temperature-related trends in the vertical position …

INTERNATIONAL JOURNAL OF CLIMATOLOGY

Int. J. Climatol. 26: 1977–1997 (2006)

Published online 5 May 2006 in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/joc.1344

TEMPERATURE-RELATED TRENDS IN THE VERTICAL POSITION OF THESUMMER UPPER TROPOSPHERIC SURFACE OF MAXIMUM WIND OVER

THE NORTHERN HEMISPHERE

COURTENAY STRONGa* and ROBERT E. DAVISb

a University of Missouri-Columbia, Department of Soil, Environmental, and Atmospheric Sciences, 302 Anheuser-Busch NaturalResources Building, Columbia, Missouri 65211, USA

b The University of Virginia, Department of Environmental Sciences, 386-B Clark Hall, P.O. Box 400123, University of Virginia,Charlottesville, Virginia 22904-4123, USA

Received 5 March 2005Revised 3 March 2006

Accepted 4 March 2006

ABSTRACT

The surface of maximum wind (SMW) is used to examine spatial and temporal variability in the vertical position of jetstreams and fast upper tropospheric wind maxima over the Northern Hemisphere (NH) in the NCEP/NCAR Reanalysisfor summers 1958–2004. At a given observation time in a gridded data set, the SMW is defined as the surface passingthrough the fastest analyzed wind above each grid node, with a vertical search domain restricted to the upper troposphereand any tropospheric jet streams extending into the lower stratosphere. The 47-year mean summer SMW generally residesbelow the tropopause, undulates in the tropics, and descends poleward in middle and high latitudes. Trends in the pressureof the summer SMW are primarily positive, exceed 30 hPa/decade at some locations, and are found over most of thetropics and subtropics for the period 1958–2004. The thermal wind relationship is used to establish links between theSMW pressure trends and temperature gradient changes in the upper troposphere. The changing temperature gradientsare consistent with nonuniform tropospheric warming and are correlated with sea surface temperature (SST) variabilityrelated to El Nino and the Pacific Decadal Oscillation. Copyright 2006 Royal Meteorological Society.

KEY WORDS: jet stream; circulation; climatology; climate change

1. INTRODUCTION

1.1. Perspective

Following the discovery of concentrated bands of fast upper tropospheric winds (Ooishi, 1926; Lewis, 2003and references therein), research was undertaken to document and understand their position in the atmosphere.The location and strength of upper tropospheric jet streams were theorized and confirmed to be linked tothe position and strength of underlying frontal boundaries (Staff Members, Department of Meteorology,University of Chicago, 1947; Palmen, 1948; Palmen and Newton, 1948). On the seasonal timescale, the winterintensification and equatorward expansion of the polar front jet (PFJ) in conjunction with the intensificationand equatorward expansion of the polar high (Namias and Clapp, 1949) demonstrates the linkage betweenthe thermal structure of the atmosphere and fast upper tropospheric wind features.

On the interannual timescale, the position and strength of fast upper tropospheric winds vary with the ArcticOscillation (AO) – the meridional seesaw of atmospheric mass between the polar region and midlatitudes(Lorenz, 1951; Kutzbach, 1970) apparent in the leading principal component of wintertime sea-level pressure(Thompson and Wallace, 1998). The positive phase of the AO is associated with slower (faster) zonal wind

* Correspondence to: Courtenay Strong, University of Missouri-Columbia, Department of Soil, Environmental, and AtmosphericSciences, 302 Anheuser-Busch Natural Resources Building, Columbia, Missouri 65211, USA; e-mail: [email protected]

Copyright 2006 Royal Meteorological Society

1978 C. STRONG AND R. E. DAVIS

speeds near 30 °N (60 °N) from the surface into the stratosphere for data zonally averaged over the NorthernHemisphere (NH) (Thompson and Wallace, 2000). The phase of the winter AO also influences atmosphericcirculation and the distribution of air temperature during the subsequent summer (Ogi et al., 2003a, 2004a),particularly during maxima in the 11-year solar cycle (Ogi et al., 2003b). The lagged correlation stems inpart from the persistence of winter AO-related anomalies of sea ice cover (Rigor et al., 2002), snow cover,and sea surface temperatures (SSTs) in circumpolar regions (Ogi et al., 2003a). The AO is also discernible insummertime geopotential height fields as an annular structure with smaller meridional extent than its winteranalog (Ogi et al., 2004b).

Interannual variability in the horizontal position of fast upper tropospheric winds can also be traced to SSTvariability. The Pacific jet stream is intensified and displaced equatorward during the warm phase of the ElNino/Southern Oscillation (ENSO), accompanied by changes in wind speed (Arkin, 1982; Rasmusson andMo, 1993; Yang et al., 2002), the circumpolar vortex (Angell and Korshover, 1985; Frauenfeld and Davis,2000; Angell, 2001), middle and upper tropospheric geopotential height topography (Horel and Wallace,1981; Hastenrath, 2003), and storm tracks (Straus and Shukla, 1997; Chen and Van den Dool, 1999). Thejet stream and general tropospheric circulation can also be influenced by SST variability with periods longerthan the ENSO cycle. Multidecadal oscillations have been identified in Atlantic SST fields north of theequator in records dating to 1850 (Schlesinger and Ramankutty, 1994) and in climate reconstructions datingto 1650 (Delworth and Mann, 2000). The ability of this ‘Atlantic Multidecadal Oscillation’ (AMO) (Kerr,2000) to account for tropospheric geopotential height variability extends beyond the Atlantic (Mestas-Nunezand Enfield, 1999; Delworth and Mann, 2000; Enfield et al., 2001). Sutton and Hodson (2005) recentlydocumented the AMO’s influence on summer sea-level pressure, precipitation, and surface air temperatureover North America and Europe.

Changing tropospheric carbon dioxide and stratospheric ozone concentrations (Tett et al., 1996) maypotentially influence the strength and position of upper tropospheric jet streams. Tropospheric warming andlower stratospheric cooling are robustly observed (Angell, 1991; Parker et al., 1997; Hansen et al., 1997;Santer et al., 1999; Gaffen et al., 2000; Ramaswamy et al., 2001) and consistent with theory and modelingwork concerning changing concentrations of tropospheric carbon dioxide and stratospheric ozone (Schlesingerand Mitchell, 1987; Tett et al., 1996; Ramaswamy et al., 2001). The significance, magnitude, and sign of uppertropospheric and lower stratospheric temperature trends vary in the literature (e.g. Ramaswamy et al., 2001)in part because of variability in spatial coverage among data sets, regional temperature variations, analysismethods, and equipment changes (Gaffen, 1994; Parker et al., 1997; Basist and Chelliah, 1997; Stendel et al.,1998; Hurrell and Trenberth, 1997, 1998; Santer et al., 1999; Chase et al., 2000).

Jet stream-level temperature fields frequently exhibit temporal variability associated with the 11-year solarcycle (e.g. Labitzke and van Loon, 1988; van Loon and Shea, 1999, 2000) and volcanic emissions (e.g.Robock and Mao, 1992; Kelly et al., 1996; Kirchner et al., 1999). Some air temperature records near the jetstream level contain steplike changes (Pawson et al., 1998; Gaffen et al., 2000) temporally coincident withthe mid-to-late 1970s climate shift documented in atmospheric fields (Nitta and Yamada, 1989; and Trenberth,1990; Graham, 1994; Miller et al., 1994) and biological data (Ebbesmeyer et al., 1991; Francis and Hare,1994; McGowan et al., 1998; Hare and Mantua, 2000). There is evidence that, rather than being unique, themid-to-late 1970s climate shift was one of a sequence of shifts associated with at least two interdecadal climateoscillations: one with 50–70 year periodicity and the other with 15–25 year periodicity (Minobe, 1999, 2000).The interdecadal variability is often referred to as the Pacific Decadal Oscillation (PDO) (Mantua et al., 1997;Zhang et al., 1997), and its underlying physical mechanism is a focus of ongoing research (Mantua and Hare,2002; Seager et al., 2004). The presence of a distinctly decadal signal in the climate of the Pacific Ocean wasrecently established by examining the coupled behavior of SSTs and NH atmospheric circulation (Frauenfeldet al., 2005).

The vertical positions of upper tropospheric wind maxima also relate to the thermal structure of theatmosphere as indicated by the thermal wind equation (Section 1.2). Jet stream vertical position variability,although less commonly studied than speed or horizontal position variability, has received some researchattention. Mintz (1954) produced cross sections of zonally averaged summer and winter wind speed data fromall latitudes, revealing the approximate vertical position of persistent jet cores. A distribution of pressure at

Copyright 2006 Royal Meteorological Society Int. J. Climatol. 26: 1977–1997 (2006)DOI: 10.1002/joc

SUMMER SURFACE OF MAXIMUM WIND VERTICAL POSITION TRENDS 1979

the maximum wind was developed for 261 soundings in jet stream regions over the United States, revealinga peaked distribution centered on 215 hPa with a 225 hPa range (Endlich et al., 1955). The U.S. Navy beganoperationally mapping the altitude and vertically averaged speed of the three-dimensional layer of maximumwind (LMW) beginning at the end of 1958 for aviation purposes (Reiter, 1958, 1961). The LMW was found toprotrude into the lower stratosphere on the poleward flank of the subtropical jet (STJ), sometimes manifestinga distinct wind maximum above the PFJ (Newton and Persson, 1962).

