4.1. Model versus measurements
In the study area, we first use in situ hourly wind measurements (speed and direction) from three stations (Pula Airport (1), Rijeka (2) and Senj (3)) to validate the simulation results in the fine-grid domain. Table II shows statistics for the employed model, namely root mean square error (RMSE), mean absolute error (MAE) and the index of agreement, d-index, (e.g. Willmott, 1982; Mastura, 2009), while Figure 3 illustrates a comparison between 10 m measured and modelled wind (diagnosed by the MO similarity theory from model fields). Throughout the period, the wind at Pula Airport is simulated satisfactorily, while in Rijeka and Senj, the modelled wind speed was overestimated compared to measured wind speed during the bora event. It is important to note that Rijeka and Senj are situated in very complex topography, so the model overestimation is partly due to the smoothed topography used in our simulation. However, similar wind speed overestimations have occurred in bora wind simulations performed by other mesoscale models: MEMO6 (Klaić et al., 2003) at the same 1 km horizontal grid spacing and RAMS (Gohm and Mayr, 2005; Gohm et al., 2008) at a higher horizontal grid spacing than here. Klaić et al.(2003) reported that, at these sites, measured maximal values are questionable due to the lee-side positions of the two measuring sites. In Figure 3(c), apart from the standard anemometer measurements in Senj, additional special measurements are displayed for the same town. They were performed by the WindMaster ultrasonic Gill anemometer placed at a nearby location, approximately 300 m toward the coast (44.99°N, 14.90°E: Orlić et al., 2005). The additional measured wind speed is 22% higher, on average, than the standard measured one during the examined bora event. For larger bora speeds, deviations between the two measuring sites are even higher—approximately 30–40% (Belušić and Klaić, 2004; Klaić et al., 2009). Modelled wind speed is closer to the associated special measurements (marked by 2 in Table II) than to the standard measured data (d-index_1 in Table II). The special measuring site is probably better suited for comparison with model results because of weaker local influences.
Figure 3. Modelled (grey line) versus measured (black line) surface winds (10 m above the ground) from 0600 UTC (0800 CEST) 28 June, until midnight 30 June 2004 for three stations: Pula Airport, Rijeka and Senj. The positions of the measuring sites are indicated in Figure 1(b). In order to show exact discrepancies between the measured and modelled directions, wind directions spanning a range of 0–90° are sometimes expanded by 360°.
Download figure to PowerPoint
Table I. Stations in the north-eastern Adriatic used in the study. The table shows the type of station, where M = main meteorological station, C = ordinary meteorological station, AQ = air-quality monitoring and R = radiosonde station, and the geographic characteristics (latitude, longitude and height above sea level, ASL). The sites are also shown in Figure 1.
|No.||Station (type of station)||lat||long||ASL (m)||No.||Station (type of station)||lat||long||ASL (m)|
|1||Pula Airport (M)||44°54′||13°55′||63||16||Učka (C)||45°17′||14°12′||1372|
|2||Rijeka (M)||45°20′||14°27′||120||17||Volosko (C)||45°22′||14°19′||46|
|3||Senj (M)||45°0′||14°54′||26||18||Kukuljanovo (C)||45°20′||14°32′||355|
|4||Novigrad (C)||45°20′||13°35′||20||19||Crikvenica (C)||45°10′||14°42′||2|
|5||Poreč (C)||45°13′||13°36′||15||20||Rijeka Airport (M)||45°13′||14°35′||85|
|6||Sveti Ivan na Pučini (C)||45°3′||13°37′||8||21||Malinska (C)||45°07′||14°32′||1|
|7||Rovinj (C)||45°6′||13°38′||20||22||Ponikve (C)||45°04′||14°35′||25|
|8||Pula (C)||44°52′||13°51′||43||23||Krk (C)||45°02′||14°35′||9|
|9||Pazin (M)||45°14′||13°56′||291||24||Rab (C)||44°45′||14°46′||24|
|11||Botonega (C)||45°20′||13°55′||50||26||Rijeka (AQ)||45°19′||14°25′||20|
|12||Cres (C)||44°57′||14°25′||5||27||Krasica (AQ)||45°18′||14°33′||186|
|13||Labin (C)||45°11′||14°4′||316||28||Senj_additional (M)||44°59′||14°54′||2|
|14||Čepić (C)||45°12′||14°9′||30||28||Udine (R)||46°3′||13°18′||94|
|15||Letaj brana (C)||45°16′||14°8′||120||29||Zagreb (R)||45°49′||16°2′||123|
| || || || || ||30||Zadar Airport (R)||44°07′||15°23′||88|
Table II. Some statistical indices—root mean square error (RMSE), mean absolute error (MAE) and the index of agreement (d-index)—for wind speed (WS; m s−1) and wind direction (WD; deg) between the model and measuring sites Rijeka, Pula Airport and Senj. In Senj, the comparison between the basic model simulation and the standard measuring site is marked by 1 and that between the basic model simulation and a special measuring site is marked by 2.
