In this paper, we propose a new method based on the detection of jet cores with the aim to describe the climatological features of the jet streams and to estimate their trends in latitude, altitude, and velocity in the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) and 20th Century reanalysis data sets. Due to the fact that the detection method uses a single grid point to define the position of jet cores, our results reveal a greater latitudinal definition allowing a more accurate picture of the split flow configurations and double jet structures. To the best of our knowledge, these results provide the first multiseasonal and global trend analysis of jet streams based on a daily-resolution 3-D detection algorithm. Trends have been analyzed over 1958–2008 and during the post-satellite period, 1979–2008. We found that, in general, trends in jet velocities and latitudes have been faster for the Southern Hemisphere jets and especially for the southern polar front jet which has experienced the fastest velocity increase and poleward shift over 1979–2008 during the austral summer and autumn. Results presented here show an acceleration and a poleward shift of the northern and southern winter subtropical jets over 1979–2008 that occur at a faster rate and over larger zonally extended regions during this latter period than during 1958–2008.
 Jet streams are fast and relatively narrow air currents thousands of kilometers long, which are found near the tropopause level. Although there are several jets, at global scale they are often classified into two main types: the subtropical jet (STJ) and the eddy-driven jet or polar front jet (PFJ). Occasionally, another high-latitude jet, the so-called Arctic jet, can be found in the troposphere [Shapiro et al., 1987], although this is not a permanent feature. The STJ tends to form along the poleward side of the Hadley cell due to two different mechanisms: the formation of sharp gradients of temperature due to the warm air carried by the Hadley cell from the tropics and also because of angular momentum transport. On the other hand, the PFJ occurs along the region of the polar front where there is a sharp temperature contrast that enhances the formation of baroclinic eddies driving this PFJ.
 Jet streams are probably the most important dynamical systems in the troposphere, since they have a key role in the formation and development of middle-latitude cyclones [Holton, 2004] and in the stratosphere-troposphere exchange. During recent years, tendencies in jet stream characteristics are being considered as potential indicators of climatic change. Although there is wide agreement that jets have shifted poleward in the last two or three decades, there is still some controversy about the magnitude of this migration. Archer and Caldeira  described a poleward shift of the Northern Hemisphere (NH) jets in the period 1979–2001 of 0.16–0.18° per decade. However, Hu and Fu  and Seidel et al.  reported an expansion of the Hadley circulation ranging between 2° and 4.5° of latitude and between 1° and 8°, respectively during 1979–2005, which is equivalent to a poleward shift of the jets ranging between 0.2° and 1.6° per decade [Archer and Caldeira, 2008]. Regarding the Southern Hemisphere (SH), Gallego et al.  and Archer and Caldeira  found a poleward displacement of the southern jets, although only the latter work offers a quantitative analysis. The lack of agreement on the magnitude of the latitude trend extends to the analysis of jet velocities. Again, although there is wide agreement on the strengthening of the SH PFJ [Thomson and Solomon, 2002; Gallego et al., 2005, Russel et al., 2006; Archer and Caldeira, 2008; Polvani et al., 2011], several inconsistencies have been reported for the NH jets and the SH STJ. In this manner, while Archer and Caldeira  found a weakening of the NH jets and the SH STJ, Gallego et al.  reported an increase of the velocity of the southern STJ, and Lorenz and DeWeaver  predicted a strengthening of the tropospheric zonal jets in response to increased CO2 concentration in the Intergovernmental Panel on Climate Change (IPCC) models.
 Probably, one of the reasons for these inconsistencies lies in the intrinsic difficulty in constructing a climatology of the jet streams. Jet streams are so diverse in structure and change so fast that their climatic characterization based in the study of individual “events” is very difficult. The position and strength of the jets vary spatially and temporally, and their meandering structure can be fragmented or can split in branches that may join together again thousands of kilometers away. On occasions, two theoretically contiguous jet systems (Arctic-Polar or Polar-Subtropical) will merge into a single jet. In addition, jet properties may change very fast, making it difficult to follow their structure by using objective methods. First jet climatologies were based on the average wind displayed in a latitude-height cross-section or on the average wind at some predefined level near the tropopause [Blackmon et al., 1977; Kidson, 1999]. These methods are useful to describe the basic characteristics of the zonal circulation but not several jet properties since the instantaneous distributions of the jet streams can be much more complex (see Koch et al.  or Limbach et al. ). Some interesting attempts have been made to associate the jets with structures with physical meaning as an isoheight contour [Davis and Benkovich, 1994; Burnettt and McNicoll, 2000] or geostrophic streamlines of maximum average velocity [Gallego et al., 2005]. To complicate the matter further, more recently, the vertical structure of the jets has also been taken into account. Thus, Koch et al.  described the climatological features of jet streams by defining a jet event as the occurrence of horizontal wind averaged between 100 and 400 hPa above 30 m/s. Strong and Davis  defined a jet core as a local wind maxima above 25.7 m/s belonging to what they define as “surface of maximum wind” in order to exclude stratospheric wind maxima not associated with tropospheric jet streams. Finally, Limbach et al.  developed a segmentation method with per-grid-point localization of merging and splitting events that is able to detect three-dimensional atmospheric features and their development over time and that has been applied to the detection of jet streams. On the other hand, Archer and Caldeira  performed the first study to make a quantitative analysis of the trends in the height, latitude, and velocity of the jets introducing the mass weighted average of wind speed and mass-flux weighted average of pressure and latitude between 400 and 100 hPa.
