1853Bulletin of the American Meteorological Society
1. Introduction
Increasing the horizontal resolution of the GoddardEarth Observing System (GEOS) general circulationmodel (GCM) has been a high priority in the DataAssimilation Office at the National Aeronautics andSpace Administration Goddard Space Flight Center.
Within the last two years, substantial increases in hori-zontal resolution have been implemented in GEOS.These increases have led to a very significant improve-ment in the ability of the GEOS GCM to represent thestructure and evolution of extratropical baroclinic sys-tems. As an illustration of this improvement, Fig. 1shows near-surface (975 hPa) winds over the NorthAtlantic Ocean produced by the GEOS GCM run insimulation mode for three progressively finer horizon-tal resolutions. The top panel corresponds to a gridspacing of 2° latitude × 2.5° longitude (hereafter 2° ×2.5°). At this resolution, fronts consist of broad tran-sition zones and are barely discernible. The middlepanel in Fig. 1 shows the same cyclone–frontal fea-tures at 1° × 1° resolution. At this resolution, the GEOSGCM is able to resolve a cyclone–frontal family thathas formed along a boundary characterized by sharpturning of the wind. The bottom panel shows that whenthe grid spacing is even finer [0.5° × 0.5° in an ex-
The Structure and Evolutionof Extratropical Cyclones, Fronts, JetStreams, and the Tropopause in the
GEOS General Circulation ModelA. L. Conaty,*,# J. C. Jusem,*,# L. Takacs,*,# D. Keyser,+ and R. Atlas*,#
ABSTRACT
The realism of extratropical cyclones, fronts, jet streams, and the tropopause in the Goddard Earth Observing Sys-tem (GEOS) general circulation model (GCM), implemented in assimilation and simulation modes, is evaluated fromclimatological and case-study perspectives using the GEOS-1 reanalysis climatology and applicable conceptual modelsas benchmarks for comparison. The latitude–longitude grid spacing of the datasets derived from the GEOS GCM rangesfrom 2° × 2.5° to 0.5° × 0.5°. Frontal systems in the higher-resolution datasets are characterized by horizontal potentialtemperature gradients that are narrower in scale and larger in magnitude than their lower-resolution counterparts, andvarious structural features in the Shapiro–Keyser cyclone model are replicated with reasonable fidelity at 1° × 1° resolu-tion. The remainder of the evaluation focuses on a 3-month Northern Hemisphere winter simulation of the GEOS GCMat 1° × 1° resolution. The simulation realistically reproduces various large-scale circulation features related to the NorthPacific and Atlantic jet streams when compared with the GEOS-1 reanalysis climatology, and conforms closely to aconceptualization of the zonally averaged troposphere and stratosphere proposed originally by Napier Shaw and re-vised by Hoskins. An extratropical cyclone that developed over the North Atlantic Ocean in the simulation features sur-face and tropopause evolutions corresponding to the Norwegian cyclone model and to the LC2 life cycle proposed byThorncroft et al., respectively. These evolutions are related to the position of the developing cyclone with respect toupper-level jets identified in the time-mean and instantaneous flow fields. This article concludes with the enumerationof several research opportunities that may be addressed through the use of state-of-the-art GCMs possessing sufficientresolution to represent mesoscale phenomena and processes explicitly.
1854 Vol. 82, No. 9, September 2001
perimental stretched-grid version (Fox-Rabinovitzet al. 1997)], the GEOS GCM produces features in thewind field that are reminiscent of those seen in high-resolution scatterometer wind data (e.g., Atlas et al.1999, their Fig. 9d).
The improvement in the ability of the GEOS GCMto resolve cyclone–frontal features realistically forincreasing horizontal resolution is also seen when theGEOS system is run in data-assimilation mode.Figure 2 depicts sea level pressure, along with the
wind, potential temperature, andhorizontal potential temperaturegradient at 950 hPa, for a NorthPacific cyclone–frontal systemat 1800 UTC 12 December 1991produced by the GEOS DataAssimilation System (DAS).Results for 2° × 2.5° resolutionare shown in the top panel andfor 1° × 1° resolution in the bot-tom panel. The cold front in thetop panel is depicted as a broad,diffuse zone of weak horizontalpotential temperature gradientand a gradual turning of the windacross the frontal zone. The coldfront in the bottom panel con-sists of a long, narrow zone withlarge, localized horizontal po-tential temperature gradient anda more abrupt change in winddirection across the frontal zonethan in the lower-resolution as-similation. At this particular stageof cyclone–frontal evolution, thewarm front bends around thesurface cyclone, suggestive of thebent-back warm front and fron-tal T-bone stage of the Shapiro–Keyser cyclone model (Shapiroand Keyser 1990). Figure 3shows an extratropical cycloneat 1800 UTC 26 October 1998,also over the North Pacific Oceanbut at a late stage in its evolu-tion, produced by the GEOSDAS at 1° × 1° resolution. In thiscyclone, the 900-hPa potentialtemperature pattern exhibits alocal maximum immediately tothe west of the cyclone center, a
signature of the warm-core seclusion stage of theShapiro–Keyser conceptual model.
The remainder of this article focuses on the sub-jective evaluation of the realism of extratropical cy-clones, fronts, jet streams, and the tropopause in a3-month Northern Hemisphere winter (Dec–Feb, DJF)simulation of the GEOS GCM at 1° × 1° resolution.The characteristics of the GEOS GCM and the datasetsto be analyzed are described in section 2. Since thestructure and evolution of cyclones and fronts are in-
FIG. 1. GEOS simulations for three progressively finer resolutions showing cyclones andfronts over the North Atlantic Ocean. Near-surface (975 hPa) winds are plotted at every gridpoint; vector scale (40 m s−1) is displayed beneath each panel. (top) Results for a uniform 2°× 2.5° latitude–longitude grid, (middle) results for a uniform 1° × 1° latitude–longitude grid,and (bottom) results for an experimental stretched latitude–longitude grid with maximumresolution of 0.5° × 0.5°.