| ||Rijeka||Pula Airport||Senj Standard measuring site 1||Senj Special measuring site 2|
The wind direction is reproduced very well at each of these three stations (Figure 3). More variations in wind direction can be noted for very low wind speeds when most numerical models fail to reproduce wind completely over the very complex terrain (e.g. Baklanov and Grisogono, 2007; Grisogono and Belušić, 2009). At Pula Airport on 30 June, some wind direction discrepancies exist between the measurements and the model, influenced by the convergence zone position above Istria. At all stations, the model successfully predicts the timing of the bora breakthrough near the coast, as well as the SB timing. In contrast to the Pula Airport and Rijeka stations, the bora wind started very suddenly at Senj, reaching moderate bora speeds with the longest duration under the same upstream conditions.
The radiosondes launched from Zadar Airport (Figure 1(a)), the only one at the coast, are shown in Figure 4, allowing a vertical comparison of the model results with measurements. During the bora event, two layers can be observed (visible in both the measurements and the model); the bora wind blew at about 4 m s−1 from the east-northeast (around 65°) in the lowest 1600 m in the stable boundary layer, with west-north-westerly wind above (Figure 4, upper row). The maximum bora wind speed was 7.2 m s−1 at 300 m height, observed only in measurements. Above the bora layer, the westerly wind, which the model reproduced satisfactorily compared to the measurements, increases with height. On the next day (Figure 4, lower row), the potential temperature profile shows a convective boundary layer approximately 2 km deep. Inside the boundary layer, in the lowermost 700 m, the SB developed as a weak north-westerly wind. The wind changes its direction with height (up to 90°) in the next 1000 m. These SB characteristics agree with the SB climatology (Prtenjak and Grisogono, 2007). The wind is west-north-westerly above the boundary layer. The model reproduced the local circulation cell satisfactorily in the somewhat lower boundary layer. The discrepancies in comparisons of the radiosondes from Udine and Zagreb (Figure 1(a)) were very similar to those shown here for Zadar Airport and thus not shown.
Figure 4. Modelled (solid line with circles) versus measured (grey line) vertical profiles of (a), (d) wind speed, (b), (e) wind direction and (c), (f) potential temperature at Zadar Airport from the radiosondes launched on 29 June 2004 at 1200 UTC (top) and 30 June 2004 aSt 1200 UTC (bottom). Position of the measuring site is indicated in Figure 1(a).
Download figure to PowerPoint
We can conclude that despite both model limitations (e.g. model grid spacing or the MO scheme used) and sparse high-quality measurements, the model satisfactorily reproduced the case analysed.
4.2. Near-surface wind field characteristics
The previous section introduced certain model capabilities of the simulation of bora/SB exchange. Here, we continue to analyse and verify model results for the near-surface flow pattern using wind measurements from the 27 stations (Figure 1(b)). Unfortunately, most of the wind data are collected from ordinary meteorological stations (Table I). Since these stations provide only the strength of the wind in three terms (according to the Beaufort scale), discrepancies between measurements and the model may also be due to wind strength conversions into wind speed (m s−1).
Figure 5 shows the 10 m wind field and potential temperature field on 28 June 2004 at 1300 UTC. The measurements and simulation results both show the prevailing relatively weak southerly winds over the larger part of the Istrian peninsula and Kvarner Bay. The simulated wind reached its highest speed within the Great Gate and the Senj Gate due to channelling and downslope airflow in the hinterland. In the study area, certain known mesoscale features (Prtenjak et al., 2006) developed, e.g. convergence zones over the Istrian peninsula and the island of Krk, as well as a clockwise mesoscale eddy inside Rijeka Bay. The Istrian convergence zone is located about 35 km east from the western Istrian coast, where SB speeds were below 4.5 m s−1. Comparing measured and simulated wind (Table III) at 1300 UTC, the WRF model reproduced the wind direction satisfactorily and the wind speed somewhat less well. The surface air potential temperature rises toward the centre of both the Istrian peninsula and the island of Krk, showing SB formation at the coasts (Figure 5(c)). Although the model correctly simulates the coastal surface air temperatures, in general, a slight underestimation of the measured temperature values is noted in the middle of Istria.
