacific toward the United States(Figure 8a) with changes in mean pressure as large as27 hPa (Figure 8b). At 7.5�N, the warm phase of ENSO isassociated with up to 60 hPa of SMW pressure increase near120�E and 160�W with relatively low SMW pressures over

Figure 8. (a) Pearson correlation coefficient (r) for winter (December–February) mean surface ofmaximum wind pressure (eP) versus the Multivariate El Nino/Southern Oscillation Index (MEI) for1958–2004. Contours are at 0.2 with negative values dashed. Results locally significant at a = 0.05account for 38% of the map area (including, but not limited to, all of the contoured area), and the map ofresults is field significant at a = 0.05. (b) Mean winter eP for MEI � 1 (El Nino warm phase) minus meanwinter eP for MEI � �1 at 7.5�N and 30�N. Differences significant in a t test at a = 0.05 are shown withcircles.


11 of 15

D10106

intervening longitudes (Figure 8b). The low-eV , high-eP de-pression in the west-Pacific SMW topography introduced insection 3.1 (Figure 3a) is a stable feature with lower variancethan the surrounding SMW (not shown), and exhibits a weakpressure correlation with the MEI (Figure 8a).[37] The relationship between SMW pressure variability

and ENSO shown in Figure 8 can be understood in thecontext of atmospheric temperature changes. Beginningwith the east Pacific at 140�W (Figures 9a–9c), the warmphase of ENSO is associated with tropospheric warmingcentered at 15�N and 250 hPa and tropospheric coolingcentered at 40�N and 400 hPa (Figure 9c). The east Pacificwarming center at 15�N is consistent and collocated withENSO-related downward vertical velocity anomalies in theHadley circulation cell [Wang, 2002]. Likewise, the coolingat 40�N in Figure 9c is consistent with ENSO-relatedupward vertical velocity anomalies identified by Wang[2002]. @Tv/@n is strengthened in the intervening columnat 25�N (Figure 9b), supporting the relatively low SMWpressures and faster wind speeds in the overlying jet stream(Figure 9a). Equatorward of the tropospheric warmingcenter at 15�N, @Tv/@n is weakened at 7.5�N and 300 hPa(Figure 9b), resulting in SMW pressure increases and sloweroverlying wind speeds (Figure 9a). Poleward of the coolingcenter at 40�N, @Tv/@n is weakened (Figure 9b) and theoverlying SMW is slower and lower (Figure 9a).[38] While ENSO-related changes in the east Pacific

SMW are significant as far north as 65�N (Figure 9a),ENSO-related changes in the west Pacific are more tropical,extending instead to 30�N (Figures 9d and 8a). The uppertroposphere in the west Pacific exhibits ENSO-related cool-ing at 30�N and 300 hPa (Figure 9f), collocated andconsistent with an ENSO-related upward vertical velocityanomaly in the Hadley circulation cell [Wang, 2002]. @Tv/@nis weakened (strengthened) in the column immediatelypoleward (equatorward) of the cooling. Upper troposphericwinds are faster and SMW pressures are lower over thecolumn of increased @Tv/@n at 20�N. Warming near thelapse-rate tropopause at 25�N associated with ENSO weak-ens @Tv/@n in the equatorial troposphere above whichnear tropopause winds are slower and SMW pressures areincreased. Considering Figures 9a and 9d together, thewarm phase of ENSO is associated with a flattening of theSMW topography equatorward of 30�N when viewed alongmeridians. Warm phase related flattening of the SMWtopography is also apparent in Figure 8b which views theENSO-related changes in SMW pressure along parallels.

4. Summary and Discussion

[39] Mapping the mean three-dimensional structure of thewinter SMW reveals a surface that generally slopes upwardfrom the pole to the equator with a steep midlatitudetransition close to the polar front jets. The SMW is generally50 to 100 hPa below the lapse-rate tropopause except withinfast jet cores that extend laterally into the stratospherethrough tropopause folds or breaks (e.g. the winter jet streamover the western Pacific). The SMW’s structural relationshipwith the dynamic tropopause is similar to the SMW’sstructural relationship with the lapse-rate tropopause.[40] The dominant pattern of SMW space-time variability