1855Bulletin of the American Meteorological Society
fluenced by the background flow, it isappropriate to evaluate the extent towhich the 3-month simulation exhibitsrealistic climatological patterns. Thisevaluation is conducted in section 3. Be-cause the simulation spans 3 months,comparison with actual observationswould not prove meaningful. Instead, thesimulated cyclone–frontal features arecompared with well-established concep-tual models. This comparison is per-formed in the context of a case study insection 4, which emphasizes not onlyfronts and cyclones but also jet streamsand the tropopause. The main results aresummarized and the article is concludedin section 5.
2. Model description anddatasets
a. Model hydrodynamics and physicsThe GEOS GCM is discretized in the
horizontal using the staggered ArakawaC-grid and employs version 2 of theAries/GEOS Dynamical Core for thefinite-differencing algorithm. Version 2of the Aries/GEOS Dynamical Core is afourth-order version of the Sadourny en-ergy- and potential-enstrophy-conserv-ing scheme described by Burridge andHaseler (1977), and has been generalizedto allow for a nonuniform latitude–longitude spherical grid (Suarez andTakacs 1995). This scheme conservestotal energy and potential enstrophyfor the nondivergent component of theflow. Horizontal advection of potential temperature,moisture, and passive tracers is performed using thefourth-order scheme developed at the University ofCalifornia, Los Angeles (A. Arakawa 1996, personalcommunication). In the vertical the Aries/GEOS Dy-namical Core uses a Lorenz or unstaggered grid ingeneralized sigma coordinates following Arakawa andSuarez (1983).
The GEOS GCM physics package includes a fullset of subgrid-scale parameterizations, as well as acoupled land surface model. Penetrative and shallowcumulus convection are parameterized using the re-laxed Arakawa–Schubert scheme of Moorthi and
Suarez (1992), coupled with a Kessler-type scheme forthe evaporation of falling rain (Sud and Molod 1988).The moisture scheme also predicts liquid water con-tent and fractional cloud cover associated with shallowcumulus and stratocumulus clouds. These cloud prop-erties are allowed to interact with the radiation param-eterization of the GCM, which includes effects fromboth shortwave and longwave processes. Turbulenteddy fluxes of momentum, heat, and moisture in thesurface layer are calculated using stability-dependentbulk formulas based on Monin–Obukhov similarityfunctions. Above the surface layer, turbulent fluxes ofmomentum, heat, and moisture are calculated using the
FIG. 2. Sea level pressure (black contours, interval 4 hPa) and near-surface(950 hPa) winds (vector scale, 50 m s−1), potential temperature (red contours, in-terval 2 K), and horizontal potential temperature gradient [K (100 km)−1, shadedaccording to legend on right-hand side of panel] at 1800 UTC 12 Dec 1991 pro-duced by the GEOS DAS at two different resolutions. (top) Results for a 2° × 2.5°latitude–longitude grid and (bottom) results for a 1° × 1° latitude–longitude grid.Near-surface fields in the bottom panel are not displayed where the 950-hPa sur-face is subterranean.
1856 Vol. 82, No. 9, September 2001
level 2.5 Mellor–Yamada-type closure scheme ofHelfand and Labraga (1988), which predicts turbulentkinetic energy and determines the eddy transfer coef-ficients used for a bulk formulation. This scheme hasbeen modified recently to account for buoyancy gen-erated by moist processes through inclusion of the ef-fects of small-scale condensation and evaporation.Finally, the GEOS GCM employs the gravity wavedrag scheme of Zhou et al. (1996) to provide physi-cally based momentum dissipation in the stratosphereand mesosphere due to orographic forcing.
b. DatasetsThe GEOS GCM was run at 1° × 1° resolution with
48 sigma levels in simulation mode to produce the3-month (DJF) dataset used in this study. This data-set is interpolated from sigma to pressure coordinatesand consists of 40 levels from 1000 to 0.2 hPa.Analyses from 0600 UTC 1 December 1993 are usedto initialize the simulation. Snow cover, ground wet-ness, sea ice, and sea surface temperature are speci-fied as boundary conditions in the simulation usingclimatological values.1 To compare the performanceof the GEOS GCM at 1° × 1° resolution with the
GEOS system at other resolutions, the model was runfor a few multiday segments at 2° × 2.5° horizontalresolution on a uniform grid, and for a 2-day segmenton the 0.5° × 0.5° experimental stretched grid. Thesemultiday segments are used to produce the outputshown in Fig. 1. In addition to the 3-month simula-tion and the multiday segments, the GEOS system wasrun in data-assimilation mode at 2° × 2.5° and 1° × 1°horizontal resolution. Sample output from the DAS isshown in Figs. 2 and 3, which have been discussed inthe introduction.
In the climatological evaluation and the case studythat follow, we make extensive use of the dynamictropopause (DT). The DT is defined in terms of theErtel potential vorticity (PV), which will be expressedin PV units [PVU, defined as 10−6 m2 s−1 K kg−1 fol-lowing Hoskins et al. (1985, section 2a)]. Here we de-fine the DT as the 2.5-PVU surface in theclimatological evaluation and as the 1.5-PVU surfacein the case study. These values lie within the range of1–3.5 PVU cited by Morgan and Nielsen-Gammon(1998, section 2b) as having been used in earlier stud-ies to represent the DT.
