Classification of Meteorological Patterns Associated with the Ozone Categories in Kelowna, British Columbia

Transcription

1 MARCH 2000 SCHWARZHOFF AND REID 463 Classification of Meteorological Patterns Associated with the Ozone Categories in Kelowna, British Columbia PETER J. SCHWARZHOFF Environment Canada, Kelowna, British Columbia, Canada PETER D. REID* British Columbia Ministry of Environment, Lands, and Parks, Kamloops, British Columbia, Canada (Manuscript received 1 June 1998, in final form 25 January 1999) ABSTRACT To develop a highly skillful, semiautomated means of forecasting periods of high ozone concentrations, it is critical to understand how ozone concentrations depend on weather patterns, especially in extreme cases. In this work, the synoptic-scale weather patterns that are optimal for photochemical ozone production in the southern interior region of British Columbia are discerned. Rather than generate meteorological composites for days on which an arbitrary 1-h ozone concentration is exceeded, the authors isolate these days through use of the diurnalcurve space defined by M. Böhm et al. This novel technique defines a set of diurnal ozone patterns to characterize ozone exposure at a location. It is applied to hourly ozone data in Kelowna, British Columbia, for the period Each day in the May October ozone season was categorized using the Böhm classification. The meteorological data for each group of dates were then aggregated to generate a set of corresponding composite atmospheric patterns. These composite synoptic maps represent the typical meteorological scenarios that lead to the corresponding diurnal ozone curve experienced at Kelowna. Of special interest are the C2 (urban/small) and C4 (urban/medium) categories. 1. Introduction To fulfill a growing demand for air quality forecasts, researchers have investigated a number of automated forecast techniques. The goal is to develop a highly skillful, automated means of determining periods of increased ozone concentrations, applying human expert input only when appropriate. To accomplish this goal, it is critical to understand how ozone concentrations depend on weather patterns, especially in extreme cases. This work attempts to uncover the synoptic-scale weather patterns that produce different diurnal trends in ozone concentrations and thus determine which cases require more attention. In-depth analysis of the meteorological conditions of each case of high ozone concentration will be the basis of future work. To classify ozone regimes for the study area, classic * Current affiliation: Westcoast Energy, Inc., Fort St. John, British Columbia, Canada. Corresponding author address: Peter Schwarzhoff, Mountain Weather Services Office, Environment Canada, 3140 College Way, Kelowna, BC V1V 1V9, Canada. peter.schwarzhoff@ec.gc.ca univariate statistics were discarded in favor of the diurnal-curve space developed by Böhm et al. (1991; 1995a,b). Böhm s technique was used to classify the ozone regimes at Kelowna, British Columbia, for the period 1984 through 1996 by Reid and Hepworth (1997). This dataset is used in the current work to determine the degree of photochemical activity. Exceedance frequencies are of limited use for this goal. Kelowna, with a 1996 population of , is located in the semiarid, intermountain region of western Canada. It lies 250 km east of Vancouver and is separated from the Pacific Ocean by the 3000-m-high Coast mountain range. A rapid population increase within a narrow mountain valley and a parallel growth in precursor emissions threaten to deteriorate air quality greatly over time at this location. By tracking the spatial and temporal variations of ozone regimes, the authors seek to identify meaningful trends in the data before the area experiences regular occurrences of ozone concentrations elevated above background. Taylor (1991), Lord (1993), McCollor (1993), and Saunders (1997) illustrate the progressive trend toward the application of increasingly sophisticated statistical modeling techniques to the prediction of air quality factors. The latest methodology to be employed by Environment Canada is a nonlinear regression version of

