The influence of mesoscale flow on the elevated O3 concentrations recorded during 13–19 August 2000 in the coastal town of Rijeka (45.33°N, 14.45°E) has been examined. Although high levels of ozone concentration are often observed there, this episode was unique due to hourly afternoon concentrations persistently higher than 180 µg m−3 and nocturnal levels above 100 µg m−3. To study this episode, available meteorological and ozone measurements were analysed, along with supplemental numerical simulations and trajectories. The measurements revealed favourable atmospheric conditions for the production of ozone, i.e., high air temperatures associated with (1) the exchange of two dominant (although relatively weak) large-scale flows: southwesterly versus northeasterly bora winds, and (2) local thermal circulation. The bora wind was relatively weak and short lasting. Although the bora transported the polluted air mass away along the western Adriatic coast, the southwesterly flow returned ozone-rich air toward the Alps, thus contributing to the overall pollution over the northern Adriatic Sea. Models revealed the superposition of the southwesterly wind and local, thermally induced winds that caused the regional transport of ozone toward the Alps and the eastern Adriatic coast. However, while the regional transport of ozone from the northern Adriatic was found to enhance somewhat the ozone concentrations in Rijeka at the end of period studied, its surrounding local emission sources, uniform local thermal circulation systems, and recirculation of pollutants were crucial factors in the formation of the large daytime and night time ozone levels observed in Rijeka.
During the last few decades, many papers have addressed episodes of high ozone/photochemical pollution in areas along the Mediterranean, including Greece (e.g. Lagouvardos et al., 1996; Ziomas, 1998; Moussiopoulos et al., 2006; Im et al., 2011), Spain (e.g. Millan et al., 2000, 2002; Adame et al., 2010), France (e.g. Bastin et al., 2006), Italy (e.g. Fortezza et al., 1995; Mangia et al., 2010), Tunisia (e.g. Bouchlaghem et al., 2007) and Slovenia (e.g. Žabkar et al., 2008, 2011). These studies have revealed some specific meteorological conditions on both the local and synoptic scales that affect the air quality in southern Europe. Affected mainly by the complex non-linear interactions of nitrogen oxides (NOx) and volatile organic compounds (VOCs), levels of ground level ozone (as a secondary pollutant) were particularly high during the summer when intense solar radiation enhances photochemical activity. Very high insolation levels were associated with anticyclonic situations and weak pressure gradients, which enabled the development of local winds, particularly sea/land breezes (SLBs), and slope winds over the complex Mediterranean coast. These studies also showed that breeze circulations were often accompanied by severe pollution. These circulations sometimes contribute to the (re)distribution of pollutants, such as in Busan, Korea (Oh et al., 2006); Perth, Australia (Leslie and Speer, 1998; Ma and Lyons, 2003) and the USA (Pack and Angell, 1963; Darby et al., 2007). European urban monitoring networks and experimental campaigns have revealed that on the large scale, polluted air masses that were enriched by ozone over the Mediterranean were mostly advected in central Europe (Millan et al., 2002; Lindskog et al., 2007) by long-range transport (Carvalho et al., 2010). Large-scale meteorological processes can also create conditions favourable to the stratospheric intrusion of ozone (e.g., Klaić et al., 2003) that can result in a local increase in tropospheric ozone concentrations. While daytime ozone production is dominated by photochemical reactions, an increase in night time ozone levels is solely the result of specific wind circulation systems on local and regional scales (Strassburger and Kuttler, 1998). The role of regional transport in photochemical pollution (specifically, nitrogen dioxide and ozone pollution) was recently corroborated by Finardi et al. (2011). They showed that Po Valley emissions, and specifically NO2 emissions, affect the regional air quality at distances of 500 km or more, thus stretching over the entire Adriatic Sea and the western portion of the Balkan Peninsula.
One of the conditions necessary for recirculation of pollutants in complex terrain is the significantly inhibited rotation of the local diurnal winds. Next, as a result of updrafts at the sea breeze (SB) front, pollutants may be lifted, vertically, hundreds of metres and advected in many different directions (Millan et al., 2002). Some pollutants may recirculate vertically to their coastline of origin during the daytime by the return current in the SB circulation pattern, or, horizontally, during the night time by the land breeze. For example, in Athens (the eastern Mediterranean), the seaward transport of NOx and hydrocarbons by the land breeze (LB) is followed by the daytime return of photochemically produced ozone over the city. The result is a considerable increase in the daytime local ozone concentration (at least by a factor of 3) within several hours (Ziomas, 1998; Grossi et al., 2000). During the day, over the central part of the Iberian Peninsula (western Mediterranean), a large-scale deep convection occurs together with compensatory subsidence over the surrounding seas, which allows for the vertical recirculation of ozone (Millan et al., 2002). As a result, polluted coastal air penetrates more than 80 km inland. The same is observed in experimental data. Over the northwestern Adriatic (central Mediterranean), the diurnal ozone variation is strongly affected by intense breeze circulations. Fortezza et al. (1995) have observed that coastal ozone concentrations are particularly high in the evening (from 1800 to 2200 h local time), due to the horizontal recirculation of oxidants between the land and the sea.
