Downloaded 10/01/21 08:47 AM UTC 2062 JOURNAL of the ATMOSPHERIC SCIENCES VOLUME 57 Et Al

1JULY 2000 BRYAN AND FRITSCH 2061

Diabatically Driven Discrete Propagation of Surface Fronts: A Numerical Analysis

GEORGE H. BRYAN AND J. MICHAEL FRITSCH Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania

(Manuscript received 2 July 1998, in ®nal form 30 August 1999)

ABSTRACT Discrete frontal propagation has been identi®ed as a process whereby a surface front discontinuously moves forward, without evidence of frontal passage across a mesoscale region. Numerical simulations are employed to examine the upper-level evolution of a discrete frontal propagation event and to explore the processes that were responsible for the discrete movement. Model results indicate that a frontal pressure trough was not able to penetrate through a deep surface-based layer of cool air created by a precipitating convective system several hundred kilometers in advance of the front. Meanwhile, a new low-level baroclinic zone formed well ahead of the front along the southern side of the cool layer. As the midlevel front moved continuously over the cool layer, a new low-level front developed in the new baroclinic zone and the original low-level front dissipated. At the surface, the simulated front did not pass through the cool layer. Frontogenesis terms reveal that the prefrontal circulation that becomes the new frontal circulation initially forms directly from diabatic frontogenesis. Daytime heating in the prefrontal boundary layer and cooling from thunderstorms combine to create a thermal gradient and a mesoscale pressure perturbation. Winds turn in response to the altered pressure ®eld and form a convergent boundary, resulting in kinematic frontogenesis. The boundary subsequently undergoes rapid intensi®cation. Sensitivity studies were conducted in which latent heating due to precipitation was withheld and the in¯uence of clouds on the radiation scheme was ignored. In a simulation with both of these effects withheld, the original front passes continuously through the region, that is, there is no discrete propagation. Thus, diabatic processes associated with a large complex of thunderstorms were necessary to induce the discrete frontal propagation in this case. This conclusion contrasts with previous studies, where fronts were observed to propagate discretely in dry environments.

1. Introduction disrupted by the deep, stable cold dome created by the convection, the midlevel front propagated continuously. Discrete surface frontal propagation is de®ned as dis- This paper presents a numerical simulation of the case continuous movement of a front at the surface. Bryan studied by Bryan and Fritsch (2000, hereafter referred and Fritsch (2000) examined a case of discrete frontal to as BF). A number of aspects of this case warrant propagation in a convective environment. In their case, further study. In particular, calculations of surface fron- a prefrontal convective system generated a large, deep togenetical tendencies in BF suggest that diabatic pro- cold dome just ahead of an approaching cold front. The cesses played an important role in initiating the pre- low-level cold front was observed to stall and undergo frontal trough. Output from a numerical simulation can frontolysis as it encountered the cold dome. Meanwhile, be used to con®rm or refute this suggestion. Further- a trough formed on the downstream side of the cold more, it remains to be seen whether the diabatic pro- dome and underwent rapid frontogenesis. Eventually, cesses were completely necessary in inducing the dis- the original surface frontal trough disappeared, the me- crete frontal movement in this case. In a study of discrete soscale wind ®eld adjusted to the new pressure pattern, frontal propagation in stable winter conditions, Charney and the prefrontal trough replaced the original front as and Fritsch (1999) found that cold fronts can propagate the dominant frontal feature. Their analysis suggests discretely without the need for precipitation processes. that, although the low-level frontal properties (e.g., pres- On the other hand, there is evidence that diabatic pro- sure trough, wind shift, and temperature gradient) were cesses may be important in initiating frontogenesis by creating mass ®eld changes and, eventually, wind ®eld changes as the system adjusts toward geostrophic bal- ance. The mesoscale wind shift lines and troughs may Corresponding author address: George H. Bryan, 503 Walker Building, University Park, PA 16802. then undergo frontogenesis. For example, several stud- E-mail: [email protected] ies (e.g., Purdom 1982; Pielke and Segal 1986; Segal

᭧ 2000 American Meteorological Society

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2062 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57 et al. 1986) note that gradients in cloud cover can induce mesoscale circulations similar to sea and land breezes. Another diabatic process that can generate mesoscale boundaries is precipitation. Doswell (1982) showed that thunderstorm complexes can form intense out¯ow boundaries with properties similar to cold fronts. Ryan et al. (1989) presented a case where precipitation generated a new baroclinic zone several hundred kilometers in advance of an approaching cold front. This paper examines the 5±6 September 1994 discrete frontal propagation event using a mesoscale-resolution simulation. The model output complements the observational analysis and adds insight that cannot be drawn from standard analyses. The role of physical processes are explored by examining output from the various model equations. Section 2 describes the numerical model con®guration. Section 3 examines the simulated evo- FIG. 1. Con®guration of model domains. lution of the event, and compares the results with observations. Section 4 presents the vertical structure and evolution of discrete frontal propagation. Analysis of the scheme developed by Kain and Fritsch (1990). Of frontogenesis terms and quanti®cation of diabatic pro- particular value to this study, the Kain±Fritsch scheme cesses are presented in section 5. The results of several includes parameterized downdrafts necessary for sim- sensitivity experiments using the numerical model are ulating the convectively modi®ed low levels. included in section 6. A conceptual model for discrete The model initial conditions are derived from ®rst- frontal propagation is presented in section 7, followed guess ®elds supplied by the National Centers for En- by conclusions in section 8. vironmental Prediction that are then objectively enhanced with observed surface and upper-air data (Man- ning and Haagenson 1992). At the beginning of the 2. Model design forecast, the four-dimensional data assimilation tech- The ®fth-generation Pennsylvania State University± nique of Stauffer and Seaman (1994) is utilized to bring National Center for Atmospheric Research Mesoscale the dynamic and thermodynamic ®elds into closer agree- Model (MM5) is utilized for this investigation (Dudhia ment with the observations while still satisfying the gov- 1993; Grell et al. 1994). MM5 is a nonhydrostatic model erning hydrodynamical system. with prognostic equations for temperature, horizontal, The numerical simulation of the 5±6 September 1994 and vertical components of motion, pressure, water va- discrete frontal propagation event was initialized at 0000 por mixing ratio, cloud water mixing ratio, and rain- UTC 5 September 1994 (0000/5) and run for 36 h, end- water mixing ratio. Twenty-®ve sigma (␴) levels are ing at 1200/6. For the ®rst 12 h, four-dimensional data used in this study with 11 levels below 700 hPa where assimilation was applied. This simulation is hereafter boundary layer processes are important. Two model do- referred to as the control simulation (CTL). mains are used: a coarse mesh of 61 ϫ 73 grid points For cross-section analyses of model output, averages with 60-km grid spacing, and a two-way interactive of several cross sections oriented perpendicular to the nested ®ne mesh domain of 64 ϫ 76 grid points with fronts are constructed. The averages eliminate high-fre- 20 km grid spacing (Fig. 1). The location of the nested quency perturbations (such as those associated with ®ne mesh domain was chosen such that information moist parameterized convection at individual grid ele- from the outer boundaries of domain 1 would not be ments) but still retain the mesoscale structure and ver- advected into domain 2 during the model simulation. tical circulations associated with the fronts. The meth- This is important for dry run experiments since the odology of the cross section averaging is described in boundary conditions for domain 1 are observed con- the appendix. ditions that include precipitation processes. Model physics used in these simulations include the 3. Surface and upper-air evolution explicit bulk precipitation scheme with simple ice physics based on the studies of Lin et al. (1983), Rutledge Despite minor differences in the timing and magni- and Hobbs (1983), Hsie et al. (1984), and Dudhia tude of features, the model is able to capture the basic (1989). The planetary boundary layer (PBL) parame- sequence of events in this case (Fig. 2). At the surface, terization is based on the work of Blackadar (1976, a prefrontal mesohigh and trough form at about the cor- 1979) and adapted for MM5 by Zhang and Anthes rect time and location. Over time, the surface trough (1982) and Zhang and Fritsch (1986). The subgrid-scale intensi®es and moves southward while the original front convective parameterization used in the simulation is slows, stalls, weakens, and dissipates. After the original

