The Australian Air Quality Forecasting System. Part III: Case Study of a Melbourne 4-Day Photochemical Smog Event

K. J. Tory Bureau of Meteorology Research Centre, Melbourne, Victoria, Australia

Search for other papers by K. J. Tory in
Current site
Google Scholar
PubMed
Close
,
M. E. Cope CSIRO Atmospheric Research, Aspendale, Victoria, Australia
CSIRO Energy Technology, Newcastle, New South Wales, Australia

Search for other papers by M. E. Cope in
Current site
Google Scholar
PubMed
Close
,
G. D. Hess Bureau of Meteorology Research Centre, Melbourne, Victoria, Australia

Search for other papers by G. D. Hess in
Current site
Google Scholar
PubMed
Close
,
S. Lee CSIRO Atmospheric Research, Aspendale, Victoria, Australia

Search for other papers by S. Lee in
Current site
Google Scholar
PubMed
Close
,
K. Puri Bureau of Meteorology Research Centre, Melbourne, Victoria, Australia

Search for other papers by K. Puri in
Current site
Google Scholar
PubMed
Close
,
P. C. Manins CSIRO Atmospheric Research, Aspendale, Victoria, Australia

Search for other papers by P. C. Manins in
Current site
Google Scholar
PubMed
Close
, and
N. Wong Environment Protection Authority of Victoria, Melbourne, Victoria, Australia

Search for other papers by N. Wong in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A 4-day photochemical smog event in the Melbourne, Victoria, Australia, region (6–9 March 2001) is examined to assess the performance of the Australian Air Quality Forecasting System (AAQFS). Although peak ozone concentrations measured during this period did not exceed the 1-h national air quality standard of 100 ppb, elevated maximum ozone concentrations in the range of 50–80 ppb were recorded at a number of monitoring stations on all four days. These maximum values were in general very well forecast by the AAQFS. On all but the third day the system predicted the advection of ozone precursors over Port Phillip (the adjacent bay) during the morning, where, later in the day, relatively high ozone concentrations developed. The ozone was advected back inland by bay and sea breezes. On the third day, a southerly component to the background wind direction prevented the precursor drainage over the bay, and the characteristic ozone cycle was disrupted. The success of the system's ability to predict peak ozone at individual monitoring stations was largely dependent on the direction and penetration of the sea and bay breezes, which in turn were dependent on the delicate balance between these winds and the opposing synoptic flow.

Corresponding author address: K. J. Tory, Bureau of Meteorology Research Centre, GPO Box 1289K, Melbourne, VIC 3001, Australia. k.tory@bom.gov.au

Abstract

A 4-day photochemical smog event in the Melbourne, Victoria, Australia, region (6–9 March 2001) is examined to assess the performance of the Australian Air Quality Forecasting System (AAQFS). Although peak ozone concentrations measured during this period did not exceed the 1-h national air quality standard of 100 ppb, elevated maximum ozone concentrations in the range of 50–80 ppb were recorded at a number of monitoring stations on all four days. These maximum values were in general very well forecast by the AAQFS. On all but the third day the system predicted the advection of ozone precursors over Port Phillip (the adjacent bay) during the morning, where, later in the day, relatively high ozone concentrations developed. The ozone was advected back inland by bay and sea breezes. On the third day, a southerly component to the background wind direction prevented the precursor drainage over the bay, and the characteristic ozone cycle was disrupted. The success of the system's ability to predict peak ozone at individual monitoring stations was largely dependent on the direction and penetration of the sea and bay breezes, which in turn were dependent on the delicate balance between these winds and the opposing synoptic flow.

Corresponding author address: K. J. Tory, Bureau of Meteorology Research Centre, GPO Box 1289K, Melbourne, VIC 3001, Australia. k.tory@bom.gov.au

Introduction

The Australian Air Quality Forecasting System (AAQFS) currently issues 36-h forecasts 2 times per day for two domains—one domain encompasses the state of Victoria, and the other most of New South Wales. The AAQFS consists of five major components: a numerical weather prediction (NWP) system [Limited Area Prediction System (LAPS); Puri et al. 1998], an emissions inventory module, a chemical transport module (CTM) for air quality (AQ) modeling, an evaluation module, and a data archiving and display module. The meteorological conditions are modeled with a 0.05° horizontal grid with 29 vertical levels (nested in a 0.375° limited-area model, which in turn is nested in a global model); the chemistry is modeled with a 0.01° horizontal grid with 17 vertical levels (grid 2), which is nested in a 0.05° horizontal grid (grid 1). The topographical maps presented in Fig. 1 illustrate the coverage of grid 1 (Fig. 1a) and grid 2 (Fig. 1b). A description of the system and a statistical evaluation of the performance are presented in Cope et al. (2004). In addition to statistical verification, the performance of the AAQFS has been examined through a number of case studies (Cope et al. 1999; Hess et al. 2000a,b; Tory et al. 2000). In this paper we explore a photochemical smog case study for the Victorian domain, which complements a case study by Hess et al. (2004) for the NSW domain. We believe it is important to document the ozone cycles in both the Victorian and NSW domains because topographic differences between the two domains lead to the generation of ozone cycles unique to Melbourne and Sydney.

In March 2001 the Melbourne region experienced a 4-day period of elevated ozone concentrations (maximum concentrations 50–80 ppb). The four days together provide an excellent illustration of the Melbourne ozone cycle, the day-to-day variation of the cycle, the sensitivity of the cycle to variations in the synoptic flow, and, most important, the sensitivity of the ozone forecasts to the NWP model's ability to accurately predict local-scale winds and the fine balance between these winds and synoptic flows. These four days included the longest period of elevated ozone concentrations recorded at all operating monitoring stations during the 2000/01 photochemical smog season (there were no extreme events during this season).

The NWP model is arguably the most critical component of the system. Errors in the forecast AQ concentrations may result from errors in the emissions inventory, in the forecast meteorological fields, or in the chemical mechanism. Of these three components, the NWP component is the most critical for Melbourne. Validation studies of the emissions inventory (e.g., Environment Protection Authority of Victoria 1999) have confirmed its general accuracy. (One possible area of uncertainty is the characterization of biogenic emissions, but there are no far-field monitoring stations available capable of detecting inadequacies in these emissions.) Direct comparisons of predicted ozone concentrations with the present condensed chemical mechanism, Generic Reaction Set (GRS), and a comprehensive mechanism, Carbon Bond IV, have shown generally good agreement. On occasions GRS underestimates ozone peaks by about 10 ppb. Another potential source of error is the ozone initial and boundary conditions. To minimize initialization errors, background values of 20 ppb are initially applied everywhere and the AAQFS is run for 24 h prior to the investigation period to allow the system to develop a realistic ozone distribution at the initial time. (Operationally, pollutant species from the previous day's forecast are used to initialize the current forecast.) Errors at the boundaries are not a particular concern in the two AAQFS domains, because Australian urban centers are often fairly isolated and long-range transport of pollutants is not as significant as it is in other parts of the world (Tory et al. 2003). Thus, background ozone values of 20–25 ppb are applied at the boundaries throughout the simulation.

Forecast errors associated with the emissions inventory or chemical mechanism will lead to errors in magnitude. Whereas, there is potential for 100% “hit or miss” forecast errors in the NWP model from relatively minor wind direction errors. In Melbourne the hit-or-miss error can be particularly apparent. An accurate meteorological forecast over the bay is crucial to a good AQ forecast. If the location of the precursors or ozone over the bay is near the point of flow divergence of the bay breeze, then a slight error in the forecast bay breeze divergence location could be the difference between plume transport to the east or west side of the bay. Results presented in this study do show all of the larger ozone forecast errors can be explained by errors in wind direction. For this reason the study focuses on the performance of the NWP component of the system, and the impact of the meteorological conditions (local-scale winds in particular) on the ozone cycle. It should be noted that day-to-day differences in emissions were insignificant because all four days were weekdays (emissions are lower on weekends and public holidays), and the daily temperature cycle was similar on each day [the emissions model is a function of temperature (Ng and Minchin 2000)].