The upper tropospheric surface of maximum wind (SMW) (Strong and Davis, 2005) has been used tostudy how the vertical positions of tropospheric jet cores and other upper tropospheric wind maxima varywith changes in the atmospheric thermal structure. At a given observation time in a gridded data set, theSMW is defined as the surface passing through the fastest analyzed wind above each grid node, with a verticalsearch domain restricted to the upper troposphere and any tropospheric jet streams extending into the lowerstratosphere. The dominant patterns of joint space-time variability in the winter SMW are related to the AOand ENSO (Strong and Davis, 2006).

The focus of this research is the joint variability of the atmospheric thermal structure and the verticalposition of the SMW over the NH during summer. Section 1.2 covers background theory that informs theformat of the results sections. Data and the method for locating the SMW are described in Section 2, andthe mean spatial structure of the summer SMW is provided in Section 3.1. The leading principal componentsof joint space-time variability in summer SMW pressure are identified in Section 3.2. Patterns of SMWpressure variability associated with the leading principal component are explored on the hemispheric andregional scale in Sections 3.3–3.8. Section 3.9 explores links between the SMW-related upper tropospherictemperature variability and SST variability. Summary and discussion are provided in Section 4.

1.2. Theory

In this section, we provide a brief overview of the relationship between SMW pressure (denoted P ) and thethermal structure of the upper troposphere, developing a conceptual framework that will inform the formatof the results sections. Figure 1(a) and (b) shows idealized geostrophic wind speed profiles for one node attime t1 and time t2, and their local maxima (150 hPa in 1a; 250 hPa in 1b) mark P . Expressing verticalderivatives using −∂/∂ ln p so that the vertical coordinate increases in value upward, P is the pressure where−∂Vg/∂ ln p = 0 and −∂2Vg/∂ ln p2 < 0. The first derivative −∂Vg/∂ ln p is commonly referred to as thethermal wind shear (e.g. Bluestein, 1992), and can be expressed as a function of the spatial distribution oftemperature and geopotential height in the atmosphere as

− ∂Vg

∂ ln p= Rd

f‖ �∇Tv‖ cos β (1)

100

150

Pre

ssu

re (

hP

a)

200

250

300350

(a) (b) (c) (d) (e)

P (t1)∼

P (t2)

00

t2t1t2t1 t2 - t1

000 Vg Vg ∂Vg

∂ ln p

∂Vg

∂ ln p∆

∼

Figure 1. For a node at two times (t1 and t2), idealized profiles of (a, b) geostrophic wind speed Vg , (c, d) the magnitude of thethermal wind shear −∂Vg/∂ ln p, and (e) the t2 minus t1 difference in the thermal wind shear �[−∂Vg/∂ ln p]. Filled circles indicate

the pressure of the surface of maximum wind (P ) at t1 and t2



where p is pressure, Rd is the dry air gas constant, f is the Coriolis parameter, ‖ �∇Tv‖ is the magnitudeof the virtual temperature gradient (−�∇Tv), and β is the angle between −�∇Tv and the geopotential heightgradient (derivation in the Appendix).

The thermal wind shear profiles in Figure 1(c) and (d) were developed by differentiating the idealized Vg

profiles in Figure 1(a) and (b) in accordance with Equation (1). Because the thermal wind shear is positivebelow P and negative above P (Figure 1(c), (d)), changes in P should be accompanied by changes in thethermal wind shear on nearby isobaric surfaces. This is illustrated in Figure 1(e), which shows the decreasein the thermal wind shear after the SMW descends from 150 hPa at t1 to 250 hPa at t2. In the resultssection of this manuscript, we will show the relationship between P variability and the spatial distribution ofatmospheric temperature by analyzing time series of seasonal mean P and seasonal mean thermal wind shearon an isobaric surface within the vertical range of the SMW motion.

2. DATA AND METHODS

2.1. Data

From 2.5° grids in the NCEP-NCAR Reanalysis (Kalnay et al., 1996), we use six-hourly wind speed,geopotential height, humidity, and air temperature on seven isobaric surfaces from 500 to 100 hPa, and thepressure of the lapse-rate tropopause. Artificial discontinuities may exist in the Reanalysis because of variationsin the observing system including changes in the density of rawinsonde data and the use of satellite data after1978 (Kistler et al., 2001). Kanamitsu et al. (1997) found that satellite data impacted the NH troposphereReanalysis less than they impacted the data-sparse stratosphere and eastern oceanic areas of the SouthernHemisphere. Conclusions presented here include references to shifts in upper tropospheric temperature thatoccurred in the mid-to-late 1970s. Although the shifts nearly coincide with the introduction of satellite data,they appear in the radiosonde record as well (Lindzen and Giannitsis, 2002) and coincide with the mid-to-late1970s climate shift identified by Nitta and Yamada (1989) and Trenberth (1990).

The Climate Prediction Center (CPC) AO index (AOI) values used here are developed by projecting themonthly mean 1000 hPa geopotential height anomalies onto the first principal component of the monthlymean 1000 hPa geopotential height field north of 20 °N. Southern Oscillation Index (SOI) values are from theClimate Research Unit and are the normalized pressure difference between Tahiti and Darwin (Ropelewskiand Jones, 1987; Allan et al., 1991; Konnen et al., 1998). PDO index (PDOI) values are obtained fromthe Joint Institute for the Study of the Atmosphere and Ocean, and are the leading principal component ofmonthly ‘residual’ SST anomalies in the North Pacific poleward of 20 °N, in which residual indicates thatthe monthly mean global average SST anomalies are removed prior to performing the principal componentsanalysis (PCA) (Zhang et al., 1997; Mantua et al., 1997). AMO values obtained from NOAA-CIRES ClimateDiagnostic Center (CDC) are the 10-year running mean Atlantic SST anomalies north of the equator (Enfieldet al., 2001) based on the Met Office Hadley Center’s sea ice and SST data set (Rayner et al., 2003). Alsofrom the CDC, we use a multivariate El Nino/Southern Oscillation Index (MEI) that captures variability insea-level pressure, zonal and meridional components of the surface wind, SST, surface air temperature, andtotal cloudiness fraction of the sky over the tropical Pacific (Wolter and Timlin, 1993, 1998). Finally fromthe CPC, we use the monthly mean SST for the global equatorial belt 10 °S–10 °N (denoted SST10°−10 °N)(Smith and Reynolds, 1998).

2.2. The surface of maximum wind

At a given observation time in a gridded data set, the SMW is defined as the surface passing through thefastest analyzed wind above each grid node, with a vertical search domain restricted to the upper troposphereand any tropospheric jet streams extending into the lower stratosphere. Figure 2 is used to illustrate howSMW pressure is identified in gridded data such as the Reanalysis (or in rawinsonde data that are analogousto one column of gridded data). An example from January is used because the utility of the domain restrictiondescribed above is most apparent in winter (when the polar night jet stream is present and the tropospheric



50

70

100

150

200

250300

400

500600700

90 80 70 60 50

Latitude (°N)

MWL

SMW

Tropopause

40 30 20 10 0

Pre

ssu

re (

hP

a)

Figure 2. The pressure of the lapse-rate tropopause, the maximum wind level (MWL), and surface of maximum wind (SMW) along135 °E at 0600 UT on 20 January 2004. Open circles show data that are SMW candidates, filled circles show the SMW. Gray shading

indicates areas with winds ≥25.7 m/s (50 knots)

jet streams extend poleward into the lower stratosphere through tropopause folds or breaks). In the columnabove each node (each tick mark on the abscissa in Figure 2), the wind speed data points are flagged ascandidates or noncandidates for the SMW (with the distinguishing criteria to be described presently). AfterSMW candidates are flagged (open circles, Figure 2), the fastest speed from among the candidates above eachnode defines one point on the SMW (filled circles, Figure 2).

Any datum from 500 to 100 hPa is a SMW candidate unless the point is above the level closest to thelapse-rate tropopause and outside a 25.7 m/s (50-knot) contour that captures at least one tropospheric datum.In other words, SMW candidates are points in the upper troposphere or within upper tropospheric jet streamsextending into the lower stratosphere. For example, the datum at 40 °N and 150 hPa is above the measurementlevel closest to the tropopause, but remains a SMW candidate because the point is within a 25.7 m/s contourcapturing at least one tropospheric datum. The datum at 75 °N and 100 hPa is not a SMW candidate becauseit is above the measurement level closest to the tropopause and the 25.7 m/s contour encircling the point doesnot capture a tropospheric datum (i.e. is not associated with a tropospheric jet stream). Properties associatedwith the SMW will be denoted by a tilde (SMW pressure is P and SMW wind speed is V ).

The SMW is distinguished from the operationally defined LMW (Reiter, 1958) and the Reanalysis maximumwind layer (MWL) because the upper bound of the SMW search domain varies spatially and temporallywith the position of the tropopause and the bounds of upper tropospheric jet streams. The MWL has afixed search domain from 500 to 70 hPa and is shown for comparison in Figure 2. The sometimes largedifferences between the SMW and MWL poleward of 40 °N result largely from the SMW’s intentional focuson upper tropospheric winds, whereas the differences equatorward of 40 °N result from the use of cubic splineinterpolation in the MWL.

2.3. Derived variables

The magnitude of a gradient is its Euclidian norm

‖ �∇Z‖ = ([∂Z/∂x]2 + [∂Z/∂y]2)1/2 (2)

where first derivatives are numerically approximated using second-order central differences for most nodes,second-order backward differences at the pole, and second-order forward differences at the equator. Thecosine of the angle β between two gradients is

cos β = �∇Z · �∇T (‖ �∇Z‖‖ �∇T ‖)−1. (3)



If the gradient at a node rotates temporally, the relative importance of its components changes. To conveythis graphically a ‘mean unit-length vector’ is calculated for a block of time and a series of mean unit-lengthvectors is presented on a timeline. The mean unit-length vector is calculated in three steps: six-hourly vectorsare normalized to magnitude 1 (converted to unit-length vectors), the mean x-component and y-componentof the unit-length vectors are calculated for the block of time (yielding the mean unit-length vector), and themean unit-length vector is normalized to magnitude 1 for display. Using virtual temperature as an example, themean unit-length vector normalized to magnitude 1 is denoted −�∇Tv/‖ �∇Tv‖ (temporal averaging is implicit).