in the winter extratropics is related to the AO such that,

during the AO positive phase, mean SMW pressures aredecreased at high latitudes and increased at middle andsubtropical latitudes. The magnitude of the SMW responseto the AO is largest in the European and East Pacific sectorsand smallest in the Central Pacific. The SMW response tothe AO is elliptical with the centers of SMW pressurechange displaced most equatorward over Asia and theAtlantic. The position and strength of the tropospheric andstratospheric temperature changes associated with the AOare used to explain why the belts of SMW pressure changeare arranged elliptically about the pole, and why each is notconsistently associated with speed changes. Over the At-lantic and Europe, AO-related temperature changes causedeep tropospheric changes in @Tv/@n and variation of theoverlying upper tropospheric winds is correspondinglylarge. Over the east Pacific and Asia, AO-related tempera-ture changes cause largely stratospheric changes in @Tv/@n,resulting in SMW pressure changes with comparably smallchanges in upper tropospheric wind speeds close to theSMW.[41] The dominant pattern of SMW space time variability

in the tropics is related to ENSO. Changes in the tropo-spheric temperature structure alter @Tv/@n such that thetropical SMW topography is somewhat flatter (more iso-baric) during the warm phase of ENSO. Changes in tropo-spheric @Tv/@n align well with changes in uppertropospheric wind speed, and the well-known eastwardexpansion and strengthening of the Pacific jet stream isassociated with SMW pressures up to 27 hPa lower near30�N. The east Pacific SMW pressure decrease is flankedby slower upper tropospheric winds and SMW pressureincreases near the equator and the Gulf of Alaska.[42] Some of the interannual differences in SMW pres-

sure declared significant in section 3 are small (20 to 50 hPa)relative to the vertical spacing of the Reanalysis data withinthe SMW domain (50 or 100 hPa). The Reanalysis verticalresolution is sufficient to support the statistical significanceof these conclusions for two reasons. First, a large samplesize informs each seasonal mean: 360 (or 364 for leap years)individual SMW pressure measurements are used for eachseasonal mean calculation at each node. Second, and moreimportant, the within-season pressure variability of theSMW is sufficiently large relative to the vertical spacingof the Reanalysis data. Using winter 2004 as an example,calculation of the seasonal mean SMW pressure involvedseven isobaric surfaces at nodes accounting for more thanhalf the NH surface area, and no single node incorporatedfewer than four isobaric surfaces in its seasonal mean SMWpressure calculation. For winter 2004, the 95% confidenceintervals on the estimate of the seasonal mean are typicallyeP ± 8.1 hPa and rarely exceed eP ± 16.0 hPa. Theseasonal mean SMW pressure results also exhibit alogical relationship to seasonal mean wind speed onnearby isobaric surfaces [Strong and Davis, 2005], andthe seasonal mean SMW pressure results correlate logi-cally with independently-developed climate indices suchas the AO (section 3.3).[43] The geostrophic thermal wind example discussed in

the Introduction (Figure 1 and section 1.2) makes clear thedependence of SMW pressure on the temperature structureof the underlying atmosphere. PCAwas used to identify theAO and ENSO as the two leading contributors to SMW


12 of 15

D10106

Figure 9. The symbol D indicates mean value for winters when multivariate El Nino/SouthernOscillation Index (MEI) � 1 minus mean value for years when MEI � �1 for 1958–2004. Each panelshows the 47-year mean tropopause with a thin gray line. Figures 9a–9c are for 140�W, and Figures 9d–9f are for 120�E. (a, d) DV contoured gray at 4 m/s with negative values dashed. Mean winter surface ofmaximum wind pressure (eP) for winters with MEI � 1 (solid black line) and MEI � �1 (dashed blackline) with circles showing differences significant in a t test at a = 0.05. (b, e) D[�Rdf

�1@Tv=@n]contoured at 3 K with negative values dashed. (c, f) DT contoured at 0.5 K with negative values dashed.


13 of 15

D10106

pressure variability over the NH, and their influences on theSMW were studied in the context of associated changes intropospheric and stratospheric temperature. The upwardtrend in mean NH surface temperature could also contributeto SMW and jet core pressure variability. Surface temper-ature increases have been largely confined to the dry, cold-core anticyclones of North America and Siberia [Balling etal., 1998; Michaels et al., 2000], likely altering the thick-ness structure and distribution of isobaric temperaturegradients in the overlying atmosphere. Indeed, for thewinter mean SMW pressure PCA performed using the entireNH as input, the third principal component is significantlycorrelated with NH mean surface temperature, identifying arelationship needing further study. The SMW sensitivity toupper tropospheric and lower stratospheric temperaturegradients motivates its investigation in the context of trendsand variability in the atmospheric thermal structure [Angell,1998; Pielke et al., 1998; Christy et al., 2000] and theirrelationship to surface temperature variability [Hurrell andTrenberth, 1998; Gaffen et al., 2000; Michaels et al., 2000].[44] Although the mean wind speed on the SMW is

