-------
-a
4->
(/I
4->
I/)
01
S-
o
01
a
QJ
to
c
o
L.
O>
ra
0>
S-
o
Q.
O)
s_
CD
\
11
-------
order to study the visibility/pollutant relationship. For these locations, we
contacted the local monitoring agencies which operate the NASN samplers. The
purpose of these contacts was to assess the utility of the NASN TSP (total
suspended particulates) data for visibility/pollutant studies.
The survey of the NASN TSP monitoring sites included the following
questions:
How long has the TSP Hi-Vol been operated? Has it been relocated?
t What is the height of the Hi-Vol above the ground? Is the sampler
exposed to air flow in all four directions?
Is the sampler exposed to significant local sources of dust (e.g.,
unpaved roads)?
Is the NASN site representative of the area wide pollution levels?
In particular, is it representative of the air mass at the NCC site?
Are there any suggestions or comments in regard to our visibility/
pollutant studies?
The survey of airport observers and NASN monitoring agencies indicated
that the visibility/pollutant studies had a fair to good chance of being
successful at all ten locations. Because of budgetary constraints, however,
the analysis was restricted to six locations where conditions (e.g., distance
between airport and NASN site, data quality, etc.) appeared to be best for the
study. These six locations were Chicago, Newark, Cleveland, Lexington,
Charlotte, and Columbus (OH).
Initial Data Processing
For each of the airport locations studied, complete tapes of all surface
weather data were obtained from NCC in the CD-144 format. These tapes were
*
processed to extract data for the four daytime hours. With data for these
hours, we formed a "processed visibility data base" for each location; this
data base included the date, hour, visibility, relative humidity, and special
notations (storms, liquid precipitation, frozen precipitation, fog, blowing
dust, smoke, haze, etc.).
The nationwide NASN data for TSP, sulfate, nitrate, etc. were obtained
in tape form from EPA's SAROAD data bank. We reorganized the original EPA
data to create a "processed pollutant data base". For each site, this data
base listed the date and the various pollutant measurements in a consistent,
easy-to-access format.
*For some years, the original NCC tapes contained data for every hour rather
than every third hour. For consistency, we extracted the same four daylight
hours in all years. -
12
-------
In order to investigate visibility/pollutant relationships at six
locations, the "processed visibility data base" was combined with the "pro-
cessed pollutant data base" for those locations. The resultant data base
listed, for each day, the 24 hour average pollutant concentrations, the
daytime averages of visibility and relative humidity, and special weather
notations.
FREQUENCY DISTRIBUTIONS OF VISIBILITY DATA
*
Because of the nature of the reporting methods, visibility data are most
appropriately summarized by cumulative frequency distributions of the form
"percent of time visibility is greater than or equal to X miles." Figures 2,
3, and 4 present recent cumulative frequency distributions for all the sites
studied. Figure 2 is for metropolitan locations; Figure 3 is for urban/
suburban locations; and Figure 4 is for nonurban locations.
When analyzing cumulative frequency distributions for visibility, it is
important to use only those visibilities that are routinely reported by the
observer. For instance, it is not uncommon to see the following type of
situation:
Visibility % of Time Reported Cumulative Frequency
15 miles 20% 20%
12 miles 1% 21%
10 miles 29% 50%
7 miles 20% 70%
In this case, the 12 mile recordings produce a "kink" in the cumulative fre-
quency distribution. It is obvious, in this example, that the 12 mile visi-
bilities are not routinely reported but happened to be recorded a few times by
a member of the observation team. In our analysis of frequency distributions
for visibility data, we took care to use only those visibilities that are
routinely reported.
*When an airport observer reports a visibility of X miles, this usually means
that visibility is at least X miles, not that visibility is exactly X miles.
13
-------
\
01
jQ
1/5
00
0)
20-
15
10-
o
20_
15
10.
5
Washington D.C. 1970-1972
10%
100%
Cumulative Frequency (percent)
Chicago 1970-1972
10% 20 30
S 50 60 70
Cumulative Frequency (percent)
90 100%
Figure 2. Cumulative frequency distributions of visibility at
metropolitan locations.
14
-------
OJ
20_
15
10
Newark 1970-1972
10%
I
20
I
30
T
40
I
50
I
60
T
70
I
80
90 100
Cumulative Frequency (percent)
.0
to
15-
10-
5-
Cleveland 1970-1972
100'
Cumulative Frequency (percent)
Figure 2. Cumulative frequency distributions of visibility at
metropolitan locations. (Continued)
15
-------
OJ
J3
r-
1/1
15
10
5
10%
Lexing.on 1970-1972
i
20
I
30
i
50
i
60
40 50 60 70
Cumulative Frequency (percent)
i
80
90 100%
O)
15
10-
5-
10%
Charlotte 1970-1972
Cumulative Frequency (percent)
100%
Figure 3. Cumulative frequency distributions of visibility at
urban/suburban locations.
16
-------
I/)
Ol
Columbus, Ohio 1970-1972
10% 20 30 ' 40 50 60 70
Cumulative Frequency (percent)
1
80
I
90
100%
to
OJ
20
15
10
5
Dayton 1970-1972
T
I
I
50
I
10% 20 30 40 50 60 70 80 90 100%
Cumulative Frequency (percent)
Figure 3. Cumulative frequency distributions of visibility at
urban/suburban locations. (Continued)
17
-------
<1>
P"
E
.a
Ul
20
15
10
30 _
Columbus, Ind. 1967-1969
10% 20 30
I
40
30
j
60
I
70
30
Cumulative Frequency (percent)
90 100%
1/1
25-
20-
\
\
\
\
\
10-
5-
Williamsport 1970-1972
iiri r i
20 30 40 50 60 70
Cumulative Frequency (percent)
90 100%
Figure 4. Cumulative frequency distributions of visibility at
nonurban locations.
\
18
-------
15.
5-
Wilmington 1966-1967
r
40
60
10% 20 30 40 50 60 70
Cumulative Frequency (percent)
80
90
100%
25-
>> 15-
10%
Figure 4.
I
20
I
30
I
40
Dulles 1970-1972
60
I
70
I
30
Cumulative Frequency (percent)
Cumulative frequency distributions of visibility at
nonurban locations. (Continued)
19
-------
The graphs in Figures 2 through 4 illustrate a property that we found
to be nearly universal among the sites studied. The cumulative frequency dis-
tribution tends to be nearly linear at the higher visibilities (i.e., the
lower percent!les). In many cases we have used this property to calculate the
10th percentile of visibility even if the actual recordings started at a
higher percentile (e.g., the farthest marker might be reported 20% or 30% of
the time). This calculation was done by linear extrapolation of the cumulative
frequencies for the two farthest markers. The extrapolation is indicated by
the dashed lines in Figures 2 through 4.
In this report, historical trends in visibility are based on changes in
visibility percent!'les: the 10th percentile (best conditions), the 50th per-
centile (median), and the 90th percentile (worst conditions). This method of
reporting visibility trends differs from the traditional method (Holzworth
1960, 1962; Neiburger 1955; Keith 1964, 1970; Green and Battan 1967; Miller
et al. 1972; Hartman 1972) which examines shifts in the fraction of days (or
hours) that visibility is in certain ranges. The units of our visibility
trend index are [ miles ], while the units of the traditional index are
[percent of days ] or [percent of hours]. Our visibility trend index can
be directly transformed into trends for "extinction coefficient" which are
linearly related to pollutant trends (see discussion in next section).
ANALYSIS OF VISIBILITY/POLLUTANT RELATIONSHIPS
Our analysis of visibility/pollutant relationships follows the statisti-
cal procedures established by Cass (1976), White and Roberts (1977), and
Trijonis and Yuan (1978). Regression equations are developed which relate
daytime average visibility to daily averages of total suspended particulate
(TSP), sulfates (SO^), nitrates (NOg), and relative humidity (RH). The coef-
ficients in these regression equations can be interpreted as estimates of
"extinction coefficient per unit mass" for each of the pollutant species.
These extinction coefficients allow us to estimate the fraction of haze (or
fraction of visibility loss) attributable to each pollutant. The following
paragraphs summarize the statistical techniques and discuss some of the
potential limitations in the methods.
20
-------
Definition of Variables
The basic data for the study of visibility/pollutant relationships con-
sist of the measurements listed in Table 3. Before conducting statistical
analysis of the data, we perform some simple changes in the forms of variables.
For instance, instead of using visual range (V) as the dependent variable, it
is convenient to use the extinction coefficient, B,
B = 24-3 . (D
41 *
where the units of B are [10 meters] and the units of V arefmiles] . The
extinction coefficient can be linearly divided into contributions from various
atmospheric components, i.e.,
Rayleigh + scat + abs-aerosol + abs-gas ^ '
where BRayleiah = ''1'9'lt scattering by air molecules (blue-sky
or Rayleigh scatter) ~.15 (Robinson 1968)
B . = light scattering by atmospheric aerosols
Babs-aerosol = 11ght absorPtion by aerosols
Babs as = light absorption by gases
TABLE 3. DATA FOR VISIBILITY/POLLUTANT STUDIES
Variable Units Averaging Time
V... visibility or visual
range miles 4 daylight measurements
RH... relative humidity percent " 4 daylight measurements
TSP... total suspended -
particulates jjg/m 24-hour average
S07... sulfates yg/m 24-hour average
3
NO^... nitrates pg/m 24-hour average
*Equation (1) is the Koschmieder formula based on a threshold brightness level
of 0.02 for the human eye. In a uniform atmosphere with extinction coefficient
B, a contrast level of 1 (black object against norizon sky) will be reduced to
a contrast level of .02 at a distance of V .= 24.3/B miles.(
,',"' y
3_0 , ., I/, 21
A r.
5)
-------
Each of the components of B should be directly proportional to aerosol
or gas concentrations (assuming other factors such as light wavelength, aero-
sol size distribution, particle shape, and refractive index remain constant).
