VT DEC Air Trajectory Analysis of

Long-Term Ozone Climatology

Status Report to OTAG Air Quality Analysis Workgroup: 11/7/96

Rich Poirot and Paul Wishinski, VT DEC


This provides an update of a backward air trajectory analysis project conducted for the OTAG Air Quality Analysis Workgroup by VT DEC. Methods and previous results (based primarily on "location-based trajectory sorting") have been reported at: http://capita.wustl.edu/otag/Reports/Restime/Restime.html and http://capita.wustl.edu/otag/Reports/vtdecair/vtdecair.html. The current report presents results of "concentration-based trajectory sorting" for 7 Summers (1989-95) for 23 ozone monitoring sites in the OTAG domain, for selected subsets of the ozone distribution (lower 50th percentile and upper 50th, 25th, 10th and 5th percentile concentrations). The results are generally consistent with previous "location-based sorting" results, and suggest that: the OTAG region is geographically well-configured to address zone transport; local stagnation is most important in the South; and that industrial Midwest exhibits a high probability contributing to peak ozone concentrations at many sites throughout the OTAG domain.


Concentration-Based Trajectory Sorting

The "Residence Time Analysis" techniques employed in these studies involves tracking, sorting and aggregating large numbers of backward HY-SPLIT trajectory calculations on a grid of 1440 80x80 km squares. Trajectories arriving at a specific receptor site are sorted either as a function of prior trajectory location, or as a function of ozone concentration at the receptor. Location-based sorting involves calculation of a descriptive statistic (for example, the mean ozone concentration, or, more recently, the mean Z-score derived from the geometric mean and standard deviation of the ozone concentration) for all trajectories passing through a specific grid square en route to the receptor. Resulting plots address the question: if the air has previously been here (or there), what's the average ozone concentration at this (or that) receptor? Results of location-based sorting are described in vtdecair.html. Advantages of location-based sorting include: use of all available trajectories (large numbers) and quantitative results (if the air was here, ozone at the receptor averaged 10 ppb higher than average). Disadvantages include: relevance only to changes in average (rather than peak) ozone, and inability to identify potential contributions from areas close to the receptor. This latter limitation is due to the constraint that all trajectories arriving at a site must pass through the grid square containing the receptor, through 1 of 8 squares surrounding the terminal square, etc. Consequently, the average ozone for trajectories passing over grid squares containing or close to the receptor site will be similar to the average at the monitoring site (ie ozone in air near Whiteface Mtn. will have ozone concentrations similar to Whiteface Mtn.). An additional disadvantage is a disregard of trajectory frequency - a relatively distant location may be 'conditionally' associated with high ozone at the receptor, but may be infrequently upwind.

Concentration-based sorting involves sorting trajectories as a function of associated ozone concentration at the receptor. Space/time characteristics of trajectories associated with subsets of high (low, very high, etc.) ozone concentrations are plotted as "residence-time probabilities". Resulting plots address the question: if ozone at the receptor was high (or low or very high), where did the air come from? One advantage of concentration-based sorting is an ability to focus on different portions of the ozone distribution (including episodes). In addition, the results reflect both concentration and frequency associated with upwind locations. A potential disadvantage is that the highest residence-time probability, regardless of which subset of trajectories is examined, will always be for grid squares containing, or adjacent to the receptor square (all trajectories must pass through the terminal square). In this analysis, we attempt to remove this "near bias" by comparing residence-time probabilities for specific subsets of the receptor ozone concentration with residence-time probabilities for all trajectories arriving at the receptor, regardless of associated ozone concentration. This concept is referred to as "incremental probability", and addresses the question: how much greater than "normal" is the probability a location was upwind if the ozone was high? Techniques for calculating residence-time probabilities are described in: Poirot and Wishinski ((1986) Atmos. Environ., 20: 1457­1469) and Wishinski and Poirot ((1996) at http://capita.wustl.edu/otag/Reports/Restime/Restime.html). An "everyday" probability field is calculated for each monitoring site by summing, for each grid square, the total hours that all trajectories over the past 7 summers spent over that grid square. That grid square's everyday probability is expressed as the ratio of residence-time hours in that square to the total hours for all grid squares. In a standard plotting routine, isopleths are drawn to bound the smallest areas accounting for 25, 50 and 75% of the total residence time probabilities. These everyday probability fields are analogous to a trajectory-based "airshed" for each receptor site. They address the question: where is the summertime air at site X most likely to have previously resided? (See http://capita.wustl.edu/otag/Reports/Restime/Images/IMG18.GIF). Similar probability fields can be calculated for any subset of the total trajectories, such as for the upper 50% of the receptor's ozone concentration (http://capita.wustl.edu/otag/Reports/Restime/Images/IMG04.GIF).

The incremental probability for any subset of trajectories is calculated by subtracting the everyday probability from the high (or low) ozone subset probability. For consistency, isopleths bound areas accounting for 25, 50 and 75% of the total incremental probability. Plotted areas are those more likely to be upwind if ozone is high (or low), regardless of their proximity to the receptor. Locations close to and moderately distant from the receptor may be implicated.