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2063

FIG. 2. Surface analysis from the ®ne mesh domain at (a) 1200 and (b) 1800 UTC 5 Sep and (c) 0000 and (d) 0600 UTC 6 Sep 1994. Sea level pressure is contoured every 1 hPa (solid) and temperature is contoured every 2ЊC (dashed). front dissipates, large-scale pressure and wind ®elds ad- porarily) reversed from what one would expect at a cold just to the new frontal feature. Based on the fact that front. discrete frontal propagation is captured by the model, At 0000/6, frontolysis is occurring at the original front it is assumed that the basic processes are represented in (Fig. 2c). The front is still identi®able by a pressure the model output. Therefore, the model output can be trough and wind shift in southeastern Kansas and north- used to draw conclusions about the mechanisms re- central Oklahoma, though no strong temperature gra- sponsible for discrete frontal propagation. dient exists. The PFT at this time continues to intensify Looking more carefully at the model output, condi- and move southward. By 0600/6, the pressure trough tions at 1200/5 capture the cold and warm fronts at the and convergence zone associated with the front are vir- surface with good placement (cf. Fig. 2a with BF Fig. tually nonexistent and identi®cation of a surface front 4a). By 1800/5 a pronounced mesohigh and cold pool can no longer be justi®ed (Fig. 2d). As the frontal trough have developed ahead of the cold front (Fig. 2b). At disappears, the interfront mesohigh becomes incorpo- this time, a prefrontal trough (PFT) is also analyzed. rated into the synoptic-scale high pressure system to the Winds have responded to the interfront pressure rises, north. Meanwhile, pressure and wind ®elds adjust to the creating con¯uence and thus frontogenetic tendencies new surface front. It is readily apparent in this sequence along the PFT. It is also important to note that the model (Fig. 2) that there is a large area through which the is able to simulate the warm pocket that develops behind original front does not propagate at the surface. the original front in central Kansas. Thus, the surface Simulated precipitation is shown in Fig. 3. The model temperature gradient at the front is locally (and tem- generates a broad swath of precipitation in the interfront

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2064 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

FIG. 3. Simulated precipitation (mm) on the ®ne mesh domain. Contours represent 1.0, 5.0, 10.0, and 20.0 mm. region, in roughly the same area that radar echoes were observed (cf. Figs. 3 and 5 of BF). The model also reproduces several speci®c convective features; for example, a convective system is predicted behind the front in Kansas. The evolution of conditions at upper levels also compares favorably with observations. At 850 hPa, a single frontal trough is apparent in the height and wind ®elds at 1200/5 (Fig. 4a). The thermal pattern, marked by a warm ribbon to the south and a strong temperature gradient to the north, also supports the existence of a single front at this level. By 1800/5, a more complicated thermal pattern emerges at 850 hPa due to the introduction of a cold anomaly in the interfront region (Fig. 4b). Now there are two warm ribbons: one above the original front, and one above the developing prefrontal trough. A weak height ridge in the same location as the cold anomaly suggests that the cold dome extends farther above 850 hPa. Over most of the PFT, a second trough and wind shift do not develop at this level; the exception is along the far eastern end of the PFT. This trough is associated with a shortwave that moves rapidly through the north- west ¯ow, mostly to the east of where the discrete propagation occurs. It may have been in¯uential along the eastern end of the PFT, but a height trough (and associated wind shift) does not emerge over the PFT in Oklahoma or Texas. Furthermore, this trough is later observed to move ahead of and away from the PFT

→

FIG. 4. Simulated ®elds at 850 hPa at (a) 1200 and (b) 1800 UTC 5 Sep and (c) 0000 UTC 6 Sep 1994: geopotential height (solid contours) every 15 m, and temperature (dashed) every 2ЊC. Positions of surface fronts are indicated in gray. A low-pass ®lter (Barnes 1964) was applied to the ®elds.

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2065

(new front), further suggesting that it was not critical in the development of the PFT. This evolution at 850- hPa contrasts with BF, which shows a double structure at 850 hPa as well as the surface. It will be shown later that a second upper-level trough does form in this simulation, but it does not initially reach 850 hPa. Thus, the developing new front in the simulation is not as deep as the observed system. Six hours later at 0000/6, the cold anomaly at 850 hPa is not as well pronounced, as it begins to merge with the postfrontal temperature gradient (Fig. 4c). A warm pocket over eastern Kansas and northern Missouri is all that remains of the warm ribbon that was located along the original front. The shortwave trough at the eastern end of the new front continues to deepen. Later, this shortwave moves away from the PFT. During the next 12 h, the thermal ®eld continues to adjust to the new frontal location (not shown). Rapid changes also occur at 700 hPa. Like at lower levels, there is a single frontal structure at 1200/5 (Fig. 5a). However, there is already an indication of a cold anomaly just ahead of the front. By 1800/5, there is a pronounced interfront cold anomaly (Fig. 5b), as seen at 850 hPa. Unlike at lower levels, locally higher heights do not develop in the interfront region. Instead, lower heights prevail, which is consistent with the observations (BF) and Zhang and Harvey (1995, a study of convection ahead of a cold front, which will be dis- cussed later). Like at 850 hPa, deepening heights along the eastern end of the PFT can be seen through this sequence. A height trough and wind shift do not develop over most of the PFT at this level. Rather the original height trough of the front moves continuously through Oklahoma and western Texas (Fig. 5c). Farther aloft at 500 hPa, it is dif®cult to identify a strong frontal structure (Fig. 6). This is consistent with observed warm season fronts in other studies (e.g., Sanders 1955; Ogura and Portis 1982; Reeder and Smith 1992). However, the strongest temperature gradient in the postfrontal region at 1200/5 (Fig. 6a) moves southward into the interfront region. This pattern is inter- rupted at 0000/6 by the appearance of a warm anomaly, apparently induced by the convection (Fig. 6b). In addition, a weak shortwave trough intensi®es in the interfront region. The emergence of a shortwave trough during the period of discrete frontal propagation has implications for the conceptual model of this event. In a case study of convection ahead of a cold front, Zhang and Harvey (1995) show that a mesoscale convective system acted to intensify the midlevel height trough.