Until 1999, research on air quality along the northeastern Adriatic (Figure 1) was based only on campaigns of sporadic ozone measurement (Cvitaš et al., 1997), which were characterized by low spatial resolutions and relatively short durations. These limited ozone measurements were monitored at two stations that were established in the summer of 1990: one in Istria and the other in Rijeka Bay. The authors of this study concluded that ozone concentrations decreased southward and pointed to the northern Adriatic Sea as the potential source of ozone. Since 1999, Rijeka (45.33°N, 14.45°E) is the only site that has had continuous ozone monitoring in Rijeka Bay, as it is the most polluted coastal town in Croatia. Measurements were originally collected in the city centre, which was followed by an extension to five more sampling sites from 2003 onward. A declining trend in ozone levels, as also observed by Ainslie and Steyn (2007), is preserved over the whole monitoring period (1999–2010), with occasionally high concentrations at/exceeding the national daily limit value during summer time. For example, in 2009 (from 28 July to 3 August), daily ozone concentrations increased to 140 µg m−3, exceeding the daily limit value of 110 µg m−3 (http://www.zzjzpgz.hr/zrak/index.php) according to national standards. However, one notable episode occurred between 13 and 19 August 2000 (Figure 2). During this period both daily and hourly ozone concentrations were close to or exceeded the daily limit value and hourly warning level, respectively. During the night time, the hourly ozone concentrations were unusually high (100 µg m−3) as well. Conversely, NOx levels were not significantly higher than usual. Alebić-Juretić (2012) also showed that for Rijeka during the 6 year period, the highest ozone concentrations correspond to southwestern winds. A recent study of 4.5 year's of data (Jelić and Klaić, 2010) also showed similar ozone concentration patterns. In a recent study of the coastal area north of Rijeka (Žabkar et al., 2008), observed ozone levels exceeded threshold values up to 25 days every summer. The authors argued that local emission sources do not explain the measured concentration levels and that the high concentrations were associated with slow-moving air masses approaching from the southwest. They concluded that the northern Adriatic Sea was rather unexpectedly a common origin of ozone-rich air but did not notice the significant precursor emissions over the sea. They also concluded that this ozone phenomenon of the northern Adriatic Sea needed to be investigated in more detail. The latter, modelling, study (Žabkar et al., 2011) suggested accumulations of ozone-rich air over the Adriatic Sea, which may last for days, resulting in a pollution reservoir and consequent ozone episodes over Slovenia.
Millan et al. (2002) pointed out that the atmospheric processes responsible for the formation and transport of ozone differ significantly across the Mediterranean Basin. For example, in the investigation of air quality resulting from the interaction between the onshore thermal breeze and the cold down slope wind, Bastin et al. (2006) found that the combined sea breeze/mistral event in France was associated with high ozone concentrations in the densely inhabited coastal area. Conversely, during similar wind conditions (sea breeze and moderate bora) above Rijeka Bay, hourly ozone concentrations did not show any significant deviation from normal monthly averages (Prtenjak et al., 2010). As generalizations of known results for unexamined topographic areas and events are always questionable, the points of investigation for this study are threefold: (1) the existence of ozone-rich air above the northern Adriatic, west of Istria; (2) the high ozone levels in Rijeka that were observed between 13 and 19 August 2000, and, finally, (3) the possible connection between these two points, are examined in detail. To these ends, available measurements and two models, a standard (50 km) EMEP model (Simpson et al., 2003) and a mesoscale WRF model (Skamarock et al., 2008) were applied. A trajectory-based method was also used to show possible ozone transport into the region of interest.
2. Specifications of the studied area
More than the rest of Mediterranean region, the northeastern Adriatic in Croatia (Figure 1(a)) has a notably complex coastline. This area is characterized by the large Istria Peninsula and Kvarner Bay, as well as by the nearby mountain ranges parallel to the coast (Učka and Risnjak). Within the larger Kvarner Bay there are several islands, including Cres and Krk, while the smaller Rijeka Bay faces the urban area of Rijeka. Rijeka is the main industrial town in the region and is located on the northeast coast of Rijeka Bay on the steep slopes of Risnjak Mountain. The major anthropogenic individual sources of pollutants near the town of Rijeka in the range of 8–25 km to the southeast (among others, an oil refinery and a thermal power plant; see squares in Figure 1(b)) as well as on the nearby island of Krk (a petrochemical plant and an oil terminal; Figure 1(b)). Other possible sources of pollutants are a thermal power plant and a cement factory on the eastern coast of Istria, about 40 km southwest of Rijeka, and sources in Trieste Bay, approximately 60 km northwest of Rijeka. Finally, since the analysed episode occurred during the summer, biogenic emissions could also be important.
