Paul Wishinski and Rich Poirot, VT Department of Environmental
Conservation, 2/20/95 Summary Draft
"Trajectory Analysis" has been identified on the task
list of the OTAG Air Quality Analysis Workgroup. VT DEC had agreed
to devote some in-kind resources to assist in conducting this
task. We have a copy of the NOAA HY-SPLIT trajectory model (Draxler,
1992), and NOAA NGM meteorological data sets for the years 1989
through 1995 for Eastern North America (East of 105 degrees Longitude).
We have modified the original (PC) HY-SPLIT fortran code to run
on a UNIX workstation, and have developed graphic output routines
to display forward or backward trajectory results on a UNIX platform.
Our current operating mode for backward trajectory calculates
4 trajectories per day (with arrival times of 3 AM, 9 AM, 3 PM
and 9 PM, EST) for up to six individual receptor locations. Individual
back trajectories with a maximum duration of 108 hours are stored
for later display and analysis.
For analysis of large numbers of back trajectories, we employ
a technique known as "Residence Time Analysis" (adopted
from Ashbaugh, 1984). Our particular techniques for residence
time analysis are described in Poirot and Wishinski (1985) and
Wishinski and Poirot (1986). The general approach is to grid the
trajectory domain, and keep track of the "residence time
hours" of each trajectory over each grid square in its path.
We can then examine the spatial characteristics of the long-term
Initial Site Selection
For an initial group of six receptor sites, we selected the following group of high elevation sites, all of which have reasonably complete ozone data sets for the past 7 summers (1989-95 - for which we have NGM data.
|Latitude||Longitude||Elev. (m)||AIRS Site #|
|Whiteface Mtn., NY||44.365||73.902||1480||360310002|
|Mt. Greylock (Adams), MA||42.637||73.169||1140||250034002|
|Roof of World Trade Ctr., NYC||40.711||74.013||503||360610063|
|Shenandoah NP (Big Mead.), VA||38.522||78.436||1073||511130003|
|Greenbriar County, WV||37.819||80.513||829||540240001|
|Gt. Sm. Mt. NP (Look Rock), TN||35.631||83.944||793||470090101|
These high "elevation sites" were selected for the following
- They are (with exception of World Trade Center Rooftop) inherently remote, and relatively free from local source influences (much of what affects these sites is obviously transported).
- They exhibit minimal diurnal variation, as they are relatively free from nocturnal ozone destruction (by NOx scavenging and dry deposition) which is characteristic of most lower elevation sites.
- They provide a good opportunity to characterize nocturnal (and
daytime) ozone transport (or lack thereof) in layers above the
surface (where nocturnal decoupling, lack of scavenging, and higher
wind speeds may make transport most important).
Unfortunately, sites in excess of 500 meters above sea level are
primarily limited to the Appalachian Mountain range, and are not
widely distributed throughout the OTAG domain. On the other hand,
these sites are generally located to the West of the East Coast
urban corridor, and may therefore be good locations to characterize
transport flows to/from the corridor to other OTAG subregions.
Future analyses will focus on sites in other areas of the OTAG
For each selected site, backward trajectories, with a maximum
duration of 108 hours, were calculated 4 times each day, for the
months June through August, 1989 through 1995. An example individual
trajectory arriving at Whiteface Mtn., NY at 3 PM on 7/7/94 is
displayed in Figure 3. The Whiteface ozone concentration at this
arrival time was 51 ppb - about average for this site for the
last 7 Summers.
The individual trajectory segments (between dots) represent 3-hour
time intervals. When the segments are packed closely together,
the air parcel is moving slowly. When the segments are farther
apart (over Illinois, for example) the air parcel movement is
Figure 3 also displays the grid of 80 x 80 km squares employed
in our Residence Time Analysis calculations. For each trajectory,
we keep track of the "residence time" over each grid
square in the trajectory's path (white squares in this figure).
Figure 4 displays 8 back trajectories from Whiteface Mountain
which arrived at 9 AM with an ozone concentration of greater than
90 ppb. These were the 8 highest 9 AM concentrations at Whiteface
(for which we also have trajectory data) during the months of
June through August, 1989 through 1995. As such, these might be
considered to represent extreme events which occur on average
of about once a year (although there were seven 9 AM ozone values
> 90 ppb in 1988 - for which we have no trajectory data).
These morning ozone levels, in excess of 75% of the standard are
unquestionably transported, from ozone production at least a day
(or more) earlier. The associated trajectories suggest a wide
range of regional influences, including recirculation within the
OTR region (red), several episodes moving along the Ohio River
Valley (black and blue), and several more rapidly moving events
which have passed over northern, western and southwestern regions
of the OTAG domain.
Figure 5 shows all back trajectories arriving at Whiteface at
3 AM, 9 AM, 3 PM and 9 PM during June - August, 1989 - 1995. (Aside
from draining your ink jet printer with one picture), this figure
provides useful information in two areas:
(1)It clearly shows the eastern and western limits of the meteorological
data employed in these model runs. Many trajectories exceed these
limits, and are subsequently truncated.