←

FIG. 5. Simulated ®elds at 700 hPa at (a) 1200 and (b) 1800 UTC 5 Sep and (c) 0000 UTC 6 Sep 1994: geopotential height (solid contours, 180 ϭ 3180) every 15 m, and temperature (dashed) every 2ЊC [an intermediate contour of 7ЊC is included in (b) and (c)]. Po- sitions of surface fronts are indicated in gray. A low-pass ®lter (Barnes 1964) was applied to the ®elds.

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2066 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

FIG. 7. Response curves used for two low-pass ®lters (B1 and B2) and the response curve used for the bandpass ®lter (BP). The BP is obtained by subtracting B2 from B1 and then normalizing the resulting curve to a maximum value of 1.0. See Maddox (1980) for additional details.

eliminates wavelengths less than 900 km (B2 in Fig. 7). The bandpass response curve is obtained by subtracting B2 from B1. The resulting response curve is then normalized to a maximum value of 1.0. The bandpass response curve used here (BP in Fig. 7) highlights features with a wavelength of about 400 km, that is, mesoscale wavelengths. Vertical cross sections of bandpass geopotential height reveal that the frontal trough has a forward-sloping structure at 1300/5 (Fig. 8a), similar to that noted by BF in their analyses of the observations. Later, the midlevel trough ampli®es and detaches from the surface trough (Fig. 8b). As this occurs, the PFT forms and FIG. 6. Simulated ®elds at 500 hPa at (a) 1200 UTC 5 Sep and (b) intensi®es. Notice that the trough that was observed over 0000 UTC 6 Sep 1994: geopotential height (solid contours, 580 ϭ the far eastern end of the PFT at 850 and 700 hPa (Figs. 5580) every 20 m, and temperature (dashed) every 2ЊC. Positions of 4 and 5) does not appear over the PFT in this cross surface fronts are indicated in gray. A low-pass ®lter (Barnes 1964) was applied to the ®elds. section; this observation further supports the conclusion that the shortwave trough was not instrumental in the development of the PFT. Later, the midlevel trough con- They further show that this convectively induced inten- nects with the PFT (Fig. 8c). Low heights associated si®cation acts to enhance synoptic-scale frontogenesis with the original low-level front ®ll in and disappear. and cyclogenesis. The resulting structure is a continuous height trough from the surface to midlevels, which is more tilted back than was noted earlier. These bandpass height analyses 4. Vertical structure suggest that the front aloft propagates continuously a. Synoptic view while discrete frontal propagation occurs in low levels. This agrees with the observational study (BF), where a To highlight the vertical structure of the frontal double-frontal structure was found only in the lower trough, a bandpass ®lter was utilized. Following Mad- troposphere. dox (1980), two low-pass ®lters are applied (Barnes Large-scale cross sections of potential temperature 1964, 1973). One ®lter effectively eliminates wave- further reveal the overall structure and evolution of the lengths less than 150 km (B1 in Fig. 7); the second ®lter frontal features (Fig. 9). Early in the simulation, there

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2067

is a typical frontal structure with sloping isentropes behind the front and level isentropes ahead of the front (Fig. 9a). A deep and broad region of ascent also de®nes the position of the front. Four hours later at 1600/5, a low-level cold dome has developed ahead of the front in the convectively active region (Fig. 9b). Temperatures in the interfront region are as cold as those immediately behind the front. To the south of the cold dome (to the right on the cross section), a deep well-mixed boundary layer is present in the clear air. Thus, there is a new baroclinic zone between the cold dome and the mixed PBL. A vertical circulation develops in this baroclinic zone, such that two areas of ascent are apparent: one along the original front, and one along the PFT. By 2200/5 the near-surface potential temperature gradient of the original front is dissipating and lifting (Fig. 9c). As this occurs, positive vertical motion weakens and disappears. Meanwhile, ascent at the PFT becomes more pronounced than in the previous analysis. Furthermore, the horizontal temperature gradient from the surface to 800 hPa is stronger and more coherent at the PFT. By 0000/6, the developing frontal structure becomes the dominant frontal feature (Fig. 9d). This thermal evolution resembles the case studied by Ryan et al. (1989, particularly their Fig. 8). In the case they studied, precipitation ahead of a cold front generated a deep, surface-based cold layer. A new baroclinic zone developed several hundred kilometers ahead of the front, between the cold dome and a well-mixed boundary layer. Frontogenetical tendencies were observed in this baroclinic zone. The case of 5±6 September 1994 in the present paper and BF show that these prefrontal baroclinic zones can become the major synoptic-scale front. Several studies of Australian fronts identify mesoscale ``change lines'' within a broader frontal transition zone (e.g., Berson et al. 1957; Clarke 1961; Garratt et al. 1985; Garratt 1988; Smith et al. 1995). These change lines are observed to develop, intensify, and decay while the overall frontal transition zone propagates continuously. Bryan and Fritsch (2000) argue that this case is fundamentally different from these change lines in two ways: the Australian change lines are of smaller scale, and are associated with mesoscale features (such as sea breezes and out¯ow boundaries); and the PFT in this study intensi®es at the same time that the original front dissipates, that is, the PFT replaces the original front. The cross sections in this section support these two points, especially since these cross-section averages re- move small-scale structures such as gust fronts. Only

←

FIG. 8. Average cross sections of bandpass height (contour interval 2 m, zero contour excluded) at (a) 1300/5, (b) 2300/5, and (c) 1100/6. Data are from the coarse mesh domain. The location of the cross section is approximately from western Nebraska to southern Loui- siana.