shown here, its variability is not explicitly considered.When the geostrophic thermal wind equation in naturalcoordinates (equation (A3)) is integrated from the surfaceto the SMW, we may expect a faster (slower) wind maxi-mum when the integration is performed over a deeper(shallower) layer unless there is a compensating change inthe horizontal gradient of temperature. Indeed, when uppertropospheric wind speed changes accompanied the exam-ples of SMW pressure change examined here, the underly-ing changes in @Tv/@n were of sufficient magnitude andvertical depth that deceleration most often occurred withSMW pressure increases and acceleration most often oc-curred with SMW pressure decreases. Future work willinclude more thorough consideration of the climatologicalcovariance between SMW speed and altitude.

Appendix A

[45] Assuming that the magnitude of the geostrophicwind Vg = (ug

2 + vg2)1/2 is nonzero and its vector components

are differentiable functions of pressure, we can write

@Vg

@p¼ 1

2u2g þ v2g

� ��1=22ug

@ug@p

þ 2v@vg@p

� �: ðA1Þ

Substituting the magnitude of the geostrophic wind(f�1k ~rFk) for (ug2 + vg

2)1/2, the geostrophic wind compo-nents for ug and vg, and the geostrophic thermal windcomponents in isobaric coordinates for @ug/@p and @vg/@p,we obtain

@Vg

@p¼ f

~rF�� 1

f

@F@y

� �Rd

fp

@Tv@y

� ��þ 1

f

@F@x

� ��Rd

fp

@Tv@x

� ��

p@Vg

@p¼ � Rd

f ~rF��

@F@y

@Tv@y

þ @F@x

@Tv@x

� �

@Vg

@ ln p¼ � Rd

f ~rF�� rF � rTvð Þ ¼ �Rdf

�1 ~rTv

�� cos bðA2Þ

where b is the angle between ~rTv and ~rF. Since k ~rTvkcosb is the portion or component of the virtual temperaturegradient vector that operates in the direction of thegeopotential height gradient, (A2) may be equivalentlyexpressed in natural coordinates

@Vg

@ ln p¼ �Rd

f

@Tv@n

ðA3Þ

where n is oriented normal to ~Vg and defined positive to theright of the flow direction. Equation (A3) is also obtained ifthe isobaric geostrophic wind magnitude expressed innatural coordinates (Vg = �f�1@F/@n) is differentiated withrespect to pressure.

[46] Acknowledgments. Comments from three anonymous reviewershelped to improve this manuscript. Discussions with J. W. Nielsen-Gammon regarding the dynamic tropopause were very helpful and aregratefully acknowledged. C. Strong was supported under a NationalScience Foundation Graduate Research Fellowship.

ReferencesAngell, J. K. (1998), Contraction of the 300 mbar north circumpolar vortexduring 1963–1997 and its movement into the eastern hemisphere,J. Geophys. Res., 103(D20), 25,887–25,894.

Angell, J. K. (2001), Relation of size and displacement of the 300 mbarnorth circumpolar vortex to QBO, El Nino, and sunspot number, 1963–2000, J. Geophys. Res., 106, 31,787–31,794.

Angell, J. K., and J. Korshover (1977), Variation in size and location of the300 mb North circumpolar vortex between 1963 and 1975,Mon. WeatherRev., 105, 19–25.

Angell, J. K., and J. Korshover (1985), Displacements of the north circum-polar vortex during El Nino, 1963–1983, Mon. Weather Rev., 113,1627–1630.

Arkin, P. A. (1982), The relationship between interannual variability in the200 mb tropical wind field and the Southern Oscillation, Mon. WeatherRev., 110, 1393–1404.

Balling, R. C., P. J. Michaels, and P. C. Knappenberger (1998), Analysis ofwinter and summer warming rates in gridded temperature timeseries,Clim. Res., 9, 175–181.

Bjerknes, J. (1969), Atmospheric teleconnections from the equatorial Pa-cific, Mon. Weather Rev., 97, 163–172.