 As a result of all these research efforts, it seems evident that, to construct a jet stream climatology in a three-dimensional scheme, it is desirable to use both the day-to-day structure of the jet streams and some kind of categorization able to separate the jets (polar, subtropical, or any other type). In this regard, because of the relative smoothness of the SH jets, the association of the jet to relatively simple structures as the previously commented isoheight contour of geostrophic streamlines works well, but in the much more complicated NH, several of the methods proposed up to date have severe problems to deal with the multiple-jet structure. The method based on geostrophic streamlines of Gallego et al.  which achieved a net separation between the two main jet structures in the SH simply did not work for the NH. Even the approach proposed by Archer and Caldeira  considers a single jet for the entire NH while for the SH is clearly able to separate the dual jet structure.
 In this paper, we propose a new method based on the detection of the jet cores to describe the climatological features of the jet streams and to estimate the trends in their latitude, altitude, and velocity. We analyze the trends seasonally looking at their spatial structure and also estimating their global averages. The data and methodology are explained in section 2. In section 3, a new climatology and an analysis of the trends in the jet streams velocity and position are shown. Finally, a summary and discussion is presented in section 4.
2 Data and Method
 In this study, the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) Reanalysis (NCEP) data set [Kalnay et al., 1996] as well as the 20th Century Reanalysis V2 (20CR) data set [Compo et al., 2011] have been used to analyze past changes in jet stream properties. The 20CR data set has been produced assimilating only surface pressure reports and using observed monthly sea-surface temperature making it less susceptible to the influence of the satellite data [Compo et al., 2011]. This makes this data set of special interest for intercomparisons with other reanalyses that make use of satellite data as NCEP. In this way, the comparative analysis of the results obtained from both data sets allows the detection of possible spurious trends due to the introduction of satellite data starting in 1979 in NCEP.
 The availability of different atmospheric variables in these data sets allowed the testing of a wide range of atmospheric fields traditionally related to the presence of a jet stream. Indirect parameters as geopotential height gradient or maxima in potential vorticity gradient related with tropopause foldings were tested as candidates to construct the jet climatology. These approaches have demonstrated its potential for individual case studies [Morgan and Nielsen-Gammon, 1998; Nielsen-Gammon, 2001]. However, when tested, they showed important caveats from a computational point of view, limiting too much the kind of detectable structures, as for example, subtropical jets with no evident tropopause fold or high-wave number PFJ. Perhaps, more important is the fact that a highly derived variable as the vorticity or even a gradient showed not to be adequate to implement automatic tracking techniques because of the “spatial noise” of the resulting fields. Finally, a “the simpler the better” approach was followed by focusing on looking only at the basic definition of a jet as a “fast and narrow air current,” i.e., just by looking at the wind vector. Thus, the daily wind speed at six (seven in the case of 20CR) different pressure levels between 400 and 100 hPa with a horizontal spatial resolution of 2.5° longitude by 2.5° latitude (2° by 2° in the case of 20CR) was used to construct the jet-detection algorithm.
 In this approach, the horizontal wind speed has been computed for each day and at each grid point. Then, at each longitude, every local wind maximum in the region between 400 and 100 hPa is identified. We refer to local wind maxima as those grid points surrounded, in the latitude-height plane, by points with a slower wind speed (Figure 1). The frequency of occurrence of local maxima above 30 m/s during a certain period, expressed as the number of days per month in which wind maxima occurs at each grid point, is interpreted as the probability of jet cores to occur at each location. The use of a threshold criterion in the wind speed to define the jets is somewhat arbitrary, however, in order to minimize the local maxima not being part of any jet stream structure; we decided to use this threshold (30 m/s) as previous studies by Koch et al.  and by Strong and Davis , who used thresholds of 30 m/s and 27.5 m/s, respectively. The average jet core speed and height for a certain period are calculated as the average strength and height of the wind maxima at each grid point.