2 464 JOURNAL OF APPLIED METEOROLOGY VOLUME 39 Classification and Regression Trees (CART) entitled CANFIS (CART Neuro-Fuzzy Inference System) (Burrows 1995, 1997). Although not yet thoroughly tested, even this advanced expert system is expected to exhibit the fundamental disadvantage of statistical forecasts, which are weakest at the times when their output is most valuable the extreme cases. The evaluation report of the first operational trials of the CANFIS methodology states that this model should not be used to attempt to predict values outside of the extremes contained in the learning database, and indeed that the verification statistics indicate the model shows low skill in predicting extreme values (Cote et al. 1998). Taylor and Lord also discuss the meteorological scenarios that produce increased ozone concentrations. The factors leading to increased radiative loss and reduced dispersion are well known to meteorologists. A compilation of these scenarios together with a history of efforts to analyze them objectively can be found in Taylor (1991) and Elder (1994). Although Kelowna does not share the marine influences experienced in Vancouver, its close proximity means that the two cities share similar synoptic atmospheric circulation patterns. Several previous efforts have been made to relate the ozone concentrations recorded at Vancouver, British Columbia, to the synoptic weather patterns, each effort with its own methodology. Taylor (1991), for example, subjectively examines the meteorological conditions for each observed exceedance of the national standard and attempts to draw conclusions about the synoptic requirements for increased ozone concentrations. Mc- Kendry (1994) first applies the Kirchhofer synoptic classification procedure for surface and 500-hPa data over the region and then relates the observed exceedances to the resultant classifications. Pryor et al. (1995) used principal component analysis of three meteorological fields to examine the impact of changes in the frequency of dominant atmospheric circulation patterns on ozone concentrations. Two fundamental approaches to synoptic classification were identified by Yarnal (1993). The circulation-to-environment approach first identifies synoptic regimes and then relates these to observed environmental variables (Yarnal 1984; El-Kadi 1992). This work uses the second, or environment-to-circulation, approach first to identify patterns of ozone concentrations and then to determine associated synoptic weather patterns. A greater understanding of the underlying meteorological conditions associated with different ozone regimes exists, and this greater knowledge can be applied in an operational ozone forecast setting; in this work, however, detailed case studies to dissect the meteorological conditions of each instance of increased ozone were not performed. Efforts were directed only at uncovering those synoptic-scale weather scenarios that are typical of certain diurnal ozone patterns to identify the scenarios that require greater scrutiny. A reliable link between high ozone concentration days and synoptic Averaging period TABLE 1. Environment Canada ozone objectives. Level A [ g m 3 (ppb)] Level B [ g m 3 (ppb)] Level C [ g m 3 (ppb)] 1 h 100 (51) 160 (82) 300 (153) 24 h 30 (15) 50 (25) meteorological patterns is required so that air quality management plans can be implemented proactively. 2. Ozone regime classification Traditional analyses of ozone data, such as the exceedance of objective thresholds (1-h mean, 24-h mean, annual mean) reveal very little about the differences between sites with substantially different ozone regimes. For example, using traditional ozone summaries alone, it may be impossible to distinguish large urban areas themselves from areas affected by transport from large urban areas. The ozone summary for a small urban area may be indistinguishable from that of a remote rural site. Researchers employ a wide variety of summarization techniques to gain a better understanding of ozone dynamics over space and time. A common technique employs exceedance summaries (1-h, 24-h, etc.) and histograms. Utilization of the concept of diurnal-curve space is less common. All are important methods of analysis, but each yields different kinds of information about ozone behavior. The 1- and 24-h exceedance summaries are simple, univariate statistical methods while the technique using diurnal-curve space is based on a variety of multivariate statistical methods. The 1- and 24-h exceedances provide information on how often the concentration of ozone exceeds a certain threshold. The exceedances are calculated by comparing 1- and 24-h (midnight to midnight) measurements to the air quality objectives established by Environment Canada. These objectives are defined as maximum desirable (A), maximum acceptable (B), and maximum tolerable (C) levels. The concentrations assigned to these objectives (referred to hereinafter as levels A, B, and C) for two averaging periods are given in Table 1. The percent frequency of exceedance of the above objectives for Kelowna for the period 1984 through 1996 is given in Table 2. The frequency of exceedances observed in Kelowna fluctuates considerably from year to year. For example, over the period from 1984 to 1996, Kelowna experienced a range of from 8 to 288 exceedances of the 1-h level-a objective. The year with the most 1-h level-a exceedances in Kelowna was There were only 2 yr where 1-h level-b exceedances were recorded in Kelowna: one in July 1984 and two in August By any classic definition (e.g., 1-h level-b exceedance frequency) Kelowna does not have a problem with ozone pollution. The application of diurnal-curve space involves ex-

3 MARCH 2000 SCHWARZHOFF AND REID 465 TABLE 2. Summary of exceedance frequencies for Kelowna, British Columbia ( ). Year 1-h objectives Level A (h yr 1 ) Level B (h yr 1 ) 24-h objectives Level A (days yr 1 ) Level B (days yr 1 ) Total Annual average Annual percent 1.12% 0.00% 53.42% 22.80% amining the form of the daily ozone concentration curve. The diurnal pattern of hourly ozone concentrations provides insight into the formation, scavenging, and transport of ozone at a site or aloft within an area. Diurnal-curve space is described as a two-dimensional space in which 64% of the variance in ozone exposure regimes can be expressed (Böhm et al. 1991; 1995a,b). Use of diurnal-curve space allows large amounts of data to be reduced with limited loss of information. The various univariate statistical techniques fail to preserve information on common phenomena such as ozone scavenging. In an urban area, scavenging [ozone destruction by reaction with nitric oxide (NO)] usually takes place overnight so that nighttime hours typically have very low ozone concentrations and afternoon hours have high concentrations. In remote regions, scavenging by NO is limited, leaving only deposition as a sink process, and so ozone concentrations are almost constant throughout the day and night. With the use of univariate statistical methods, this scavenging phenomenon is obscured. Scavenging is a feature of many of the urban curves described by Böhm et al. (1991) and is important to discern from other phenomena. Air quality objectives are useful for communicating health risks to the public; however, the use of exceedance frequencies is adequate for developing an understanding of long-term trends and regional patterns. For example, if there are no exceedances, it is implied that ozone pollution is not a problem. An area with one or two exceedances every 2 3 yr is assumed to be better off than an area with several exceedances every year. Through use of diurnal-curve space, it is possible to distinguish between mild ozone episodes and severe ozone episodes. Diurnal-curve space also allows for comparison between areas that is based on the number of days that photochemical reactions are not occurring. It can reveal that an area is on the threshold of routinely exceeding the 1-h level-b objective. None of these details are revealed through simple univariate statistical analyses. 3. Kelowna s ozone regime Reid and Hepworth s (1997) study examined hourly ozone averages from Kelowna over the period Only data from the May October ozone season were examined, to remain consistent with Böhm et al. s work. These data were run through a pattern-matching algorithm coded in Visual Basic and derived from Böhm s work (Böhm et al. 1991). Each day in the 13-yr record was classified as one of Böhm s 17 characteristic diurnal curves. Four representative curves are shown in Fig. 1. These curves are given a letter and number combination, which varies from A curves exhibiting little diurnal fluctuation to E curves having a large diurnal fluctuation and from 1 curves having low relative variations in ozone concentration to 6 curves having the highest relative daily variation. Each of these curves is assigned to one of six categories. The relative frequency of occurrence of each of these categories determines which of eight regimes is representative of a given site. Reid and Hepworth have determined that Kelowna fits into the small-urban regime, dominated by urban/ small categories (C1, C2, and C3 curves) and remote categories (B1, B2, A3, and A4), but with some urban/ medium (C4) and urban/transport categories (B5). Other categories were not identified at Kelowna during