Bluestein, H. (1992), Synoptic-Dynamic Meteorology in Midlatitudes: Prin-ciples of Kinematics and Dynamics, vol. 1., p. 448, Oxford Univ. Press,New York.

Burnett, A. W. (1993), Size variations and long-wave circulation within theJanuary North Hemisphere circumpolar vortex: 1946–89, J. Clim., 6,1914–1920.

Buss, S., A. Hertzog, C. Hostettler, T. P. Bui, D. Luthi, and H. Wernli(2004), Analysis of a jet stream induced gravity wave associated withan observed stratospheric ice cloud over Greenland, Atmos. Chem. Phys.,4, 1183–1200.

Chen, W. Y., and H. M. Van den Dool (1999), Significant change of extra-tropical natural variability and potential predictability associated with theEl Nino/Southern Oscillation, Tellus, 51A, 790–802.

Christoph, M., U. Ulbrich, and P. Speth (1997), Midwinter suppression ofNorthern Hemisphere storm track activity in the real atmosphere and inGCM experiments, J. Atmos. Sci., 54, 1589–1599.

Christy, J. R., R. W. Spencer, and W. D. Braswell (2000), MSU tropo-spheric temperatures: Dataset construction and radiosonde comparisons,J. Atmos. Oceanic Technol., 17, 1153–1170.

Danielson, E. (1968), Stratospheric-tropospheric exchange based on radio-activity, ozone, and potential vorticity, J. Atmos. Sci., 25, 502–518.

Danielson, E., and R. Hipskind (1980), Stratospheric-tropospheric ex-change at polar latitudes in summer, J. Geophys. Res., 85, 393–400.

Davis, R. E., and S. R. Benkovic (1992), Climatological variations in theNorthern Hemisphere circumpolar vortex in January, Theor. Appl. Clima-tol., 46, 63–74.

Davis, R. E., and S. R. Benkovic (1994), Spatial and temporal variations ofthe January circumpolar vortex over the Northern Hemisphere, Int. J.Climatol., 14, 415–428.

Defant, F., and H. Taba (1957), The threefold structure of the atmosphereand the characteristics of the tropopause, Tellus, 9, 259–274.

Fawbush, E. J., R. C. Miller, and L. G. Starrett (1951), An empiricalmethod of forecasting tornado development, Bull. Am. Meteorol. Soc.,32, 1–9.


14 of 15

D10106

Frauenfeld, O. W., and R. E. Davis (2000), The influence of El Nino-Southern Oscillation on the Northern Hemisphere 500 hPa circumpolarvortex, Geophys. Res. Lett., 27, 537–540.

Frauenfeld, O. W., and R. E. Davis (2003), Northern Hemisphere circum-polar vortex trends and climate change implications, J. Geophys. Res.,108(D14), 4423, doi:10.1029/2002JD002958.

Gaffen, D. J., B. D. Santer, J. S. Boyle, J. R. Christy, N. E. Graham, andR. J. Ross (2000), Multidecadal changes in the vertical temperaturestructure of the tropical troposphere, Science, 287, 1242–1245.

Hastenrath, S. (2003), Upper-air circulation of the Southern Oscillationfrom the NCEP-NCAR Reanalysis, Meteorol. Atmos. Phys., 83, 51–65.

Hitchman, M. H., M. L. Buker, G. J. Tripoli, E. V. Browell, W. B. Grant,T. J. McGee, and J. F. Burris (2003), Nonorographic generation of Arcticpolar stratospheric clouds during December 1999, J. Geophys. Res.,108(D5), 8325, doi:10.1029/2001JD001034.

Hoerling, M. P., T. K. Schaack, and A. J. Lenzen (1991), Global objectivetropopause analysis, Mon. Weather Rev., 119, 1816–1831.

Hoerling, M. P., T. K. Schaack, and A. J. Lenzen (1993), A global analysisof stratospheric-tropospheric exchange during northern winter, Mon.Weather Rev., 121, 162–172.

Horel, J. D., and J. M. Wallace (1981), Planetary-scale atmospheric phe-nomena associated with the Southern Oscillation, Mon. Weather Rev.,109, 813–829.

Hurrell, J. W., and K. E. Trenberth (1998), Difficulties in obtaining reliabletemperature trends: Reconciling the surface and satellite microwavesounding unit records, J. Clim., 11, 945–967.