 In a second step, the selected grid points were categorized as belonging to a STJ or a PFJ by defining latitude limits based on the latitudinal grid point distribution (see section 'Jet Categorization and Global Trends' for a complete discussion). Once the latitudinal criterion was defined, the average latitude, altitude, and velocity of the STJ and the PFJ for each season from 1958 to 2008 were computed. It is important to notice that most of the previous works using reanalysis to quantify jet trends seems quite sensitive to the inclusion of satellite data in 1979 so the trends based on a linear regression of any of these variables were separately evaluated for 1958–2008 and 1979–2008.
3.1 Climatology of the Jet Streams
 Figure 2 depicts the spatial distributions of the seasonally averaged occurrence of jet cores, and Figure 3 shows the corresponding average velocity. In both figures, only the regions with frequencies higher than 1.5 days/month at each grid point have been colored.
 The presented climatology of the jet stream is in good accordance with previous studies [Archer and Caldeira, 2008; Koch et al., 2006; Gallego et al., 2005]. In this way, Figures 2 and 3 are consistent with Koch et al.  showing that in the NH, except for the boreal summer, the jet is strongly asymmetric and evolves from subtropical to middle and high latitudes forming a spiral-like structure. However, since our method evaluates the location of the jets from the jet cores using a single grid point to define the position of the jet cores at each longitude, it is able to localize the jets with a greater latitudinal precision, allowing a more accurate picture of the split flow configurations and double jet structures. For example, in Figure 2 and in the NH, it can identify a strong split flow configuration of the jets over mid- and western Pacific in December-January-February (DJF) and March-April-May (MAM) that converge over North America and split again over the Atlantic Ocean. Here the jet divides in two branches. The fastest (with jet core velocities between 60 and 70 m/s) and most zonally contracted branch, flowing from the tropical western Africa to the western Pacific where it reaches latitudes slightly northward from 30 N, is evident in the previous climatologies [Koch et al., 2006; Archer and Caldeira, 2008]. However, the present climatology also identifies a second, well-separated, weaker branch corresponding to the PFJ over northern Europe and Eurasia (with jet core velocities between 35 and 45 m/s). These two branches merge together at subtropical latitudes over eastern Asia where the jet reaches maximum velocities above 85 m/s. The split flow configuration over the Pacific and the Atlantic oceans decays in the boreal summer and autumn but not over Eurasia where the high-latitude jet reaches its maximum strength in September-October-November (SON) with velocities around 40 m/s.
 In the Southern Hemisphere, the circulation is much more zonally symmetric than in the northern one (Figure 2), and thus, it is possible to define separately subtropical and midlatitude jets in all seasons. The double-banded structure of the SH jets extends from 0° to 250° of longitude during MAM and June-July-August (JJA). However, over the eastern South Pacific, the Southern Cone, and the western South Atlantic Ocean, the frequency of occurrence of jet cores strongly weakens southward from 40°S, and a maximum in the occurrence of jet cores is found around 25°S, indicating a more persistent STJ over these longitudes where the jet core reaches average velocity between 60 and 40 m/s. During the austral summer, the subtropical jet strongly weakens, but it is still clearly defined over Australia and the western South Pacific. In general, the SH STJ is rather stable in latitude, while the SH PFJ displays a noticeable migration toward the pole during the austral winter. Figure 3 shows that SH jets reach the highest velocities at subtropical latitudes over the region extending from the Southwest Indian Ocean to the Southeast Pacific where they can be faster than 65 m/s during the austral winter. The change of the SH PFJ cores velocity with seasons is more reduced than for the STJ and ranges between 40 m/s and 55 m/s, reaching its maximum over the eastern Atlantic and western Indian oceans.
 The jet climatology obtained from the 20CR (not shown) shows almost identical features to those depicted in Figures 2 and 3 for NCEP/NCAR. The consistency in the climatological characteristics of the jets obtained from each data set allows a comparative analysis of the trends in jets position and velocity performed in the following sections.