Figure 5. (a) 10 m wind vectors (m s−1) and surface air potential temperature in K (numbers near vectors) from meteorological and climatological stations (in Figure 1(b)); (b) modelled WRF wind field; and (c) modelled surface air potential temperature depicted every 0.5 K for 28 June 2004 at 1300 UTC. The wind vectors are given at a horizontal resolution of 4 km, with reference vectors near the upper right-hand corner. The wind speed is depicted by filled areas (grey-scale legend on the right) with a 1 m s−1 interval.
Download figure to PowerPoint
Table III. Some statistical indices for the basic numerical simulation—root mean square error (RMSE), mean absolute error (MAE), the index of agreement (d-index) and the correlation coefficient (r)—between measured and modelled 10 m wind speed (WS; m s−1) and direction (WD; deg) provided by measurements from the 27 stations in Figure 1(b).
|Time||28 June 2004 1300 UTC||29 June 2004 0600 UTC||29 June 2004 1300 UTC||30 June 2004 1300 UTC|
Around 2200 UTC on the same day, the bora wind started to blow along the north-eastern Adriatic coast (not shown), after the cold air outbreak began in the hinterland. However, due to the low-level cold-front transverse, the surface wind field during the bora event varies in time and space. In the early morning on 29 June (Figure 6), the bora blew over nearly the entire area, except for two regions: within the valley between Ćićarija and Risnjak and within Rijeka Bay. Such a wind distribution means that the bora wind did not become fully established at the surface near the high coastal mountains within Rijeka Bay. There, due to flow separation from the leeward slope of Risnjak Mountain, the bora layer with north-easterly winds was lifted above 300 m off the sea. In the lowermost layer, a cold pool existed over Rijeka Bay which was blocked in front of the coastal mountains of Istria (not shown). The whole wind distribution was reproduced very well by the model (Table III), although wind speed was somewhat overestimated (Figure 6). In general, the bora wind varies spatially in strength, with the formation of several bora jets through the mountain passes, north and south of Istria. The northern jet formed (near the towns of Trieste and Koper), which is sited in a topographic incision between the Dinaric and the Julian Alps. The southern jets are associated with mountain gaps in the Dinaric Alps range, the stronger one (around 14 m s−1) through Vratnik Pass and the weaker one through Gornje Jelenje. Over the Istrian peninsula (according to both the measurements and model), a moderate bora dominated, due to passage of the cold front. The maximum modelled bora wind speed of more than 16 m s−1 formed east of the Istrian peninsula, near the island of Cres. Above the island of Krk, a moderate easterly bora veered toward another strong narrow bora jet originating through Vratnik Pass. This jet achieved its maximum strength within the Senj Gate and the eastern coast of the island of Cres. The event can be classified as shallow, since the cross-mountain flow (i.e. the bora layer with north-easterly winds aloft) was restricted to the lowermost 2.5 km of the atmosphere.
Figure 6. (a) 10 m wind from meteorological and climatological stations and (b) from model simulations on 29 June 2004 at 0600 UTC. The wind vectors are given at a horizontal resolution of 4 km, with reference vectors near the upper right-hand corner. The wind speed is depicted by filled areas (grey-scale legend on the right) with a 2 m s−1 interval.
Download figure to PowerPoint
Figure 7 displays the surface wind field at 1000 UTC, around its maximum. The modelled wind at 1000 UTC is only somewhat similar to modelled wind at 0600 UTC. Figure 7 shows very well formed bora jets; the northern one (near Trieste, maximum wind speed ∼14 m s−1), the primary southern stronger (∼15 m s−1) bora jet through the Vratnik Pass with a width of 25 km, and the second southern, weaker (∼13 m s−1) bora jet through Gornje Jelenje with a width of 15 km. Both narrow jets over Kvarner Bay join together, forming one broad bora jet about 50 km wide near the surface, several kilometres downstream of the coast. This merged bora jet stretches from the middle of the peninsula to the island of Lošinj, with its centre above the tip of Istria. This wind distribution agrees very well with the results of other observational and numerical winter severe bora studies (e.g. Jurčec, 1980; Grubišić, 2004; Gohm et al., 2008). The modelled wind distribution reveals the western part of Rijeka Bay, especially around the town of Opatija (station 25 in Figure 1(b), at the foot of Uč ka Mountain) and at the north-western coast of the island of Krk around the town of Malinska (station 21 in Figure 1(b)), as a sheltered area. Since the bora typically weakens somewhat during the daytime due to the evolution of a convective boundary layer over land (Grisogono and Belušić, 2009), the bora decreases its speed (over Istria) causing a bora wake region over the flat western Istrian coast. This area is affected by the convergence zone from the westerly onshore flow and the easterly bora flow. The model shows that the bora wind at the southern part of the jet north of Istria was considerably whirled, supporting the onshore flow formation along the western Istrian coast. In section 4.3, we discuss this feature as the result of the unsteady bora/SB interaction.
