A dynamically based climatology is derived for Northern Hemisphere atmospheric blocking events. Blocks are viewed as large amplitude, long-lasting, and negative potential vorticity (PV) anomalies located beneath the dynamical tropopause. The derived climatology [based on the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40)] provides a concise, coherent, and illuminating description of the main physical characteristics of blocks and the accompanying linear trends. The latitude–longitude distribution of blocking frequency captures the standard bimodal geographical distribution with major peaks over the North Atlantic and eastern North Pacific in all four seasons. The accompanying pattern for the age distribution, the genesis–lysis regions, and the track of blocks reveals that 1) younger blocks (1–4 days) are more prevalent at lower latitudes whereas significantly older blocks (up to 12 days) are located at higher latitudes; 2) genesis is confined predominantly to the two major ocean basins and in a zonal band between 40° and 50°N latitude, whereas lysis is more dispersed but with clear preference to higher latitudes; and 3) the general northeastward–west-northwest movement of blocks in the genesis–lysis phase also exhibits subtle seasonal and intra- and interbasin differences. Examination of the intensity and spatial-scale changes during the blocking life cycle suggests that in the mean a block’s evolution is independent of the genesis region and its eventual duration. A novel analysis of blocking trends reveals significant negative trends in winter over Greenland and in spring over the North Pacific. It is shown that the changes over Greenland are linked to the number of blocking episodes, whereas a neighboring trend signal to the south is linked to higher-frequency anticyclonic systems. Furthermore, evidence is adduced that changes in blocking frequency contribute seminally to tropopause height trends.
An atmospheric blocking episode is characterized by a disruption of the prevailing westerly circumpolar flow in the extratropics through the presence of a major quasi-stationary high pressure system that persists for at least several days (e.g., Rex 1950a; Dole and Gordon 1983; Lupo and Smith 1995b). The block’s large amplitude signifies a major synoptic event with an intermediate life span from day-to-day synoptic-scale to weeks-to-months climate variations. Hence, blocking as phenomenon is of intrinsic dynamical interest and its prediction relevant both to medium-range weather forecasting and to short-term climate prediction (e.g., Pavan et al. 2000; Shabbar et al. 2001: Trigo et al. 2004; Scherrer et al. 2006). Moreover since blocking exhibits a preferred spatial distribution [the Euro–Atlantic and Pacific sectors in the Northern Hemisphere (e.g., Dole and Gordon 1983; Tibaldi and Molteni 1990)] a significant and sustained trend in its amplitude and/or location would impact upon the signature of climate change.
In light of the above a detailed climatology of atmospheric blocks could help provide a better understanding of the phenomenon’s dynamics and also be useful for assessing its link to and implications for climate and climate change. In particular it would be helpful if the climatology cataloged the phenomenon’s structure, spatial scale, intensity, duration, location, and space–time evolution as well as supplying information on any accompanying trends. Extant climatological studies have addressed many of these topics, and the main results are noted below. Most emphasis has been placed upon deriving a detailed climatology for the frequency of blocking and this has entailed constructing an index to objectively detect block occurrence [cf. the pioneering contributions of Elliott and Smith (1949) and Rex (1950a, b)].
One set of studies has yielded a quasi one-dimensional blocking detection from the latitudinal gradient of the 500-hPa geopotential height field (Z500) (e.g., Lejenäs and Økland 1983; Tibaldi and Molteni 1990; Lupo and Smith 1995a). In a refined approach (Pelly and Hoskins 2003) block occurrence was inferred by a reversal of the latitudinal gradient of potential temperature on the dynamical tropopause [2-PVU isosurface (1 PVU = 10−6 m2 s−1 K kg−1)]. The resulting climatologies show a similar overall pattern with a bimodal longitudinal distribution of pronounced peaks centered in the Euro–Atlantic and the Pacific. However, there are some distinctive interstudy variations in terms of secondary peaks and seasonality such as the markedly increased blocking frequency (Pelly and Hoskins 2003) over most of the extratropics and a marked longitudinal shift of the peak over the Pacific. A recent study by Barriopedro et al. (2006) addresses Northern Hemispheric blocking climatology using a modification of the Tibaldi blocking index (Tibaldi and Molteni 1990).
Another set of studies provides a latitude–longitude blocking climatology (e.g., Dole and Gordon 1983; Shukla and Mo 1983; Sausen et al. 1995). These are derived with indices that assess a block’s spatial extent, and are for the most part based upon detecting significant and sustained positive departures from the mean climatological 500-hPa height field. The resulting climatologies are in broad agreement, showing two major blocking regions that are, in addition, in harmony with the one-dimensional climatologies.
Significant attention has also been devoted to examining the lifetime of blocks. Some of the extant climatologies show a peak in the persistence distribution of 5 days (Lejenäs and Økland 1983; Lupo and Smith 1995a) and 8 days (Treidl et al. 1981) with an extended tail toward longer lasting blocks. A related aspect is the number of blocking events lasting at least a given number of days (Dole and Gordon 1983; Pelly and Hoskins 2003). The compiled climatologies indicate, on the one hand, that the probability that an anomaly will continue is initially increasing with increasing duration (up to ∼5 days) and is thereafter nearly constant (Dole and Gordon 1983) and, on the other hand, highlight the different persistence characteristics for anomalies lasting shorter/longer than 4 days (Pelly and Hoskins 2003).