3.2 Analysis of the Spatial Trends
3.2.1 Northern Hemisphere
 The characterization of the jet cores as a set of grid points of maximum velocity allows the estimation of the spatial patterns of the tendencies in jets position. Figure 4 depicts the tendency in the frequency of occurrence of jet cores for each season for the period 1958–2008 (this period would be referred to as P5808 hereafter) and 1979–2008 (P7908). In all cases, the 95% confidence interval for the slope of the fitted line is defined as the two-sigma uncertainty of the slope parameter (striped areas).
 Figure 4a illustrates the trend in the frequency of occurrence of jet cores in DJF and during P5808 for NCEP. It reveals statistically significant trends along the NH STJ branch over the Middle East and Southern Asia. Here, a significant decrease in the frequency of occurrence of jet cores of −0.4 days/month per decade is detected around 25°N–30°N, while an increasing trend of 0.5 days/month per decade is observed over a parallel band around 30°N–35°N. This spatial distribution with positive (negative) trends, forming parallel bands at the northern (southern) sides of the jet branch, suggests a significant poleward shift of the NH STJ in this region during this 51 year period. Furthermore, Figure 4b reveals that the poleward shift of the NH STJ has been more intense and has affected the entire NH STJ during P7908. For this period, the two-band structure with negative (positive) trends over the southern (northern) limit of the NH STJ band extends eastward over the Pacific and North America and westward along North Africa to the eastern North Atlantic where the trends reach absolute values above 0.6 days/month per decade.
 Figure 4a also reveals the evolution of the jets over the North Atlantic. During P5808 (figure 4a), a negative trend is observed over the North Atlantic basin between 20°N and 40°N, while positive trends are found southward and northward from this region, suggesting a general equatorward shift of the Atlantic NH STJ and a poleward shift of the Atlantic NH PFJ and, in consequence, a significant trend toward the separation of the jets. However, a notable change in this pattern is observed over P7908 (Figure 4b). In this latter period, it seems clear that both the NH STJ and PFJ are experiencing a poleward shift (statistically significant in the case of the NH STJ).
 Figure 4c, depicting the trends computed from NCEP and for MAM, shows increasing (decreasing) frequencies of NH STJ cores southward (northward) from 15 N over the North Pacific, suggesting an equatorward shift of the NH STJ over this region during P5808. An equatorward shift of the NH PFJ is also observed over the North Pacific Ocean. During P7908, an increasing trend in the frequency of jet cores is found just south of the Bering Strait, between 50 N and 70 N and across the United States, while a decreasing trend is observed over the Pacific at latitudes around 40 N and across northern Canada. This result suggests that the NH PFJ tended to be more zonally asymmetric over P7908, reaching high latitudes over the Pacific and turning toward middle latitudes over North America.
 Finally, the spatial patterns of the trends obtained from NCEP data for the northern summer and autumn (Figures 4e and 4g) show rather low and hardly significant tendencies that do not indicate any latitudinal shifts.
 The spatial patterns of the trends obtained from 20CR (Figure 5) are consistent with the NCEP results for the NH. They show almost identical trends of spatial patterns for the NH winter and spring during P7908, corroborating the poleward shift of the NH winter jet over zonally extended regions, and the poleward (equatorward) shift of the spring PFJ over the North Pacific (North America and the North Atlantic). The only remarkable difference is observed in SON when 20CR results show a poleward shift of the jet over the Pacific and the Atlantic that cannot be recognized in NCEP results.
3.2.2 Southern Hemisphere
 The SH STJ trends for P5808 (Figures 4a, 4c, 4d, and 4g) show opposite shifts over two different regions. In all seasons except in DJF, when the SH STJ is very weak, the trends suggest a clear poleward shift of the SH STJ over Australia and the Pacific Ocean indicated by negative (positive) trends in the frequency of occurrence of jet cores at around 20°S–25°S (30°S–35°S), while an equatorward shift of the SH STJ is found over the Indian Ocean and southern Africa. However, during P7908, no equatorward shift of the SH STJ is observed over any region (Figures 44b, 4d, 4f, and 4h), and on the contrary, the trends suggest a poleward shift of the STJ along extended regions and occurring at faster rates than during P5808. This is evident in the austral winter when a significant and zonally symmetric poleward shift of the SH STJ is suggested by a decreasing (increasing) probability of jet cores at around 20°S (30°S) of latitude (Figure 4f). In SON, the poleward shift of the SH STJ observed over the South Pacific during P5808 intensifies during P7908 (Figure 4h), and in MAM (Figure 4d), a poleward shift of the SH STJ is observed over the Indian Ocean and Western Australia where equatorward displacements were found during P5808.