2.4. Statistical methods

PCA is used to identify patterns of simultaneous anomalies that account for variability in summer meanP . The PCA variables are Reanalysis grid nodes and the PCA cases are the 47 summer (June–August) orwinter (December–February) seasonal mean SMW pressures at each node for the years 1958–2004. Seasonalmean P values are weighted by the square root of the cosine of latitude to ensure proper representation inthe correlation matrix. The unit-length eigenvectors and eigenvalues of the correlation matrix are obtained viasingular value decomposition. The mth principal component time series (‘PC m’) is produced by projectingthe standardized, weighted seasonal mean P data onto the mth eigenvector and then standardizing the timeseries by subtracting its mean and dividing by its standard deviation. When mapped, the mth unit-lengtheigenvector is referred to as the PC m ‘loading pattern.’ The total variance explained by PC m is the PC m

eigenvalue divided by the summation of the correlation matrix eigenvalues.Correlations are measured with least squares linear regression and, unless otherwise specified, statistics

for all reported correlations are significant at α = 0.05 based on effective degrees of freedom ‘df’ (adjustedto account for Pearson lag-1 autocorrelation in the regression residuals). MSE is used to denote the meansquared error of the regression. When the results of statistical tests at multiple locations are mapped, thecollection of results is tested for field significance at the 95% confidence level using Monte Carlo simulations(Livezey and Chen, 1983). Passing the field significance test indicates there is less than a 5% probabilitythat the amount of area with significant local results occurred by chance given the field’s spatial degrees offreedom. Time series change points are detected using an iterative, multiple change point detection algorithmbased on data ranking (Siegel and Castellan, 1988; Lanzante, 1996).

3. RESULTS

3.1. Summer SMW spatial variability

The summer mean SMW pressure is highest near the pole and lowest over the tropical eastern hemisphere,especially in the vicinity of the tropical easterly jet (TEJ) (dark shading centered over India in Figure 3(a)).Figure 3(b) shows a cross-section from Figure 3(a) at 45 °E, illustrating how the SMW relates to themean tropopause and isotachs. The SMW slopes down toward the equator within the TEJ (12 °N 150 hPa,Figure 3(b)). The subtropical jet core at 40 °N marks a local minimum of SMW pressure flanked by zones offairly steep SMW pressure increase. In higher latitudes, the mean SMW generally follows the slope of themean tropopause, residing approximately 60 hPa below it. The spatial relationship between the high-latitudeSMW and tropopause is fairly robust from summer to summer. For all latitudes poleward of 45 °N, zonallyaveraged summer SMW pressure and zonally averaged summer lapse-rate tropopause pressure (pLRT) arecorrelated, accounting for up to 42% of the SMW pressure variance (Figure 3(c)). Equatorward of 45 °N, theSMW offset from the tropopause is generally larger and more variable (Figure 3(b)), and the P versus pLRT

correlation is not significant (Figure 3(c)).

3.2. Summer P principal components

Results of the PCA of summer P are shown in Figure 4. Summer P PC 1 accounts for more than three timesthe variance of the second PC (Figure 4(a) vs 4(c)). The loading pattern of summer P PC 1 is concentrated



(a)P (hPa)

340320300280260240220200180160140120

100V (m/s)

10

10

20 20

20

10

20

30

SMW

Tropopause200

(b)

Pre

ssu

re (

hP

a)

300

400(c)

0.6

0.3

r (P

, pLR

T)

∼ 0.0

90 75 60 45

Latitude (°N)30 15 0

∼

Figure 3. (a) Mean surface of maximum wind pressure (P ) for summers 1958–2004. The line marks the meridian used for thecross-section in the lower panel. (b) Cross-section showing P from upper panel (bold dashed line) with the mean pressure of thesummer lapse-rate tropopause (bold solid line) and mean isotachs (light gray lines showing ‖ �V ‖ at 10 m/s interval) along 45 °E forsummers 1958–2004. Dashed lines bound the 10th to 90th percentile of P and the gray area bounds the 10th to 90th percentile ofseasonal mean tropopause pressure. (c) Pearson correlation coefficient r for regression of zonally averaged summer mean P versus

zonally averaged summer mean lapse-rate tropopause pressure pLRT, with circles indicating correlations significant at α = 0.05

in the tropics and subtropics, with positive values over most of the eastern hemisphere and Atlantic, andnegative values appearing over the equatorial East Pacific and South America (Figure 4(a)). Summer P PC 1has a positive trend (Figure 4(b)) that indicates a decadal-scale SMW pressure increase (descent) over mostof the eastern hemisphere tropics and subtropics, and a SMW pressure decrease (ascent) over the equatorialcentral and eastern Pacific. The SMW variability associated with summer P PC 1 is examined in more detailin Sections 3.3–3.9.

Summer P PC 2 is correlated with the summer AOI (Figure 4(d)). P PC 2 has a zone of positive loadingsnear 35 °N from the East Pacific across the United States and from northern Africa across central Asia(Figure 4(c)). During the positive phase of the summer AO, SMW pressures are anomalously high in thiszone (positive loadings) and zonal wind speeds are anomalously low (Ogi et al., 2004b). Consistent with the



(a) PC 1−0.01

0.02

0.01

0.02

0.02

0.010.02

0.03

0.03

0.01−0

.01

0.03

0.01

−0.01

0.02

−0.01

−0.02

0.02 0.05

0.01

−0.02

−0.03

−0.01

0.01

(c) PC 2

(e) PC 3

19%

6%

5%

1960 1970

r = 0.51, df = 33MSE = 0.74

r = 0.59, df = 16MSE = 0.65

r = 0.86, df = 20MSE = 0.27

Summer AOl

Trend1

0

(b)

(d)

(f)

P P

C 1

∼

−1

−2

2

P P

C 2

∼

1

0

−1

−2

3

P P

C 3

∼

2

0

1

−1

−2

−3

PC 1

PC 2

June AMOPC 3

1980

0.6

0.4

0.2

0.0

−0.2 Jun

e A

MO

−0.4

−0.6

1.2

0.8

0.4

0.0

−0.4 Su

mm

er A

Ol

−0.8

−1.2

Time (year)1990 2000

Figure 4. Leading three principal components (PC 1–3) of summer mean surface of maximum wind pressure (P ). Loading patterns(panels (a), (c), and (e)) are contoured at 0.01 with negative values dashed. Percentages indicate the fraction of total variability accountedfor by each PC. (b) Time series of the P PC 1 score with least squares linear regression shown, (d) time series of the summer P PC 2score and the summer (mean June–August) Arctic Oscillation Index(AOI), (f ) time series of the summer P PC 3 score and the JuneAtlantic Multidecadal Oscillation (AMO) index. The Pearson correlation coefficient is denoted r , df is the degrees of freedom adjusted

to account for Pearson lag-1 autocorrelation in the regression residuals, and MSE is the mean squared error

thermal wind relationship, this zone of increased SMW pressure and decreased zonal winds is located justequatorward of a belt where temperatures are anomalously high during the positive phase of the summer AO(Ogi et al., 2004b).

Summer P PC 3 is correlated with the AMO (Figure 4(f )). P PC 3 has strong negative loadings in theNorth Atlantic and Northeast Pacific (where SSTs are anomalously high during the AMO warm phase). P

PC 3 also has strong loadings through portions of the tropics, particularly over the tropical Pacific. The AMOreflects variations in the thermohaline circulation (THC) of the North Atlantic (Delworth and Mann, 2000),and recent modeling work (Zhang and Delworth, 2005) links changes in the THC to shifts in the positionof the intertropical convergence zone and changes in the strength of the Walker and Hadley circulations.The presence of strong, opposite loadings in the eastern and western tropical Pacific may also suggest a



relationship between P PC 3 and the Southern Oscillation. P PC 3 is significantly correlated with the SOI(r = 0.38, df = 24, MSE = 0.86), but its relationship to the AMO is stronger.

3.3. Summer P trend

The trend in P PC 1 is associated with significant SMW pressure trends (primarily positive) as large as30 hPa/decade at nodes accounting for more than two thirds of the area from 0 to 30 °N (Figure 5(a)). Since P

changes are related to temperature gradient changes (Section 1.2), we expect regions with large P trends to beflanked by air temperature trends of opposing sign near the jet stream level. Indeed, the axis of large positiveP trends near 15 °N from the Atlantic east toward India in Figure 5(a) is bounded by equatorward temperatureincreases and poleward temperature decreases (compare Figure 5(a), (b)). The relationship between jet-levelair temperature and the P trends to the west of 60 °W in Figure 5(a) is explored in the regional analysis inSection 3.7.

Although the focus of this research is variability in the pressure of the SMW, speed trends on the SMW areshown in Figure 5(c) and briefly discussed in this paragraph to provide a context for the observed changesin P . Many of the areas of SMW descent in Figure 5(a) also exhibit decreasing wind speeds on the SMW(negative V trends, Figure 5(c)). Wind speed deceleration often accompanies descent of the SMW becauseVg on the SMW is the integral of the thermal wind Equation (1) from the surface to P , and convergenceof the limits of integration (SMW descent toward the surface) yields a smaller integral unless there is acompensating increase in the integrand ‖ �∇Tv‖ cos β. In other words, a jet stream core that descends will haveslower wind speeds unless the underlying temperature gradients strengthen. Variability in V may thus dependon changes in the thermal structure of the entire underlying troposphere, whereas P is sensitive to relativelyshallow upper tropospheric temperature changes that relocate the point where ‖ �∇Tv‖ cos β decreases throughzero.

