EPA-450/2-77-024a
October 1977
(OAQPS NO. 1.2-083)
GUIDELINE SERIES
GUIDELINE ON PROCEDURES
FOR CONSTRUCTING AIR
POLLUTION ISOPLETH
PROFILES AND POPULATION
EXPOSURE ANALYSIS
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
-------
EPA-450/2-77-024a
(OAQPS NO. 1.2-083)
GUIDELINE ON PROCEDURES
FOR CONSTRUCTING AIR POLLUTION
ISOPLETH PROFILES AND POPULATION
EXPOSURE ANALYSIS
Monitoring and Reports Branch
Monitoring and Data Analysis Division
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
October 197?
-------
OAQPS GUIDELINE SERIES
The guideline series of reports is being issued by the Office of Air Quality
Planning and Standards (OAQPS) to provide information to state and local
air pollution control agencies; for example, to provide guidance on the
acquisition and processing of air quality data and on the planning and
analysis requisite for the maintenance of air quality , Reports published in
this series will be available - as supplies permit - from the Library Services
Office (MD-35), Research Triangle Park. North Carolina 27711; or, for a
nominal fee, from the National Technical Information Service, 5285 Port
Royal Road, Springfield. Virginia 22161,
Publication No. EPA-450/2-77-024a
(OAQPS No. 1.2-083)
11
-------
PREFACE
Development of air pollution isopleths and population exposure analysis
are components of the "Air Monitoring Strategy for State Implementation Plans"
(EPA-45Q/2-77-Q10). The guideline document is intended to provide assistance
in performing these activities. A companion "Users Manual" describes existing
computer software applicable to these tasks.
iii
-------
ABSTRACT
This guideline document provides an overview of methodologies that exist
for constructing pollutant isopleth displays and for estimating population
exposure to air pollutants from air monitoring data. Actual examples of the
methodologies are presented for applications to data for the New York-New Jersey-
Connecticut area and for the Los Angeles area. This report is to assist the EPA's
Regional Offices and States in reviewing their data bases to determine feasibiltiy
of performing isopleth/population exposure analyses, and to guide regional, state,
and local air pollution control agencies in actually conducting such analyses.
Spatial presentation of air quality monitoring information in the form of
isopleth maps is useful to air pollution control agencies and highly informa-
tive to the public. Steps that must be taken to obtain an isopleth map from
air monitoring data measured at widely separated monitoring stations as well
as from dispersion-moael-simulation outputs are described. Currently available
methods of developing an isopleth map are documented for three types: manually
drawn maps, character-printed maps, and line-drawn maps. Problems associated
with each type of isopleth map are discussed.
A population exposure analysis, which combines air quality data and demo-
graphic data to estimate population exposure to air pollution, is outlined.
Data-set preparation and analysis methods are explained for estimating both
long-term and short-term exposure of the population to air pollution. Exposure
to long-term average concentrations is described by the percentage of the
population exposed above an annual National Ambient Air Quality Standard
(NAAQS). Exposure to short-term (e.g., 1 hr or 24 hr) concentrations is
quantified by the distribution of the population who are exposed above an
hourly or 24-hour NAAQS at various percentages of the time (hours or days),
IV
-------
TABLE OF CONTENTS
ABSTRACT ii
CHAPTER 1 - INTRODUCTION 1
1.1 ANOTHER WAY TO STUDY AIR QUALITY 2
1.2 POTENTIAL BENEFITS 3
1.3 PERFORMING A SPATIAL ANALYSIS . 4
CHAPTER 2 - SPATIAL DISPLAY OF AIR QUALITY INFORMATION 6
2.1 FIRST STEP IN SPATIAL ANALYSIS - FEASIBILITY 7
2.1.1 Spatial Coverage of Monitoring Data 9
2.1.2 Completeness of Air Quality Data 13
2.1.3 Other Considerations 15
2.2 MANUALLY DRAWN ISOPLETH MAPS 16
2.3 SPATIAL REFINEMENT OF AIR QUALITY INFORMATION 19
2.3.1 Spatial Interpolation of Air Monitoring Data ... 20
2.3.2 Air Quality Simulation by Dispersion Model .... 27
2.4 COMPUTER-DRAWN ISOPLETH MAPS 29
2.4.1 Character-Printed Maps 31
2.4.2 Line-Drawn Maps 36
2.4.3 Hybrid Isopleth Map 40
CHAPTER 3 - POPULATION EXPOSURE ANALYSIS 44
3.1 DEVELOPMENT OF RECEPTOR NETWORKS 45
3.2 PREPARATION OF DATA SETS 49
3.2.1 Population Data 49
3.2.2 Receptor Point Data 52
-------
TABLE OF CONTENTS (Continued)
3.3 CHARACTERIZATION OF POPULATION EXPOSURE TO AIR POLLUTION. . 53
3.3.1 Analysis of Long-Term Population Exposure 53
3.3.2 Analysis of Short-Term Population Exposure 58
REFERENCES 63
APPENDIX A. Locating Monitoring Stations on a Map 65
ACKNOWLEDGMENTS
This guideline document has been prepared by Neil Frank, Monitoring
and Reports Branch, Monitoring and Data Analysis Division, Office of Air
Quality Planning and Standards and Yuji Horie, Data Sciences Division,
Technology Service Corporation, 2811 Wilshire Boulevard, Santa Monica,
California 90403, under contract No. 68-02-2318, project No. DU-76-C190.
VI
-------
CHAPTER 1. INTRODUCTION
This guideline document is to assist the U.S. EPA, States and local air
pollution control agencies in displaying air quality information in a geo-
graphical context and reporting air quality progress to the public in a more
meaningful and understandable form. This report provides guidelines on
developing contour lines of air pollution levels (isopleth map) from air
quality data and presents an outline of the methodology that has been
developed for analyzing population exposure to air pollution. To assist air
pollution control agencies in actually doing such analysis, examples of the
methodology and the computer programs developed in recent applications to
the New York-New Jersey-Connecticut area and to the Los Angeles area are
presented.
Isopleth maps are often used to illustrate the spatial variation of air
quality and other environmental variables. An isopleth map can be drawn
manually, or it can be drawn mechanically by using a readily available com-
puter program such as SYMAP [1], Because environmental variables are measured
at widely separated monitoring stations, the performance of the spatial-inter-
polation scheme being used may significantly affect the isopleth map. Therefore,
numerous interpolation schemes employed in meteorological mapping and in other
areas are reviewed to determine their suitability for air pollution applica-
tions. The state of the art of spatial interpolation is discussed in regard
to preparing isopleth maps for air pollution studies.
Although several reports [2,3,4,14] have recently been published on popu-
lation exposure, the methods used need to be thoroughly explained before
they can find wide application in various geographical regions. This report
-------
outlines the population exposure methodology that we have developed and
discusses problems that have arisen in two applications -- one for the New
York-New Jersey-Connecticut area and one for the Los Angeles area. Because
a population exposure analysis may be performed under a variety of situations,
alternative methods, as well as the currently applied methods, are discussed.
WAY TO STUDY AIR QUALITY
The nation's air quality is routinely measured by some 4,000 air
monitoring stations all over the United States. While most studies of
air quality data have been made for individual stations, important knowledge
can be gained by collectively studying air quality information provided by
an entire network of monitors in a specific geographic area. Conventional
tabulations and graphs of air quality summary statistics, given separately
for individual stations, do not provide a convenient basis for studying
these spatial patterns of air quality. When air-quality monitoring infor-
mation is presented in a geographical context, say through use of an isopleth
map, one can readily comprehend the spatial variation in air quality.
A few economists and geographers have combined air quality data with
socio-economic data, studying air pollution in relation to economic and
demographic variables [5,6,7]. Following their lead, air pollution researchers
have developed methods to quantify ambient air quality in terms of population
exposure to air pollution instead of concentration units [2,3,8,9,14], The
methodology to perform population exposure analyses is described in this
report.
-------
1.2 POTENTIAL BENEFITS
Geographically oriented air-quality presentations supplement conventional
tabulations of air-quality sunmary statistics. Therefore, an isopleth map
of air quality data is very effective for providing:
• A visual perspective of the spatial variation in air quality and
patterns of human exposure levels;
t A meaningful reference for evaluating air quality trends in relation
to population-growth patterns, arid tempos1 arid spatial emission
trends;
• A spatially representative baseline of air quality for evaluating
arid modifying r-gulsco! / *,. ogratrc (e.g., State Implementation Plans
and new source reviews).