Less attention has been placed upon compiling climatologies of other ancillary blocking characteristics (e.g., spatial scale, intensity, genesis and lysis, and movement). Estimates of the spatial scale, as measured by the maximum width (i.e., largest longitudinal extension) during a blocking episode, yielded a frequency diagram (Lejenäs and Økland 1983) with a peak at 30° with an attendant wide spread from 10° to 120°. Consideration of block intensity indicated that winter blocking events are more intense and larger than their summer counterparts with a significant correlation between scale and intensity (Lupo and Smith 1995a). Examination of block genesis (Lupo and Smith 1995a) indicated that the preferred locations are on the eastern end of the three major storm track regions and west of the time-mean positive 500-hPa height anomalies. Examination of the movement suggests that short duration episodes tend to move eastward, while longer-lasting episodes can experience a westward component (Rex 1950a; Lejenäs and Økland 1983).
Other aspects of blocking can now be examined with the availability of longer-term reanalysis datasets [the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) (1948–2005) and the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) (1958–2002)], and include the variability and trends in block frequency. Note however that there are limitations to the use of these particular datasets for examining linear trends (cf. Bengtsson et al. 2004a, b; Bromwich and Fogt 2004), and moreover caution is required for their use at higher latitudes or the Southern Hemisphere. Notwithstanding, linear trend analysis can be helpful in assessing the recent evolution of the climate. In this vein two recent trend calculations (Wiedenmann et al. 2002; Chen and Yoon 2002) examined, respectively, the time periods 1968–98 and 1954–97 using the NCEP–NCAR reanalysis dataset. The trends, derived for either the entire Northern Hemisphere or for relatively large regions such as the Atlantic and Pacific basin, were weak and not statistically significant. In contrast a detailed trend analysis based upon subdividing the basins into western and eastern parts (Luo and Wan 2005) reveals significant negative blocking trends, both in the Atlantic and Pacific basin. To the present, no trend analysis has been derived based upon a two-dimensional distribution of blocking frequency.
The present study is designed to complement and extend the aforementioned studies. The overarching aim is to derive a dynamically based set of interrelated climatologies useful for assessing seminal block characteristics and trends. To this end the study is based upon the adoption of a recently introduced (Schwierz et al. 2004) quasi-three-dimensional dynamically based index for identifying a block (section 2). The use of this index is motivated by the fact that it captures the core PV anomaly of the block at the tropopause region and can thereby provide further insight on the dynamical evolution and maintenance of blocks. This index is deployed herein to illustrate an archetypical block evolution in the Euro–Atlantic sector incorporating a comparison with other indices (section 3); establish a two-dimensional climatology of blocks including a comparison with comparable shorter-lived systems (section 4); examine the spatial distribution of genesis and lysis (section 5) and the intervening movement (section 6), intensity, and scale evolution (section 7); and evaluate and analyze the two-dimensional linear trend (section 8). In addition some of the implications of the derived results are discussed in the context of the atmosphere’s overall dynamics and climate (section 9).
2. Data and methodology
The database for this study is the ERA-40 reanalysis of the ECMWF (Uppala et al. 2005) for the time period December 1957 to February 2002. It is used with the set’s full temporal resolution of 6 h and the original model fields are interpolated horizontally onto a 1° × 1° grid. Analysis is undertaken for the four individual seasons but with a predominant emphasis on the autumn [September–November (SON)], winter [December–February (DJF)], and spring [March–May (MAM)].
The present study is cast in the potential vorticity (PV) framework that provides a transparent and rational basis for formulating a blocking index. During deep atmospheric blocking episodes a strong negative PV anomaly is located in the immediate vicinity of, and below, the tropopause (e.g., see Fig. 2 of Schwierz et al. 2004). Indeed a major negative anomaly in this layer connotes strong anticyclonic circulation at and below that layer that can counter the westerly flow on its equatorward rim, and also connotes an underlying region of locally lower temperature and higher pressure.
The foregoing constitutes the dynamical basis for the averaged PV (APV*) blocking indicator (Schwierz et al. 2004) that is adopted herein. To calculate the indicator the hydrostatic form of the Ertel PV (Ertel 1942) is first computed and then averaged vertically between 500 and 150 hPa. The resulting two-dimensional field is referred to hereafter as the vertically averaged PV (VAPV), and the anomaly at a specific location is defined as the difference between the instantaneous VAPV and its in situ climatological mean value. Thereafter a tracking algorithm is applied to follow the entire blocking life cycle. In the present study a block is required to satisfy a threshold value of −1.3 PVU for the closed contour in the anomaly field and to persist for at least 5 days (for further details of the index and the procedure to derive the climatology see Schwierz et al. 2004).
For each identified block a record is kept throughout its life cycle of its location (i.e., center of mass), intensity [mean of the lowest 10 gridpoint values in the VAPV anomaly field (PVU)], and spatial scale (km2). The first and last registration of a block (assessed at 6-h intervals) is taken to correspond with the block’s genesis and lysis, and the time difference defines the block’s duration. The resulting registry forms the database for our compilation of the geographical distribution of block frequency, mean block age, genesis and lysis, movement, spatial-scale evolution, and trends.