the period. A summary of the frequency of occurrence of these categories is given in Table 3. The C2 characteristic ozone curve is a member of Böhm s urban/small category. The vertical axis shows hourly mean ozone concentration (ppb) plus standard deviation, and the horizontal axis is the time of day (h). The C designates an average diurnal variation in ozone concentration, and the 2 designates the second lowest category of the relative variation of the 24 hourly ozone concentrations during the day. One goal of the Reid and Hepworth study was to make a comparison with the 63 sites characterized by Böhm et al. (1995b). Management decisions taken by U.S. land managers in airsheds with ozone regimes similar to that of Kelowna might then offer guidance for local airshed management decisions. A direct comparison of the ozone category mix in Kelowna with those of 10 U.S. small-urban sites identified by Böhm et al. (1995b) resulted in four close, but no exact, matches. This lack of exactness may be attributable in part to mismatched study intervals (Reid and Hepworth: , Böhm et al.: ). The closest matches to Kelowna were Marion County, Medford, and Eugene, all located in Oregon. Kelowna most closely matched Marion County. The relative mixes of categories for these four sites are given in Table 4. Reid and Hepworth also determined through a climatological analysis that there is a strong relationship

4 466 JOURNAL OF APPLIED METEOROLOGY VOLUME 39 FIG. 1. Representative diurnal curves selected from Böhm et al. (1995b). The ordinates are ozone concentrations (ppb) and the abscissas are time of day (LST). Vertical lines represent std dev of ozone concentrations. between generally hot and sunny ozone seasons and the prevalence of urban/small and urban/medium curves. Although there were only three exceedances of the 1-h level-b objective logged in Kelowna during the period, there were 23 C4 curves that tended to cluster in the hottest summers (1987, 1994). Observations from this past year confirm this trend. In 1998 Kelowna experienced three C4 curves. The summer of 1998 ranks as one of the hottest on record in Kelowna. It also has been marked by unprecedented ozone exceedances. There were three exceedances of the level-b hourly objective on a single day. Also occurring was the highest hourly ozone concentration reading recorded in 15 yr of operation (96 ppb; 7 ppb above the previous record high) at that site. Given a trend of increasing ozone precursor concentrations caused by rapid population growth and given the warmer summers predicted by global warming studies (Taylor 1997), Kelowna can be expected to experience greater ozone pollution in the future. Unless there is some dramatic change in either population dynamics, precursor emissions, or climate, the frequency of C4 curves will increase at Kelowna. The appearance of more C4 curves, a rise in the urban/small curves (C1 C3), and a commensurate fall in remote curves (B1, B2, A3, and A4) will push Kelowna from a small-urban regime to a medium-urban regime, representing a substantial deterioration in air quality. 4. Synoptic patterns associated with the ozone categories The Reid and Hepworth (1997) dataset consisting of Böhm characteristic patterns for the period was sorted by classification. Gridded meteorological data at TABLE 3. Frequency of occurrence of diurnal curve classifications for Kelowna, British Columbia ( ). Class Days in class Days total Frequency (%) Remote Inversion Urban/medium Urban/small Urban/large Urban/transport TABLE 4. Relative mix of Böhm et al. categories for selected smallurban regime sites (%). Kelowna Marion County Medford Eugene Remote Inversion Urban/medium Urban/small Urban/large Urban/transport

5 MARCH 2000 SCHWARZHOFF AND REID 467 FIG. 2. Composite 500-hPa (a) height field (dam) occurring for the C2 urban/small diurnal curves and (b) mean climate corresponding to the days of the year that yield the C2 urban/small composite, as explained in the text. several standard pressure levels were composited using 0000 UTC data from all dates in each of the categories. The synoptic composites were compiled from meteorological data from the National Meteorological Center (NMC) available on a CD-ROM produced jointly by the National Center for Atmospheric Research (NCAR) and the Department of Atmospheric Sciences at the University of Washington (UoW). Data in this set come from varying periods of record, but all data cover the period used in this study. Details are found in the documentation that