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2068 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

FIG. 9. Average cross sections of potential temperature (contours, K) and vertical velocity (shaded, representing values greater than 2 cm sϪ1) from the coarse mesh domain at (a) 1200/5, (b) 1600/5, (c) 2200/5, and (d) 0000/6. The location of the cross sections is from South Dakota (left) to the Gulf of Mexico south of Louisiana (right). synoptic-scale and the largest mesoscale features re- frontal trough dissipates and becomes incorporated into main. Furthermore, these ®gures show that the new front the synoptic-scale high pressure to the north. As this becomes the only large-scale frontal feature. occurs, low-level convergence and ascent aloft cease. Further insight can be drawn by inspecting the potential temperature pattern together with the horizontal b. Mesoscale structure wind ®eld. Along the PFT, frontogenetic tendencies are As noted in BF, a sudden deep-layer wind adjustment strong due to a collocation of temperature gradient and to the new low-level front occurs near the completion convergence (cf. Figs. 10b,c and 12b,c). In contrast, of discrete surface frontal propagation. This process can convergence along the original low-level front is not be seen in the model simulation also. Cross sections of frontogenetic since virtually no favorable horizontal wind speed normal to the fronts show that the original temperature gradient exists there. Thus, frontogenetical convergent pattern associated with one front (Fig. 10a) support for the original front is disrupted by the very is replaced by a double structure (Fig. 10b). As the cold (relative to temperatures behind the cold front) pre- original low-level front dissipates, winds just ahead of frontal environment generated by the persistent con- the front turn to northerly (Fig. 10c), resulting in only vection. one region of contrasting cross-frontal windsÐalong the Another potential factor in frontolysis along the orig- PFT. In this same time period, vertical motion at the inal front is the introduction of substantially cooler air original low-level front undergoes a rapid transition just ahead of it. The presence of this cool air weakens from ascent to descent (Figs. 11b and 11c). The as- the horizontal temperature gradient and thereby reduces cending circulation at the PFT intensi®es throughout this the rate of frontogenesis. In addition, the model reveals time period. By 2300/5, the only appreciable region of that ascent is quickly replaced by descent at this time, ascent is associated with the new frontal circulation. so tilting frontogenesis ceases along the original front. Combining these vertical and horizontal motions to- Thus, there is no mechanism to support low-level front- gether, the mesoscale ¯ow surges downgradient and to- ogenesis and the surface front weakens. Hypothetically, ward the PFT as the discrete frontal jump occurs. This if the original frontal trough was more intense, the fron- sequence occurs as heights fall along the PFT and as tal trough might not have been canceled by the cold heights rise along the original front; that is, the original pool±induced high pressure anomaly, and the front

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2069

FIG. 10. Average cross sections of horizontal wind speed normal FIG. 11. Average cross sections of vertical velocity (w, cm sϪ1) to the front (u, msϪ1) from the ®ne mesh domain: (a) 1200, (b) 1600, from the ®ne mesh domain: (a) 1200, (b) 1600, and (c) 2300 UTC and (c) 2300 UTC 5 Sep 1994. Contour interval is 2 m sϪ1. 5 Sep 1994. Contour interval is 2 cm sϪ1. The zero contour is excluded.

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2070 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

Doswell 1982). The key difference in this case appears to be the depth and intensity of the prefrontal stable layer, and a shallow front that does not penetrate through the stable layer. Thus, the surface frontal trough weakens on one side of the cold dome and emerges on the other side.

5. Analysis of frontogenesis To quantify the frontogenetical tendencies that were inferred from the thermal and wind ®elds, the frontogenesis equation (Miller 1948) is utilized: D (١␪|, (1| F ϭ Dt where ␪ is potential temperature. For a cross section oriented perpendicular to the front (i.e., for the cross sections that are presented here), a two-dimensional frontogenesis equation can be used: ␪ץ D F ϭ , (2) xץDtΗΗ which can be expanded to yield qץ ␪ץ wץ ␪ץ ␷ץ ␪ץ uץ F ϭϪ Ϫ Ϫ ϩ , (3) xץ zץ xץ yץ xץ xץ xץ where positive x points toward the warm air (to the right on the cross sections), positive y points into the cross section, positive z points upward, and q is the heating tendency due to diabatic processes. The ®rst term quanti®es horizontal con¯uence deformation, and will hereafter be referred to as F1. The second term is horizontal shearing deformation. Since this term has been found to be negligible compared to the other terms in this case, it will not be presented. The third term is the tilting term (F2). The fourth term is the diabatic term (F3). Calculations of q were obtained directly from the numerical model. Temperature tendencies from the following effects were summed to yield q: the explicit moisture scheme, the Kain±Fritsch convective parameterization, the Blackadar PBL parameterization, long- wave and shortwave radiation, and turbulent diffusion. Fifteen-minute averages of these terms were computed, where the average comprises data from every time step during model integration. At 1600 UTC, the prefrontal cold dome is beginning to emerge (Fig. 12b). The q ®eld supports the hypothesis made earlier that diabatic cooling generated this feature (Fig. 13a). Analysis of the individual heating terms (not FIG. 12. Average cross sections of potential temperature (␪, K) from the ®ne mesh domain: (a) 1200, (b) 1600, and (c) 2300 UTC shown) con®rms that evaporative cooling from the ex- 5 Sep 1994. Contour interval is 2 K. plicit moisture scheme and parameterized cooling from the Kain±Fritsch scheme are responsible for this cooling. Near the surface on either side of the cold dome, would have propagated continuously through the cold there exists warming from the PBL scheme, associated pool. In fact, this is observed to occur: fronts can en- with turbulent mixing of sensible heat ¯ux from the counter low-level, convectively generated cold pools surface. The strong heating in the PBL behind the front and pass through the area seemingly unaffected (e.g., agrees with observations in BF, and is the cause of the