During the warm season, weak, large-scale pressure gradients allow the formation of local thermal circulations along the northeast Adriatic coast on approximately half of all summer days (Prtenjak and Grisogono, 2007). Recent observational (Prtenjak and Grisogono, 2007) and numerical modelling studies (Nitis et al., 2005; Prtenjak et al., 2006) have demonstrated several low-level flow formations, mostly the result of local thermal circulations developed in the area, which include: (1) the almost periodic, mesoscale daytime and night time eddies inside Rijeka Bay that last several hours; (2) convergence zones above Istria and over the nearby island(s), and (3) a channelling effect on the near-surface flow between the mainland and adjacent islands (Cres and Krk). The last of these contributes to the formation of anticyclonic daytime eddies within the Rijeka Bay area. Conversely, night time cyclonic eddies develop there, due to katabatic flow from the surrounding mountains. Rijeka Bay and the town of Rijeka are characterized by low wind speeds (Bajić et al., 2007; Prtenjak and Grisogono, 2007). Under these light wind conditions, airborne pollutants are not easily ventilated out of the bay, due to specific surrounding topography (Prtenjak et al., 2009).
3.1. Meteorological model
In this study, meteorological measurements were supplemented by results of the research mesoscale version of the Weather and Research Forecasting (WRF-ARW 3.1) model (Skamarock et al., 2008). This WRF-ARW model has already been tested and applied in investigations of different atmospheric phenomena above the Adriatic (e.g., Belušić and Strelec Mahović, 2009; Prtenjak et al., 2009). A two-way nested configuration using three of the model's domains (from the largest (A) to the smallest (C), in Figure 1(a)) was employed using a Lambert conformal projection and 27, 9 and 3 km grid spacings, respectively. The vertical grid is divided into 75 terrain-following hydrostatic pressure co-ordinate levels gradually increasing from 50 m near the surface to 350 m near the top of the model, which is located at 50 hPa. Initial and boundary conditions were updated every 6 h with analysed data from the European Centre for Medium-Range Weather Forecasts (ECMWF) at a 0.25° resolution. Topography and land-use data were taken at 30″ resolution from USGS 24 category data. The physical schemes used in these studies are: a Mellor-Yamada-Janjic scheme for the planetary boundary layer; a Rapid Radiative Transfer Model for long wave radiation and a Dudhia scheme for short wave radiation; a single-moment, three class microphysics scheme with ice and snow processes; an Eta surface layer scheme based on Monin–Obukhov theory, and a five-layer thermal diffusion scheme for soil temperatures.
3.2. EMEP model
An Eulerian photochemical grid model (the Unified EMEP model (http://www.emep.int/)) was used to calculate ozone concentrations. This model is fully documented (Simpson et al., 2003; Fagerli et al., 2004) and widely validated (e.g. Jeričević et al., 2009). It simulates the long-range transport and deposition of air pollutants, including photo-oxidants and particulate matter. The chemical scheme is identical to that of the EMEP Unified Model (Simpson et al., 2003). The chemical mechanism used is based upon ozone chemistry from the Lagrangian photo-oxidant model (Andersson-Sköld and Simpson, 1999) but with additional reactions introduced to extend the model's capabilities to include acidification and eutrophication.
A PARalell version of the HIRLAM (HIgh Resolution Limited Area Model) using a Polar-Stereographic map projection (PARLAM-PS) model (Tsyro and Støren, 1999) was used as the meteorological driver. It has a horizontal resolution of 0.5° (50 km) with 170 × 133 grid points. Numerical solutions for the advection terms are based on the scheme by Bott (1989). A fourth order scheme is used in the horizontal direction and, in the vertical direction, a second order version applicable to variable grid distances is employed.
The emission input required by the EMEP model consists of gridded annual national emissions of SO2, NOx, NH3, NMVOC, CO and particulates. The monthly and daily distribution of emissions is provided in the model according to time factors for emissions of the seven compounds (CO, NH3, NOx, PM2.5, PMco, SOx and VOC). Emissions of isoprene, probably the most important biogenic compound with regard to ozone formation, are calculated every hour in the EMEP model using the model's radiation, temperature and land-use data. The emission rates used in the EMEP model are based upon the algorithms of Guenther et al. (1993), and take into account recent measurements and evaluations from both the USA and Europe (Guenther et al., 1993; Simpson et al., 1995; Seufert et al., 1997).
4. Meteorological conditions during the case
4.1. Synoptic event overview
The period simulated was from 12 to 19 August 2000. Surface diagnostic charts (some are shown in Figure 3), upper-level charts at 700 hPa, and operational radiosonde data in Udine (46.03°N, 13.18°E) were used in order to investigate synoptic conditions in detail. On 12 and 13 August, a ridge of the Azores High reached the central Mediterranean, forming a field with almost no gradient surface pressure (Figure 3(a)). Along the northern Adriatic, the large scale winds were weak and of variable direction, while the northwesterly flow known as Etesians (e.g. Klaić et al., 2009) dominated over the middle and southern Adriatic. According to the radiosoundings at Udine, in the lowermost 3000 m of the atmosphere, the wind was southwesterly and rarely above 6 m s−1 (not shown). During 14–15 August 2000, a shallow cold front passed over the southeastern part of Alps and brought cold air (Figure 3(b)). This cold air outbreak over inland Croatia resulted in a short-lasting, light, northeasterly wind (bora) along the east Adriatic coast. On 16 August, the surface pressure conditions returned to those with weak pressure gradients (Figure 3(c) and (d)). Although a prominent, elongated ridge stretching in a NE–SW direction over southeastern Europe is created, very close to northern Italy, a shallow and small low-surface-pressure disturbance formed. The wind flow over the northeast Adriatic was southwesterly, and the wind speed below the 3000 m level was mostly below 10 m s−1.