The GEOS dataset used in the climatologicalevaluation refers to the simulated 3-month (DJF)Northern Hemisphere winter season derived from theGCM run at 1° × 1° horizontal resolution with 48sigma levels. These time-averaged fields are comparedto a 14-yr DJF climatology based on the earlierGEOS-1 reanalysis for the 1981–94 period, which isan extended version of the 5-yr reanalysis from March1985 through February 1990 performed by Schubertet al. (1993). The horizontal and vertical resolutionsof the DAS used to produce the GEOS-1 reanalysis are2° × 2.5° and 20 sigma levels, which are coarser thanthose of the current GEOS GCM. The case study cov-ers a 36-h period in the life cycle of a cyclone–frontalsystem over the North Atlantic Ocean beginning in itsdeveloping stage and ending after the cyclone hasdeepened considerably and the frontal system has be-gun wrapping around the cyclone. This cyclone–frontal system is selected because of its distinctivesurface evolution. Output from 0000 UTC model day54 through 1200 UTC model day 55 is used for thecase study.
3. Climatological evaluation
As a first step in assessing the realism of the 3-monthGEOS simulation, in Figs. 4–6, we compare global
FIG. 3. Sea level pressure (contour interval 4 hPa) and 900-hPapotential temperature (K, shaded according to legend at bottomof panel; shading intervals differ for “cool” and “warm” values)at 1800 UTC 26 Oct 1998 produced by the GEOS DAS for a 1° ×1° latitude–longitude grid.
1 Specification of snow cover and ground wetness is required be-cause the GEOS GCM used in this study is an earlier version inwhich the land surface model was not yet implemented.
1857Bulletin of the American Meteorological Society
distributions of the time-mean (DJF) sea level pres-sure, 500-hPa geopotential height, and 250-hPa windspeed with their counterparts derived from the longer-term (14 yr) GEOS-1 reanalysis dataset. Although glo-bal patterns are presented, we focus on the Northern(winter) Hemisphere, given our emphasis on extratro-pical baroclinic systems. Background material docu-menting the structure and behavior of the wintertimeclimatological fields serving as the basis for the presentcomparison can be found in standard textbooks (e.g.,Palmén and Newton 1969, sections 3.1, 3.4, and 3.5;Holton 1992, sections 6.1 and 10.5.2; Bluestein 1993,section 1.4.4).
The time-mean sea level pressure fields from thesimulation (Fig. 4, top panel) and the reanalysis(Fig. 4, bottom panel) both capture the Aleutian andIcelandic lows. The GEOS simulation exhibits lowercentral pressures than the GEOS-1 reanalysis, and the
cyclone centers in the simulation are located farthereast than in the reanalysis. The Siberian high is ap-parent in both the simulation and reanalysis, and, asis the case for the climatological lows, this high issomewhat stronger in the simulation than in the re-analysis. The time-mean 500-hPa geopotential heightfields from the simulation (Fig. 5, top panel) and thereanalysis (Fig. 5, bottom panel) exhibit prominenttroughs over eastern Asia and eastern North America,and ridges over western North America and the east-ern North Atlantic Ocean. The positions of the troughsand ridges in the simulation match well with theircounterparts in the reanalysis, and, consistent with therelative magnitudes of the sea level pressure centers,the troughs and ridges are amplified in the simulationrelative to the reanalysis. The time-mean 500-hPa geo-potential height gradients are larger in the bases of thetroughs over eastern Asia and eastern North Americain the simulation (Fig. 5, top panel) compared withthe reanalysis (Fig. 5, bottom panel). In agreementwith this comparison, the maximum wind speeds inthe Pacific and Atlantic jet streams at 250 hPa are
FIG. 5. As in Fig. 4 except for 500-hPa geopotential height (con-tour interval 120 m).
FIG. 4. Global distributions of time-mean (DJF) sea level pres-sure (contour interval 8 hPa). (top) Results from the GEOS GCMfor a 1° × 1° latitude–longitude grid; the time-mean refers to thesimulated 1993–94 Northern Hemisphere winter season. (bottom)Results from the GEOS-1 reanalysis for a 2° × 2.5° latitude–longitude grid; the time-mean refers to a 14-yr climatology for the1981–94 period.
1858 Vol. 82, No. 9, September 2001
slightly stronger in the simulation (Fig. 6, top panel)than in the reanalysis (Fig. 6, bottom panel). Theconfluence and diffluence patterns at 500 hPa (Fig. 5)coincide well with the entrance and exit regions of thejet streams at 250 hPa (Fig. 6): the Pacific jet exhibitsconfluence over eastern Asia and diffluence over theeastern North Pacific Ocean upstream of the westernNorth American ridge, whereas the Atlantic jet exhib-its confluence over North America and diffluence overthe central North Atlantic Ocean upstream of the east-ern North Atlantic ridge.
The foregoing comparison indicates reasonableagreement between the current GEOS simulation andthe GEOS-1 reanalysis. The overall greater strengthand sharper definition of circulation features in thesimulation compared with the reanalysis appear to berelated primarily to the finer horizontal and verticalresolutions of the former (1° × 1° and 48 sigma lev-els) compared with the latter (2° × 2.5° and 20 sigmalevels). Additional factors contributing to the discrep-ancies noted between the simulation and the reanaly-sis are the representativeness of a single season withrespect to a longer-term (14 yr) climatology; differ-
ences in the formulations of the respective modelingsystems used in the simulation and in the reanalysis;and the inherent distinction between a simulation anda data assimilation, whereby the former is uncon-strained by observations following the initialization ofthe model and the latter is continually influenced byobservations of the evolving atmosphere.