 The spatial trends obtained from the 20CR data set (Figure 5) also show opposite shifts of the SH STJ over the two analyzed periods with equatorward (poleward) shifts during P5808 (P7908). However, in contrast to the trends in NCEP, the equatorward shift of the STJ observed in JJA and SON during P5808 is observed over zonally symmetric regions.
 Tendencies corresponding to the SH PFJ for the period P5808 reveal a poleward shift of this jet during DJF and MAM (Figures 4a and 4c). For these seasons, these figures exhibit positive trends to the south of the SH PFJ branch, at latitudes between 5°S and 60°S, while at latitudes around 40S, negative trends are observed over certain regions as the Indian Ocean and the Southern Cone. This poleward shift is neither observed in JJA nor in SON (Figure 4g) when a significant reduction in the probability of PFJ cores is observed at high latitudes, along the limit between the Atlantic and the Southern Oceans. Again, as in the case of the SH STJ, the trend pattern changes during P7908. Figures 4b and 4d are in qualitative agreement with their counterpart for P5808 but make evident an intensification of the SH PFJ poleward shift during DJF and MAM. The pattern of the trends for JJA over P7908 (Figure 4f) change importantly with respect to P5808 (Figure 4e) and reveal a zonally symmetric increase of the frequency of jet cores at latitudes poleward from 60°S to the pole, suggesting a poleward shift of the winter SH PFJ not observed during P5808. On the contrary, the pattern of the trends for SON suggests a rather general equatorward shift of the SH PFJ cores as indicated by the decreasing (increasing) probability of jet cores at around 50°S–60°S (40°S–50°S) of latitude.
 Important differences are found between 20CR and NCEP in the case of the SH PFJ spatial trends during the longer period, P5808 (Figure 5). During this time, the frequency of occurrence of jet cores significantly increases southward from 50°S in DJF and MAM in both data sets. However, in the case of 20CR, the occurrence of jet cores extensively increases southward from 50°S and to 90°S also in JJA and SON, while a reduction of jet core frequency is observed when NCEP data are used. During P7908, the spatial trends for DJF and MAM are again consistent in both data sets (Figures 5b and 5d); however, in JJA and SON (Figures 5f and 5h), the increase in the occurrence of PFJ cores is higher and extends more poleward in the case of 20CR.
3.3 Jet Categorization and Global Trends
 Results presented in Figures 4 and 5 show that there have been some significant changes in the location of the jet cores during the last decades. However, in order to globally evaluate these trends and to make the comparison with other similar studies easier, a categorization of the jet cores in terms of subtropical and polar circulations has been carried out when possible.
 The categorization of the jets has been based on a latitudinal criterion. The latitude limits were defined using the zonal integral of the probability of jet cores occurrence (not shown). Minima in this integral were taken as the latitude values separating polar and subtropical circulations. Table 1 shows the resulting thresholds. In the SH, it is possible to define the 40°S threshold as the separation limit between subtropical and polar latitudes, with this value being independent of the season. Unfortunately, for the NH it has been possible to define a separate STJ only for DJF. For the equinoxes, a single jet has been defined between 10°N and 70°N, while limits for the northern summer jet are 30°N and 60°N. In order to test the sensitivity of the results to the latitude criterion used for the separation of STJ and PFJ, we also evaluated the trends changing these thresholds in 5° of latitude northward and southward. Results (not shown) showed, in general, very slight differences in the magnitude of the trends. This meets the expectations since, for the cases where a categorization has been performed, the occurrence of most of the STJ and PFJ cores is confined to relatively narrow latitudinal bands (Figure 2), and so, as far as the selected latitude criteria keep these bands within the limits, the computed trends will not differ significantly.
Table 1. Range of Latitudes for the Categorization of Velocity Maxima as Belonging to a Subtropical or Polar Jet
 Table 2 shows global average trends for latitude, pressure level, and velocity of the jets classified by season and latitude according to the values given in Table 1. This table includes the trends for P5808 and P7908 and for both NCEP and 20CR data sets. Those trends which are statistically significant at the 95% confidence level are in bold, and those cases where the trends are significant and of the same (opposite) sign depending on the used data set are represented in bold blue (bold red). Results represented in Table 2 make evident that the highest level of agreement between both data sets is reached for the SH PFJ. In this case, trends show an increase in the velocity of the jet over both periods and at almost all seasons as well as a poleward shift during the southern summer and autumn (DJF and MAM). The values of the velocity trends obtained from NCEP and 20CR are very similar over P7908 when the strongest accelerations of the SH PFJ are observed during DJF, MAM, and SON and range between 0.57 and 0.81 m/s per decade. The velocity trend is not so intense during JJA when it reaches values around 0.3 m/s/decade in both data sets. Although both data sets show general positive velocity trends over P5808, they disagree about its magnitude, and in this regard, while NCEP shows smoother trends during this longer period than over P7908, 20CR shows velocity trends between 1 and 1.4 m/s per decade that are well above the values obtained for P7908. With regard to the trends in the latitudinal position of the SH PFJ, both data sets concur that the fastest poleward shift occurred in the southern summer (DJF) and that it intensifies over P7908, when the trend reaches values around 0.5° per decade in both data sets. Results shown in Table 2 suggest a slower poleward shift of the SH PFJ in MAM and JJA, although this is not statistically significant in most cases.