In polluted urban air, it is thought that aerosol light scattering (B$cat)
tends to dominate over the other contributions to the extinction coefficient
(Charlson 1969).
Slight transformations are also performed on the independent variables.
Following White and Roberts (1977) we define
\
and
S = SULFATE = 1.3 SOJ
N = NITRATE = 1.3 N0~ (3)
in order to account for the mass of cations (presumably ammonium) associated
with the measured values of SOT and NOr. The variable,
T = TSP - SULFATE - NITRATE = TSP - S - N (4)
is used to represent the non-sulfate, non-nitrate fraction of TSP.
Multi-Variate Regression
When several independent variables (RH, SULFATE, NITRATE, and T) are
affecting a dependent variable (B) it is important to perform a multi-variate
analysis that can separate out the individual impact of each independent
variable, discounting for the simultaneous effects of other independent
variables. Uni-variate analyses, -based on simple one-on-one relationships,
can lead to spurious results because of intercorrelations among the independent
variables. For instance, in some cases we found that NITRATE and T apparently
correlated with B only because they were correlated with SULFATE which, _ir[
turn, was significantly related to B.
An appropriate tool for multi-variate analysis is multiple regression.
Following the precedure of Cass (1976), White and Roberts (1977), and Trijonis
and Yuan (1978) we perform multiple linear regressions of the form
B = a + bjRH + b2(TSP - SULFATE - NITRATE) + b3SULFATE + b4NITRATE,
or B = a + bjRH + b2T + b3S + b4N.
These regressions are run stepwise, retaining only those terms which are
greater than zero at a 95% confidence level. The regression coefficients
22
-------
bo, and bj represent the extinction coefficient per unit mass for each pol-
A
lutant species, in units of (104 meters)~V(ug/m3).
For all the regressions according to Equation (5), the constant term "a*
turns out to be a number on the order of minus 1 to minus 4 [10 m]~l. The
constant "a" represents the scattering when all four variables (RH, T, 5, and
N) are zero. It is reasonable to consider the possibility that T, S, and N
are zero, but it is not reasonable to extrapolate the linear regression
equation to values of zero relative humidity. To make the constant term
better-behaved, and to facilitate interpretation of the results, we choose to
write the results of the linear regressions as
B = a' + b:(RH - RH") + bgT + b3S + b4N, (6)
where RH" = average relative humidity for the location, and a" = a + bjRH".
The constant term "a'" now represents the scattering coefficient when the
three pollutant variables are zero and relative humidity is at its average
value.
We also perform regressions which include relative humidity effects in
a nonlinear manner. Cass (1976) indicates that light scattering by a sub-
RH a
micron, hygroscopic aerosol might be proportional to (l-fro) , where the
exponent a is expected to occur in the range -0.67 to -1.0. To account for
this type of effect, we attempt regressions of the form
D =^K TSP-SULFATE-NITRATE ^ u SULFATE . . NITRATE m
B = a + bj - - - flj- - + b - - jgr + b3- - m . (7)
(1 T"oo) (1 TTo]
For most locations, the constant "a" in Equation (7) turns out to be approx-
imately the same as the constant "a'" in Equation (6).
Average Extinction Budget
The regression equations can be used to compute the fraction of visibi-
lity loss, on the average, that is due to each pollutant species. These
calculations are best illustrated by examples.
The linear regression [Equation (6) ] for Columbus, Ohio results in the
formula,
B = 1.33 + .089(RH - RJT) + .120 SULFATE + .091 NITRATE, (8)
. - _. ". , . f
with a total correlation coefficient of 0.81.' The average value for the ex-
,' - 23
-------
4 -1
tinction coefficient at Columbus is B = 3.56 [10 meters] , corresponding to
a visibility of 6.8 miles. Using Equation (8), the average extinction (haze)
level at Columbus can be disaggregated into components by substituting in
average values for the variables. With average values for B, RH, SULFATE, and
NITRATE, Equation (8) reduces to
Average SULFATE Average NITRATE
^3 / 3
3.56 = .15 + 1.18 + .120(15.7yg/nT) + .091(3.9yg/nT)
Blue-sky /"^ Remainder Contribution Contribution
scatter by of 1.33 of sulfates of nitrates
air molecules constant
term
or 3.56 = .15 + 1.18 + 1.88 + .35 (9)
Equation (9) indicates that, on the average for Columbus, 53% of the extinction
is from sulfates, 10% is from nitrates, 4% is from air molecules, and 33% is
unaccounted for.
Alternately, we can compute an average extinction budget using the non-
linear RH regression model. For Columbus, Equation (7) reduces to
P _ n og . n df- SULFATE noo NITRATE ,,n,
B - 0.98 + 0.46(1 _ >01 RH) + .022 (l _ >01 RHj . (10)
Substituting average values for B, SULFATE/(1 - .01 RH), and NITRATE/(1 -
.01 RH), we obtain
3.56 = .15 + .83 + .046(50.0) + .022(12.5)
Blue-sky
scatter
Contribution Contribution
of sulfates of nitrates
Remainder
of constant
or 3.56 = .15 + .83 + 2.30 + .28 (11)
Equation (11) indicates that, on the average for Columbus, 65% of the
extinction is from sulfates, 8% is from nitrates, 4% is from air molecules,
and 23% is unaccounted for.
Limitations of the Regression Approach
There are several limitations to using regression models to estimate the
contribution of various pollutants to visibility loss. One limitation is that
the NASN site and airport are not co-located. Random errors introduced by
24
-------
differences in the air masses at the two locations would tend to weaken the
statistical relationship, leading to a lower correlation coefficient and lower
regression coefficients. This could cause us to underestimate the extinction
coefficients per unit mass for the pollutant species, and therefore to under-
estimate the contributions of the pollutant species to the total extinction
budget.
A systematic error could result if the pollutant concentrations at the
downtown NASN sites are consistently higher than the pollutant concentrations
averaged over the visual range surrounding the airport. The bias caused by
relatively high pollutant measurements would also result in an underestimate
of extinction coefficients per unit mass for the pollutant species. A reverse
type of bias, e.g. an overestimate of extinction coefficients per unit mass,
would result if daytime pollution levels (corresponding to the time period of
the visibility measurements) were higher than 24-hour average pollutant
levels.
An overestimate of extinction coefficients per unit mass could also be
produced by the loss of water associated with the aerosol during equilibria-
tion of the Hi-Vol filter. The ambient aerosol mass tends to be greater than
the measured aerosol mass because more water is usually attached to the former.
Lie low estimate of actual aerosol mass could lead to a high estimate of ex-
tinction coefficient per unit mass. This effect should not bias the ex-
tinction budgets, however, because the extinction budgets are based on a
product of extinction coefficient per unit mass and the measured mass of
aerosol.
The statistical regression equation may also overstate the importance
of the pollutant variables if these variables are correlated with other pol-
lutants which are not included in the analysis. Nitrates (and possibly sul-
fates) may act, in part, as surrogates for other related photochemical pol-
lutants, such as secondary organic aerosols and nitrogen dioxide.
Potential errors in measurement techniques also raise a caution flag.
Artifact sulfate (formed by S02 conversion on the measurement filter) may
cause us to underestimate slightly the extinction coefficient per unit mass
for sulfates. The greatest measurement concern, however, involves nitrates
(Spicer and Schumacher 1977). Nitrate measurements may represent gaseous
compounds (N02 and especially nitric acid) as well as nitrate aerosols.
25
-------
Also, high sulfate concentrations may negatively interfere with nitrate
measurements (Marker et al. 1977).
A final difficulty in the regression analysis is the problem of coline-
arity, i.e. the interconnections that exist among the "independent" variables
(sulfates, nitrates, remainder of TSP, and relative humidity). Although the in-
tercorrelations among these variables are not extremely high, they usually are
significant (correlations on the order of 0.2 to 0.6). Multiple regression is
designed to estimate the individual effect of each variable, discounting for
the simultaneous effects of other variables, but the colinearity problem can
still lead to distortions in the results. In particular, the effect of nitrate
and the remainder of TSP may sometimes be lost because these variables are co-
linear with sulfate which appears to be the predominant pollutant variable re-
lated to extinction.
Although the regression models are subject to several limitations, it
should be noted that the basic conclusions resulting from these models have
proven to be quite reasonable. Chapter 5 demonstrates that our results are
consistent with the published literature and with known principles of aerosol
physics.
CONSIDERATION OF METEOROLOGY
In analyzing aerometric data, it is often important to make allowances
for the effects of meteorology. This section discusses our treatment of
meteorology in both the visibility pollutant regressions and the visibility
trend studies.
Visibility/Pollutant Regressions
Special weather events, such as fog or precipitation, can have signifi-
cant effects on the visibility/pollutant regression analyses. Regression,
based on minimization of squared errors, is sensitive to outliers in the data.
The extremely low visibilities (extremely high extinction coefficients)
associated with special weather conditions might substantially affect the
results of the regressions.
To help minimize the effects of special weather events, we eliminated all
days with precipitation from the visibility/pollutant regression analysis. We
did not, however, eliminate days with fog. The reasons for including days
26
-------
with fog in the regression studies were fourfold:
In the Northeast, haze is often so intense that it is difficult to
distinguish from fog (Holzworth 1977). Eliminating days with fog
might entail the loss of the "very hazy" days that are important to
the visibility/pollutant regressions.
Eliminating days of fog would reduce the size of the data base by
about 10 to 30% and would slightly reduce the statistical signifi-
cance of the results.
The presence of relative humidity as a variable in the regressions
should help to minimize distortions that might be produced by in-
cluding days with fog.
A sensitivity analysis indicated that the results of the regressions
were not very sensitive to the inclusion of days with fog.
Visibility Trend Studies
The historical visibility trends presented in this report are based on
data for all daylight hours, without sorts for meteorology. As will be
demonstrated in Chapter 4, these data indicate that visibility has deteriorated
in the Northeast, especially in suburban and nonurban locations. It is of
interest to determine whether the decreasing trend in visibility might be due
to meteorology. To investigate this issue, we have conducted special trend
studies for Lexington (Ky.) and Charlotte (N.C.), the two sites that ex-
hibited the greatest historical decrease in visibility.