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2071

suggested earlier, con¯uence frontogenesis is negligible along the front (Fig. 14a), despite well-de®ned convergence there (Fig. 10a). Over the next several hours, kinematic frontogenesis (F1 and F2) dramatically increase in magnitude and area along the prefrontal trough (Fig. 15). The terms F1, F2, and F3 all contribute to frontogenesis along the PFT. The corresponding plot of total frontogenesis shows an intense maximum near the surface. Near the original front, no signi®cant values of frontogenesis exist. These data support the PFT as the dominant frontogenetic feature by this time. Hourly plots of the frontogenesis terms near the PFT show that positive values of F3 appear along the PFT before the other two terms (Fig. 16). This time-height plot neglects the complete three-dimensional structure of frontogenesis, including the typical dipole structure of the tilting term [as shown in Figs. 14b and 15b, as well as other studies (e.g., Sanders 1955; Baldwin et al. 1984; Koch et al. 1995)]. Nevertheless, these data support the conclusion inferred from the thermal and wind evolution that diabatic processes initiated the creation of a new low-level thermal gradient. Then the pressure and therefore wind ®eld adjusted to the new thermal pattern, initiating con¯uence deformation and tilting frontogenesis, which further intensi®ed the thermal gradient. This model contrasts with the quasigeostrophic view of frontogenesis (e.g., Hoskins 1982; Bluestein 1986), where large-scale cyclogenetic processes impose horizontal deformation onto a preexisting low-level baroclinic zone. In the present case, there is no preexisting baroclinic zoneÐit is created by the diabatic processes before the usual frontogenetic processes can ensue. These conclusions agree with several studies that FIG. 13. Average cross sections of diabatic heating rate (q, KhϪ1) show how sensible heating from PBL processes can from the ®ne mesh domain: (a) 1200, (b) 1600, and (c) 2300 UTC enhance frontal structure (e.g., Keyser and Anthes 1982; 5 Sep 1994. Contour interval is 0.25 K hϪ1. The zero contour is excluded. Thorpe and Nash 1984; Shapiro 1983; Keyser 1986, section 10.3.1; Moore 1991; Blumen et al. 1996; Miller et al. 1996; Koch et al. 1997). A recent numerical study reversal of temperature gradient there. In contrast, by Koch et al. (1995) is highly relevant to this work; strong differential heating exists across the zone where they found that differential PBL heating enhanced front- the prefrontal trough forms. Later, at 2300 UTC, a deep ogenesis when compared to numerical simulations of layer of cooling continues in the interfront zone, al- uniform PBL heating. The results of the present paper though the PBL heating becomes negligible (Fig. 13b). con®rm that differential PBL processes are strongly The frontogenesis terms at 1600/5 show two low- frontogenetic. Furthermore, this paper shows that PBL level maxima: one behind the original front, and one processes can play a role in creating new ageostrophic where the PFT develops (Fig. 14d). There is a deep circulations by creating horizontal temperatures gradi- layer of positive F3 about 300 km ahead of the front, ents that did not previously exist. where the PFT forms. Above the surface, F2 mostly cancels with F3, which is consistent with other studies 6. Sensitivity studies (e.g., Ballentine 1980; Orlanski et al. 1985; Koch et al. 1995). However, near the surface where w is necessarily To see how sensitive discrete frontal propagation is small, F3 is not counteracted by tilting. Thus, diabatic to diabatic processes, additional simulations were con- frontogenesis is a strong contributor to the creation of ducted in which one or more physical parameters were the PFT. An aspect of frontogenesis that was inferred withheld from the model solution. The ®rst of these is from other ®gures but quanti®ed here is frontolysis at a dry run simulation (hereafter referred to as DRY1) in the original front caused by cooling ahead of the front which latent heating due to precipitation is neglected. and by warming behind the front (Fig. 14c). Also, as Speci®cally, heating tendencies from the explicit mois-

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2072 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

FIG. 14. Average cross sections at 1600 UTC 5 Sep 1994: (a) F1, (b) F2, (c) F3, and (d) F. Contour interval is 5 ϫ 10Ϫ9 KmϪ1 sϪ1. Values greater than ϩ5 ϫ 10Ϫ9 KmϪ1 sϪ1 are shaded. The zero contour is excluded. ture scheme (which includes evaporation and melting Thus, there is an indirect effect of condensation/subli- of hydrometeors) and the convective parameterization mation on the generation of a prefrontal boundary: cloud are not considered. Moisture still exists in the simulation patterns induce differential solar heating, and thus dia- and is included in virtual temperature calculations and batic frontogenesis. This result is similar to studies that cloud-radiative effects. Thus, this simulation is not truly have noted developing mesoscale circulations at cloud dry. To make sure that the presence of moisture does boundaries (e.g., Purdom 1982; Segal et al. 1986). not affect the momentum ®elds, water loading effects Based on these results, a second sensitivity experi- are not considered. ment includes latent heating from phase changes, but Output from DRY1 clearly retains a signal of discrete neglects clouds in the radiation and PBL schemes. In propagation (Fig. 17a). The primary factor leading to effect, this is a clear sky radiation experiment (hereafter discrete frontal propagation can be seen in the surface referred to as CLR). This simulation also contains a thermal ®eld; a temperature gradient develops well signal of discrete frontal propagation in the solution ahead of the front. Analysis of the model output reveals (Fig. 17b). Once again, a strong thermal gradient is that the presence of cloud water in conjunction with the created well ahead of the approaching front. This time radiation and PBL schemes is primarily responsible for the gradient is created primarily by the precipitation- creating the thermal gradient. Although latent heating induced cooling that occurs in the interfront region. due to phase changes is not considered, clouds are still The ®nal sensitivity experiment eliminates heating created by the explicit and implicit moisture schemes. due to phase changes and imposes clear sky radiation Other physical processes (such as radiation) interact conditions, that is, no clouds or precipitation. This ex- with these clouds, so warming occurs in the prefrontal periment (hereafter referred to as DRY2) does not fea- clear air while conditions under the clouds remain cool. ture a discrete propagation of the front (Fig. 17c). Rath-

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2073

FIG. 15. Same as Fig. 14 but for 2300 UTC 5 Sep 1994. er, the original low-level front moves continuously reveal that the primary cold anomaly (the interfront cold through Oklahoma and Missouri. dome) extends to about 700 hPa. This depth is consistent From these sensitivity experiments, it is clear that with other studies (e.g., Lewis 1975; Kuo and Anthes diabatic processes were necessary for the discrete fron- 1984; Cotton et al. 1989; Ryan et al. 1989). Above 600 tal propagation in this case. This is in contrast to the hPa, warm anomalies from latent heat release are ap- case studied by Charney and Fritsch (1999), which prop- parent. The strongest forcing for the discrete frontal agated discretely through adiabatic processes alone. propagation, revealed by the highest perturbation values To highlight the magnitude and extent of diabatic and strongest gradients, develops near the surface, but processes, difference ®elds between the various exper- extends into the midtroposphere. iments were calculated. By subtracting DRY2 from A cross section of the geopotential height difference CTL, the effects that condensation and sublimation have between CTL and DRY2 shows the vertical extent of on the model solution, both direct and indirect, are re- convective effects. The cross section at 1700/5 shows vealed. The interfront mesohigh and anticyclonic cir- a typical convective signature, with high heights near culation are clearly evident in the difference ®elds (Fig. the surface, low heights in midlevels, and high heights 18a). The mesohigh is coincident with a broad region farther aloft in the upper troposphere (Fig. 19a). This of cooler temperatures throughout the interfront zone plot also suggests that ampli®cation of midlevel (700 (Fig. 18b). As suggested earlier, the depth and strength and 500 hPa) height troughs over the interfront region of the prefrontal cold anomaly should play a role in is a direct result of convective warming above these discrete frontal propagation by determining whether levels [which agrees with Zhang and Harvey (1995) and frontogenetical tendencies can continue unabated at low Stensrud (1996)] and is consistent with the quasigeo- levels. Cross sections of temperature difference taken strophic height tendency equation (Hirschberg and through the center of the surface high pressure anomaly Fritsch 1993). Later, at 2300/5, the same pattern can be