National Ambient Air Qun't'ty Standards (NAAQSs) have been set to protect
the public health and welfare, "-nibient air quality data provide the basic
means of quantifying whether th' public benefits by trends in air pollution
levels. To make this quantification directly relatable to public health, it
is useful to report air quality in terms of population exposure levels. To
do so requires synthesis of population anj air quality data bases.
Air pollution concentration? alone are not indicative of population
health risks, because they do not specify what fraction of the population
is exposed to various concentration levels. Fo>~ example, it may be more
meaningful for residents of an urban area to learn that the percentage of
the urban population living in areas above the primary national standard
for particulates decreased from 60* in 1970 to 10";. in l^S, than to learn that
"*
the annual mean pollutant concentrations decreased from 69.5 ug/m0 to 64.3 ;,q/m
in the same period. In the latter statement, even tho people whc know the
-------
3 3
NAAQS (75 vg/m for the primary and 60 pg/m for the secondary) may misinterpret
the implications of meeting the primary NAAQS but failing to meet the secondary
standard in both 1970 and 1975. The former statement tells people, regardless
of their knowledge, that the particulate air pollution in the urban area
improved from 1970 to 1975 and that the number of people exposed to health
risks from particulate matter was less in 1975 than in 1970.
The Population Exposure Methodology discussed in this report focuses on
exposure to air pollution above established air quality standards. Exposure
to long-term average concentrations is characterized by that percentage of
the population exposed to air pollution above the yearly average NAAQS. Ex-
posure to short-term (e.g., one hr or 24 hr) concentrations is quantified by
the average percentage of the time (hours or days) that people were exposed
above the short-term NAAQS. A more complete characterization of population
exposure is given by cumulative distributions of people's exposures to long-
term average or short-term concentrations.
1.3 PERFORMING A SPATIAL ANALYSIS
To develop an isopleth map from air monitoring data, a network of
air monitoring stations must adequately cover the area of interest. The
required size may vary from one pollutant to another. If sufficient
monitoring sites are not available, isopleth analysis might depend on the re-
sults of a calibrated dispersion model. As few as three stations may
be needed for estimating a spatial concentration distribution of photochemical
oxidant and other secondary pollutants. For the primary pollutants such as
-------
CO, TSP, and S02, a larger number of stations are required to obtain a
meaningful isopleth map over a study area. In addition, these stations
must be clustered so that concentration in the neighborhood of the monitor-
ing sites can be inferred from those measured by the stations.
Isopleth maps can be used in several ways to describe air quality.
First, an isopleth map for a single year can be used to inform the public
of the spatial variation in air quality. Second, a series of isopleth maps
for selected years can be used to present the trends in the spatial patterns
of air quality over time. Third, by computing the land area within each
isopleth level, both the regional air quality level and the percentage of
land area above the NAAQS can be determined. Fourth, by combing air quality
data with demographic data, exposure of the population to air pollution can
be determined.
In obtaining an accurate isopleth map or in determining population
exposure to air pollution, a rather ']-ate jwlysts method and some
computer software are required. The analyses method for dev:-]opiny an
isopleth rr:ap Is discussed in Chapter :/; that for perfonpbig a population
exposure analysis ~'3 describe-"' in C 3p+er 3. The computer programs for
actually doing such analyses are presanted in a separate volume of the
User's Manual.
-------
CHAPTER 2 SPATIAL DISPLAY OF AIR QUALITY INFORMATION
Ambient air quality in most major U.S. cities is routinely measured
by a network of air monitoring stations. The display of air quality monitor-
ing information in a spatial context is useful to air pollution control
agencies and highly informative for the public. However, many factors have
to be considered in displaying air quality data from a finite set of mon-
itors on a spatial scale. Considerations may include the number of stations,
configuration of monitoring sites, spatial gradient of pollutant concentra-
tions, emission source inventory, geographical features, and local climatic
characteristics.
When air monitoring stations are too few, too far apart or clustered
in too small an area, a spatial display of air quality monitoring informa-
tion cannot be made in a meaningful manner. A spatial analysis might depend
on the results of a validated dispersion mode. The first section discusses
items to be included in checking feasibility of a spatial display of monitor-
ing information.
The simplest method of displaying monitoring information in a spatial con-
text is to plot monitoring sites and air quality values on a map and to manual 1
draw an air-quality isopleth map. In drawing an isopleth line, one employs
either consciously or unconsciously, an "eyeball estimate" of air quality
at nonstation locations from air quality values measured at nearby monitoring
stations. This eyeball interpolation involves all the considerations of
pollutant-concentration dispersion characteristics that the person perceives.
The importance of these subjective factors in drawing an isopleth map is dis-
cussed in Section 2.2.
-------
Isopleth displays have been used more frequently for dispersion-
model outputs than for air monitoring data. The main reason is that air-
quality monitoring information was not adequate for developing an isopleth
map in many urban areas. In recent years, however, the air monitoring network
in metropolitan areas has expanded considerably and its data quality has
improved measurably. Section 2.3 discusses two alternate methods of estimating
air quality at nonstation locations by 1) spatially interpolating air-quality
monitoring information, and 2) applying an air quality dispersion model. Pro-
cedures involved in each of the two methods are described in separate sub-
sections. Advantages and disadvantages of each method under various situations
are also stated.
Modern computer graphic techniques can be used for drawing an isopleth
map after air quality values are estimated at sufficiently dense grid points.
Section 2.4 presents an overview of currently available computer mapping
techniques and several examples of computer-drawn isopleth maps of air
quality parameters.
2.1 FIRST STEP IN SPATIAL ANALYSIS - FEASIBILITY
Given a study pollutant and a geographical area, the first thing to
be considered is whether an isopleth analysis will be feasible. It is
suggested that the isopleth analysis be based on monitoring data. This is
the easiest way to establish trends over time and provides a basis for
convenient updates. If this is not possible, then the isopleth analysis
might be based on the believable results of a calibrated dispersion model.
-------
8
Several items are useful for establishing the feasibility of an
isopleth analysis based on monitoring data. These items include previously
published isopleth maps, monitoring location maps and a check of data com-
pleteness.
As obvious as it may sound, a good starting point is to check if
isopleth maps have been previously published. These could be based on a
past year of monitoring data or the results of a modeling exercise. This
basically indicates that an isopleth analysis can be done and will provide
a basis for comparison of relative spatial patterns with new analysis. This
comparison will also ensure that contradictory new results won't be pro-
duced.
Another important item is a map which shows the location of existing
monitors plus those which previously produced historical data. This will
show if monitoring data might serve as the basis for the isopleth analysis.
This map will help to establish the spatial coverage of the monitors and
also help to define the boundaries of the possible study area. Some hints
on locating monitoring sites on a map is discussed in Appendix A.
Next, air quality data completeness must be checked for all candidate
monitors. The check list must include:
a. Validity .of air monitoring data over a time period of interest,
say a year or a season, and
b. Historical continuity of data over several years for a trend
analyses.
-------
jLl.l Spatial Coverage of Monitoring Data
Suppose that 5 stations report hourly CO concentrations with an
acceptable data quality at three locations near highway and at two urban
background sites. In this case, it may not be feasible to draw a concen-
tration isopleth map from the CO monitoring data because each station read-
ing of CO concentration is representative for such a small area (less than
1 mi around a station) that CO concentrations at a place apart from a station
site may differ considerably from concentration readings at that station.
On the other hand, if the above example is for 0 , it is quite possible
/\
to develop an isopleth map by interpolating the station readings to places
between stations. The reason is that the concentration gradients of Ox are
generally so small that Ox concentration at any place within the urban area
can be estimated reasonably well from concentration readings at the nearby
stations.
A "representative area" of a monitoring station may be defined as a
distance from the station at which a pollutant concentration remains almost
constant, say it does not change more than ±20% from the reading at that
station. Then, based on expected pollution gradient of long-term concen-
tration statistics (e.g., annual mean and annual percentile concentration),
the "representative area" presented in Table 1 may be used as a guide of how
far from a station air monitoring data can be extrapolated for various
pollutants. The representative areas specified in Table 1 are only a guide
and are open to user modifications. For example, a site affected by local
emission sources may have a much smaller representative area than the repre-
sentive area listed in Table 1. Similarly, a monitoring site in a high
-------
10
pollution gradient area (e.g., due to complex terrain or due to high emission
density) may have a smaller representative area. On the other hand, a
representative area by a monitor located in a uniform land use zone (e.g.,
agricultural land) may be substantially larger than the typical representative
area of Table 1. As a matter of fact, areas with a similar land use tend to
have a similar level of air pollution. In many cases, a land use map will
be quite useful for refining a representative area on a site-by-site basis.