3. An illustrative example
In this section consideration is given to the evolution of one individual block over its entire life cycle. The purpose is twofold. First to illustrate the main characteristics of blocking as captured by the APV* index, and second to provide a first comparison of the new index’s performance relative to that of other extant indices. The blocking episode occurred over the North Atlantic during the period 9–19 December 1975. Figure 1 shows the block’s evolution (red contour) in a sequence of daily (1200 UTC) charts including the corresponding blocking track (black–white). In addition the Z500 distribution (gray contours) gives information for the traditional blocking indices. Beneath each panel is the signature that would be recorded by the quasi one-dimensional indices of Tibaldi and Molteni (1990) (TM, blue curve) and Pelly and Hoskins (2003) (PH, black curve). Note that no time persistence criterion is applied to the latter indices. The discussion below considers in turn the block’s genesis, mature, and lysis phases.
a. Genesis (10–11 December)
In this phase the block increases its spatial scale and intensity. The negative PV anomaly characterizing the block forms over Nova Scotia and then moves slowly northeastward. During this period weak positive VAPV anomalies emerge to the southwest and to the southeast of the block. The accompanying underlying pattern in Z500 comprises a somewhat later but rapidly developing meridionally orientated trough–ridge system over and off the eastern seaboard of the United States. The APV* block itself is located on the northwestern flank of the ridge.
b. Mature (12–17 December)
The mature phase can be subdivided into two periods. During the first period (12–14 December) the slowly eastward moving block initially undergoes a decrease in intensity while it retains its spatial scale, and the accompanying positive VAPV anomaly to the southwest weakens. At the end of this period a tripolar pattern again forms as a positive VAPV anomaly reemerges to the southwest and the one to the southeast intensifies. This structural transition in the VAPV pattern heralds here, and is often a signal for, an extended existence for a block. In line with the establishment of the tripolar VAPV pattern the geopotential height field forms a distinctive Ω-shaped pattern.
During the second period (15–17 December) of the mature phase the block retrogresses westward, retains its spatial scale, and reintensifies. These changes are closely allied to the formation of the tripolar VAPV pattern. This pattern is an integral feature of an Ω-type block, and it can counter and overcome the prevailing westerly flow while retaining its spatial structure, and it also allows for the advection of low and high VAPV air toward respectively the block’s core and the southeastern positive VAPV anomaly. The positive VAPV anomalies are also evident as troughs in the Z500 field. At the end of this period there is a reversal to a dipole structure resulting from the coalescence of the two positive anomalies.
c. Lysis (18 December)
For this particular event, lysis is a comparatively rapid process and is characterized by a synchronous and strong decrease in the strength of both the block and the positive anomaly. The negative VAPV anomaly decreases to a relatively weak value. Hence, in the postblocking phase there remains a sizeable region of negative VAPV between Europe and Greenland, but its weak intensity is such that it does not match the stipulated blocking criterion.
A comparison of the APV*’s detection of the block with that of the other two indices (see the signatures beneath each panel of Fig. 1) shows good temporal agreement. The APV* index identifies the genesis slightly earlier (10 December 1975) than the other two indices. Note also that, based on Z500 and the isentrope field (not shown) the PH index at 15–16 December 1975 and TM index at 15 December 1975 register two separate block signatures on the eastern and western arms of the Ω-shaped pattern. The APV* index locates the block slightly northwestward of both the closed or quasi-closed Z500 contours and this results from the use of anomaly as opposed to absolute fields (see, e.g., Sausen et al. 1995; Doblas-Reyes et al. 2002). In addition, the application of a persistence criterion to the blocked longitudes in both the PH and TM index may result in a more stringent spatial constraint in comparison to the tracking of the entire blocking region with the APV* index.
The foregoing case study serves to illustrate the physical characteristics captured by the APV* indicator and the cartographic basis for determining various blocking parameters (e.g., genesis, lysis and block duration; spatial extent and intensity; movement). The comparison with the other blocking indicators shows their underlying compatibility while pinpointing some interesting differences.
4. Long-term APV* climatology
To establish the APV* climatology the entire domain covered by an identified block is recorded (i.e., area enclosed by the red contour in Fig. 1) and this procedure is repeated at the 6-h time interval throughout the block’s life cycle. Figure 2 shows the geographical distribution of this APV* blocking frequency in the Northern Hemisphere for the four seasons. The number of blocking episodes per season is 476 (DJF), 424 (MAM), 205 [June–August (JJA)], and 471 (SON).
Overall there are two major regions of peak frequency. They are located over the North Atlantic and the Pacific basins with maxima in the winter season. A third and weaker region of peak frequency occurs over the Asian continent, and it is most evident in the equinoctial seasons.
The two major blocking areas are located downstream of pronounced maxima in the time-mean jet stream (i.e., isentropic PV gradient or the tropopause height gradient) in each season (see Fig. F2 in Kållberg et al. 2005) and correspond in essence to the core of the Atlantic and Pacific storm tracks. To a measure the seasonal changes in frequency also mirrors changes in the upstream jet strength.
In addition to the foregoing general features there are some rich regional variations. Consider first the Atlantic. In winter (Fig. 2a) the frequency peak is located south of Greenland and Iceland with three bands of enhanced frequency extending toward Davis Strait, Nova Scotia, and Scandinavia, respectively. Note that the maximum value of 11% corresponds to about 10 blocked days per season. In spring (Fig. 2b) the peak is reduced to 8%, and has a primary extension toward Davis Strait and a secondary toward Scandinavia. A third separate maximum is evident over northern Asia. In summer there are very few Atlantic blocks (3%). Finally for autumn (Fig. 2d) the frequency values are similar to those for winter, but the enhanced bands now extend toward the British Isles and Scandinavia, rather than toward Davis Strait. The net effect is that in the autumn there is a stronger and more zonally elongated distribution than for the other seasons. Now turning to the other regions we note that over the Pacific the peak frequency is located south of the Aleutians during the three major blocking seasons, and for each season has a value of ∼9%. The Asian block region is at 90°E and is most evident in spring and autumn but with frequency values of only 5%.
a. Comparison to existing climatologies
Comparison of the climatology presented above to other extant climatologies is hampered by various factors. These include use of different meteorological fields (e.g., geopotential height at various levels), adoption of different indicators to identify blocks (e.g., threshold criteria for duration of blocks, and deployment of anomaly versus actual fields), and use of different datasets (e.g., variety of time periods and spatial resolution).