Figure 5 compares visibility trends for all hours at Lexington and
Charlotte to trends for hours sorted by meteorology (hours of fog and pre-
cipitation deleted). It is apparent that the trends for the meteorologically
sorted hours are nearly parallel to the trends for all hours. For all hours,
the 50th and 90th percentile visibilities at Lexington decreased by 41% and
47%, respectively, from 1953-1955 to 1970-1972.* For the meteorologically
sorted hours, the 50th and 90th percent!les decreased 35% and 29%. For
all hours, the 50th and 90th percentile visibilities at Charlotte decreased
by 31% and 33%, respectively, from 1955-1957 to 1970-1972. For the
meteorologically sorted hours, the 50th and 90th oercenti^es decreased 22'i
and 29%. Although the net percent reductions ir, visibility are slightly less
*These percent reductions are based on net changes in three-year averages.
Note that 10th percentile visibility is not considered here because computing
the 10th percentile for the meteorologically sorted hours would involve ex-
cessive extrapolation of the cumulative frequency distribution.
27
-------
Hours of Fog and Precipitation Deleted
All Hours
20-
LEXINGTON
^ 15-
£ 10-
5 -
1950
1955
1 I
1960
Year
"I
1965
1970
20-
» 10-
5-
1950
Hours of Fog and Precipitation Deleted
All Hours
\ CHARLOTTE
\
90tii Percentile
1955
1960
1965 1970
Figure 5. Long-term visibility trends at Lexington and Charlotte,
raw trends compared to trends sorted for meteorology.
28
-------
for the meteorologically sorted data, a definite decrease in visibility is
still apparent in the meteorologically sorted data.
Another way to determine whether meteorology might have affected histor-
ical visibility trends is to examine trend data for meteorological variables.
A key meteorological variable with respect to visibility in the Northeast is
relative humidity, which typically exhibits a correlation coefficient of -0.3
to -0.7 with visibility on a day-to-day basis. Figure 6 illustrates long-term
trends in median daytime relative humidity at Lexington and Charlotte. An up-
ward trend in relative humidity is apparent. At Lexington median daytime RH
was 69% in 1970-1972, compared to 60% in 1953-1955; at Charlotte it was 65%
in 1970-1972 compared to 61% in 1955-1957.
The upward trend in relative humidity from the middle 1950's to the
parly 1970's raises the question as to whether this might have been the cause
of the observed visibility decrease. To answer this question, we examined
visibility trends for constant values of relative humidity. As illustrated
in Table 4, visibility decreased significantly from the middle 1950's to early
1970's within each fixed range of relative humidity.
The last column of Table 4 presents trends in visibility that have been
normalized for the historical relative humidity changes according to meteoro-
logical normalization procedures developed by Kerr (1974) and Zeldin and
Meisel (1977). The meteorologically normalized trends show a 39% decrease in
median visibility from 1953-1955 to 1970-1972 at Lexington and a 24% decrease
in median visibility from 1955-1957 to 1970-1972 at Charlotte. These de-
creases are almost as great as the decreases observed in the raw visibility
trend data (41% at Lexington and 31% at Charlotte). Thus, we conclude that
the upward trend in relative humidity had only a very slight effect on the
visibility trends.
The above discussion leads to a new and very intriguing question. We
have found that the historical increase in relative humidity was not the
*
There are at least three plausible explanations for the slightly lesser de-
crease in visibility for the meteorologically sorted hours. First, there may
have been more occurrences of low visibility due to fog and precipitation con-
ditions in later years. Second, the cause of the decrease in visibility may
have been such that it had a relatively greater effect for hours of fog and/or
precipitation. Third, because of the general increase in haziness, more hours
may have been classified as foggy in the later years.
29
-------
70 _
-u
c
ia
"oj
r<3
o 50
OJ
-. Yearly Values
-, Three-Year Moving
Averages
LEXINGTON
T - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - ] - 1 - 1 - 1 - 1 - ] - 1 - T
1950 1955 1960 1965
1970
Year
.. Yearly Values
rr 70
c
01
CL
r-
-a
1 60 -
O)
to
K.
c
ra
^ 50 -
Thre^-Y^'ir flovin" Averan°s
A
t \
i \
^""^ *-^** \
^^^^^
' \ i
/ V
CHARLOTTE
.
1950 1955 1960 1965 1970
Year
Figure 6. Long-term trends in median relative humidity for
daylight hours at Lexington and Charlotte.
30
-------
VI
Ol
VI
0)
in
0)
Sri
ti CO
O to
-j >i
O >
CC
O
E
in
10
cn
ro
I
E
to
in
CO
CXI
i
a>
JC
C7l*->
4-> O
.c v-
cn
a! c
2 o>
o> ro
£ >
3
.c cr
4-> 0)
v
cn
CO
CM
in
O
Q
cn
co
cc.
v
cn
rH VO
00
10
co
ro
I
u E
a 3
a> -c
c ai
i- >
O H-
I -
S5 Q
II
X
! , (_>
CD
O
cn
VI
:c
ac
v
o
o
r-
VI
CO
c^ o
Q
Z 01
LU O
c; u.
o s:
z a:
O O
O
10
V
O
in
o
in
VI
CO
00
os
V
«t
ro
CO
VI
ro
in
CM
cn
CO
O
ro
1 ««
o
O) ro
cn
CM
i
in
o
CM
CM
co
CM
O
r^
CM
en M
in
t-l CO
-" i
>e
ro
10
«»
i
»
CM
o
CM
co
10
x.
r**
CM
»*
ro
ro
vo
CM
I
X
o
CM
in c
i- O
O) 1
ai ro ai
cn x £
ns
t- VIT3
O) -r- c
> -C O
a *>
c
T3 C i-
-
cn ivi
r* IO
ai +>
S t/> I.
01 S- CD
r- ro i£
r- X
J3 t-
ia v
c i-
ro c -c
^ o «
o
a; -3 *-»
E 0) w
t/> 13
£) in .a CM
-r- m -r- rx.
vi cn vi cn
> i > i
ro o
cm c r^
C
ro O
*-> U1
c cn
«
u i
i. ro
ai in
a. cn
,
^ in
i/> cn
> T
m
cm
ro Cn
QJ
^i
>»
o'
J3 CM
M- r^
w cn
VI CM
> cn
r- O
rs.
O) cn
m
Q. cn
Cue
O ID
c .c
u 10
>, (O I.
a> ro
«<
ro !- >,
U O
^ <*- 4)
cn cu
O >, i.
^- o j;
o c *J
1- IV
O 3 *J
a> cr vi
*J 0) t-
OJ 1-
31
-------
basic cause of the visibility decrease. However, could the converse be true?
Possibly, the historical increases in haze levels have affected daytime
relative humidity. In this regard, we note that Husar and Patterson (1978)
in a companion study focusing on interactions among visibility, meteorology,
and other parameters, have found distinct temperature trends at Lexington and
Charlotte. As illustrated in Figure 7, mid-day (1 P.M.) temperatures at
Lexington and Charlotte have decreased approximately 3 to 4°F from the middle
1950's to the early 1970's. Early morning (4 A.M.) temperatures show no
change over this period. Possibly, as hypothesized by Bo!in and Charlson
(1976), the increased haze levels have interacted with incoming solar radiation
during the day and irradiation away from the earth during the night to produce
these changes in temperature patterns. Long-term cycles in climatology consti-
tute an alternative possible explanation for the temperature trends.
In summary, our cursory analysis of weather data indicates that changes
im meteorology do not appear to be the cause of increased haze levels in the
Northeast. However, questions are raised concerning the possibility that the
haze levels may have affected climatology. Both of these issues should be
addressed more thoroughly in future research. With respect to interactions
between temperature and visibility, data for more locations, with varied
trends in visibility, should be analyzed. The visibility trends should be
normalized for long-term temperature changes, and the temperature trends
should be normalized for long-term visibility changes. A theoretical analysis
of the interactions among haze, solar radiation, and temperature should also
be conducted.
The decrease in daytime temperature may, in fact, explain the increases in
daytime relative humidity, because Husar and Patterson (1978) have noted that
dew point remained constant.
32
-------
70-
_^ Yearly Values
_, Three-Year Moving Averages
O>
(O
01
D.
65-
1950
1955
1960
Year
1965
1970
Yearly Values
Three-Year Moving Averages
CHARLOTTE
60
1950
Figure 7.
1955
1960
'1965
1970
Year
Long-term trends in yearly averaoe 1 P.M. temperature at
Lexington and Charlotte,(Husar and Patterson 1978).
33
-------
CHAPTER 3
EXISTING VISIBILITY LEVELS
A very basic issue that needs to be resolved is "what are existing
visibility levels in the Northeast?" Specifically, we would like to quantify
visual range on average days, "best-case" days, and "worst-case" days; we
would also like to know if any large-scale geographic patterns exist in visi-
bility within the Northeast. These questions can be answered by an analysis
of airport visibility measurements.
VISIBILITY VERSUS DEGREE OF URBANIZATION
Figures 2, 3, and 4 (in Chapter 2) presented recent cumulative frequency
distributions for visual range at four metropolitan locations, four urban/
suburban locations, and four nonurban locations. From these figures, one can
easily read the 10th percentile (best), 50th percentile (median), and 90th
percentile (worst) visibilities for each location. These percentile visi-
bilities are listed in Table 5.
Table 5 indicates that visual range is rather low in the Northeast and
that visibility does not depend a great deal on the degree of urbanization.
Median visibility ranges from 8 to 12 miles among the metropolitan locations,
8 to 10 miles among the urban/suburban locations, and 9 to 14 miles among the
nonurban locations. Best 10th percentile visibility is 15 to 22 miles for the
metropolitan sites, 16 to 21 miles for the urban/suburban sites, and 14 to 27
miles for the nonurban sites. Worst 90th percentile visibility ranges from
2 to 4 miles among all the sites.
