Evolution and Structure of Severe Cyclonic Bora

EVOLUTION AND STRUCTURE OF SEVERE CYCLONIC BORA:

CONTRAST BETWEEN THE NORTHERN AND SOUTHERN ADRIATIC

Kristian Horvath1, Stjepan Ivatek–Šahdan1, Branka Ivančan-Picek1 and Vanda Grubišić2

1Meteorological and Hydrological Service, Zagreb, Croatia

2Desert Research Institute, Reno, NV, USA ABSTRACT

While statistical analyses and observations show that severe Bora with maximum gusts exceeding 40 m s-1 can occur in all parts of the Adriatic, the Bora research to date has been mainly focused on the dynamics and structure of severe Bora in the northern Adriatic.

Examined to a significantly lesser degree is a less predictable counterpart in the southern

Adriatic, where the Dinaric Alps are higher, broader, and steeper, and where the upwind Bora layer is generally less well defined. This study identifies main differences in the sequence of events leading to the onset of Bora in the northern and southern part of the eastern Adriatic, based on a comparative analysis of the evolution and structure of two typical severe cyclonic

Bora events, one “northern” (7-8 November 1999) and one “southern” (6-7 May 2005) event.

The analysis utilizes airborne, radiosonde, and ground-based observations, as well as the hydrostatic ALADIN/HR and non-hydrostatic COAMPS simulations.

It is shown that the development of a severe Bora in both the northern and southern

Adriatic is dependent on the synoptic setting to create an optimal set of environmental conditions. However, whereas the southern Adriatic Bora is crucially dependent on the presence of a cyclone in the southern Adriatic to give rise to a deep impinging upstream flow that is able to traverse the southern Dinaric Alps and overcome a strong blocking potential there, the cyclonic Bora in the northern Adriatic is less sensitive to the exact position of a cyclone in the Mediterranean. That is true as long as mesoscale conditions are conducive to a strong forcing of the northeasterly low-level jet and pronounced discontinuities in the upstream flow, such as inversions and environmental critical levels, which can significantly modulate Bora behavior. Furthermore, presence of an upper-level jet appears to impede severe Bora development in both the northern and southern Adriatic.

2 1. INTRODUCTION

Bora is a gusty north-easterly downslope wind along the eastern Adriatic that is strongly influenced by local processes and orography. Severity of Bora is well known and exemplified by its gusts, which can exceed 50 m s-1 (Belušić and Bencetić Klaić, 2004). Because of its severity and very frequent occurrence, in particular in the cold season (Bajić, 1989), Bora substantially influences the way of life of local people. As the statistical analysis over a 30- year period (1958-1987) shows, severe Bora with maximum gusts over 40 m s-1 may appear along the entire Adriatic coast, but its duration and frequency decrease from north to south

(Bajić, 1989; Vučetić, 1991). Despite the climatological preference for the northern Adriatic as the locus of strong Boras, events do occur in which Bora reaches higher strength in the southern Adriatic.

Most of the Bora research to date has focused on the severe Bora in the northern Adriatic, in particular that at the locus of the climatological maximum near Senj, at the northern end of the Velebit range, where hydraulic theory appears to capture dynamical essence of the Bora phenomenon (e.g. Smith, 1985; Grubišić, 1989; Ivančan-Picek and Vučetić, 1990; Klaić et al.,

2003; Grubišić, 2004; Ivatek-Šahdan and Tudor, 2004; Göhm et al., 2008). Less studied and less predictable is the Bora in the southern Adriatic, south of the Velebit range, where the

Dinaric Alps are higher, broader, and steeper and where the upwind Bora layer is less well defined. A few studies to date that have explicitly addressed the Bora in the southern Adriatic challenge the traditional view of Bora as a local small-scale phenomenon, and reveal a multi- scale nature of the Bora-related airflow (Jurčec, 1989; Ivančan-Picek and Tutiš, 1996).

The main aim of this study is to identify differences in the sequence of events favorable for the generation of severe cyclonic Bora in the southern and northern Adriatic. It is well known that the synoptic pressure pattern over Europe can be used to distinguish between a cyclonic and an anticyclonic Bora (e.g. Defant, 1951). The cyclonic Bora is dependent on the

3 existence of a low-pressure area over the central Mediterranean, with an especially warm

Sirocco or Jugo1 blowing from the south over the front portion of this low, topping the near- surface Bora layer. Clouds, precipitation, and strong winds over the entire Adriatic Sea basin typically characterize cyclonic Bora. The anticyclonic Bora develops under the influence of a high-pressure system over Central Europe (upstream of the Dinaric Alps) and strong lee-side winds are restricted to the near costal region. Local population has always distinguished “dark

Bora” (cyclonic), with cloudy skies and precipitation, from a cloudless or “clear

Bora” (anticyclonic).

The identification of key differences in the mesoscale structure and evolution of airflow over the northern and southern part of the eastern Adriatic shore in this study is based on a comparative analysis of two severe cyclonic Bora events, one “northern” and one “southern” event. As the knowledge of mesoscale flow features generally leads to increase in the forecasting skill for surface winds, especially in hazardous situations, we pay a particular attention to the generally less explored upstream Bora conditions. The analysis is based largely on the numerical model simulations with the operational ALADIN/HR hydrostatic model run at 8 km horizontal resolution, but the use is also made of the available airborne, radiosonde and ground-based observations. The ability of the ALADIN/HR model to accurately reproduce the observed Bora spatial structure was established using the MAP IOP

15 flight-level data obtained in the northern Adriatic (Ivančan-Picek et al., 2005). A good agreement between the IOP 15 data and the ALADIN/HR model simulation in the northern

Adriatic forms a basis for extending our analysis of the evolution and structure of the Bora flow to the southern Adriatic using the ALADIN/HR model results.

The paper is organized as follows. Section 2 delivers a description of the ALADIN/HR model and numerical simulations. In Section 3, we present synoptic and mesoscale

1 Jugo is a local warm and humid southeasterly wind that belongs to the family of sirocco winds, but is channeled along the Adriatic basin by the Dinaric Alps (e.g. Jurčec et al., 1996).

4 environment during two chosen severe Bora events. A comparison of the model results with observations is given in Section 4. Spatial structure of the upstream, downstream and cross- mountain flow during these Bora events is examined in Section 5. Section 6 concludes the paper.