Ichiye, T., and J. Petersen (1963), The anomalous rainfall of the 1957–58winter in the equatorial central Pacific arid area, J. Meteorol. Soc. Jpn.,41, 172–182.

Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull.Am. Meteorol. Soc., 77(3), 437–471.

Kanamitsu, M., R. E. Kistler, and R. W. Reynolds (1997), NCEP/NCARReanalysis and the use of satellite data, Adv. Space Res., 19, 481–489.

Kistler, R., et al. (2001), The NCEP-NCAR 50-Year Reanalysis: Monthlymeans CD-ROM and documentation, Bull. Am. Meteorol. Soc., 82(2),247–267.

Kloth, C. M., and R. Davies-Jones (1980), The relationship of the 300-mbjet stream to tornado occurrence, NOAA Tech. Memo., ERL NSSL, 88.

Knox, J. A. (1997), Possible mechanisms of clear-air turbulence in stronglyanticyclonic flows, Mon. Weather Rev., 125, 1251–1259.

Krishnamurti, T. N. (1961a), The subtropical jet stream of winter, J. Atmos.Sci., 18, 172–191.

Krishnamurti, T. N. (1961b), On the role of the subtropical jet stream ofwinter in the atmospheric general circulation, J. Atmos. Sci., 18, 657–670.

Kutzbach, J. E. (1970), Large-scale features of monthly mean NorthernHemisphere anomaly maps of sea-level pressure, Mon. Weather Rev.,98, 708–716.

Langford, A. O. (1999), Stratosphere-troposphere exchange at the subtro-pical jet: Contribution to the tropospheric ozone budget at midlatitudes,Geophys. Res. Lett., 26, 2449–2452.

Livezey, R. E., and W. Y. Chen (1983), Statistical field significance and itsdetermination by Monte Carlo techniques, Mon. Weather Rev., 111, 49–59.

Lorenz, E. N. (1951), Seasonal and irregular variations of the NorthernHemisphere sea-level pressure profile, J. Meteorol., 8, 52–59.

Michaels, P. J., P. C. Knappenberger, R. C. Balling Jr., and R. E. Davis(2000), Observed warming in cold anticyclones, Clim. Res., 14, 1–6.

Mintz, Y. (1954), The observed zonal circulation of the atmosphere, Bull.Am. Meteorol. Soc., 35, 208–214.

Morgan, M., and J. Nielsen-Gammon (1998), Using tropopause maps todiagnose midlatitude weather systems, Mon. Weather Rev., 126, 2555–2579.

Nakamura, H. (1992), Midwinter suppression of baroclinic wave activity inthe Pacific, J. Atmos. Sci., 49, 1629–1642.

Namias, J., and P. F. Clapp (1949), Confluence theory of the high-tropo-spheric jet stream, J. Meteorol., 7, 130–139.

Newton, C. W. (1959), Axial velocity streaks in the jet stream: Ageos-trophic ‘‘inertial’’ oscillations, J. Atmos. Sci., 16, 638–645.

Palmen, E. (1948), On the distribution of temperature and wind in the upperwesterlies, J. Meteorol., 5, 20–27.

Palmen, E., and C. W. Newton (1948), A study of the mean wind andtemperature distribution in the vicinity of the polar front in winter,J. Meteorol., 5, 220–226.

Pielke, R. A., Sr., J. Eastman, T. N. Chase, J. Knaff, and T. G. F. Kittel(1998), The 1973–1976 trends in depth-averaged tropospheric tempera-ture, J. Geophys. Res., 103, 16,927–16,934.

Platzman, G. (1949), The motion of barotropic disturbances in the uppertroposphere, Tellus, 1, 53–64.

Rasmusson, E. M., and K. Mo (1993), Linkages between 200 mb tropicaland extratropical circulation anomalies during the 1986–89 ENSO cycle,J. Clim., 6, 595–616.

Rasmusson, E. M., and J. M. Wallace (1983), Meteorological aspects of theEl Nino/Southern Oscillation, Science, 222, 1195–1202.

Reiter, E. R. (1958), The layer of maximum wind, J. Meteorol., 15, 27–43.Reiter, E. R. (1961), Jet Stream Meteorology, 515 pp., Univ. of ChicagoPress, Chicago, Ill.

Reynolds, R. W., and T. M. Smith (1994), Improved global sea surfacetemperature analyses, J. Clim., 7, 929–948.