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2074 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

FIG. 16. Time series of frontogenesis terms along the prefrontal trough. Here F1, F2, and F3 are con¯uence frontogenesis, tilting frontogenesis, and diabatic frontogenesis, respectively; and F is the sum of F1, F2, and F3. Contour interval is 4 ϫ 10Ϫ9 KmϪ1 sϪ1.

seen. But, in addition, low heights are evident over a cold dome generated by moist convection. The frontal deep layer at the location of the PFT (Fig. 19b). This trough did not penetrate through the deep stable layer. ®gure shows that a deep layer of height falls occur at Rather, in effect the low-level front discretely moved the prefrontal trough during discrete frontal propaga- into a new, more favorable position for development on tion. Also note that high height anomalies exist at low the south side of the cold dome rather than continuously levels near the location of the original front, showing propagating through the unfavorable intervening area. how the height trough associated with the original front Based on observations and a numerical simulation of ®lls in. the event, a conceptual model is proposed for discrete frontal propagation in a convective environment. Since other cases of discrete frontal propagation will likely 7. Conceptual model differ from this event in minor ways, this model only An example of discrete frontal propagation induced stresses the important processes involved and the basic by convection has been presented. Based on this anal- structural evolution. The ®rst step in the conceptual ysis, discrete frontal propagation can be viewed as a model is the creation of a deep (Ͼ1 km) and long-lived process whereby a midlevel front moves through an area (Ͼ3 h) convectively induced cold dome ahead of an aloft without evidence of a front moving through the approaching cold front. Convection must be persistent, same area at the surface, while the surface frontal prop- so that a deep, coherent mesoscale cold dome can be erties (e.g., pressure trough, wind shift, and thermal gra- created and maintained by individual thunderstorms or dient) dissipate at one location and simultaneously de- thunderstorm complexes. The cold dome must be long- velop at another location. Discrete movement in this lived, since if it is not (i.e., if the cold pool and surface case is initiated when the cold front encounters a deep mesohigh quickly dissipate), the original surface trough

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2075

will simply slow and then move continuously. The cold anomaly must be deep/strong enough to induce a high pressure anomaly at the surface. Low-level winds turn in response to this newly formed mesohigh, creating convergence and, thus, a focused region for new frontogenesis. The cold dome is created ahead of the approaching front. When the front encounters the cold dome, the cross-frontal temperature gradient is reduced, inhibiting horizontal con¯uence frontogenesis. Tilting frontogenesis is also reduced, since ascent above the front must lift higher static stability air than previously. The low- level frontal structure thus weakens. A crucial element of discrete frontal propagation in- volves the creation of a new low-level frontogenetic trough. Initially, diabatic frontogenesis occurs along the boundary between the convectively created cold air and a prefrontal mixed boundary layer. As winds adjust to the new trough, con¯uence frontogenesis intensi®es the thermal gradient. A vertical, thermally direct, ageostrophic circulation is created. Near the completion of discrete movement, the original frontal trough disappears and is incorporated into the postfrontal high pressure system. As the original frontal trough weakens, a mesoscale adjustment of the low-level wind ®eld occurs.

8. Conclusions A numerical simulation is used to investigate the vertical structure of discrete frontal propagation on 5±6 September 1994. The results con®rm BF's observational analysis that the front aloft propagates continuously over a deep, convectively generated cold dome. A new baroclinic zone forms between the cold dome and a well- mixed planetary boundary layer. The new low-level front develops within this baroclinic zone while the original low-level front dissipates. The important role of diabatic processes in discrete frontal propagation has been revealed in this study. A prefrontal trough that becomes a new frontal boundary was initiated by diabatic processes. Once a horizontal temperature gradient was created by differential heating, winds adjusted to the new pressure pattern and induced con¯uence and tilting frontogenesis. This is in contrast to other studies of the initiation of fronts, in which dynamic terms act to intensify preexisting baroclinic temperature gradients. In this case, diabatic processes create the initial gradient. Sensitivity experiments reveal that the front does not propagate discretely without the effects of precipitation and clouds. Differential diabatic processes are identi®ed as strong contributors to frontogenesis here. FIG. 17. Surface analyses from sensitivity experiments at 1800 UTC Further study into discrete frontal propagation should 5 Sep 1994 (6 h into the forecast). Sea level pressure (solid) is con- be carried out. Several dynamic considerations have toured every 1 hPa and temperature (dashed) is contoured every 2ЊC. been noted in this study, but have not been rigorously (a) DRY1, (b) CLR, and (c) DRY2. Shading in (a) represents ver- tically integrated cloud water greater than2gkgϪ1. Shading in (b) investigated. For example, the role of the depth and is accumulated precipitation from 1200 to 1800 UTC. intensity of the prefrontal cold anomaly versus the intensity of the original frontal circulation may be a key

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2076 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

FIG. 19. Cross sections of geopotential height difference (contour interval 5 m) between CTL and DRY2 (i.e., CTL-DRY2) at (a) 1700 and (b) 2300 UTC 5 Sep 1994. The surface locations of the original front and the prefrontal trough in CTL are indicated by the arrows.

←

FIG. 18. Difference ®elds between CTL and DRY2 at 1800 UTC 5 Sep 1994: (a) sea level pressure (contour interval 0.25 hPa) and surface wind vectors (kt), (b) surface temperature (contour interval 1ЊC), and (c) cross section of temperature (contour interval 1ЊC) where the location of the cross section is indicated in (a). Negative contours are dashed. Frontal symbols denote the location of fronts in CTL.

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2077 in differentiating between discretely propagating fronts and continuously propagating fronts that encounter prefrontal thermal anomalies. Idealized variations in the depth, location, and intensity of prefrontal thermal anomalies, similar to the studies of Hoskins et al. (1984), might reveal the sensitivity of discrete frontal propagation to differences in the mesoscale environment. In addition to dynamical studies, a thorough climatological study of discrete frontal propagation in observations could be performed. Also, upstream discrete propagation (i.e., discrete backward movement of cold fronts) as well as discrete propagation of warm fronts should be explored.