Figure 8(a) shows the measured bora wind and the associated potential temperature pattern at 1300 UTC, which agree satisfactorily with the modelled ones (Table III). Modelled wind in Figure 8(b) shows, apart from bora jets (the much weaker Trieste jet and the temporally almost unchanged southern jets), several simultaneous enlarged areas of bora minima (Figure 8(a) and (b)). They are the western Istrian coast, the sheltered areas in Rijeka Bay (the western sides of Rijeka Bay and the island of Krk) and southern part of the island of Rab. Depending on the bora strength, these areas of bora minima vary in space. At the western Istrian coast, the SB develops in the narrow area despite the fact that the bora brings cold and dry continental air and suppresses a daytime temperature rise (Figure 8(a) and (c)). There, the wind direction changes over time from south-west to north-west. Here, the stronger bora wind speed (>6 m s−1) south of Rovinj (station 7 in Figure 1(b)) did not allow daytime SB penetration over the land, maintaining the SB front over the sea (45.4°N, 13.5°E). In Rijeka Bay near Opatija, the low bora speed allows the formation of the weak thermally-induced perturbation, but within the sheltered areas near the town of Malinska and along the narrow area on the western coast of the island of Rab, only a redirection and weakening of the bora wind occurs.
Figure 8. (a), (b), (c) Same as Figure 5 except on 29 June 2004 at 1300 UTC. (d) Satellite image taken on 29 June 2004 at 1238 UTC in the visible spectra by the NOAA16 satellite.
Download figure to PowerPoint
A satellite image (Figure 8(d)) confirms the bora and SB interaction. The satellite image was produced about twenty minutes earlier than ground observations (Figure 8(a)), the modelled WRF field (Figure 8(b)) and surface potential temperature (Figure 8(c)). A wide stratocumulus field that spreads from the Croatian coast inland corresponds well with the lower surface potential temperatures in Figure 8(c). The convergence in the wind field results in air convection, visible as convective cloud development above the western coast of the Istrian peninsula. Cumulus clouds are high enough to produce clearly visible shadows even at an image resolution of 1 km. The cold-front passage (which is outside of Figure 8(d), to the east) produced stronger convective development, seen as dense stratocumulus over the high coastal mountains (Risnjak, Velika Kapela and Velebit). A wide stratocumulus layer with sharp edges is an example of a typical orogenetic cloud induced by bora wind. The plume of higher clouds moves in westerly winds, opposite to the mostly north-eastern direction of the surface WRF wind modelled in Figure 8(b).
At 1600 UTC, in western Istria, the SB reaches a maximum speed of about 6 m s−1 (Figure 9(a)). For the same wind field, the convergence and vorticity are displayed in Figure 9(b) and (c). The moderate grey-filled areas show where convergence (and consequently convection) occurs during the bora and SB interaction (Figure 9(b)). Comparing the values of convergence during the study period above Istria, the highest values were on 29 June. These areas are also associated with significant vorticity conditioned earlier by the bora jets (Figure 9(c)). The cyclonic vorticity (light grey) formed south-west of the Trieste bora jet maximum. At the same time, south of Rovinj, moderate bora winds determined the formation of the anticyclonic vorticity (moderate grey). The penetration of the mature western SB was approximately 20 km east (Figure 9(a) and (b)). Over the north-western coast of Rijeka Bay and the islands of Krk and Rab, onshore flow developed. However, the moderate to strong bora (>6 m s−1) still obstructed SB formation above the rest of the north-eastern Adriatic coast. In the evening, on the western Istrian coast, the SB vanished and a land breeze developed, blowing toward the sea. It coincides with the reinforced bora flow, especially above the tip of Istria (not shown).
Figure 9. (a) Modelled 10 m wind (m s−1), (b) divergence of the near-surface wind field (s−1), and (c) vorticity of the near-surface wind field (s−1) for 29 June 2004 at 1600 UTC. In the divergence field, divergence is depicted as light grey areas and convergence by medium grey colours. The positive (cyclonic) vorticity is light grey and the negative (anticyclonic) vorticity is medium grey.