Although a representative area for the primary pollutant is influenced
strongly by a particular site characteristic, this may not be a problem
with the secondary pollutants (e.g., Ox and SO^) whose concentration
gradients tend to be moderate. One exception is that 0Y concentration
/\
gradient near a heavily traveled highway is quite sharp. Therefore, 0
J\
monitoring site near a busy highway will have a much smaller representative
area than the typical representative area for 0 monitors.
/\
Using the representative area listed in Table 1, one can check the
feasibility of an isopleth analysis for a given set of air monitoring stations
and a given pollutant. Various configurations of air monitoring stations
are given in Figure 1 to illustrate how to check the feasibility of an
isopleth analysis and to determine the study area.
Suppose that, after quality and completeness of air monitoring data
have been checked, the stations whose locations are designated by dots in
Figure 1 have been selected for a spatial analysis. By drawing a circle around
each station, with the radius being equal to the representative area, one can
determine a potential study area for a given configuration of station sites.
Figure 1-a shows that the stations are located so far apart from each other
-------
11
compared to the representative area for TSP that one cannot form a contiguous
study area. On the other hand, Figure 1-b shows that although two stations
are too distant from any other station, the rest of the stations form a con-
tiguous study area.
Table 1. Typical "Representative Area" of a Monitoring Station Estimated
for Each Major Pollutant from the Expected Pollution Gradients
of Long-Term Concentration Statistics.
Pollutant
CO
S02
TSP
N02
°x
so=
Representati
Distance from a
Less than 1 mi
3 mi
3 mi
5 mi
10 mi
20 mi
ve Area
Station
Figures 1-c and 1-d are for 0 . Although the stations of Fig. 1-c are
X
scattered as far apart from each other as those of Fig. 1-a, the large
representative area for Ox monitoring sites makes it possible to form a
contiguous study area. In Fig. 1-d, the three stations are so far apart
from the rest of the stations that the study area is split into two areas.
-------
Figure 1. Determination of Study Area from a Set of Monitoring Stations by
Using the Range of Influence (a and b for TSP, c and d for 0 ).
-------
13
When a set of monitoring stations cannot form a contiguous study area
covering the region of interest, the spatial distribution of pollutant
concentrations must be estimated by knowledge about pollution behavior
in addition to the air monitoring data. If it is known that the pollution
levels are intermediate between those of adjacent monitoring stations, then
these areas can be estimated by a credible interpolation formula. Also a
dispersion model may be helpful for depicting a spatial concentration dis-
tribution of a primary pollutant. In addition, pollution level in areas
with no monitors may be guessed from the concentration readings of monitors
in areas with the same or similar land use and similar adjacent emission
sources. AQDM [10] and COM [11] computer models are useful for S02 and
TSP, while APRAC [12] and HIWAY [13] computer models are useful for CO.
The use of these dispersion models for estimating a spatial distribution
of concentration is discussed in Section 2.3.2.
2.1.2 Completeness of Air QualityData
When air monitoring data are used for developing an air quality isopleth
map, credibility of the monitoring data must be checked. The check list
must include a) completeness of air monitoring data over a time period of
interest, and b) historical continuity of data over several years for a
trend analysis.
In reporting a seasonal or annual air quality, we have to check, for
each station, the number of valid concentration measurements during the
period of interest. In distinguishing valid stations from invalid stations,
we need a specific criterion as to the minimum number of valid concentration
measurements per station-year or station-season. The criterion for a valid
-------
14
station-year used by the U.S. Environmental Protection Agency's National
Aerometric Data Bank (NADB) will serve as a preliminary data screening
tool.
The NADB uses separate criteria for continuous and intermittent moni-
toring. For continuous monitoring, a year (or quarter) of data is valid
if at least 75% of the total number of possible observations were recorded.
For intermittent monitoring a year of data is valid if it has four valid
quarters of data. For a valid quarter, a minimum of five observations is
required. In addition, if one month has no measurements, the other two
months must have at least two observations.
In order to establish air quality trends, spatial analysis of air
monitoring data would be performed for a multi-year period rather than for
a single year. This would entail at least two isopleth maps within a
several year study period. A series of maps would show the change in the
spatial pattern of air quality and can specifically show the change in the
study area having air quality values above established air quality standards.
Ideallys only "trend" stations meeting the valid Nation-year criteria
in every year during the study period should be selected for the analysis.
In actual situations, however, such stations are usually too few to ade-
quately depict the spatial variation in air quality. One solution to this
problem is to use those stations that reported air quality data in several
years. Then the air quality at stations in the missing years can be estimated
from other data. If isopleth maps were developed only from the reporting
stations in a given year some bias might be introduced into the trend
analysis.
-------
15
There are several possible approahces for estimating missing values.
If the changes in air quality is gradual, then linear interpolation of air
quality values between neighboring years might be used. Because of the
problem of what to do with missing data in the start and end years, an
alternate approach is to make use of the region-wide average change in air
quality between successive years.
The average year-to-year change can be established from all sites with
data for each successive two year period. This change can then be applied
to estimate missing values. A third potential approach is to use a gener-
alized linear model with class variables for years and sites. Unfortunately,
this analysis of variance procedure will probably involve an unbalanced
design due to the missing values and this may not be computationally con-
venient without the aid of a statistical analysis computer package. The
missing values would be estimated from a term which describes the average
air quality value for each year and from a term which describes the average
air quality value for each site. For each of the above procedures, we can
make use of emission density and/or land use maps to select an appropriate
subset of sites that should be used for the estimation procedures.
2.1.3 Other Considerations
An adequate "area representativeness" of each monitoring site is
essential to correctly estimate a spatial distribution of air quality from
air monitoring data. If a monitoring station reports an extraordinary con-
centration caused by a local "hot soot", that station must be given special
consideration in the analysis. The "hot spot" might be indicated on the
isopleth maps as a localized condition. Because it is not representative
-------
16
of a wide area, it should not be included in a spatial analysis that
estimates air quality by spatial interpolation from neighboring monitoring
stations or in a population exposure analysis to be described in later
sections of this guideline.
Site characteristics are another important factor. Since a pollutant
concentration varies greatly with height above the ground, some adjustment
of air monitoring data might be necessary to account for the height effects
on concentration readings. Concentration readings at an elevated monitoring
site, say, 50 ft above the ground, could be low because of the greater dilu-
tion at that elevation compared to the one at 5 ft above the ground. The
height effect is great for CO and the other primary pollutants, but is not
so great for 0 arid the other secondary pollutants. Therefore, careful con-
A
si deration of site characteristics such as height and local emission influ-
ence must be given for selecting a set of stations reporting concentrations
of a primary pollutant.
2.2,_ MANUALLY DRAWN ISOPLETH MAPS
When air monitoring data are available at several station sites, the
first step necessary for obtaining an isopleth map is to plot the station
locations and concentration (or other air quality parameter) levels on a
skeleton map as shown in Fig. 2.
Figure 2-a illustrates a standard procedure for drawing an isopleth map
when one has concentration/station site information only. Starting with the
stations where the highest concentration occurs, straight lines are drawn
connecting adjacent station sites. When the concentration levels at two
-------
17
036
o
42
50
o
o 53
037
lu
o
15
o
a. Without Subjective
Considerations
36
o se
53
o
o 37
22
o
42
ll
b. With Subjective
Considerations
Figure 2. Illustration of the Stepwise Procedure for Preparing
Isopleth flaps with and without Subjective Considerations,
-------
adjacent station sites are considered, an intermediate grid point is assigned
to one or more location on the edge where one or more isopleth levels fall.