Acknowledgments. We would like to thank Jay Char- ney for his assistance during this work, especially for his help with MM5 programs, data display, and his com- ments in interpreting the results. Valuable input was also added by Greg Forbes, Rob Rogers, Jack Kain, and Chuck Doswell. Reviews by Roger Smith and an anon- ymous reviewer strengthened the content and organization of this manuscript. This work was supported by NSF Grant ATM 92-22017. The ®rst author was supported for one year by an American Meteorological So- ciety Graduate Fellowship sponsored by the National Weather Service.

APPENDIX Description of Cross-Section Averaging Technique In cross section analyses of model output, averages of several cross sections oriented perpendicular to the FIG. A1. Schematic representation of cross-section averaging tech- fronts are constructed. The averages eliminate high-fre- nique, where m is the horizontal dimension of the cross section, z is the vertical dimension, and n refers to the individual cross sections quency perturbations (such as those associated with that the average comprises. (a) Horizontal view and (b) vertical view moist parameterized convection at individual grid ele- of one cross section illustrating the three regions: postfrontal (m ϭ ments) but still retain the mesoscale structure and ver- 1tom ϭ F), interfront (m ϭ Ftom ϭ PFT), and pre-PFT (m ϭ tical circulations associated with the fronts. PFT to m ϭ 151). Cross sections are spatially normalized before averaging to account for variations in the separation between the original front and the new front. The normalization ®elds are then added together and averaged. Since the is achieved in the following manner. First, the surface averages are computed on sigma surfaces, the boundary locations of the front (F) and prefrontal trough (PFT) layer structure is retained. An average terrain for the are subjectively determined. The center of the pressure cross sections is computed from the mean terrain of the trough was primarily used to de®ne the locations of the individual cross sections. front and PFT (see, e.g., Fig. 2). For normalization, the postfrontal region (the cold air mass behind the original REFERENCES cold front) is de®ned as the left part of the normalized cross section, the interfront region (the intervening zone Baldwin, D., E.-Y. Hsie, and R. A. Anthes, 1984: Diagnostic studies between the original cold front and the PFT) is de®ned of a two-dimensional simulation of frontogenesis in a moist at- as the middle part of the cross section, and the pre-PFT mosphere. J. Atmos. Sci., 41, 2686±2700. Ballentine, R. J., 1980: A numerical investigation of New England region (the warm sector ahead of the PFT) is de®ned coastal frontogenesis. Mon. Wea. Rev., 108, 1479±1497. as the right part of the cross section. This methodology Barnes, S. L., 1964: A technique for maximizing details in numerical is illustrated in Fig. A1a. For each of the individual weather map analysis. J. Appl. Meteor., 3, 396±409. cross sections that comprise the average, values of se- , 1973: Mesoscale objective map analysis using weighted time series observations. NOAA Tech. Memo. ERL NSSL-62, 60 pp. lected variables (such as temperature, vertical velocity, [Available from National Technical Information Service, Sills etc.) are interpolated to equally spaced points across the Building, 5285 Port Royal Road, Spring®eld, VA 22161.] three regions (Fig. A1b). The individual cross section Berson, F. A., D. G. Reid, and A. J. Troup, 1957: The summer cool

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 2078 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 57