Nevertheless there is a reasonable overall agreement with earlier two-dimensional climatologies (e.g., Dole and Gordon 1983; Shukla and Mo 1983; Sausen et al. 1995; Schwierz et al. 2004). In particular the two distinct frequency maxima over the North Atlantic and Pacific are almost collocated. Minor but interesting differences include the amplitude of the third blocking maximum over the Asian continent. For DJF this peak in frequency is located at 90°E by Dole and Gordon (1983) and at ∼45°E by Sausen et al. (1995), whereas it is not particularly evident in the APV* climatology. The reverse is true for MAM and SON, and this suggests that Asian blocks might have a significant seasonal difference in their vertical structure. Another notable difference is the markedly different seasonal dependency recorded in Shukla and Mo (1983), but this is probably attributable to the latter’s use of a seasonally varying threshold value for block detection. Finally, the close agreement of our climatology with that presented in the foreshadowing study of Schwierz et al. (2004) based upon the 15-yr ECMWF Re-Analysis (ERA-15) dataset for the winter season reflects both the consistency of the ERA-40 and ERA-15 datasets, and the relative insensitivity of the climatology to mildly differing threshold values for block identification.
There is also qualitative agreement with the quasi-one-dimensional-based climatologies (e.g., Lejenäs and Økland 1983; Tibaldi and Molteni 1990). Both approaches identify the two regions of peak frequencies in the Atlantic and Pacific sector, but with differences in location and intensity. In relation to location there is a 20°–30° westward shift of the APV* maximum in the Atlantic compared to the corresponding one-dimensional results, and this is attributable in part to the approach adopted to identify blocks (see section 3). In relation to frequency we note that the relatively higher values of >20% (Tibaldi and Molteni 1990) and up to 30% (Pelly and Hoskins 2003) in the Euro–Atlantic sector compared to the present study’s APV* maxima of ∼13% were derived with different threshold values for block duration. This suggests that the temporal threshold values adopted for block identification in the one-dimensional approaches are less stringent.
b. Comparison to transient negative VAPV anomalies
It is of intrinsic interest to compare the frequency climatology presented above for blocks persisting for at least five days with the corresponding climatology for structurally similar (i.e., equivalent VAPV anomaly values) but more transient systems with a duration of up to three days (ANOM3). For the latter systems, the analog of Fig. 2 is provided in Fig. 3 for each of the three active blocking seasons (SON, DJF, and MAM).
It is evident that the pattern formed by the frequency distributions of these more transient systems has a relatively similar form. There are again peaks in the Pacific and Atlantic, but these are displaced somewhat upstream in the Atlantic and downstream in the Pacific. They are more confined zonally to the prevailing storm tracks and, in general, shifted more to the south. A notable feature is the slightly lower peak frequency during all three seasons. The Atlantic distribution further bears comparison to the tracked anticyclones displayed in Hoskins and Hodges (2002). In effect, the patterns signify that these transient negative VAPV features are reminiscent of embryonic blocks both in terms of location and intensity, but they failed to attain a sustained mature phase.
5. Mean age, genesis, and lysis
Figure 4 displays, for each of the three active blocking seasons (SON, DJF, and MAM), the mean age of blocks as a function of geographical location, and also the genesis and lysis locations. For the mean age pattern the contour spacing represents the age in days, and the display is limited to those grid points with a climatological blocking frequency exceeding 1% (cf. Fig. 2).
In general the distribution of the mean blocking age shows that “younger” and “older” mean ages correspond to extratropical and more poleward latitudes, respectively. In DJF the distribution of older mean ages is comparatively widespread with a maximum over the pole and two swathes stretching southeastward from peaks astride respectively of Bering Strait and the Mackenzie River basin. The maxima are ∼10 days. Notable domains of younger mean ages neighbor the forementioned swathes and there is a further broad band across northern Asia.
In the equinoctial seasons the peak age values are higher (∼12 days, mainly confined to areas with low blocking frequencies) but more spatially confined. In MAM there are three maxima astride the pole with a swath of high values again extending southeastward from the Mackenzie maximum to the mid-Atlantic coastline of North America, while the younger age domains are centered over the extratropical Pacific and the Atlantic. In SON there is a polar maximum and the one over northeast Asia extends southeastward to the mid-Pacific, interrupting the quasi-zonal band of the younger age domain. The relatively strong gradient in mean blocking age between Nova Scotia and the Davis Strait that is clearly visible during DJF and MAM is shifted poleward during SON.
Also in Fig. 4 the seasonal locations of blocking genesis (Figs. 4a,c,e) and lysis (Figs. 4b,d,f) are indicated as black circles. These locations are in line with the mean age distribution and consistent with the mobility of the systems. Note that the variation in the number density merely reflects the variation in the number of blocking events (cf. the previous section).
Genesis occurs predominantly within a zonal circumpolar band between 40° and 50°N latitude during all three seasons (Figs. 4a,c,e). There is a predilection for the band to be more populated in three regions, the northwest and northeast Pacific and the northwest Atlantic. This banded structure with maxima over the two major ocean basins points to the aforementioned connection between blocking genesis and the storm tracks. Note also that there are other, but less populated, preferred domains of genesis (e.g., Scandinavia and northwest Asia).