To find such an intermediate point between two stations, we use an eyeball
interpolation of the two station values, which is usually similar tc a linear
interpolation. As seen in the second illustration of Figure 2-a, we car, draw
an isopleth line around the station with the highest concentration bj con-
necting the intermediate points with the isopleth level of 50. Then, proceed to
the next isopleth level, 40. Now, we consider all stations with concentrations
near 40, plus their neighboring stations. By repeating the above procedure,
an isopleth map is completed in the third illustration of Figure 2-a. Here,
intermediate points and station values are removed to have a clear isopleth map.
The procedures described above are quite similar to those used in computer
software for drawing an isopleth map. Figure 2-b illustrates a typical
isopleth-drawing procedure employed by an air pollution expert and an
experienced meteorologist. These experts also plot the station locations
and concentration levels on a skeleton map. However, because they know the
geographical feature of the area and its effect on the concentration field,
they incorporate their subjective considerations in drying an isopleth map.
In Figure 2-b, the locations of mountains are designated by hatched areas.
Because an air pollution expert knows that a polluted air mass on one side
of the mountains does not mix with the one on the other side of the mountains,
he may draw an isopleth line separately for each of the two polluted valleys
as shown in the second illustration of Figure 2-b. With his subjective con-
sideration of the mountains, his final isopleth map, shown in the third
-------
19
illustration of Figure 2-b, is totally different from the isopleth map obtained
without considering the mountains (Figure 2-a).
The above example considers only the effect of mountains on a concentration
field. An air pollution expert may also consider potential effects on a con-
centration field of wind-flow patterns and configuration of major emission
sources. Because the effects of mountains, emissions, and wind patterns on
a pollutant concentration field are difficult to quantify, very little ela-
boration has been made in incorporating the subjective considerations men-
tioned above in a computer-drawn isopleth map. A technique to incorporate
some of the subjective considerations in a spatial interpolation scheme is
discussed in the next section.
2.3 SPATIAL REFINEMENT OF AIR QUALITY INFORMATION
The estimate of the spatial variation of air quality can be refined
by two methods — spatial interpolation of air monitoring data, and air
quality simulation by a dispersion model. The former method is appropriate
when air quality data are available at closely spaced network points. The
latter method is more suitable when air quality data are not available at
all (e.g., future air quality) or are available only at a few sites.
For most applications, a spatial distribution of pollutant concentra-
tion was obtained from dispersion-model outputs. The reason has been that,
in early years of air quality surveillance, there were only a few air
monitoring stations even in a large metropolitan area. Therefore, it was
almost impossible to estimate a spatial distribution of pollutant concen-
tration from concentration readings at a few widely separated monitoring sites.
-------
20
The monitoring data have been used to validate the model calculations. How-
ever, in recent years, most major U.S. cities and metropolitan areas have
a large and dense network of air monitoring stations. Because of this
development of air monitoring networks, we can now develop an isopleth TOP
of various air quality parameters from air monitoring data and monitoring
site information.
2.3.1 Spatial Interpolation of Air Monitoring Data
To obtain a reproducible air quality isopleth map from air monitoring
data, a credible spatial-interpolation scheme must be used to estimate con-
centrations at places other than monitoring sites.
The most commonly used interpolation scheme is linear interpolation.
There are two types of linear interpolation formulas; true linear interpo-
lation and pseudo linear interpolation. When applied to estimate a con-
centration at a place other than monitoring sites, the linear interpolation
formulas may be written as
(1)
where d. is the distance between the ith grid (or receptor) point and the jth
station site. C., j=l,2,3s is the observed air quality at each of the three
J
stations nearest to the i grid point. The sign parameter a^ takes only
-------
21
+1 inside the triangle formed by the three stations for both the true
linear and pseudo linear interpolation formulas. However, outside the
triangle, a- of the true linear interpolation formula takes either +1
j
or -1 depending upon the position of the grid point, while a- of the
J
pjeudjo linear formula takes +1 only irrespective of the position of the
grid point.
The performance of the linear interpolation formulas, Eq. (1), is
shown in Figs. 3 and 4 for a hypothetical case where observed concentrations
of 85, 15, and 60 (in arbitrary units) are assumed at points A, B, and C,
respectively. In both figures the spatial distribution of the interpo-
lated concentrations inside the triangle is generally in fair agreement
with what we would expect from the concentrations at A, B, and C. Outside
the triangle, however, the true interpolation formula tends to over-
extrapolate the concentration values assigned at A, B, and C (Fig. 3).
This over-extrapolation tendency of the true linear interpolation formula
is very undesirable for air pollution applications. The pseudo linear
interpolation formula does not show such over-extrapolation tendency (Fig. 4),
In the resulting isopleth i,;ap, however, we miss both the highest observed
concentration, 85, and the lowest, 15. Therefore, the pseudo 1inear inter-
polation formula is said to be over-smoothing particularly for its applica-
tion to air quality mapping.
To avoid the problems of over-smoothing as well as over-extrapolation,
the performance of various candidate interpolation formulas was examined.
Among the interpolation formulas examined, the parabolic interpolation
formula, given by Eq. (2), showed the most satisfactory performance [2].
-------
22
LLJ
O
cn
DISTANCE X
Figure 3. Performance of the True Linear
Interpolation Formula [2].
-------
23
UJ
O
DISTANCE X
Figure 4. Performance of the Pseudo Linear
Interpolation Formula [2].
-------
24
Cn- = 7 C, J. Vd • (2)
The concentration isopleth map obtained by the parabolic inf-: po^
formula is shown in Fig. 5 . It shews the isopleths of the highes" 85,
lowest, 15, observed concentrations. Although the parabolic -inte^po
formula is not perfect in its performance, it certainly mininiz- rhc prob-
lem of over-smoothing as seen in the pseudo linear interpolation formula
(Fig. 4). The conservative characteristic that no estimated value exceeds
the maximum observed concentration nor is any estimated value less than
the minimum observed concentration is also desirable for air pollution
applications.
In Eq. (2) k the s'jrT>piat"'0" can hp
-------
25
Ul
O
CO
o
DISTANCE X
Figure 5. Performance of the Parabolic
Interpolation Formula [2].
-------
26
grid points should be closer to the real spatial distribution of air quality
than a computer map drawn to the air monitoring data at the stations alone.
Any interpolation formula by itself cannot Incorporate, the effects
of mountains, prevailing wind patterns, and configuration of major emission
sources on the concentration pattern. Some computer algorithms (e.g., SYMAP,
LPEM, and SHEW in user's Manual; can incorporate into air quality Interpo-
lation schemes the effects which geographical features exert on pollutant
concentration fields. For example, when stations are located on boeh sides
of mountains, the interpolated concentration at a receptor point will be
determined primarily from the concentration readings at the stations located
on the same side of the mountains as that receptor grid point. This effect
can be accomplished by imposing a penalty distance to the stations on the
opposite side of the mountain. This penalty distance simulates the difficulty
of an air mass crossing the mountain. The location of mountains or other
barriers would be approximated by a series of piecewise™linear penalty
functions having a particular penalty value at each node. The technique can
be applied to take into account the effects of the prevailing wind pattern,
the configuration of large bodies of water and major emission sources.
These techniques are incorporated into spatial interpolation schemes used
in LPEM and SPEM computer models as well as the SYMAP graphics package (see
User's Manual).
Spatial interpolation of air monitoring data should not be applied
without constraints imposed by the physical reality of air pollution. The
concentration reading at a station is representative only of a specified
area about that station. As discussed in Section 2.1, the "representative
area" depends on the type of pollutant and emission source pattern and is
-------
27
inversely associated with concentration gradient. A yajid interpolated
concentration will be limited to a receptor point lying within the repre-
sentative area from the nearest station. Interpolated concentrations at
receptor points outside the representative area of the nearest station
should be denoted by some warning symbol (e.g., asterisks (*)). In this
manner, the air quality in the study area will be estimated from concen-
tration readings at appropriate monitoring stations. The study area
may turn out to be one of the four cases illustrated in Fig. 1. In case
(1-a), we would have to resort to a dispersion model or some other method
then spatial interpolation of air monitoring data to estimate air quality
dt intermediate nonstation grid point,
2.3,2 Air Q'.«a^iiy SlwulaTior1 by inversion Mode1
Dispersion models ara widely urcd to simulate tr.e spatial distribution
of air quality unde" various situation. The so-r;Jlc:-d ,-.-l iinatological dis-
persion irodel—-which uses, as imji cM,3, the Piiii^iu'; inventory and the
joint frequency distribution cf V'in- ..t;eed} wind direct/: jn, and atmospheric
stability—1:; considered useful lor *-:;tirr
-------
28
A dispersion model should first be calibrated by comparing the simu-
lated values with the observed air quality at the monitoring station(s).