Download figure to PowerPoint
During the night, the bora was stronger and still blew mostly in the form of jets, with the strongest one in Senj. In the morning on 30 June (Figure 10), the bora weakened and gradually stopped (until 1000 UTC), allowing simultaneous SB development on the north-eastern Adriatic coast. A moderate bora blew only through Vratnik Pass toward the tip of Istria and north of Trieste. In Rijeka Bay, the north-westerly SB formed in the presence of the weak bora limited to the very narrow coastal area. Above Istria, the western coast was under the SB (starting at 0700 UTC), which penetrated ∼10 km over the land. The SB flow diverged there, toward the bora jets, less northward and more southward of Istria, forming an anticlockwise whirl in front of the coast between Novigrad and Rovinj (sites 4 and 7 in Figure 1(b), respectively). This feature in the wind corresponds closely to the swirling in the surface currents detected during the NASCUM project (Cosoli et al., 2008). It seems that the wind distribution during the SB/bora interplay forced this small-scale phenomenon in the surface currents. On the eastern Istrian coast, a south-easterly SB overlapped with the weak bora wind, forming an onshore flow that was twice as strong as the western SB. Over the northern part of the island of Krk, an SB developed as well.
Figure 11 shows the measured and modelled wind and temperature distributions at 1300 UTC on 30 June, which agree quite well (Table III). Still, the model somewhat underestimates the maximum daily temperatures. Mature SBs developed at the coastlines of Istria as well as on the coastlines of the islands. Thus, during the day (Figure 11(a) and (b)), the western and south-easterly Istrian SBs (with a maximum speed of 4.5 m s−1) merged, forming a convergence zone along the peninsula. The convergence zone moved eastwards (with an average speed of 0.5 m s−1), although it was somewhat slower than in reality indicated by measurements. The potential temperature distribution followed these merged SBs (Figure 11(c)). Over the island of Krk, two convergence zones were generated: the weak convergence zone due to the north-westerly SB around Malinska and another above the hilly area along the island (Figure 11(b)). In Rijeka Bay, a weak clockwise eddy developed that is in agreement with Prtenjak et al.(2006). The islands of Cres and Rab also generated weak thermally-induced circulations. Inside Velebit Channel, a weak southerly channelled flow met the weak northerly wind (near the town of Senj). The existence of a convergence zone over the Istrian peninsula was confirmed by all three sources: ground observations (Figure 11(a)), the WRF wind field (Figure 11(b)) and satellite imagery (Figure 11(d)). The image shows convective clouds developed all over the area of convergence in the WRF model: over the central part of the Istrian peninsula and over the slopes of Ćićarija and Velebit Mountains.
Figure 11. (a), (b), (c) Same as in Figure 5 except for 30 June 2004 at 1300 UTC. (d) The satellite image was taken at 1227 UTC in the visible spectra by the NOAA16 satellite.
Download figure to PowerPoint
Figures 8 and 11 allow us to compare the SB on the western Istrian coast during 29 and 30 June, respectively. On 30 June, in the prevailing undisturbed synoptic conditions, the western SB was earlier and generally weaker than on 29 June. The bora mostly limited the SB horizontal extent over the study area, like the mistral in Provence (Bastin et al., 2006). However, the overlap of the weak bora and SB directions, as occurred on the eastern Istrian coast, resulted in a more time-persistent final south-easterly onshore flow that slowed down the Istrian convergence zone displacement eastward.
4.3. Vertical structure
As we said in the previous section, on 29 June, at the western Istrian coast, the onshore flow formed after 0800 UTC in the lowermost 350 m. However, until noon, this flow did not represent the westerly SB alone, and instead there was the superposition of the dominant swirled bora flow north of Istria and a very weak SB wind (see Figure 7). From 1200 UTC, as the bora weakened in the Gulf of Trieste, the impact of the northern bora jet on the onshore westerly (now thermally driven) wind became negligible. Here, this unsteady SB/bora interaction is discussed in more detail.