Lysis (Figs. 4b,d,f) shows two major differences in geographical distribution compared to the genesis regions. First lysis is located much more poleward and there is no evidence of a banded structure. Second there are no distinct clusters of enhanced lysis, and it is much more randomly dispersed between 50°N and the pole. Together the genesis and lysis distributions imply a poleward tendency of the tracks, and this is in line with the general latitudinal gradient in the mean age.
In this section the compiled database, with its record of the track of each block, is exploited to determine the mean track of blocks. Two refinements are incorporated to the quantitative representation of the mean block movement. First, since by prescription every recorded blocking event persists for at least 5 days, a distinction can be drawn between the mean movement during the genesis and lysis phases. Second, motivated by the clusters identified in the previous section a geographical distinction can be made between blocks. To the latter end we consider separately the following four sectors: west Atlantic (90°–30°W), east Atlantic (30°W–50°E), west Pacific (110°E–180°), and east Pacific (110°W–180°). The number of blocking events considered amounts to 136 (west Atlantic), 78 (east Atlantic), 97 (west Pacific), and 101 (east Pacific). Compilation of the track climatology involves 1) standardizing the mean track in each sector such that the blocking genesis (lysis) is located at a zero pseudo longitude and latitude and 2) examining each track during the first (last) 5 days after (before) blocking genesis (lysis).
The results are shown in Fig. 5 with the panels displaying the mean tracks for each of the three active blocking seasons. In each panel the genesis tracks (left) and lysis tracks (right) are given for each sector, and one standard deviation is indicated by the “error bars” based on a daily resolution.
In the genesis phase there is a general northeastward movement in all four sectors and this prevails during all three seasons. In DJF there is a pronounced spread of the tracks (Fig. 5a) with events originating in the Pacific having a more northward track than their Atlantic counterpart. In the Atlantic, events originating in the western sector take a much more northward orientation than those originating in the eastern sector, whereas the reverse is true in the Pacific. The general northeastward movement is in line with the seasonal mean wind distribution at tropopause levels (see Fig. F2 in Kållberg et al. 2005) and the interbasin differences can also be linked to the in situ prevailing mean wind directions. In MAM and SON (Figs. 5b,c) the mean track during the first 2–3 days does not vary significantly for all genesis regions. However, beyond the mean lifetime of 2–3 days eastern Atlantic and western Pacific tracks tend to be more zonally oriented than their counterpart in the other basin. A more zonal flow field in these two basins can also be observed in the seasonal mean wind direction.
In the lysis phase there is a marked difference in the mean movement. In particular blocks in the east Atlantic (DJF, MAM, and SON) and west Pacific (MAM and SON) track from the west-northwest in comparison to the more southerly track of the blocks in the other sectors during all three seasons. This tendency of west-northwestward movement is in line with the wind vectors at the tropopause region (cf. Fig. F2 in Kållberg et al. 2005), in particular at the eastern flank of their basin.
7. Life cycle studies
Here we present a climatology of the intensity and spatial scale of blocks during their genesis, maturity, and lysis phases. The mean value of the intensity and spatial scale of a block is evaluated at 6-h intervals from first identification until lysis, and the blocks are further subdivided into two classes: short (<10 days) and long (≥10 days) lasting events. No distinction is made between the east–west sectors of the major ocean basins, and results are shown only for the winter season (supplementary calculations revealed no significant intrabasin variations, and no marked seasonal differences).
a. Intensity evolution
Figure 6a depicts the relationship between blocking lifetime and intensity. Note that the uncertainty in estimating the intensity ranges between 0.15 and 0.30 PVU standard deviation at every phase of the life cycle.
In the genesis phase blocks evolve rapidly in intensity (i.e., toward lower VAPV anomaly values) and decrease to about 50% of the first detected value before saturating after about 2.5 days. The initial value of the intensity is similar both for long and short and for Euro–Atlantic and Pacific blocks, and they all undergo a similar temporal evolution. This indicates that the dynamics of the evolution is not sensitively dependent upon the final lifetime or location of the block. At the end of the lysis phase short-lasting events begin to decay on attaining their peak intensity and do not exhibit a sustained mature phase. In contrast long-lasting events are characterized by a prolonged mature phase with a sustained period at their peak intensity. In the lysis phase the blocks undergo a similar decay over the final 2.5 days attaining a final recorded value that is typically about 30% less than the initial value and is independent of the lifetime and block location.
b. Spatial-scale evolution
Figure 6b depicts the relationship between blocking lifetime and spatial scale. The uncertainty in scale ranges between 0.40 × 106 km2 and 1.30 × 106 km2 standard deviation at every phase of the life cycle.
In the genesis phase blocks grow rapidly in scale during the first 3 days irrespective of their type (long/short) or geographical location. Thereafter short events decrease in scale on attaining their maximum extent (∼3 × 106 km2) at around 4 days, whereas long events continue to increase their scale during the mature phase but at a significantly lower rate than earlier. In the mature phase the long-lived Pacific blocks possess a marginally larger spatial extent (up to ∼3.9 × 106 km2) than their Euro–Atlantic counterparts (up to ∼3.4 × 106 km2).
The lysis phase is characterized by a strong decrease of the spatial scale irrespective of the block type and geographical location. At the final time the mean spatial scale is only about 30% below the maximum extent. Note, however, the impact of two constraints intrinsic to the compilation of the climatology. First, the final time is determined by the stipulated threshold intensity value; hence the spatial scale can remain large beyond the prescribed lysis time and that in the decaying phase the block can become highly distorted and thereby pose problems for the tracking algorithm.
