DECEMBER 1999 GARREAUD 2823

Cold Air Incursions over Subtropical and Tropical South America: A Numerical Case Study

RENEÂ D. GARREAUD Department of Atmospheric Sciences, University of Washington, Seattle, Washington

(Manuscript received 6 June 1998, in ®nal form 24 November 1998)

ABSTRACT Synoptic-scale incursions of cold, midlatitude air that penetrate deep into the Tropics are frequently observed to the east of the cordillera. These incursions are a distinctive year-round feature of the synoptic climatology of this part of South America and exhibit similar characteristics to cold surges observed in the lee of the Rocky Mountains and the Himalayan Plateau. While their large-scale structure has received some attention, details of their mesoscale structural evolution and underlying dynamics are largely unknown. This paper advances our understanding in these matters on the basis of a mesoscale numerical simulation and analysis of the available data during a typical case that occurred in May of 1993. The large-scale environment in which the cold air incursion occurred was characterized by a developing midlatitude wave in the middle and upper troposphere, with a ridge immediately to the west of the Andes and a downstream trough over eastern South America. At the surface, a migratory cold over the southern plains of the continent and a deepening cyclone centered over the southwestern Atlantic grew mainly due to upper-level vorticity advection. The surface anticyclone was also supported by midtropospheric subsidence on the poleward side of a jet entrance±con¯uent ¯ow region over subtropical South America. The northern edge of the anticyclone followed an anticyclonic path along the lee side of the Andes, reaching tropical latitudes 2± 3 days after its onset over southern . The concomitant cold air produced low-level (surface to ϳ800 hPa) cooling on the order of 10ЊC over the subtropical part of the continent (as far north as 10ЊS). Based on the observations and model results, a three-stage evolution of the cold air incursion is suggested. The initial cooling to the south of 30ЊS and far from the Andes is mainly produced by the geostrophic southerly winds between the continental anticylone and the developing low off the coast of Argentina. As the surface pressure increases over southern Argentina, a large-scale meridional pressure gradient is established between the migratory anticyclone and the continental trough farther to the north. The blocking effect of the Andes leads to an ageostrophic, low-level southerly ¯ow that advects cold air into the subtropics. Finally, as the cold air moves to the north of 18ЊS, the blocking effect of the Andes weakens (because the adjustment back to geostrophy is quite slow at these low latitudes) and the cold air spread out over the Tropics. In the last two stages of the incursion the strong pressure (temperature) gradient drives the northward accelaration of the low-level winds, while horizontal advection of cold air by southerly winds maintains the strong temperature gradient against the dissipative effects of the strong surface heat ¯uxes.

1. Introduction in¯uence on the weather over the western side of the Synoptic-scale incursions of cold air that penetrate great plains of (e.g., Tilley 1990; Blue- deep into the Tropics are frequently observed to the east stein 1993; Colle and Mass 1995) and their subsequent of major north±south-oriented mountain ranges, pro- impact on Central America and the Caribbean [see ducing dramatic weather changes over a wide range of Schultz et al. (1997, 1998) for an extensive review of latitudes. Particular emphasis has been placed on cold the literature on Central American cold surges]. Surges surges bounded by the Himalayan Plateau over south- of cold air of somewhat smaller scale have also been east Asia and their in¯uence on the Tropics during the documented along the east side of the Appalachians winter monsoon (e.g., Ramage 1971; Lau and Chang (e.g., Bell and Bosart 1988) and along the east coast of 1987; Boyle and Chen 1987; Wu and Chan 1995, 1997). (e.g., Baines 1980; McBride and McInnes Cold surges along the east side of the Rocky Mountains 1993). have also attracted considerable interest due to their In South America, episodic incursions of midlatitude air to the east of the subtropical Andes (also referred to as South American cold surges) are a distinctive year- round feature of the synoptic climatology, whose ex- Corresponding author address: Dr. Rene D. Garreaud, Departa- mento de Geo®sica, Universidad de , Casilla 2777, Santiago, istence and relevance has been recognized by synopti- Chile. cians for a long time. Statistical analyses by Kousky

᭧ 1999 American Meteorological Society 2824 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 1. Topography of South America. Elevation scale in meters above sea level. The outer and inner boxes indicate the outer and inner domain used in the numerical simulation. and Cavalcanti (1997), Compagnucci and Salles (1997), or so of moderate low-level southerly winds (ϳ10 m Garreaud and Wallace (1998), Vera and Vigliarolo sϪ1) characterize the passage of cold surges over sub- (2000), and Garreaud (1998) also demonstrate that tropical South America during wintertime. Extreme ep- South American cold surges dominate the synoptic var- isodes produce widespread freezing from central Ar- iability of the low-level circulation, air temperature, and gentina to southern Brazil, which has motivated some rainfall over much of the continent to the east of the case studies (Hamilton and Tarifa 1978; Fortune and Andes. The prevalence of this phenomenon seems at Kousky 1983; Marengo et al. 1997; Bosart et al. 1998). least partially related to the favorable continental-scale Summertime episodes produce less dramatic ¯uctua- topography: the Andes cordillera extends continuously tions in temperature and pressure, owing to the smaller from the southern tip of the continent (ϳ50ЊS) to the seasonal temperature gradient between mid- and low north of the equator with an almost straight north±south latitudes, but they are accompanied by a band of en- orientation, elevations in excess of 3000 m along most hanced convection and rainfall at the leading edge of of its extension, and a steep eastern slope (ϳ300 km the cool air (Ratisbona 1976; Parmenter 1976; Kousky wide). Along its central portion (15Њ±22ЊS) the Andes 1979; Kousky and Ferreira 1981; Garreaud and Wallace holds a high-level plateau known as the Altiplano, about 1998). These synoptic-scale bands of organized deep 300 km wide and at a mean elevation of 3800 m (ϳ620 convection account for nearly half of the summertime hPa). To the east of the Andes, the terrain is low and rainfall over the subtropical plains of the continent and ¯at, at an average elevation of 500 m, with the exception produce some of the heaviest rainfall episodes during of the Brazilian highland near the eastern tip of the summer (Garreaud and Wallace 1998). continent (see Fig. 1 for a topographic map of South The above cited case studies and statistical analyses America and Fig. 16 for a west±east cross section along provide a comprehensive picture of the mean, synoptic- 20ЊS). scale structure of South American cold surges. The up- Moderate surface pressure rises (ϳ10 hPa dayϪ1), per-level circulation is typically characterized by a de- low-level air temperature drops (ϳ5ЊC dayϪ1) and a day veloping midlatitude wave, with a ridge immediately to DECEMBER 1999 GARREAUD 2825 the west of the Andes and a downstream trough over 20ЊS) and tropical (20Њ±10ЊS) South America to the east eastern South America, which provides the large-scale of the Andes that took place in the autumn of 1993 forcing of the system predominantly in the form of vor- between 11 and 16 May. The analysis is based on con- ticity advection (Marengo et al. 1997; Garreaud and ventional synoptic data (surface and radiosonde station Wallace 1998; Garreaud 1998). At lower levels, the reports stored by NCAR), and the National Centers for equatorward incursion of cold air is initiated by the Environmental Prediction (NCEP)±NCAR gridded (2.5Њ movement of a migratory cold anticyclone from the lat ϫ 2.5Њ long) reanalyzed ®elds (Kalnay et al. 1996). southeastern Paci®c onto the southern plains of the con- Because the 6-hourly reanalyses were produced using tinent (to the east of the Andes between 40Њ and 30ЊS) a state-of-the-art weather prediction model and data as- and the deepening of a low-pressure center over the similation system, and an enhanced observational da- Atlantic off the coast of southern Argentina. Along 35ЊS tabase (including conventional, aircraft, and satellite the anticyclone moves slowly to the east, while its north- data), they are considered to compose one of the most ern leading edge becomes increasingly detached from complete and physically consistent datasets of the at- the eastward drifting upper-level ¯ow and moves north- mospheric circulation (Kalnay et al. 1996). Furthermore, ward to reach the subtropical and tropical (20Њ±10ЊS) the most important features of South American cold part of the continent within 2±3 days, collocated with surges take place over a continental area to the north a sharp transition in low-level meridional wind and a of 40ЊS, where the assimilation of conventional data low-level baroclinic zone. should strongly damp any errors that may propagate in The mesoscale structural evolution and dynamics of from the data-void southern Paci®c. We were unable to this phenomenon, however, are largely unknown. Two gather enough precipitation data, but GOES-7 infrared important issues need to be examined. First, is the north- images (from a Geostationary Operational Environ- ward advance of cold air produced by orographically trapped waves (such as Kelvin waves or shelf waves) mental Satellite) were used as a proxy for convective or, alternatively, is horizontal advection the dominant cloudiness. Speci®cally, we have used 3-hourly maps mechanism? The compositing analysis of Garreaud and of brightness temperature (Tb) on a regular 0.5Њ lat ϫ Wallace (1998) indicates a dominant role of the hori- 0.5Њ long grid obtained from the reduced radiance im- zontal advection of air temperature and anticyclonic vor- ages (the International Satellite Cloud Climatology Pro- ticity in supporting the advance of the incursion, but ject B3 product). Further details of this dataset are given their results are based on coarse data (twice daily, 2.5Њ in Rossow and Schiffer (1991) and the gridding pro- lat ϫ 2.5Њ long reanalysis), so they miss subsynoptic cedure is described in Garreaud and Wallace (1997). details. In the second place, can the large-scale circu- The traditional threshold T*b ϭ 235 K (e.g., Negri et al. lation forcing be isolated from mechanisms operating 1993), corresponding to cloud-top height above the 300- in the mesoscale, in shaping the evolution and structure hPa level, was used to identify areas of deep convection. of the cold air incursion? To improve our understanding As noted above, wintertime episodes are associated in these matters, we analyzed the synoptic observations with a pronounced decrease of low-level air temperature and performed a mesoscale numerical simulation of a but are essentially ``dry,'' in contrast to the summertime typical case that occurred in May of 1993. In this study cases. By selecting an episode during the austral au- we used the Pennsylvania State University±National tumn, we were able to explore an intermediate condition Center for Atmospheric Research (PSU±NCAR) me- that exhibits sharp signatures in the temperature, pres- soscale model (MM5) with interior ®ne-mesh domain sure, and wind ®elds, as well as a relevant cloudiness resolution of 20 km. Our primary goal is to document pattern. To place this episode in context, Fig. 2 shows the structure and evolution of this incursion on the syn- time±latitude sections along 60ЊW (a few hundreds of optic- and meso-␣-scales (200±2000 km) and to identify kilometers to the east of the Andes) of 925-hPa tem- the underlying dynamics responsible for its equatorward perature and temperature anomalies, the meridional propagation. Whenever possible, we compared the mod- wind component, and sea level pressure (SLP) for late el outputs with observations in order to evaluate the fall and early winter of 1993. The frequent occurrence ability of the MM5 to simulate cold air incursions over of cold surges is marked by the northward displacement South America. An observational overview of the ep- of the isotherms, the occurrence of cold anomalies in isode is given in section 2 of this paper. The model excess of Ϫ3ЊC, the northward advance of the southerly results, including the synoptic and mesoscale structure wind (␷ Ͼ 0), and the rise of SLP over the subtropical of the cold surge, are presented in section 3. Section 4 plains of the continent (40Њ±20ЊS). In the case analyzed includes several diagnoses of the model results. Finally, in this work, the cold air began to move northward on a summary and the formulation of a conceptual model 11 May, tightening the temperature gradient over the are presented in section 5. subtropical latitudes (as far north as 15ЊS), and accom- panied with southerly winds. At 30ЊS, SLP increases by 2. Observational overview of the 12±15 May 1993 more than 10 hPa between the onset and mature stage episode (15 May) of the episode. Comparison with incursions This section provides a synoptic overview of the in- of midlatitude air during this season (Fig. 2) and other cursion of cold, midlatitude air over subtropical (30Њ± years (not shown) indicates that the strength, duration, 2826 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 2. Latitude±time diagrams of several variables along 60ЊW during the autumn of 1993. (a) 925-hPa air temperature and temperature anomaly (deviation from seasonal mean). Contour interval 4ЊC; shading indicates cold anomalies algebraically less than Ϫ3ЊC. (b) 925-hPa meridional wind component. Contour interval 6 m sϪ1, negative values indicated by dashed lines, and the zero line is omitted. (c) Sea level pressure (SLP). Contour interval 5 hPa. Data from NCAR±NCEP re- analysis. The dashed box indicates the episode analyzed in this study. and equatorward extent of the 11±16 May episode are mediately to the west of the Paci®c coast of the con- rather typical of wintertime cold surges over this region. tinent, while the trough extends from central Argentina The large-scale circulation at middle levels during the to the southern Atlantic. The overall pattern remains incursion life cycle is shown in Fig. 3 in terms of the similar over the next 24 h, as the wave propagates east- 500-hPa geopotential height and observed winds. The ward at ϳ12 m sϪ1. The pronounced northwest±south- pattern is dominated by a midlatitude ridge-trough cou- east orientation of the ridge and trough axes produces plet with a wavelength of ϳ3000 km. By the beginning strong meridional ¯ow over southern South America, of the signi®cant subtropical low-level cooling (13 with southerly winds from the tip of the continent to May), the ridge has strengthened and is located im- about 25ЊS where they turn eastward into a con¯uent DECEMBER 1999 GARREAUD 2827