5 2. ALADIN/HR SIMULATIONS

ALADIN (Air Limitee Adaptation Dynamique development InterNational) is a limited- area model built on top of the IFS/ARPEGE (ARPEGE – Action de Recherche Petite Echelle

Grande Echelle, IFS – Integrated Forecast System) global model, which uses spectral representation of predicted fields (Bubnova et al., 1993). The physical parameterization package includes the vertical diffusion parameterization (Louis et al., 1982), shallow convection (Geleyn, 1987) and a modified Kuo-type deep convection (Bougeault and Geleyn,

1989) schemes, as well as a Kessler-type large-scale precipitation parameterization (Kessler,

1969) for treatment of stratiform cloud processes. The vertical transport of moisture and heat in the soil uses a two-layer parameterization scheme (Giard and Bazile, 2000).

ALADIN was one of the limited-area models used operationally at the MAP Operational

Centre (MOC) during the MAP SOP. At the Meteorological and Hydrological Service of

Croatia, ALADIN/HR is run operationally at 8 km horizontal resolution twice daily at 00 and

12 UTC using a 72-hour integration of the ALADIN full physics package. For simulations in this study, initial and boundary conditions were taken from the IFS/ARPEGE global atmospheric model with horizontal resolution of approximately 40 km over the target area.

The ALADIN/HR domain with the model representation of orography is shown in Fig. 1.

3. SYNOPTIC AND MESOSCALE CONDITIONS

3.1 The “Northern” Bora Event

The “northern” cyclonic Bora event selected for this study is that documented during

MAP IOP 15 (7-8 Nov 1999). An in depth analysis of this case is presented in Bencetić-Klaić

6 et al. (2003) and Grubišić (2004), where the dynamics and structure of the lee-side and cross- mountain flow were investigated using the aircraft and other in situ data.

The synoptic situation during MAP IOP 15 (7-8 Nov 1999), as shown in the ECMWF analysis (Fig. 2), was characterized by an explosive lee cyclogenesis over the Tyrrhenian Sea

(Fig. 2a). The lee cyclogenesis, low-level blocking, and deflection of airflow by the Alps were associated with a primary cyclone north to the Alps and the related cold front. At upper levels, a deep trough extending from the North Sea to the central Mediterranean developed into an upper-level cut-off low (Fig. 2b,d). With a strong anticyclone over north-western Europe, and the cyclone over the Tyrrhenian Sea, the situation was characterized by strong low-level pressure gradients across the Alps as well as the Dinaric Alps (Fig. 2c). The mesoanalysis of the pressure field in such situations has shown a very high correlation between the enhanced pressure gradients across the Dinaric Alps and the Bora strength (e.g. Ivančan-Picek and

Tutiš, 1995). The largest surface pressure difference between Zagreb (continental part of

Croatia) and Senj (at the northern Adriatic coast) was ~12 hPa (~10 hPa / 100 km) during morning hours of Nov 7, leading to the onset of Bora in the northern Adriatic. At that time, strong Jugo was still present over the southern part of the Adriatic. Later that day, Jugo was displaced by Bora that gradually extended to the middle and the southern Adriatic.

Nevertheless, this Bora was significantly stronger in the northern Adriatic (wind gusts of over

45 m s-1) than farther south, where wind gusts were approximately twice weaker than in the north.

3.2 The “Southern” Bora Event

The “southern” Bora event selected for this study is that of 6-7 May 2005. From the outset in the northern Adriatic, this Bora gradually spread southward too, but was significantly

7 stronger in the southern Adriatic, causing a significant amount of infrastructural damage there. The highest values of 10-min average wind speed and wind gusts during this event, of

22 m s-1 and 38 m s-1, respectively, were measured in Makarska in the south, exceeding by a factor of three the respective values at the Krk Bridge site in the north (7 m s-1 and 11 m s-1 at

KB in Fig. 1).

The synoptic situation for period 6-7 May 2005 was characterized by a cyclone development over the central Mediterranean and an anticyclone over the eastern Atlantic and central Europe (Fig. 3). As demonstrated by Večenaj (2005), who examined the synoptic situations in twenty strongest southern Adriatic Bora events in the period 1991–2000 using the ECWMF analyses, this represents a typical case of Bora in the southern Adriatic. At 00

UTC 6 May upper-level trough was situated southwest of the Alps, slowly moving towards the east with increasingly negative meridional tilt of the trough axis (Fig. 3b). At the same time, centre of the surface cyclone was located in the southern Mediterranean moving towards

NE (Fig. 3a). At 18 UTC 6 May 2005, centre of the upper-level cyclone was positioned above the centre of the surface cyclone, which had reached the Ionian Sea. At 00 UTC 7 May 2005, the surface cyclone experienced a rapid deepening above the southern Adriatic and southeastern Europe and the central pressure fell slightly below 996 hPa, as inferred from the

ECMWF analysis (Fig. 3c). This synoptic setting has created a non-uniform pressure gradient across the Dinaric Alps that increased in strength from north to south. As shown by Jurčec and Visković (1994), based on their analysis of the mean pressure distribution at surface and

500 hPa for 15 severe southern Adriatic Bora cases, such a synoptic situation is favorable for generation of severe southern Adriatic Bora, and is linked to the existence of a mesoscale cyclone in the southern Adriatic, which may or may not be evident in large-scale analysis maps.

8 Despite the absence of a primary cyclone north of the Alps and the associated cold front impinging on the Alps (such as in MAP IOP 15 case), there was a strong NW flow present there. A pronounced low-level splitting of this NW flow had produced sheltered conditions in the northern Adriatic. As the northern branch of this split flow turned around the eastern end of the Alps, the flow with a component perpendicular to the southern portion of the Dinaric

Alps developed. While this flow splitting around the Alps bears some resemblance to that in

MAP IOP 15, thermal properties of these flows were quite different, as is discussed farther below. By the end of 7 May, the center of the surface low had moved eastward and was positioned over the Black Sea, resulting in weakening of the pressure gradients over the

Dinaric Alps.

4. COMPARISON OF OBSERVATIONS AND MODEL SIMULATIONS

During these two selected Bora cases, standard surface wind measurements were available at several coastal stations (Krk Bridge, Maslenica Bridge, Hvar and Makarska). To determine the upstream conditions we have utilized radiosonde data from the station Zagreb-Maksimir.