To summarize the effects of spatial and temporal variations of the large-scale pressure field, over the entire period studied, the mostly weak, synoptic-scale wind dominated above the northeast Adriatic coast. Such stagnant synoptic conditions greatly facilitated the occurrence of photochemical pollution episodes (e.g., Ainslie and Steyn, 2007). Furthermore, the prevailing large-scale southwesterly winds above the investigated region favoured the high ozone levels in the northeastern Mediterranean Basin, as observed by Žabkar et al. (2008, 2011).
4.2. Near surface wind field characteristics above the northern Adriatic
On 13 August, the pressure field over the area of interest allowed the development of mesoscale thermally induced flows (Figure 4). These comprise the onshore, sea breeze, and upslope flows. Over the northeast Adriatic coast, modelled (Figure 4(b)) winds altered between northwesterly and westerly breezes with significant wind direction deviations between the islands. The northern Adriatic wind speed field was characterized by an almost dipole-like behaviour: eastward, the largest breeze speeds (up to 6 m s−1) occurred near the tip of Istria, while westward, weak winds (below 1 m s−1) formed over the Adriatic near the Po River Delta. This wind field agreed reasonably well with the available measurements (Figure 4(a)) and with modelled flows obtained by Klaić et al. (2009). Such airflow pattern favoured the persistence and accumulation of the air pollutants over the Northern Adriatic (Žabkar et al., 2011). Over the western Istrian coast, northwesterly flow dominated within the lowermost 700 m of atmosphere, and it penetrated up to about 30 km inland. Similar to the results in Prtenjak et al. (2006), a shallow, daytime, clockwise eddy in the lowermost 300 m layer was formed above Rijeka Bay.
Over the land, areas with strong divergence in wind field (Figure 4(b)) indicate convective activity in the domain. Such divergences could develop in the regions of the downdrafts under strong Cb clouds. The same is corroborated by (1) the satellite image (Figure 4(c)) showing convective clouds; (2) large values of convective indices based on the radiosonding data in Udine, and (3) surface observations of short, early afternoon lightning and thunder over Istria and Rijeka. A deep moist convection could decrease the land–sea temperature difference over Istria and consequently onshore wind strength (Babić et al., 2012).
The bora wind occurred over the Adriatic during 14–15 August (Figure 5). It exhibited several well-known jets (north of Istria near Trieste over Kvarner Bay and southward along the east Adriatic coast) and wakes (e.g., Belušić and Klaić, 2006; Grisogono and Belušić, 2009). Along the western Adriatic coast, bora jets whirled into northwesterly flow, while along the eastern coast sporadic breeze/bora interactions were seen similar to those found by Prtenjak et al. (2010). Although the model overestimated bora strength, it still reproduced typical bora behaviour.
On 16 August, the synoptic forcing is weaker compared to that on 14 and 15 August. The wind field pattern became similar to those for 13 August. Still, the onshore flow was somewhat stronger. Over the northeast Adriatic coast, the daytime southwesterly synoptic wind was the strongest on 17 August 2000 (Figure 6(a) and (b)). The inland penetration occurred within the first 1000 m of the atmosphere, reaching above Istria up to 45 km onshore. By the end of the studied period, the large-scale wind over Istria had weakened slightly and inland penetration had lowered. There was no significant convective activity which would interfere with the penetration of the SB.
During the night time, the low-level winds were weak over the entire domain (Figure 6(c)). The mountainous area of the northern Apennine Peninsula was affected by downslope winds. Still, due to a shallow low pressure system over northern Italy, a southeasterly onshore wind was observed above the wide Po River Valley. This part of the region was exposed to the constant southeasterly onshore wind without a reversal in the wind direction between daytime and night time. Over the Adriatic, a large, almost calm area existed, while along the northeastern Adriatic coast, northwesterly wind dominated. The wind was channelled by the valley between the Ćićarija and Risnjak Mountains, thus near Rijeka exceeding a speed of 5 m s−1. Simultaneously, a weak land breeze blew with maximum strength over the southern tip of the Istria Peninsula (Figure 6(c)). An anticlockwise eddy at around 600 m appeared above Rijeka Bay, similar to that found by Prtenjak et al. (2006).