Building on the preliminary observational studies of Spillane (1978), Tapp (1985, 1992), Evans et al. (1985), and Abbs (1986), and the early modeling studies of Abbs (1986), McGregor and Kimura (1989), Hess (1989a,b), and Cope (1999), a conceptual model of the AQ and the AQ sensitive meteorological conditions has been developed after more than six years of AAQFS modeling and monitoring of meteorological and AQ observations. The ozone conceptual model is introduced in this paper as the “typical ozone cycle,” which includes the associated mesoscale wind features, drainage flow, and bay and sea breezes.

In Melbourne, the typical ozone cycle is similar to ozone cycles around the world at other coastal cities, or cities located on the shores of large lakes, in which ozone precursors are advected offshore by early morning drainage flow, and high concentrations of ozone are advected back inland by the sea breeze later in the day (e.g., Simpson 1994). This cycle can be greatly complicated by nonuniform coastal geography, such as complex coastlines (e.g., bays and islands) and mountainous terrain. Examples of very complex coastal flow patterns can be found in a number of Japanese regions (e.g., Kurita et al. 1985; Mizuma 1995, 1998). The Melbourne coastal geography, although not as complex as some regions of Japan, is also influenced by an irregular coastline and nearby hills and mountains. The Melbourne metropolitan area is spread around the edge of a large bay (Port Phillip, see Fig. 1) with hills surrounding the northern half of the bay and mountains to the northeast. At night, drainage flows carrying ozone precursors tend to converge over the bay. In addition, this nocturnal offshore flow can be enhanced by the Melbourne eddy and possibly land breezes. The Melbourne eddy is generated by vorticity shed from the upwind mountains (Great Dividing Range to the east of Melbourne, Fig. 1a) and can play a significant role in recirculating pollutants in the Melbourne region (e.g., see Spillane 1978; Tapp 1985; Baines and Manins 1989; McGregor and Kimura 1989; Hess 1989a,b). A “bay breeze” often develops in the late morning to early afternoon, in which the winds appear to diverge outward from some central point over the bay. The bay breeze typically lasts for a few hours before it is overpowered by the sea breeze that develops on the coastal boundary south of the bay. Another confounding factor is that the inland penetration of the bay and sea breezes is often opposed by background northerly flow. A fine balance between the opposing flows may exist, which can lead to slowed or stalled inland penetration of bay and sea breezes. When ozone is transported by the bay and sea breezes in a sensitive situation such as this, ozone forecasts can become hit or miss. Thus, a good ozone forecast can demand a high level of forecast accuracy of the relative strengths of these opposing winds.

In this paper we demonstrate the ability of the AAQFS to capture the Melbourne ozone cycle over four consecutive days. We provide a detailed description of the modeled ozone cycle—more detailed than is possible to verify because of a relatively small and localized ozone observation network. Performance of the forecasting system is assessed by comparing observed and forecast ozone and NOy time series. We show the ozone cycle is well forecast, and where inconsistencies exist we examine the spatial and temporal ozone distribution, and, with a knowledge of wind errors, speculate on what might have happened in reality. A secondary purpose of the paper is to show that many finer-scale features can be captured by the AAQFS (i.e., show that they are within the limits of predictability).

In the next section the synoptic background and vertical atmospheric structure for the 4-day period are presented. Using the 4-day period for illustration the typical Melbourne ozone cycle is presented in section 3, and the day-to-day variation of the cycle is introduced. In section 4 the vertical structure of the modeled ozone, winds, and potential temperature is examined for a typical ozone day (day 2). In section 5 the ozone forecasts are validated against monitoring station data. The case study is discussed in section 6 and summarized in section 7.

Meteorological conditions

Synoptic background

The synoptic situation is presented in Fig. 2, which shows the manual analyses of mean sea level pressure at 0000 UTC [1100 Australian Eastern Daylight Time (AEDT)] for 6–9 March 2001. Over the 4-day period the synoptic flow was dominated by a high pressure cell centered to the south and southeast of mainland Australia, which maintained a gradient wind direction between easterly and northerly throughout the period. The background wind direction, combined with vertical stability provided by subsidence within the high pressure cell, are conditions favorable for the development of the Melbourne eddy (Spillane 1978; Tapp 1985; Baines and Manins 1989; McGregor and Kimura 1989; Hess 1989a,b). The northerly component of the synoptic wind typically opposes the inland penetration of the bay and sea breezes. On the morning of 6 March a northeasterly gradient wind over Melbourne can be inferred from the pressure pattern evident in Fig. 2a. A broad trough covered much of northern and western Australia punctuated by a series of low pressure centers. A very similar pattern was present on the following day (Fig. 2b), except that the pressure gradient over Melbourne was particularly weak and the gradient wind direction had shifted to a more easterly direction. Twenty-four hours later the pressure had increased over the Tasman Sea and coastal southeastern Australia, with the approach of a cold front from the south (Fig. 2c). This led to a slight rotation and an increase in the pressure gradients there, which resulted in a rotation to northeasterly and strengthening of the gradient wind. In the next 24 h, ridging over the Tasman Sea (east of southeastern Australia) became more pronounced and the gradient wind became more northerly in direction over southeastern Australia (Fig. 2d). On the following day a trough and cold front approached the Melbourne area (not shown). Stronger winds, enhanced by the approaching front, increased the ventilation and disrupted the ozone cycle of the previous few days (not shown).

Vertical structure

Throughout the 4-day period the atmosphere at Melbourne Airport was strongly stable above about 2000– 2500 m. This can be seen in the time-versus-height profiles of the atmosphere presented in Fig. 3. During this period the stable layer, which acts to trap pollutant species vertically and, thus, to limit vertical dilution, was relatively deep as compared with other high-ozone events we have investigated; this might partly explain the observations of “high” rather than “extreme” maximum ozone concentrations (i.e., greater dilution). The observed structure (Fig. 3a) was constructed from data collected by domestic aircraft as they take off from Melbourne Airport [Aircraft Meteorological Data Relay (AMDAR) data]. The equivalent profile generated from LAPS (NWP model) data is presented in Fig. 3b. A comparison of Figs. 3a and 3b shows the model forecast captured the strong diurnal pattern of low-level nocturnal cooling followed by surface heating, and the development of a daytime mixed layer that extended to 2000 m on the first, second, and fourth days, and about 1500 m on the third day. On each day the forecast depth of the mixed layer (thick line) compares very well to that observed. The changing direction of the low-level winds was also well represented. During the early morning of the first and second days the low-level winds were northerly between 1900 and 2300 UTC (0600– 1000 AEDT in the model, because of drainage flow and possibly downstream turning of the upstream northeasterly winds flowing parallel to the Great Dividing Range) and shifted to southerly later in the day. The timing of the forecast wind shift to a southerly direction was excellent on both days (about 0500 and 0100 UTC, 1600 and midday AEDT, respectively). The model southerly winds were due to the arrival of the bay and sea breezes, which on day 2 were reinforced by larger-scale southeasterly winds that continued throughout the night and into the next day. The later arrival of the model bay/sea breeze on the first day was due to opposing northerly flow that slowed the advance of the cooler marine air (described later in section 3). The same pattern was evident in the observations, although the observed southerly winds had more of an easterly component on both the first and second days. Throughout the third day the observed and model winds were predominantly southeasterly and southerly, respectively. Despite no obvious bay-/sea-breeze arrival, because the background wind was from the bay direction, the winds did intensify during the middle of the day, possibly enhanced by the daytime land–sea pressure gradient. Between 1900 and 0300 UTC (0600 and 1400 AEDT) on day 4 both model and observations showed stronger northerly winds (due presumably to the strengthening northerly gradient winds, Fig. 2d). Soon after, Fig. 3a shows a bay or sea breeze arrived at the airport [see the wind shift to southeasterly and an accompanying drop in temperature after about 0500 UTC (1600 AEDT) 9 March]. The model bay or sea breeze did not penetrate as far as the airport (although the northerlies did weaken because of the close proximity of the wind shift farther south). Last, both the model and the observations show the northerlies reintensified between 1900 and 2300 UTC 9 March (0600 and 1000 AEDT 10 March) as the cold front approached from the west.