FIG. 3. Reanalyzed 500-hPa geopotential height contoured every 100 m, relative vorticity (shaded) and observed 500-hPa winds at 1200 UTC (a) 13, (b) 14, (c) 15, and (d) 16 May 1993. The light (dark) shading indicates relative vorticity greater (less) than ϩ1.5 10Ϫ5 sϪ1 (Ϫ1.5 10Ϫ5 sϪ1). Wind symbols with one pennant, full barb, and half-barb represent speed of 50, 10, and 5 m sϪ1, respectively. jet entrance region over the subtropical part of the con- over South America described elsewhere (e.g., Marengo tinent, more evident in the 300-hPa winds shown in Fig. et al. 1997) and summarized in the introduction, with 4. By 15 May the upper-level wave weakens as it moves a surface anticyclone developing and expanding north- into the southern Atlantic and 24 h later the ¯ow over ward over the continent and a cold front (particularly southern South America becomes mostly zonal. well de®ned over the western half of the continent) mov- Throughout the incursion's life cycle, the cross-barrier ing into lowlatitudes. With the exception of the Andes, ¯ow at the Andes' crest level (ϳ600 hPa) remains more the South American terrain is below 1000 m (cf. Fig. or less uniform, with a maximum around 25ЊS, just un- 1), so we are con®dent on the reduction of station pres- der the subtropical jet axis, and weaker westerlies else- sure to sea level. Notice that after 1200 UTC 13 May, where. the near-surface winds over southern and central Ar- The daily SLP and frontal analyses (based primarily gentina are close to geostrophic balance in a region of on surface reports) and surface wind observations in moderate pressure gradient, while the ¯ow near the Fig. 5 feature the typical evolution of cold air incursions northern ¯ank of the anticyclone becomes increasingly 2828 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 4. Reanalyzed 300-hPa winds and isotachs at 1200 UTC 14 May 1993. Reference vector at the bottom of the ®gure. Light and dark shading indicate wind speeds in excess of 30 and 45 msϪ1, respectively. ageostrophic and parallel to the barrier as one moves cold air has crossed this latitude on 0000 UTC 14 May, closer to the Andes. while ahead of the cold surge SLP increases only a few Further details concerning the synoptic evolution of hPa (Fig. 6). the cold surge can be inferred from the near-surface The warm side of the baroclinic zone and the wind potential temperature (␪) analyses and surface reports shift line are collocated during most of the surge life presented in Fig. 6. Because of the large difference in cycle (e.g., Figs. 6a,b). As the cold air reaches as far elevation between the subtropical plains (ϳ600 m above north as 12ЊS, however, the surface reports suggest that sea level) and the central Andes (ϳ3500 m asl), we wind transition occurs ahead of the temperature front have preferred to use ␪ instead of air temperature. In (top of Fig. 6c). The increasing separation of these two spite of the low density of stations, the analyses in Fig. low-level features has also been noted in cold surges 6 show a well-de®ned baroclinic zone [ϳ4 K (100 moving into Central America and the Caribbean km)Ϫ1] to the east of the Andes, accompanied by a sharp (Schultz et al. 1997). transition from weak northerlies (over the tropical part The potential temperature over the central Andes ex- of the continent) to moderate southerlies (over the sub- periences little change during the evolution of the sys- tropical plains). Both features are better de®ned near tem, suggesting that the cold air did not reach the South the eastern slope of the Andes, but they can be traced American Altiplano. The shallow character of the cold more than 900 km southeastward from the mountain surge relative to the Andes is con®rmed by the tem- range, presumably extending well into the Atlantic (as perature, dewpoint, and wind pro®les from two radio- inferred from the reanalysis data). Hereafter, we will use sonde stations a few hundred kilometers to the east of the narrow belt of surface wind transition to trace the the eastern slope (Fig. 7). In both cases, the largest leading edge of the incursion. The advance of the cold changes before and after the passage of the leading edge air (␪ Ͻ 15ЊC) from central Argentina into subtropical of the surge are found between the surface and the 800- latitudes is also accompanied by a marked increase in hPa level, and they tend to vanish above the 600-hPa SLP. Around 18ЊS, SLP increases by more than 10 hPa level. in 24 h, reaching values in excess of 1015 hPa after the Simultaneous GOES infrared images were superim- DECEMBER 1999 GARREAUD 2829