In anticipation of the use of the DT in the forth-coming case study to illustrate the relationship be-tween the tropopause and jet streams, we present zonalaverages of the simulated DJF time-mean potentialtemperature and wind speed2 for the Northern Hemi-sphere in Fig. 7. Also shown are the positions of theDT (2.5-PVU contour) and the “dynamic equator”(0-PVU contour). This display is motivated by theschematic diagrams shown by Hoskins (1991, hisFig. 1) and Holton et al. (1995, their Fig. 3), which sub-divide the zonally averaged troposphere and strato-sphere into the so-called underworld, middleworld,and overworld. As stated by Hoskins (1991), this sub-division extends the underworld–overworld separation
FIG. 6. As in Fig. 4 except for 250-hPa wind speed (contourinterval 10 m s−1).
FIG. 7. Meridional cross section of zonally averaged time-meanpotential temperature (black contours, interval 5 K) and windspeed (greater than or equal to 35 m s−1 shaded in red and con-toured at a 2 m s−1 interval) for the Northern Hemisphere. Alsoshown are the 298- and 380-K isentropes (purple contours), andthe 0- and 2.5-PVU surfaces (blue contours). Cross section isbased on results from the GEOS GCM for a 1° × 1° latitude–longitude grid, the ordinate and abscissa correspond to pressure(hPa) and latitude (°N), and the time-mean refers to the simulated1993–94 DJF winter season.
2 Here we use the term zonally averaged (time-mean) wind speedto refer to the speed of the zonally averaged (time-mean) horizon-tal vector wind.
1859Bulletin of the American Meteorological Society
introduced by Napier Shaw(1930, 316–318) through theaddition of the middleworld. Inthis subdivision, the overworldlies above the lowest isentropicsurface that intersects the DTwhere it becomes vertical in theTropics [380 K, the nominalvalue given by Holton et al.(1995), is highlighted in Fig. 7].This particular isentropic sur-face, which may be taken to de-fine the tropical tropopause,delineates the boundary aboveand below which the extratropi-cal stratosphere is and is not incommunication with the tropo-sphere through mass and chemi-cal constituent transport alongisentropic surfaces. The regionwhere such transport can occur(i.e., where isentropic surfacescross the DT) is referred to as themiddleworld. The middleworldlies beneath the overworld and isbounded from below by the isen-tropic surface that is tangent tothe earth’s surface in the Trop-ics (298 K in Fig. 7). In additionto being tangent to the earth’ssurface in the Tropics, thisbounding isentropic surface approaches the DT nearthe North Pole (see Fig. 7). The region beneath thisisentropic surface is referred to as the underworld, whereisentropic surfaces come into contact with the earth’ssurface.
Although not shown in the above-cited schemat-ics, Fig. 7 includes the zonally averaged wind speedmaximum, the core of which straddles the DT whereits slope attains a local maximum at the inflection lo-cated slightly poleward of 30°N. This latitudinal po-sition reflects the dominant influence of the Pacific andAtlantic jet streams during the Northern Hemispherewinter season (Fig. 6). In view of the remarkable con-sistency between the time-mean fields in Fig. 7 andthe under–middle–overworld conceptualization of thezonally averaged troposphere and stratosphere, weexamine in Fig. 8 the zonal variability of selected sur-faces shown in Fig. 7 through a three-dimensional vi-sualization produced using Vis5D graphical displaysoftware.3 Figure 8 shows that the “topography” of the
DT and of the respective isentropic surfaces bound-ing the middleworld varies relatively slowly in thezonal direction. The Pacific (background) and Atlan-tic (foreground) jet streams, indicated by the 35 m s−1
isosurface (corresponding to the outermost isotachbounding the jet core in Fig. 7), straddle the DT andstand out dramatically in Fig. 8.
4. Case study
Continuing with the evaluation of the 3-monthGEOS simulation, we focus on a 36-h period in thelife cycle of an extratropical cyclone that developedoff the east coast of North America, beginning at
FIG. 8. Three-dimensional time-mean perspective for the Northern Hemisphere of thedynamic tropopause (2.5-PVU surface) with tropopause potential temperature (K) shadedaccording to legend. Also shown are the 35 m s−1 isosurface of wind speed (red shading),and the 298- and 380-K isentropic surfaces (yellow and blue shading, respectively). Forpurposes of orientation, the view is toward the southeast with the Atlantic jet appearing inthe foreground and the Pacific jet in the background; the ordinate corresponds to geometricheight ranging between 0 and 25 km. Perspective is based on results from the GEOS GCMfor a 1° × 1° latitude–longitude grid, and the time-mean refers to the simulated 1993–94DJF winter season.
3 Documentation of Vis5D is available from the Space Scienceand Engineering Center, University of Wisconsin—Madison,through their Web site at http://www.ssec.wisc.edu/∼billh/vis.html.
1860 Vol. 82, No. 9, September 2001
0000 UTC model day 54 (00Z/54) and ending at1200 UTC model day 55 (12Z/55). The sea level pres-sure and 900-hPa potential temperature fields areshown in Fig. 9 at 12-h intervals. At 00Z/54, the cy-clone center, located over the western North AtlanticOcean, coincides with the peak of the warm sectorevident in the 900-hPa potential temperature field. Thecold front stretches from the center of the cyclone, thecentral pressure of which is 992 hPa, southwestwardtoward the North Carolina coast. At 12Z/54, the cy-clone has deepened to 984 hPa and the cold front haselongated meridionally. At this time, the warm frontis oriented zonally, its length is shorter than that ofthe cold front, and its cross-frontal potential tempera-ture contrast is weaker than that of the cold front. By00Z/55, the central pressure has decreased to 968 hPa,and the cold front is approaching the warm front suchthat the warm sector is beginning to narrow. At12Z/55, the central pressure has reached 960 hPa, the
cold front continues to approach the warm front, andthe warm sector narrows further, reminiscent of theevolution in the Norwegian cyclone model (e.g.,Schultz et al. 1998).