The locations of these measuring sites are shown in Fig. 1. In addition to the routine operational data, we have used the aircraft data collected during MAP IOP 15 for the validation of the operational hydrostatic ALADIN model and the comparison of its performance against the non-hydrostatic COAMPS model (Hodur, 1996).

4.1. Flight-Level Data

The observed flow structure derived from the flight-level data from a coast-parallel track of NCAR Electra on 07 Nov 1999 (15-16 UTC) and that simulated by COAMPS model at the horizontal resolution of 3 km (Grubišić, 2004) are compared here with the results from the

9 ALADIN/HR simulation at the horizontal resolution of 8 km. Figures 4a-b show the magnitude of the 10-min wind speed component normal to the flight track AX (cf. Fig. 1) at

330 m and 660 m a.g.l. as a function of latitude2. The main feature in these diagrams is a strong and wide NE jet between 44.6º and 45.1º N. As discussed in Grubišić (2004), this jet is a downwind extension of the lee-side maximum associated with the strong flow through

Vratnik Pass (near Senj in Fig. 1). Two weaker jets are also visible, to the north and south of the main jet, as well as the wake (weak wind) regions in between the jets, owing their origin, respectively, to other prominent passes (Postojna Pass to the north and Oštarije Pass to the south) and peaks in the coastal range (Grubišić, 2004). The main jet is 5 m s-1 stronger at 660 m than at 330 m and displaced ~10 km to the northwest, revealing the change with height of the exact low-level jet location and its intensity.

The model predictions and flight-level data at these two altitudes show a reasonably good agreement. At 660 m a.g.l. the agreement between the model-predicted and observed flow structures is better than at 330 m. This is true for both the ALADIN/HR and COAMPS models. The two model predictions show slight differences in the predicted spatial flow structure and the degree of agreement with observations. Specifically, while the ALADIN/HR predictions capture the spatial structure of the observed flow at 660 m a.g.l. very well, the extrema are over-predicted, in particular the wake zone centered at 44.5º N. On the other hand, although the fine-scale spatial variation of the Bora flow at 330 m a.g.l seems to be better reproduced by the COAMPS simulation, the horizontal displacement of low-level jets is more pronounced than in the ALADIN/HR results. For the most part, the discrepancies between the observed and simulated wind speed magnitudes are likely related to the local differences between the height of the terrain and its representation in these two models, but the hydrostatic approximation (and its influence on the mountain-induced pressure

2The right orthogonal coordinate frame has positive x-axis aligned with the flight track and directedd toward NW.

10 perturbation) is further constraining the ALADIN/HR model in its ability to reproduce the finer-scale details of the Bora flow.

4.2. Surface Data

Comparisons of the time series of the observed and simulated surface wind direction, wind speed, and gusts (from 1 Hz data) from four coastal weather stations for the selected

“northern” and “southern” Bora events are shown in Fig. 5. The model fields are obtained from the lowest model level at 10 m.

Panels a-d of this figure show the data for the northern Bora event, for the period 00 UTC

7 November to 00 UTC 9 November 1999. The Bora breakthrough, marked by a rapid increase of wind speed in the northern Adriatic, occurred in the early morning hours of 7

November. Throughout that day Bora strengthened and reached its maximum strength around

00 UTC 8 November while gradually spreading southward. An evidence of the southward spread is a time delay (of a few hours) in the onset of Bora between the Krk and Maslenica

Bridge sites (Fig. 5a,b) as well as an even later onset of NE winds at the southern Adriatic sites (Fig. 5c,d). The highest mean hourly wind speeds were measured at the two bridge sites, exceeding 20 m s-1 (with gusts over 25 m s-1) at the Krk Bridge and 25 m s-1 (with gusts over

40 m s-1) at the Maslenica Bridge location. While Bora was quite strong at the Krk Bridge site already in the early morning hours of November 7, a strong Jugo (mean wind over 12 m s-1 with gusts over 20 m s-1), much stronger than the subsequent Bora at the same location, was measured at the same time at the island of Hvar (Fig. 5c). Similarly, in the morning hours of 7

November, Jugo was replaced by moderate Bora winds at the coastal station of Makarska, a favorable site for severe Bora in the southern Adriatic, where wind gusts during this event barely exceeded 20 m s-1– the mean hourly values at the Krk site. The noted simultaneous

11 existence of Bora in the northern Adriatic and Jugo in the southern Adriatic is quite common in the case of surface lows over southern Italy (Jurčec et al, 1996; Brzović, 1999; Pasarić et al., 2007).

The ALADIN/HR model prediction of the mean wind speed agrees well with the observed one except for the Krk Bridge site, where the model appears to under-predict the mean wind speed and has clearly not captured a short-lived cessation of Bora around 12 UTC on Nov 7.

While the discrepancy between the modeled and observed wind speed can be attributed in large part to the height of the measurement site (~50 m a.g.l.), the missing prediction of the short-lived Bora cessation is more likely related to a missed prediction of the weakened intensity of a low-level jet upstream of the coastal mountains (cf. Section 4.3).

In contrast to the MAP IOP 15 Bora, during the “southern” event on 6-7 May 2005, the

10-min wind speed maximum of 22 m s-1 (with gusts reaching up to 40 ms-1) was measured in

Makarska, in the southern Adriatic. This maximum significantly exceeded the maximum 10- min wind measured to the north at the Krk Bridge site (<6 m s-1)(Fig. 5e,f). From the outset in the northern Adriatic, this Bora gradually spread southward, too, but the initial phase of this

Bora was not preceded by Jugo, since there was no cyclone over the Tyrrhenian Sea, which is the usual driver of this wind in the Adriatic. The model simulation captures the onset, timing and cessation of Bora well, although its intensity was underestimated by roughly 50%. There is likely more than one contributing reason for this significant under-prediction. One of them is again the model representation of orography, which in addition to the terrain height reduces its steepness, the latter being the most important mountain shape parameter controlling the strength of downslope windstorms (Miller and Durran, 1991). For example, for the Biokovo part of the Dinaric Alps (Makarska station lying in its foothills), the mountain top height in the ALADIN/HR simulation is 668 m, in contrast to 1762 m in reality. In addition, as shown recently by Horvath and Ivančan-Picek (2008), the predictability of the Mediterranean

12 cyclones and Bora in the southern Adriatic are strongly correlated. As inferred from ECMWF analyses (Fig. 2) and ALADIN/HR predictions (Fig. 5), the cyclone in this event was under- predicted, which led to a weaker simulated background flow and contributed to weaker Bora winds in the model prediction. Another possible reason for this under-prediction is discussed in section 5.1.3.