4.3. Surface meteorological measurements in Rijeka
The set of surface wind data in Rijeka provided an insight into the local temporal variation of the wind speed and direction (Figure 7(a) and (b)), enabling simultaneous assessment of the performance of the model. From 14 August at 1500 CET (Central European Time = UTC + 1 h) to 16 August at 0700 CET, the measured weak bora wind ranged from 4 to 6 m s−1. During the rest of the period, the southwesterly daytime landward flow altered the offshore night time northeasterly flow, which hardly exceeded both measured and modelled speeds of 3 m s−1. A transition from night time to daytime was marked by a southeasterly flow lasting for up to 3 h. It was followed by the fully formed, almost perpendicular onshore flow from the southwest, which lasted from 1000 to 1900 CET, on average. The daytime breeze is in agreement with the typical local circulation regime in Rijeka (Pandžić and Likso, 2005; Prtenjak and Grisogono, 2007). Finally, by the end of the period, maximum air temperatures continuously increased from 30 to 36 °C. While during 13 and 14 August clouds covered the sky almost completely, high air temperatures at latter days were associated with low hourly cloudiness (less than 2/10).
Table 1 shows statistical indices in Rijeka for both empirical measurements and the meteorological model at a 3 km resolution using the closest land grid point in corresponding time. The quantities shown are as follows: mean value, mean bias, mean absolute error (MAE), root mean square error (RMSE), and the index of agreement (e.g., Willmott, 1982; Mahmud, 2009). The WRF model simulated wind directions relatively well (Figure 9(b) and Table 1), while the wind speed in stable atmospheric conditions was sometimes overestimated (e.g., at night and during the bora wind, Figure 9(a)). Bearing in mind the domination of weak wind speeds, which are generally hard to model (e.g. Baklanov and Grisogono, 2007), modelled and measured values agree reasonably well (Table 1). Discrepancies between the modelled and measured winds are similar to those found in other mesoscale studies (Bao et al., 2008; Jiménez et al., 2008; Mangia et al., 2010). The largest discrepancies for the temperature field were found in daily minimum and maximum values which were overestimated and underestimated, respectively. Summarizing, the ability of the WRF model to simulate the mesoscale atmospheric conditions was adequate.
Table 1. Statistical indices (mean value, mean bias, root mean square error (RMSE), mean absolute error (MAE), the index of agreement, and correlation co-efficient) for meteorological variables, including wind speed (WS; m s−1), wind direction (WD; °) and temperature ( °C) and for ozone concentrations (µg m−3) between the models (WRF and EMEP) and measuring (M) site in Rijeka (45.33°N, 14.45°E) for the studied period: 13–19 August 2000
Ozone levels (µg m−3)
Mean value (M)
Mean value (WRF)
Index of agreement
5. Ozone levels during the measured time period
5.1. Surface ozone measurements in Rijeka
As seen in Figure 2, very high ozone levels (much higher than the mean value of daily concentrations for August 2000, 88 µg m−3) in Rijeka went along with the meteorological conditions discussed above. The period with high ozone levels may be divided into two parts (Figure 8), before and after 16 August. The main differences between these two time intervals are seen in the different patterns of diurnal variations prior and after 16 August, and in the different daytime ozone concentration values. The hourly ozone concentrations in the second time interval were higher than in the first one. On 13 August, the maximum ozone level occurred in the late morning, at 1100 CET, with a significant decrease over the next several hours. To a certain extent, this can be attributed to significant cloudiness (∼9/10, Figure 4(c)), with thunder and lightning recorded above Rijeka. During the following days, weak and sporadic (so called ‘dark’) bora (again, with ∼9/10 cloudiness) occurred. Consequently, the afternoon maximum ozone concentrations on 14–15 August slightly decreased. From 16 August onward, the diurnal cycle of ozone levels represented the typical pattern of a polluted atmosphere (Simpson, 1994). The ozone accumulation started before noon and lasted until 1900 CET. The lowest levels of ozone concentrations were recorded around sunset and sunrise, which coincided with the rush hours (Jelić and Klaić, 2010).
Each day, two pronounced peaks (one daytime and another night time), characterized the daily course of ozone levels over the investigated period. Occasionally, a third, secondary daytime ozone peak occurred. While the first peak was recorded around 1100 CET, the second daytime peak (most frequently, exceeding the hourly warning level) was recorded around 1500–1600 CET. Generally, photochemical ozone production occurs only during the day, while significant ozone loss occurs through dry deposition, titration with NO, or thermal decay during the night time. However, during this measured period, the nocturnal maxima were high (that is, two times lower than the daytime maxima), which is similar to the results obtained for the city of Essen (Strassburger and Kuttler, 1998). Thus, according to Strassburger and Kuttler (1998), these high night time ozone concentrations could be due to advection by the local breeze or/and nocturnal low level jet.
Figure 8 also shows the comparison between modelled and measured concentrations. Similar to the WRF evaluation, the statistical performance for the measured and simulated hourly ozone concentrations was also assessed (Table 1). Despite large horizontal grid spacing, the EMEP model reproduced the diurnal cycle of ozone levels satisfactorily. However, it mostly underestimated the afternoon O3 concentration maxima and the night time secondary peaks, although the modelled night time O3 levels were quite high (up to 70 µg m−3) between 13 and 17 August. Obviously, meteorological effects that contributed to the higher night time ozone levels were acting on a smaller scale. Thus, these sub-grid scale effects were not captured by the standard (50 × 50 km2) EMEP model. Still, the mean bias smaller than 2, the index of agreement of around 0.7 and correlation coefficient of 0.6 may be considered as acceptable. The more so, these values are similar to those obtained by Mangia et al. (2010) for substantially finer horizontal resolution.