Melbourne ozone cycle

Precursor drainage

Figure 1 shows the topography of the Melbourne area used in the NWP component of the AAQFS. The topography has been smoothed to avoid numerical instability problems and, thus, does not represent the true mountain heights, valley depths, or local-scale structure. The Great Dividing Range runs approximately east– west to the north of Melbourne and Port Phillip (the bay), although it curves slightly to match the concavity of the northern half of the bay. This figure shows a relatively gentle slope from the coast to the base of the mountains of about 200 m over 30–60 km, and the contours are roughly parallel to the coastline of the northern part of Port Phillip. A detailed study of low-level winds during light wind conditions by Tapp (1992) produced a climatology of light-wind flow patterns, including nocturnal drainage flows, in the Melbourne area. Tapp (1992) showed that under clear-sky conditions with light background winds, drainage flow converged on the northern half of the bay. Anemometers located in open and relatively flat terrain around the northern half of the bay identified preferred flow directions that radiated inward toward the bay.

Figure 4 shows the ozone precursors NOy1 with model (thin regularly spaced) and observed (thick) wind vectors overlaid at 0700 AEDT, for each of the four days. The converging drainage flow is most apparent on the second day (Fig. 4b). The modeled wind directions are generally in very good agreement with the observed wind directions. The magnitude, however, can differ considerably. This wind speed difference is largely due to differences between the modeled roughness lengths and roughness lengths at observation sites. Observation sites tend to be chosen in grassy areas free of large obstacles, where the typical roughness lengths are on the order of 0.01 m, whereas the typical grid-averaged roughness length of the area (used in the NWP model) could be orders of magnitude higher. In addition to this, the NWP model appears to have a low-level wind speed bias of about 1 m s−1 (too slow) over land.

A counterclockwise rotation is evident in the forecast wind field on all four days, and on all but the last day the rotation completes 360°. On the first three days, model conditions were favorable for the development of the Melbourne eddy (northeasterly to easterly background wind, stable stratification). Note the generally stronger wind magnitudes from the east in the southern part of the domain. The northern part of the bay is typically in the lee of the upstream mountain barrier; consequently, the Melbourne eddy, formed by downstream vorticity shedding, is typically centered in the vicinity of the bay. On the fourth day (Fig. 4d) the wind shifted to the north, and the northern part of the bay was no longer in the lee of the mountain barrier.

On days 3 and 4 the modeled wind directions are well supported by the observed flow (Figs. 4c,d). However, the observed wind vectors on the western side of the bay on days 1 and 2 are more consistent with drainage flow rather than an eddy. Errors associated with an incorrectly forecast Melbourne eddy would be most greatly felt on the western side of the bay where the eddy circulation is in the opposite direction to the drainage flow. Thus, the majority of the emissions (Melbourne source) would at least be advected in the right general direction. The smaller emissions source at Geelong would be advected inland rather than over the bay. This is likely to result in an underprediction of ozone later in the day near Geelong, but because there are no monitoring stations to the northwest of the Geelong area, the errors would go unnoticed.

The advection of ozone precursors in the rotating flow is illustrated by the curved shape of the ozone NOy plume (see, e.g., the Geelong area). The modeled Melbourne eddy differed in structure on the three days that it developed. On the first two days the flow direction in the southeast corner of the domain (a reasonable indication of the background flow) was similar, and the eddy centers were located on the northern and southern coasts of Corio Bay, respectively (see Fig. 1 for location). The difference in eddy center location may be due to the strength of the background wind. The stronger easterly wind on day 1 appears to have produced an eddy of larger horizontal scale, which may have led to the more northerly center of rotation. On the third day the background flow was stronger than the previous two days and had a more southerly component. This resulted in a shift of the rotation center considerably farther north and greatly reduced the bayward advection of the NOy plume (Fig. 4c).

The effect of these different morning flow patterns on the air quality of the day was largely dependent on whether the ozone precursors were advected over the bay. Figure 5 is the same as Fig. 4, except 3 h later (1000 AEDT). On all but day 3, forecast ozone precursors in significant concentrations were advected over the bay by this time. A movie loop of the predicted NOy transport on day 3 showed that the greatest concentrations were advected in a circular pattern toward the north and then to the east and remained north of the bay. On this day the ozone concentrations were noticeably weaker than on the other days.

Bay-/sea-breeze ozone transport

A significant gradient in mixing depth between the land and sea, due to land–sea heating contrasts, led to greater dilution of ozone precursors and ozone over the land surface (presented in section 4). This led to ozone development in the highest concentrations, over the precursor-rich areas of the bay. The ozone-rich air was then transported inland by the bay and sea breezes. Figures 6 and 7 are the same as Figs. 4 and 5, except ozone is contoured and the times are 1300 and 1700 AEDT, respectively. These are typical times when the bay and sea breezes, respectively, are well established. Evidence of the bay breeze can be seen in Figs. 6a, 6b, and 6d. Note the point of divergence in the wind vectors, over the bay, and the good agreement with the observed wind direction. Typically, the apparent divergence point first develops in a more central location over the bay and migrates with time toward the south (toward the advancing sea breeze). Like the Melbourne eddy, the forecast and observed bay breeze structure varies significantly from day to day and is influenced by the background flow. A strong opposing background flow may slow the bay-breeze advance or inhibit the bay breeze completely. Or, as occurred on day 3, the advection of cooler marine air by a southeasterly background wind was sufficiently strong to impede the development of a well-defined land–sea temperature gradient, particularly over the Mornington Peninsula (see Fig. 1 for location). Further evidence of the day-to-day variation of the bay-breeze structure and development is present in the location of the apparent divergent point of the vectors. (Because the land–sea temperature gradient was insignificant over the Mornington Peninsula on day 3, there was no wind acceleration from the bay to the peninsula and, hence, no apparent divergence pattern of the wind vectors over the bay.) At 1300 AEDT on days 1, 2, and 4, the apparent point of divergence was located near the southeastern, southern, and eastern edges of the bay, respectively (Figs. 6a, 6b, and 6d). A highly accurate forecast of the bay breeze structure and development can be crucial for an accurate ozone forecast. Because the bay breeze is a radial flow, small errors in the background wind and relative land–sea temperature gradient strengths around the bay can shift the apparent divergence slightly, which in turn can lead to precursor and ozone advection to the opposite side of the bay.

At 1300 AEDT the ozone development was at an early stage on all days except day 2, where the concentrations were particularly high offshore of the Bellarine Peninsula (Fig. 6b). Note the relatively clean air over the bay on the day ozone precursors were confined to the north by the Melbourne eddy (day 3). The bay breeze typically lasts only a few hours before the stronger and larger-scale sea breeze arrives from the southeast (Fig. 7). In the NWP model the transition from bay to sea breeze tends to be relatively gentle and occurs over a period of an hour or two. On the western side of the bay the transition is most difficult to identify because the direction of both winds tend to be similar, whereas the transition on the other side of the bay typically involves a 180° rotation from the northwesterly bay breeze to the southeasterly sea breeze. The observed wind directions again show the model performance was generally good (Fig. 7). An exception is in the vicinity of the northeastern bay coastline. On day 1 (Fig. 7a) the observed wind vectors show the bay breeze was still prominent, which suggests the model sea-breeze arrival was a little early. On days 2 and 3 the modeled winds were biased slightly to the west, and on day 4 the bias was more pronounced because of the more easterly modeled sea breeze. The sea breeze generally transports the ozone-rich bay air inland toward the north and northwest. However, on day 4 a relatively strong opposing northerly flow was present (cf. the forecast and observed winds presented in Figs. 6 and 7d with Figs. 6 and 7a– c) and a more easterly forecast sea breeze failed to transport ozone to the north (Fig. 7d). This is a likely explanation for the underprediction at the Alphington monitoring station (about 10 km inland, longitude 144.95°E) mentioned below in section 5. The strength and direction of the opposing flow not only influence the development of the bay breeze, but also affect the timing and direction, and the ultimate penetration of the sea breeze. Because the bay and/or sea breezes often transport ozone-rich air inland, the success of an ozone forecast often hinges on the success of the bay- and/or sea breeze forecast. This is a result we have found repeatedly in both Melbourne and Sydney (e.g., Hess et al. 2004).