accompanies the CD-ROM set (NCAR and UoW 1996). The synoptic composite patterns represent the typical meteorological scenarios leading to the Böhm et al. ozone categories experienced at Kelowna. Many studies of the connections between surface ozone and synoptic circulation have examined surface pressure patterns; McKendry (1994) suggested that the synoptic patterns be examined in three dimensions. Composite 500-hPa height charts were constructed for all of the classifications identified at Kelowna. Additional composites have been constructed of selected fields such as mean sea level pressure and relative humidity for those classifications deemed to be of special interest. Not all were considered in detail. The remote case, for example, was not explored, since so many different meteorological scenarios could lead to this classification. The authors have elected to illustrate the synoptic situation through charts of the geopotential height of the 500-hPa level. Charts of mean sea level (MSL) pressure are not included here to conserve space but also because mesoscale features in mountainous terrain make the identification of typical MSL pressure patterns complex. Charts of mean climate pattern are created from data for all years in the NMC dataset corresponding to each yearday used in creating each ozone category composite. For example, if 4 June 1991 was used in creating a composite, then all 4 Junes in the NMC dataset are used to create a chart of mean meteorological conditions. Figure 2a shows the composite of 500-hPa height pattern associated with the C2 ozone profile, one of three members of the urban/small (C1, C2, and C3) ozone category. This category typically is observed at Kelowna during times when meteorological conditions are conducive to ozone formation. A weak, blocking upper ridge over the area maintains wide-scale subsidence, creating an inversion near the top of the valley that limits vertical dispersion. Light winds limit horizontal transport and dispersion while clear skies promote photochemical ozone production. The upper ridge axis should be over Kelowna to produce this diurnal ozone curve, rather than over eastern British Columbia as is shown to be normal in the mean 500-hPa composite (Fig. 2b), and must be of greater amplitude (5720 m over Kelowna vs 5680 m average). Figure 3a illustrates the composite 500-hPa patterns that lead to Böhm s urban/medium (C4) classification. Figure 3b shows the corresponding 500-hPa mean climate. Here, the upper ridge is much sharper than that illustrated in Fig. 2a, with heights about 60 m greater. The associated subsidence inversion is much stronger. In an air quality forecast program, a forecast of this scenario would warrant a closer analysis by an air quality meteorologist. The NMC gridpoint dataset includes mean relative humidity in the hPa layer but not mixing ratio. During periods of increased ozone concentration, the air mass typically is dry in this layer, with mean values near 40% relative humidity for the C4 composite

6 468 JOURNAL OF APPLIED METEOROLOGY VOLUME 39 FIG. 3. Same as Fig. 2 but for the C4 urban/medium diurnal curves. (Fig. 4a), as compared with the 50% 55% relative humidity range found in the C2 composite (Fig. 4b). Although relative humidity is not the best choice as a measure of atmospheric moisture because of its inverse relationship to temperature, experience in weather forecasting shows that decreased relative humidity is another indicator of the increased strength of the subsidence, which causes the air to become warmer and dryer. This enhanced subsidence leads to a reduced mixing layer and ensures a cloud-free sky. Elder (1994) also found a high negative correlation between dewpoint and ozone concentration in his study. The urban/medium classification occurs infrequently at Kelowna. Global climate model predictions (Taylor 1997) have raised concern that high-amplitude ridges will become more frequent and long lived, leading to an increase in the number of days with hot, dry, and stagnant conditions, and, thus, to deteriorating air quality in this area (Thomson 1997). The C4 classification is based upon a large diurnal variation with a fairly high maximum ozone concentration. A likely mechanism that could contribute to such a pattern would be the maintenance of increased ozone concentrations in a residual layer above the nocturnal FIG. 4. (a) Composite mean relative humidity (%) in the hPa layer for the (a) C4 urban/medium diurnal curves and (b) C2 urban/small diurnal curves.