Air quality at places including but not limited to the station sites would
then be given by the predictions of the calibrated model. For additional
details on modeling, refer to an upcoming OAQPS Guideline on Air Quality
Models.
A dispersion model can be used to estimate the spatial distribution
of future air quality when a proposed project is completed or when a pro-
posed control strategy is implemented. First, the dispersion model is
applied to the jsresent emission inventory and present meteorology to simu-
late present air quality. Second, the simulated values are compared with
the observed air qualities at several monitoring stations to calibrate
the dispersion model. Then, the calibrated model is applied to the
postulated emission inventory and multi-year average meteorology to simulate
the air quality likely in the future.
A critical limitation of a climatological dispersion model is that it
cannot be applied to complex terrain. When applied to flat or gentle terrain,
a climatological dispersion model does generate useful spatial information of
pollutant concentration, although it may not predict accurately a level of
pollutant concentration. Therefore, there exists a possibility of obtaining
a more accurate isopleth map by combining the spatial information from air
monitoring data and from dispersion-model outputs than an isopleth map ob-
tainable from air monitoring data or dispersion model outputs alone.
The following discussion presents a conceptual approach to accomplish
the above objective. Specific guidelines are not yet available at this time.
-------
29
Suppose that a predicted value o ve re sti "is. *.?><•: a concentration level at
one monitoring site in the south and underestimates at the other two monitoring
sites in the north. Further, suppose that the emission inventory data used
for a dispersion model are so complete that at least the spatial distribution
pattern of predicted concentration is believable. One could then obtain a
better isopleth map by superimposing the spatial distribution over an error
field, as shown in Fig. 6. The error field is a plane that passes through
the three points (xm, ym, em),m=l,2,3, where (xm»ym) is the x-y coordinates
of the m monitoring station and em is the prediction error at that station,
i.e., the observed concentration minus the predicted concentration.
When there are more than three monitoring stations, the study area should
be divided into several sub-areas formed by three adjacent stations. Then,
the above adjustment can be made separately on the predicted spatial distri-
bution for each sub-area. When a dispersion model is applied to complex
terrain, the study area should be divided into several flat segments. Then,
the adjustment can be made separately for each flat segment.
2.4 COMPUTER-DRAWN ISOPLETH MAPS
Before deciding to use a computer for drawing an air quality isopleth
map, one has to consider advantages and disadvantages of employing a computer-
graphics technique to his particular problems. Although a computer-drawn
isopleth map is neat and objective, it requires some familiarity with computer
graphics and a lot of data-base preparation. When prepared carelessly, a
computer-drawn isopleth map can be quite erroneous. As mentioned in Section 2.2,
one can incorporate into his hand-drawn isopleth all the considerations of
pollutant-dispersion characteristics in a particular geographical area.
These subjective factors often play a vital role in determining a pattern
of isopleth lines, as demonstrated in Figure 2.
-------
30
Adjusted Concentration Field
Simulated Concentration Field
Error Field
Figure 6, Adjustment of a Simulated Concentration Field According
to Differences Between the Observed and Predicted Concentrations,
-------
31
To draw an isopleth map by computer, all the data points and data values
must be numerically specified in either computer cards or magnetic data tapes.
In addition, isopleth levels and boundary of a study area must be numerically
specified. This preparatory work requires substantial time and effort.
Therefore, it may not be practical to use a computer for drawing a single
isopleth map. Computer mapping is the most advantageous for drawing many
isopleth maps repetitiously for similar applications. Here, a study area
does not have to be the same as long as the type of isopleth mapping is
similar.
Two types of computer-drawn maps are discussed in this report: character-
printed maps and line-drawn maps. Character-printed maps are produced by a
standard computer printer, which prints typewriter-like characters or,
standard corT!pijt9r-pr''ntOL|t. paper. Line-drawn maps ?,^e produced by eithe1"
a pen plotter or a cathode ray tube (CRT). A brief description of the two
types of computer mapping is presented in the first two subsections. In
the third subsection, a hybrid method incorporating manually drawn and computer-
drawn isopleth maps is explained. Computer software for these applications
is presented in the User's Manual of this guideline document.
2.4.1 Character-Printed Maps
Character~printing--given access to a standard highspeed computer—is
probably the r,;ost cc~;ncn nsans of drav/ing an iscpleth rrap, SYMAP is a
widely used corrpjter progrcr-, for era,/ing a variety of naps [1]. When SYMAP
is used, the basic requirements are to specify the cap boundaries numeri-
cally and to select the type of isopleth map from among three options:
Conformal, Contour and Proximal [1].
-------
32
In each of these maps, the study region is represented as a series of
polygons. The polygons can be used to depict sub-areas such as counties.
These polygons must be numerically specified and are used as basic input
data of the package. Ideographical features can also be incorporated into
the resultant figure.
When the Conformal option is chosen, each polygonal area must be
associated with a data value. Figure 7 is an example of the Conformal
option. In this case, each polygonal area represents a sub-county census
division. Solid lines, showing the boundaries of various counties, are
overlayed over the Conformal SYMAP output. The SYMAP program simply
compared each data value with the ranges of average earnings of male workers
and printed the appropriate character in the polygonal area.
The Contour option is based on a concentration field established by
linear interpolation among existing data points. With this option, the
points at which data values are available must be specified by the X-Y
coordinates. The contour intervals are pre-specified and interpolated
values at intermediate data points are compared with the specified ranges
of the variable. Then, the computer prints the appropriate symbol at
each data point. The print positions at boundaries of one contour level
and another can be left blank so that the shaded region of one contour
level is clearly separated from the other shaded regions by a blank space.
Due to the use of linear interpolation, spurious results can occur in areas
of sharp pollutant gradients among existing data points.
Figure 8 is an example of the Contour option. In-the figure, the
geographical boundaries of the United States are featured by including Mexico
-------
33
A
DISTORTION IS CAUSED BY COMPUTER SYSTEM LIMITATION,
RESULTING IN APPROXIMATE SCALES OF 1 IN. = 19Mi,(VERTICALt
AND ? IN. - 24 ML(HOBIZONTAL).
TRI-8TATE REGIONAL PLANNING COMMISSION
County Boundaries Overlayed on
Conformal SYMAP Output
1970 AVERAGE EARNINGS OF MALE
:" Less than $ 8000
H* $ 8000 to $ 8999
ooo $ 9000 to $10499
000
|«» $10500 to $12999
$13000 or More
Figure 7. Example of Conformal Option, Showing Spatial
Distribution of 1970 Average Earnings of Male
Workers in the New York-New Jersey-Connect!cut
Tri-State Region [12],
-------
0 OOOOOOOP 00
OPOOOOO OOCO
ooooo oooeeee o
t'lOOO
nooo
ooo e
ooo
4- oooo
ocoo e
ooooo &
OOOOO 96866866666
t »00
,100 9966899699 KHf
OOOO 99998699666 Mil
-OOOO ., „_,-, _„-,-
OOOO
4- OOOO 8686666696 •••VCIIIIIIIM 6668666666866 0000
OOOO 66696699069 •fllllllllli 66884689668868 000'
oooooo eees&eeaeee ••••••§••* aeeeeweeeaaee ooooo
OOOOO 869969866998 •&••••• 8966969ft9*68966 OOOOO
+ OOOftO 609966^66669 6686666666664666 HQQOO
noOGOQOO 66660666688886866866986898668^696 OOOOO
OOOOOOOO 99609666^0666869666666666966666 OOOOO
ooooooooo eeef0e6eeee0eie&ee9909e9eeee9e oooooo
OOOOOOO 09868866896869996966916969 OOOOOOO
OOOOOOO.
OOOOO .