Vertical ozone structure

In section 3 the basic Melbourne ozone cycle was introduced, in which precursors were advected over the bay by drainage flows, and ozone was transported inland by the bay and sea breezes. Unlike the last two days, days 1 and 2 are good examples of the basic Melbourne ozone cycle, relatively free of external complicating factors. This cycle is well illustrated by surface plots alone; however, it is useful to examine the vertical structure as well. Unfortunately there were no above-surface ozone data available (no above-surface AQ monitoring is performed in the Melbourne airshed), and the upper-air meteorological data were limited to the AMDAR data presented in Fig. 2. (Twice-daily radiosonde data were available at Melbourne airport, but they offered very little extra information.) The NWP model forecast of the vertical structure at Melbourne Airport, discussed in section 2, will not be addressed further in this section, except to say that the airport location is representative of the northernmost parts of the cross sections presented in Fig. 8 (introduced below). The remainder of this section is a discussion of the modeled above-surface ozone. Because no observations are available to verify this discussion, the scenario described should be considered as a tentative explanation of the real ozone distribution and development.

The two north–south cross sections for Fig. 8 were chosen to best illustrate the vertical structure of the basic ozone cycle. They are taken from day 2 and pass through longitudes 145.0° and 144.8°. The two sections provide a good representation of the modeled precursor transport, and ozone development and transport, respectively. They pass through points close to the Brighton and Point Cook monitoring stations. The ozone precursors NOy are shaded in Figs. 8a–c, at 0700, 1000, and 1300 AEDT. Potential temperature contours and horizontal wind vectors are also included to illustrate the static stability and horizontal wind direction. Strong static stability in the lowest 500 m and drainage flow (between 60 and 110 km) is evident in Fig. 8a (0700 AEDT). The highest NOy concentrations were present in the Melbourne metropolitan region (between 60 and 90 km) north of the bay, and these precursors were largely contained in a very shallow layer by the strong static stability. By 1000 AEDT (Fig. 8b) the stable layer and drainage flow were still present below about 500 m, although the development of the daytime mixed layer had begun. At this time relatively high NOy concentrations had been advected over the bay, and the depth of the NOy plume had increased with the growing mixed layer. Three hours later (Fig. 8c, 1300 AEDT), the mixed layer had grown to about 1500 m between 90 and 110 km and remained quite shallow over the bay. At this time the bay breeze had penetrated about 20 km inland (60–80 km), and carried with it relatively high NOy concentrations in the cooler maritime air. Note that the NOy concentrations decrease and the mixed-layer depth increases with distance from the coast. (North of 90 km the NOy concentration dropped below the lowest shaded range of 10 ppb.)

At the same time approximately 17 km farther west, the same basic potential temperature structure was present, although the shallow stable layer over the bay was stronger (cf. Figs. 8c with 8d). The northern bay coastline is located about 10 km farther north on this more westerly cross section. The shading in Figs. 8d–f represents ozone concentrations. Note the greatest (weakest) values over the bay (land) where the mixed layer is shallowest (deepest). Note also the intermediary region where the mixed-layer depth ramps up with distance from the coast and the ozone concentrations decrease over the same area. The elevated ozone plume (between 1000 and 1500 m, of Fig. 8d) is a remnant from the previous day. Two hours later (Fig. 8e, 1500 AEDT), the mixed layer had continued to grow in excess of 2000 m over the land near 110 km, and approached 3000 m farther inland (not shown). Ozone concentrations in this relatively deep mixed layer ranged between 35 and 50 ppb, with greater values closer to the coast. By 1700 AEDT the sea breeze was established and stronger south-southwesterly winds had begun to clear the bay of ozone. The clearing is visible between 0 and 10 km in Fig. 8f. At this time ozone concentrations between 50 and 90 ppb had been advected northward over the land between 70 and 85 km (Fig. 8f).

Verification: O3, NOy

As mentioned in the previous section, no above-surface AQ observations were available to verify the vertical AQ structure. Thus, the AQ forecasts can only be verified at the surface monitoring stations. Figure 1 shows the locations of the eight monitoring stations that recorded O3 and NOy during the 4-day period [Box Hill, Mount Cottrell, Alphington, Royal Melbourne Institute of Technology (RMIT), Geelong South, Paisley, Point Cook, and Brighton]. They are all located relatively close to the bay (approximately 15, 15, 10, 5, 5, 1, 1, and 1 km, respectively). All stations are located near the northern parts of the bay except Geelong South, which is close to the western extreme (see Fig. 1). Ozone time series for the eight stations are presented in Fig. 9.

In general the daily maximum ozone forecasts were good to excellent, with the exceptions of Alphington on days 1, 3, and 4, Paisley on day 1, Geelong South on day 2, and Box Hill and RMIT on day 4, where differences between forecast and observed maximum ozone concentrations exceeded 15 ppb. In each case the system underpredicted the maximum ozone concentrations, except Paisley on day 1.

Time series of observed and modeled NOy at the same eight monitoring stations are presented in Fig. 10. There is no consistent correlation between under- and overprediction of NOy and ozone at the eight stations during the 4-day period. For example, at Brighton NOy was under- and overpredicted on the first two days, respectively, and well forecast on the remaining two days, whereas the ozone was overpredicted on day 1, well forecast on days 2 and 3, and underpredicted on day 4. This is of course a product of the nonlocal nature of the ozone cycle (i.e., precursors and ozone are transported significant distances during the cycle, so that precursors measured in the early morning at a particular station may not have any bearing on the ozone concentration measured at the same station later in the day). Figure 10 also shows each station has a unique signature as compared with the ozone traces of Fig. 9 (i.e., more variability among stations), particularly the forecast concentrations.

Many stations exhibit a daily double peak associated with the morning and afternoon peak commuting periods. This feature is most prominent at the inner-city (RMIT, where motor vehicle sources are highly concentrated) and more-inland stations (Alphington and Box Hill, where the sea-breeze strength is comparatively weak). At Brighton and Paisley the morning peaks tend to be strong, and the afternoon peaks are weak. These coastal stations tend to be downwind of the large inner-suburban source areas in the morning (Fig. 4), and influenced by the stronger coastal sea breeze in the afternoon (Fig. 7). The AAQFS forecast concentrations at Point Cook and Mount Cottrell show near-background levels. Figure 1b shows these two stations are located in low-emissions source areas. However, Point Cook is located relatively close to higher source areas, and the discrepancy between the forecast and observed concentrations can be explained by relatively small trajectory errors (e.g., cf. observed and forecast wind vectors near the Point Cook location in Fig. 4). Trajectory errors can also explain the underprediction at Geelong South, possibly because of Melbourne eddy forecast errors. Other discrepancies between forecast and observed NOy concentrations are generally related to trajectory errors. However, the overprediction bias at RMIT is more likely to be associated with insufficient ventilation, because this station is located within the more highly concentrated motor vehicle source region. Another possible cause of the RMIT overprediction is model representation error. The RMIT site is located at the top of a building about 20 m above ground level, whereas the majority of sources are at ground level. The Paisley AQ monitoring site is located close to industrial point sources that may not be well resolved by the AAQFS and could have a bearing on the Paisley underprediction bias.