In summary, the intensity and spatial-scale evolution are basically independent of the location and final duration of the block during both the blocking onset and breakdown phases. A key difference between long- and short-lasting blocks is that the former display a distinctive and sustained mature phase.
8. Trend analysis
One feature of intrinsic interest given the compilation of a multidecadal climatology is to examine the temporal evolution and the possible trends of the derived field. This is particularly pertinent since, as was noted earlier, the spatial scale, geographical location, intensity, and duration of blocks is such that a trend in blocking could have significant impact upon weather and climate.
Here the linear blocking frequency trend is calculated for the entire time span of the ERA-40 dataset. The analysis is performed on a monthly basis to determine the linear trend at every grid point. More specifically for the calculation 3 × 44 = 132 monthly files are used (1957/58–2001/02) for each season. Thereafter the individual trends are displayed on a latitudinal–longitudinal plot to yield a two-dimensional trend distribution. Significance of the linear trend (slope) is tested against the zero trend at the 95% level. Trend calculations are restricted to the seasons DJF, MAM, and SON.
a. APV* blocking trends
Figure 7 shows the blocking frequency trend analysis of the monthly mean APV* blocking distribution for the period 1957/58–2001/02. In DJF (Fig. 7a) there is a major region with a significant negative blocking trend located southwest of Greenland in the Labrador Sea area. The peak value of ∼2.5% decade−1 equates to about 10 less winter blocking days in 2002 compared to 1958. This region has also been pinpointed in other trend studies using the Z500 (e.g., Raible et al. 2005; Schneider et al. 2003), sea level pressure (SLP) (Wu and Straus 2004), and 2-m temperature (Simmons et al. 2004) fields that exhibit reduced geopotential, pressure, and temperature signals, respectively. All three trends are consistent with the signal expected below a region with a local increase in the vertically averaged PV, and a plausible tentative inference is that together the trends point to a change in the tropopause-level flow related to a reduction in block frequency. There is a second region with a negative winter blocking trend in the North Pacific (Aleutians), but the decrease (∼1% decade−1) is less pronounced and the statistical significance weak. Local regions of positive APV* blocking trends over North America and Scandinavia border the negative trend over Greenland but their statistical significance is weak.
In MAM (Fig. 7b) the pattern is markedly different. In particular over the Atlantic region there is a shift to positive values (∼1.5%). The pattern is a more zonally elongated band of positive trend values compared with the meridionally aligned negative trend during DJF. (A caveat is that the significant areas are not spatially continuous). In the Pacific the negative blocking trend strengthens significantly, and the shape is also consistent with a more zonally elongated negative trend spanning the entire North Pacific Ocean. Additionally a region of positive trend is apparent over northern Asia. In SON (Fig. 7c) the amplitude of the trends and their spatial extent are much less pronounced, and it has been shown (Croci-Maspoli 2005) that other variables (e.g., the Z500 and SLP fields), also display relatively weak trends in this season.
The foregoing findings point to some significant regional trends in blocking. Concomitantly and most interestingly they also point to a possible link between these trends in the frequency of occurrence of a phenomenon with trends in the underlying primitive meteorological variables (i.e., Z500, SLP, and temperature fields). However, apart from discrepancies arising from different methodologies, database fields, and periods, note that these findings differ considerably from the trends recorded in some earlier blocking studies (e.g., Wiedenmann et al. 2002; Chen and Yoon 2002) that assessed the trend for the entire Northern Hemisphere and detected no significant hemisphere-wide blocking trend. However, a study of wintertime regional blocking by Luo and Wan (2005) is in line with our findings and shows significant negative trends in the Atlantic region.
b. Trend diagnosis
From a purely phenomenological standpoint the detected blocking trends can be related to either changes in the characteristic lifetime of blocks and/or changes in the number of the blocking events. In effect, in the absence of other phenomenological changes, a trend toward a longer/shorter block lifetime or a trend toward more/less blocking events would manifest itself as an increase/decrease in the APV* trend. Here we assess in turn these two contributory factors.
1) Lifetime trends
Figure 8 shows the trend derived from the lifetime climatology presented earlier for the three active seasons. It has been calculated at every grid point on a monthly basis, and the linear increase/decrease of the blocking age is shown in contours of 0.25 days decade−1.
In DJF (Fig. 8a) the negative trends in the North Atlantic and Pacific sector agree in sign with the corresponding APV* blocking trends, but they are weak and barely statistically significant. In effect the weak decrease in the age trend over southern Greenland can hardly account for the corresponding strong negative APV* blocking trend. In MAM (Fig. 8b) significant age trends occur over both the North Atlantic and northern Asia, whereas there is no signal in the Pacific sector. Again age trend in blocks over the North Pacific do not appear to account for the corresponding strong negative APV* trend. Finally, for SON (Fig. 8c) the blocking age tends to be significantly reduced in the North Pacific.
In summary blocks do not exhibit major lifetime trends in regions of significant APV* trends, and the inference is that the latter trends must be attributable to other factors.
2) Episode trends
Figure 9 provides an indication of the change in the number of blocking episodes. The distribution of this episode trend pattern is fairly consistent with that of the APV* trends (cf. Fig. 7). For example, for the North Atlantic/Greenland region during DJF (Fig. 9a), significantly less individual blocks have occurred in recent years than at the beginning of the dataset (1958). A similar result can be observed in both the Atlantic and Pacific basin during MAM (Fig. 9b). Here the number of blocking episodes has significantly increased over the Atlantic and decreased over the Pacific, respectively. Changes in the episode number are comparabily small during SON (Fig. 9c).