45

30

15

0

45

30

15

0

45

30

15

0180 120W 60W

Longitude (°)

Lat

itu

de

(°N

)

0 60E 120E

−55

10

−10

20

−30

5

2520

155

5

1015

0.3 0.3

0.20.2 0.1

0.1 −0.1−0.2

−0.1 −0.3

−0.5

−0.5

−0.5

−0.5

−0.5

−0.5

0.1

0.5

0.5 −1−1

−1

−2

P trend (hPa/decade)∼

−T 250 to 150 hPa (K/decade)

∼V trend (ms−1/decade)

(b)

(a)

(c)

Figure 5. 1958–2004 trend in three summer variables: (a) surface of maximum wind (SMW) pressure (P ) (contour interval is5 hPa/decade with negative values dashed), (b) air temperature for the 250 to 150 hPa layer (contour interval is 0.1 K/decade withzero contour bold and negative values dashed), and (c) SMW wind speed (V ) (contour interval is 0.5 m/s per decade with negativevalues dashed). Panels (a–c) are each field significant at α = 0.05. The small gray boxes in panel (a) mark locations where regional

analyses are performed in Sections 3.4–3.7



Atlantic (30°W, 10−15°N)

r = −0.52, df = 25

MSE = 763.2 hPa2

100

150

200P

(hP

a)∼

250

300

2.2

2.0

cos b

P (hPa)∼

1.8

1.6

1.4

1.2

1960 1970 1980


Longitude (°W)60

0.0

0.2

0.4

cos

b

0.6

0

2

4

6

8

10

12

14 (d) T250hPa,1958–1962 (K)−

(e) T250hPa trend (K decade−1)−

(f) T250hPa,1998–2004 (K)−

30

231.5

231

230

231.5

232

0

15

30

00.2

0.1

0.1−0.2

231231

231

232

233

230

−0.1

0.2

15

Lat

itu

de

(°N

)

30

0

15

30

0

(a)

(b)

(c) N

∇T

(×1

0−3 K

km

−1)

−∇T / ∇T

−∇Z / ∇Z

∇T

Rdf −1 ∇Tv cos b Rdf−1

∇T

v c

os b

(m

/s)

Figure 6. For each summer (June–August) 1958–2004 over the tropical Atlantic (30 °W, 10–15 °N): (a) mean surface of maximumwind pressure P and mean 250-hPa thermal wind shear ( −∂Vg/∂ ln p = Rdf −1‖∇Tv‖ cos β), (b) mean 250-hPa temperature gradientmagnitude ‖ �∇Tv‖ and cosine of the angle β between −�∇Tv and −�∇Z, and (c) 250-hPa mean unit-length vectors of geopotential heightgradient −�∇Z/‖ �∇Z‖ and temperature gradient −�∇T /‖ �∇T ‖ for 5-year blocks centered on 1960, 1970 . . . 2000. (d) 1958–1962 mean250-hPa temperature with the 1960 −�∇T /‖ �∇T ‖ shown, (e) trend of summer mean 250-hPa temperature contoured at 0.1 K/decadewith negative values dashed, and (f ) 1998–2002 mean 250-hPa temperature with the 2000 −�∇T /‖ �∇T ‖ shown. The small boxes in

panels (d–f) show the averaging domain for data in panels (a–c)

Regions around the locations marked with small gray boxes in Figure 5(a) (the tropical Atlantic, tropicalEast Africa, and the tropical West and East Pacific) are further examined in Sections 3.4–3.7 to morequantitatively demonstrate the relationship between SMW pressure trends and upper tropospheric thermalstructure changes. Changes in the isobaric thermal pattern responsible for the observed temporal patterns ofP and the thermal wind shear are then mapped and discussed.

3.4. P and temperature pattern trends over the tropical Atlantic

The SMW over the tropical Atlantic descended from 150 to 250 hPa between 1958 and 1990, and residedclose to 250 hPa thereafter (Figure 6(a)). The descent of the SMW was accompanied by a decrease of the250-hPa thermal wind shear toward zero (Figure 6(a)), consistent with the thermal wind equation as discussedin Section 1.2. The decrease of the thermal wind shear resulted from weakening of the 250-hPa temperaturegradient (‖ �∇Tv‖ → 0, Figure 6(b)) and reorienting of the temperature gradient toward orthogonality withthe geopotential height gradient (cos β → 0, Figure 6(b); orientation of −�∇Tv/‖ �∇Tv‖ and −�∇Z/‖ �∇Z‖,Figure 6(c)). For the five-summer period around 1990, the SMW periodically descended below 250 hPa(Figure 6(a)) and, consistent with thermal wind theory, the 1990 −�∇Tv/‖ �∇Tv‖ was at an angle larger than90° from −�∇Z/‖ �∇Z‖ (Figure 6(c)).

Figure 6(d–f) illustrate the changes in 250 hPa temperature that caused the weakening and reorientationof −�∇Tv . Early in the record (1958–1962), the 250 hPa temperature gradient was approximately 2.2 × 10−3

K/km and was directed toward the south-southwest (Figure 6(d)). A 1600 km-wide cooling trend southwest



of the Cape Verde Islands interacted with a warming trend spanning 0–7.5 °N (Figure 6(e)) to essentiallyeliminate the meridional temperature gradient (∂T /∂y component of −�∇Tv) by the year 2000 (Figure 6(f )).Late in the record, residual −�∇Tv was approximately 30% weaker (Figure 6(b)) and nearly orthogonal to −�∇Z

(Figure 6(c), (f )), no longer supporting wind speeds increasing with height through 250 hPa. Summarizingfor the tropical Atlantic, the upper tropospheric temperature pattern early in the record supported wind speedsincreasing up to 150 hPa, but was later modified by a dipole of warming and cooling that weakened androtated the temperature gradient toward orthogonality with the geopotential gradient such that the SMWdescended to 250 hPa.

3.5. Tropical East Africa

The SMW over tropical East Africa descended gradually between 1960 and 1975, then abruptly descendedtoward 200 hPa coincident with the mid-to-late 1970s climate shift, and ended the period of record near160 hPa (Figure 7(a)). Considering the entire period of record, SMW motion is well-correlated with thethermal wind shear at 150 hPa (time series and statistics in Figure 7(a)). Comparison of the 150 hPatemperature structure associated with P before and after the 1970s climate shift provides further insight.Between 1960 and 1990, the thermal wind shear decrease resulted from the 150 hPa temperature gradientweakening (‖ �∇Tv‖ → 0, Figure 7(b)) and rotating toward orthogonality with the geopotential height gradient(cos β → 0, Figure 7(b); orientation of −�∇Tv/‖ �∇Tv‖ and −�∇Z/‖ �∇Z‖, Figure 7(c)).

Figure 7(d–f) illustrate the changes in upper tropospheric temperature that caused the 1960–1990 weak-ening and reorientation of −�∇Tv . Early in the record (1958–1962), the 150-hPa temperature gradient was

Africa (32.5°E, 10−15°N)

r = −0.89, df = 30

MSE = 128.5 hPa2

100

120

140

P (

hPa)

∼ 160

180

2.8

2.4

cos b

P (hPa)∼

2.0

1.6

1.2

1960 1970 1980

Time (year)

1990 2000

Longitude (°E)60

0.0

0.2

0.4 cos

b0.6

0.8

0

−5

5

10

15

20

25

30 (d) T150hPa,1958–1962 (K)−

(e) T150hPa trend (K/decade)−

(f) T150hPa,1988–1992 (K)−

30

209

209

209209

210211 212

0

15

30

00.5

0.7

0.5

0.10.1

−0.2−0.3

211213

209

207

206 206207

−0.1

15

Lat

itu

de

(°N

)

30

0

15

30

0

(a)

(b)

(c) N

2001.0

0.3

−∇T/ ∇T

∇T

(×1

0−3 K

km

−1)

Rdf −1 ∇Tv cos b

∇T

Rdf

−1 ∇

Tv

cos

b (

m/s

)

−∇Z/ ∇Z

Figure 7. For each summer (June–August) 1958–2004 over the tropical East Africa (32.5 °E, 10–15 °N): (a) mean P and mean 150-hPathermal wind shear, (b) mean 150-hPa ‖ �∇Tv‖ and mean cos β, and (c) mean 150-hPa −�∇Z/‖ �∇Z‖ and −�∇T /‖ �∇T ‖ for 5-year blockscentered on 1960, 1970 . . . 2000. (d) 1958–1962 mean 150-hPa temperature with the 1960 −�∇T /‖ �∇T ‖ shown, (e) trend of summermean 150-hPa temperature contoured at 0.1 K/decade with negative values dashed, and (f ) 1988–1992 mean 150-hPa temperature with

the 1990 −�∇T /‖ �∇T ‖ shown. The small boxes in panels (d–f) show the averaging domain for data in panels (a–c)



directed almost due south (Figure 7(d)) with a magnitude of approximately 2.8 × 10−3 K/km (Figure 7(b)).Widespread 150-hPa warming over Africa from the equator to 15 °N and cooling from 15 to 30 °N (Figure 7(e))interacted to weaken the meridional temperature gradient (Figure 7(b)) and orient it nearly orthogonal to −�∇Z

by 1990 (Figure 7(c), (f )). The near orthogonality of −�∇Tv and −�∇Z no longer reliably supported windspeed increasing with height through 150 hPa, and the SMW fluctuated around 170 hPa (Figure 7(a)). Sum-marizing for tropical East Africa, spatially extensive (11 × 106 km2) regions of cooling and warming in theupper troposphere caused a rather abrupt rotation and weakening of the temperature gradient coincident withthe mid-to-late 1970s climate shift, and the SMW descended to around 170 hPa.