FIG. 5. SLP analyses, surface winds, and frontal position at 1200 UTC (a) 12, (b) 13, (c) 14, and (d) 15 May 1993. The SLP ®elds are based on synoptic reports over the continent and complemented with NCEP±NCAR reanalysis over sea. The heavy, notched line indicates the surface front position, inferred from wind and temperature surface observations. Winds (full barb and half-barb representing 5 and 10 msϪ1, respectively) are shown for selected stations. posed on the analyses in Fig. 6. They show a synoptic- edge of the surge, as the cold air advances into lower scale band of cold cloudiness that moves northward over latitudes. the continent, in association with the leading edge of the cold surge, and limited by the eastern slope of the Andes. The band extends toward the low-pressure center 3. Numerical simulation over the Southern Atlantic, and is followed by an ex- a. Model description tensive cloud-free area (more evident in the visible im- agery, not shown). The overall pattern inferred from Tb In this section we describe the synoptic and mesoscale is very similar to the equatorward propagating convec- structure and evolution of the cold air incursion during tive bands described by Garreaud and Wallace (1998) 11±16 May, obtained from a numerical simulation using for the summer season. Notice that the position of the the PSU±NCAR MM5. The MM5 is a nonhydrostatic, coldest (i.e., highest) cloudiness and the surface weather fully compressible numerical model formulated in ter- reports (not shown) suggest that the position of active rain-following vertical coordinate (␴) [see details in Du- convection moves from behind to ahead of the leading dhia (1993) and Grell et al. (1994)]. The physical pa- 2830 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 6. Near-surface potential air temperature analysis (thin lines, contoured every 4ЊC), near-surface winds (full and half-barb representing 5 and 10 m sϪ1, respectively), frontal analysis and surface

reports [␪ (ЊC), Td (ЊC), SLP (hPa)] at selected stations on (a) 0000 UTC 14 May, rametrizations used in this simulation include the simple in the lee of the Andes to the south of 30ЊS and the ice microphysics (Dudhia 1989), the Kain±Fritsch con- surface front intersected the eastern slope of the Andes vective scheme (Kain and Fritsch 1993), a cloud±ra- at about 23ЊS (Fig. 6b), so that the simulation focuses diation scheme, and the Blackadar scheme for the treat- on the mature stage of the incursion, as the cold air ment of the planetary boundary layer (PBL) (Blackadar moves from the subtropical plains into the tropical part 1979). of South America. The model domain (see Fig. 1) consists of a coarse mesh (⌬x ϭ⌬y ϭ 60 km) that covers most of South b. Synoptic-scale structure and model validation America and the adjacent Paci®c and Atlantic Oceans, and a two-way nested ®ne mesh (⌬x ϭ⌬y ϭ 20 km) Figure 8 shows the model SLP and 500-hPa geopo- over the central part of the continent. Outputs from the tential height at 2, 12, 24, and 36 h into the simulation outer and inner domain are referred to as synoptic and over the coarse domain. The synoptic-scale evolution mesoscale results, respectively. The simulation used 37 of the surface pressure ®eld simulated by the model is full-sigma levels in the vertical, with highest resolution very similar to that depicted by the observations (Fig. in the PBL; the lowest half-sigma level (␴ ϭ 0.995) 5), and the modeled-minus-observed difference in sur- corresponds roughly to 30 m above the ground. Initial face pressure was within Ϯ3 hPa throughout the sim- and lateral boundary conditions for the model integra- ulation period. Even at this coarse resolution, the ridging tion were obtained by interpolating NCEP surface- and along the eastern slope of the Andes as the cold-core upper-level analyses each 12 h, enhanced with synoptic anticyclone expands northward is very clear. The overall observations assimilated using a Cressman scheme. The middle-level circulation simulated by MM5 closely re- simulation was initialized at 1200 UTC 13 May (0900 sembles the pattern derived from the reanalysis, with LST at 60ЊW) and the model was integrated for 36 h. the correct amplitude and position of the ridge-trough At the initial time, high pressure was already established couplet. At 500 hPa, the largest differences between the DECEMBER 1999 GARREAUD 2831

FIG.6(Continued) (b) 1200 UTC 14 May, and (c) 0000 UTC 15 May. Shading indicates cloud-top

temperature (Tb Ͻ 235 K) at the corresponding synoptic time, according to gray scale at the bottom of panel (a). The 2000-m topographic contour is indicated by a thick, solid line. 2832 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 7. (a) Air temperature, dewpoint, and wind pro®les at Asuncion (25.3ЊS, 57.6ЊW) at 1200 UTC 13 and 14 May 1993. (b) Air temperature and wind pro®les at Vilhena (13.2ЊS, 60.1ЊW) at 1200 UTC 14 and 15 May 1993. Wind symbols with one pennant, full barb, and half-barb represent speeds of 50, 10, and 5 msϪ1, respectively. DECEMBER 1999 GARREAUD 2833

FIG. 8. Model SLP and 500-hPa geopotential height at 2, 12, 24, and 36 h into the simulation over the outer (coarse) domain. SLP contoured every 2.5 hPa in dashed lines. 500-hPa geopotential height contoured every 100 m in solid lines. Shaded area indicates terrain elevation in excess of 1500 m. In panel (c), Hc and Hn indicate the positions of two boxes used in diagnosing the low-level vorticity tendency. reanalyzed and modeled ®eld are on the order of 60 m slower than observed, so the largest errors in the SLP ®eld and occur after 24 h into the simulation over the sub- (ϳ3 hPa) occur by the end of the simulation over southern tropical Andes, as a result of the slower eastward move- Brazil and eastern Bolivia. ment of the simulated ridge. The simulated pattern of 400-hPa vertical motion (␻) Figure 9 shows the observed and simulated near-surface 24 h into the simulation (Fig. 11a) exhibits a broad region winds at 1800 UTC on 13 and 14 May. Although there of midlevel downward motion south of 24ЊS, and a narrow, are vast areas without surface observations, the equator- northwest±southeast-oriented band of upward motion at ward advance of the southerly wind to the east of the lower latitudes. Also included in Fig. 11a are the Q-vectors subtropical Andes is evident in the observed and simulated at the same level (e.g., Durran and Snellman 1987). The winds, with a good agreement between both ®elds. We agreement between the actual ®eld of ␻ and the divergence also performed model veri®cation on some key points (as of the Q-vectors (proportional to the quasigeostrophic ver- the one shown in Fig. 10) and found a good agreement tical velocity ␻g) is indicative of the major role of the between the observed and simulated evolution of other midlatitude, large-scale circulation in producing the ver- near-surface variables, both in the amplitude and the timing tical motion pattern over much of the continent. A rough of the changes associated with the passage of the cold estimate of ␻g at the center of the area of maximum con- Ϫ1 surge. However, the simulated surface anticyclone to the vergence yields ␻g ϳϩ0.2 Pa s , close to the maximum lee of the subtropical Andes expands northward slightly value of ␻ ϳϩ0.23 Pa sϪ1. 2834 MONTHLY WEATHER REVIEW VOLUME 127

The intensity of the subtropical jet and the position of the con¯uent jet entrance region are realistically sim- ulated by the model (not shown). As the upper-level ¯ow accelerates in the jet entrance region, transverse ageostrophic ¯ow generates a direct secondary circu- lation cell in the plane normal to the jet axis, with down- ward (upward) motion on the poleward (equatorward) ¯ank of the jet. The midtropospheric downward motion in the poleward side of the jet entrance described by Uccellini and Johnson (1979) for a straight-oriented jet streak is further enhanced by the cyclonic curvature of the subtropical jet over South America due to the along- contour component of the ageostrophic wind (Keyser and Shapiro 1986; Moore and VanKnowe 1992). In this case, the secondary circulation is mostly in the merid- ional plane, and is illustrated in Fig. 11b by a vertical

cross section of (u, ␷ a, w) along 60ЊW. Midtropospheric subsidence on the poleward side of the jet tends to in- crease the anticyclonic vorticity at lower levels, and therefore the presence of a subtropical jet entrance re- gion acts in concert with anticyclonic vorticity advection ahead of the upper-level ridge in building and main- taining the surface high pressure center over the con- tinent south of 25ЊS.

c. Mesoscale results: Horizontal structure The near-surface structure and evolution of the in- cursion is illustrated by a sequence of maps of SLP and surface (lowest sigma level) air temperature and wind over the inner domain at selected times during the in- tegration (Figs. 12 and 13). At 0000 UTC (2000 LST at 60ЊW) 14 May (12 h into simulation) the SLP ®eld reaches a maximum of 1019 hPa just to the east of the Andes in the southern border of the domain, decreasing monotonically northeastward into a ¯at trough (ϳ1007 hPa) that covers most of the tropical sector (Fig. 12a). The gentle gradient in air temperature over northern Argentina and Paraguay collapses into a narrow baro- clinic zone at about 22ЊS collocated with a sharp tran- sition of surface wind and approximately the 1005-hPa isobar (Fig. 13a). Strong southerly winds (ϳ10 m sϪ1) over the cold air region are highly ageostrophic (down- gradient) both along the eastern slope of the Andes and farther east over the subtropical plains, while weak northerly winds (ϳ3msϪ1) dominate the preincursion sector. Recall that the narrow belt of surface wind tran- sition (maximum horizontal relative vorticity) is used FIG. 9. Near-surface wind observed (thick arrows) and simulated (thin arrows) at 1800 UTC (a) 13 and (b) 14 May 1993. Simulated to trace the leading edge of the incursion. wind vectors are shown every three grid points. Only winds in excess During the next 24 h (Figs. 12b,c and 13b,c), the core of 2.5 m sϪ1 are shown. Shaded area indicates terrain elevation in of the surface anticyclone intensi®es up to 1021 hPa excess of 1500 m. The reference vector for wind is indicated at the and moves slowly toward the east over central and bottom of the panels. northern Argentina, while its leading edge reaches lat- itudes as low as 13ЊS, raising SLP by about 10 hPa dayϪ1 in a broad area over the central part of the con- tinent to the east of the Andes (central Bolivia, Para- guay, and southern Brazil). Daytime heating raises the low-level air temperature over the tropical part of the DECEMBER 1999 GARREAUD 2835