4.3. Radiosonde Data

In order to determine the accuracy of simulated upwind conditions leading to and throughout the selected Bora events, we have examined the radiosonde data from Zagreb-

Maksimir, located about 130 km upstream of the northern part of the coastal mountain range.

Figure 6 shows a comparison of the observed upstream flow profile in Zagreb and the simulated profiles at the closest model grid point. The interval between the individual soundings at Zagreb was 6 hours during MAP IOP 15 and 12 hours during the May 2005 case.

The appearance of a shallow layer (<1000 m) of cold NE flow at 00 UTC on 7 November

1999 marks the beginning of this Bora event (Fig. 6 a). At the same time, the northern

Adriatic coastal stations also show the onset and gradual increase of Bora flow (cf. Fig. 5). In the subsequent period, the time-height diagram of wind at Zagreb-Maksimir indicates the existence of an environmental critical level, where the wind turns to the mountain-parallel SE.

This type of wind reversal or inversion formation is an important part of the Bora dynamics

(Smith, 1987; Glasnović and Jurčec, 1990). The upstream Bora layer above Zagreb had a depth of 2-4 km and was characterized by strong wind shear (synoptic wind veering to mountain parallel SE) and a maximum of tropospheric wind speeds at 1–1.5 km altitude throughout the mature phase of Bora development (e.g. at 8 November 1999, 06 UTC). The

13 ALADIN/HR model seems to capture these features well (Fig. 6 a), although the very initial phase of low-level jet creation and the later short-lived decrease of the jet intensity at 12 UTC

7 November (with a coincident strong reduction of Bora intensity at Krk Bridge) appear to be forecasted with reduced accuracy.

The vertical wind structure from the Zagreb sounding data and ALADIN/HR simulations during the “southern” Bora event is presented in Fig. 6b. The NE low-level jet, present in the

Zagreb sounding during the “northern” event, is not present in this case. The rather calm near- surface conditions as well as a gradual intensification of the upper-tropospheric NW winds were simulated with reasonable accuracy with some more significant discrepancies already after the end of the Bora event by mid-day of May 7. This whole Bora episode is marked by a strong wind shear in the lowest 4 km that extended throughout the depth of the troposphere toward the end of the event.

In summary, the above verification shows that the ALADIN/HR model captures the onset and cessation of Bora, as well as its spatial structure, sufficiently accurately to aid us in examining the main differences in temporal evolution and spatial structure of severe Bora in the northern and southern Adriatic, as well as environmental factors favorable for its generation.

5. TEMPORAL EVOLUTION AND SPATIAL STRUCTURE OF THE BORA FLOW

The comparative analysis in this section is based on the ALADIN/HR model predictions and available sounding data at Zagreb-Maksimir. As the sounding data from this location is representative of the upstream conditions for the northern part of the coastal mountain range only and a radiosonde site does not exist in the eastern part of Croatia, we utilize the model

14 soundings near town of Vukovar (cf. Fig. 1) as representative for upstream conditions for the southern part of the coastal mountain range.

5.1 The "Northern" Bora Case

5.1.1 Surface Wind and Pressure Gradient Fields

The spatial structure of the ALADIN/HR simulated horizontal wind fields at 10 m a.g.l. on the regional scale during 7 and 8 November is shown in Fig. 7. The left branch of the blocked

NW flow that impinged on the Alps turned from northwesterly to northeasterly over northern

Croatia. It is the crossing over the coastal mountain range of this layer of well-defined NE winds that gave rise to a development of strong Bora in the northern Adriatic. The region of increased surface winds in the upstream environment was confined to a narrow (~50 km wide) zone over central Croatia, including the Zagreb-Maksimir radiosonde site. Surface winds over Vukovar, approximately 400 km farther east, were significantly weaker, especially in a later phase of this Bora event. On the lee-side, Bora winds extended over the Adriatic, downstream of the northern coastal mountains and the Velebit range, during both days of this event. In the initial phase, Bora was met by strong Jugo farther south, creating a shear zone across the central Adriatic. As previously discussed in Section 4.2, this was followed in the late afternoon of November 7 by the onset of Bora in the southern Adriatic, where it was significantly weaker than in the north.

The cross-mountain pressure gradient is typically a very good indicator of the lee-side wind speed (Žagar and Rakovec, 2007). Although the cross-mountain surface pressure gradient in this event was fairly uniform along the Dinaric Alps (~6-8 hPa/100 km; not shown here), there was a clear gradient in the strength of this Bora along the length of the mountain

15 range. As shown in the next section, it is the difference in the upstream conditions for the northern and southern part of the Dinaric Alps that had led to less optimal conditions for the

Bora severity in the southern Adriatic in this case.

5.1.2. Upstream Vertical Flow Structure

Unlike the vertical wind profile over Zagreb (Fig. 6a and Section 4.3), which shows evidence of a strong and sustained NE low-level jet, the lower tropospheric winds over

Vukovar were mostly easterly, shifting between SE and NE during this event (Fig. 8a). Again, in contrast to the Zagreb-Maksimir sounding, a shallow upstream NE flow of only 1 to 2 km in depth was present upstream of the high and wide southern Dinaric Alps in the mature phase of this Bora event.

Looking at the thermal structure of the upstream flow, the Brunt-Väisälä frequency profiles for both Zagreb and Vukovar sounding show high static stability of the low-level tropospheric air (Fig. 9a). The shallow layer of very high stability in the lowest 1 km of the atmosphere, having its origin in shallow inversions, was clearly deeper in the Vukovar sounding, leading to potential for stronger blocking upstream of the southern part of the

Dinaric Alps. On the other hand, an elevated zone of the enhanced stability (due to a thermal inversion) near 2-3 km a.g.l., which after the initial Bora breakthrough gradually strengthened over time, was more pronounced in the Zagreb sounding.

Temporal evolution of the Scorer parameter3 profile at Zagreb clearly shows two distinct phases of this Bora event (Fig. 10a). In the first (initial) phase, through approximately

18 UTC 7 November, the Scorer parameter profile at Zagreb was rather uniform with height.

At subsequent times, the profile is characterized by the presence of a local maximum near 3

3 Scorer parameter was approximated by l2=N2/U2, where N = N(z) is the Brunt–Väisälä frequency and U = U(z) is the vertical profile of the horizontal wind, both upstream of the barrier.