Riehl, H., M. A. Alaka, C. L. Jordan, and R. J. Renard (1954), The jetstream, Meteorol. Monogr., 2.

Shapiro, M. A., and P. J. Kennedy (1981), Research aircraft measurementsof jet stream geostrophic and ageostrophic winds, J. Atmos. Sci., 38,2642–2652.

Skaggs, R. H. (1967), On the association between tornadoes and 500-mbindicators of jet streams, Mon. Weather Rev., 95, 107–110.

Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. C. Stokes (1996),Reconstruction of historical sea surface temperature using empiricalorthogonal functions, J. Clim., 9, 1403–1420.

Smith, W., and R. J. Younkin (1972), An operationally useful relationshipbetween the polar jet stream and heavy precipitation, Mon. Weather Rev.,100, 434–440.

Sobel, A. H., I. M. Held, and C. S. Bretherton (2002), The ENSO signal intropical tropospheric temperature, J. Clim., 15, 2702–2706.

Soden, B. J. (2000), The sensitivity of the tropical hydrological cycle toENSO, J. Clim., 13, 538–549.

Spaete, P., D. Johnson, and T. Schaak (1994), Stratospheric-troposphericmass exchange during the Presidents’ Day storm, Mon. Weather Rev.,122, 424–439.

Staff Members (1947), On the circulation of the atmosphere in middlelatitudes, Bull. Am. Meteorol. Soc., 28, 255–280.

Straus, D. M., and J. Shukla (1997), Variations of midlatitude transientdynamics associated with ENSO, J. Atmos. Sci., 54, 777–790.

Strong, C., and R. E. Davis (2005), The surface of maximum wind as analternative to the isobaric surface for wind climatology, Geophys. Res.Lett., 32, L04813, doi:10.1029/2004GL022039.

Strong, C., and R. E. Davis (2006), Temperature-related trends in the ver-tical position of the summer upper tropospheric surface of maximumwind over the Northern Hemisphere, Int. J. Clim., in press.

Thompson, D. W. J., and J. M. Wallace (1998), The Arctic oscillationsignature in the wintertime geopotential height and temperature fields,Geophys. Res. Lett., 25, 1297–1300.

Thompson, D. W. J., and J. M. Wallace (2000), Annular modes in extra-tropical circulation, Part 1: Month-to-month variability, J. Clim., 13,1000–1016.

Uccellini, L. W., and D. R. Johnson (1979), The coupling of upper andlower tropospheric jet streaks and implications for the development ofsevere convective storms, Mon. Weather Rev., 107, 682–703.

Uccellini, L. W., and P. J. Kocin (1987), The Interaction of jet streakcirculations during heavy snow events along the east coast of the UnitedStates, Weather Forecasting, 2, 289–308.

Wang, C. (2002), Atmospheric circulation cells associated with the El Nino-Southern Oscillation, J. Clim., 15, 399–419.

Wimmers, A. J., J. L. Moody, E. V. Browell, J. W. Hair, W. B. Grant, C. F.Butler, M. A. Fenn, C. C. Schmidt, J. Li, and B. A. Ridley (2003),Signatures of tropopause folding in satellite imagery, J. Geophys. Res.,108(D4), 8360, doi:10.1029/2001JD001358.

Wolter, K., and M. S. Timlin (1993), Monitoring ENSO in COADS with aseasonally adjusted principal component index, paper presented at 17thClimate Diagnostics Workshop, Okla. Clim. Surv., Sch. of Meteorol.,Univ. of Okla., Norman.

Wolter, K., and M. S. Timlin (1998), Measuring the strength of ENSO -how does 1997/98 rank?, Weather, 53, 315–324.

Yang, S., K.-M. Lau, and K.-M. Kim (2002), Variations of the East Asianjet stream and Asian-Pacific-American winter climate anomalies, J. Clim.,15, 306–325.

Yulaeva, E., and J. M. Wallace (1994), The signature of ENSO in globaltemperature and precipitation fields derived from the microwave sound-ing unit, Clim, J., 7, 1719–1736.

��R. E. Davis, Department of Environmental Sciences, University of

Virginia, Charlottesville, VA 22904, USA.C. Strong, Department of Soil, Environmental, and Atmospheric

Sciences, University of Missouri-Columbia, 302 Anheuser-Busch NaturalResources Building, Columbia, MO 65211, USA. ([email protected])


15 of 15

D10106

Documents

Variability in the altitude of fast upper tropospheric