FIG. 10. (a) Six-hourly air temperature at station San Jose, Bolivia (17.5ЊS, 60.4ЊW) (solid circles) and 2-hourly model air temperature at the lowest ␴-level at 18ЊS, 60ЊW (open circles). (b) As (a) but for SLP. domain to about 30ЊC, tightening the temperature gra- tion), and the transition from southerly to weak north- dient at the leading edge of the cold air outbreak to erly ¯ow is collocated with the leading edge of the ϳ3ЊC 100 kmϪ1 (Fig. 13c). The belt of strong temper- baroclinic zone. The SLP cross section reveals two dis- ature and pressure gradients acquires a crescent shape tinctive stages in the evolution of the meridional pres- that mimics the surface observations (cf. Figs. 6b and sure gradient. During the ®rst half of the simulation, a 13c) and is consistent with dif¯uent surface winds on synoptic-scale pressure gradient extends from the de- both sides of the southerly jet over the cold air region veloping anticyclone south of 25ЊS to the tropical and about 250 km to the east of the Andes. The dif¯uent trough, with the area of high pressure rapidly expanding ¯ow occurs as the southerly winds turn to the southeast northward (the phase speed of the isobars is ϳ25 m following the Andes near 18ЊS, while southerly and sϪ1). In the second half of the simulation, the increase southwesterly winds spread out over the Amazon basin in SLP over northern and central Argentina is less pro- to the east of the jet axis. Meanwhile, the near-surface nounced, its northward propagation acquires the same ¯ow over the subtropical plains south of 25ЊS becomes phase speed as the southerly winds, and the pressure closer to geostrophic balance, turning from purely gradient becomes collocated with the baroclinic zone southerly into southeasterly. The northerly ¯ow over the and the meridional wind transition. preincursion sector becomes very weak. Further details of the near-surface evolution of the d. Mesoscale results: Vertical evolution incursion can be obtained from latitude±time cross sec- tions of the surface meridional wind component, air The vertical structure of the cold air outbreak is il- temperature, and SLP along 62ЊW shown in Fig. 14. lustrated in Fig. 15 by means of pressure±latitude cross The northward propagation of the southerly winds is sections of the meridional circulation and potential tem- very uniform during the 36-h simulation at a phase perature averaged between 63Њ and 61ЊW. At 12 h into speed of c ϳ 10msϪ1 (very close to the meridional the simulation (0000 UTC 14 May), the band of strong surface wind speed just behind the sharp wind transi- baroclinity at the leading edge of the incursion extends 2836 MONTHLY WEATHER REVIEW VOLUME 127 DECEMBER 1999 GARREAUD 2837 from the surface to 800 hPa, with nearly vertical is- the simulation (Fig. 16b) the incursion is crossing 20ЊS, entropes. Strong southerly ¯ow in the cold side is also with a rapid cooling below 800 hPa and a narrow, bar- restricted below 800 hPa, capped by northerly winds in rier-bounded southerly jet (␷ Ͼ 10 m sϪ1). At this time, the middle and upper troposphere, while a deeper layer there is clear evidence of cold air damming, with tilting of rather uniform northerly ¯ow (still strong at this time) isentropes over the slope of the Andes and a weak east- exists in the preincursion sector. Since the zonal ¯ow erly ¯ow above the southerly jet. By 1200 UTC 14 May is small, the low-level meridional wind convergence (Fig. 16c) the layer of strong southerly ¯ow and cold produces an intense meso-␣ updraft extending up to 400 air has expanded about 1000 km from the eastern slope hPa. Over the cold-air sector, the depth of the cold air of the Andes, but still exhibits a clear westward slope. and the southerly winds gradually increase rearward, Finally, Fig. 16d shows the structure of the cold air dome and low-level southerly winds in excess of 14 m sϪ1 when the leading edge of the incursion is more than extend for more than 1000 km behind the leading edge 500 km to the north of the cross section. A cold (␪ ϳ of the incursion. As the incursion progresses northward 295 K) well-mixed PBL with a rather uniform top near (Figs. 15b,c), the edge of the southerly winds and the 850 hPa has established to the east of the Andes, and baroclinic zone remain together and vertically oriented, the southerly low-level jet began to separate from the moving at a uniform speed of c ϳ 10msϪ1 throughout eastern slope of the Andes, suggestive of a decaying the lower troposphere. Coincident with the developing stage of the cold air damming and the topographic trap- of the dif¯uent surface wind pattern described previ- ping. ously, the leading edge of the incursion becomes shal- lower (southerly winds under 850 hPa), especially after e. Rainfall pattern 24 h into the simulation. The southerly low-level jet embedded in the cold air dome persists during the whole During the life cycle of the cold air incursion, the simulation, while the northerly winds over the prein- model simulates light rainfall (ϳ10 mm dayϪ1) near the cursion sector weakened, reducing the low-level con- leading edge of the incursion, in agreement with the vergence and the intensity of the meso-␣ updraft. signature of cold cloudiness in the brightness temper-

The leading edge of the incursion also exhibits a ature (Tb) maps in Fig. 6. The rainfall ®eld is shown in strong horizontal gradient of moisture, as shown in Fig. Fig. 17 by a sequence of 12-h accumulated rainfall. The 15d by the water vapor mixing ratio in the same previous relationship between rainfall, cold cloudiness, and the cross section. The humid and warm conditions prevalent leading edge of the surge is better captured in the lat- in the preincursion sector are conducive of deep con- itude±time sections along 60ЊW of the simulated 1-h vection, while the moisture ®eld within the cold out- accumulated rainfall and 3-hourly GOES Tb in Fig. 18. break is dominated by the advection of dry air from the Throughout the ®rst 24 h of simulation, the bulk of south. There is also a clear signature of an intrusion of the precipitation occurs over the cold air region (behind dry, midtropospheric air near the rear end of the incur- the wind shift line), consistent with the distribution of sion, in association with gentle subsidence that takes the cloud ϩ rainwater mixing ratio and vertical motion place south of 25ЊS. shown in Fig. 15d. In the horizontal plane, the rainfall The highly ageostrophic ¯ow near the leading edge is organized in a postfrontal band about 300 km wide, of the incursion suggests a strong blocking of the zonal extending southeastward from the eastern slope of the ¯ow and consequently the damming of the cold air by Andes (Figs. 17a,b), in qualitative agreement with the the steep slope of the Andes. This phenomenon is il- position of the coldest cloudiness in Figs. 6a,b. Follow- lustrated in Fig. 16 by means of vertical cross sections ing the horizontal vorticity arguments put forward by of horizontal wind and potential temperature taken nor- Rotunno et al. (1988), we can interpret the rearward mal to the Andes at 20ЊS. Before the passage of the ascent over the cold air dome of air parcel trajectories incursion (e.g., at 0000 UTC 14 May), northwesterly originating in the boundary layer immediately ahead of ¯ow exists in the lower and middle troposphere over the cold front, and the subsequent postfrontal rainfall. the plains, and there is clear indication of downslope Because of the negligible low-level shear in the presurge ¯ow over the eastern slope of the Andes. The prein- environment (Fig. 15), moist air approaching from low- cursion environment is characterized by a warm (␪ ϳ latitudes ¯ows toward the rear of the cold air dome 307 K), well-mixed PBL with its top near 820 hPa over under the in¯uence of the negative vortex at the edge the lowlands to the east of the Andes. After 18 h into of the cold air, so that the updraft leaned back over the

FIG. 11. (a) Model vertical velocity (contours) and Q-vectors (arrows) at the 400-hPa level on 1200 UTC 14 May 1993 (24 h into the simulation). Contour interval is 0.1 hPa sϪ1, the zero line is omitted. Shaded indicates downward motion. The thick, dashed arrow indicates the upper-level jet axis. (b) Model zonal wind (u, shaded) and ageostrophic meridional circulation (␷ a, w) along 60ЊW (results from the outer domain) on 1200 UTC 14 May 1993 (24 h into the simulation). Light to dark shading indicates zonal wind from 30 to 60 m sϪ1 and

J indicates the position of the subtropical jet. Reference vector for (␷ a, w) at the bottom of the ®gure. 2838 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 12. Model SLP and wind vectors at the lowest ␴-level (␴ ϭ 0.995) at ϩ12, ϩ18, ϩ24, and ϩ30 h into the simulation. Corresponding dates are indicated at the top of each panel. Results from inner domain. SLP contoured every 2 hPa. Reference vector at the bottom of the ®gures. Shading indicates terrain elevation in excess of 2000 m. cold dome and the rainfall fell over the cold sector [see this hypothesis, we show in Fig. 19 the model temper- Fig. 18b in Rotunno et al. (1988)]. ature and dewpoint pro®les at 18ЊS, 61ЊW before and The postfrontal rainfall is mostly grid solved, sug- at the time of the frontal passage (Figs. 19a and 19b, gestive of a relative minor role of the deep convection respectively). The small convective available potential during the ®rst half of the simulation. To further explore energy (CAPE ϳ 700 J kgϪ1) and elevated level of free DECEMBER 1999 GARREAUD 2839