Table 2. Slope Parameter for the Latitude, Pressure Level, and Velocity Trends of the Jet Coresa
aExpressed in degrees/decade, hPa/decade, and m/s/decade, respectively, as computed from the 20th Century and NCEP/NCAR reanalyses for the periods 1958–2008 and 1979–2008. Bold values indicate those trends that are statistically significant at the level of 95%. Those cases where the trends are significant and of the same (opposite) sign in each of the two data sets are represented in bold blue (bold red).
 Figure 6 compares the changes in the mean zonal wind and the temperature for NCEP and 20CR over P5808 and P7908. It evidences that discrepancies found in the trends of the PFJ properties are fully consistent with different temperature trends in NCEP and 20CR data sets. In the case of 20CR, temperature trends over both periods show a warming of the upper troposphere and lower stratosphere between 40°S and 40°N and a cooling of the polar stratosphere that extends downward to the upper troposphere. This temperature trend pattern enhances the poleward temperature gradient which leads to an increase of westerly winds with height at around 50°S consistent with the observed acceleration of the PFJ. Temperature trends in NCEP show a somehow different picture. For example, the tropospheric warming does not extend so high into the stratosphere, and a much smaller cooling of the southern polar upper troposphere over P5808 is found. This leads to a weaker increase of the wind velocity at mid-high latitudes that is consistent with the weaker PFJ velocity increase obtained from NCEP over P5808. However, in spite of these differences, Figure 6 suggests that also in NCEP data, the acceleration of the SH PFJ and its poleward shift observed in DJF and MAM can be associated with an increase in zonal mean temperature difference between the middle and high latitudes related to the cooling of the polar upper troposphere and lower stratosphere. Both NCEP and 20CR concur that the greatest cooling of the polar upper troposphere occurs in DJF and over P7908, which is consistent with the fact that the greatest poleward shift of the PFJ occurs during this period.
 Table 2 makes evident an important lack of agreement on the altitude trends in the NH jets and the SH STJ, which is particularly significant over P5808. 20CR results depict a significant rise in the altitude of the NH jets, at almost all seasons and over the two analyzed time periods, ranging from −1.0 to −2.1 hPa per decade. Furthermore, except for MAM, 20CR trends make evident a faster rise of the NH jets over P7908. The 20CR depicts a similar pattern for the SH STJ altitude with a rise ranging between −0.4 and −1.4 hPa per decade over P5808 and between −1 and −1.7 hPa per decade over P7908. On the contrary, NCEP results show significant descents of the NH jets and of the SH STJ over P5808 at almost all seasons with rates as high as 3 hPa per decade in the case of the SH STJ. Over P7908, NCEP shows a flatter behavior and, even in some cases, a nonsignificant rise of the jets. Figure 6 makes evident that the observed trends in the temperature of the upper troposphere between subtropical and polar latitudes are quite different for each data set over P5808. In the case of 20CR, temperature trends change sign at around 30° of latitude in each hemisphere with warming (cooling) at lower (higher) latitudes leading to a strengthening of the poleward temperature gradient and the westerly wind shear at subtropical latitudes in both hemispheres, which is consistent with a rise in the altitude of the subtropical jets. However, temperature trends obtained from NCEP show a tropospheric warming that extends further northward in the NH and southward in the SH and a cooling of the upper troposphere which is much lower than in 20CR and only noticeable at polar latitudes. This temperature trend pattern leads to a weakening of the poleward temperature gradient and the westerly wind shear at subtropical latitudes in both hemispheres, resulting in a descent of the subtropical jet cores.