Figure 10 shows the sea level pressure field and thedistribution of precipitation rate calculated over a 3-hperiod ending at the output time for the same timesshown in Fig. 9. At 00Z/54, the heaviest precipitationis concentrated to the north and east of the cyclonecenter, while a band of lighter precipitation extendsalong the trough of low pressure corresponding to thecold front. At 12Z/54, the most intense precipitationis occurring just east of the cyclone center along thewarm front and southwest of the center along the coldfront. At 00Z/55, a short band of intense precipitationto the north and east of the cyclone lies along the warmfront and has begun to wrap cyclonically around thecenter. At this time, a long narrow band of precipita-tion stretches from east of the cyclone center along the
FIG. 9. Sea level pressure (contour interval 4 hPa) and 900-hPa potential temperature (K, shaded according to legend) from theGEOS simulation for a 1° × 1° latitude–longitude grid. Meridional and zonal extent of the display domain is between 15° and 70°N,and between 100°W and 10°E, respectively. (top left) Results for 0000 UTC model day 54, (top right) for 1200 UTC model day 54,(bottom left) for 0000 UTC model day 55, and (bottom right) for 1200 UTC model day 55.
1861Bulletin of the American Meteorological Society
cold front to the south and west. At 12Z/55, the shortband of precipitation continues to wrap cyclonicallyto the north and east of the cyclone, while the longnarrow band of precipitation continues to extend southand west along the cold front.
Figure 11 shows the sea level pressure, 500-hPageopotential height, and 300-hPa wind speed fields.At 00Z/54, the surface cyclone is positioned betweenthe left-exit and right-entrance regions of upper-leveljet streaks situated upstream and downstream of thecyclone, respectively, a configuration favoring focusedascent in the vicinity of the cyclone center and con-comitant deepening (Uccellini and Kocin 1987). Ashort-wave trough, evident in the 500-hPa geopoten-tial height field and accompanying the upstream jetstreak, is approaching from the west at this time. At12Z/54, the surface cyclone, which has deepened 8 hPa(to 984 hPa) during the previous 12 h, remains posi-tioned between the two upper-level jet streaks. The500-hPa short-wave trough has amplified, acquired anegative tilt, and undergone a shortening of the half
wavelength between the short-wave trough and thedownstream ridge, all signatures of continuing cyclo-genesis. At 00Z/55, the surface cyclone has deepenedan additional 16 hPa (to 968 hPa) and is located be-neath the inflection in the 500-hPa geopotential heightfield between the axes of the short-wave trough anddownstream ridge. This location corresponds to thecyclonic-shear side of the downstream jet streak, as thecyclone and the trough–ridge couplet have begun toseparate from the left-exit region of the upstream upper-level jet streak. The trough–ridge couplet continues toamplify as the ridge builds to the north-northeast ofthe surface cyclone. At 12Z/55, the surface cyclone hasdeepened by 8 hPa (to 960 hPa) during the previous12 h, but is showing signs of the cessation of deepen-ing, including the increasing separation of the cyclonecenter from both upper-level jet streaks and the verti-cally stacked orientation of the 500-hPa trough rela-tive to the surface cyclone.
Having documented the development of the simu-lated North Atlantic cyclone from a conventional, iso-
FIG. 10. As in Fig. 9 except for sea level pressure (contour interval 4 hPa) and precipitation rate (mm day−1, shaded according tolegend).
1862 Vol. 82, No. 9, September 2001
baric perspective, we reexamine this developmentfrom an isentropic PV perspective in the spirit ofMcIntyre (1999, p. 350), who predicts that in the fu-ture one might be able to use “‘virtual reality’ tech-niques . . . to fly oneself very quickly, supermanlike,around some kind of 3D representation of significantfeatures such as the tropopause and other regions ofsteep isentropic gradients of PV.” In Figs. 12–15, weuse Vis5D software to illustrate the three-dimensionalstructure of the DT, tropopause folds, upper-levelfronts, and jet streaks. The top panels of Figs. 12 and13 depict the evolution of the sea level pressure andtropopause potential temperature fields from a two-dimensional perspective, whereas the bottom panelsof these figures depict the topography of the DT abovea two-dimensional sea level pressure map. The latterdepiction is motivated by the displays of Bleck (1974,his Fig. 2) and Uccellini (1990, his Fig. 6.12). As intheir displays, the DT shown in Figs. 12 and 13 fea-tures various downward-protruding vortexlike
structures, one of which (that highlighted by the redarrows) is associated with the development of theNorth Atlantic cyclone. This particular structure willbe shown to be the three-dimensional manifestationof a tropopause fold represented in a two-dimensionalvertical cross section at 00Z/55 (Fig. 14, bottompanel). Given this correspondence, we shall refer tothis structure as a “vortical feature” and a “tropopausefold” interchangeably throughout the remainder of thissection.
At 00Z/54, the vortical feature corresponding tothe tropopause fold associated with the developmentof the North Atlantic cyclone is located upstream ofthe surface center. From the depiction of the DT to-pography, we can follow the vortical feature in timeand observe its interaction with the surface cyclone.At 12Z/54, the region of low values of tropopausepotential temperature, the southern tip of which co-incides with the tropopause fold, begins wrappingaround the surface cyclone in a cyclonic sense (Fig. 12,
FIG. 11. As in Fig. 9 except for sea level pressure (black contours, interval 4 hPa), 500-hPa geopotential height (blue contours,interval 60 m), and 300-hPa wind speed (shaded for values greater than or equal to 50 m s−1 according to legend beneath each panel).