(2) It shows that the summertime air on Whiteface Mtn. Has on
various occasions previously resided over (and might be potentially
influenced by emissions from) any and all locations in the OTAG
domain. Equivalent plots for the other 5 selected sites are quite
Similar to Figure 5.
There are 2,120 trajectories included in Figure 5, which represents
82% of the possible 2,576 trajectories (4 trajectories/day x 92
days/summer x 7 years) during this time period. The difference
is caused by time periods for which either NGM meteorological
data or local ozone data is missing. Collectively, these trajectories
have resided for 153,737 hours over the model domain - which represents
67% of the possible 228,960 hours (2,120 trajectories x maximum
of 4.5 days per trajectory). The difference is caused by truncated
trajectories which exceed the model domain, or for which missing
NGM data is encountered within the model domain.
In Figure 6, the total 153,737 trajectory residence time hours
for Whiteface Mtn. for the past 7 Summers have been aggregated
into the grid squares displayed in Figure 4. Contours have been
selected for plotting so that they bound the smallest numbers
of grid squares accounting for 25% , 50%, and 75% of the total
residence time hours. The area shaded pink, grey and tan each
include 25% (38,434 hours) of the total residence time hours for
this scenario, and represent areas of decreasing residence time
probabilities away from the receptor. The least probable 25% of
scenario residence time hours are in the unshaded, white area
of the map.
This residence time probability plot may be thought of as providing
an answer to the following question: "Where is the summertime
air at Whiteface Mountain most likely to have previously resided"?
Perhaps the answer ("Somewhere to the west") did not
require this level of effort, but now we have a specific statistical
definition based on a long-term "trajectory climatology",
and can begin asking other questions of the trajectory database.
For the residence time plots in Figures 7 and 8, the Whiteface
Mtn. trajectories have been sorted into two groups, depending
on whether the resultant ozone concentration was low or high.
The definition of high or low in this case is based on determining
the ozone level for which half the total cumulative, seven-summer
ozone dose is contributed by hours with higher concentrations,
and half by hours with lower concentrations. In this case the
50% cut point is an ozone concentration of 51 ppb. Figure 7 shows
the residence time probability for "low" Whiteface Mtn.
ozone concentrations and is based on 1323 trajectories, each with
resultant ozone levels of less than 51 ppb. Figure 8 shows similar
residence time probabilities, but in this case, based on 791 trajectories
with resultant ozone concentrations of 51 ppb or higher (roughly
40% or more of the standard). Figures 9 and 10 are similar to
figure 8, but are based on increasingly higher ozone thresholds.
Figure 9 is based on 323 trajectories with resultant ozone concentrations
greater than 62.5 ppb (50% of standard) which collectively contributed
25% of the total 7-summer ozone dose at Whiteface Mtn.. The upper
10 % of the ozone dose at Whiteface was associated with 103 trajectories
with ozone levels greater than 75.5 ppb. Thus, Figure 10 shows
the most probable locations of air arriving at Whiteface Mtn.
With ozone levels in excess of 60% of standards.
Whiteface Mtn., NY Residence time plots depict residence time
probabilities for trajectories associated with lower 50% of 7-Summer
cumulative ozone dose (Fig.7), upper 50% (fig. 8), upper 25% (Fig.
9) and upper 10 % of cumulative 1989-95 ozone dose (Fig. 10).
Figures 11 through 15 are similar to Figures 7 through 10, except
that they are based on trajectories and ozone concentrations from
Look Rock Mtn., TN. They show residence time probabilities associated
with ozone levels of < 52.5 ppb (Fig. 11), > 52.5 ppb (Fig.
12), > 62.2 ppb (Fig. 13) and > 71 ppb (Fig. 14).
By comparison to the Whiteface Mtn., the Look Rock Mtn. residence
time plots show much less of a westerly orientation - and are
more uniformly distributed in all directions from the receptor.
Also, the differences at Look Rock between the plots of relatively
low (< 52.5 ppb) and relatively high (> 52.5 ppb) ozone
are less distinctly different from each other - showing large
regions around the receptor which are associated with both high
and low ozone levels. There is a distinct south-southwesterly
orientation to the Look Rock trajectories with low resultant ozone
levels, while the higher concentrations at Look Rock appear to
be more frequently associated with transport from the north-northwest.
Figures 15 and 16 display average ozone levels at Whiteface (Figure
15) and Look Rock (Figure 16) as a function of prior trajectory
location. The shaded areas in both plots represent grid squares
where the residence time-weighted average of all trajectories
passing through that square and arriving at the receptor is greater
than 50 ppb (approximately the average ozone concentration at
both receptor sites). Grid squares are only included for these
plots if they contained at least 100 hours of trajectory residence
times (which is approximately the average number of residence
time hours for each of the 1440 grid squares in our OTAG Domain
map for our seven-summer data set). So the shaded areas are both
frequently upwind of the receptors, and are also areas which are
associated with higher than average ozone levels at the receptors.
Trajectories passing through areas shaded green result in average
ozone concentrations of > 55 ppb at the respective receptor
sites. Areas shaded pink result in ozone greater than 52.5 ppb.