7 MARCH 2000 SCHWARZHOFF AND REID 469 FIG. 5. Composite 500-hPa height field (dam) occurring for a subset of C4 urban/medium diurnal curves. inversion, and the subsequent mixing of ozone to the surface on the following morning. There unfortunately have been no vertical ozone profiles measured at this location with which to demonstrate this process. The urban/medium (C4) category was examined in greater detail, since all exceedances of the 1- and 24-h ozone limits observed at Kelowna over the period fall into this category. As with other categories, the pattern shown in Fig. 3a easily could have been anticipated by air quality meteorologists. The composites, for the most part, illustrate what had been expected, but not previously verified. It must be remembered that composites such as this one may contain inherent flaws. It is possible that what is represented as a typical scenario may in fact be an aggregate of disparate atmospheric patterns (Yarnal 1993). This likelihood increases as the sample size is reduced. The C4 composite is based on only 23 occurences. Manual inspection of the 500-hPa field for each date upon which the C4 classification was identified showed a subset (four occurrences) of synoptic patterns that combine to produce the composite shown in Fig. 5. This synoptic pattern was not anticipated, since despite the high heights (greater than 60 m above normal), an upper low in this location would suggest to a meteorologist that conditions at Kelowna would be unsettled, with bands of cloud and showers circulating northeastward in the flow about the upper low not ideal for the production and concentration of ozone. It also might be expected that there would be sufficient atmospheric motion to remove ozone and its precursors from the source area. Closer examination revealed that the upper low off the coast was stationary for several days and had pumped warm air over the Kelowna area, generating stable, stagnant conditions near the surface. Frequently, under this scenario, the low also pumps moisture over the area, combined with a series of short-wave troughs that typically trigger convection based above the stable surface layer. In the cases composited here, however, conditions remained clear, warm, and dry, with light surface winds. The urban/transport category also was inspected closely. Although ozone or its precursors can be transported a considerable distance from large sources (such as Vancouver or the Pacific Northwest), the authors speculate that the 3000-m mountains effectively bar such transport to Kelowna. It is more likely that Kelowna may be responsible for its own ozone transport. Under stable conditions, katabatic anabatic winds along the valley slopes, land lake breezes, and the formation of small, thermally induced low pressure systems either to the north or south can lead to extremely complex wind patterns that could transport the ozone up or down the valley, out over the lake, or into a nocturnal residual layer aloft, only to recirculate it back to the source some time later. Smaller sources to the north and south also likely play a role in this case. The latter discussion highlights the shortcomings of synoptic pattern typing. A forecast of increased ozone concentrations does not depend solely upon synopticscale meteorological conditions, but also upon mesoscale meteorological conditions, complex photochemical reactions, and varying precursor generation. Pattern typing, however, can highlight the occasions when the smaller scales warrant investigation. 5. Conclusions A set of synoptic composites has been constructed for each curve of the Böhm et al. urban/small and urban/ medium ozone categories that were identified by Reid and Hepworth (1997) at Kelowna during the May October ozone season. The remote category composite is deemed to be of little value because of the diversity of meteorological scenarios that lead to this classification. The urban/transport category composite has not yet yielded new insight into the mechanisms responsible for the classification. The urban/small category composite illustrates that (as expected) a weak, persistent upper ridge with a ridge axis over or slightly west of the area typically leads to increased ozone concentrations. The urban/medium category composite shows that a sharp, high-amplitude, upper ridge with a ridge axis over or slightly east of the area provides the meteorological conditions necessary to produce ozone exceedances. The only surprise of this exercise was in a subset of this classification in which an atypical scenario was identified. Further investigations of this scenario will be undertaken, as will be further studies into mesoscale meteorological effects such as recirculation of ozone by local winds.