1863Bulletin of the American Meteorological Society
top-right panel). At 12Z/54(Fig. 12, bottom-right panel)and 00Z/55 (Fig. 13, bottom-leftpanel), the fold has penetratedinto the lower troposphere. By00Z/55, a hook-shaped region oflow-tropopause potential tem-perature has wrapped cycloni-cally around the surface cyclone(Fig. 13, top-left panel), whichhas deepened 16 hPa since theonset of cyclonic wrapping 12 hearlier (Fig. 12, top-right panel).At 12Z/55, coinciding with thecessation of surface deepening,the cyclonic wrapup of the re-gion of low-tropopause poten-tial temperature has progressedto the point where the tip of thehook overlies the surface cy-clone (Fig. 13, top-right panel),and the vortical feature corre-sponding to the tropopause foldis retracting upward (i.e., “un-folding”) (Fig. 13, bottom-rightpanel). It is apparent that thedownward penetration of thetropopause fold and the associ-ated cyclonic wrapup of thetropopause potential tempera-ture minimum are playing animportant role in the deepeningof the cyclone. In this regard,Browning (1999, p. 282) com-ments that, “The effect of thehigh-PV air [associated with adescending tropopause fold],especially where it overruns alow-level baroclinic zone, is totrigger (or enhance) cyclogen-esis and ascent.”
The cyclonic wrapup evidentin the tropopause potential tem-perature field documented inFigs. 12 and 13 is suggestive ofthe LC2 life cycle proposed byThorncroft et al. (1993). TheLC2 life cycle is favored on thecyclonic-shear side of the large-scale jet stream providing theenvironmental flow for the cy-
FIG. 12. (top) Sea level pressure (contour interval 4 hPa) and tropopause potential tem-perature (K, shaded according to legend); (bottom) three-dimensional perspective from asouthern viewpoint of the dynamic tropopause (1.5 PVU) with tropopause potential tem-perature (K) shaded according to legend, and sea level pressure contoured (interval 4 hPa)on lower horizontal plane. Geographical region displayed is the same as in Figs. 9–11; theordinate in the bottom panels corresponds to geometric height ranging between 0.46 and16.67 km. Panels show results from the GEOS simulation for a 1° × 1° latitude–longitudegrid: (left) for 0000 UTC model day 54 and (right) for 1200 UTC model day 54. Red arrowsin bottom panels indicate tropopause fold discussed in text.
FIG. 13. As in Fig. 12 except for model day 55.
1864 Vol. 82, No. 9, September 2001
clone evolution.4 The cyclone in the present caseevolves in the left-exit region of both the simulatedDJF time-mean North Atlantic jet (Fig. 6, top panel)and the instantaneous upstream upper-level jet streak(Fig. 11); this region is characterized by cyclonic shearand diffluence. Although neither the simulated DJFtime-mean flow nor the instantaneous flow might be
expected to correspond closely to the environmentalflow for the cyclone evolution in the present case (be-cause the former is an average applying to an entireseason, and the latter is unaveraged and thus containsthe cyclone as well as its environment), the locationof the developing cyclone on the cyclonic-shear sideof the time-mean and instantaneous jets is consistentwith the cyclonic wrapup in the tropopause potentialtemperature field (Thorncroft et al. 1993; Shapiro et al.2001). Similarly, the location of the developing cy-clone in the diffluent exit region of these respectivejets is consistent with the evolution of the 900-hPapotential temperature field in accord with the Norwe-gian cyclone model documented in Fig. 9, includingthe meridionally extensive cold front, the zonally com-pact warm front, and the narrowing warm sector(Schultz et al. 1998).
Further documentation of the interaction betweenthe tropopause fold and the developing cyclone at00Z/55, coinciding with the end of the 12-h period ofmost rapid surface deepening, is provided in Figs. 14and 15. The top panel of Fig. 14 reproduces the sealevel pressure and tropopause potential temperaturepatterns that are shown in the top-left panel of Fig. 13,with the added depiction of the upstream upper-leveljet streak. The bottom panel of Fig. 14, consisting ofa meridionally oriented cross section along 48°W, re-veals a deep tropopause fold extending below the700-hPa level along a strong frontal zone. This fron-tal zone, which extends from the depressed tropopauseto the earth’s surface, corresponds to the cold frontshown in the bottom-left panel of Fig. 9. The bisec-tion of the jet core by the DT recalls the relationshipbetween the time-mean DT and jet stream shown inFig. 7, where the time-mean jet core straddles the DT.In contrast to the approximately zonally symmetricstructure of the time-mean DT in the latitudinal vicin-ity of the Pacific and Atlantic jet streams evident inFig. 8, the tropopause fold is highly localized in space,consistent with its prior characterization as a vorticalfeature (Fig. 13, bottom-left panel). The three-dimensional nature of the tropopause fold is illustratedfurther in Fig. 15, which is a rotated version of thebottom-left panel of Fig. 13 that includes depictionsof the jet streak associated with the tropopause foldand the meridional cross section along 48°W (seeFig. 14). From this perspective, it is evident that thejet streak straddles the DT where the latter becomesnearly vertical in the form of a “PV wall,” which con-stitutes the equatorward boundary of the stratosphericpolar vortex. Also apparent is the vortical nature of the
FIG. 14. (top) Sea level pressure (contour interval 4 hPa), tropo-pause potential temperature (K, shaded according to legend), andprojection of the 60 m s−1 isosurface of wind speed onto the hori-zontal plane (red shading); (bottom) meridional cross section along48°W of potential temperature (blue contours, interval 5 K), windspeed (greater than or equal to 60 m s−1 shaded in red and con-toured at a 2 m s−1 interval), and the 1.5-PVU surface (black con-tour). Geographical region displayed in the top panel is the sameas in Figs. 9–13; the ordinate and abscissa in the bottom panel cor-respond to pressure (hPa) and latitude (°N). The vertical line inthe top panel indicates the location of the cross section shown inthe bottom panel. Panels show results from the GEOS simulationfor a 1° × 1° latitude–longitude grid at 0000 UTC model day 55.