66960 OOPO
feepeo roo
ooo
000 K.
ooooo erteeaeeaeueewee ooo 4 «
ooooooo eee«e66ee**ee OPO 4»JKN
ooroooco eeeeaetfee oooo 2.
oooooooooooo e oooo
00000000000,10000 ooooo
.oooooooonoooooorooooo
BeaebAda&eeaoefii>eoe Ov>'.""oo UJf*
seeoeeye'jee^tibt'deoeH oo^oo i-li ft
6Si83ftl:J60Ortfl63H08Ht*f TOJO1" * t { 4 r (
(iB*»8ei*-*O»it'titvi*es!(iiJflt,*i CU.J -) i ^ ( J , 1
$ •*•* fleeetoaetfj^ ooi : 4-ftl;f-
eeeeeeeo^
66046096 6P
SO000 69666666666669086868696 OOCOOCOO
003000 6866£666666669866 OCOOO003O
OOOOOO 666666666 OOOOCOOOOOOOO
OOOOOOOO OOOOOOOOOOOOCOOOOO
t ooocoooooooooooooooooooooooccooo
oooooooocoooo
ooooooco
oooooo
ooooooooc
ooooooooo
ooooocoo
ooooocon
ooooooooo
0 10OOOOOO
OQOOOOOOO ,
oooooooooo
ooocoooooo
oooooon
ooao-joo
OOtlOPO
0
__,...} ^ 2 + 3 -t U + 5 4 6 + 7 4 8
"oorjocnr
ro .coooar
ooooooooo
r?t<-}, j 15
*Bsoi!)^r fiice HIKCU
c «*7* SO'IPCh: >PCO, fcaT'I- NtTWC^K, F. TADPfi AND B. 1CCOPHICK
1-XJtO Df SIGHl •'•S'j RY K» «"*TflCKTTT PY », HIP CAT* HXTVSAS PY H.
MAX.
PPM(SOj)
Figure 8. Example of Contour Option, Showing Spatial Variation of 504 =
Washout Concentration Over the Contiguous United States M3l
-------
35
(designated by M), the Nantucket Islands (designated by N), and Cape Hatteras
(designated by H). The spatial variation of sulfate washout concentration
over the United States is illustrated with the use of five different
characters, each corresponding to a certain range of sulfate concentration.
In the Proximal option, the specifications of map boundaries, data
points, and grid-point interval are made in the sar
-------
36
Line-drawn isopleth maps can be generated by a number of plotting
devices. These devices can be separated into two basic categories, Cathode
Ray Tubes (CRTs) and Pen Plotters. CRTs are usually equipped with a photo-
graphic device so that the desired picture can also be printed on film or
photo-sensitive paper.
Whereas a character-printed map is obtained by using a simple proce-
dural language described in the user's manual of SYMAP, a line-drawn map
usually requires special programming. To generate a character-printed map,
a computer simply calculates a variable value at every grid point, using a
certain interpolation formula (e.g., linear interpolation), and prints a
particular symbol according to the range in which the computed value falls.
In contrast, to obtain a line-drawn map, the computer must command the
plotting device to draw a contour line connecting the points at which the
computed values are equal to the isopleth value.
Data points can be connected by a number of methods. From among the
many methods, we recommend using either a piecewise linear fit, which is
used in the computer program called TRICON listed in the User's Manual, or
a splines-under- tension method, computer program for which is obtainable from
the University of Colorado [17]. Under no circumstance should polynomial
fit be used for drawing isopleth lines. Polynomial fit exhibits very poor
performance with undesirable oscillations.
Figure 9 shows two weather maps produced by a CRT. In this figure,
the first contour map was drawn using a piecewise linear fit; the second
-------
37
a. Piecewise linear fit
b, Splines under tension
Figure 9. Contour Maps of Barometric Pressure Drawn by a
Piecewise Linear Fit and by Splines under Tension p8].
-------
38
contour map was drawn using splines under tension [18]. The former map
was probably drawn from regularly spaced data points of barometric pressure
by either writing a small computer program or simply using a procedural
language of some preprogrammed package for computer graphics. Although the
latter contour map has more aesthetic appeal than the former, it requires
a large program and considerably longer computing time.
In the above examples, the data values were given at regularly spaced
data points. The mechanism for determining an isopleth line for regularly
spaced data points is illustrated in Figure 10a. For every edge whose two
vertices have a data value greater than or equal to the isopleth level at
one end and less than or equal to the isopleth level at the other, the
isopleth point is determined by linearly interpolating the data values at
the two vertices to that isopleth level. The isopleth line is then drawn
by connecting all those isopleth points.
Air monitoring data (and population exposure variables to be discussed
later) are available only at scattered data points, i.e., monitoring station
sites and/or receptor points. A method of obtaining an isopleth map from
such randomly spaced data points involves triangulation of data points as
shown in Figure lOb. The triangulation can be performed in many ways. How-
ever, it is recommended that the outlying data points be connected first by straight
lines to form a polygon, and the triangulation proceed inward. In forming
a polygon by connecting the outlying data points, it is, to a certain extent,
possible to make the polygonal area resemble the study region.
Once the triangulation is completed for randomly spaced data points,
the steps for drawing an isopleth map are similar to those used for regularly
-------
26 25 24 23
14
V
21 I 23
18
16
_J 1
15 18
a. Regularly spaced data points
24
24
b. Randomly spaced data points
co
Figure 10. Concentration Isopleths Resulting from Regularly and
Randomly Spaced Data Points.
-------
40
spaced data points. An isopleth point is determined on each edge whose two
vertices have a data value greater than or equal to the isopleth level at
one end and less than or equal to the isopleth level at the other end.
Then, the isopleth line is drawn by connecting all those isopleth points
(Figure 10b). A computer program doing the above tasks has been developed
by Martin Cohen at Technology Service Corporation. Listings of the computer
program called TRICON are given in the User's Manual.
Figure 11 shows an example of a computer-drawn isopleth mao for randomly
spaced data points. In the figure, the boundary of the study area is drawn
separately from the isopleth lines. The map boundary was drawn for the
input data produced by a digitizer whose pencil-like probe traced the
boundary of the original geographic map, starting from the bottom (note
the small gap there). When such a convenient digitizing device is not
available, the boundary can be numerically specified by reading off the
X-Y coordinates of major vertices on the boundary line from an overlayed
graph paper. A simpler approach, as mentioned before, is that of overlaying
a transparent map of the study area on the computer-drawn isopleth lines.
The isopleth lines obtained from the randomly spaced data points cover
the study area quite well. When data points sufficiently cover a study
area, as seen in Figure 11, data values need not be computed for every
regularly spaced data point; an isopleth map obtained from triangulation
of randomly spaced data points is sufficient for most cases.
2.4.3 Hybrid Isopleth Map
Even when one cannot trust a computer-drawn isopleth map or computer
software to draw an isopleth map is not readily available, a computer
plotting device still can be a useful tool to draw an isopleth map. For
-------
Figure 11- Computer-Drawn Isopleth Map, Showing Oxidant Air Quality
over the Los Angeles Region in Percentage of Days on
which the NAAQS was Exceeded During 1973/74 [3].
-------
42
instance, a pen plotter will draw a specific symbol at an exact data point
(monitoring station and/or receptor grid point) according to a range of
data values. Then, one can easily draw an isopleth line by following the
same symbols and procedures described in Section 2.2.
Figure 12 shows an example of a hybrid isopleth map. Three data sets
are required:
(1) A set of annual air quality statistics (here, percentage of days the
Ox NAAQS was exceeded during 1973/74 period),
(2) X-Y coordinates of non-uniformaly spaced data points (here, receptor
grid points at which the air quality was estimated from air monitoring
data), and
(3) A set of X-Y coordinates describing polygonal areas that represent
the study region.
The computer sorts each data value into an appropriate range of data values
and commands the pen plotter to draw the corresponding symbol at the data
point. Note that Figure 12 is drawn for the same data used to produce
Figure 11.
With this hybrid approach, one can save some time by using a computer
and at the same time can manually draw an isopleth map that the person
perceives to be the most likely. The listing of the computer program to
produce the hybrid map is given in the User's Manual.
-------
Symbol
9
8
7
6
5
4
3
% of days
above Ox NAAQS
50-59
40-49
30-39
20-29
15-20
10-14
5-9
-p*
co
Figure 12. Hybrid Isopleth Map, Showing Oxidant Air Quality over the Los Angeles
Region in Percentage of Days on Which the NAAQS was Exceeded During
1973/74
-------
44
CHAPTER 3. POPULATION EXPOSURE ANALYSIS
One can extend a spatial analysis of air quality information a step
further by combining air quality data and demographic data. Ambient air
quality can be quantified by the percentage of the population living in
areas above an air quality standard instead of concentration units.