FIG. 13. Model potential temperature and wind vectors at the lowest ␴-level (␴ ϭ 0.995) at ϩ12, ϩ18, ϩ24, and ϩ30 h into the simulation. Corresponding dates are indicated at the top of each panel. Results from inner domain. Potential temperature contoured every 3ЊC. Reference vector at the bottom of the ®gures. Shading indicates terrain elevation in excess of 2000 m. convection (LFC ϳ 3 km above the ground) on the (Fig. 19b, midnight), the troposphere became condi- presurge environment (Fig. 19a) preclude the outbreak tionally neutral and saturated in the lowest 200 hPa. of intense, deep convection at and to the south of this Figure 19 also includes the temperature and dewpoint latitude (the local time in Fig. 19a is 1900). Once the pro®les at the same previous point about 18 h after the cold air reached 18ЊS and precipitation was occurring leading edge of the surge has crossed 18ЊS (Fig. 19c), 2840 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 14. Model (inner domain) latitude±time sections of the surface (␴ ϭ 0.995) meridional wind component (every 2.5 m sϪ1), air temperature (every 2ЊC), and SLP (every 2 hPa) taken along 62ЊW. The dashed lines trace the propagation of the meridional wind transition (leading edge of the incursion), with a uniform phase speed of c ϳ 10msϪ1. showing the cold and dry air far behind the leading edge 4. Model diagnosis of the surge. As the integration progresses into the second daytime We begin the diagnosis of the cold air incursion by cycle and the incursion moves into an increasingly un- stable environment, cores of convective rainfall become quantifying the relative importance of the large-scale organized in a compact, crescent-shaped band at and processes in the development of the surface cold anti- ahead of the leading edge of the cold air (maximum cyclone over the continent, based on a simpli®ed form low-level convergence) (Figs. 17c and 18a), somewhat of the Zwack±Okossi (Z±O) equation (Zwack and Okos- si 1986). The Z±O equation diagnoses all the processes resembling of the Tb map at 0000 UTC 15 May on Fig. 6c. Larger CAPE (ϳ1900 J kgϪ1) and lower LFC (ϳ900 (vertically integrated) that contribute to the geostrophic m above the ground) in the presurge, tropical environ- vorticity tendency near the surface. Neglecting the ment (see model sounding at 12ЊS, Fig. 19d) are also smaller contributions from the tilting terms and vertical consistent with the development of intense, deep con- advection, Lupo et al. (1992) proposed the extended vection at the leading edge of the surge. form of the Z±O equation:

pppooo po ␨go RQdpÇץ 2 (ϫ Fr) dp, (1 ١ ´ ١T) ϩϩS ␻ dp ϩ pÄ (k ´١␨ ) dp Ϫ pÄ ٌϪ(Vhp ´ഠ pÄ Ϫ (Vha ͵ tfcp͵ ͵͵ pץ ppptttΌ΍[] pt (A) (B) (C) (D) (E)

where ␨go is the geostrophic relative vorticity at a near- vorticity advection; (B) temperature advection, hori- surface level (925 hPa in this application), Vn is the zontally nonuniform; (C) diabatic heating; (D) adiabatic horizontal wind vector, ␨a is the absolute vorticity, Q is warming; and (E) friction. the diabatic heating rate, Sp is the static stability, ␻ is Using the model output ®elds from the coarse domain the vertical motion in isobaric coordinates, and Fr is the interpolated every 25 hPa, we calculated the spatial av- frictional force. The limit po is the near-surface pressure, erage of the observed 925-hPa geostrophic vorticity ten- and pt is an upper level chosen so that the integration dency, and terms A, B, and D on the rhs of Eq. (1), includes most of the atmospheric mass (100 hPa in this over two 1Њ lat ϫ 1Њ long boxes indicated in Fig. 8c by Ϫ1 application) and p ϭ (po Ϫ pt) . Thus, the extended the labels Hc and Hn at 1200 UTC 14 May (24 h into Z±O equation states that changes of near-surface geo- the simulation). Terms (C) and (E) were not evaluated strophic vorticity are forced by vertically integrated (A) in this calculation, but the effect of the diabatic heating DECEMBER 1999 GARREAUD 2841

FIG. 15. Model (inner domain) meridional circulation [(␷,w), vectors], potential temperature (contoured every 2.5 K), and meridional wind speed (shaded) along 60ЊW at: (a) 12, (b) 24, and (c) 30 h into the simulation. Reference vectors at the bottom of the ®gures. Light and dark shading indicate meridional wind speed in excess of 8 and 14 m sϪ1, respectively. (d) Model meridional circulation [(␷,w), vectors], water vapor mixing ratio (contoured every2gkgϪ1) along 60ЊW at 18 h into the simulation. Light and dark shading indicate cloudϩrainwater mixing ratio in excess of 0.25 and 0.75 g kgϪ1. and friction will be explicitly considered in the ther- processes can be learned from Fig. 20, which shows modynamic and momentum balances, respectively. Here pro®les of the integrand terms in Eq. (1). Figure 20 also Hc is located over central Argentina near the core of shows the vertical pro®le of vorticity and vorticity ten- the surface anticyclone, while Hn, in the lee of the An- dencies over Hc and Hn. The differences in relative des at 20ЊS, is close to the leading edge of the anticy- vorticity above the 500-hPa level over the two sectors clone at this time. Numerical values of the vertically of the surface anticyclone (Fig. 20a) arise mainly from integrated contributions are presented in Table 1, while the position of Hc and Hn on the poleward and equa- some details about the vertical distribution of the leading torward sides of the subtropical jet, respectively. Also 2842 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 16. Model (inner domain) cross section at 20ЊS of potential temperature (contoured every 2.5 K), horizontal wind vectors (reference vector at the bottom of the ®gure), and meridional wind speed (shaded, reference at the bottom of the ®gure) taken at 12, 18, 24, and 32 h into the simulation. note that while anticyclonic vorticity is increasing 300 hPa (Fig. 20c). Similarly, the deepening of the sur- throughout the entire tropospheric column over the an- face cyclone off the coast of southern Argentina during ticyclone core (Hc), positive vorticity tendencies over the early stages of a cold surge is dominated by upper- its leading edge (Hn) are largely restricted to levels level cyclonic vorticity advection, according to the qua- below 700 hPa (Fig. 20b). sigeostrophic analysis of an intense surge in June 1994 The development of the central part of the anticyclone documented in Marengo et al. (1997) and the compos- appears to be dominated by anticyclonic vorticity ad- iting analysis in Garreaud (1998). In contrast, strong vection (AVA), which contributes nearly two thirds of cold advection, mainly below 600 hPa (Fig. 20d), is the the total vorticity change at Hc and is maximum near leading positive term in Eq. (1) at the northern edge of DECEMBER 1999 GARREAUD 2843

FIG. 17. Model 12-h accumulated rainfall over the inner domain during the period ending at 0000 UTC 14 May (12 h into the simulation), 1200 UTC May 14 (24 h into the simulation), and 0000 UTC 15 May (36 h into the simulation). Reference scale in mm. The 1500 topographic contour is indicated by a thin solid line. The position of the leading edge of the incursion is indicate by a thick dashed line and the numbers indicate the time step (in h) into the simulation.

anticyclone (Hn), explaining ϳ75% of the increase of where ⌬ph is the surface pressure change corresponding low-level vorticity. to a temperature change of ⌬T throughout a column of The effect of the incursion of cold air in raising the depth H(x, y, t) and mean temperature T(x, y, t), R is surface pressure over the continent can be estimated the ideal gas constant, and p 0 ϭ 1010 hPa is a reference directly using the hydrostatic equation integrated over surface pressure. Here H is de®ned as the height of the the depth of the cold air: cold air dome (see Fig. 15): H ϭ z(␪ ϭ 300 K) if H Ͻ 2000 m or H ϭ 2000 m otherwise (notice that H ϭ 0 pgpT ⌬ h 0 ⌬ ahead of the cold outbreak), and T is the vertically av- ϭϪ 2 H, (2) ⌬t RT ⌬t eraged temperature within the cold air: 2844 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 18. (a) Model (inner domain) latitude±time section of 1-h accumulated rainfall taken along 62ЊW. The dashed lines trace the propagation of the meridional wind transition (the leading edge of the incursion), with a uniform phase speed of c ϳ 10msϪ1. (b) Latitude±time section of the 3-hourly brightness temperature (infrared channel, GOES-7) along 62ЊW. The solid circles indicate the meridional wind transition estimated from the surface 6-hourly observations.