Maximum ozone concentrations at Alphington were underpredicted on all four days, and on all but the third day the discrepancy can be explained by the timing and penetration of the model bay breeze. On day 1 the predicted bay breeze arrived near midday, whereas the observed shift in wind direction occurred a few hours later. The later-observed bay-breeze arrival may have provided an extra few hours of ozone development that could have further increased the ozone concentrations over the bay, before the inland transport began. Apart from the 1800 AEDT Alphington observation (42 h; Fig. 9), the day-2 ozone forecast was very good. It is unclear what mechanism is responsible for this isolated peak. Meteorological observations at nearby stations (not shown) identify a shift in wind direction from south-southwesterly to southeasterly, a drop in temperature, and a rise in dewpoint temperature, which suggest the peak coincides with the arrival of the sea breeze. Supporting meteorological observations at the Alphington station (not shown) include a drop in temperature, and a southeasterly wind direction. Missing wind data between 1400 and 1700 AEDT make it impossible to identify the timing of the wind shift from bay to sea breeze, and dewpoint temperature is not collected at the air quality monitoring stations. It is likely that the 1800 AEDT peak resulted from higher concentrations of ozone that were transported inland with the sea breeze leading edge. Similar day-2 late peaks were present at all other stations except Geelong South. The underprediction at Geelong South is likely to be related to small wind direction errors. Figure 6b shows a region of ozone in excess of 120 ppb immediately east of the Bellarine Peninsula. This plume was advected to the northwest and concentrations were diluted where it passed over the Bellarine Peninsula and the edge of the diluted plume passed to the north of the monitoring station. Modeled concentrations of 58 ppb (which matches the observed value) passed within a few kilometers of the site.

On day 3 the AAQFS forecast maximum ozone concentrations were quite low (close to 40 ppb) at all stations, which was an excellent forecast at all stations except Mount Cottrell (slightly underpredicted) and Alphington. (Box Hill did not record ozone on that day.) The stronger southerly component to the wind during the early morning, mentioned in section 3, prevented precursor drainage over the bay, which disrupted the typical ozone cycle. However, just north of the grid-2 domain, model ozone concentrations reached 50–60 ppb (not shown), which is of similar magnitude to the Alphington observations. On day 4 the maximum observed Alphington ozone concentration was about 23 ppb higher than the predicted concentration. This discrepancy was most likely due to the failure of the modeled ozone-rich sea and bay breezes to penetrate inland as far as Alphington (as commented on in sections 2 and 3). The limited bay-/sea-breeze penetration is also the most likely explanation for the ozone underprediction at Box Hill and RMIT on day 4. At RMIT the ozone plume “missed” by less than 2 km.

The overpredictions of maximum ozone concentrations at Paisley, and to a lesser extent at Point Cook on day 1, appear to be related to a shift in the observed wind direction during the middle of the afternoon, which was not present in the modeled winds. Nearby meteorological observations suggest the shift was associated with the sea-breeze arrival, which is likely to have brought cleaner air to these stations earlier in the afternoon, once the leading edge of the sea breeze had passed.

In the above few paragraphs we have attempted to explain the differences between the AAQFS ozone forecasts and the observations, where they occur. In each case it seems plausible that the discrepancies resulted from relatively minor forecast wind errors. This focus on the discrepancies draws attention away from the many excellent peak ozone forecasts (e.g., RMIT on day 1; Brighton, Box Hill, Paisley, and Point Cook on day 2; Geelong South, Paisley, and RMIT on day 3). This study shows that the AAQES is potentially able to provide air quality forecasts with high temporal and spatial resolution, accurate to within a few kilometers and a few hours. Case studies of this type highlight weaknesses within the system and help to establish the system's limits of predictability. The ozone concentration time series depicted in Fig. 9 show that this degree of spatial and temporal accuracy is within the limits of predictability. However, at a number of stations and on some of the days, it is clear that these high demands of accuracy were not met, largely because of the fickle timing and direction of the bay and sea breezes, which varied spatially over the domain. This result shows how a focus on performance at individual stations, particularly in the delicate environment of opposing local and synoptic-scale winds, can be an overly demanding expectation of the system. Such a narrow focus also fails to distinguish between degrees of forecast accuracy (i.e., it can not identify between an ozone plume miss of 1 or 100 km). Thus, for the AAQFS to be of use as a forecasting tool, it would be necessary to view a sequence of surface ozone charts that illustrate the ozone cycle as a whole, rather than individual ozone time series charts at selected stations.

Discussion

Of the many ozone events that we have observed using the AAQFS during the last few photochemical smog seasons, we have identified a number of fundamental properties that are common to many ozone events that make up a “basic” Melbourne ozone cycle. These include the following: the advection of ozone precursors over the bay by nocturnal drainage flow (enhanced by the Melbourne eddy on occasions); the development of high ozone concentrations over the bay confined to a shallow layer by stable stratification; the inland advection of cooler ozone-rich air by the bay breeze; inland ozone dilution with distance from the bay; and the flushing of ozone clear of the Melbourne area by the sea breeze, after a possible initial increase in ozone concentration. It has been demonstrated above that slight variations in background wind can significantly affect the precursor drainage, Melbourne eddy, and bay breeze, which all have a crucial bearing on the ultimate location of ozone precursors and ozone later in the day. Thus, an accurate forecast of the synoptic conditions is also of great importance for the production of ozone forecasts with high temporal and spatial resolution.

Accurate forecasts of local scale flows, such as nocturnal drainage and bay and sea breezes, are even more critical than the forecast of light synoptic flows, because the former provide the necessary gas transport of the ozone cycle, whereas the latter tend only to perturb these local flow patterns. (Strong synoptic flows are not considered here because they tend to eliminate any local winds and clear the region of concentrated pollutants.)

Because NWP models are employed to perform more specific tasks and at an ever-increasing resolution, the relative importance of many components of the NWP system increases. We have shown in the previous sections that flows generated by horizontal density gradients (drainage flows, bay and sea breezes) are of particular importance for the current task. These modeled processes are greatly influenced by the interaction between the model interior and the model surface, where accurate modeling of soil temperature and moisture and surface radiation is of particular importance. Accurate initialization of soil properties is also of importance. Errors in the assimilation and evolution of all these modeled properties and errors in the way the various meteorological processes (e.g., land surface scheme, boundary layer parameterization, radiation) are implemented into the AAQFS, have the potential to significantly affect the ozone forecast. It is also important to have sufficient resolution to be able to resolve the most critical aspects of the flow.

At this stage the relative importance of each of the NWP components is uncertain. We have shown in this study (and other case studies, Cope et al. 1999; Hess et al. 2000a,b, 2004; Tory et al. 2000) that the system is capable of forecasting the mesoscale flows important to the ozone cycle. Preliminary studies investigating the sensitivity of the results to model diffusion and roughness have been performed. As expected, differences in wind speed and direction were relatively small (not shown). However, these small wind differences had a nontrivial impact on the ozone distribution over the bay, because of differences in precursor drainage speeds and the relative positions of the “apparent” divergent point of the bay-breeze winds. In one sensitivity study the diffusion was reduced by a factor of 10 and an experimental effective roughness scheme was applied (greater spatial resolution in the effective roughness led to reduced effective roughness in the vicinity of grid 2). The most highly concentrated day-2 precursor drainage in this case had extended a farther 50 km south on the eastern side of the bay by 1000 AEDT (the time shown in Fig. 5b). This led to a more southerly and easterly position of the ozone plume at 1300 AEDT (see Fig. 11, which shows the ozone structure differences between the standard and sensitivity runs). This example serves to illustrate the sensitivity of the ozone plume to small differences in the NWP component of the AAQFS.

Although these significant differences in ozone plume structure, due to relatively minor differences in the NWP model, suggest a chaotic nature to the problem, we suspect a large level of determinism is still involved. Based on the evidence so far we believe that we should be able to accurately predict air quality on scales smaller than the airshed, but we are not sure that we will be able to attain scales as small as the suburban level. Future sensitivity studies aimed to address this uncertainty will be performed to identify the relative importance of many components of the AAQFS and to identify where best to focus on improving the system. Other studies will investigate the sensitivity of forecasts to small changes in initial conditions. Together these studies will begin to establish the limits of predictability.