A population exposure analysis can report ambient air quality by a
statement such as, "The percentage of the urban population living in areas
above the primary national ambient air quality standard for particulates
decreased from 60% in 1970 to 10% in 1975." This statement may be more
informative to the public than telling them that the annual-mean-particulate
3 3
concentrations decreased from 69.5 yg/m to 64.3 ug/m in the same period.
The following three different methods have been used to quantify
population exposures from air monitoring information:
1. Associating populations to a monitoring site [5,6,7,14],
2. Transcribing air pollution isopleth patterns on a map of census
tracts [8,9].
3. Interfacing air quality and population data at strategically located
receptor grid points [2,3,14],
This guideline document describes the third method that provides the most
objective estimates of population exposure and lends itself to computer
processing.
The proposed population exposure analysis requires several steps. The
first step is a search for a proper regional map and population data in a
convenient form. The second step is development of a receptor network of
artificial grid points that are used to approximate the spatially distributed
-------
45
population. The study area boundary is determined by the methods of
Chapter 2. The location of the receptor points and the monitoring stations
must be specified in a digital form, based on the coordinate system of the
regional map. The third step involves the preparation of computer-ready
data sets and computation of population exposure variables by computer
software. The fourth step included analysis of the computed population
exposure variables and presentation of the results.
The details of each step are discussed in the sections which follow.
3.1 DEVELOPMENT OF RECEPTOR NETWORK
Instead of employing the assumption that the population exposure level
in various parts of a region is represented by the concentration reading at
a single air monitoring station, we propose the use of a receptor network of
artificial grid points. The development of this network is based on the
spatial distribution of the population and is used to interface the population
data and the air quality data at each receptor point. A receptor point is
used to represent the local population in the areas where they reside. The
location and density of receptor points are selected so that the distribution
of receptor points reflects the spatial distribution of population density
and adequately covers the land area throughout the study region. Air quality
is estimated at each receptor point by spatial interpolation of air quality
observed at neighboring monitoring sites.
To develop a receptor network, one has to have population data and a
regional map. Population data are always obtainable from census population
summaries. However, because it takes time and effort to reduce these data
into a form that is practical for population exposure analysis, it may be
-------
46
worthwhile to contact a regional planning agency. The regional planning
agency often has not only conveniently aggregated population data into
properly aggregated statistical areas, but has also prepared a regional map
showing boundaries of the statistical areas, cities, and counties.
When such a convenient source of information is not available, census
population data must be used. Although every census summary contains maps
showing tracts and statistical areas, a regional map covering the entire
study area, say for an AQCR, may not be found in the census summary. There-
fore, it is necessary either to look for other map sources (such as street
and geographical maps) or to create a new regional map by combining several
maps found in the census summaries.
The census population summarires range from detailed demographic data
for each census tract to aggregated population data for each state and each
standard metropolitan statistical area (SMSA). For a single city or small
SMSA, the census tracts provide a basis for establishing a receptor network.
The center of each census tract will serve as a receptor point. For a
larger study area covering many cities and/or counties, the number of census
tracts included in the study area may become too large. For the larger areas,
the most convenient census statistics may be the population summaries for
subcounties [19]. These summaries provide the population size of statistical
areas that often correspond to administrative jurisdictions such as cities,
towns, villages, or boroughs.
When a planning agency's statistical areas, census subcountry or census
tract population statistics are found to be suitable for the analysis, re-
ceptor points are assigned to each "statistical" area according to the size
-------
47
of the population and the land area. Assuming that the population is
uniformly distributed within each statistical area, the location of re-
ceptors (in most cases just one receptor) is determined such that the
receptor points represent properly the spatial location and boundary of
that statistical area. Empirically, we have found that the following
method yields the proper number of receptor points to each statistical area:
1, Regardless of the size of the population and the land area, each
statistical area is represented by at least one receptor point.
2
2. An additional receptor point is assigned for each 50 mi increment
land area or each 200,000-person increment of resident population.
3. Take the number of receptor points representing the larger increment
(land or population).
A question arises here as to why the monitoring stations themselves are
not included as receptor points in computing population exposure variables.
The following two paragraphs provide some explanation.
Because the spatial interpolation scheme used in the population exposure
methodology smoothes out the observed air qualities at the three nearest
neighboring stations to each receptor point, the highest and lowest observed
air qualities among the monitoring stations sometimes do not appear on the
concentrations at the receptor points. Therefore, a population exposure
analysis may indicate that the study population is exposed to concentrations
either slightly above the lowest observed concentration or slightly below
the highest observed concentration.
-------
48
The problem of missing the highest and lowest concentrations, however,
is already inherent in the limited number of monitoring sites. Although
the spatial interpolation scheme used may further smooth out those highest
and lowest concentrations, the conservative nature of population exposure
methods should be a merit rather than a demerit. The population exposure
analysis is less affected by the extreme values observed than is the air
quality analysis that is applied directly to the air monitoring data. The
smoothing effect of a spatial interpolation of air monitoring data is not
necessarily undesirable for estimating population exposure because it is
partially equivalent to people moving around and smoothing out their exposure
levels.
When the receptor network is developed, both the receptor points and
the monitoring stations have to be located on the same digitized regional
map, e.g., a map on graph paper. To correctly locate a monitoring station
on the map is not as easy as one might think; the regional map used may not
contain any landmarks helpful in locating the station at the right place.
The larger the study region, the more difficult the placement of the stations
on the regional map. For a large, ummarked regional map, one's prior know-
ledge of the station sites does not work effectively. A mathematical method
of correctly locating monitoring stations on a skeleton map, discussed in
Appendix,A.should be used for such a case.
-------
49
3.2 PREPARATION OF DATA SETS
The population exposure analysis requires careful preparation of several
data sets. These include data on the resident population, receptor points,
air quality, and monitoring stations. The data sets of air quality and
monitoring stations similar to those required for performing a spatial analysis
of air quality. Air quality data used for a population exposure analysis are
commonly available statistical summaries: annual mean concentrations for
analyzing long-term average population exposure, and percentage of the time
above the air quality standard or percentile concentrations for analyzing
short-term population exposure.
Population and receptor data sets to be prepared for performing a
population exposure analysis are described in the following two subsections.
These data sets, together with those of air quality and monitoring stations,
are used as input data to one of the two computer software systems (LPEM and
SPEM) described in the User's Manual.
3.2.1 Population Data
Sources of Population Data
National census data are the primary source for almost all demographic
data. Detailed demographic data are available in both magnetic data tapes
and published reports for some 35,000 census tracts in 241 Standard Metropolitan
Statistical Areas (SMSAs) and the remaining unincorporated census statistical
areas. However, it is rather difficult to transform these census data into
a form useful for a population exposure analysis. Unless an in-house capa-
bility of handling the census data tapes exists, a search for pre-existing
sources of conveniently aggregated population data is recommended.
-------
50
Census statistical summaries issued from the Bureau of Census and
local planning agencies are the two major sources of such aggregated popu-
lation data. Among the census statistical summaries, the following two
series provide useful population data for performing a population exposure
analysis: PC(1)-A series [19], which summarizes the number of inhabitants
for political jurisdictions (boroughs, towns, cities, counties, etc.), and
PHC(l) series [20J, which provides data on population characteristics
(e.g., sub-populations by age group, nativity, place of work, worker and
nonworker status) for each census tract. These two series also contain
maps showing boundaries of political jurisdictions arid/or census tracts.
Some of these maps may be useful for a population exposure study.
Local planning,agencies that may have useful population data include
state highway departments, regional planning agencies, state housing
administrations, and civil defense agencies. These agencies may have not
only conveniently aggregated population data for regional statistical areas
but also a high quality map showing boundaries of counties and regional
statistical areas.
Sub-Population Data
Population exposure methodology discussed in this text assumes that
people are locationally fixed at their residence location. This simplifying
assumption may not be valid for working-age population, but may be good
for sub-populations such as school-age and elderly who tend to stay near
their residence location most of the time. Furthermore, these sub-populations
are believed to be the most susceptible to air pollution. Therefore, population
-------
51
exposure analysis should be performed not only for total population but
also for sub-populations including school-age children and elderly people.
Because all statistics of the sub-populations mentioned above are given
for place of residenee, the data for sub-populations should be converted to
a percentage of the total resident population in each aggregated statistical
area. Then, we can perform a population exposure analysis for various sub-
populations by using the same receptor network developed for the total
resident population.