Upper 50% Probability minus Everyday Probability = Upper 50% Incremental Probability

Figure one.

Incremental Probability plots for all 23 sites and for several different percentiles of ozone concentration are stored in a series of .avi animations (for which CAPITA Movie Program is recommended) as follows:

50hilo.avi - shows incremental probabilities for upper and lower 50th % ozone at all sites

25hi10.avi - shows incremental probabilities for upper 25th % and 10th% ozone at all sites

50hi5.avi - shows incremental probabilities for upper 50th % and 5th% ozone at all sites

Example frames from 50hilo.avi for selected sites and site groups are pasted below in Figures 2 - 7. In each figure, the left panel shows locations associated with the lowest 50% of ozone values at the receptor(s); the right panel shows locations associated with the highest 50% of ozone values.

Figure 2. Incremental Probability: Grafton, WI
for Lower 50% (left) and Upper 50% Ozone (right)

Figure 3. Incremental Probability for Bennington, VT
for Lower 50% (left) and Upper 50% Ozone (right)

Figure 4. Incremental Probability: Iberville Parish, LA
for Lower 50% (left) and Upper 50% Ozone (right)

Figure 5. Incremental Probability: Seminole Co., FL
for Lower 50% (left) and Upper 50% Ozone (right).

Figure 6 Incremental Probability: 6 Midwestern Sites
for Lower 50% (left) and Upper 50% Ozone (right)

Figure 7. Incremental Probability: 6 Northeastern Sites
for Lower 50% (left) and Upper 50% Ozone (right)

The Incremental Probability plots based on all 23 trajectory sites are displayed in Figure 8. As with the previously-reported results of location-based sorting, these results suggest that the OTAG domain is geographically well-formulated for addressing ozone transport. Clean air at sites throughout OTAG is associated with locations external to OTAG, while higher ozone concentrations are associated with areas internal to OTAG. By contrast, the 13-State OTR is not so well configured (see Figure 7), as high ozone levels at OTR sites include high incremental residence-time probabilities in areas external to the OTR.

Figure 8. Incremental Probability: All 23 Trajectory Sites, for Lower 50% (left) and Upper 50% Ozone (right)

Back to Contents

Incremental Probabilities for Upper Percentile Ozone Concentrations

Incremental probability plots for upper 10th percentile ozone concentrations for 12 individual sites are displayed in Figure 9. These are similar to the right-hand sides of Figures 2 - 5, but are based on fewer of trajectories, associated with higher ozone levels the receptors. The upper 50th percentile plots are each based on about 1200 trajectories (half of total trajectories) associated with above median ozone at each site. The upper 10th percentile plots are based on only one fifth (about 240) as many trajectories. At most sites, however, there are not large differences in the spatial patterns associated with moderately high (upper 50%) and very high (upper 10th percentile) ozone concentrations. Note also that the incremental probability areas for the most southerly sites (Figures 4, 5 and bottom row of Figure 9) are smaller than for other sites - suggesting that these southern sites may be more heavily influenced by local stagnation.

Direct comparisons between upper 50th and upper 5th percentile incremental probabilities for all 23 sites and selected groups of sites are presented in the movie 50hi5.avi. The upper 5% ozone values represent relatively extreme concentrations - which might be expected to occur on 4 to 5 days a year at these sites. The 5th percentile plots are based on only 1 tenth as many trajectories as the 50th percentile plots, and so tend to be more tightly focused. For most sites, however, the patterns of incremental probability for areas upwind for moderately high and extremely high ozone are quite similar (see Figures 10).

Figure 9. Incremental Residence-Time Probabilities for Upper 10% Ozone vs. Every Summer Day

Figure 10. Examples of Sites with Similar Incremental Residence-Time Probabilities for Upper 50 Percentile (left) and Upper 5 Percentile (right) Ozone Concentrations

As displayed above in Figure 10, for a majority of sites there are relatively small differences between incremental probabilities for areas upwind for moderately high (upper 50%) and extremely high (upper 5%) ozone levels. Two notable exceptions to this pattern are at the high elevation Shenandoah National Park and Great Smokey Mtn. National park sites. During the highest episode days, Shenandoah exhibits relatively more influence from the due west and due south, and relatively less influence from the southwest direction, while Gt. Smokey mountain NP shows an increased prevalence of southwesterly flows during episodes (see Figure 11).

Figure 11. Examples of Sites with Different Incremental Residence-Time Probabilities for

Upper 50 Percentile (left) and Upper 5 Percentile (right) Ozone Concentrations

Figure 10 shows sites with very similar incremental probabilities for moderately high and extremely high ozone. Figure 11 results represent the most extreme exceptions to this general pattern of similarity. A third group of sites show both similarities and differences between upwind areas for moderately high and extremely high ozone Specifically, the areas with highest probability of being upwind for upper 5%, also have high incremental probability for upper 50%. However, other areas upwind for upper 50% are not implicated during episodes (see Figure 12).