These results for the Northeast contrast strikingly with results for the
Southwest. In the Southwest, median visual range is 30 to 55 miles in large
urban areas and 65 to 80 miles in nonurban areas (Trijonis and Yuan 1978).
Thus, visibility is 4 to 8 times greater in the Southwest. Also, unlike the
Northeast, a distinct urban-nonurban difference sxists in the Southwest.
34
-------
TABLE 5. MEDIAN, 10TH PERCENTILE, AND 90TH PERCENTILE
VISIBILITY AT TWELVE NORTHEASTERN LOCATIONS
VISIBILITY PERCENTILES (1970-1972)
t
LOCATION
METROPOLITAN
Washington, D.C.
Chicago, IL
Newark, NJ
Cleveland, OH
URBAN/SUBURBAN
Lexington, KY
Charlotte, NC
Columbus, OH
Dayton, OH
NONURBAN
Columbus, IN
Williamsport, PA
Wilmington, OH
Dulles, VA
10th% (Best)
22 miles
*
19
*
21
15
*
16 miles
*
19
*
16
*
21
22 miles
*
27
*
14
*
27
50th% (Median)
12 miles
9
10
8
10 miles
10
8
10
13 miles
11
9
14
90th%
4
3
3
2
3
3
3
2
4
3
3
3
(Worst)
miles
miles
miles
Data are for the three years 1970-1972 with the exceptions of Columbus,
Indiana (1967-1969) and Wilmington, Ohio (1966-1967).
*
Estimated by linearly extrapolating the cumulative frequency distribution.
35
-------
GEOGRAPHICAL PATTERNS IN VISIBILITY
Figures 8, 9, and 10present maps of 50th, 10th, and 90th percentile
visibilities, respectively. The maps distinguish between metropolitan, urban/
suburban, and nonurban locations. It is apparent that the state of Ohio ex-
hibits the lowest visibilities. The exceptionally low visibility in the upper
Ohio river valley has also been noted in a previous study (Husar et al 1976).
SEASONAL PATTERNS IN VISIBILITY
*
Figure 11 summarizes recent seasonal patterns in visibility. The most
notable feature in Figure 11 is that the summer (third) quarter exhibits visi-
bilities about 2 to 3 miles lower than the other seasons. Median visibility,
averaged over the twelve locations, is 11.2 miles in the first quarter, 11.7
miles in the second quarter, 8.5 miles in the third quarter, and 10.5 miles
in the fourth quarter. The dip in visibility during the summer is especially
apparent at the urban/suburban and nonurban locations.
The peaking of haze levels during the summer season has special signifi-
cance. As will be demonstrated in the next chapter, this is a rather new
feature of visibility patterns. By contrast, in the 1950's, visibility during
the summer tended to be better than average visibility during the rest of the
year.
*
All data in Figure 11 are for the years 1970-1972 except Columbus, Indiana
(1967-1969) and Wilmington, Ohio (1966-1967).
36
-------
I/)
0)
E
C
If)
o
O)
C3
00
01
i-
C71
37
-------
\
CO
O)
QJ
CO
(O
CJ
CL
(O
s_
cn
O
cu
en
OJ
3
38
-------
to
0)
to
CD
JD
to
O)
-P
c
OJ
O
O)
ex
-u
O
en
O
c
O
to
TT
01
O
Ol
QJ
s_
3
39
-------
CD
(X.
co
ca
CD
ZD
co
co
ca
o
a.
o
ca
eC
LU
>-
Li-
CS
OL
o-
LU
O
a:
a
>
a;
.a
to
(O
a
01
E
s-
(1)
ra
a.
o
I/I
fl3
OJ
CO
c
O)
U ro
J- CVJ
C£.
-
U.
O
o;
s-
3
O)
O
CVJ
T
IT)
I
O
(S9L.LIU)
I
if)
40
-------
CHAPTER 4
HISTORICAL VISIBILITY TRENDS
The airport observations offer a unique opportunity to examine histori-
cal changes in visibility within the Northeast. This chapter uses the airport
data to document changes in visibility from the late 1940's to the early
1970's.
YEARLY TRENDS IN VISIBILITY
Figures 12 through 21 illustrate historical visibility trends at four
metropolitan locations (Figures 12 to 15), four urban/suburban locations
(Figures 16 to 19), and two nonurban locations (Figures 20 and 21). Trends
are presented for the 10th percentile (best visibility), 50th percentile
(median visibility), and 90th percentile (worst visibility). For each year,
the percent!les are computed from cumulative frequency plots, such as Figures
2 through 4, based only on those visibilities that are routinely reported.
Most of the 10th percentiles have been estimated by linear extrapolation of
the cumulative frequency distribution.
The basic period for which trend data are available is 1949 to 1972.
For Charlotte (Figure 17), the trend analysis is started in 1955 because of a
change in the observation site in 1954. Data for Dulles do not start until
1963. As explained in Chapter 2, the two Air Force bases (Columbus, Ind. and
Wilmington, Ohio) have been excluded from the trend study because of missing
data and inconsistencies over time in reporting practices.
As evidenced by Figures 12 to 21, most of the sites show improvement
in visibility from the late 1940's to the middle 1950's, followed by either
decreasing or nearly constant visibility from the middle 1950's to the early
1970's. The improving trend in the late 1940's and early 1950's has been
noted earlier by Holzworth (1962), who attributed the improvement to the
switch toward cleaner fuels (from coal to oil and gas) during that period.
Some of the improvement may also have been due to meteorological trends.
Table 6 summarizes net changes in visibility from the middle 1950's to
the early 1970's. As indicated by the table, little change occurred in the
41
-------
30-i
- Yearly Values
- Three-Year Moving Averages
1/5
(1)
-0
l/l
25 -
10th Percent!le
(Estimated)
20-
15-
10-
5-
90th Percent!le
i i r i r I IT i i [ i i i i i i r i i i i
1950 1955 1950 1965 1070
Year
Figure 12. Long-term visibility trends at Washington National
42
-------
30 _
25
20
O)
f
E
15
10
j» Yearly Values
Three-Year Moving Averages
10th Percent!le
(Estimated)
A _.--
50th Percent!le
90th Percentile
«. ^
"i | i i i i\iiii|\itr
1950 1955 1960 1965
Year
Figure 13. Long-term visibility trends at Chicago.
1 I ' '
1970
43
-------
30-,
_ ^ Yearly Values
, Three-Year Moving Averages
25
20
10th Percentile
(Estimated)
OJ
15
10
50th Percent!1e
f.
90th Percent! le
i r i
1950
1955
^ r i r i i I |
1960 1965
Year
i | ' '
1970
Figure 14. Long-term visibility trends at Newark.
44
-------
30*
« Yearly Values
.« Three-Year Moving Averages
25 _
20 _
OJ
E
15
10th Percent!le
(Estimated)
50th Percentile
b_
90th Percenti le
| i . i . |
1950 1955
I I I I 1 I I I I I | T 7
1960 1965 1970
Year
Figure 15. Long-term visibility trends at Cleveland.
45
-------
30-1
25-
20
cu
15
10
. Yearly Values
Three-Year Moving Averages
K
i *
/ \
i '
, i
10th Percentile
(Estimated)
90th Percent!le
T I '
1950
I 117 I 1 1 l I I
1955 1960 1965
1960
Year
Figure 16. Long-term visibility trends at Lexington.
1970
46
-------
30-1
Yearly Values
Three-Year Moving Averages
,*-.
25-
20
10th Percentile
(Estimated)
15
10
50th Percentile
90th Percent!le
»
-*.
| - 1 I - 1 I |
1950 1955
I - 1 I | - 1 - 1 - 1 I | - 1 - 1 I - 1 j T
1960 1965 1970
Year
Figure 17. Long-term visibility trends at Charlotte.
47
-------
30 -r Yearly Values
Three-Year Moving Averages
25-
20
CJ
15
10th Percent!le
(Estimated)
10
50th Percent!le
5
90th Percent!le
i I i I 1
1950
i 1 i I i r i i i I i i r
1960 1965 1970
Year
Figure 18. Long-term visibility trends at Columbus.
48
-------
30-T
Yearly Values
Three-Year Moving Averages
25 _
20
10th Percentile
(Estimated)
>> 15
10
v *^N>*-*-'?<^t~
SOth Percentile
^-t^^^tr^r^^r
i960
1955
i | i
1960
Year
196
1970
Figure 19. Long-term visibility trends at Dayton.
49
-------
30-i
25-
20
- Yearly Values
-« Three-Year Moving Averages
10th Percentile
(Estimated)
to
OJ
-Q
to
15
50th Percent!le
10
5
90th Percent! le
1950
1955
|
1960
Year
1965
1970
Figure 20. Long-term visibility trends at Williamsport.
50
-------
30 ^ ^
Yearly Values
Three-Year Moving Averages
25-
/ \
10th Percentile
(Estimated)
20 _
15
-O
r-
LO
50th Percent!le
10
90th Percent!le
iiijiiiijr
1955 1960
\Ii 1 \| 1I 1I|
1965 1970 1975
Year
Figure 21. Long-term visibility trends at Dulles.
51
-------
TABLE 6. NET PERCENT CHANGES IN VISIBILITY,
1953-1955 TO 1970-1972
LOCATION
CHANGES IN THREE-YEAR AVERAGES, 1954 to 1971
Best (10th«) Median Worst (90th%)
Visibility Visibility Visibility
METROPOLITAN
Washington, DC
Chicago, IL
Newark, NJ
Cleveland, OH
Average for Metropolitan Sites
URBAN/SUBURBAN
Lexington, KY
Charlotte*, NC
Columbus, OH
Dayton, OH
Average for Urban/Suburban Sites
NONURBAN
Williamsport, PA
Dulles , VA
Average for Nonurban Sites
-2%
-12%
+3%
-16%
-7%
-38%
-24%
-16%
-10%
-22%
+39%
+44%
+41%
-8%
-6%
+14%
-10%
-2%
-41%
-33%
-11%
-9%
-23%
-9%
-25%
-17%
+5%
-3%
+21%
-24%
0%
-47%
-36%
-30%
-31%
-36%
-37%
-42%
-39%
Trends for these two locations are extrapolated to cover the period
1954-1971.