3.6. Tropical West Pacific

The seasonal mean SMW in the West Pacific near 15 °N has been descending at approximately16 hPa/decade in concert with the 250-hPa thermal wind shear (Figure 8(a)). The decreasing trend in thethermal wind shear is driven in large part by rotation of the temperature gradient toward orthogonality withthe geopotential height gradient (cos β, Figure 8(b); −�∇Tv/‖ �∇Tv‖ and −�∇Z/‖ �∇Z‖, Figure 8(c)), and isdriven in small part by a modest weakening trend in ‖ �∇T ‖ (Figure 8(b)). The weakening and eastward rota-tion of the temperature gradient is also made apparent by comparing the early versus late 250 hPa temperaturefield (Figure 8(d) vs 8(f)). The temperature changes underlying the weakening and rotation of −�∇Tv consistof a cooling over East Asia and equatorial warming over the ocean east of Southeast Asia (Figure 8(e)). Insummary for the tropical West Pacific, upper tropospheric cooling and warming trends caused 100 hPa ofSMW descent by weakening the temperature gradient and aligning it almost orthogonal to the geopotentialheight field.

West Pacific (132.5°E, 15−20°N)

r = −0.74, df = 23

MSE = 354.1 hPa2120

150

180

P (

hPa)

∼

210

240

2.4

2.0

cos b

P (hPa)∼

1960 1970 1980


Longitude (°E)

0.1

0.2

0.4

cos

b

0.5

2

4

6

8

10

12 (d) T250hPa,1958–1962 (K)−


(f) T250hPa,1998–2004 (K)−

203

204202205

0

15

30

0

0.1

−0.2

207206

205204

203

202

−0.1

0.2

15

Lat

itu

de

(°N

)

30

0

15

30(a)

(b)

(c) N

0.3

120 150

−∇T/ ∇T

∇T

(×1

0−3 K

km

−1)


∇T

Rdf

−1 ∇

Tv

cos

b (

m/s

)

−∇Z/ ∇Z

Figure 8. The same as Figure 6 except for the tropical West Pacific (132.5 °E, 15–20 °N)



3.7. Tropical East Pacific

Recall from Figure 5(a) that the East Pacific has a zonally oriented band of SMW descent (‘descent-region’)poleward of a zonally oriented band of SMW ascent (‘ascent region’). Both phenomena are associated with anupper tropospheric warming trend located over the ocean near 10 °N (0.2 K/decade contour, Figure 9(e)). Thetendency toward more frequent El Nino events during the second half of the record accounts for a portion ofthe warming. The 250-hPa temperature within the region bounded by the 0.2 K/decade contour in Figure 9(e)is positively correlated with the March MEI (r = 0.75, df = 31, MSE = 0.14). The relationship between SSTvariability and SMW-related upper tropospheric temperature changes is explored further in Section 3.9.

The 250 hPa temperature topography was relatively flat prior to the 0.2 K/decade warming trend, and thedescent-region gradient had an eastward orientation (97.5 °W, 15 °N in Figure 9(d)). The 250-hPa temperaturetrend reoriented the descent-region temperature gradient away from the warming center (Figure 9(f )) andout of alignment with the geopotential gradient (Figure 9(c)). The resultant decrease in cos β at 250 hPa(Figure 9(b)) weakened the thermal wind shear and the SMW descended (Figure 9(a)).

In the ascent region, the 250-hPa temperature trend rotated the temperature gradient away from thewarming center (Figure 9(f )) and into alignment with the geopotential gradient (Figure 10(c)). The resultantincrease in 250-hPa cos β (Figure 10(b)) caused the thermal wind shear to increase and the SMW to ascend(Figure 10(a)). The magnitude of the temperature gradient lacks a significant trend in the descent and ascentregions (Figures 9(b), 10(b)), and so played a minimal role in the SMW pressure changes. Summarizingfor the tropical East Pacific, a center of upper tropospheric warming reoriented the temperature gradient onits poleward side out of alignment with the geopotential field, causing SMW descent. The same warmingreoriented the temperature gradient on its equatorward side into alignment with the geopotential field, causingSMW ascent.

East Pacific (97.5°W, 12.5−17.5°N)

r = −0.77, df = 34

MSE = 394.2 hPa2

160

180

P (

hPa)

∼

220

200

240

260

2801.8

1.5

1.2cos b

P (hPa)∼

1960 1970 1980

Time (year)

1990 2000

Longitude (°W)

−0.1

0.1

0.0

0.2

0.4

cos

b

2

4

6

5

(d) T250hPa,1958–1962 (K)−


(f) T250hPa,1998–2004 (K)−

232

231 230

231.5

231.5

0

15

30

0

0.2

0.2

0.1

230

230

231

231.5 231.

5

0.1

0.1

15

Lat

itu

de

(°N

)

30

0

15

30(a)

(b)

(c)

N

0.3

120 90

3232

1

0

−1

−∇T/ ∇T

∇T

(×

10−3

K/k

m)


∇T

Rdf−1

∇T

v c

os b

(m

/s)

−∇Z / ∇Z

Figure 9. The same as Figure 6 except for the tropical East Pacific (97.5 °W, 12.5–17.5 °N)



East Pacific (80°W, 2.5−7.5°N)

r = −0.56, df = 27

MSE = 1042.2 hPa2

160

200

240

P (

hPa)

∼

280

320

1.8

1.2

cos b

P (hPa)∼

1960 1970 1980

Time (year)

1990 2000

0.1

0.2

0.4

cos

b

0.5

5

10

20

15

(a)

(b)

(c)

N

0.3

1.5

0.9

−∇T/ ∇T

∇T

(×1

0−3 K

/km

)

Rdf −1 ∇Tv cosb

∇T

Rdf

−1 ∇

Tv

cos

b (

m/s

)

−∇Z/ ∇Z

Figure 10. For each summer (June–August) 1958–2004 over the equatorial East Pacific (80 °W, 2.5–7.5 °N): (a) mean P and mean250-hPa thermal wind shear, (b) mean 250-hPa ‖ �∇Tv‖ and mean cos β, and (c) mean 250-hPa −�∇Z/‖ �∇Z‖ and −�∇T /‖ �∇T ‖ for 5-year

blocks centered on 1960, 1970 . . . 2000

3.8. Relative importance of the temperature gradient’s magnitude and orientation

Trends in the thermal wind shear and P may result from a trend in only ‖∇Tv‖, a trend in only cos β,or reinforcing (like-signed) trends in both variables. We now consider the relative importance of ‖∇Tv‖ andcos β trends on a summary level for the portion of the hemisphere where the SMW pressure and the thermalwind shear exhibit their strongest trends (0–25 °N). Two hundred hPa is a useful surface for this analysisbecause most of the SMW pressure time series in Sections 3.4–3.7 pass through or reach this pressure, andit is the closest data surface to the zonally averaged SMW pressure from 0 to 25 °N.

Approximately 66% of the significant SMW pressure trends (by area) are driven by reinforcing, significanttrends in both ‖ �∇Tv‖ and cos β (e.g. the tropical Atlantic). Approximately 23% of SMW pressure trends aredriven by a cos β trend without a reinforcing ‖ �∇Tv‖ trend (e.g. tropical East Pacific). SMW pressure trendsdriven by a ‖ �∇Tv‖ trend without a reinforcing cos β trend are less common (11% of cases) and tend to beweak relative to the cos β-driven changes (<5 hPa/decade) in part because seasonal mean ‖ �∇Tv‖ is typicallyof the order 10−6/Km, whereas cos β is of the order 100. Locations lacking significant SMW pressure trendsmost often (88%) lack significant cos β trends.

3.9. Link to sea surface temperature

We now present a cross-section view of zonally averaged temperature gradient trends in the vicinity ofthe SMW, and explore how these trends relate to SST variability. Figure 11(a) shows the trend in zonallyaveraged summer thermal wind shear. A large region with thermal wind shear decreases exceeding 1 m/s perdecade is found in the upper troposphere between 5 and 20 °N. The summer SMW resided in this region of



100

150

200

250

300

Pre

ssu

re (

hP

a)P

tre

nd

(hP

a/d

ecad

e)

∼

350

400450500

1086420

−210 20

Latitude (°N)30

(b)S

umm

er P range

∼

(a)−0.8

r = 0.98, df = 46,MSE = 0.01

−1.2

TE

qu -

TS

ub (

K)

−1.6 PC 1

TEqu - TSub

TSub

TEqu

July PDOI

1960 1970 1980

Year1990 2000

−2

−1

0

July

PD

OI

1

2

3

−2.0222

221

220

T (

K)

219

218

(c)

(d)

−2

−1

P P

C 1

∼

0

1

2

0.00.0

0.2

0.2

−0.2

0.0

−0.2TSub

TEqu

−0.8

−1.2

−0.4

40 50

Rdf −1 ∇Tv cos b (ms−1/decade)

Figure 11. (a) Trend in the zonally averaged thermal wind shear for summers (June–August) 1958–2004 contoured at 0.2 m/s perdecade with negative values dashed. The bold gray line indicates the pressure of the 47-year mean tropopause and the gray boxes showdomains used to produce the zonal mean air temperatures TEqu and TSub. (b) Trend in the zonally averaged surface of maximum windpressure (P ) with filled circles indicating results significant at α = 0.05 (c) P PC 1 (from Figure 4) shown with the difference TEqu

minus TSub. (d) TEqu, TSub, the July Pacific Decadal Oscillation Index (PDOI) and a third-order harmonic fit to the July PDOI

decreasing thermal wind shear between 1958 and 2004 (right axis of Figure 11(a)), resulting in increasingtrends in zonally averaged P ranging from 3 to 10 hPa/decade (Figure 11(b)).