Figure 12. Examples of Sites with "Similar and Different" Incremental Probabilities for

Upper 50 Percentile (left) and Upper 5 Percentile (right) Ozone Concentrations

For Whiteface Mtn., NY, moderately high ozone is spatially associated with a broad area extending south to southwest to west of the site. For extremely high ozone, there is relatively low probability that the air resided due south of the site, with higher probability from areas to the west (along the lower Great Lakes) ans southwest (Ohio River Valley). Farther east, at Port Clyde ME, there's a high probability for moderately high (upper 50%) ozone concentrations that the air had previously resided to the south of the site (east coast corridor). But this 'southerly' influence is less evident for extremely high concentrations, when southwesterly flows (residence over Midwestern industrial areas) are more probable. For moderately high ozone at Mark Twain State Park, MO, incremental probabilities are high for areas extending both to the south and northeast of the site. During episodes, the southerly influence is much less evident, while the northeasterly influence (industrial Midwest) remains. A similar pattern is evident for Boone County, KY. Thus, for a number of sites in different regions, moderately high ozone is associated with prior residence time over a broad range of upwind areas. But when only peak ozone concentrations are considered, the industrial Midwest appears to be the region most consistently upwind for a number of different sites and regions throughout the northern half of the OTAG domain (see Figures 12, and 13).

Figure 13. Comparative Incremental Probabilities for Groups of 6 Midwestern & 6 Eastern Sites for

Upper 50 Percentile (left) and Upper 5 Percentile (right) Ozone Concentrations

Figure 14. Incremental Residence-Time Probabilities for 23 Sites for
Upper 50% (left) and Upper 5% (right) Ozone Concentrations

Upper 50th and upper 5th percentile incremental probabilities for all 23 ozone monitoring sites are combined in Figure 14. Note that at both the upper 50% and 5% ozone levels, there are areas of relatively high incremental probability surrounding/adjacent to all the most southerly sites (consistent with local stagnation influences). Note also that at moderately high and extremely high ozone levels, there are no areas of high probability at/near any of the New England sites (consistent with a predominant influence of transport). As increasingly high ozone levels at all sites are considered, the area of highest incremental probability becomes increasingly focused on the industrial Midwest (Ohio River Valley).

It should be noted that the synoptic-scale trajectory model and associated NGM windfields are not well-suited for evaluation of local-scale influences (nor is evaluation of local influences the objective of this study). It is also emphasized that the methods employed here to process ensemble trajectory results do not explicitly include estimates of emissions and chemical reactions (nor is it the intent of this analysis to distinguish among the influences of emissions, chemistry and meteorology). The basic assumption employed here is that the ambient ozone at a given place and time represents (exactly) the sum total of all forces which have acted upon it (emissions - anthropogenic and natural; chemistry - formation and destruction; and meteorology - including stagnation, mixing, transport, wet and dry deposition, etc.). If a specific "source region" is persistently upwind of one (or many) receptors when ozone concentrations are highest, this does not necessarily implicate any specific emissions sources or source categories in that source region. Note, for example in Figures 3 and 5, high probabilities for upper 50th percentile ozone levels at VT and FL receptor sites over marine areas where emissions are presumably quite low, but which may act to protect transported ozone levels from chemical and physical destruction processes.

Conceivably, the Ohio River Valley may be "innocently" upwind of the various receptor sites. The region is spatially large (compared to individual urban areas) and centrally located (geographically and meteorologically) relative to many ozone non-attainment areas. Adverse meteorological conditions such as a high potential for subsiding high pressure systems may cause this region to act as a sort of accumulation area for ozone and precursors from surrounding areas. However, the Ohio River Valley is also characterized by high densities of elevated stationary source NOx emissions (and current geography and adverse meteorology are beyond our control). Historical modeling results (as summarized in the 2/96 ENVIRON review) suggest that reductions in elevated stationary source NOx emissions are relatively ineffective in reducing surface ozone concentrations. The long-term trajectory-based results presented here are clearly at odds with that observation, and suggest that control of stationary source NOx emissions may be an effective means of reducing peak and chronic ozone levels over a broad range of receptors.

Back to Contents


1. The OTAG region (unlike the OTR) is geographically well-configured to address ozone transport, as low ozone concentrations are consistently associated with prior airmass residence over areas (north, south, east and west) external to OTAG, and high ozone is consistently associated with prior airmass residence over areas (throughout and) internal to OTAG (Figures 2-8).

2. Local stagnation (high residence time near the receptor) appears to be a much more important characteristic of high ozone concentrations in the South than in the Northeast (Figures 4, 5, 9, 10, 14).

3. For most receptor sites, upwind areas associated with moderately high ozone are also associated with extremely high ozone concentrations at these sites (Figure 10).

4. While there are unique upwind areas associated with moderately high (above median) ozone levels for each site, the industrial Midwest appears to be a common source area for moderately high ozone levels many sites throughout OTAG (Figures 6-8).

  1. For many sites and sub-regions throughout OTAG, the probability that the industrial Midwest was upwind increases as the ozone concentration increases - i.e. when only the highest 5% of ozone values are considered (Figures 12-14).

Back to Contents

Submit your comments, feedback, questions, and ideas pertaining this page. Your input will be automatically added to the existing annotations. In order to add a new comment, you must be registered with the OTAG/AQA Peoples Page.