52
-------
haze levels at the metropolitan sites. Visibility increased slightly at
Newark, while visibility decreased slightly at Washington, Chicago, and
Cleveland. In aggregate, the metropolitan sites show approximately a 5% de-
crease in visibility from 1953-1955 to 1970-1972.
*
With the exception of the 10th percent! les at Williamsport and Dulles,
the urban/suburban and nonurban locations show considerable decreases in
visibility, on the order of 10 to 40%, from 1953-1955 to 1970-1972. The great-
est decrease in visibility, approximately 30 to 40%, occurred at the two
southernmost locations, Lexington and Charlotte. The increase in haze at
urban/suburban and nonurban locations is consistent with other findings pub-
lished in the literature. Miller et al (1972) reported substantial decreases
in summertime visibility from 1962 to 1969 at three suburban airports (Akron,
Ohio; Lexington, Kentucky; and Memphis, Tennessee). Using sun photometers,
Peterson and Flowers (1977) found increases in atmospheric turbidity from the
1960's to the 1970's at four suburban/nonurban locations (Meridian, Mississippi;
St. Cloud, Minnesota; Oak Ridge, Tennessee; and Raleigh, North Carolina).
Another way of expressing visibility trends is to compute changes in
the extinction coefficient. Here it is useful to examine only "extra" ex-
tinction, the fraction of extinction above and beyond the constant contri-
bution from blue-sky (Rayleigh) scatter. Given visibility, V in [miles],
extra extinction is computed according to the expression
=
with units of [10 meters]" .
Table 7 summarizes the net changes in extra extinction from 1953-1955 to
1970-1972. Viewed as a whole, the net change in extra extinction at the
metropolitan sites was quite small, an increase of about 5%. With the ex-
ception of the 10th percentiles at Williamsport and Dulles, the urban/suburban
and nonurban sites showed substantial increases in extra extinction, on the
order of 10 to 80%. The two largest increases occurred at Lexington (approx-
imately 80%) and Charlotte (approximately 50%).
*
The anomalous trends in the 10th percentiles at Williamsport and Dulles may,
in part, be an artifact produced by the extrapolation techniques used to es-
timate 10th percentile visibility.
53
-------
LOCATION
TABLE 7. NET PERCENT CHANGES IN EXTRA EXTINCTION,
1953-1955 TO 1970-1971
CHANGES IN THREE-YEAR AVERAGES, 1954 to 1971
METROPOLITAN
Washington, DC
Chicago, IL
Newark, NJ
Cleveland, OH
Average for Metropolitan Sites
URBAN/JSUBLIRBAN
Lexington, KY
Charlotte , NC
Columbus, OH
Dayton, OH
Average for Urban/Suburban Sites
NONURBAN
Williams port, PA
Dulles*, VA
Average for Nonurban Sites
Best (10th%)
Extinction
+2%
+17%
-4%
+21%
+9%
+74%
+37%
+21%
+13%
+36%
-32%
-35%
-33%
Median
Extinction
+9%
+7%
-13%
+12%
+4%
+79%
+55%
+12%
+11%
+39%
+11%
+37%
+24%
Worst (90th%)
Extinction
-5%
+3%
-18%
+33%
+3%
+90%
+59%
+43%
+47%
+60%
+58%
+74%
+66%
Trends for these two locations are extrapolated to cover the period
1954-1971.
54
-------
SEASONAL TRENDS IN VISIBILITY
It is of interest to examine historical trends in visibility according
to season. Figures 22 through 31 present historical visibility at the ten
study locations, disaggregated by quarter of the year. The striking feature
of Figures 22 to 31 is the strong downward trend in visibility during the
summer (third) quarter. In the early 1950's, the third quarter tended to be
either the best or second best season for visibility. By the early 1970's,
the third quarter was almost invariably the worst season for haze.
The historical changes in seasonal visibility from 1954 to 1971 are
highlighted in Table 8. Visibility decreased at every location during the
summer, and the summer decrease at each location was greater than the decrease
in any other season. The net percentage reductions in summer visibility from
1953-1955 to 1970-1972 are approximately 5 to 25% for metropolitan locations
and 25 to 60% for urban/suburban and nonurban locations. At four locations
(Lexington, Charlotte, Dayton, and Dulles) summer visibility decreased by
approximately a factor c-" 2 to 2.5 from 1953-1955 to 1970-1972.
Table 8 indicates that the slight downward trend in yearly visibility
at metropolitan sites is composed of moderate visibility decreases during
the summer which more than negate slight to moderate visibility increases
during the winter. The moderate downward trend in yearly visibility at
urban/suburban and nonurban locations is basically composed of substantial
visibility decreases during the summer and slight to moderate decreases during
other seasons.
Table 9 presents seasonal trends in extra extinction (above-and-beyond
blue-sky scatter), calculated according to Equation (12). Qualitatively, the
trends in extra extinction are essentially reverse images of the trends in
visibility. It is notable that, at urban/suburban and nonurban locations,
extra extinction during the summer approximately doubled from 1953-1955 to
1970-1972. The largest summertime increases were at Lexington (161%),
Charlotte (136%), Dulles (120%), and Dayton (102%).
55
-------
30-r
_ 1st Quarter
_, 2nd Quarter
_ 3rd Quarter
_, 4th Quarter
to
-------
30_
25-1
2(H
CO
OJ
E
.
to
15H
_« 1st Quarter
+ 2nd Quarter
.« 3rd Quarter
- 4th Quarter
1950
1955
~n | i i
1960
Year
~I I T
1965
I | I I
1970
Figure 23. Seasonal visibility trends at Chicago
(median level, 3 year averages).
-------
30 -i
1st Quarter
2nd Quarter
3rd Quarter
4th Quarter
25 -
20
CD
15
-Q
i/l
10
5
1950
Figure 24.
1955
1960
1 T r
1965
1970
Year
-Seasonal visibility trends at Newark
(median level, 3 year averages).
58
-------
30 -,
1st Quarter
- 2nd Quarter
.« 3rd Quarter
4th Quarter
25
20
0)
:*>
M
15
10
5
I I
1950
I I I I
1 T
1955
1960
i I
1965
Year
Figure 25. Seasonal visibility trends at Cleveland
(median level, 3 year averages).
1970
59
-------
30-] 1st Quarter
2nd Quarter
3rd Quarter
4th Quarter
251
0)
E
20
15
10-
5
1950
1955
1960
1965
Year
Figure 26.' Seasonal visibility trends at Lexington
(median level, 3 year averages).
60
. i
1970
-------
30,
25
20-
QJ
15
1950
_. 1st Quarter
._« 2nd Quarter
3rd Quarter
_ 4th Quarter
1
4
1955
1960
Year
1965
1970
Figure 27. Seasonal visibility trends at Charlotte
(median level, 3 year averages).
61
-------
30 -i
25 -
20 -
to
a;
15 -
10-
5
1st Quarter
2nd Quarter
3rd Quarter
4th Quarter
i i I i i
T I I I
1950
1955
1960
Year
i i I I
1965 1970
\ \
Figure 28. Seasonal visibility trends at Columbus
(median level, 3 year averages).
62
-------
-------
O)
30^
25-
20-
>, 15-
10
5-
-- 1st Quarter
- 2nd Quarter
3rd Quarter
* 4th Quarter
1950
1955
1960
Year
1965
1970
Figure 30. Seasonal visibility trends at Williamsport
(median level, 3 year averages).
64
-------
30,
25_
20_
Ol
i
X)
r-
1/1
- 1st Quarter
_. . 2nd Quarter'
3rd Quarter
4th Quarter
2
4;
1'
4
1
- 2
1950
I I I | 1 1 I I | I I I
1955 1960
1965
197 (
Year
Figure 31.
Seasonal visibility trends at Dulles
(median levels, 3 year averages).
65
-------
I 1
en ID
l-H CU
i >-
LO CU
en i-
r 1 !
>- UJ
1
oo u
u~> O
1 *
> > 1
z
O Z 4J
OO -t O
UJ O
oo uj
s:
OO LJ_ Q.
O CU
>- oo
h- 00
>-i UJ 1
_i O
1-1 S >»
oo ex r-
1-1 UJ 3
oo > *o
^^
CC CU
z CM - rs
oo i i i i
UJ 1 UJ
to o uj s-
z r^ a: a.
< en z <
Z -! 1
o
o z
1 » -« S.
Z (0
UJ LO OO SI
C_> LO UJ
cc en cu i
UJ r-H Z
Cu 1
r~
^
o
CL.
O
£_
+£
e_> cu
zr o z s:
"^ °
i i C i i ~3 « O
_1 O Z -0 4-
O -M " C
r> rr» Q A (O CU
O c cn j«: r- cn
o: >- S-
LU in -r- s cu cu
2! (0 .C 01 r >
3 o z o <:
I-H co ? i en co
«3- CO ^H 1 CM
lit i
- Z 3
oa i<: z
Q; O $-
ro »* z o
OO C CU " O 4-
OO 4-> -4-> 3 " CU
^ O1 O JO C O>
z c i E o J-
OQ X (O r >, O)
Q£ O) JZ O (C >
"~1 1 C_J (_J Q C3C
en to r^
1 CM t-l
<£> LO O
1 1
as as as
LO CM ex>
CM LO CO
as as as
o en <3-
i i t i i
+ i
o*s as as
oo en co
1 CM I-H
m
o>
^B
oo
g*
re
i.
<: 3
Q- C
O
> z:
z i- o
CO Q. 4-
Qi W >
=3 E *
3 Q et
t I
1^
cn
r-l
I
LO
en
«^
o
o
r-
S.