The measured ozone levels in Rijeka are subsequently compared with those available for neighbouring sites in Croatia (Zavižan, Dinaric Alps), Slovenia (Celje and Iskrba) and Austria (Villach) (Figure 9). While ozone levels were highest in Rijeka, especially during 17–19 August, Slovenian and Austrian stations showed similar increases in ozone concentrations with an approximately 3 day lag (from 19 to 21 August 2000). It should be noted that the Villach station is an urban one and, thus, it is generally exposed to traffic emissions. Accordingly, the low night time values, lower than observed, are expected. Similar comments are also valid for Rijeka, where elevated night time values (higher than generally expected) occurred. Therefore, the possibility of a regional transport of ozone, which could contribute to elevated night time values, was investigated (Section '6. Trajectories'). Furthermore, in Iskrba, which is a background rural station, the typical daily course of urban ozone levels was seen. As there are no significant local emission sources close to Iskrba, the transport of polluted air masses might play a dominant role in this case (Žabkar et al., 2008). Finally, very small temporal variations in ozone levels in nearby background site Zavižan (Cvitaš et al., 2007) suggest that elevated night time concentrations in Rijeka the most probably could not be attributed to stratospheric intrusion.
5.2. Horizontal near-surface distribution of the ozone concentrations
As shown in Figure 10, near-surface ozone concentration fields revealed temporal and large-scale spatial variations above the central Mediterranean.
On 13 August, an elongated area of enhanced ozone concentrations that extends from the Po River Valley to Croatia, reaching a maximum value over the northern Adriatic, was prominent (Figure 10(a)). The position of the core of high ozone concentrations corresponded to the area of weak winds (Figure 4(b)), as well as with the region of elevated pollutant concentrations shown by Finardi et al. (2011). Over the northeast Adriatic and inland Croatia, VOC concentrations were low (Figure 10(b)). In time, the bora transported ozone-rich air mass away from the eastern Adriatic coast, especially over its middle and southern parts (compare Figure 10 parts (a) and (c)). The bora wind, however, brought the air with the elevated VOC concentrations (compare Figure 10 parts (b) and (d)) from woody inland areas. Over the northern Adriatic, relatively high ozone levels (Figure 10(c)), in accordance with the wind fields in Figure 5, were still found, while a confined strip of enhanced ozone concentrations along the western Adriatic coast could be attributed to a swirled bora jet. As of 16 August, over the Po River Valley, a significant decrease in ozone concentrations, accompanied by a shallow low, was found (Figure 10(e)). The largest ozone values and still elevated levels of VOC were confined to the northern Adriatic and northeastern inland regions over Slovenia and Austria (Figure 10(f)). A local, thermally induced onshore flow superimposed to the large-scale, southwesterly wind, controlled the ozone transport along the northeast Adriatic and surrounding regions. The northern Adriatic region as a permanent source of high ozone concentrations was also suggested by the studies of Cvitaš et al. (1997), Kaiser et al. (2007, see their figure 7e) and Žabkar et al. (2011). Although this area of high ozone levels was observed here at other times during August (not shown), it was much less pronounced than during the studied period.
The hourly mean ozone field modelled for selected days is shown in Figure 11. During the daytime, particularly in the early afternoon (∼1600 CET), high ozone production was evident over the Mediterranean and the larger part of Europe (Figure 11(b)). This remained almost constant throughout the entire investigated period (not shown). As expected, the night time ozone levels were low, especially over the continent. Thus, no high ozone concentrations were recorded during the night time over the northern Adriatic, either before (Figure 11(a)) or after (Figure 11(d)) the episode. Conversely, during the episode of high ozone concentrations (17 August), the persistent high ozone values at night were found mainly above the northern Adriatic, while concentrations were low above the rest of the Adriatic (Figure 11(c)). During and after this episode, the larger night time (Figure 11(d)) ozone levels over the highlands east of the Northern Adriatic, (above Slovenia and Austria) indicated a regional ozone transport that corresponded to the observed elevated concentrations (Figure 9).
The pathway of ozone transport can be elucidated by tracing the trajectory of a hypothetical air parcel into the location of interest, assuming that the air parcel collects pollutants while passing above emission areas. To this end, three-dimensional, 48 h backward trajectories were calculated for each hour using version 4 of RIP (Read/Interpolate/Plot), a post-processing program of the WRF model with two arbitrary endpoints (one at the northwestern Istria, and the other above the northern Adriatic Sea) and the third point in Rijeka. These three points are denoted by I, N and R respectively (Figure 12).