Modelled vertical cross-sections (A1B1 in Figure 1(b)) of wind, potential temperature and temperature over Istria above Poreč (station number 5 in Figure 1(b)) are presented in Figure 12. The arrows represent wind vectors with along-section horizontal (vh) and vertical (w) wind components. The potential temperature is shown by filled areas, and the horizontal pressure gradient is shown by dashed lines. Along the cross-section A1B1 on 29 June 2004 at 1300 UTC (Figure 12(a)), the SB penetrated over the land only 6 km from the western Istrian coast. The SB extended into the 400 m deep layer and collided with the bora at a well-marked SB front. At the front (that is ∼1.5 km high), the updraught was characterised by a maximum vertical velocity of 0.8 m s−1 and turbulent kinetic energy (TKE) of 0.5 m2 s−2. In the next 1400 m above the SB, the return current and the north-easterly wind coincided, resulting in a higher wind speed than in the SB. Above the lowermost 2 km, a westerly flow dominated in the study area. During the afternoon, at 1600 UTC, as the bora continuously weakened, the only slightly deeper SB (500 m) reached further above Istria, 10 km from the coast (Figure 12(b)). On the next day at 1300 UTC (Figure 12(c)), the SB circulation existed and in a somewhat deeper layer (600 m), with weaker SB maximum speeds than on 29 June, penetrating ∼16 km over the land. Comparing Figure 12(a) and 12(c), the return air current in the undisturbed synoptic conditions was weaker as well, squeezed into the 800 m deep layer between the SB below and the westerly winds above 1800 m.
Figure 12. Vertical cross-sections (A1B1 in Figure 1(b)) of the modelled wind (m s−1), the potential temperature (K) and horizontal pressure gradient (hPa m−1) on 29 June 2004 at (a) 1300 UTC and (b) 1600 UTC, and (c) on 30 June 2004 at 1300 UTC. The arrows represent the wind vectors (vh, w) with the along-section horizontal (vh) and vertical (w) wind components. The wind vectors are given at a horizontal grid spacing of 1 km, with reference vectors near the lower right-hand corner. The potential temperature is depicted by the filled areas (legend on the right) and the horizontal pressure gradient is shown by dashed lines (every 0.000015 hPa m−1). The position of Poreč station (dot 5 in Figure 1(b)) is shown by an arrow. (d) Changes with time during 29 June 2004 (grey) and 30 June 2004 (black) of the potential temperature gradient (K m−1) and divergence (s−1) across the sea-breeze front along A1B1 vertical cross-section.
Download figure to PowerPoint
In order to argue for the stronger afternoon SB during the bora, we evaluated the SB frontogenesis. The leading edge of the SB (the SB front), which affects the wind speed behind, can be described by the frontogenesis function, e.g. (d/dt)(∂Θ/∂x), as in Arritt (1993). The so-called ‘convergence frontogenesis’ is very important for SB front formation or intensification (defined as a product of divergence (div) and the potential temperature gradient (Θ-gradient) along a cross-section = − ∂u/∂x × ∂Θ/∂x)). The Θ-gradient and divergence across the SB front were estimated (along the vertical cross-section A1B1 (Figure 1(b)), and they are shown in Figure 12(d). During the daytime on 29 June, both parameters varied in strength, reaching two maxima: the first one at 0900 UTC and the second one in the afternoon (around 1400 UTC). In the morning, the narrow zone of interaction between the westerly onshore and the easterly bora flows was characterised by a significant Θ-gradient, and it was followed by a strong low-level convergence. Generally, a strong convergence can occur with strong opposing offshore winds and strong onshore winds (Miller et al., 2003). Later, the bora strength within jets redistributed in the area, weakening near the city of Trieste. The onshore western flow (now less influenced by the bora) decreased over time. As the onshore flow decreased, the low-level convergence became weaker as well. Around noon, the easterly bora pushed the onshore flow backward over the sea. At the same time, the Θ-gradient across the front reached its minimum value, which was also smaller than for the pure SB case. After 1200 UTC, with the overall bora weakening, the convection over the land decreased the stable stratification of the air, but less than in undisturbed synoptic conditions (Figure 12). However, despite these temperature characteristics, the afternoon convection activities enhanced the Θ-gradient as well as the low-level convergence (Figure 12(d)). Because of moderate offshore bora flow, the locations of the strongest Θ-gradients and the maximum low-level convergence occurred in the same place, in the very narrow zone. In general, the reduced bora intensity (as the offshore synoptic-scale wind) decreased the width of the SB front, making stronger temperature gradients across it than during an undisturbed SB event. Due to larger temperature gradients, the horizontal pressure gradients (which actually drive the SB) were higher as well. Figure 12(a) and (c) clarify how the pressure gradients depend on the structure of the SB front. In the presence of bora, the orientation of the gravity current head in respect to the x-axis (α) was more in the vertical (e.g. α in Figure 12(a) versus α in Figure 12(c)), and the horizontal pressure gradients (along the SB front) were two times larger than in the pure SB case. The result was a stronger maximum wind speed behind the SB front in the mature SB during the bora. This result agrees with Arritt (1993), Miller et al.(2003) and Bastin et al.(2006). At the same time, inland penetration of the SB front was slower (∼0.35 m s−1) than in undisturbed conditions (∼0.5 m s−1). In undisturbed synoptic conditions, the location of the maximum near-surface wind convergence differs somewhat from the location of the sharp Θ-gradient across the SB front by as much as several km during the day. Therefore, the orientation of the gravity current head was less in the vertical; the horizontal pressure gradients, and consequently, the SB speeds were weaker on 30 June.