change of south-eastern Australia. I. General behaviour. Tech. merical study of the effects of differential cloud cover on cold Paper No. 8, CSIRO Division of Meteorological Physics, Mel- frontal structure and dynamics. J. Atmos. Sci., 52, 937±964. bourne, 48 pp. [Available from CSIRO Publishing, PO Box 1139, , A. Aksakal, and J. T. McQueen, 1997: The in¯uence of me- Collingwood, Victoria 3006, Australia.] soscale humidity and evapotranspiration ®elds on a model fore- Blackadar, A. K., 1976: Modeling the nocturnal boundary layer. Pre- cast of a cold-frontal squall line. Mon. Wea. Rev., 125, 384± prints, Third Symp. on Atmospheric Turbulence, Raleigh, NC, 409. Amer. Meteor. Soc., 46±49. Kuo, Y.-H., and R. A. Anthes, 1984: Mesoscale budgets of heat and , 1979: Modeling pollutant transfer during daytime convection. moisture in a convective system over the central United States. Preprints, Fourth Symp. on Turbulence, Diffusion, and Air Pol- Mon. Wea. Rev., 112, 1482±1497. lution, Reno, NV, Amer. Meteor. Soc., 443±447. Lewis, J. M., 1975: Tests of the Ogura±Cho model on a prefrontal Bluestein, H. B., 1986: Fronts and jet streaks: A theoretical per- squall line case. Mon. Wea. Rev., 103, 764±778. spective. Mesoscale Meteorology and Forecasting, P. S. Ray, Lin, Y. L., R. D. Farley, and H. D. Orville, 1983: Bulk parameteri- Ed., Amer. Meteor. Soc., 173±215. zation of the snow ®eld in a cloud model. J. Climate Appl. Blumen, W., N. Gamage, R. L. Grossman, M. A. LeMone, and L. T. Meteor., 22, 1065±1092. Miller, 1996: The low-level structure and evolution of a dry Maddox, R. A., 1980: An objective technique for separating mac- arctic front over the central United States. Part I: Mesoscale roscale and mesoscale features in meteorological data. Mon. observations. Mon. Wea. Rev., 124, 1676±1692. Wea. Rev., 108, 1108±1121. Bryan, G. H., and J. M. Fritsch, 2000: Discrete propagation of surface Manning, K. W., and P. L. Haagenson, 1992. Data ingest and fronts in a convective environment: Observations and theory. J. objective analysis for the PSU/NCAR modeling system: Pro- Atmos. Sci., 57, 2041±2060. grams DATAGRID and RAWINS. NCAR Tech. Note Charney, J. J., and J. M. Fritsch, 1999: Discrete frontal propagation 376ϩIA, National Center for Atmospheric Research, 209 pp. in a non-convective environment. Mon. Wea. Rev., 127, 2083± [Available from NCAR, P.O. Box 3000, Boulder, CO 80307- 2101. 3000.] Clarke, R. H., 1961: Mesostructure of dry cold fronts over featureless Miller, J. E., 1948: On the concept of frontogenesis. J. Meteor., 5, terrain. J. Meteor., 12, 715±735. 169±171. Cotton, W. R., M.-S. Lin, R. L. McAnelly, and C. J. Tremback, 1989: Miller, L. J., M. A. LeMone, W. Blumen, R. Grossman, N. Gamage, A composite model of mesoscale convective complexes. Mon. and R. Zamora, 1996: The low-level structure and evolution of Wea. Rev., 117, 765±783. a dry arctic front over the central United States. Part I: Mesoscale Doswell, C. A., 1982: The operational meteorology of convective observations. Mon. Wea. Rev., 124, 1648±1675. weather. Vol. I: Operational mesoanalysis. NOAA Tech. Memo Moore, G. W. K., 1991: Frontogenesis in the presence of surface NWS NSSFC-5, 158 pp. [Available from National Technical heating. J. Atmos. Sci., 48, 63±75. Information Service, Sills Building, 5285 Port Royal Road, Ogura, Y., and D. Portis, 1982: Structure of the cold front observed Spring®eld, VA 22161.] in SESAME-AVE III and its comparison with the Hoskins± Dudhia, J., 1989: Numerical study of convection observed during the Bretherton frontogenesis model. J. Atmos. Sci., 39, 2773± winter monsoon experiment using a mesoscale two-dimensional 2792. model. J. Atmos. Sci., 46, 3077±3107. Orlanski, I., B. Ross, L. Polinsky, and R. Shaginaw, 1985: Advances , 1993: A nonhydrostatic version of the Penn State±NCAR Me- in the theory of atmospheric fronts. Advances in Geophysics, soscale Model: Validation tests and simulation of an Atlantic Vol. 28B, Academic Press, 223±252. Pielke, R. A., and M. Segal, 1986: Mesoscale circulations forced by cyclone and cold front. Mon. Wea. Rev., 117, 1493±1513. differential terrain heating. Mesoscale Meteorology and Fore- Garratt, J. R., 1988: Summertime cold fronts in southeast AustraliaÐ casting, P. S. Ray, Ed., Amer. Meteor. Soc., 516±548. Behavior and low-level structure of main frontal types. Mon. Purdom, J. F. W., 1982: Subjective interpretation of geostationary Wea. Rev., 116, 636±649. satellite data for nowcasting. Nowcasting, K. A. Browning, Ed., , W. L. Physick, R. K. Smith, and A. J. Troup, 1985: The Aus- Academic Press, 149±166. tralian summertime cool change. Part II: Mesoscale aspects. Reeder, M. J., and R. K. Smith, 1992: Australian spring and summer Mon. Wea. Rev., 113, 202±223. cold fronts. Aust. Meteor. Mag., 41, 101±124. Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of Rutledge, S. A., and P.V. Hobbs, 1983: The mesoscale and microscale the ®fth-generation Penn State/NCAR Mesoscale Model (MM5). structure and organization of clouds and precipitation in mid- NCAR Tech. Note NCAR/TN-398 ϩ 1A, 107 pp. [Available latitude cyclones. VIII: A model for the ``seeder-feeder'' process from National Center for Atmospheric Research, P.O. Box 3000, in warm-frontal rainbands. J. Atmos. Sci., 40, 1185±1206. Boulder, CO 80307-3000.] Ryan, B. F., K. F. Wilson, and E. J. Zipser, 1989: Modi®cation of the Hirschberg, P. A., and J. M. Fritsch, 1993: On understanding height thermodynamic structure of the lower troposphere by the evap- tendency. Mon. Wea. Rev., 121, 2646±2661. oration of precipitation ahead of a cold front. Mon. Wea. Rev., Hoskins, B. J., 1982: The mathematical theory of frontogenesis. Annu. 117, 138±153. Rev. Fluid Mech., 14, 131±151. Sanders, F., 1955: An investigation of the structure and dynamics of , E. C. Neto, and H.-R. Cho, 1984: The formation of multiple an intense surface frontal zone. J. Meteor., 12, 542±552. fronts. Quart. J. Roy. Meteor. Soc., 110, 881±896. Segal, M., F. W. Purdom, J. L. Song, R. A. Pielke, and Y. Mahrer, Hsie, E.-Y., R. A. Anthes, and D. Keyser, 1984: Numerical simulation 1986: Evaluation of cloud shading effects on the generation and of frontogenesis in a moist atmosphere. J. Atmos. Sci., 41, 2581± modi®cation of mesoscale circulations. Mon. Wea. Rev., 114, 2594. 1201±1212. Kain, J. S., and J. M. Fritsch, 1990: A one-dimensional entraining/ Shapiro, M. A., 1983: Mesoscale weather systems of the central detraining plume model and its application to convective param- United States. The National STORM Program: Scienti®c and eterization. J. Atmos. Sci., 47, 7284±2802. Technological Bases and Major Objectives, R. A. Anthes, Keyser, D., 1986: Atmospheric fronts: An observational perspective. Ed., University Corporation for Atmospheric Research, 3.1± Mesoscale Meteorology and Forecasting, P. S. Ray, Ed., Amer. 3.77. Meteor. Soc., 216±258. Smith, R. K., M. J. Reeder, N. J. Tapper, and D. R. Christie, 1995: , and R. A. Anthes, 1982: The in¯uence of planetary boundary Central Australian cold fronts. Mon. Wea. Rev., 123, 16±38. layer physics on frontal structure in the Hoskins±Bretherton hor- Stauffer, D. R., and N. L. Seaman, 1994: Multiscale four-dimensional izontal shear model. J. Atmos. Sci., 39, 1783±1802. data assimilation. J. Appl. Meteor., 33, 416±434. Koch, S. E., J. T. McQueen, and V. M. Karyampudi, 1995: A nu- Stensrud, D. J., 1996: Effects of persistent, midlatitude mesoscale

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC 1JULY 2000 BRYAN AND FRITSCH 2079

regions of convection on the large-scale environment during the parison with SESAME-79 data. J. Appl. Meteor., 21, 1594± warm season. J. Atmos. Sci., 53, 3503±3527. 1609. Thorpe, A. J., and C. A. Nash, 1984: Convective and planetary bound- , and J. M. Fritsch, 1986: A case study of the sensitivity of ary layer parameterizations in a diagnostic model of atmospheric numerical simulation of mesoscale convective systems to vary- fronts. Quart. J. Roy. Meteor. Soc., 110, 443±466. ing initial conditions. Mon. Wea. Rev., 114, 2418±2431. Zhang, D.-L., and R. A. Anthes, 1982: A high-resolution model , and R. Harvey, 1995: Enhancement of extratropical cyclogenesis of the planetary boundary layer sensitivity tests and com- by a mesoscale convective system. J. Atmos. Sci., 52, 1107±1127.

Unauthenticated | Downloaded 10/01/21 08:47 AM UTC