16 km a.g.l., owing its existence to the aforementioned inversion but also to an environmental critical level created by the synoptic-scale advection of the secondary upper-level PV anomaly. This sharp peak in the Scorer parameter defines the top of the Bora layer in the mature phase of this event, and is a good indicator for a potential for layered behavior of the atmosphere, i.e., uncoupling of the lower- and upper-tropospheric flow upstream of the northern part of the Dinaric Alps. In contrast, the Scorer parameter profile for Vukovar does not change significantly over time during this event (Fig. 10b). Additionally, its gradual decrease with height is conducive neither of wave energy trapping at low levels nor amplification of vertically propagating internal gravity waves with height.

5.1.3. Cross-Mountain Flow Structure

Vertical cross-sections of the potential temperature and airflow in the initial and mature stages of this Bora event are shown in Fig. 11. The initial phase is characterized by the lee- side wave activity in both the northern and southern part of the Dinaric Alps. Due to the strong NE low-level jet below 2 km and the synoptically-induced critical level around 4 km in the upstream environment, strong mountain waves and wave-breaking activity is indicated in the north, creating a region of overturned statically unstable air and reversed flow over the northern lee foothills (Fig. 11a). In the south, only weak wave activity is present due to a much weaker NE low-level flow there being in large part blocked by the 1.5 km high mountains, leading to only a shallow layer of the lower atmosphere involved in the mountain wave dynamics and downslope flow (Fig. 11b).

In the mature phase of this event, the northern cross-section reveals the presence of a strong inversion upstream of the coastal mountains, which together with the stagnant overturned region above the lee slope and the mean-flow critical level completely defines the

17 Bora layer (Fig. 11c). The hydraulic character of the flow is evident, resembling that from theoretical studies (e.g. Smith, 1985). In the south, weak NE winds in the lower troposphere were almost completely blocked by the higher southern Dinaric Alps (Fig. 11d). Weak quasi- linear gravity waves continued to be present with no wave breaking occurring over the lee slopes in spite of the presence of a synoptically-induced critical level (at ~3 km a.g.l.) there as well. In addition, the absence of Bora appears to coincide with the presence of the upper-level jet stream aloft, which will be additionally discussed in the next subsection.

5.2 The “Southern” Bora Event

5.2.1 Surface Wind and Pressure Gradient Fields

During the “southern” event, the eastern branch of the Alpine split flow was weaker than in the “northern” case (Fig. 12a,b). Also, the northwesterly direction of surface winds did not change significantly over northern Croatia, resulting in a weak northerly flow upstream of the Dinaric Alps. A stronger surface background flow was generated only in the southern

Adriatic, seemingly aided by the cyclone presence over the southern Adriatic and southeastern

Europe. This synoptic situation produced negligible mean sea-level pressure gradients in the northern Adriatic. In the southern Adriatic, the inland-seaside cross-mountain mean sea-level pressure gradient near Makarska was 5.4 hPa/70 km. As nearly the same pressure gradient of

5.9 hPa/70 km was detected there during the MAP IOP 15 case, in which the Bora strength was almost twice lower than in this case, this putting forward an indication that additional controlling mechanisms exist for the strength of the southern Adriatic Bora.

18 5.1.2. Upstream Vertical Flow Structure

The lower tropospheric winds over Vukovar during the “southern” event show prevalence of northerly and NW directions, the latter being nearly parallel to the coastal mountains farther west (Fig. 8b). The upstream flow was characterized by positive wind speed shear throughout the troposphere, with some degree of wind backing with height from northerly to NW to SW farther aloft. The simulated wind profiles at Vukovar were more uniform than the profiles above Zagreb, leading to a deeper Bora layer upstream of the southern Adriatic (cf. Fig. 6).

The lower troposphere over northern Croatia was less stable in this event due to the lack of the parent cyclone and the associated frontal system north to the Alps. To some degree the weaker stability can also be attributed to a warmer season in which this event occurred

(Fig. 9). Furthermore, there were no strong inversions in the upstream soundings throughout this whole event, although the static stability profile above Vukovar shows an increase of stability in the later phase of this event. Similarly, vertical profiles of the Scorer parameter show more uniform tropospheric conditions, in particular during the later stages of this event

(Fig. 13). As during the initial, weaker phase of this Bora event (6 May, 12 UTC), the Scorer parameter profiles show a peak at ~6 km a.g.l., indicating an existence of a relatively deep

Bora layer, this implies that a certain transition of the Bora behavior might have occurred during the “southern” event as well. However, the nature of this transition is less clear and is definitely opposite to that in the “northern” event, leading to the vertical homogenization of the upstream conditions in the “southern” event. In the next section, we focus only on the part of this event with a well-defined upstream Bora layer.

5.1.3. Cross-Mountain Flow Structure

19 Vertical cross-sections of the potential temperature and airflow in the initial and mature stages of this event are shown in Fig. 14. At 21 UTC 06 May 2005, during the initial phase, northerly low-level flow was present all along the upstream side of the Dinaric Alps. This flow was confined to altitudes below 3 km in the north and 6 km in the south (Fig. 14a,c). In the near upstream region in the north, these northerly winds turned into NE in the lowest 1.5 km, producing a shallow cross-mountain flow at these altitudes. An environmental critical level, where winds turned to NW mountain-parallel direction, was indicated at ~3-4 km a.g.l., at nearly the same height as in the “northern” event (Fig. 14a). However, in contrast to the

MAP IOP 15 case, only weak quasi-linear waves were present over the lee slopes in this case, with no evidence of wave breaking and strong downslope winds. In the south, the synoptically-induced critical level at 6-7 km a.g.l. allowed most of the deep northerly wind layer to pass over the highest mountain peaks and descend towards the coast, changing its direction in turn to the NE, resulting in significant wave activity over the lee slopes. On the upstream slopes, some evidence of low-level blocking is present at the lowest levels in form of NW winds (Fig. 14d). From this, it appears that wave generation and wave steepening leading toward wave breaking are common ingredients of the initial phases of strong Bora events in both the southern and northern Adriatic.

Six hours later, an upper-level jet streak was located above the northern part of the coastal mountain range, contributing to the reduction of static stability of the weak cross- mountain flow there (Fig. 14c). This cross-section reveals presence of weak wave activity only – a development of strong downslope winds has not ensued in the northern Adriatic.