Figure 12 shows daytime (1600 CET on 13 August) and night time (at 0400 CET on 14 August) 48 h backward trajectories for selected points. Trajectories were calculated for the near surface wind (10 m a.g.l.) at a 3 km resolution. Trajectories arriving in point N show that heavily industrialized Po River Basin contributed considerably to the overall pollution over the northern Adriatic. This is in accordance with results of Žabkar et al. (2011, Episode III). Apart from the Po Valley, the contribution of coastal emissions within the Trieste Bay and surrounding industrialized coastal area was significant (particularly during night time). However, only the western Istrian coast was affected by the air advection from the Italian coast since a strong convective activity prevented the deep inland SB penetration during the daytime. As suggested by trajectories arriving at point R (Figure 12), the elevated levels of ozone observed in Rijeka were mostly detached from the high ozone concentrations found in the northern Adriatic areas. Thus, the elevated ozone levels in Rijeka were primarily due to local sources in Kvarner Bay accompanied by mesoscale thermally induced circulations.
During the bora, trajectories clearly followed the dominant northeasterly winds and almost never varied over time (not shown). The bora wind transported ozone offshore towards the western Adriatic coast, simultaneously enhancing the VOC concentrations at the eastern Adriatic coast. This could soon affect the production of ozone in the Rijeka Bay area since some authors (e.g. Symeonidis et al., 2008; Poupkou et al., 2009) found that the biogenic emissions can lead to an increase of maximum ozone levels up to 20 ppb in the Mediterranean area during summer. On 17 August, the large-scale wind changed, turning to southwesterly flow.
The trajectories in Rijeka area again were not influenced by the northern Adriatic region (Figure 13). During the night time (Figure 13(a)) the seaward flow carried the air from the highlands (enriched by the VOC) over Rijeka town and further over Rijeka Bay. During the daytime hours (Figure 13 parts (b) and (c) ∼1000 CET) the polluted air was close to sea level, due to subsidence in the thermal circulation cell over Rijeka. In that location, the ozone levels remained relatively high, due to stronger stratification, deficient vertical diffusion and slower destruction (Fortezza et al., 1995). Because the fair-weather conditions allowed photochemical ozone production over the bay, ozone that was formed in the last few hours returned over the coastline via onshore flow in the form of a shallow (∼200–300 m deep), clockwise, afternoon eddy. Further, the ozone was moved uphill by the onshore wind flow toward the mountain ridge. Pollutants that had been advected offshore during the daytime could be very easily returned over Rijeka during the night time. The model reproduced an almost equal horizontal extent of the onshore/offshore flow, which, according to Fortezza et al. (1995), creates favourable conditions for horizontal ozone recirculation. Therefore, according to the observed trajectories (Figure 13), the recirculation occurred in both the horizontal plane (between seaward and landward flow exchanges) and in the vertical plane in the thermal circulation over the bay (by the subsidence between return current and onshore breezes). Aged and polluted advected air and in situ production of pollution are the main reasons for the periodic afternoon high ozone levels.
The domination of the southwesterly large-scale flow marked the shape of the trajectories in Figure 14. This flow allowed the return of the aged polluted air toward the northern Adriatic (Figure 14(a)). Thus, the distribution of the ozone and its precursors was controlled by a combination of mesoscale (i.e., regional landward breeze) and synoptic-scale southwesterly winds. Therefore, the trajectories point to the possible transport of ozone and/or its precursors from the northern Adriatic and its coastal regions (compare to Figure 10) toward the Alps (observed in Figure 9), Istria, and Rijeka Bay. Accordingly, above the northeast Adriatic, the daytime ozone originating from the northern Adriatic was carried by the southwesterly wind (black line in Figure 14) in the lowermost 1500 m far inland over the Istria. During the night, the ozone was transported from Istria towards Kvarner Bay by offshore flow (e.g. hours from 1 to 7 in Figure 14(a)). In latter times, it was brought from the Kvarner Bay to Rijeka by the channelled, thermally induced daytime flow (from 0700 to 1600 in Figure 14(a)). During their transport, air parcels within the lowermost 1000 m of the atmosphere passed near a thermal power plant (square numbered 3 in Figure 1(b)) that could have an impact on the photochemical activity in Rijeka Bay. Although trajectories showed the prevailing influence of local sources in Rijeka, toward the end of the study period a certain influence of the regional transport on concentrations measured in Rijeka can be seen (Figure 14).
This finding is also in agreement with related studies by Kallos et al. (2007) and Lindskog et al. (2007), who have found that ozone plumes can pass long distances over the Mediterranean in the lower troposphere, reaching urban remote locations and affecting their air quality over a time scale of several days.
7. Summary and concluding remarks
In this study, the influence of meteorological conditions on the ozone episode over 13–19 August 2000 in Rijeka, a coastal town located in an area with complex topography, is examined. During the episode, peak ozone concentrations exceeded the national daily limit values. Throughout the episode, the synoptic forcing was relatively weak, thus facilitating the development of a local thermal circulation. These predominantly fair-weather conditions were occasionally interrupted by short-lived and sporadic bora wind. At the beginning of the studied time, maximum air temperatures higher than 30 °C in Rijeka were influenced by high convective activity and cloudiness that decreased incoming irradiation. After 16 August, air temperatures gradually rose to 36 °C, and hourly cloudiness did not exceed levels of 2/10.