4 The environmental flow for an extratropical cyclone may bedefined as one in which the cyclone has been extracted from thetotal flow, which may be accomplished through PV inversion (e.g.,Hakim et al. 1996) or digital filtering (e.g., Lackmann et al. 1997;Schultz et al. 1998).
1865Bulletin of the American Meteorological Society
tropopause fold, which tilts downward along the re-gion of strong horizontal potential temperature gradi-ent corresponding to the cold front.
5. Summary and conclusions
Increasing the horizontal resolution of the GEOSGCM through a reduction of the latitude–longitudegrid spacing from 2° × 2.5° to 1° × 1° has resulted in asubstantial improvement in the representation of ex-tratropical baroclinic systems. These systems includecyclones, fronts, jet streams, and the tropopause—thestructure and evolution of which are evaluated subjec-tively from climatological and case-study perspec-tives. The former evaluation is conducted usingclimatological fields for the Northern Hemispherewinter season (Dec–Feb) derived from the GEOS-1 re-analysis for the 1981–94 period, and the under–middle–overworld conceptualization of the zonallyaveraged troposphere and stratosphere described by
Hoskins (1991), as benchmarks for comparison.Various large-scale circulation features related to thePacific and Atlantic jet streams are reproduced realis-tically in the time-mean fields simulated by the GEOSGCM for a single 3-month winter season.
A case study of an extratropical cyclone that de-veloped off the east coast of North America revealed anevolution of the low-level potential temperature fieldsuggestive of the Norwegian cyclone model, featur-ing a meridionally extensive cold front, a short warmfront, and a narrowing warm sector. Analysis of thedynamic tropopause indicates that the cyclogenesisinvolves the interaction between a deep tropopausefold and the surface cyclone; three-dimensional visu-alization of the tropopause fold using Vis5D graphi-cal display software shows this feature to be highlylocalized, exhibiting a vortical structure. The evolutionof the potential temperature field on the dynamic tropo-pause features a cyclonic wrapup, reminiscent of theLC2 life cycle proposed by Thorncroft et al. (1993).The surface cyclone intensifies in the left-exit region
FIG. 15. Three-dimensional perspective from a southwestern viewpoint of the dynamic tropopause (1.5 PVU surface) with tropo-pause potential temperature (K) shaded according to legend. Also shown are 60 m s−1 isosurface of wind speed (red shading), meridi-onal cross section along 48°W of potential temperature (purple contours, interval 5 K), and sea level pressure (contour interval 4 hPa)on lower horizontal plane. Geographical region displayed is the same as in Figs. 9–14; the ordinate corresponds to geometric heightranging between 0.46 and 16.67 km. Perspective is based on results from the GEOS GCM for a 1° × 1° latitude–longitude grid at0000 UTC model day 55.
1866 Vol. 82, No. 9, September 2001
of both the time-mean North Atlantic jet stream and anupper-level jet streak depicted in the instantaneous flowfield. The cyclonic wrapup in the tropopause potentialtemperature field may be related to the cyclonicallysheared environment of the left-exit region, whereas theevolution of the low-level potential temperature fieldin accord with the Norwegian cyclone model may berelated to the diffluent character of the jet-exit region.
This evaluation indicates that high-resolution state-of-the-art GCMs are on the verge of resolving meso-scale phenomena and processes explicitly. As such,they offer the unprecedented opportunity to investigatethe structure and dynamics of extratropical baroclinicsystems from a multiscale perspective over intrasea-sonal timescales. When run in assimilation mode,GCMs possessing mesoscale resolution will be ableto take advantage of the fine structure presently avail-able in satellite data such as scatterometer winds.Specialized observations collected during intensivefield programs also may be assimilated into GCMs,yielding unique high-resolution experimental datasetsfor analysis and diagnosis. Finally, high-resolutionGCMs are well suited to addressing issues related toclimate change. As an example, it should be possibleto simulate regional weather and climate anomalieswith increasing certainty because of the improvedresolution of individual weather systems in GCMsapplied to various climate change scenarios.
Acknowledgments. The authors thank the three anonymousreviewers for their advice and suggestions for improving the pre-sentation, and Professor Chris Thorncroft for sharing his insightson the relationship between extratropical cyclone life cycles andthe large-scale flow. Support for D. Keyser’s participation in thepreparation of this article was provided by NASA through GrantNAG5-7469, awarded to the University at Albany, State Univer-sity of New York.
References
Arakawa, A., and M. J. Suarez, 1983: Vertical differencing of theprimitive equations in sigma coordinates. Mon. Wea. Rev., 111,34–45.
Atlas, R., S. Bloom, R. N. Hoffman, E. Brin, J. Ardizzone, J. Terry,D. Bungato, and J. C. Jusem, 1999: Geophysical validation ofNSCAT winds using atmospheric data and analyses. J.Geophys. Res., 104 (C5), 11 405–11 424.
Bleck, R., 1974: Short-range prediction in isentropic coordinateswith filtered and unfiltered numerical models. Mon. Wea. Rev.,102, 813–829.