Summary and conclusions

A detailed study of a 4-day photochemical smog event for Victoria has been described here. The study focused on the NWP component of the system and identified a complex ozone cycle, involving precursor drainage, ozone production over the bay, onshore ozone advection by the bay and sea breezes, and, finally, ozone clearance by the sea breeze. Sensitivity of ozone plume development to drainage flows and the development of the bay breeze was identified. The system in general performed very well. It captured the diurnal evolution of local winds and the ozone cycle, and the day-to-day variations of the cycle. Variations in the background wind perturbed the local-scale winds that comprised the ozone cycle, and they were largely responsible for the differences in the ozone cycle from day to day. The best example of this occurred on day 3 when a southerly background wind greatly reduced the precursor drainage over the bay, resulting in reduced ozone production there, and the inland transport of relatively clean air by the bay/sea breeze. Point-by-point ozone forecasts were generally good to excellent. The exceptions were attributed to errors in timing and direction of local-scale flows. This study showed that air quality forecasts with high spatial and temporal resolution were within the limits of predictability, and, as the system currently stands, it can be used to provide valuable information to assist in the production of air quality forecasts.

Many components of the NWP model influence the development, structure, and evolution of the local-scale winds that play a crucial role in the ozone cycle. The sensitivity of these components is yet to be examined.

Acknowledgments

The Australian Air Quality Forecasting System was partially funded by the Air Pollution in Major Cities Program (sponsored by Environment Australia through the Natural Heritage Trust).

REFERENCES

  • Abbs, D. J. 1986. Sea-breeze interactions along a concave coastline in southern Australia: Observations and numerical modeling study. Mon. Wea. Rev 114:831848.

    • Search Google Scholar
    • Export Citation
  • Baines, P. G. and P. C. Manins. 1989. The principles of laboratory modeling of stratified atmospheric flows over complex terrain. J. Appl. Meteor 28:12131225.

    • Search Google Scholar
    • Export Citation
  • Cope, M. E. 1999. Mathematical modelling of photochemical smog processes in the Port Phillip control region. Ph.D. thesis, Earth Sciences Department, University of Melbourne, 283 pp.

    • Search Google Scholar
    • Export Citation
  • Cope, M. E., G. D. Hess, S. Lee, M. Azzi, J. Carras, N. Wong, and M. Young. 1999. Development of the Australian air quality forecasting system: Current status. Proc. Int. Conf. on Urban Climatology, Sydney, NSW, Australia, World Meteorological Organization, 595–600.

    • Search Google Scholar
    • Export Citation
  • Cope, M. E. Coauthors 2004. The Australian Air Quality Forecasting System. Part I: Description of the system and statistical verification. J. Appl. Meteor 43:649662.

    • Search Google Scholar
    • Export Citation
  • Environment Protection Authority of Victoria, 1999. Air emissions inventory: Port Phillip region. Environment Protection Authority of Victoria Rep. 632, 48 pp.

    • Search Google Scholar
    • Export Citation
  • Evans, L. F., I. A. Weeks, and A. J. Ecclestone. 1985. Measurements of photochemical precursors in the Melbourne atmosphere. Clean Air 19:2129.

    • Search Google Scholar
    • Export Citation
  • Hess, G. D. 1989a. Photochemical model for air quality assessment: Model description and verification. Atmos. Environ 23:643660.

  • Hess, G. D. 1989b. Simulation of photochemical smog in the Melbourne airshed: Worst case studies. Atmos. Environ 23:661669.

  • Hess, G. D., M. E. Cope, S. Lee, P. C. Manins, G. A. Mills, K. Puri, and K. J. Tory. 2000a. The Australian Air Quality Forecasting System. AMOS Bull 13:6773.

    • Search Google Scholar
    • Export Citation
  • Hess, G. D., M. E. Cope, S. Lee, P. C. Manins, G. A. Mills, K. Puri, and K. J. Tory. 2000b. The development of the Australian Air Quality Forecasting System: Current status. Proc. Millennium NATO/CCMS Int. Technical Meeting on Air Pollution Modeling and Its Application, Boulder CO, Amer. Meteor. Soc., 276–283.

    • Search Google Scholar
    • Export Citation
  • Hess, G. D., K. J. Tory, M. E. Cope, S. Lee, K. Puri, P. C. Manins, and M. Young. 2004. The Australian Air Quality Forecasting System. Part II: Case study of a Sydney 7-day photochemical smog event. J. Appl. Meteor 43:663679.

    • Search Google Scholar
    • Export Citation
  • Kurita, H., K. Sasaki, H. Muroga, H. Ueda, and S. Wakamatsu. 1985. Long-range transport of air pollution under light gradient wind conditions. J. Climate Appl. Meteor 24:425434.

    • Search Google Scholar
    • Export Citation
  • McGregor, J. L. and F. Kimura. 1989. Numerical simulations of mesoscale eddies over Melbourne. Mon. Wea. Rev 117:5066.

  • Mizuma, M. 1995. General aspects of land and sea breezes in Osaka Bay and surrounding areas. J. Meteor. Soc. Japan 73:10291040.

  • Mizuma, M. 1998. General aspects of land and sea breezes in Western Seto Inland Sea and surrounding areas. J. Meteor. Soc. Japan 76:403418.

    • Search Google Scholar
    • Export Citation
  • Ng, Y. L. and M. Minchin. 2000. Spatial and temporal allocation of emissions from wood combustion. Proc. Int. 15th Clean Air Environment Conf., Brighton Beach, NSW, Australia, Clean Air Society of Australia and New Zealand, 288–291.

    • Search Google Scholar
    • Export Citation
  • Pitts, B. J. and J. N. Pitts. 2000. Chemistry of the Upper and Lower Atmosphere: Theory, Experiments and Applications. Academic Press, 969 pp.

    • Search Google Scholar
    • Export Citation
  • Puri, K., G. S. Dietachmayer, G. A. Mills, N. E. Davidson, R. A. Bowen, and L. W. Logan. 1998. The new BMRC Limited Area Prediction System, LAPS. Aust. Meteor. Mag 47:203223.

    • Search Google Scholar
    • Export Citation
  • Simpson, J. E. 1994. Sea Breeze and Local Wind. Cambridge University Press, 234 pp.

  • Spillane, K. T. 1978. Atmospheric characteristics on high oxidant days in Melbourne. Clean Air 12:5056.

  • Tapp, R. G. 1985. Indications of topographically-induced eddies in stratified flow during a severe air pollution event. Bound.-Layer Meteor 33:283302.

    • Search Google Scholar
    • Export Citation
  • Tapp, R. G. 1992. Characteristics of near surface air movement. The meteorology of the Melbourne urban area. Meteorology Section, School of Earth Sciences, University of Melbourne Final Rep. Publ. 32, 222–417.

    • Search Google Scholar
    • Export Citation
  • Tory, K. J., G. D. Hess, G. A. Mills, and K. Puri. 2000. Verification of the meteorological component of the Australian Air Quality Forecasting System. Proc. Int. 15th Clean Air Environment Conf., Brighton Beach, NSW, Australia, Clean Air Society of Australia and New Zealand, 221–226.

    • Search Google Scholar
    • Export Citation
  • Tory, K. J., M. E. Cope, G. D. Hess, S. Lee, and N. Wong. 2003. The use of long-range transport simulations to verify the Australian Air Quality Forecasting System. Aust. Meteor. Mag 52:229240.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Topographic maps centered on Melbourne. (a) Much of the state of Victoria is represented, and (b) a close-up of the Melbourne area, corresponding to the rectangle in (a), is represented. The locations of the eight ozone-monitoring stations are indicated by black triangles, and the location of Melbourne Airport is marked with an aircraft symbol. Height above sea level is contoured and shaded (contour interval 100 m, first contour 50 m). The stations are labeled with the following abbreviations: ALPH = Alphington, BRTN = Brighton, BXHL = Box Hill, GSTH = Geelong South, MTCL = Mount Cottrell, PAIS = Paisley, PTCK = Point Cook, and RMIT = Royal Melbourne Institute of Technology. RMIT is located within the Melbourne central business district. The Melbourne metropolitan area and other pollutant-source regions, including shipping routes, are represented by cross-hatching. The two north–south lines show the location of the Fig. 8 vertical sections

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 2.
Fig. 2.