This “episodic” trend pattern differs considerably from that for the “blocking age” trend pattern. In effect the change in number of block occurrences might well account for a major portion of the APV* trend.
c. Tropopause height trend
From a methodological standpoint a systematic change in the height of the time-mean dynamical tropopause could itself impact upon the APV* blocking climatology. The APV* is defined as the difference between the instantaneous VAPV field and the corresponding in situ time-mean VAPV, and hence an increase/decrease of tropopause height over an extended time period at a particular location would yield more/less intense APV* blocks at that location. This in turn would favor the detection of a (possibly spurious) increase/decrease in blocking events. Examination of the change in tropopause height indicates that it has undergone a height rise over the last two decades (Santer et al. 2003, 2004), and this has in turn been linked to external effects and to stratospheric temperature changes (e.g., Highwood et al. 2000; Seidel et al. 2001). These findings and the trend in APV* prompts us to examine the trend in tropopause height.
The trend in dynamical tropopause height, defined as the 2-PVU isosurface, is depicted in Fig. 10 for all three seasons. In the usual fashion the trend has been calculated at every grid point, and is shown in contours of 2 hPa decade−1. Note that a negative pressure trend is associated with a trend toward a higher tropopause.
Inspection of Fig. 10a shows that in DJF the height trend exhibits a coherent pattern with a dipole structure over the western Atlantic and a monopole over the central Pacific. The dipole in the western Atlantic connotes a sharper slope of the extratropical time-mean tropopause in the vicinity of the jet stream, and hence can be related to other flow signatures. This dipole tropopause trend behavior is in line with the studies by, for example, Hoinka (1998) and Santer et al. (2004) (note their different tropopause definition). Indeed the main features are in line with the trend patterns of some other meteorological variables. For example, the corresponding Z500 and SLP trends during DJF (not shown) both exhibit a meridional dipole structure with negative trends in the north (southern Greenland) and positive trends to the south (cf. Hurrell 1996). A related issue is the possible linkage of the APV* trends to that of the North Atlantic Oscillation (NAO) since the latter can exhibit a life cycle of growth and decay within two weeks (Feldstein 2003) and this is comparable to that of major blocks. Studies of NAO variability show a trend toward positive index values over the last few decades (e.g., Hurrell 1995).
In the present context it is pertinent to note that the trends in the tropopause height over the Atlantic and Pacific also overlap significantly with allied structures in the APV* trend, whereas the dipole in the Mediterranean sector is not apparent in the APV* trend. Over the Atlantic and Pacific an increase/decrease in tropopause height corresponds to locations of more/less blocks (see Fig. 7). Thus the present findings are consistent with a realized (as opposed to a spurious methodological) blocking trend, and suggests that there are regions where it is the change in block frequency that impacts upon the tropopause height.
2) Linkage to Atlantic blocks
To explore this issue further we repeat the calculation leading to Fig. 10 but for the tropopause height trend that results when block days over the Euro–Atlantic sector (90°W–0°) are specifically excluded from the dataset. The results for DJF (Fig. 11a) show little indication of a dipole pattern in the North Atlantic, in comparison with the corresponding trend for the full dataset (cf. Fig. 10a). The negative trend in the south remains significant, but only a small portion of positive trend is present west of southern Greenland. Note, however, that a positive tropopause trend establishes northwest of the British Isles. The corresponding tropopause trend pattern without blocking days for MAM and SON (Figs. 11b,c) remain relatively similar to the net trends. In summary this component of the diagnosis suggests that, during DJF, Atlantic APV* blocking episodes contribute to a considerable fraction to the tropopause height trend over the North Atlantic region.
3) High-frequency contribution
Further insight on the tropopause height trend can be gleaned by examining the trend of the higher frequency analog of the blocking systems, that is, negative VAPV anomalies that satisfy the same threshold criterion but possess a lifetime of <3 days (ANOM3).
The climatology of these systems was considered in Fig. 3 and their trend distribution is shown in Fig. 12 for the three selected seasons. In DJF a significant and pronounced positive trend is located in the North Atlantic east of Nova Scotia (Fig. 12a), whereas the negative trends have a smaller spatial scale and their interpretation is more problematic. In MAM a zonal dipole-trend structure is evident in the northeastern Pacific (Fig. 12b), but again the significant areas are small in spatial extent. In SON (Fig. 12c) the trends are relatively weak and similar to those displayed in Fig. 7c.
There is a reasonable overall agreement between the ANOM3 trends and the APV* blocking trends (cf. Fig. 7) both in location and sign, but with some distinctive regional differences. The coherent areas of significance are much smaller for all seasons, and the absolute trend values lower. Intercomparison of the two trends with the tropopause height trend suggests that the ANOM3 trends are of less importance and tend merely to augment the contribution of the APV* trend. However, there are again regional differences. For example consider the relative contributions of the APV* and ANOM3 trends to the dipole pattern in the tropopause height trend between Greenland and Nova Scotia during DJF. The signal over Greenland is consistent with the APV* trend while the negative over Nova Scotia is consistent with the ANOM3 trend. In other words, these phenomena, with different intrinsic time scales, appear to contribute differentially to the tropopause height trend. The trends for the ANOM3 systems also bear comparison with their cyclonic counterparts, and there is evidence of a significant upward trend of intense cyclones in the Atlantic region (e.g., Gulev et al. 2001).