The weakening of the thermal wind shear is in part attributable to weakening of the meridional componentof ‖ �∇T ‖ by convergence of the air temperature in the latitude belts to the north and south of the weakening.To demonstrate this quantitatively, two boxes are drawn flanking the region of decreasing thermal windshear in Figure 11(a): an equatorial box (100–300 hPa, 0–5 °N with zonally averaged summer meantemperature denoted TEqu) and a subtropical box (100–300 hPa, 20–30 °N with zonally averaged summermean temperature denoted TSub). The difference TEqu – TSub approaches zero over the period 1958–2004 andaccounts for more than 95% of the variability in the leading principal component of summer SMW pressure (PPC 1, Figure 11(c)). The strong correlation in Figure 11(c) indicates that P PC 1 represents a pattern of jointspace-time variability driven largely by the difference in zonally averaged temperature between the equatorialand subtropical upper troposphere. TEqu and TSub converge during the first 20 years of the record because TSub

decreases more rapidly than TEqu (Figure 11(d)). TEqu and TSub undergo positive shifts significant at α = 0.05in 1978, and the faster increase in TEqu produces a convergence of the TEqu and TSub time series (Figure 11(d)).

The rapid convergence of TEqu and TSub in 1978 is coincident with the decadal climate shift documentedin various atmospheric fields (Nitta and Yamada, 1989; and Trenberth, 1990; Graham, 1994; Miller et al.,1994) and biological data (Ebbesmeyer et al., 1991; Francis and Hare, 1994; McGowan et al., 1998; Hareand Mantua, 2000). The multidecadal variability associated with the climate shift is often quantified usingthe SST-based PDOI. The temporal patterns of TEqu and TSub share some similarities with the July PDOI,including a significant change point in 1978 (Figure 11(d)).



The PDOI discussed in the preceding paragraph is based on SST anomalies poleward of 20 °N in thePacific. The notion that SST variability may influence the temporal patterns observed in TEqu and TSub isfurther strengthened by considering SST-related time series geographically closer to TEqu and TSub. Stronglinks have been established between air temperature in the deep tropical troposphere and SST anomaliesassociated with El Nino (Horel and Wallace, 1981; Yulaeva and Wallace, 1994; Sobel et al., 2002). Here, theMEI accounts for variability in summer TSub at the α = 0.05 level with r ≥ 0.50 at lags of −10 to −3 months,where negative numbers indicate that the MEI lags TSub (Figure 12(a)). The TSub versus MEI correlation peakswhen the MEI is taken from March of the contemporaneous year (lag −4 months, Figure 12(a), (b)). For theequatorial latitudes, SST10 °S−10 °N accounts for variability in summer TEqu at the α = 0.05 level with r ≥ 0.50

0.8(a)

(b)

(c)

0.6TEqu vs. SST

TSub vs. MEI

0.4

0.2

0.0−16−14−12−10 −8 −6 −4 −2 0 2 4 6

Lag (months)

Month

Sep

Nov

Jul

Mar

May

May

Mar

Jan

Nov

Jan

Sep

Jul

Co

rrel

atio

n r

3TSub

MEI

r = 0.62, df = 12,MSE = 0.12

r = 0.69, df = 9,MSE = 0.21

2

1

0

−1

−2

3TEqu

SST2

1

0

−1

−2

1960 1970 1980 1990 2000

Year

TS

ub

, Mar

ch M

EI

TE

qu

, May

SS

T10

°S-1

0°N

Figure 12. (a) Bold line shows lagged correlation between TEqu and mean sea surface temperature around the hemisphere between10 °S and 10 °N (SST10 °S−10 °N) where negative lags indicate that SST10 °S−10 °N lags TEqu and symbols show correlations significantat α = 0.05. The thin line shows lagged correlation between TSub and the multivariate El Nino/Southern Oscillation Index (MEI) wherenegative lags indicate that the MEI lags TSub and symbols show correlations significant at α = 0.05, (b) TSub and March MEI, and

(c) TEqu and May SST10 °S−10 °N



at lags of 0 to −10, with peak correlation when SST10 °S−10 °N is taken from May of the contemporaneousyear (lag −2 months, Figure 12(a), (c)).

Additional phenomena potentially influencing upper troposphere air temperature near the SMW includesolar output variability, volcanic emissions, and changing greenhouse-gas concentrations. Considering the neteffect of changes in sulfate aerosol, ozone, and carbon dioxide concentrations over the past several decades,modeling work indicates a maximum net warming in the upper troposphere over the equator, transitioningpoleward to a maximum net cooling in the lower stratosphere over the polar latitudes (Santer et al., 1996).Poleward decreases in the net warming rate may contribute to the decadal SMW pressure changes reportedhere. Such a temperature change pattern is consistent with the observed convergence of TEqu and TSub, becauseTSub is higher than TEqu during the summer (Figure 11(d)).

4. SUMMARY AND DISCUSSION

The summer SMW generally slopes up equatorward, following and residing below the tropopause in polarlatitudes and exhibiting undulations associated with jet cores in middle and lower latitudes. Changes in thealtitude of fast upper tropospheric wind features are climatologically important because of the role of thejet streams in global circulation and surface weather (Riehl et al., 1954; Skaggs, 1967; Nakamura, 1992;Christoph et al., 1997; Rao et al., 2004). The documentation of vertical displacement of upper troposphericwind maxima is also important because such motion may complicate the use of isobaric surfaces for climatestudies of the jet stream (Strong and Davis, 2005). In this research, we used the SMW to map the 47-yearmean position of fast upper tropospheric winds during summer. Variability of summer SMW pressure wasthen documented and linked to variability in the thermal structure of the upper troposphere in accordancewith the relationships expressed in the thermal wind equation. The temperature gradient changes influencingthe SMW were then linked to SST variability on quasi-biennial to decadal timescales, ENSO variability, andthe AO.

The dominant pattern of variability (PC 1) is related to spatially expansive trends as large as 30 hPa/decadeover regions of the tropics and subtropics. Over the Atlantic, a dipole of cooling poleward of warmingweakened and reoriented the upper tropospheric temperature gradient, causing the SMW to descend from150 hPa to below 250 hPa. Similar temperature field changes over Africa caused the SMW to descend from120 to 180 hPa, including a rather abrupt transition coincident with the mid-to-late 1970s climate shift.Over the Pacific, changes in the magnitude of the temperature gradient were less important than changesin its orientation. Rotation of the West Pacific temperature gradient caused approximately 100 hPa of SMWdescent. In the tropical East Pacific, a warming trend changed the horizontal orientation of the temperaturegradients such that zonally oriented bands of descent and ascent were generated.

The gradient of zonally averaged temperature in the upper troposphere between equatorial and subtropicallatitudes accounts for more than 95% of the variability of the leading principal component of SMW pressure.Subtropical upper tropospheric air temperatures are significantly correlated with the MEI, and equatorialupper tropospheric air temperatures are significantly correlated with the mean SST between 10 °S and 10 °N(SST10 °S−10 °N). SST10 °S−10 °N rapidly increased coincident with the decadal climate shift during the mid-to-late 1970s, and the correlated weakening of upper tropospheric temperature gradients accounts for much ofthe observed SMW pressure increases.

The results for the summer SMW presented here include P PC 1 pressure trends in the tropics andsubtropics, and extratropical oscillations associated with the AO appearing as P PC 2. This is in contrast tothe winter season in which the leading pattern of P variability from the hemispheric PCA is more oscillatoryand AO-like, with ENSO-related tropical changes appearing as PC 2 (Strong and Davis, 2006). In addition toproviding a framework for objectively documenting the speed and vertical position of fast upper troposphericwinds, the SMW serves as an informative climate diagnostic when temperature gradients in the stratosphereand mid-to-upper troposphere may be changing. Interesting directions for future work include querying aglobal climate model for SMW pressure trends and oscillations to reinforce understanding of the underlyingmechanisms and determine the sensitivity of the SMW speed and pressure to global climate change. Research



is currently underway to study the joint variability of SMW speed and pressure, particularly in the context ofgreenhouse-gas related temperature variability at continental middle and high latitudes during winter.

ACKNOWLEDGEMENTS

Comments from two anonymous reviewers helped to improve this manuscript. C. Strong was supported undera National Science Foundation Graduate Research Fellowship.

APPENDIX

Given that the magnitude of the geostrophic wind Vg = (u2g + v2

g)1/2 is nonzero and its vector components

are differentiable functions of pressure,

∂Vg

∂p= 1

2(u2

g + v2g)

−1/2(

2ug

∂ug

∂p+ 2v

∂vg

∂p

)(A1)

Substituting the magnitude of the geostrophic wind (f −1‖ �∇Z‖) for (u2g + v2

g)1/2, the geostrophic wind

components for ug and vg, and the thermal wind components in isobaric coordinates for ∂ug/∂p and ∂vg/∂p,we obtain

∂Vg

∂p= f

‖ �∇Z‖[(

− 1

f

∂Z

∂y

) (Rd

fp

∂Tv

∂y

)+

(1

f

∂Z

∂x

) (−Rd

fp

∂Tv

∂x

)]

p∂Vg

∂p= − Rd

f ‖ �∇Z‖[∂Z

∂y

∂Tv

∂y+ ∂Z

∂x

∂Tv

∂x

]

∂Vg

∂ ln p= − Rd

f ‖ �∇Z‖ ( �∇Z · �∇Tv) = −Rdf−1‖ �∇Tv‖ cos β (A2)

where β is the angle between �∇Tv and �∇Z. The quantity −‖�∇Tv‖ cos β is the virtual temperature gradientprojected onto the geopotential height gradient, so Equation (A2) may be converted to natural coordinates

∂Vg

∂ ln p= Rd

f

∂Tv

∂n(A3)

where n is oriented normal to �Vg and defined positive to the left of the flow direction.