1 H ing southeastward from the eastern slope of the An- T ϭ Tdz. (3) H ͵ des, as evidenced by the local rate of change of 925- 0 hPa air temperature, 24 h into the simulation (Fig. Using the model results from the inner domain at 30- 22). To determine the origin of these tendencies we min intervals, the hydrostatic surface pressure was cal- diagnose the horizontal temperature advection culated at several stages of the simulation. Figure 21 ١T ) and adiabatic cooling (␻S p ) terms in the ´ ϪV) shows the resulting ®eld (⌬ph) together with the total thermodynamic energy equation at the 925-hPa level local rate of change of surface pressure ( p )at24h ⌬ t using centered differences in time (⌬t ϭ 30 min) and into the simulation. At this time, surface pressure is space (⌬x ϭ⌬y ϭ 20 km). The resulting ®elds, rising over the whole region occupied by the cold air smoothed using a distance-weighted nine-point ®lter, mass. The largest positive tendencies, exceeding 1 hPa are presented in Fig. 22 at 24 h into the simulation. (2 h)Ϫ1, are observed over a belt just behind the leading Within the cold outbreak region, horizontal advection edge of the incursion, and are largely explained (qual- itatively and quantitatively) by the low-level, hydro- of cold air dominates the local cooling; adiabatic cool- statically induced pressure change. Farther south (ϳ400 ing is dominant only within a very narrow belt just km behind the leading edge), the surface pressure is ahead of the cold front in association with the intense increasing over an area of negligible or even negative meso-␣ updraft. Surface heat ¯uxes are very small at t) andץ/Tץ) this time (0800 LST), and the observed ⌬ph. Results at other times during the simulation are ١T ϩ ␻S p ) cooling rates agree ´ similar, and overall they support the hypothesis derived calculated (ϪV from the Z±O diagnosis. The central part of the surface closely. Results at other times reveal that horizontal anticyclone is generated and maintained primarily by advection of cold air is the dominant term in the low- middle- and upper-troposphere processes, with little (if level thermodynamic energy balance throughout the any) contribution from the lower-troposphere temper- model simulation. Over 700 hPa the release of latent ature change. In contrast, the hydrostatic pressure heating ahead of the cold air dome is larger and com- change produced by the advance of the shallow (lower- pensates for the adiabatic cooling in the prefrontal tropospheric) cold air dome into a warmer environment environment. accounts for most of the surface pressure rise at the As discussed in section 4e, most of the light, night- forward ¯ank of the anticyclone. time rainfall during the ®rst half of the simulation fell The intense local cooling at the leading edge of the within a cold, conditionally neutral environment (see incursion takes place over an elongated area extend- model sounding in Fig. 19b), in which evaporative cool- DECEMBER 1999 GARREAUD 2845

FIG. 19. Model air temperature, dewpoint, and wind pro®les at 18ЊS, 61ЊW at (a) 12, (b) 18, and (c) 36 h into the simulation, and at 12ЊS, 63ЊW at (d) 30 h into the simulation. Wind symbols with one pennant, full barb, and half-barb represent speeds of 25, 5, and 2.5 m sϪ1, respectively. The corresponding date is indicated at the top of each panel. In (a), (b), and (d) the thick curved line indicates the moist adiabatic that will follow an air parcel lifted from the surface.

ing and downdrafts are relatively unimportant in in- offset these terms. The sounding in Fig. 19d also sug- ducing temperature perturbations. However, as the sim- gests a downdraft-induced low-level cold pool anomaly ulation progresses into the second diurnal cycle (after of about Ϫ7ЊC. Nevertheless, the role of the rain-in- 24 h into the simulation), it is possible that evaporative duced diabatic cooling in wintertime cold surges seems cooling becomes more important in the maintenance of secondary, as some of the strongest cold surges (in terms the cold air pool. At 32 h into the simulation (1400 LT) of temperature changes) are not accompanied by sizable the residual term in the low-level energy balance is still rainfall. On the other hand, the evaporative cooling very small in spite of the strong latent and sensible heat could be more signi®cant in warm season incursions ¯uxes, suggesting that evaporative cooling may partially (when there is intense rainfall at and near the leading 2846 MONTHLY WEATHER REVIEW VOLUME 127

TABLE 1. The 925-hPa geostrophic relative vorticity tendency and the Coriolis force in the momentum balance. The in- vertically integrated contributions from the leading terms in the ex- crease of F in this region is not an artifact of the method tended Zwack±Okossi equation over boxes Hc and Hn at 1200 UTC r 14 May 1993 (12 h into the simulation). See text for de®nition of of calculation but is consistent with the maximum sur- the terms A, B, and D. Calculations based on model results from the face wind stress within the baroclinic zone. Finally, over outer domain. Units are 10Ϫ9 sϪ2,(ϩ) anticyclonic tendency, (Ϫ) a narrow belt across the leading edge of the incursion, cyclonic tendency. the frictional force is insuf®cient to balance the large 925-hPa pressure gradient force, resulting in a net northward geostrophic A: Absolute acceleration of the surface winds. vorticity vorticity B: Temperature D: Adiabatic In the later section (Fig. 23b) in which the cold air Box tendency advection advection cooling has advanced beyond 18ЊS, the zonal component of the Hc ϩ0.92 ϩ0.69 ϩ0.13 Ϫ0.03 pressure gradient near the Andes is much weaker and Hn ϩ0.95 ϩ0.13 ϩ0.83 ϩ0.08 the Coriolis force induced by the southerly winds is out of balance, resulting in a strong eastward acceleration and the development of a dif¯uent ¯ow pattern to the edge of the cool air) in a similar fashion as described west of the southerly jet. The downgradient ¯ow at and for Appalachians cold air damming (e.g., Bell and Bos- near the leading edge of the cold air is still maintained art 1987; Fritsch et al. 1992). by the balance between the pressure gradient force and Most of the horizontal advection of cold air is ac- friction, both near the Andes and far away from the complished by low-level southerly winds, which exhibit barrier, as cold air spreads out over the tropical part of a distinctive cross-isobaric, mountain parallel character the continent. over the plains to the east of the Andes, which is evident A key element in our analysis is the damming of the both in observations and the model results. To illuminate cold air by the Andes in a similar fashion to that ob- the dynamical forcing of this largely ageostrophic ¯ow served along the lee side of the Rockies (e.g., Colle and and its feedback upon the temperature and pressure Mass 1995) and the Appalachians (e.g., Bell and Bosart ®elds, we diagnosed the various terms in the horizontal 1988) during cold outbreaks, which is studied analyti- momentum equation at 950 hPa over the inner domain: cally by Xu (1990). The evidence of this phenomenon in the vertical evolution of the incursion (Fig. 16) is Vץ Vץ ,١V ϩ ␻ complemented by the trace of the surface wind at 20ЊS ´ hhϩ V -p 61ЊW (Fig. 24), which exhibits a characteristic simulץ t hhץ taneous acceleration of the wind parallel (␷) and normal (ϭϪf kˆ ϫ V Ϫ g١z ϩ F , (4 hr (u) to the barrier, followed by a steady-state, geostrophic where Vh ϭ ui ϩ ␷j is the horizontal wind vector and southerly ¯ow (␷ ϳ ␷ g) and a rapid weakening of the z is the 950-hPa geopotential height. The accelerations zonal component. Scale analysis provides further sup- produced by vertical motion are at least one order of port of the blocking effect of the subtropical Andes (with magnitude smaller than the other terms and therefore a mean height hm ϳ 3300 m), since for the observed Ϫ1 neglected, the frictional term (Fr) is calculated as a re- range of upstream ¯ow (U ϳ 10ms ) and static sta- sidual in Eq. (4), and, as in our previous diagnosis, the bility within the cold air (N ϳ 1.8 10Ϫ2 sϪ1) the ratio derivatives are evaluated using ®nite differences cen- of kinetic energy of the approaching ¯ow to the potential tered in space and time. Results at 18 and 30 h into the energy required to pass over the mountain range (the simulation are displayed in Fig. 23. Froude number Fr ϭ U/hmN) is much less than unity The momentum balance is accomplished in different (Fr ϳ 0.2). ways over different parts of the domain. Let us ®rst Numerical simulations by Pierrehumbert and Wyman consider the balance within the cold air south of 20ЊS (1985) suggest a horizontal scale of the upstream effect (Fig. 23a). Within the cold air region and far enough of the mountain range: for steep topography [cross-bar- (ϳ500 km) behind the leading edge of the surge, the rier length scale (l)/along-barrier length scale (L) K 1] surface winds are more or less uniform. The Lagrangian a perturbed zone (primarily mountain-parallel winds) acceleration is small compared with the other terms in extends to about one Rossby radius of deformation lR the momentum balance. The low-level ¯ow is close to ϭ Nhm/ f from the barrier. In the cold air region between geostrophic balance between the zonal Coriolis force 20Њ±25ЊS, l ϳ 300 km, f ϳ 5.5 10Ϫ5 sϪ1, N ϳ 1.8 10Ϫ2 induced by the strong southerly winds and the mountain- sϪ1, and V ϳ 12msϪ1. Using the average height of the normal pressure gradient force. The frictional term, al- Andes (ϳ3300 m) for hm, the Rossby radius is lR ϳ though small, de¯ects the wind toward the lower pres- 1000 km. However, the modeled dome of cold air is sure and balances a small meridional component of the restricted to the layer below 800 hPa, and therefore it

Coriolis and pressure gradient forces. Farther north, but seems more appropriate to use hm ϳ 2000 m, which still within the cold air mass, the along-barrier pressure implies lR ϳ 600 km. However, the observed and sim- gradient is close to being in balance with a southerly ulated length scale of the barrier-parallel ¯ow and the directed frictional force, since the increasingly blocked cold air damming in this range of latitudes is on the mountain-normal (zonal) ¯ow precludes a major role of order of 1000 km (e.g., Fig. 16c). Nevertheless, given DECEMBER 1999 GARREAUD 2847

FIG. 20. (a) Model (outer domain) vertical pro®les of the relative vorticity over the boxes Hc and (solid line) Hn (dashed line) at 1200 UTC 14 May (24 h into the simulation). See Fig. 8 and text for position of Hn and Hc. Units are 10Ϫ5 sϪ1. (b) As (a) but for the relative vorticity tendency. Units are 10Ϫ9 sϪ2. (c) As (a) but for the horizontal advection of absolute vorticity. Units are 10Ϫ9 sϪ2. (d) As (a) but for the horizontal advection of temperature. Units are 10Ϫ4 KsϪ1. the strong strati®cation of the cold air and the marked be blocked during their ascent. This ``second-order tilting of the isentropes close to the terrain (i.e., within blocking'' against the preexisting cold air may be less a Rossby radius from the steep slope), eastward moving effective than the topographic damming, but it might air parcels need to move upward over the cold air to be responsible for widening the topographically per- conserve their potential temperature, and will eventually turbed zone observed over the subtropical plains. 2848 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 21. Model (inner domain) (a) low-level hydrostatic and (b) total surface pressure change at 1200 UTC 14 May (24 h into the simulation). Contour interval is 0.5 hPa (2 h)Ϫ1, negative values indicated by dashed lines, and the zero line is omitted. The leading edge of the incursion is indicated by a thick, dashed line. Shaded area indicates terrain elevation in excess of 1500 m.