 Also, Table 3 shows some cases where consistency is found between the global trends obtained with both data sets but where these trends are not statistically significant. In these cases, we can compare these global trends with the spatial trends in the frequency of jet cores depicted in Figures 4 and 5 in order to check if they show consistent results. In this regard, patterns of spatial trends obtained for both data sets (Figures 4 and 5) showed a poleward shift of the northern winter jet and of the southern winter STJ over P7908 that was evident at all longitudes (section 3.2). Global trends for this time period (Table 2) are consistent with this picture showing a poleward shift of the northern winter jet occurring at a rate around 0.1° per decade in both data sets and a poleward displacement of the southern winter STJ occurring at a rate of −0.33° per decade in NCEP and of −0.11° per decade in 20CR. In both data sets, the poleward shift of the winter jet cores was accompanied by a velocity increase. In the case of the NH jet, the acceleration was of 0.29 and 0.41 m/s per decade in 20CR and NCEP, respectively. Concerning the SH STJ, the velocity increase is statistically significant in both data sets during the southern winter when it reaches 0.6 m/s per decade. These trends can be observed at all seasons over P7908. Also with regard to the SH STJ, global trends obtained from NCEP show a significant poleward shift also during SON, not observed in 20CR. This can be due to the fact that this shift occurs only over the southern Pacific (Figures 4 and 5), and then it need not be observed in global averages.
4 Summary and Discussion
 This paper proposes a direct jet-detection method purposely designed to be as direct as possible. This method is based in the premise that any jet stream has a core in the upper troposphere where the wind speed is maximum in the height-latitude plane. Since any local maximum is not necessarily part of a jet stream, we restricted our search to wind maxima between 400 and 100 hPa and above 30 m/s, and we interpreted the frequency occurrence of these wind maxima as the probability of jet cores to occur at each grid point. Based on the detection of these jet cores, trend analysis of the latitude, altitude, and velocity of the jet streams have been performed. This method has some advantages with respect to those approaches that use a single isobaric surface [Gallego et al., 2005] since, potentially, this should capture the short-term temporal and spatial variations of the jet streams altitude. This is particularly important when computing the trends in the position and speed of the jets, since not taking into account their altitude can give rise to wrong estimations [Strong and Davis, 2006b]. On the other hand, in contrast to those climatologies based on temporal averages of any flow variable, this approach estimates the average features of the jet streams from the frequency of occurrence of local wind maxima evaluated with daily resolution at each grid point, allowing to detect even the less frequent or weaker jet branches, which otherwise could hide behind temporal or spatial wind velocity averages, and reflecting both the steady and transient variability of the jets and, thus, offering a more accurate picture of the split flow configurations and double jet structures.
 With regard to the trends in the position and velocity of the jets, a global overview of the results reveals that the change has been greater for the SH PFJ. It is also for the SH PFJ that the obtained results from both NCEP and 20CR reach the highest level of agreement. NCEP and 20CR concur with a significant and zonally symmetric poleward shift of the SH PFJ and with a significant velocity increase reaching 0.7 m/s per decade in DJF and MAM during P7908. The acceleration and poleward shift of the SH PFJ is associated with an increase in zonal mean temperature difference between the middle and high latitudes due to the cooling of the polar upper troposphere, which is most intense in DJF when the acceleration and the poleward shift of the SH PFJ also maximizes. This is consistent with Polvani et al. , who related the poleward shift of the southern summer PFJ and the strengthening of the southern annular mode to the cooling of the lower-stratospheric polar cap caused by ozone depletion.
 Regarding the SH STJ, our results show an equatorward shift in JJA and SON occurring over different longitudinally extended regions during P5808 that turns into a zonally symmetric poleward shift during P7908 in both NCEP and 20CR data sets. The fact that both data sets agree with this change in the trends along the second half of the 20th century suggests that it is not associated with NCEP assimilation of satellite data in 1979 since this kind of data is not included in the 20CR reanalysis. The poleward shift of the southern winter STJ over P7908 is also noticeable in the global-average trends that show a displacement of 0.3 and 0.1°/decade in NCEP and 20CR, respectively, and occurs together with an acceleration of the jet cores of around 0.6 m/s per decade.
 Poleward shifts of the SH STJ and PFJ shown in this study are stronger than those reported by Archer and Caldeira  that found annually average poleward rates of around −0.1° per decade over 1979–2001. This discrepancy may be due to the different time period or the different nature of the methodology used but also to the strong seasonal changes of the Southern Hemisphere trends. We show that the seasonal variability of the trends is greater in the SH, especially for the PFJ that reaches a poleward trend of 0.5° per decade in DJF, while it shows a nonsignificant equatorward shift in SON. On the other hand, the SH jet cores velocity increase is consistent with Lorenz and DeWeaver  that analyzed the zonal wind response to increased CO2 concentration in the IPCC models and predicted a strengthening of the tropospheric zonal jets. Archer and Caldeira  obtained an increase in the velocity of the PFJ ranging from 0.2 to 0.4 m/s per decade in annual averages, which is slightly below the seasonal values estimated here ranging between 0.3 and 0.8 m/s per decade. However, with regard to the SH STJ, Archer and Caldeira  showed a velocity decrease of around −0.4 m/s per decade, which contrasts with the acceleration of the SH STJ cores estimated here.