Areas shaded dark tan, purple and red are associated with average
receptor ozone concentrations of 55 ppb, 57.5 ppb and 60 ppb,
There are some interesting similarities and differences between
Figures 15 and 16. One obvious difference is that substantially
larger areas are associated with higher average ozone levels (>
57.5 and >60 ppb) at Whiteface Mtn. than at Look Rock. As noted
with the residence time plots, a broader range of areas are associated
with higher ozone levels at Look Rock, and many of these areas
were associated with both high and low ozone levels at this site
(presumably due to influences of other meteorological factors
- such as precipitation, stagnation, cloud cover, etc.). The highest
Whiteface ozone levels are associated with flows from the southwest,
while the highest ozone levels at Look Rock are associated with
flows from the northwest. Consequently, there is a large area
of overlap in the regions associated with the highest ozone levels
on these two mountaintop sites, located approximately 800 miles
Figure 17 is based on all the back trajectory calculations for
the summers of 1989 through 1995 for each of four high elevation
sites at Whiteface Mtn., NY, Mt. Greylock, MA, Big Meadows at
Shenandoah National Park, VA and Look Rock Mtn. in Great Smokey
Mountain National Park, TN. This represents a total of 7,782 trajectories,
which have collectively resided for a total of 575,779 hours over
our gridded Eastern US (approximately OTAG) domain. Trajectories
arriving at each of these sites have (as in Figure 4 for Whiteface
Mtn.) Passed over all locations in the OTAG region.
For each of these 4 sites, we calculated an average ozone value for trajectories passing through each grid square and arriving at the receptor location. Thus, for each grid square there are 4 average ozone values - one associated with each of the 4 receptor locations. Then for each grid square, we selected the lowest of the 4 average ozone
values and plotted the results in figure 17. Trajectories passing through the areas shaded green have resulted in average concentrations of at
least 45 ppb at all 4 mountain
top sites. Trajectories passing through the areas shaded purple and red have resulted in average ozone levels of greater than 52.5 and 55 ppb at all 4 mountain sites, respectively. Thus there appears to be a large region in the Midwest which is consistently upwind prior to high, (transport-dominated) ozone levels at mountain tops throughout the Appalachians. This ozone transported aloft is approximately half the ozone standard on average, and can exceed standards on occasion - minimizing the effectiveness of local control strategies in downwind locations during typical afternoon episode periods when these elevated layers of high ozone mix downward to the surface.
Future Applications and Issues
The ensemble trajectory analyses presented here are based on a
meteorological model, without consideration of emissions or atmospheric
chemical reactions. The preliminary results do not demonstrate
cause and effect, but do identify areas with strong statistical
associations with airmasses subsequently resulting in high measured
ozone concentrations at downwind monitoring sites - at relatively
high elevation. The high elevation sites provide ideal "lamp
posts" for examining ozone transport, but such lamp posts
are limited primarily to the Appalachian Mountain range, and do
not illuminate the entire OTAG region. For future applications
we would like to include receptor sites more widely distributed
in other parts of the OTAG domain, and are seeking recommendations
from other OTAG participants. For lower elevation sites with strong
diurnal ozone patterns, we believe it will be important to examine
different hours of the day separately (and perhaps the night time
ozone values will contain little useful information.
Also, in running trajectories for these higher elevation sites,
we have eliminated (or minimized) the importance of choosing among
complex alternative options in the way the HY-Split Model is run.
One option in particular (which has not been thoroughly tested
to our knowledge) allows the modeler to specify any number of
"sub-layers" within the models lowest "sigma layer"
(below approximately 300 meters). The user can specify the height
of these sub-layer(s), and the model will interpolate a series
of new windfields at the specified elevations, by interacting
with the gridded model terrain with a modified "Ekman Spiral"
approach. Based on our limited sensitivity testing use of these
interpolated sublayers has little effect on the trajectory paths
for higher elevation sites - which range (with exception of the
World Trade Center rooftop) from about 800 to 1500 meters). For
lower elevation sites, the use of sub-layers can become quite
important. For example, Figure 18 shows two different HY-SPLIT
back trajectory results for a low elevation (near sea level) receptor
site in Rye, NH. Both trajectories arrived at 3 PM on July 20,
1991, with an ozone concentration of 156 ppb.
The grey (northerly) trajectory was calculated in the standard
HY-SPLIT default mode - without interpolated sub-layers. The orange
( southerly) trajectory was calculated with interpolated sub-layers
at 50, 100, and 250 meters above the surface. These heights were
selected for approximate consistency with the vertical layer structure
for a 7-layer UAMV run according to OTAG Modeling Protocol. Clearly
the resultant trajectoriy paths are quite different, as the terrain
interpolated sub-layers tend to slow the trajectory down, and
turn it in a counter-clockwise direction. The trajectories tend
to suggest potential transport (and local) influence from different
source regions (although for other episodes at this site, trajectories
with and without sub-layers are very similar to each other). In
a general sense, this "more than one answer" result
is probably quite realistic for many ozone episode periods. That
is, many ozone exceedances are influenced by local-scale emissions
and short-range transport, mesoscale transport from nearby urban
areas and adjacent states, and synoptic scale transport from more
distant source regions.
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.