The analysed measurements were supplemented by results obtained from the WRF meteorological model and EMEP chemical model, as well from trajectory-based analyses. Modelled meteorological fields and available measurements confirmed the domination of local, thermally induced circulations along the northern Adriatic coast. Most often, the weak, large-scale wind over the northern Adriatic was of southwestern direction. The statistical meteorological indices calculated for Rijeka showed relatively good agreement between the modelled and measured values, with the wind direction being reproduced slightly better than the wind speed. In spite of the large horizontal grid spacing, the EMEP chemical model acceptably reproduced the diurnal variation of ozone in Rijeka, showing typical urban pattern.
The obtained results suggest several factors which can be responsible for: (1) the occurrence of the wide area of ozone-enriched air over the northern Adriatic; (2) the high ozone levels in Rijeka (13–19 August 2000) and (3) the interaction of the above two.
The EMEP model results showed that coastal industrial emission sources, particularly in the heavily industrialized Po River Basin and the Trieste Bay, contributed to the overall pollution over the northern Adriatic Sea. The level of ozone concentration and the position of the concentration maximum were associated primary with (1) an exchange of two prominent flows: the large-scale southwesterly flow, and the opposite northeasterly (bora) wind, and (2) a superimposed, mesoscale, thermally induced breeze circulation along the Adriatic coasts. At the synoptic scale, the northeasterly bora affected the pollutant concentrations, in particular transporting the VOC from inland woody areas towards the eastern Adriatic coast. Above the northern Adriatic region the southwesterly flow returned ozone rich air toward the Alps. These large-scale conditions were accompanied by thermally induced circulations that markedly affected the strength and direction of the wind field. The landward breeze (up to 6–7 m s−1) along the entire Adriatic coast caused a significant divergence zone with weak winds (below 1 m s−1) along the Adriatic Sea. This region of daytime divergence (which was also found by Klaić et al., 2009) favours accumulation of ozone produced in the area. Wind trajectories showed that the nocturnal seaward transport of ozone and its precursors was presumably maintained by the subsequent daytime photochemical reactions and ozone production and its onshore return in the afternoon. This landward ozone transport along the coast accompanied by the local breeze and southwesterly large-scale wind agrees with the conclusion of Žabkar et al. (2008, 2011), who showed that the high ozone levels in nearby Slovenia are associated with the short trajectory clusters of slow-moving air masses originating from the southwest. Over the central and southern parts of the Adriatic, the weak northwesterly flow (the Etesian wind) was accompanied by lower ozone concentrations.
In Rijeka, the daytime, onshore flow from the southwest which was occasionally interrupted by the bora wind, exchanged with a night time northerly seaward flow of comparable strength and extent. This change caused dominant horizontal and sometimes vertical recirculation of the polluted air with an obvious tendency to accumulate pollutants. During nocturnal hours, the ozone and its precursors were presumably transported seaward from the mountain slopes by the offshore flow (i.e., superimposed land breeze and down-slope wind). Over the urban site (Rijeka), the destruction of ozone was fast, while over the sea, O3 levels remained high for at least 6 h. The latter was due to stronger stratification, deficiency of vertical diffusion, and slower destruction (Fortezza et al., 1995). In both the nocturnal and transition period (from 0800 to 1000 CET) the photochemical precursors were easily advected toward Rijeka by easterly winds and transformed into ozone. In the air above Rijeka, the advected ozone was added to the urban ozone, resulting in the first daytime peak (around 1100 CET). In the afternoon (especially from 1400 to 1600 CET), the ozone levels were increased by pollution brought by the southwesterly flow from the sea in agreement with results of Alebić-Juretić (2012). Simultaneously, the pollution was advected toward the mountainous hinterland without significant dilution. At night, certain amounts of pollution were returned to the town by the local offshore flow, resulting thus in significant (up to 120 µg m−3) peaks in ozone concentration. Other events of ozone recirculation captured by a SLB cell (sometimes lasting for several days) are described by e.g. Grossi et al. (2000) and Adame et al. (2010).
High ozone levels in the northern Adriatic and ozone advection towards Rijeka somewhat contribute to the extraordinarily high afternoon O3 concentrations. However, the advection from northern Adriatic to Rijeka occurred only at the end of the examined episode. The necessary conditions for the advection of oxidant-rich air masses from the northern Adriatic are a large-scale southwesterly wind superimposed to the local, thermally induced circulation above Istria during fair weather conditions without convection.
Finally, it can be summarized that both synoptic and mesoscale meteorological conditions were highly favourable for the formation of the high ozone episode. While possible ozone transport from the northern Adriatic enhanced the afternoon ozone concentrations in Rijeka, especially on 18 and 19 August, the surrounding local emission sources in association with uniform, thermally induced circulation systems and pollutant recirculation were crucial in the formation of the large ozone levels in Rijeka.
We are very grateful to Nataša Strelec Mahović for providing the satellite image. This work has been supported by the Ministry of Science, Educational and Sport (grants BORA No. 119-1193086-1311, No. 004-1193086-3036, No. 119-1193086-1323, No. 098-0982915-2947 and No. 062-0621341-0308). Anonymous referees are acknowledged for their useful comments and suggestions.