 Our results for the NH jets reveal an enhancement of the poleward shift of the winter STJ during P7908 with respect to P5808. The spatial trends for the NH jets obtained from NCEP and 20CR show a significant poleward shift of the NH STJ in DJF over Middle East and southern Asia during P5808, while over the North Atlantic, a tendency for the NH STJ and the PFJ to separate is observed, which is also consistent with Strong and Davis . However, our results reveal a poleward shift of the NH winter STJ occurring along all longitudes over P7908 in both NCEP and 20CR. In global-average terms, both NCEP and 20CR data sets show a poleward shift of the winter STJ during P7908 of around 0.1° per decade, although this is not significant in either NCEP or 20CR. Despite not being statistically significant, the magnitudes of the poleward shift estimated here compare well with those derived by Archer and Caldeira  which, in annual averages, showed a poleward shift of the Northern Hemisphere jet streams at rates of 0.165 and 0.185°/decade during the period 1979–2001 in the ERA-40 and NCEP reanalysis, respectively. Taking into account these values, the present study makes evident that the poleward shift of the jets has been faster in the Southern Hemisphere, where the poleward rates of the SH STJ and PFJ are, respectively, 3 and 5 times higher than for the northern jets.
 Previous studies have shown an increase of the tropopause altitude in the last decades at rates around −1.5 hPa/decade [Lorenz and DeWeaver, 2007; Seidel et al., 2007]. Santer et al.  found an increase in the height of the tropopause by −2.16 hPa per decade over 1979–2000 in NCEP. The tropopause rise implies a rising of the jet streams that has been reported by Archer and Caldeira . In the present study, the trends obtained from 20CR are consistent with these previous results and show an ascent of the NH jets and of the SH STJ of around −1 to −2 hPa per decade during P7908. However, 20CR trends do not confirm the rise of the SH PFJ, and, in fact, it even shows a significant descent over the period P5808. A different picture is obtained from NCEP where significant descents are found for the NH jets and the SH STJ during P5808, and nonsignificant trends are observed over P7908. On the other hand, NCEP shows a significant rise during the winter PFJ over P7908. The lack of agreement between NCEP and 20CR on the trends in the altitude of the SH STJ and the NH jets is consistent with the differences in the temperature trends obtained from each data set over P5808. For this time period, NCEP shows a warming trend in the upper troposphere that increases with latitude from the tropics to middle latitudes in each hemisphere, while 20CR is characterized by a cooling trend of the upper troposphere that intensifies poleward from 30–40° of latitude in each hemisphere. The important lack of agreement on the temperature trends in the SH between NCEP and 20CR should not be interpreted as a result of the assimilation of satellite data by NCEP from 1979. On the contrary, this discrepancy arises from the presatellite period, 1958–1978 (not shown), when NCEP shows a strong mid-high latitude tropospheric warming, while 20CR estimates a cooling trend that intensifies poleward from 30°S to 40°S.
 Finally, we would like to stress that typically the discrepancies in the assessment of trends in jet stream properties have been attributed to deficiencies in the methodology or in the data. Our study points to another possible cause. Usually, as a means for simplifying the presentation of the results, trends of different jet stream characteristics are presented as zonally averaged magnitudes (as in our Table 2). However, it is also clear that the longitudinal differences in the jet stream trends may be important (Figures 4 and 5). In view of our results, it seems evident that a precise characterization of jet stream trends is not really attainable with zonally averaged values, the use of local trends being mandatory to fully understand the changes in the complicated jet stream structure.
 This work was funded by the Ministry of Education and Science (Government of Spain) under the TRODIM project (Diagnosis and modelization of the extratropical tropopause) and the project “Diagnosis of the Northern Hemisphere jet stream: A new perspective from tropopause maps” (Ref: HP2008-0073). Part of the funding was provided by the Government of Andalusia under the Plan Andaluz de Investigación, Desarrollo e Innovación (PAIDI group RNM-356). The 20th Century Reanalysis V2 data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/