9. Overview and further remarks
This study’s overall aim was to derive a dynamically based multifaceted climatology of atmospheric blocking. To this end blocks are viewed from a PV perspective and identified as large amplitude, long-lasting, and deep negative PV anomalies located immediately underneath the dynamical tropopause (2-PVU isosurface). On this basis an objective blocking indicator (Schwierz et al. 2004) was adopted to detect the PV anomaly that takes into account its three-dimensional character. The selected indicator (APV*) yields information on the occurrence, location, amplitude, and spatial scale of individual blocks throughout their life cycle, and was applied to a Northern Hemisphere 44-yr ERA-40 reanalysis dataset.
A brief case study analysis served to demonstrate the blocking characteristics captured with the new indicator, and to compare its performance with some other extant indices. Concomitantly, it lent credence to the index’s utility and effectiveness, and diminished concern that it would capture structures that a synoptician might not necessarily classify as a typical block (Kaas and Branstator 1993; Cash and Lee 2000).
The resulting APV* climatology provides a concise, coherent, and illuminating description of the main physical characteristics of blocks, and thereby provides a springboard to elucidate the phenomenon’s dynamics. The derived results both complement and extend those of previous studies.
The new two-dimensional climatology distribution for the frequency of block occurrence captures their standard bimodal geographical distribution with two main regions of enhanced blocking over the North Atlantic and eastern North Pacific in all four seasons (but with significantly lower amplitude during JJA). Comparison of the blocking frequency distribution with that derived for shorter-lived (less than 3 days) negative PV anomaly systems (ANOM3) of comparable amplitude indicates that the latter are more confined latitudinally and only slightly displaced longitudinally. This suggests that these shorter-lived systems bear some resemblance to blocks but are more transient in character.
In addition, the present study yields a range of additional climatological information about the blocking life cycle and this is summarized below.
First, compilation of the mean blocking age as a function of location shows that young blocks (1–4 days) are more prevalent at lower latitudes whereas significantly older blocks (up to 12 days) are more prevalent at higher latitudes. Genesis is confined predominantly to the two major ocean basins and in a zonal band between 40° and 50°N latitude. The inference is that genesis tends to occur within the two major storm tracks downstream of the jet maxima of the time-mean flow. Lysis, on the other hand, is much more dispersed over the hemisphere but with clear preference to higher latitudes.
Second, a spatially differentiated analysis of the mean track of blocks for the east and west Atlantic and the east and west Pacific indicates that in the genesis phase there is a comparatively uniform mean northeastward movement within all four sectors during MAM and SON, albeit a spread is evident during DJF. There is also some indication that the eastern Pacific and western Atlantic tracks are directed systematically farther north in comparison with the more zonal orientation of the western Pacific and eastern Atlantic blocks. In the lysis phase there is also a marked difference in the mean movement with east Atlantic and west Pacific blocks tracking from the west-northwest in MAM and SON in comparison to following a more southerly track in the other sectors during all three seasons.
Third, examination of the evolution of the intensity and spatial scale during the blocking life cycle indicates that blocks undergo an intensity and scale increase during genesis and a corresponding decrease during lysis. More trenchantly these intensity and spatial-scale changes do not appear to depend upon the genesis region or the final lifetime of the blocks, and the shorter- and longer-lasting blocks differ only in that the latter have a more prolonged mature phase.
In addition, an evaluation is undertaken of blocking trends for the entire ERA-40 period. One of the key findings in regard of blocking trends is the significant negative trend detected over Greenland during DJF and over the North Pacific during MAM. Positive blocking trends are generally less significant but are still noteworthy over the entire Euro–Atlantic region during MAM.
The wintertime Greenland trend is also present in some other meteorological variables (e.g., Raible et al. 2005; Wu and Straus 2004; Schneider et al. 2003), and the latter have been linked to a variety of processes including: changes of the North Atlantic Oscillation (NAO) (e.g., Hurrell 1996), the Arctic Oscillation (Wu and Straus 2004), anthropogenic effects (e.g., Gillett et al. 2000), and warming of the tropical sea surface temperature (Hoerling et al. 2001).
Diagnosis of the present study’s APV* trends point to a linkage with changes in the number of block episodes, rather than to changes in block persistence or the background height of the tropopause. Indeed, evidence is presented indicating that blocks contribute seminally to establish the in situ tropopause height trends, and hence by default could also help account for the other changes listed above. Furthermore, intercomparison of the trends in the tropopause height, the APV* blocks, and the high-frequency (ANOM3) anomalies suggest that the wintertime negative trend over Greenland is linked to the low-frequency APV* blocks whereas the neighboring signal to the south is linked to the higher frequency ANOM3 systems.
The range and diversity of the above results provide a rich climatological base for the further examination of blocking dynamics and for assessing the performance of climate models in capturing blocks. They also serve to highlight a possible linkage between, on the one hand, blocking as a longer-lasting synoptic-scale weather phenomenon and, on the other hand, short-term climate variations and longer-term climate change.
We thank ECMWF and MeteoSwiss for providing access to the ERA-40 dataset and acknowledge that the study was partly funded through the Swiss NCCR Climate Programme.
Corresponding author address: Mischa Croci-Maspoli, ETH Zurich, Institute for Atmospheric and Climate Science, CHN M16.3, Universitaetsstrasse 16, CH-8092 Zurich, Switzerland. Email: firstname.lastname@example.org