As the cold air moves into the Tropics, downgradient erally applicable to midlatitude air incursions over the ¯ow over the crescent-shaped baroclinic zone expands South American region. eastward more than 1500 km from the slope of the An- The case described in this paper follows a synoptic- des, more than twice the Rossby radius of deformation scale evolution typical of cold air incursions over this at 15ЊS. As demonstrated in the momentum balance, the region, and similar to cold surges trapped along the lee change in direction of the Andes cordillera allows the of major north±south-oriented mountain ranges de- development of a westward component to the ¯ow, scribed elsewhere (e.g., Colle and Mass 1995). The ini- which, in conjunction with the reduction of the Coriolis tial development of the surface anticyclone over south- force, reduces the potential for cold air damming. In ern Argentina was largely controlled by the large-scale fact, a vertical cross section taken parallel to the leading forcing associated with the middle and upper-level cir- edge of the incursion at 33 h into the simulation (Fig. culation within a baroclinic wave, in agreement with 25) reveals a nearly uniform depth of the cold air (hor- previous ®ndings by Marengo et al. (1997) and Garreaud izontal isentropes) except very close to the eastern slope and Wallace (1998). In addition to the upper-level AVA, of the Andes. Thus, the ageostrophic ¯ow and the spread midtropospheric subsidence on the poleward side of an of cold air over the tropical part of the continent appears upper-level jet over subtropical South America seems to be mainly the result of the slow adjustment between important in the maintenance of the central part of the the mass and wind ®elds (as f → 0), rather than an anticyclone to the east of the Andes. A subtropical jet upstream effect of the topography. entrance region is also evident to varying degrees in the South American cold surge episodes reported by Ma- rengo et al. (1997) and Hamilton and Tarifa (1978), and 5. Conclusions the compositing analysis in Garreaud (1998). Similarly, Reanalysis ®elds, observations, and a numerical upper-level jet entrance regions over subtropical lati- weather prediction model have been used to describe tudes to the east of the relevant mountain range appear synoptic- and mesoscale features of a cold, midlatitude as important ingredients in the occurrence of strong, air incursion over subtropical and tropical South Amer- long-lived cold surges over north Central America (e.g., ica to the east of the Andes. The episode, which occurred Schultz et al. 1997) and over Southeast Asia (e.g., Lau during the austral autumn of 1993, exhibited a strength, and Chang 1987). duration, and equatorward extent typical of wintertime While at 30Њ±35ЊS the surface anticyclone moves incursions, as well as a cloudiness pattern typical of slowly to the east, its northern leading edge follows an warm season incursions. Thus, even though we have anticyclonic path along the lee side of the Andes, reach- documented the results of just one case, we are con®dent ing tropical latitudes 2±3 days later. The ridging along that some elements of its structural evolution and, more the leading edge of the incursion is largely produced by importantly, its underlying dynamics, should be gen- the hydrostatic effect of the shallow cold air dome re- DECEMBER 1999 GARREAUD 2849

FIG. 22. (a) Model (inner domain) 925-hPa local rate of change of temperature at 1200 UTC 14 May (24 h into the simulation). Contour interval is 0.5ЊChϪ1, negative values in dashed lines, and the zero line is omitted. (b) As (a) but for the horizontal temperature advection. (c) As (a) but for the adiabatic cooling. (d) As (a) but for the residual term (a)±(c). In all panels the 1000-m topographic contour is indicated by a shaded area and the leading edge of the incursion is indicated by a thick, dashed line. placing warm air, while the direct forcing by the upper- The intense local cooling at the leading edge of the level wave is small. A low-level (surface to ϳ800 hPa) incursion was dominated by low-level horizontal ad- baroclinic zone and a sharp transition of the meridional vection of cold air. Adiabatic cooling and diabatic heat- component of the wind are collocated with the band of ing associated with latent heat release were important largest pressure gradient, moving equatorward at about only within an intense mesoscale updraft just ahead of 10msϪ1 (very close to that of the low-level wind speed the cold front. The model also simulates moderate just behind it). The subsequent cooling (on the order of amounts of rainfall (consistent with the satellite imag- 10ЊC) affected most of the subtropical part of the con- ery), mostly postfrontal and with little in¯uence upon tinent, with the exception of the eastern tip of Brazil. the temperature ®eld. However, the rain-induced dia- Partial blocking of the cold air dome by the Brazilian batic cooling might be important as the cold air moves highland might be responsible for the lack of cooling into lower latitudes, or in summertime episodes when along the east coast of Brazil in this episode. there is signi®cant rainfall along the leading edge of the 2850 MONTHLY WEATHER REVIEW VOLUME 127

FIG. 23. (a) Model (inner domain) 950-hPa momentum balance at 0600 UTC 14 May 1993. (b) As (a) but at 1800 UTC 14 May 1993. The key of the force vectors is indicated in the small inset. In (a) and (b) shaded area indicates terrain elevation in excess of 1000 m, and the leading edge of the incursion is indicated by a thick, dashed line. Reference vector (in N kgϪ1) at the bottom of the ®gures. The solid line in (b) indicates the segment A±B±C used in Fig. 24. DECEMBER 1999 GARREAUD 2851

FIG. 24. Model (inner domain) 950-hPa traces of the zonal (barrier-normal, u), meridional (barrier-parallel, ␷), and

meridional geostrophic (␷ g) component of the winds at 20ЊS, 61ЊW. cool air. Sensitivity experiments (dry vs moist simula- leading edge, the southerly, barrier-parallel winds are tions) are under way to address this point. close to steady state and geostrophic balance, the Along the subtropical Andes, observations, model southerly jet at and near the northern ¯ank of the cold results, and scaling analysis suggest a complete block- air is maintained against friction by the mountain- ing of the zonal ¯ow and, consequently, cold air dam- parallel pressure gradient. The imbalance between ming by the steep slope of the barrier. The horizontal these two terms in the frontal zone leads to the down- scale of the cold dome and the perturbed low-level gradient acceleration of the ¯ow, fostering the north- wind region (ϳ900 km) is controlled by a balance ward propagation of the incursion. As the cold air between the mountain-normal component of the pres- moves to the north of 18ЊS, the blocking of the zonal sure gradient (outward from the barrier) and the Cor- ¯ow diminishes due to the change in the direction of iolis force in the cold air acting upon the southerly the Andes, but the ageostrophic balance forcing ¯ow (toward the barrier). While far back from the downgradient ¯ow at the leading edge of the cold air