REFERENCES

Allan RJ, Nicholls N, Jones PD, Butterworth IJ. 1991. A further extension of the Tahiti-Darwin SOI, early SOI results and Darwinpressure. Journal of Climate 4: 743–749.

Angell JK. 1991. Changes in tropospheric and stratospheric global temperatures, 1958–1988. In Greenhouse-Gas-Induced ClimaticChange: A Critical Appraisal of Simulations and Observations, Schlesinger ME (ed.). Elsevier Science: New York, 231–247.

Angell JK. 2001. Relation of size and displacement of the 300 mbar north circumpolar vortex to QBO, El Nino, and sunspot number,1963–2000. Journal of Geophysical Research 106: 31787–31794.

Angell JK, Korshover J. 1985. Displacements of the north circumpolar vortex during El Nino. Monthly Weather Review 113: 1627–1630.Arkin PA. 1982. The relationship between interannual variability in the 200 mb tropical wind field and the Southern Oscillation. Monthly

Weather Review 110: 1393–1404.Basist AN, Chelliah M. 1997. Comparison of tropospheric temperatures derived from the NCEP/NCAR reanalysis, NCEP operational

analysis, and the Microwave sounding unit. Bulletin of the American Meteorological Society 78: 1431–1447.Bluestein H. 1992. Synoptic-Dynamic Meteorology in Midlatitudes: Principles of Kinematics and Dynamics, Vol. 1. Oxford University

Press: New York, 448.Chase TN, Pielke RA Sr., Knaff JA, Kittel TGF, Eastman JL. 2000. A comparison of regional trends in 1979–1997 depth-averaged

tropospheric temperatures. International Journal of Climatology 20: 503–518.Chen WY, Van den Dool HM. 1999. Significant change of extratropical natural variability and potential predictability associated with

the El Nino/Southern Oscillation. Tellus 51A: 790–802.



Christoph M, Ulbrich U, Speth P. 1997. Midwinter suppression of Northern Hemisphere storm track activity in the real atmosphere andin GCM experiments. Journal of Atmospheric Sciences 54: 1589–1599.

Delworth TL, Mann ME. 2000. Observed and simulated multidecadal variability in the Northern Hemisphere. Climate Dynamics 16:661–676.

Ebbesmeyer CC, Cayan DR, McLain DR, Nichols FH, Peterson DH, Redmond KT. 1991. 1976 step in the Pacific climate: fortyenvironmental changes between 1968–1975 and 1977–1984. Proceedings of the Seventh Annual Climate (PACLIM) Workshop, April1990. Department of Water Resources: California, CA, 115–126, Interagency Ecological Studies Program Technical Report 26.

Endlich RM, Solot SB, Thur HA. 1955. The mean vertical structure of the jet stream. Tellus 7: 308–313.Enfield DB, Mestas-Nunez AM, Trimble PJ. 2001. The Atlantic multidecadal oscillation and its relation to rainfall and river flows in

the continental U.S. Geophysical Research Letters 28: 2077–2080.Francis RC, Hare SR. 1994. Decadal-scale regime shifts in the large marine ecosystems of the North-east Pacific: a case for historical

science. Fisheries Oceanography 3: 279–291.Frauenfeld OW, Davis RE. 2000. The influence of El Nino-Southern Oscillation on the Northern Hemisphere 500 hPa circumpolar

vortex. Geophysical Research Letters 27: 537–540.Frauenfeld OW, Davis RE, Mann ME. 2005. A distinctly interdecadal signal of Pacific ocean–atmosphere interaction. Journal of Climate

18: 1709–1718.Gaffen DJ. 1994. Temporal inhomogeneities in radiosonde temperature records. Journal of Geophysical Research 99: 3667–3676.Gaffen DJ, Sargent MA, Habermann RE, Lanzante JR. 2000. Sensitivity of tropospheric and stratospheric temperature trends to

radiosonde data quality. Journal of Climate 13: 1776–1796.Graham NE. 1994. Decadal-scale climate variability in the tropical and North Pacific during the 1970s and 1980s: observations and

model results. Climate Dynamics 10: 135–162.Hansen JM, Sato M, Ruedy R, Lacis A, Asamoah K, Beckford K, Borenstein S, Brown E, Cairns B, Carlson B, Curran B, de Castro S,

Druyan L, Etwarrow P, Ferede T, Fox M, Gaffen D, Glascoe J, Gordon H, Hollandsworth S, Jiang X, Johnson C, Lawrence N,Lean J, Lerner J, Lo K, Logan J, Luckett A, McCormick MP, McPeters R, Miller R, Minnis P, Ramberran I, Russell G, Russell P,Stone P, Tegen I, Thomas S, Thomason L, Thompson A, Wilder J, Willson R, Zawodny J. 1997. Forcings and chaos in interannualto decadal climate change. Journal of Geophysical Research 102: 25679–25720.

Hare SR, Mantua NJ. 2000. Empirical evidence for North Pacific regime shifts in 1977 and 1989. Progress in Oceanography 47:103–145.

Hastenrath S. 2003. Upper-air circulation of the Southern Oscillation from the NCEP-NCAR Reanalysis. Meteorology and AtmosphericPhysics 83: 51–65.

Horel JD, Wallace JM. 1981. Planetary-scale atmospheric phenomena associated with the Southern Oscillation. Monthly Weather Review109: 813–829.

Hurrell JW, Trenberth KE. 1997. Spurious trends in satellite MSU temperatures from merging different satellite records. Nature 386:164–167.

Hurrell JW, Trenberth KE. 1998. Difficulties in obtaining reliable temperature trends: reconciling the surface and satellite MSU record.Journal of Climate 11: 945–967.

Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M,Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D. 1996. TheNCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society 77: 437–471.

Kanamitsu M, Kistler RE, Reynolds RW. 1997. NCEP/NCAR Reanalysis and the use of satellite data. Advances in Space Research 19:481–489.

Kelly PM, Jones PD, Pengqun J. 1996. The spatial response of the climate system to explosive volcanic eruptions. International Journalof Climatology 16: 537–550.

Kerr RA. 2000. A North Atlantic climate pacemaker for the centuries. Science 288: 1984–1986.Kirchner I, Stenchikov GL, Graf H-F, Robock A, Antuna JC. 1999. Climate model simulation of winter warming and summer cooling

following the 1991 Mount Pinatubo volcanic eruption. Journal of Geophysical Research 104: 19039–19055.Kistler R, Kalnay E, Collins W, Saha S, White G, Woollen J, Chelliah M, Ebisuzaki W, Kanamitsu M, Kousky V, van den Dool H,

Jenne R, Fiorino M. 2001. The NCEP-NCAR 50-year reanalysis: Monthly means CD-ROM and documentation. Bulletin of theAmerican Meteorological Society 82: 247–267.

Konnen GP, Jones PD, Kaltofen MH, Allan RJ. 1998. Pre-1866 extensions of the Southern Oscillation Index using early Indonesianand Tahitian meteorological readings. Journal of Climate 11: 2325–2339.

Kutzbach JE. 1970. Large-scale features of monthly mean Northern Hemisphere anomaly maps of sea-level pressure. Monthly WeatherReview 98: 708–716.

Labitzke K, van Loon H. 1988. Association between the 11-year solar cycle, the QBO and the atmosphere. Part I: The troposphere andstratosphere in the Northern Hemisphere winter. Journal of Atmospheric and Terrestrial Physics 50: 197–206.

Lanzante J. 1996. Resistant, robust and nonparametric techniques for the analysis of climate data: Theory and examples, includingapplications to historical radiosonde station data. International Journal of Climatology 16: 1197–1226.

Lewis JM. 2003. Ooishi’s observation viewed in the context of jet stream discovery. Bulletin of the American Meteorological Society84: 357–369.

Lindzen RS, Giannitsis C. 2002. Reconciling observations of global temperature change. Geophysical Research Letters 29: 1583DOI:10.1029/2001GL014074.

Livezey RE, Chen WY. 1983. Statistical field significance and its determination by Monte Carlo techniques. Monthly Weather Review111: 49–59.

Lorenz EN. 1951. Seasonal and irregular variations of the Northern Hemisphere sea-level pressure profile. Journal of Meteorology 8:52–59.

Mantua NJ, Hare SR. 2002. The Pacific decadal oscillation. Journal of Oceanography 58: 35–44.Mantua NJ, Hare SR, Zhang Y, Wallace JM, Francis RC. 1997. A Pacific interdecadal climate oscillation with impacts on salmon

production. Bulletin of the American Meteorological Society 78: 1069–1079.



McGowan JA, Cayan DR, Dorman LM. 1998. Climate–ocean variability and ecosystem response in the Northeast Pacific. Science 281:210–217.

Mestas-Nunez AM, Enfield DB. 1999. Rotated global modes of non-ENSO sea surface temperature variability. Journal of Climate 12:2734–2746.

Miller AJ, Cayan DR, Barnett TP, Graham NE, Oberhuber JM. 1994. The 1976–77 climate shift of the Pacific ocean. Oceanography7: 21–26.