O)
CL
O)
JZ
S-
o>
o
o
o
+J
o
Ol
re
o
Q.
(O
i.
X
cu
eu
i.
to
in
O
r-
^_>
re
u
o
o
.j^J
Ol
in
Ol
JZ
i-
o
4-
to
o
C
cu
S-
1
K
66
-------
rH
CD
i 1
O
to
LO >-
o^
r-H CO
i-
Z 4J
O C
i i LU
CJ
Z
> r-
Z X O
O LU Q)
IO Q
> LU O
CO O
p»
z <
o -
H O 4J
1 LU CX
o s: >
LU e
-
C9 O 1 S-
zr^ LU Q.
^^ O^ LiJ >^C
a: rn a:
o 3:
o H-
h- 1 S-
Z Z 10
LU LO ii s:
CJ LO
OC CT> 00 1
LU 1 LU
Q- 1 CD C
ro Z IO
» LO «C 0
LU cr> a:
Z 1 CJ
*
O^
LU
1
CO
ef.
I
Z
O
i »
t
0
o
I
s« &s a« »s 5^
CTI r*^ co CM it"
+ + «-H I-I +
1 +
3"^ ^^ S^ d^ 5^
CM oo »-< cn o
+ + CM +
3^ &R &£ ^^ ^5
»-i . i «3- co r^
»a- co + co CM
d^ ^^ ^% d^ &^
VD ijD O O LH
+ 1 1 t 1 t 1 +
+ 1 +
d*£ ^^ %v^ ^9 &%
O V£> CO CO i-i
t ( i I CM + « 1
III 1
I/)
O C CT) ^ r O)
a: -i- 10 s- ai 10
1 ^ 0 ro > i-
LU to !- 2 a> a>
S fO -C 0) >
3 <_> z o <:
3* o-S &5 &S 6-5
Cf* LO CM « i en
r^ LO i-i -( co
+ + + + +
%^ &^ d^ ^^ ^^
CM I «^- CO CM
LO CO + + CM
"t" "^ ^
d^ d^ 3^ ^^ 3^
t » ID ^. CM t I
vo co «a- o »-H
1 « t-H + t 1 t 1
^ ^* ~f" "f"
^Q d^ d^ &^ %^
CM i < O CO «tf"
< i r^ CM f-n LO
-H + + + ^»
^
$% ^^t d^ d^ &C
r>. co vo o> r^
CO -f 1 1 +
in
O)
M
r-
10
C
(^
S-
3
3
00
x^
c
10
Z 0 S-
- Z r>
OO ^ "X.
a: o s-
r> >* a: o
m c GJ * O **
^ O *-* to
oo
^-^ O> o -Q c cn
Z C r- £ O <0
^C *" ^ 3 4J ^.
CO X IO r >, O)
or a» ^r o 10 >
r) _i <_>(_) o
r
IO
c
(T3
^
i-
a: 3
a. c
o
« z
4^
Z S- ct S_
«f O S> O
CO Q. M-
Q^ I/) *
r? E *
3 Q
-C
"^
a;
o
<_>
o
4->
o
OJ
4^
IO
r^
O
CX
f3
^
4^
X
OJ
a*
^
ro
to
C
O
r
4J
ro
O
0
^»
O
3
4->
ai
I/)
cu
.c
4->
S_
O
4-
1/1
a
c
0)
Sb.
r
*
67
-------
DISCUSSION OF VISIBILITY PATTERNS AND TRENDS
A natural question to pose at this point is "What caused the historical
visibility changes in the Northeast?" Because the results of our visibility/
pollutant regressions (Chapter 5) indicate that sulfate aerosol is the major
contributor to haze in the Northeast, one might hypothesize that the visibility
changes were related tosulfate changes. Indeed, this hypothesis becomes very
plausible if one considers the following similarities between sulfate patterns
and visibility patterns:
From the early 1960's to the early 1970's, sulfate concentrations
did not change greatly at most urban sites in the Northeast,
but increased substantially at nonurban sites (EPA 1975; Trijonis
1975; Frank and Posseil 1976). Visibility has changed very little
at metropolitan locations, but has decreased substantially at urban/
suburban and nonurban locations.
By the early 1970's, nonurban sulfate concentrations in the Northeast
(averaging ~10 yg/m^) were nearly as great as urban sulfate con-
centrations (averaging -14 ug/m3) (EPA 1975; Frank and Posseil 1976).
Visibility in nonurban areas tended to be only slightly better than
visibility in metropolitan areas.*
a From the middle 1960's to the early 1970's sulfate concentrations in
the Northeast increased substantially during the third calendar
quarter relative to sulfate concentrations in other seasons (Frank
and Posseil 1976). By the early 1970's, these trends had produced
a distinct third quarter maximum in the seasonal pattern for sulfates.
Likewise, the decreasing trend in visibility was especially pro-
nounced during the third quarter. By the early 1970's, the seasonal
pattern for visibility exhibited a distinct minimum in the third
quarter.
In the middle 1960's, the area of greatest sulfate concentrations
centered around the Ohio Valley. By the early 1970's, the area of
greatest sulfate concentrations had expanded in a southeasterly
direction (Frank and Posseil 1976). The largest decreasing trends
in visibility were observed in the southeasterly part of the North-
east quadrant (i.e. at Lexington and Charlotte).
*
Actually, the urban/nonurban difference in visibility is even smaller than
the urban/nonurban difference in sulfates. This may be explained, in part,
by the fact that visibility observations represent integrals over several
miles, while sulfate data are point measurements.
68
-------
An area of decreasing sulfate trends existed in New York State, and
the New York City, northern New Jersey metropolitan area extending to
Philadelphia (Frank and Posseil 1976). This was the only area where
we found an increasing trend in visibility, (i.e. at Newark).
A companion report to this project (Husar and Patterson 1978) provides
data on historical SO emission trends, by source type and by season. The
/\
patterns in these emission trends lead us to propose the following hypotheses
as explanations for the sulfate and visibility trends:
The increases in sulfates (ana decreases in visibility) at nonurban
locations in the Northeast are related to the substantial increase
that occurred in SOX emissions from nonurban, tall-stack sources
(power plants).
In most metropolitan areas, sulfates (and visibility) have remained
approximately constant because the increase in background sulfates
was negated by the effect of reduced local SOX emissions from res-
idential, commercial, and industrial sources.
Sulfates (and visibility) did not show strong trends in the winter
because the power plant emission increase was not as large in the
winter as in the summer and because most of the SOX reduction from
commercial and residential sources occurred in the winter.
Sulfates rose (and visibility fell) dramatically during the summer
because the growth in power plant SOX was especially pronounced dur-
ing the summer and because there was little SOX reduction from other
sources during the summer. The increase in summertime sulfate may
also have been related to increases in photochemical smog, from
hydrocarbon and NOX emission growth, which would promote more rapid
and complete oxidation of SC.
69
-------
CHAPTER 5
VISIBILITY/POLLUTANT RELATIONSHIPS
Before control strategies can be planned for maintaining or improving
visibility in the Northeast, the atmospheric components that contribute to
visibility reduction must be identified. This chapter relates airport visi-
bility measurements to Hi-Vol particulate measurements in order to gain in-
sight as to the causes of haze in the Northeast. The analysis is based on
regression equations relating daily estimates of extinction coefficient to
TSP, sulfate, nitrate, and relative humidity. The statistical methods and
their limitations are discussed in detail in Chapter 2.
DATA OVERVIEW
In this report, visibility/pollutant regression studies are conducted
for three metropolitan locations (Chicago, Newark, and Cleveland) and three
urban/suburban locations (Lexington, Charlotte, and Columbus). The data base,
summarized in Table 3 (page 21), consists of daytime visibility and relative
humidity measurements taken at airports, combined with daily TSP, sulfate and
nitrate measurements taken at nearby NASN monitoring sites (from 2 to 10 km
away from the airports). All days of NASN sampling for the years 1966
through 1972 are included, eliminating only those days when precipitation
*
was reported at the airport.
Table 10 lists the number of data points and the average values for the
pertinent variables at each location under study. The definitions of the
variables, total extinction (B), relative humidity (RH), SULFATE (S), NITRATE
(N), and remainder of TSP (T), are discussed at length in Chapter 2.
Table 11 summarizes the linear correlation coefficients among the
variables at each location. Only two pairs of variables correlate signifi-
Out of the remaining data (over 700 days), we eliminated one day: July 8,
1972 at Lexington. Visibility was very good tt~at day, but the NASN readings
for TSP and sulfate were higher than for any other day at Lexington. The NASN
recordings appeared to be invalidated by a measurement taken on the same day
and at the same location by the Kentucky Division of Air Pollution Control
which resulted in a TSP value less than one-fourth of the NASN value.
70
-------
oo
ui
ii
Q
r>
OO
o
Q.
CO
Q
LU
O
OO
o
o:
o
oo
OO
>-
ca
oo
O
CO
71
-------
CO
O C\J O CO O
O ii CM O O
OO
l-«t
OO
>-
i
£
«C
O
a.
Z O O 1-1
o
o
o
. en \cn
ool PO
o
o
oo
Z3
CQ
O
o
O)
>
OJ
oo
II
CO
o
ea o
o
CQ O
o
CQ O
a*
o
c
OJ
Q
LU
OO
^3
OO
CQ
=c
OO
o
CO
OO
CO OH OO
CM r-v LD o
O O i-H O
ca
OO
C
o
o
IT)
03
o
s_
OJ
o
4-
c
(T3
4-
c
QC
a:
o
o
o
o:
LU
II
O
CM \O
CO 1O
o
CQ O
O
CQ O
O
CQ O
co o: oo z: t
ca
72
00
co a: oo
-------
cantly at all six locations; these pairs, both exhibiting linear correlations
from approximately 0.3 to 0.6, are extinction vs. relative humidity and ex-
tinction vs. sulfates. Two other pairs of variables (remainder of TSP vs.
sulfates, and remainder of TSP vs. nitrates) correlate significantly at five
of the six locations.