Similar conditions were found in a number of other Bora studies, in which the appearance of the upper-level jet was hypothesized to impede wave breaking (Durran, 2003; Belušić, 2004) or attenuate the existing Bora flow (Vučetić, 1984). In the southern Adriatic, the critical level

20 upstream had descended to 5 km a.g.l., leading to further channeling of the flow over the mountains and intensification of wave breaking over the lee slopes (Fig. 14d). The jet over the

Adriatic farther west is a reflection of mid-tropospheric northwesterly jet that had approached the region. This synoptically-induced wind shear in the lee, together with wave-breaking and overturning over the lee slopes, had created a critical layer that was well defined both in the upstream and downstream environments and might have additionally contributed to Bora severity in the southern Adriatic in this case. In this stage of the event, the simulation results indicate that the Bora flow has a hydraulic character, with a hydraulic jump located over the lower lee slopes. Slight differences in the exact location of the hydraulic jump and the downwind extent of the shooting flow, could potentially lead to large differences in predicted winds at the ground – another possible contributing factor in this case leading to the under- prediction of winds at the coastal sites (cf. Section 4.2).

6. SUMMARY

Two events of strong cyclonic Bora that occurred in different parts of the eastern

Adriatic coast were analyzed to assess the differences in the environmental conditions, spatial flow structure, and sequence of events leading to the appearance of severe Bora in the northern and southern Adriatic.

At the synoptic scale, the examined case of the northern Adriatic Bora (MAP IOP 15) was characterized by the existence of a primary and a secondary cyclone, located over central

Europe and in the Alpine lee, respectively. On mesoscale, strong cross-mountain pressure gradients were established along the entire length of the eastern Adriatic coast. This Bora, which displaced strong Jugo that preceded it over the Adriatic basin, developed into a severe event in the northern Adriatic only. Weaker upstream flow and stronger low-level stability

21 upstream of the higher and wider southern Dinaric Alps led to a strong blocking there, resulting in a much weaker Bora in the southern Adriatic. The main feature of the environmental conditions during the two-day event was the presence of a persistent northeasterly background low-level jet at 1–2 km a.g.l. over the western part of Croatia, in the upstream environment of the northern part of the coastal range. While initially there were no sharp discontinuities in the upstream vertical atmospheric structure, both an inversion and an environmental critical level formed during the course of the event, capping a well-defined upstream Bora layer (at ~3 km a.g.l.) in its mature phase. Highly variable wind speeds at the coastal sites were a hallmark of the outbreak stage, whereas persistent and sustained wind speeds were observed following the establishment of the inversion and the critical level in the upstream environment. This illustrates a transition between the wave-dominated stage of the

Bora during the outbreak, with possibly temporally variable wave breaking over the lee slopes, to a sustained wave breaking and the Bora flow that more clearly displays the hydraulic character.

The synoptic situation for the southern (May 2005) case was characterized by a rapid development of a cyclone over central Mediterranean and southeastern Europe. On mesoscale, cross-mountain pressure gradients in this case increased from north to south. The prevailing northerly and northwesterly winds in the upstream environment at low levels were sheared to the favorable northeasterly direction in the southern Adriatic only due to the proximity of the cyclone there. The northern Adriatic was located in the near wake of the Alps with an upper- level trough and jet streak over the area, leading to generally unfavorable conditions for wave breaking and Bora development. For the southern Adriatic, this synoptic-scale forcing and mesoscale environment produced a deep layer of moist upstream flow that was topped by the environmental critical level at 6 km and that was able to cross the southern Dinaric Alps, resulting in wave-breaking, formation of local critical layers and a severe Bora development

22 there. However, as indicated by the Scorer parameter analysis, the deep upstream layer was dynamically rather uniform with height in the mature phase of this Bora. Nevertheless, the analysis of dry dynamics presented here might prove insufficient for the southern Bora, given the amount of moisture and the fact that moist processes significantly increase the complexity of the flow and lower its predictability (Miglietta and Rotunno, 2005, 2006; Zhang et al.,

2003).

Overall, the development of severe Bora in both the southern and northern Adriatic appears critically dependent on the synoptic setting to create an optimal set of mesoscale conditions, including the cross-mountain pressure gradients and the favorable vertical wind and stability structure in the upstream environment. Our analysis also suggests that subtle differences in environmental conditions, such as presence or absence of pronounced discontinuities in the upstream flow, can potentially lead to significant differences in the Bora behavior, such as in the two phases of the northern event, the wave-dominated outbreak stage and the hydraulic flow in the fully-developed stage. The presence of upper-level jets appears to impede severe Bora development, although exact physical processes associated with that are not yet fully understood. Finally, while not examined in this study, moist processes have a potential to significantly alter the flow over mountains and could significantly impact the severe Bora development, in particular in less predictable southern Adriatic events. We plan to investigate this aspect of severe Bora development in the future, within both idealized simulation and operational prediction frameworks in order to increase the understanding and quality of forecasting of severe Bora winds in the Adriatic basin.

23 Acknowledgement: This study was partly supported by the Croatian Ministry of Science,

Education and Sport through grant 004-1193086-3036. The last author acknowledges support of the US National Science Foundation through grant ATM-0524891 to the Desert

Research Institute.

REFERENCES:

Bajić, A., 1989: Severe Bora on the northern Adriatic. Part I: Statistical analysis. Rasprave – Papers, 24, 1-9.

Belušić, D. and Z. Bencetić Klaić, 2004: Estimation of Bora wind gusts using a limited area model. Tellus, 56A, 296-307.

Belušić, D., 2004: Quasi-periodic bora gusts related to the structure of the troposphere. Q. J. R. Meteorol. Soc., 130, 1103-1121.

Bencetić Klaić, Z., D. Belušić, V. Grubišić, L. Gabela, and L. Ćoso, 2003: Mesoscale airflow structure over the northern Croatian coast during MAP IOP15 – a major bora event. Geofizika, 20, 23-60.

Bougeault, P. and J.-F. Geleyn, 1989: Some problems of closure assumption and scale dependency in the parameterization of moist deep convection for numerical weather prediction. Meteorol. Atmos. Phys., 40, 123-135.

Brzović, N., 1999: Factors affecting the Adriatic cyclone and associated windstorms. Contr. Atmos. Phys., 72, 51-65.