Bluestein, H. B., 1993: Synoptic-Dynamic Meteorology in Mid-latitudes, Observations and Theory of Weather Systems, Vol.II, Oxford University Press, 594 pp.
Browning, K. A., 1999: Mesoscale aspects of extratropical cy-clones: An observational perspective. The Life Cycles of Ex-tratropical Cyclones, M. A. Shapiro and S. Grønås, Eds.,Amer. Meteor. Soc., 265–283.
Burridge, D. M., and J. Haseler, 1977: A model for medium-rangeweather forecasting—Adiabatic formulation. ECMWF Tech.Rep. 4, Bracknell, Berkshire, United Kingdom, 46 pp.
Fox-Rabinovitz, M. S., G. L. Stenchikov, M. J. Suarez, and L. L.Takacs, 1997: A finite-difference GCM dynamical core witha variable-resolution stretched grid. Mon. Wea. Rev., 125,2943–2968.
Hakim, G. J., D. Keyser, and L. F. Bosart, 1996: The Ohio Valleywave-merger cyclogenesis event of 25–26 January 1978. PartII: Diagnosis using quasigeostrophic potential vorticity inver-sion. Mon. Wea. Rev., 124, 2176–2205.
Helfand, H. M., and J. C. Labraga, 1988: Design of a nonsingularlevel 2.5 second-order closure model for the prediction of at-mospheric turbulence. J. Atmos. Sci., 45, 113–132.
Holton, J. R., 1992: An Introduction to Dynamic Meteorology. 3ded. Academic Press, 511 pp.
——, P. H. Haynes, M. E. McIntyre, A. R. Douglass, R. B. Rood,and L. Pfister, 1995: Stratosphere–troposphere exchange. Rev.Geophys., 33, 403–439.
Hoskins, B. J., 1991: Towards a PV-θ view of the general circu-lation. Tellus, 43AB, 27–35.
——, M. E. McIntyre, and A. W. Robertson, 1985: On the use andsignificance of isentropic potential vorticity maps. Quart. J.Roy. Meteor. Soc., 111, 877–946.
Lackmann, G. M., D. Keyser, and L. F. Bosart, 1997: A charac-teristic life cycle of upper-tropospheric cyclogenetic precur-sors during the Experiment on Rapidly Intensifying Cyclonesover the Atlantic (ERICA). Mon. Wea. Rev., 125, 2729–2758.
McIntyre, M. E., 1999: Numerical weather prediction: A visionof the future, updated still further. The Life Cycles of Extrat-ropical Cyclones, M. A. Shapiro and S. Grønås, Eds., Amer.Meteor. Soc., 337–355.
Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa–Schubert:A parameterization of moist convection for general circulationmodels. Mon. Wea. Rev., 120, 978–1002.
Morgan, M. C., and J. W. Nielsen-Gammon, 1998: Using tropo-pause maps to diagnose midlatitude weather systems. Mon.Wea. Rev., 126, 2555–2579.
Napier Shaw, W., Sir, 1930: Manual of Meteorology, The Physi-cal Processes of Weather. Vol. III, Cambridge UniversityPress, 445 pp.
Palmén, E., and C. W. Newton, 1969: Atmospheric CirculationSystems: Their Structure and Physical Interpretation. Inter-national Geophysics Series, Vol. 13, Academic Press, 603 pp.
Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimi-lated dataset for earth science applications. Bull. Amer. Me-teor. Soc., 74, 2331–2342.
Schultz, D. M., D. Keyser, and L. F. Bosart, 1998: The effect oflarge-scale flow on low-level frontal structure and evolutionin midlatitude cyclones. Mon. Wea. Rev., 126, 1767–1791.
Shapiro, M. A., and D. Keyser, 1990: Fronts, jet streams and thetropopause. Extratropical Cyclones, The Erik Palmén Memo-rial Volume, C. W. Newton and E. O. Holopainen, Eds., Amer.Meteor. Soc., 167–191.
——, H. Wernli, N. A. Bond, and R. Langland, 2001: The influ-ence of the 1997–99 El Niño–Southern Oscillation on extrat-
1867Bulletin of the American Meteorological Society
ropical baroclinic life cycles over the eastern North Pacific.Quart. J. Roy. Meteor. Soc., 127, 331–342.
Suarez, M. J., and L. L. Takacs, 1995: Documentation of the Ar-ies/GEOS dynamical core: Version 2. NASA Tech. Memo.104606, Vol. 5, NASA Goddard Space Flight Center,Greenbelt, MD, 45 pp.
Sud, Y., and A. Molod, 1988: The roles of dry convection, cloud-radiation feedback processes and the influence of recent im-provements in the parameterization of convection in the GLAGCM. Mon. Wea. Rev., 116, 2366–2387.
Thorncroft, C. D., B. J. Hoskins, and M. E. McIntyre, 1993: Twoparadigms of baroclinic-wave life-cycle behaviour. Quart. J.Roy Meteor. Soc., 119, 17–55.
Uccellini, L. W., 1990: Processes contributing to the rapid devel-opment of extratropical cyclones. Extratropical Cyclones, TheErik Palmén Memorial Volume, C. W. Newton and E. O.Holopainen, Eds., Amer. Meteor. Soc., 81–105.
——, and P. J. Kocin, 1987: The interaction of jet streak circula-tions during heavy snow events along the east coast of theUnited States. Wea. Forecasting, 2, 289–308.
Zhou, J., Y. C. Sud, and K.-M. Lau, 1996: Impact of orographi-cally induced gravity-wave drag in the GLA GCM. Quart. J.Roy. Meteor. Soc., 122, 903–927.