FIG. 25. Model (inner domain) vertical section of potential temperature (contoured every 2.5 K) and horizontal wind vectors (reference vector at the bottom of the ®gure) taken parallel to the leading edge of the incursion (segment A±B±C) at 33 h into the simulation (see Fig. 22b). The wind vectors were partitioned into their components parallel to the cross section (horizontal component of the plotted vectors) and normal to the cross section (vertical component of the plotted vectors). Shaded areas indicate wind speed normal to the cross section in excess of 8 m sϪ1. 2852 MONTHLY WEATHER REVIEW VOLUME 127 incursion still prevails because the adjustment back Boyle, J. S., and T. J. Chen, 1987: Synoptic aspects of the wintertime to geostrophy is quite slow at these low latitudes. East Asian monsoon. Monsoon Meteorology, C. P. Chang and T. N. Krishnamurti, Eds., Oxford University Press, 125±160. Thus, the topographically trapped advance of the in- Colle, B., and C. Mass, 1995: The structure and evolution of cold cursion along the subtropical Andes and the subsequent surges east of the Rocky Mountains. Mon. Wea. Rev., 123, 2577± spread of the cold air over tropical South America stems 2610. from a two-way interaction between the mass and wind Compagnucci, R. H., and M. A. Salles, 1997: Surface pressure pat- ®elds: the strong pressure (temperature) gradient drives terns during the year over southern South America. Int. J. Cli- matol., 17, 635±653. the northward acceleration of the low-level winds, while Dudhia, J., 1989: Numerical study of convection observed during the the horizontal advection of cold air by southerly winds winter monsoon experiment using a mesoscale two-dimensional maintains the strong temperature gradient against front- model. J. Atmos. Sci., 46, 3077±3107. olysis by the strong surface heat ¯uxes. A similar con- , 1993: A nonhydrostatic version of the Penn State/NCAR me- soscale model: Validation test and simulation of an Atlantic cy- ceptual and analytical model were proposed in Garreaud clone and cold front. Mon. Wea. Rev., 121, 1493±1513. and Wallace (1998). In this respect, topographically Durran, D., and L. Snellman, 1987: The diagnosis of synoptic-scale trapped Kelvin or Rossby waves appear as not being vertical motion in an operational environment. Wea. Forecast- relevant for the case examined in this paper. ing, 2, 17±31. The short-term simulation in this study did not in- Fortune, M. A., and V. E. Kousky, 1983: Two severe freezes in Brazil: Precursors and synoptic evolution. Mon. Wea. Rev., 111, 181± cluded the demise of the cold air incursion. However, 196. the observations suggest that as the cold air moves deep Fritsch, J. M., J. Kapolka, and P. A. Hirschberg, 1992: The effect of into the Tropics the cold advection no longer compen- the subcloud-layer diabatic processes on cold air damming. J. sates the strong surface heat ¯uxes, so that the local Atmos. Sci., 49, 49±70. time rate of change of air temperature at the leading Garreaud, R. D., 1998: Cold air incursions into low latitudes: Global perspective and regional analysis over South America. Ph.D. edge of the incursion becomes positive. Therefore, while thesis, University of Washington, Seattle, WA, 160 pp. [Avail- the details of the advance of the cold air incursion into able from Dept. of Atmospheric Sciences, University of Wash- low latitudes are controlled by elements acting in the ington, P.O. Box 354235, Seattle, WA 98195-4235.] mesoscale (e.g., surface drag, surface heat ¯uxes, rain- , and J. M. Wallace, 1997: The diurnal march of the convective fall, etc.) the overall structure and longevity (equator- cloudiness over the Americas. Mon. Wea. Rev., 125, 3157±3171. , and , 1998: Summertime incursions of mid-latitude air into ward extent) of the incursion depend upon the avail- tropical and subtropical South America. Mon. Wea. Rev., 126, ability of cold air over the subtropical plains of the 2713±2733. continent, which in turn is controlled by the evolution Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of of the midlatitude, large-scale ¯ow. the ®fth-generation Penn State/NCAR mesoscale model. NCAR Tech. Note NCAR/TN-398ϩIA, 107 pp. [Available from NCAR, P.O. Box 3000, Boulder, CO 80307.] Acknowledgments. The author gratefully acknowl- Hamilton, M. G., and J. R. Tarifa, 1978: Synoptic aspects of a polar edges many helpful comments and advice provided by outbreak leading to frost in tropical Brazil, July 1972. Mon. Wea. Drs. John M. Wallace and Brian Colle. Drs. Cliff Mass Rev., 106, 1545±1556. and Brad Smull also offered many useful comments. Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain±Fritsch scheme. The Representa- The comments of two anonymous reviewers lead to a tion of Cumulus Convection in Numerical Models, Meteor. Mon- substantial improvement in presentation and contents of ogr., No. 46, Amer. Meteor. Soc., 165±170. this work. NCEP±NCAR reanalysis data were provided Kalnay, E., and Coauthors, 1996: The NCEP±NCAR 40-Year Re- by the NOAA Climate Diagnostics Center. The author analysis Project. Bull. Amer. Metor. Soc., 77, 437±472. Keyser, D. A., and M. A. Shapiro, 1986: A review of the structure is partially supported by the National Science Foun- and dynamics of upper-level frontal zones. Mon. Wea. Rev., 114, dation under Grant 9215512 and by the Department of 452±499. Geophysics of the University of Chile. Kousky, V. E., 1979: Frontal in¯uences on Northeast Brazil. Mon. Wea. Rev., 107, 1140±1153. , and J. Ferreira, 1981: Frontal in¯uences on Northeast Brazil: REFERENCES Their spatial distributions, origins and effects. Mon. Wea. Rev., 109, 1999±2008. Baines, P. G., 1980: The dynamics of the southerly buster. Aust. , and I. Cavalcanti, 1997: The principal modes of high-frequency Meteor. Mag., 28, 175±200. variability over the South American region. Preprints, Fifth Int. Bell, G. D., and L. F. Bosart, 1988: Appalachian cold-air damming. Conf. on Southern Hemisphere Meteorology and Oceanography, Mon. Wea. Rev., 116, 137±161. Pretoria, South Africa, Amer. Meteor. Soc., 7B.2±7B.3. Blackadar, A. K., 1979: High resolution models of the planetary Lau, K. M., and C. P. Chang, 1987: Planetary scale aspects of the boundary layer. Advances in Environmental Sciences and En- winter monsoon and atmospheric teleconnection. Monsoon Me- gineering, J. Plaf¯in and E. Ziegler, Eds., Vol. 1, Gordon and teorology, C. P. Chang and T. N. Krishnamurti, Eds., Oxford Breach Scienti®c, 50±85. University Press, 161±202. Bluestein, H. B., 1993: Synoptic±Dynamic Meteorology in Mid-lat- Lupo, A. R., P. J. Smith, and P. Zwack, 1992: A diagnosis of the itudes. Vol. 2, Observations and Theory of Weather Systems, explosive development of two extratropical cyclones. Mon. Wea. Oxford University Press, 594 pp. Rev., 120, 1490±1523. Bosart, L. F., A. R. Lupo, D. J. Knight, E. G. Hoffman, and J. J. Marengo, J., A. Cornejo, P. Satymurty, C. Nobre, and W. Sea, 1997: Nocera, 1998: Cyclonic and anticyclonic South American cold Cold surges in tropical and extratropical South America: The surges. Preprints, Eighth Conf. on Mountain Meteorology, Flag- strong event in June 1994. Mon. Wea. Rev., 125, 2759±2786. staff, AZ, Amer. Meteor. Soc., 10.2±10.3 McBride, J. L., and K. L. McInnes, 1993: Australian southerly bust- DECEMBER 1999 GARREAUD 2853

ers. Part II: The dynamical structure of the orographically mod- , , and , 1998: Planetary- and synoptic-scale signatures i®ed front. Mon. Wea. Rev., 121, 1921±1935. associated with Central American cold surges. Mon. Wea. Rev., Moore, J. T., and G. E. VanKnowe, 1992: The effect of jet-streak 126, 5±27. curvature on kinematic ®elds. Mon. Wea. Rev., 120, 2429±2441. Tilley, J. S., 1990: On the application of edge wave theory to terrain Negri, A., R. Adler, R. Maddox, K. Howard, and P. Keehn, 1993: A bounded cold surges: A numerical study. Ph.D. thesis, The Penn- regional rainfall climatology over Mexico and the southwest sylvania State University, University Park, PA, 353 pp. [Avail- United States derived from passive microwave and geosynchro- able from Dept. of Meteorology, The Pennsylvania State Uni- nous infrared data. J. Climate, 6, 2144±2161. versity, University Park, PA 16802.] Parmenter, F. C., 1976: A Southern Hemisphere cold front passage Uccellini, L. W., and D. R. Johnson, 1979: The coupling of upper at the equator. Bull. Amer. Meteor. Soc., 57, 1435±1440. and lower tropospheric jet streaks and implications in devel- Pierrehumbert, R. T., and B. Wyman, 1985: Upstream effects of me- opment of severe convective storms. Mon. Wea. Rev., 107, 682± soscale mountains. J. Atmos. Sci., 42, 977±995. 703. Ramage, C. S., 1971: Monsoon Meteorology. Academic Press, 359 Vera, C. S., and P. K. Vigliarolo, 2000: A diagnostic study of cold- pp. air outbreaks over South America. Mon. Wea. Rev., in press. Ratisbona, C. R., 1976: The climate of Brazil. Vol. 12, Climate of Wu, M. C., and J. C. L. Chan, 1995: Surface features of winter Central and South America, W. Schwerdtfeger and H. E. Lands- monsoon surges over South China. Mon. Wea. Rev., 123, 662± berg, Eds., World Survey of Climatology, Elsevier, 219±293. 680. Rossow, W., and R. Schiffer, 1991: ISCCP cloud data products. Bull. , and , 1997: Upper-level features associated with winter Amer. Meteor. Soc., 72, 2±20. monsoon surges over South China. Mon. Wea. Rev., 125, 317± Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory for 340. strong, long-lives squall lines. J. Atmos. Sci., 45, 463±485. Xu, Q., 1990: A theoretical study of cold air damming. J. Atmos. Schultz, D. M., W. E. Bracken, L. F. Bosart, G. J. Hakim, M. A. Sci., 47, 2969±2985. Bedrick, M. J. Dickinson, and K. R. Tyle, 1997: The 1993 Su- Zwack, P., and B. Okossi, 1986: A new method for solving the qua- perstorm cold surge: Frontal structure, gap ¯ow and tropical sigeostrophic omega equation by incorporating surface pressure impact. Mon. Wea. Rev., 125, 5±39. tendency data. Mon. Wea. Rev., 114, 655±666.