MULTIVARIATE REGRESSION
Stepwise multiple linear regressions relating daily extinction coef-
ficient to the other four variables are conducted according to Equation (6),
page 23. The results of these regressions, retaining only those coefficients
that are greater than zero at a 95% confidence level, are presented in Table
12. The total correlation coefficients are 0.48 at Chicago, 0.81 at Columbus,
and 0.67 to 0.70 at the other four locations. At the 95% confidence level,
the multiple linear regressions retain relative humidity at all six locations,
SULFATE at five of the locations, and NITRATE and the remainder of TSP at
only one location each. As will be demonstrated in later discussions, the
coefficients in the regression equations (extinction coefficients per unit
mass for the pollutant variables) are consistent with basic principles and
with other published values.
As evidenced by the total correlation coefficients, the results of the
regression analysis are considerably weaker at Chicago than at the other five
locations. The explanation most likely lies in air mass differences between
*
the airport (Chicago/Midway) and the NASN location (Herman 1977).
Stepwise multiple regressions are also conducted using the nonlinear
RH regression model, Equation (7), page 23. The results of these regressions,
again retaining only those coefficients that are greater than zero at a 95%
confidence level, are presented in Table 13. For each location the nonlinear
RH model attains a higher total correlation coefficient than the linear model
even though there is one less free parameter in the nonlinear RH regression
*
When we originally decided to include Chicago in the visibility/pollutant
analyses, we thought that the Chicago NASN site wa: within 3 km of the visi-
bility observation site (Midway airport). We later discovered that the la-
titude/longitude information contained in the Chicago NASN site file was
wrong and that the site is actually located nearly 20 km from the airport.
73
-------
t/>
^^
o
CO
CO
LU
rv^
CD
LU
Cg
h-
z
-
H-
t4
_J
14
CO
* -4
(/)
1 *
<:
"5*
-^
o
^-H
1
LU
Z
O
1*
co
co
1 i 1
LI '
to
LU
ry
LU
01
r
0
CO
eg
^"1
CO
23
t/)
"v^
z.
^£
co
eg
Z3
C
O
C7)
c
r-
X
O)
_J
OJ
4_J
o
S-
(O
JT
(J
(/)
3
t~t
B
3
r-«
O
O
74
-------
oo
oo
oo
LU
DC
C3
LU
oc
o
a.
oo
oo
a:
oc.
o
i
.i
\
a?
i
oo
1
-> o:
oo
> i
>
5
ex:
^^
LU
Z.
t 1
1
o
Lu.
0
OH
OO
00* 2: |_
«- o z.
^J 1 l_l
^^ LU
LU . 1 i i
i -^ <^ tJ
CO pE _ J
? Pat
1 ae LU
00
O <_)
^y
O
»-«l
1
o
o
_J
^^
LU
1
<£
U.
1
wJ
00
^J.
t 1
o
CVJ
CVI
CVJ
II
CO
CVJ
IT)
*
z
1
1
o
O- O
o o>
C£ 03
1 U
LU -i-
^ ^
O
\
LU
1
00
UD
CVJ
O
o
o
II
co
1 1
f--
*
IS
c_
oo
t£>
CVI
O
^J"
i i
II
CO
CVJ
r*^
*
co
on
13
-o to
C 3
« oo
"QJ z
>
4-* -4-J
o> o
C i
f- i.
X to
-------
equation. The total correlation coefficients are now 0.52 for Chicago, 0.90
for Columbus, and 0.71 to 0.73 for the other four locations.
The SULFATE/(1 - .01RH) variable appears in the equations for all six
locations and is the most significant variable (according to an F-test or
t-test) at all six locations. The partial correlation coefficients between
extinction and SULFATE/(1- .01RH) alone are 0.52 for Chicago, 0.89 for
Columbus, and 0.65 to 0.73 for the other four locations. At two of the sites
(Newark and Lexington), the variable (TSP - S - N)/(l - .01RH) is significant,
and at one site (Columbus), NITRATE/(1 - .01RH) is significant. As demonstra-
ted later, the regression coefficients (extinction coefficients per unit mass
adjusted for relative humidity) are again very reasonable according to
fundamental principles and other published values.
EXTINCTION BUDGETS
As explained in Chapter 2, the regression equations can be used to
derive extinction budgets which indicate the fraction of haze, on the average,
that is attributable to each pollutant species. Table 14 presents extinction
budgets for the six locations under study. In computing those extinction
budgets, we have used the nonlinear RH regression models rather than the
linear regression models because the form of the nonlinear RH models is more
reasonable on physical grounds and because the nonlinear RH models attain a
better fit to the data at all six locations.
Table 14 indicates that the extinction budgets based on the nonlinear
RH models account for the majority of extinction at each location except
Chicago. The "unaccounted for" fraction is 69% at Chicago and ranges from
19% to 43% among the other five locations. It should be stressed that the
"unaccounted for" category may represent errors in the data base (e.g. visi-
bility and pollutants measured at different locations, visibility and pollu-
tants not measured during identical portions of the day, measurement errors,
etc.) as well as extinction contributions from atmospheric constituents
omitted from the analysis. Thus, some of the "unaccounted for" category
(especially in Chicago) may, in fact, be attributable to sulfates, nitrates,
and/or remainder of TSP.
From the extinction budget, it is obvious that sulfates tend to be the
76
-------
o>
to
c
00
LU
oo
GO
et
a:
O
o
o
I
oo
z.
o
>4 I 1
( I
O CQ
O i
s.
O)
O Q-
Z C 00
O -- I
ii 0) O
z: ce
oo
z
O
ai
(O
i.
o;
o
00
a
^>
CO
0)
3
OO
oo aj
I 4->
O) 4->
3 10
O
CQ 00
DD
en cr>
CO
ro
CM
en
ro
CM
OO
CM
CO
CM
C\J
LT>
o
OO
vO
LT)
ID
LO
LO
LT>
Lf)
m
z
o
1
c^
O
o
_l
z
£
1 4
1
o
Q.
O
CE:
I
LU
s:
0
en
o
r
.C
c_?
^^
^
(O
3
cu
z
-^
c:
m
O)
>
0)
n
o
z
CQ
o:
CO
=5
OO
»^^
1
^£
CQ
fV
^)
c
O
4->
C71
r
X
CO
_j
o>
( >
o
^
-------
dominant atmospheric component related to visibility loss in the Northeast.
The estimated contributions to haze range from 26% to 65% for sulfates, 0% to
8% for nitrates, 0% to 42% for the remainder of TSP, and 4% to 5% for blue-sky
scatter by air molecules. If we take an aggregate view and average the results
for the five non-Chicago locations, we obtain the following:
Component Contribution to Total Extinction
Blue-Sky Scatter 5%
Sulfates 49%
Nitrates 2%
Remainder of TSP 16%
Unaccounted for 28%
Because the problem of colinearity between variables may distort the re-
gression results for some of the sites (e.g. by sometimes overemphasizing the
sulfate term, sometimes overemphasizing the remainder of TSP term), we tend
to have more confidence in the aggregate conclusions than we have in the
results for individual locations.
DISCUSSION OF RESULTS
Considering the potential errors in the data bases, the regression
studies for the Northeast have been quite encouraging. At all sites but
Chicago, total correlation coefficients exceeding 0.7 have been obtained
using the nonlinear RH regression model relating extinction to pollutants.
At Columbus, a total correlation of 0.9 was attained.
The results of the regressions indicate that the main contributor to
haze in the Northeast is sulfate, which typically accounts for approximately
*
50% of total extinction. This conclusion is not surprising in light of
known principles of aerosol physics. Sulfates are secondary aerosols and
tend to occur in the particle size range of 0.1 to 1 micron. In fact, field
*
Actually, the fractional contribution of sulfates to haze could be somewhat
larger than 50% if some of the "unaccounted for" category represents sulfates.
Alternatively, the fractional contribution of sulfates to haze could be some-
what smaller if sulfates are acting in part as surrogates for other pollutants
omitted from the analysis, or if sulfates have been overemphasized in the re-
gressions due to statistical problems of colinearity.
78
-------
experiments in Missouri, Arkansas, and Michigan (Charlson et al. 1974;
Weiss et al. 1977) indicate that sulfates constitute the dominant particulate
fraction (i.e. 1/2 or more) in the 0.1 to 1 micron size range at those
locations. As shown in Figure 32, light scattering per unit mass of aerosol
exhibits a pronounced peak in the 0.1 to-l micron size range, around the
wavelength of visible light. Because sulfates tend to reside in the particle
size range that is optically most important, it is not unreasonable for
sulfates to account for 50% or more of visibility reduction even though they
typically constitute only 15% of total aerosol mass in the Northeast.
Further confidence is placed in these conclusions if we compare the ex-
tinction coefficients per unit mass based on our regressions to other results
published in the literature. Table 15 indicates basic agreement that the ex-
tinction coefficients per unit mass are approximately .04 to .11 (10 m) /
(ug/m3) for sulfates, .03 to .09 (104m)"1/(yg/m3) for nitrates*, and .001 to
4-1 3
.015 (10 m) /(yg/m ) for the remainder of TSP. These values for extinction
coefficients are also consistent with Figure 32. Figure 32 indicates that
secondary aerosols (such as sulfates or nitrates) residing in the .1 to 1
micron size range should exhibit average extinction coefficients per unit
mass on the order of .06 (10 m) /(yg/m ). The remainder of TSP, residing
mostly in the size range above 3 microns, should exhibit an average extinction
coefficient per unit mass that is one order of magnitude lower.