Figure 13 shows the A2B2 transect (in Figure 1(b)) above the western part of Rijeka Bay near Opatija, on 29 June at 1300 UTC and 1600 UTC and on 30 June 2004 at 1300 UTC. The figure shows vertical cross-sections of wind, potential temperature and TKE. On 29 June, over the slopes of Risnjak Mountain at 1000 UTC, an event resembling a hydraulic jump occurred, with wake formation near Opatija. Within the wake, a thermally-induced perturbation began to develop (not shown). In the lowermost zone, 200 m deep and 9 km wide, the onshore southerly flow formed as a superposition of upslope wind (due to the mountainous coast) and the weak SB marking the front. The bora wind redirected the onshore flow westward slightly, compared to the climatological wind hodograph for Opatija (Prtenjak and Grisogono, 2007). The bora wind existed in the 700 m above the land, retarding the inland penetration of the opposing thermally-induced wind (Figure 13(a)). The onshore flow (up to ∼6 m s−1) barely penetrated 4 km inland, occupying 13 km horizontally and the first 400 m vertically. Behind the front, that occupies the lowermost 1.5 km, the head of the thermally-induced flow formed with strong upward speed (>1 m s−1). The north-easterly wind lifted up above the southerly onshore flow, suppressing its vertical extent (compare Figure 13(a) and 13(c)). The area of the SB/bora interaction was characterised by large TKE values, especially near the ground (Figure 13(a)). At 1600 UTC (Figure 13(b)), as the bora became weaker, the southerly wind (around ∼3 m s−1) weakened as well, extending over 16 km horizontally, only 2 km further inland than at 1300 UTC. The front was still visible in the meteorological fields and was associated with high TKE values (the highest at the ground), although lower than those at 1300 UTC. The onshore flow also occurred in a slightly deeper layer, 100 m thicker than at 1300 UTC, with a north-easterly wind above. At 1800 UTC, the thermally–induced onshore wind vanished and the bora started to be reinforced. The next day, the onshore flow started to develop earlier, at 0800 UTC. The well-developed onshore flow occurred in the 900 m deep layer and the vertical thermal circulation was closed at about 2 km. The weaker onshore flow (speeds ∼3 m s−1) penetrated over 30 km further inland than at the same time on the previous day (Figure 13(a) and (c)).
Figure 13. Vertical cross-sections of the modelled wind (m s−1), potential temperature (K) and turbulent kinetic energy (m2 s−2) on 29 June 2004 at (a) 1300 UTC and (b) 1600 UTC, and (c) on 30 June 2004 at 1300 UTC along A2B2 (see Figure 1(b)) transecting the western part of Rijeka Bay near Opatija. The arrows represent the wind vectors (vh, w) with the along-section horizontal (vh) and vertical (w) wind components. The wind vectors are given at a horizontal grid spacing of 1 km, with reference vectors near the lower right-hand corner. The potential temperature is depicted by the filled areas (legend on the right) with a 1 K interval and the turbulent kinetic energy is shown by the dashed lines every 0.25 m2 s−2.
Download figure to PowerPoint
The hourly concentrations of surface O3, shown in Figure 14 due to absence of high-quality wind measurements, support the hypothesis of the development of thermally-induced flow above the western part of Rijeka Bay. The concentrations were measured hourly in the greater Rijeka area, at three chosen air-quality monitoring stations (stations 25–27 in Figure 1(b) and Table I). For the pure SLB case on 28 and 30 June (see wind for Rijeka in Figure 3), O3 concentrations mostly followed a diurnal cycle of the local circulation, with winds having maxima in the early afternoon. The bora, on the other hand, is associated with low pollution levels, since it advects pollutants over Rijeka Bay. A considerable decrease in daytime O3 concentrations was observed on 29 June: only 45 µg m−3 of O3 was measured in Rijeka, which was half that in Opatija at the same time. Still, in Opatija, a daily cycle of O3 concentrations exists, in contrast to Rijeka and Krasica, which have a major bora influence. This significant deviation in O3 concentrations, between Rijeka and Krasica on one hand and Opatija on the other, support the conclusion of the evolution of a shallow local circulation above Opatija, which is suppressed by the north-easterly wind.