Bubnova, R., A. Horany, and S. Malardel, 1993: International Project ARPEGE/ALADIN. LAM Newsletter, no 22, 117 – 130.

Defant, F., 1951: "Local winds". In Compendium of Meteorology. Ed. T.F.Malone. American Meteorological Society, Boston, USA, Pp. 655-672.

Durran, D., 2003: Downslope winds. In Encyclopaedia of atmospheric sciences. Eds. J. R. Holton, J. A. Curry, and J. A. Pyle. Academic Press, London, UK, Pp. 644-650

Geleyn, J.-F., 1987: Use of a modified Richardson number for parameterizing the effect of shallow convection. Matsuno Z. (ed)., Short and medium range numerical weather prediction, special volume of J.Meteor.Soc.Japan, 141 – 149.

Giard, D. and E. Bazile, 2000: Implementation of a new assimilation scheme for soil and surface variables in a global NWP model. Mon. Wea. Rev., 128, 997-1015.

24 Glasnović, D. and V. Jurčec, 1990: Determination of upstream Bora layer depth. Meteor. Atmos. Phys., 43, 137-144.

Göhm, A., G.J. Mayr, A. Fix, and A. Giez, 2008: On the onset of Bora and the formation of rotors and jumps near a mountain gap. Q. J .R. Meteorol. Soc., 134, 21 – 46.

Grubišić, V., 1989: Application of the hydraulic theory in cases of Bora with strong upstream flow. Rasprave – Papers, 24, 21-27.

Grubišić, V., 2004: Bora-driven potential vorticity banners over the Adriatic. Quart. J. Roy. Metoerol. Soc., 130, 2571-2603.

Hodur, R.M., 1996: The Naval Research Laboratory’s Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS). Mon. Wea. Rev., 125, 1414–1430.

Horvath, K. and B. Ivančan-Picek, 2008: A numerical analysis of a deep Mediterranean lee cyclone: Sensitivity to mesoscale potential vorticity anomalies. Met. Atmos. Phy., In press.

Ivančan-Picek, B. and V. Vučetić, 1990: Bora on the northern Adriatic coast during the ALPEX SOP 20 – 25 March 1982. Rasprave – Papers, 25, 1-12.

Ivančan-Picek, B. and V. Tutiš, 1995: Mesoscale Bora flow and mountain pressure drag. Meteorol. Z., 4, 119-128.

Ivančan-Picek, B. and V. Tutiš, 1996: Severe Adriatic Bora on 28 December 1992. Tellus, 48A, 357-367.

Ivančan-Picek, B., S. Ivatek-Šahdan, and V. Grubišić, 2005: Vertical structure of the Dinaric Alps flow during MAP IOP15. Cro. Met. J., 40, 176-179.

Ivatek-Šahdan, S. and M. Tudor, 2004: Use of high-resolution dynamical adaptation. Meteor. Z., 13, 99-108.

Jurčec, V., 1989: Severe Adriatic Bora storms in relation to synoptic developments. Rasprave – Papers, 24, 11 – 20.

Jurčec, V. and S. Visković, 1994: Mesoscale characteristics of southern Adriatic bora storms. Geofizika, 11, 33-46.

Jurčec, V., Ivančan-Picek, B., Tutiš, V., and Vukičević, V., 1996: Severe Adriatic jugo wind. Meteorol. Z., 5, 67-75.

Kessler, E., 1969: On distribution and continuity of water substance in atmospheric circulations. Met. Mon. Am. Met. Soc., Vol.10, No.32, pp 84.

Klaić, Z., D. Belušić, V. Grubišić, L. Gabela, and L. Ćoso, 2003: Mesoscale airflow structure over the northern Croatian coast during MAP IOP 15 – a major bora event. Geofizika, 20, 23-61.

25 Louis, J.F., M. Tiedke, and J.-F. Geleyn, 1982: A short history of PBL parametrisation at ECMWF. Proceedings from the ECMWF Workshop on Planetary Boundary Layer Parametrisation, 25 – 27 November 1981, 59 – 79.

Miglietta, M.M. and R. Rotunno, 2005: Simulations of moist nearly neutral flow over a ridge. J. Atmos. Sci., 62, 1410–1427.

Miglietta, M.M. and R. Rotunno, 2006: Further results on moist nearly neutral flow over a ridge. J. Atmos. Sci., 63, 2881–2897.

Miller, P. and D. Durran, 1991: On the sensitivity of Downslope Windstorms to the Assymetry of the Mountain Profile. J. Atmos. Sci., 48, 1457-1473.

Pasarić, Z., D. Belušić, and Z. Bencetić Klaić, 2007: Orographic influences on the Adriatic Sirocco wind. Ann. Geophys., 25, 1263 – 1267.

Smith, R.B., 1985: On severe downslope winds. J. Atmos. Sci., 42, 2597-2603.

Smith, R.B., 1987: Aerial observations of the Yugoslavian Bora. J. Atmos. Sci., 44, 269 – 297.

Večenaj, Ž., 2005: Procesi makrorazmjera kod olujnog vjetra u Dalmaciji (Macroscale processes and severe winds in Dalmatia). BSc Thesis, University of Zagreb, pp. 59.

Vučetić V., 1984: On the time and space variations of the northern Adriatic bora. Zbornik meteoroloških I hidroloških radova, 10, 134-137.

Vučetić, V., 1991: Statistical analysis of severe Adriatic Bora. Cro. Met. J., 28, 19 – 36.

Zhang, F., C. Snyder, and R. Rotunno, 2003: Effects of moist convection on mesoscale predictability. J. Atmos. Sci., 60, 1173-1185.

Žagar, M. and J. Rakovec, 2007: Characteristics of the bora wind under different conditions of the lee background flow. 29th International Conference on Alpine Meteorology ICAM 2007, Extended Abstracts, Vol. 2, 309-312.

26 Figure captions:

Figure 1: The ALADIN/HR domain with sites of interest mentioned in the text. Solid black line AX denotes the NCAR Electra flight track and ZG, KB, VU, MA, MB and HV locations of Zagreb, Krk Bridge, Vukovar, Makarska, Maslenica Bridge and Hvar, respectively. Locations of two mountain-perpendicular vertical cross-sections (used in Figs. 11 and 14) are indicated with grey solid lines.