It should be remarked that the average extinction coefficients per unit
mass for sulfates that are determined by the regression models are often
slightly higher than would be expected according to Figure 32. The likely
explanation is that the recorded mass of sulfate aerosol is lower than the
ambient mass because some of the water associated with the ambient sulfate aero-
sol is lost during measurement procedures (i.e. filter equilibration). The low
estimate of ambient sulfate aerosol mass leads to a slightly high estimate of ex-
*
Actually, at most Northeastern locations we obtained no contribution from ni-
trates, i.e. extinction coefficients for nitrates were not greater than zero
with statistical significance. There are two plausible explanations. First,
contributions to extinction from nitrates may be so small in the Northeast
that the statistical methods are unable to discern the effect. Second, nega-
tive interference by sulfates on nitrate measurements (Marker et al 1977N
have masked the nitrate contributions; in this regard we note that negative
correlations were sometimes found between nitrates and extinction.
79
-------
to
c
o
o
i-
OJ
-»->
Ol
£
fO
o
r
S-
O)
O I
i- a;
-(-> S-
13 H
CL O
4- S_
O (O
C Q.
(O
in >,-(->
O !-> S-
i_ T- 0)
O) tO J2
(O C O
(1) CC.
>,T3
J3 T3
C C
r- 3
O -C
(J
tO
^Z CL
01 E X
r- O OJ
0-0
c
o --
i +J -r-
(O O) -M
EE o
03 tO
O-i-l.
CM
ro
a>
80
-------
REFERENCES
Bolin, B. and R.J. Charlson, "On the Role of the Tropospheric Sulfur Cycle
in the Shortwave Radiative Climate of the Earth", Ambio, Vol. 5, No. 2,
1976.
Cass, G.R., "The Relationship Between Sulfate Air Quality and Visibility at
Los Angeles", Caltech Environmental Quality Laboratory Memorandum No. 18,
Pasadena, August 1976.
Charlson, R.J., "Atmospheric Visibility Related to Aerosol Mass Concen-
tration, A Review", Environmental Science and Technology, Vol. 3, No. 10,
pp. 913-918, October 1969.
Charlson, R.J., A.M. Vanderpol, D.S. Covert, A.P. Waggoner and N.C. Ahlquist,
"H2S04/(NH4)2S04 Background Aerosol: Optical Detection in St. Louis Region",
Reprinted from Atmospheric Environment, Vol. 8, pp. 1257-1267, by Pergammon
Press, Great Britain, 1974.
EPA, "Position Paper on Regulation of Atmospheric Sulfates", Publication
No. EPA-450/2-75-007, September 1975.
Frank, N. and N. Posseil, "Seasonality and Regional Trends in Atmospheric
Sulfates", Presented before the Division of Environmental Chemistry,
American Chemical Society, San Francisco, CA, September 1976.
Green, C.R. and L.J. Battan, "A Study of Visibility Versus Population
Growth in Arizona", Journal of the Arizona Academy of Science. Vol. 4,
Arizona University Press, pp. 226-228, October 1967.
Hagen, L.J. and N.P. Woodruff, "Air Pollution from Duststorms in the Great
Plains", Atmospheric Environment. Vol. 7, pp. 323-332, 1973.
Harker, A.B., L.W. Richards and W.E. Clark, "The Effect of Atmospheric S02
Photochemistry Upon Observed Nitrate Concentrations in Aerosols", Atmos-
pheric Environment, Vol. 11, pp. 87-91, 1977.
Hartman, W.K., "Pollution: Patterns of Visibility Reduction in Tucson",
Journal of the Arizona Academy of Science, Vol. 7, pp. 101-108, Oct. 1972.
Herr;ien, J., CMrago Department of Air Pollution Control, Persona] Com-
munication, October 1977.
Holzworth, G.C. and J.A. Maga, "A Method of Analyzing the Trend in Visibi-
lity". Journal of the Air Pollution Control Association. Vol. 10, No. 6,
PP. 430-435, December 1960.
Holzworth, G.C., "Some Effects of Air Pollution on Visibility In and Near
Cities", Symposium: Air Over Cities, SEC Technical Report A62-5, pp. 69-
38, 1962.
-------
Holzworth, G.C., Meteorologist - EPA Environmental Science Research Labor-
atory, Persona] Communication, April 1977.
Husar et al., "A Study of Long Range Transport from Visibility Observations,
Trajectory Analysis and Local Air Pollution Monitoring Data", Presented at
the 7th International Technical Meeting on Air Pollution Modeling and its
Application, Airlie, Virginia, September 1976.
Husar, R. and D. Patterson, Report in preparation for EPA Office of Research
and Development under Grant No. 803896, Washington University, St. Louis,
Missouri, 1978.
Kerr, Jr., R.E., "Effects of Weather on Air Quality Measurements: San Gabriel
Valley Ozone - Meteorological Model", Working Paper, June 1974.
Keith, R.W., "A Study of Low Visibilities in the Los Angeles Basin, 1950-
1961", Los Angeles Air Pollution Control District, Air Quality Report
No. 53, 1964.
Keith, R.W., "Downtown Los Angeles Noon Visibility Trends, 1933-1969", Los
Angeles Air Pollution Control District, Air Quality Report No. 65, 1970.
Miller, M.E., N.L. Canfield, T.A. Ritter, and C.R. Weaver, "Visibility
Changes in Ohio, Kentucky, and Tennessee from 1962 to 1969", Monthly
Weather Review, Vol. 100, No. 1, pp. 67-71, January 1972.
Neiburger, M., "Visibility Trend in Los Angeles" Southern California Air
Pollution Foundation Report No. 11, 1955.
Peterson, J.T. and E.G. Flowers, "Interactions between Air Pollution and
Solar Radiation", Solar Energy. Vol. 19, 1977.
Rand McNally Company, 1977 Commercial Atlas and Marketing Guide, 1977.
Robinson, E., "Effects of Air Pollution on Visibility", in Air Pollution,
A.C. Stern, editor, Vol. 1, 2nd edition, Academic Press, New York, 1968.
Spicer, C.W. and P.M. Schumacher, "Interferences in Sampling Atmospheric
Particulate Nitrate", Submitted to Atmospheric Environment, Battelle
Columbus Laboratories, 1977.
is, J., "The Relationship of Sulfur Oxide Emissions to Sulfur
Dioxide and Sulfate Air Quality1', Air Quality and Stationary Source Emission
Control , National Academy of Sciences, Prepared for Committee on Public
Works, United States Senate, March 1, 1975.
84
-------
Trijonis, J. and K. Yuan, "Visibility in the Southwest: An Exploration of
the Historical Data Base", EPA-600/3-78-039, Environmental Protection
Agency, Research Triangle Park, North Carolina, 1978.
Ursenbach, W.O., A.C. Hill, W.H. Edwards and S.M. Kunen, Atmospheric
Particulate Sulfate in the Western United States", Paper No. 76-7.5,
Presented at the 69th Annual Meeting of the Air Pollution Control
Association, June 1976.
Waggoner, A.P., A.J. Vanderpol, R.J. Charlson, S. Larsen, L. Granat and
C. Tragardh, "Sulfate-Light Scattering Ratio as a- Index of the Role of
Sulphur in Tropospheric Optics", Nature, Vol. 261, No. 5556, pp. 120-122,
May 1976.
Weiss, R.E., A.P. Waggoner, R.J. Charlson and N.C. Ahlquist, "Sulfate
Aerosol: Its Geographical Extent in the Midwestern and Southern United
States11, Science. Vol. 197, pp. 979-981, March 11, 1977.
White, W.H. and P.T. Roberts, "On the Nature and Origins of Visibility-
Reducing Aerosols in the Los Angeles Air Basin", Atmospheric Environment,
Vol. 11, p. 803, 1977.
Williamson, S.J., Fundamentals of Air Pollution, Addison Wesley, Reading,
Massachusetts, 1973.
Zeldin, M.D. and W.S. Meisel, "Guideline Document on Use of Meteorological
Data in Air Quality Trend Analysis", Prepared at Technology Service Corpor-
ation under Contract No. 68-02-2318, for EPA Office of Air Quality Planning
and Standards, Monitoring and Data Analysis Division, Research Triangle
Park, North Carolina, November 1977.
85
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
'WW/3-78-075
2.
3. RECIPIENT'S ACCESSIOf*NO.
4. TITLE AND SUBTITLE
VISIBILITY IN THE NORTHEAST
Long-Term Visibility Trends and Visibility/Pollutant
Relationships
5. REPORT DATE
August 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
J. Trijonis and Kung Yuan
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Technology Service Corporation
2811 Wilshire Boulevard
Santa Monica, California 90403
10. PROGRAM ELEMENT NO.
1AA605 AG-17 fFY-771
11. CONTRACT/GRANT NO.
803896
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTP, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park. North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Interim 10/77 -_ 4/78
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
This research was supported under EPA grant 803896 to Washington University,
R.B. Husar, Principal Investigator.
16. ABSTRACT
The historical data base pertinent to visibility in the Northeast is analyzed.
The data base includes approximately 25 years of airport visibility observations and
more than 10 years of NASN particulate measurements. The investigation covers
existing visibility levels, long-term trends in visibility, and visibility/pollutant
relationships.
Visibility in the Northeast is rather poor, median visual range being on the order
of 10 niles. Visibility is not now substantially better in nonurban areas than in
metropolitan areas of the Northeast. From the middle 1950's to the early 1970's,
visibility exhibited only slight trends in large metropolitan areas but decreased on
the order of 10 to 40% at suburban and nonurban locations. Over the same period,
visual range declined remarkably during the third calendar quarter relative to other
seasons, making the summer now the worst season for visibility. The decrease in
visibility during the summer was especially notable at suburban and nonurban locations,
where atmospheric extinction apparently increased on the order of 50 to 150% during
the third calendar quarter.
Regression models based on daily variations in visibility and pollutant concentra
tions indicate that sulfate aerosol is the single major contributor to haze in the
Northeast. Sulfates apparently account for approximately 50% of total extinction.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
COSATl Field/Group
*Air pollution
*Aerosols
*Sulfates
Visibility
*Trends
*Haze
*Mathematical models
Northeast
13B
07D
07B
12A
04B
13. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
94
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
86
-------