Figure 14. Hourly averaged air-pollutant O3 concentrations (µg m−3) at three air-pollutant monitoring stations (see Figure 1(b) for locations): Opatija (triangles), Rijeka (solid line) and Krasica (circles) during 28 to 30 June 2004.
Download figure to PowerPoint
Figure 15, similarly to Figure 13, shows vertical cross-sections at 1300 UTC and 1600 UTC on 29 June of the meteorological fields over the next two observed bora wakes, explaining the certain role of the lee-side islands (e.g. Krk and Rab). The first row in Figure 14 corresponds to the A3B3 (see Figure 1(b)) transect above the eastern part of Rijeka Bay near Malinska. The second row belongs to A4B4 (in Figure 1(b)), which transects the island of Rab.
Figure 15. Same as in Figure 13, except on 29 June 2004 at 1300 UTC (left) and 1600 UTC (right) along (a), (b) A3B3 and (c), (d) A4B4 sections in Figure 1(b). The A3B3 section transects the eastern part of Rijeka Bay near Malinska and the A4B4 transect is over the island of Rab.
Download figure to PowerPoint
On 29 June at 1300 UTC, along A3B3 (Figure 15(a)), the bora wind blew along the lee slopes of Velika Kapela Mountain, accelerating toward the island of Krk, where a hydraulic jump occurred. The downslope wind increased its speed to near 15 m s−1 in the foothills of Velika Kapela, and, associated with the hydraulic jump-like feature, the low-level wind speed then decreased to 8 m s−1. The greatest TKE values (up to 3 m2 s−2) in the narrow, vertically-aligned band were associated with the hydraulic jump as well. The bora wind continued to blow in the lowermost 2 km toward the island of Cres. In Figure 15(b), at 1600 UTC, the hydraulic jump in the lee of Velika Kapela led to the formation of a shallow lee-wave rotor vertically, near the western coast of the island of Krk. Such an eddy has a certain resemblance to the type-2 rotor in Hertenstein and Kuettner (2005) and explains the origin of the surface onshore flow toward the island of Krk in Figure 9(a). Still, there are significant differences compared to the type-2 rotors obtained from their idealised 2D study, e.g. an absence of the near-surface jet below the reversed flow and small rotor-associated turbulence (more details in Prtenjak and Belušić (2009)). In the afternoon, the rotor forms at 1500 UTC and lasts roughly 3 hours. Since our model overestimates the bora strength, the modelled rotor formation is probably delayed relative to reality (see Figure 8(a)). Unfortunately, measurements in this narrow zone are highly limited. However, the existence and location of this kind of hydraulic-jump rotor is also somewhat in agreement with the position of rotor clouds observed by Mohorovičić during the nineteenth century (Mohorovičić, 1889; Grubišić and Orlić, 2007). It seems that the presence of the island of Krk favours the formation of lee-wave rotors: both the lee-side mountain wave-induced rotors within Velebit Channel during winter bora events (Belušić et al., 2007; Gohm et al., 2008) and the hydraulic jump-like rotor in the lee side of the island of Krk during summer bora episodes. In contrast to the island of Krk, the lee side of the nearby island of Cres was without significant influence on bora flow (Figure 15(a) and (b)).
On 29 June at 1300 UTC, along A4B4 further to the south in the study area, the north-easterly airflow was only slightly disrupted over the island of Rab, which is characterised by low topography (Figure 15(c)). The reason for this minimal disruption was the convection over land due to the daytime air temperature rise and roughness, which is higher than over the sea. The island acts to slightly increase the bora wind over the upwind half of the island, leading to a weak convergence and upward motion over the island. When the bora blows from the island toward the sea, a weak horizontal divergence and downward motion occur over the edge. This is consistent with the results of studies that examined the influence of surface roughness on wind (e.g. Yu and Wagner, 1975; Yoshikado, 1992; Prtenjak and Grisogono, 2002). At 1500 UTC, the weaker bora finally allowed a delayed SB onset above the western coast of the island. In the following hours (1600 UTC in Figure 15(d)), a weak SB developed (from 10 to 20 km in the first 500 m), and collided with the bora flow above the centre of the island. The SB almost faded completely at 1900 UTC, when the bora was reinforced. On 30 June near Malinska and the island of Rab, a pure SB developed simultaneously, similar to that obtained in Prtenjak et al.(2006).