Minobe S. 1999. Resonance in bidecadal and pentadecadal climate oscillations over the North Pacific: role in climate regime shifts.Geophysical Research Letters 26: 855–858.

Minobe S. 2000. Spatio-temporal structure of the pentadecadal variability over the North Pacific. Progress in Oceanography 47: 381–408.Mintz Y. 1954. The observed zonal circulation of the atmosphere. Bulletin of the American Meteorological Society 35: 208–214.Nakamura H. 1992. Midwinter suppression of baroclinic wave activity in the Pacific. Journal of Atmospheric Sciences 49: 1629–1642.Namias J, Clapp PF. 1949. Confluence theory of the high-tropospheric jet stream. Journal of Meteorology 7: 130–139.Newton CW, Persson AV. 1962. Structural characteristics of the subtropical jet stream and lower-stratospheric wind systems. Tellus 19:

221–241.Nitta T, Yamada S. 1989. Recent warming of tropical sea surface temperature and its relationship to the Northern Hemisphere circulation.

Journal of the Meteorological Society of Japan 67: 375–383.Ogi M, Tachibana Y, Yamazaki K. 2003a. Impact of wintertime North Atlantic Oscillation (NAO) on the summertime atmospheric

circulation. Geophysical Research Letters 30: 1704, DOI:10.1029/2003GL017280.Ogi M, Yamazaki K, Tachibana Y. 2003b. Solar cycle modulation of the seasonal linkage of the North Atlantic Oscillation (NA).

Geophysical Research Letters 30: 2170, DOI:10.1029/2003GL018545.Ogi M, Tachibana Y, Yamazaki K. 2004a. The connectivity of the winter North Atlantic Oscillation (NAO) and the summer Okhotsk

High. Journal of the Meteorological Society of Japan 82: 905–913.Ogi M, Yamazaki K, Tachibana Y. 2004b. The summertime annular mode in the Northern Hemisphere and its linkage to the winter

mode. Journal of Geophysical Research 109: D20114, DOI:10.1029/2004JD004514.Ooishi W. 1926. Raporto de la aerologia observatorio de tateno (in Esperanto). Aerological Observatory Report. 1. Central Meteorological

Observatory, Japan, 213.Palmen E. 1948. On the distribution of temperature and wind in the upper westerlies. Journal of Meteorology 5: 20–27.Palmen E, Newton CW. 1948. A study of the mean wind and temperature distribution in the vicinity of the polar front in winter. Journal

of Meteorology 5: 220–226.Parker DE, Gordon M, Cullum DPN, Sexton DMH, Folland CK, Raynes N. 1997. A new global gridded radiosonde temperature

database and recent temperature trends. Geophysical Research Letters 24: 1499–1502.Pawson S, Labitzke K, Leder S. 1998. Stepwise changes in stratospheric temperature. Geophysical Research Letters 25: 2157–2160.Ramaswamy V, Chanin ML, Angell J, Barnett J, Gaffen D, Gelman M, Keckhut P, Koshelkov Y, Labitzke K, Lin J-JR, O’Neil A,

Nash J, Randel W, Rood R, Shine K, Shiotani M, Swinbank R. 2001. Stratospheric temperature trends: observations and modelsimulations. Reviews of Geophysics 39: 71–122.

Rao BRS, Rao DVB, Rao VB. 2004. Decreasing trend in the strength of tropical easterly jet during the Asian summer monsoon seasonand number of tropical cyclonic systems over Bay of Bengal. Geophysical Research Letters 31: 14103 DOI:10.1029/2004GL019817.

Rasmusson EM, Mo K. 1993. Linkages between 200 mb tropical and extratropical circulation anomalies during the 1986–89 ENSOcycle. Journal of Climate 6: 595–616.

Rayner NA, Parker DE, Horton EB, Folland CK, Alexander LV, Rowell DP, Kent EC, Kaplan A. 2003. Global analyses of sea surfacetemperature, sea ice, and night marine air temperature since the late nineteenth century. Journal of Geophysical Research 108: 4407,DOI:10.1029/2002JD002670.

Reiter ER. 1958. The layer of maximum wind. Journal of Meteorology 15: 27–43.Reiter ER. 1961. Jet Stream Meteorology. The University of Chicago Press: Chicago, Il.Riehl H, Alaka A, Jordan CL, Renard RJ. 1954. The Jet Stream. Meteorological Monographs, Vol. 2, Iss. 7 , American Meteorological

Society: Boston, MA.Rigor IG, Wallace JM, Colony RL. 2002. Response of sea ice to the Arctic Oscillation. Journal of Climate 15: 2648–2663.Robock A, Mao J. 1992. Winter warming from large volcanic eruptions. Geophysical Research Letters 19: 2405–2408.Ropelewski CF, Jones PD. 1987. An extension of the Tahiti-Darwin Southern Oscillation Index. Monthly Weather Review 115:

2161–2165.Santer BD, Hnilo JJ, Wigley TML, Boyle JS, Doutriaux C, Fiorino M, Parker DE, Taylor KE. 1999. Uncertainties in observationally

based estimates of temperature change in the free atmosphere. Journal of Geophysical Research 104(D6): 6305–6334, DOI:10.1029/1998JD200096.

Santer BD, Taylor KE, Wigley TML, Johns TC, Jones PD, Karoly DJ, Mitchell JFB, Oort AH, Penner JE, Ramaswamy V,Schwarzkopf MD, Stouffer RJ, Tett S. 1996. A search for human influences on the thermal structure of the atmosphere. Nature382: 39–46.

Schlesinger ME, Mitchell JFB. 1987. Climate model projections of the equilibrium climatic response to increased CO2. Reviews ofGeophysics 25: 760–798.

Schlesinger ME, Ramankutty N. 1994. An oscillation in the global climate system of period 65–70 years. Nature 367: 723–726.Seager R, Karspeck AR, Cane MA, Kushnir Y, Giannini A. 2004. Predicting Pacific decadal variability. In Earth’s Climate the

Ocean–Atmosphere Interaction, Wang C, Xie S, Carton JA (eds) American Geophysical Union: Washington, DC.Siegel S, Castellan N. 1988. Nonparametric Statistics for the Behavioral Sciences. McGraw Hill: New York.Skaggs RH. 1967. On the association between tornadoes and 500-mb. indicators of jet streams. Monthly Weather Review 95: 107–110.Smith TM, Reynolds RW. 1998. A high-resolution global sea surface temperature climatology for the 1961–90 base period. Journal of

Climate 11: 3320–3323.Sobel AH, Held IM, Bretherton CS. 2002. The ENSO signal in tropical tropospheric temperature. Journal of Climate 15: 2702–2706.Staff Members, Department of Meteorology, University of Chicago. 1947. On the general circulation of the atmosphere in the middle

latitudes. Bulletin of the American Meteorological Society 28: 255–280.



Stendel M, Christy JR, Bengtsson L. 1998. How representative are recent temperature trends? Report 264. Max-Planck-Institute FurMeteorologie; Hamburg.

Straus DM, Shukla J. 1997. Variations of midlatitude transient dynamics associated with ENSO. Journal of Atmospheric Sciences 54:777–790.

Strong C, Davis RE. 2005. The surface of maximum wind as an alternative to the isobaric surface for wind climatology. GeophysicalResearch Letters 32: L04813, DOI: 10.1029/2004GL022039.

Strong C, Davis RE. 2006. Variability in the altitude of fast upper tropospheric winds over the Northern Hemisphere during winter.Journal of Geophysical Research in press.

Sutton RT, Hodson DL. 2005. Atlantic ocean forcing of North American and European summer climate. Science 309: 115–118.Tett SFB, Mitchell JFB, Parker DE, Allen MR. 1996. Human influence on the atmospheric vertical temperature structure: Detection and

observations. Science 274: 1170–1173.Thompson DWJ, Wallace JM. 1998. The Arctic Oscillation signature in the wintertime geopotential height and temperature fields.

Geophysical Research Letters 25: 1297–1300.Thompson DWJ, Wallace JM. 2000. Annular modes in extratropical circulation, Part 1: Month-to-month variability. Journal of Climate

13: 1000–1016.Trenberth KE. 1990. Recent observed interdecadal climate changes in the Northern Hemisphere. Bulletin of the American Meteorological

Society 71: 988–993.van Loon H, Shea DJ. 1999. A probable signal of the 11-year solar cycle in the troposphere of the Northern Hemisphere. Geophysical

Research Letters 26: 2893–2896.van Loon H, Shea DJ. 2000. The global 11-year solar signal in July-August. Journal of Geophysical Research 27: 2965–2968.Wolter K, Timlin MS. 1993. Monitoring ENSO in COADS with a seasonally adjusted principal component index. In Proceedings of

the 17th Climate Diagnostics Workshop, Norman, OK, 52–57.Wolter K, Timlin MS. 1998. Measuring the strength of ENSO–how does 1997/98 rank? Weather 53: 315–324.Yang S, Lau K-M, Kim K-M. 2002. Variations of the East Asian jet stream and Asian-Pacific-American winter climate anomalies.

Journal of Climate 15: 306–325.Yulaeva E, Wallace JM. 1994. Signature of ENSO in global temperature and precipitation fields derived from the microwave sounding

unit. Journal of Climate 7: 1719–1736.Zhang R, Delworth TL. 2005. Simulated tropical response to a substantial weakening of the Atlantic thermohaline circulation. Journal

of Climate 18: 1853–1860.Zhang Y, Wallace JM, Battisti DS. 1997. ENSO-like interdecadal variability. Journal of Climate 10: 1004–1020.


temperature-related trends in the vertical position …

Documents