Figure 2: The ECMWF reanalysis for the “northern” Bora event: MSLP and 10-m wind field (left) and geopotential height and wind field at 500 hPa (right) on (a-b) 07 Nov 1999 00 UTC and (c-d) 08 Nov 1999 12 UTC.

Figure 3: The ECMWF reanalysis for the “southern” Bora event: MSLP and 10-m wind field (left) and geopotential height and wind field at 500 hPa (right) on (a-b) 06 May 2005 06 UTC and (c-d) 07 May 2005 00 UTC.

Figure 4: Comparison of ALADIN/HR and COAPMS model predictions and flight data for 10-min wind speed component normal to the flight track AX (cf. Fig. 1) at 330 m (a) and 660 m a.g.l. (b) at 07 Nov 1999 15 UTC.

Figure 5: Time series of surface wind speed (m s-1), wind direction (deg), and gust (m s-1) at: (a) Krk Bridge (wind direction not available), b) Maslenica Bridge, c) Hvar, and d) Makarska during the “northern” event (7-8 Nov 1999), and at: e) Krk and f) Makarska during the “southern” event (6-7 May 2005). ALADIN/HR model prediction of the mean wind speed at 10 m level is shown in green solid line.

Figure 6: Comparison of the vertical wind profiles above Zagreb as inferred from radiosonde data (red arrows) and ALADIN model simulations (blue arrows) for the “northern” (a) and “southern” Bora event (b).

Figure 7: ALADIN/HR model prediction of 10m-wind and surface pressure (hPa, blue solid lines) at: a) 7 November 1999 09 UTC b) 8 November 1999 06 UTC.

Figure 8: ALADIN/HR model simulation of the vertical wind profile at Vukovar on (a) 7-8 November 1999 during the “northern” and (b) 6-8 May 2005 during the “southern” Bora event.

Figure 9: Simulated Brunt-Väisälä frequency for: a) the “northern” Bora event for upstream locations of Zagreb and Vukovar at 7 Nov 1999 09 UTC and 8 Nov 1999 06 UTC; b) the “southern” event for upstream locations of Zagreb and Vukovar at 6 May 2005 21 UTC and 7 May 2005 03 UTC.

Figure 10: Time evolution of the vertical profiles of the Scorer parameter above Zagreb, (a) and Vukovar (b) derived from the ALADIN/HR model simulation of for the “northern” event.

Figure 11: ALADIN-HR model simulated wind speed (m s-1) and potential temperature (K) along the northern (left) and southern (right) vertical cross-sections in the initial phase (a and b) and mature (c and d) phase of the “northern” event.

27 Figure 12: ALADIN/HR model simulation of 10-m wind at: a) 21 UTC 6 May 2005, and b) 03 UTC 7 May 2005.

Figure 13: Time evolution of the vertical profiles of the Scorer parameter above Zagreb, (a) and Vukovar (b) derived from the ALADIN/HR model simulation of the “southern” event.

Figure 14: As in Fig. 11 but for the “southern” Bora event.

28 Figure 1: The ALADIN/HR domain with sites of interest mentioned in the text. Solid black line AX denotes the NCAR Electra flight track and ZG, KB, VU, MA, MB and HV locations of Zagreb, Krk Bridge, Vukovar, Makarska, Maslenica Bridge and Hvar, respectively. Locations of two mountain-perpendicular vertical cross-sections (used in Figs. 11 and 14) are indicated with grey solid lines.

29 Figure 2: The ECMWF reanalysis for the “northern” Bora event: MSLP and 10-m wind field (left) and geopotential height and wind field at 500 hPa (right) on (a-b) 07 Nov 1999 00 UTC and (c-d) 08 Nov 1999 12 UTC.

30 Figure 3: The ECMWF reanalysis for the “southern” Bora event: MSLP and 10-m wind field (left) and geopotential height and wind field at 500 hPa (right) on (a-b) 06 May 2005 06 UTC and (c-d) 07 May 2005 00 UTC.

31 Figure 4: Comparison of ALADIN/HR and COAPMS model predictions and flight data for 10- min wind speed component normal to the flight track AX (cf. Fig. 1) at (a) 330 m, and (b) 660 m a.g.l. at 07 Nov 1999 15 UTC.

32 Figure 5: Time series of surface wind speed (m s-1), wind direction (deg), and gust (m s-1) at: (a) Krk Bridge (wind direction not available), b) Maslenica Bridge, c) Hvar, and d) Makarska during the “northern” event (7-8 Nov 1999), and at: e) Krk and f) Makarska during the “southern” event (6-7 May 2005). ALADIN/HR model prediction of the mean wind speed at 10 m level is shown in green solid line.

33 Figure 6: Comparison of the vertical wind profiles above Zagreb as inferred from radiosonde data (red arrows) and ALADIN model simulations (blue arrows) for (a) the “northern” and (b) “southern” Bora event.

Figure 7: ALADIN/HR model prediction of 10m-wind and surface pressure (hPa, blue solid lines) at: a) 7 November 1999 09 UTC, and b) 8 November 1999 06 UTC.

35 Figure 8: ALADIN/HR model simulation of the vertical wind profile at Vukovar on (a) 7-8 November 1999 during the “northern”, and (b) 6-8 May 2005 during the “southern” Bora event.

36 Figure 9: Simulated Brunt-Väisälä frequency for: a) the “northern” Bora event for upstream locations of Zagreb and Vukovar at 7 Nov 1999 09 UTC and 8 Nov 1999 06 UTC, and b) the “southern” event for upstream locations of Zagreb and Vukovar at 6 May 2005 21 UTC and 7 May 2005 03 UTC.

37 Figure 10: Time evolution of the vertical profiles of the Scorer parameter above (a) Zagreb, and (b) Vukovar derived from the ALADIN/HR model simulation of the “northern” event.

38 Figure 11: ALADIN-HR model simulated wind speed (m s-1) and potential temperature (K) along the northern (left) and southern (right) vertical cross-sections in the initial phase (a and b) and mature (c and d) phase of the “northern” event.

39 Figure 12: ALADIN/HR model simulation of 10-m wind at: a) 21 UTC 6 May 2005, and b) 03 UTC 7 May 2005.

40 Figure 13: Time evolution of the vertical profiles of the Scorer parameter above (a) Zagreb, and (b) Vukovar derived from the ALADIN/HR model simulation of the “southern” event.

41 Figure 14: As in Fig. 11 but for the “southern” Bora event.