8 470 JOURNAL OF APPLIED METEOROLOGY VOLUME 39 This set of synoptic meteorological composites has begun to assist in determining the underlying causes of past events leading to the identified ozone categories. An analysis of the differences between the meteorological composites of the urban/small and urban/medium categories forms the most important outcome of this work. The composites will act as analogs in future air quality forecasting programs in that the air quality meteorologist will be able to focus attention on synoptic patterns that are known to lead to unacceptable ozone concentrations. REFERENCES Böhm, M., B. McCune, and T. Vandetta, 1991: Diurnal curves of tropospheric ozone in the western United States. Atmos. Environ., 25A, ,, and, 1995a: Ozone regimes in or near forests of the western United States: 1. Regional patterns. J. Air Waste Manage. Assoc., 45, ,, and, 1995b: Ozone regimes in or near forests of the western United States: Part 2. Factors influencing regional patterns. J. Air Waste Manage. Assoc., 45, Burrows, W. R., M. Benjamin, E. R. Lord, D. McCollor, and B. Thomson, 1995: CART decision-tree statistical analysis and prediction of summer season maximum surface ozone for the Vancouver, Montreal, and Atlantic regions of Canada. J. Appl. Meteor., 34, , J. Montpetit, and J. Pudykiewz, 1997: CANFIS: A non-linear regression procedure to produce dynamic statistical air-quality forecast models. Environment Canada Tech. Rep. 97-TP2B.4, 15 pp. [Available from Environment Canada, Meteorological Research Branch, 4905 Dufferin St., Downsview, ON M3H 5T4, Canada.] Cote, C., M. C. Howe, and D. Waugh, 1998: Southern New Brunswick smog prediction pilot project, 1997 evaluation report. Environment Canada publication, Cat. No. En56-88/1998E. [Available from New Brunswick Weather Centre, 77 Westmorland St., Suite 40, Fredericton, NB E3B 6Z3, Canada.] Elder, B. K., J. M. Davis, and P. Bloomfield, 1994: An automated classification scheme designed to better elucidate the dependence of ozone on meteorology. J. Appl. Meteor., 33, El-Kadi, A. K. A., and P. A. Smithson, 1992: Atmospheric classifications and synoptic climatology. Prog. Phys. Geogr., 16, Lord, E., 1993: Forecasting daily maximum ground-level ozone concentrations in greater Vancouver and the Lower Fraser Valley. Environment Canada, Pacific Region internal report PAES-93-3, 8 pp. [Available from Environment Canada, 1200 W 73rd Ave., Vancouver, BC V6P 6H9, Canada.] McCollor, D., 1993: An investigation of daily maximum ground-level ozone concentration forecasting using CART statistical techniques. Environment Canada, Pacific Region internal report PAES-93-2, 10 pp. [Available from Environment Canada, 1200 W 73rd Ave., Vancouver, BC V6P 6H9, Canada.] McKendry, I. G., 1994: Synoptic circulation and summertime groundlevel ozone concentrations at Vancouver, British Columbia. J. Appl. Meteor., 33, NCAR and UoW, 1996: National Meteorological Center Grid Point Data Set version 3. [Available from Data Support Section, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO ] Pryor, S. C., I. G. McKendry, and D. G. Steyn, 1995: Synoptic-scale meteorological variability and surface ozone concentrations in Vancouver, British Columbia. J. Appl. Meteor., 34, Reid, P., and C. Hepworth, 1997: Tropospheric ozone regimes in the semi-arid interior of British Columbia. BC Ministry of Environment Internal Document, 98 pp. [Available from BC Ministry of Environment, Lands and Parks, 1259 Dalhousie Dr., Kamloops, BC V2C 5Z5, Canada.] Saunders, B., 1997: Air quality prediction initiative. Environment Canada Internal Discussion paper, 10 pp. [Available from Environment Canada, 3140 College Way, Kelowna, BC V1V 1V9, Canada.] Taylor, B., 1997: Climate change scenarios for British Columbia and Yukon. Responding to Global Climate Change in British Columbia and Yukon, E. Taylor and B. Taylor, Eds., Environment Canada, A1-1 A1-32. Taylor, E., 1991: Air quality forecast options for AES in British Columbia. Environment Canada, Pacific Region Internal Report PAES-91-1, 12 pp. [Available from Environment Canada, 1200 W 73rd Ave., Vancouver, BC V6P 6H9, Canada.] Thomson, B. 1997: Impacts of climate change on air quality in British Columbia and Yukon. Responding to Global Climate Change in British Columbia and Yukon, E. Taylor and B. Taylor, Eds., Environment Canada, Yarnal, B., 1984: A procedure for the classification of synoptic weather maps from gridded atmospheric pressure surface data. Comp. Geosci., 10, , 1993: Synoptic Climatology in Environmental Analysis. 1st ed. Belhaven Press, 195 pp.