Synoptic manual analyses produced by the Australian Bureau of Meteorology at 0000 UTC (1100 local summer time) during the Melbourne 4-day photochemical smog event of (a)– (d) 6–9 Mar 2001, respectively. Mean sea level pressure is contoured (interval 4 hPa), and cold fronts (solid line with barbs) and troughs (dashed lines) are represented. Melbourne is located in the southeast of the continent and is marked on the figures by the letter “M.”

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 3.
Fig. 3.

Time–height profiles of potential temperature (contour interval 2 K) and winds (full barb = 10 kt = 5 m s−1), for the 108-h period following 1100 UTC 5 Mar 2001 at Melbourne Airport. The thick gray line is a diagnosed mixed-layer height determined by the height at which the atmosphere is first 1 K warmer than the 10-m potential temperature. (a) Constructed from data collected by commercial aircraft as they take off from Melbourne Airport, and (b) constructed from the NWP model (LAPS) data interpolated to the flight paths and from subsections of four individual LAPS forecasts beginning at 1100 UTC on each day. The vertical lines mark the division between each run. Lines in the same positions have been added to (b) to simplify comparisons between the two

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 4.
Fig. 4.

Grid-2 NOy (shaded) at 2000 UTC (0700 AEDT) on each of the four days. The lowest concentration shaded is 10 ppb. The shading increment above 10 ppb is 30 ppb. The thin regularly spaced vectors represent the modeled 10-m winds, and the thicker vectors represent observed 10-m winds. The Melbourne and Geelong central business districts are labeled “M” and “G,” respectively, and the bay is labeled “Port Phillip.”

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 5.
Fig. 5.

As in Fig. 4 except 3 h later (2300 UTC, 1000 AEDT)

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 6.
Fig. 6.

Grid-2 ozone shaded at 0200 UTC (1300 AEDT local time) on each of the four days. The lowest concentration shaded is 35 ppb. The shading increment above 35 ppb is 15 ppb. The thin regularly spaced vectors represent the modeled 10-m winds, and the thicker vectors represent observed 10-m winds. The Melbourne and Geelong central business districts are labeled “M” and “G,” respectively, and the “bay” is labeled “Port Phillip.”

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 7.
Fig. 7.

As in Fig. 6, except 4 h later (0600 UTC, 1700 AEDT)

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 8.
Fig. 8.

North–south vertical sections of potential temperature (contour interval 1 K) and horizontal winds on day 2. Downward- (upward-) pointed vectors represent flow out of (into) the page. NOy concentrations on a cross section passing through the Brighton monitoring station (Fig. 1b) are shaded at (a) 2000, (b) 2300, and (c) 0200 UTC (0700, 1000, and 1300 AEDT), respectively. Ozone concentrations on a cross section passing close to the Point Cook monitoring station (Fig. 1b) are shaded at (d) 0200, (e) 0400, and (f) 0600 UTC (1300, 1500, and 1700 AEDT), respectively. Note that only the lowest 2000 m of the atmosphere are included in (a)–(c), whereas the lowest 3000 m are included in (d)–(f). The shading intervals are the same as Figs. 4 and 7 (30 ppb starting at 10 ppb for the ozone precursors, and 15 ppb starting at 35 ppb for ozone), respectively. The topography is indicated by the patterned shading. The thick line at the base of the images illustrates the position of the bay

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 9.
Fig. 9.

Ozone time series during the 4-day period. Observations are represented by open squares, and the solid line represents forecast concentrations. The time listed is the number of hours since 0000 AEDT 6 Mar 2001

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 10.
Fig. 10.

As in Fig. 9, except NOy is plotted

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

Fig. 11.
Fig. 11.

Difference in ozone concentrations between the base case and test runs discussed in section 6. Full shading represents test greater than base concentrations, and hatched shading represents base greater than test concentrations. Contour levels are 5, 15, 30, 45, and 60 ppb. Differences less than 5 ppb are not shaded. The wind vectors represent the vector difference between the test and base forecast winds (i.e., base subtracted from test)

Citation: Journal of Applied Meteorology 43, 5; 10.1175/2092.1

1

Ozone precursors in the AAQFS include an approximated form of NOy and reactive organic compounds (ROC). The NOy approximation is based on information available from the highly compressed GRS photochemical mechanism NOy = NOx + SGN, where SGN is the sum of the other stable gaseous nitrate species. Throughout the paper “NOy” refers to the approximated form. [A more complete definition (e.g., Pitts and Pitts 2000, p. 286) is NOy = NOx + PAN + HNO3 + NO3 (particles) + 2 N2O5 + other reactive nitrogen compounds.] Because the ROC precursor pattern (not shown) is similar to that for NOy we have chosen to use only NOy plots when illustrating the precursor production, transport, and dissipation.

Save
  • Abbs, D. J. 1986. Sea-breeze interactions along a concave coastline in southern Australia: Observations and numerical modeling study. Mon. Wea. Rev 114:831848.

    • Search Google Scholar
    • Export Citation
  • Baines, P. G. and P. C. Manins. 1989. The principles of laboratory modeling of stratified atmospheric flows over complex terrain. J. Appl. Meteor 28:12131225.

    • Search Google Scholar
    • Export Citation
  • Cope, M. E. 1999. Mathematical modelling of photochemical smog processes in the Port Phillip control region. Ph.D. thesis, Earth Sciences Department, University of Melbourne, 283 pp.

    • Search Google Scholar
    • Export Citation
  • Cope, M. E., G. D. Hess, S. Lee, M. Azzi, J. Carras, N. Wong, and M. Young. 1999. Development of the Australian air quality forecasting system: Current status. Proc. Int. Conf. on Urban Climatology, Sydney, NSW, Australia, World Meteorological Organization, 595–600.

    • Search Google Scholar
    • Export Citation
  • Cope, M. E. Coauthors 2004. The Australian Air Quality Forecasting System. Part I: Description of the system and statistical verification. J. Appl. Meteor 43:649662.

    • Search Google Scholar
    • Export Citation
  • Environment Protection Authority of Victoria, 1999. Air emissions inventory: Port Phillip region. Environment Protection Authority of Victoria Rep. 632, 48 pp.

    • Search Google Scholar
    • Export Citation
  • Evans, L. F., I. A. Weeks, and A. J. Ecclestone. 1985. Measurements of photochemical precursors in the Melbourne atmosphere. Clean Air 19:2129.

    • Search Google Scholar
    • Export Citation
  • Hess, G. D. 1989a. Photochemical model for air quality assessment: Model description and verification. Atmos. Environ 23:643660.

  • Hess, G. D. 1989b. Simulation of photochemical smog in the Melbourne airshed: Worst case studies. Atmos. Environ 23:661669.

  • Hess, G. D., M. E. Cope, S. Lee, P. C. Manins, G. A. Mills, K. Puri, and K. J. Tory. 2000a. The Australian Air Quality Forecasting System. AMOS Bull 13:6773.

    • Search Google Scholar
    • Export Citation
  • Hess, G. D., M. E. Cope, S. Lee, P. C. Manins, G. A. Mills, K. Puri, and K. J. Tory. 2000b. The development of the Australian Air Quality Forecasting System: Current status. Proc. Millennium NATO/CCMS Int. Technical Meeting on Air Pollution Modeling and Its Application, Boulder CO, Amer. Meteor. Soc., 276–283.

    • Search Google Scholar
    • Export Citation
  • Hess, G. D., K. J. Tory, M. E. Cope, S. Lee, K. Puri, P. C. Manins, and M. Young. 2004. The Australian Air Quality Forecasting System. Part II: Case study of a Sydney 7-day photochemical smog event. J. Appl. Meteor 43:663679.

    • Search Google Scholar
    • Export Citation
  • Kurita, H., K. Sasaki, H. Muroga, H. Ueda, and S. Wakamatsu. 1985. Long-range transport of air pollution under light gradient wind conditions. J. Climate Appl. Meteor 24:425434.

    • Search Google Scholar
    • Export Citation
  • McGregor, J. L. and F. Kimura. 1989. Numerical simulations of mesoscale eddies over Melbourne. Mon. Wea. Rev 117:5066.

  • Mizuma, M. 1995. General aspects of land and sea breezes in Osaka Bay and surrounding areas. J. Meteor. Soc. Japan 73:10291040.

  • Mizuma, M. 1998. General aspects of land and sea breezes in Western Seto Inland Sea and surrounding areas. J. Meteor. Soc. Japan 76:403418.

    • Search Google Scholar
    • Export Citation
  • Ng, Y. L. and M. Minchin. 2000. Spatial and temporal allocation of emissions from wood combustion. Proc. Int. 15th Clean Air Environment Conf., Brighton Beach, NSW, Australia, Clean Air Society of Australia and New Zealand, 288–291.

    • Search Google Scholar
    • Export Citation
  • Pitts, B. J. and J. N. Pitts. 2000. Chemistry of the Upper and Lower Atmosphere: Theory, Experiments and Applications. Academic Press, 969 pp.

    • Search Google Scholar
    • Export Citation
  • Puri, K., G. S. Dietachmayer, G. A. Mills, N. E. Davidson, R. A. Bowen, and L. W. Logan. 1998. The new BMRC Limited Area Prediction System, LAPS. Aust. Meteor. Mag 47:203223.

    • Search Google Scholar
    • Export Citation
  • Simpson, J. E. 1994. Sea Breeze and Local Wind. Cambridge University Press, 234 pp.

  • Spillane, K. T. 1978. Atmospheric characteristics on high oxidant days in Melbourne. Clean Air 12:5056.

  • Tapp, R. G. 1985. Indications of topographically-induced eddies in stratified flow during a severe air pollution event. Bound.-Layer Meteor 33:283302.

    • Search Google Scholar
    • Export Citation
  • Tapp, R. G. 1992. Characteristics of near surface air movement. The meteorology of the Melbourne urban area. Meteorology Section, School of Earth Sciences, University of Melbourne Final Rep. Publ. 32, 222–417.

    • Search Google Scholar
    • Export Citation
  • Tory, K. J., G. D. Hess, G. A. Mills, and K. Puri. 2000. Verification of the meteorological component of the Australian Air Quality Forecasting System. Proc. Int. 15th Clean Air Environment Conf., Brighton Beach, NSW, Australia, Clean Air Society of Australia and New Zealand, 221–226.

    • Search Google Scholar
    • Export Citation
  • Tory, K. J., M. E. Cope, G. D. Hess, S. Lee, and N. Wong. 2003. The use of long-range transport simulations to verify the Australian Air Quality Forecasting System. Aust. Meteor. Mag 52:229240.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Topographic maps centered on Melbourne. (a) Much of the state of Victoria is represented, and (b) a close-up of the Melbourne area, corresponding to the rectangle in (a), is represented. The locations of the eight ozone-monitoring stations are indicated by black triangles, and the location of Melbourne Airport is marked with an aircraft symbol. Height above sea level is contoured and shaded (contour interval 100 m, first contour 50 m). The stations are labeled with the following abbreviations: ALPH = Alphington, BRTN = Brighton, BXHL = Box Hill, GSTH = Geelong South, MTCL = Mount Cottrell, PAIS = Paisley, PTCK = Point Cook, and RMIT = Royal Melbourne Institute of Technology. RMIT is located within the Melbourne central business district. The Melbourne metropolitan area and other pollutant-source regions, including shipping routes, are represented by cross-hatching. The two north–south lines show the location of the Fig. 8 vertical sections

  • Fig. 2.

    Synoptic manual analyses produced by the Australian Bureau of Meteorology at 0000 UTC (1100 local summer time) during the Melbourne 4-day photochemical smog event of (a)– (d) 6–9 Mar 2001, respectively. Mean sea level pressure is contoured (interval 4 hPa), and cold fronts (solid line with barbs) and troughs (dashed lines) are represented. Melbourne is located in the southeast of the continent and is marked on the figures by the letter “M.”

  • Fig. 3.

    Time–height profiles of potential temperature (contour interval 2 K) and winds (full barb = 10 kt = 5 m s−1), for the 108-h period following 1100 UTC 5 Mar 2001 at Melbourne Airport. The thick gray line is a diagnosed mixed-layer height determined by the height at which the atmosphere is first 1 K warmer than the 10-m potential temperature. (a) Constructed from data collected by commercial aircraft as they take off from Melbourne Airport, and (b) constructed from the NWP model (LAPS) data interpolated to the flight paths and from subsections of four individual LAPS forecasts beginning at 1100 UTC on each day. The vertical lines mark the division between each run. Lines in the same positions have been added to (b) to simplify comparisons between the two

  • Fig. 4.

    Grid-2 NOy (shaded) at 2000 UTC (0700 AEDT) on each of the four days. The lowest concentration shaded is 10 ppb. The shading increment above 10 ppb is 30 ppb. The thin regularly spaced vectors represent the modeled 10-m winds, and the thicker vectors represent observed 10-m winds. The Melbourne and Geelong central business districts are labeled “M” and “G,” respectively, and the bay is labeled “Port Phillip.”

  • Fig. 5.

    As in Fig. 4 except 3 h later (2300 UTC, 1000 AEDT)

  • Fig. 6.

    Grid-2 ozone shaded at 0200 UTC (1300 AEDT local time) on each of the four days. The lowest concentration shaded is 35 ppb. The shading increment above 35 ppb is 15 ppb. The thin regularly spaced vectors represent the modeled 10-m winds, and the thicker vectors represent observed 10-m winds. The Melbourne and Geelong central business districts are labeled “M” and “G,” respectively, and the “bay” is labeled “Port Phillip.”

  • Fig. 7.

    As in Fig. 6, except 4 h later (0600 UTC, 1700 AEDT)

  • Fig. 8.

    North–south vertical sections of potential temperature (contour interval 1 K) and horizontal winds on day 2. Downward- (upward-) pointed vectors represent flow out of (into) the page. NOy concentrations on a cross section passing through the Brighton monitoring station (Fig. 1b) are shaded at (a) 2000, (b) 2300, and (c) 0200 UTC (0700, 1000, and 1300 AEDT), respectively. Ozone concentrations on a cross section passing close to the Point Cook monitoring station (Fig. 1b) are shaded at (d) 0200, (e) 0400, and (f) 0600 UTC (1300, 1500, and 1700 AEDT), respectively. Note that only the lowest 2000 m of the atmosphere are included in (a)–(c), whereas the lowest 3000 m are included in (d)–(f). The shading intervals are the same as Figs. 4 and 7 (30 ppb starting at 10 ppb for the ozone precursors, and 15 ppb starting at 35 ppb for ozone), respectively. The topography is indicated by the patterned shading. The thick line at the base of the images illustrates the position of the bay

  • Fig. 9.

    Ozone time series during the 4-day period. Observations are represented by open squares, and the solid line represents forecast concentrations. The time listed is the number of hours since 0000 AEDT 6 Mar 2001

  • Fig. 10.

    As in Fig. 9, except NOy is plotted

  • Fig. 11.

    Difference in ozone concentrations between the base case and test runs discussed in section 6. Full shading represents test greater than base concentrations, and hatched shading represents base greater than test concentrations. Contour levels are 5, 15, 30, 45, and 60 ppb. Differences less than 5 ppb are not shaded. The wind vectors represent the vector difference between the test and base forecast winds (i.e., base subtracted from test)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 549 382 62
PDF Downloads 215 173 44