For working-age population, its large diurnal mobility may invalidate
the stationarity assumption employed in the population exposure methodology.
However, a way exists to incorporate population mobility into the current
population exposure methodology. It can be assumed that population exposure
during working time occurs at work place, while population exposure during
non-working time,occurs at residence location[3]. To compute exposure of
the workers during working time requires employment statistics that
provide the number of workers at their place of employment. This type of
sub-population data should be given by the actual size of the sub-population
in each aggregated statistical area, instead of the percentage of the total
population.
Population Data for Off-Survey Years
When the analysis is made for many years, the population statistics should
be computed for each year by interpolating known population statistics or
population projections for other years to that year. For example, the size
of resident population in New Y ork County (Manhattan borough) in 1965 is
estimated as 1,618,757 by linearly interpolating the 1960 population, 1,698,281,
-------
52
and the 1970 population, 1,539,233, to the year 1965. The 1973 population
may be approximated by the 1970 census population or may be estimated by
interpolating the 1970 population and the population projected to 1980.
When the study period is only for a few years or when the spatial
distribution of population over the study period remains almost the seme,
the population data of a single survey year can be used for population
exposure analyses over the entire study period.
3.2.2 Receptor Point Data
In a population exposure analysis, a receptor point is used for several
purposes. A receptor point represents both the size, the spatial location,
and the spatial spread of the local population, The spatial location of the
local populace is given by the x-y coordinates of the receptor point in the
digitized regional map. Specifications for the formats are given in the
User's Manual.
Air quality data measured at air monitoring stations are spatially inter-
polated to each receptor point and are merged with the population data at the
receptor points to compute various population exposure variables. Thus, the
receptor points can also be used to develop an isopleth map of air quality
and population exposure variables from the values computed at each receptor
point. For example, the isopleth map of long-term exposure to annual mean
concentrations over the study region is developed from the values computed
at each of the receptor points. Guidelines for producing isopleth maps
of air quality and population exposure variables are given in Chapter 2.
-------
53
3.3 CHARACTERIZATION OF POPULATION EXPOSURE TO AIR POLLUTION
Pollutant concentrations over an urban area are constantly changing in
time and space. Exposure of people to these concentrations can be analyzed
with respect to both long-term average exposure and statistical distribution
of short-term exposures over a long period (usually a year). To make the
analysis tractable, the resident population (the type of data available
from the census summaries) is used, and individual members of the population
are assumed to be locationally fixed at their residence location.
3.3.1 Analysis of Long-Term Population Exposure
An annual mean concentration is used to designate the long-term average
pollution level to which people are exposed. The long-term average pollution
level at each receptor point is estimated by spatially interpolating the
annual mean concentrations observed at the three nearest neighboring stations
to that receptor point by using Eq. (2). For a given receptor point,
{x-,y-). the computer program call ad Long-Term Population Exposure Model
(LPEM) automatically searches among all the stations used for the
analysis for the three nearest stations, (x.,y.}» j - 1.2,3, to that receptor
J J
point (see User's Manual).
Once the annual mean concentration, C^, is determined for each receptor
point, population exposure can be calculated by combining the air quality
and population statistics. The LPEM compute^ program compares C- with a
*
given concentration threshold, C ; enters the local population, P.; and
determines population exposure for the entire region, R. Mathematically,
this can be expressed by
-------
54
P(C*) - £ P, U(C,-C*)/E ?< » (3)
ieR ieR 1
* *
v/here an indicator step functiontU(C.j-C ), takes the value one for C^>C
* *
and zero for C-
-------
55
O)
-------
1.0
0.8
-------
57
longer black bars indicating the percentage of the population exposed to
annual mean N02 above 130 yg/m . The decrease in population exposure in
1973/74 is noted by the shorter black bar and the longer gray bar, which
indicates the percentage of the population exposed to N02 between 100 and
130 yg/m3.
The LPEM computer program also calculates three regional indices for
characterizing long-term air quality and population exposure: station average
concentration, space average concentration, and population average concen-
tration. The station average concentration is given by an arithmetic mean
of annual mean concentrations observed at individual stations. The space
average concentration is given by an area weighted average of the concen-
trations at the receptor points. The population average concentration is
given by a population weighted average of the receptor concentrations and
indicates the air quality level most representative of the population.
Mathematically, the space average concentration^ At) . and the population
o
average concentration, AQ , are expressed respectively, by
AQ. = £ S.C./ £ S. (4)
s n n icR n
AQp = E P. C./ £ P (5)
7 1 ieR 7
where C.. is the annual mean concentration at the i receptor point,
and Si and P.. are, respectively, the land area and the local population
size represented by the i receptor point.
-------
58
3.3.2 _ Analysis of Short-Term Population Exposure
Analysis of population exposure to short-term peak concentrations is
more difficult than the analysis for long-term average concentrations. The
reason is that long-term average exposure involves only one pollution variable,
i.e., concentration level, but short-term exposure involves at least two pol-
lution variables, i.e., concentration level and frequency of occurrence or
duration of that concentration level.
It is extremely costly to compute each instantaneous exposure level In-
dividually over a long time period. Instead of using each instantaneous con-
centration value, a percentage of the time that the short-term (1-hr or 24-hr)
NAAO^ was exceeded is used to characterize short-term exposures of the popula-
tion over a long period. The quantity, "percentage of the time above the
standard,'' is termed a "risk frequency" because it indicates how frequently
^••.oplf: are exposed to a lovel of air pollution above the standard.
I he risk frequency can be computed either from the number of standard
violations or from the percent!le concentrations. When a number of standard
v;- lotions are .jvailable, the risk frequency at each monitoring station is
ustd a-, input fldt.fi t.o the I PI M computer proqram. Then, the LPHM spatially
interpo loter, thf- risk frequencies observed concentra-
i. i'jris ai f-i'jht so let, ted perceiii. i les ,ire usetl
-------
59
spatially interpolates the percentile concentrations at the monitoring sta-
tions to those at receptor points. Then, the SPEM compares each of the
eight interpolated percentile concentrations with the air quality standard
to determine the risk frequency at each receptor point (see User's Manual).
The risk frequency computed at each receptor point is used to draw an
isopleth map of risk frequencies over the study region. The risk frequency
map is an excellent aid in visualizing the spatial variation of short-term
air quality levels.
When the risk frequencies are determined for all the receptor points,
short-term exposure levels can be calculated for the entire study region.
Both computer programs (LPEM and SPEM) compare the risk frequency, F., at
each receptor point with a given frequency threshold, F*; enter the local
population, P.; and scan the entire study region, R, Mathematically, this
process can be expressed by
P(F*) - E P.UCF.-F*)/ Y P,- , (6)
i R ] ifR
where an indicator step function L:(F.~F*} takes the value one for F^>F*
and zero for F..
-------
60
illustrates how much the population exposure decreased (or increased) in
a given time period, how the decrease (or increase) in population exposure
was distributed over the population, and how the frequency of dangerous
exposures changed. For instance, Fig. 15 shows the percentage of the
population exposed to total suspended particulate above the national
secondary 24-hr standard (150 yg/m ) at least 10$ of the days dropped
from 33% in 1971 to 4% in 1974.
The bar graph shown in Fig. 16 is convenient for illustrating year-
to-year variations in population exposure over a long period. The increase
in population exposure during the middle years, 1967-72, is represented by
the rhorter white bars, indicating the percentage of the population ex-
posed to hourly NOo above the California standard less than 6% of the
days. The decrease in population exposure in 1973/74 is noted by the
longer white bar and the shorter gray bar, indicating those who were
exposed above the California standard between 6% and 12% of the days.
Both computer programs compute two regional indices for character-
izing short-term air quality and population exposure: the percentage of
time (hours or days) exceeding the standard, area weighted average over
the study region; and the percentage of time exceeding the standard, popu-
lation weighted average over the region. The latter quantity is the
average percentage of time the standard is exceeded among all the people
in the study region. Mathematically, the two regional indices, F$ and F ,
are expressed, respectively, by
Fs = £ S. F / £ S. (7)
* ieR n ieR
F _
FP "
E P, F. / L P.. (8)
ieR n 1 ifR n
-------
61
c
QJ
£
o
T3
O)
(/)
O
OL
X
c
o ,
•£> >,
03 O
r— C
3 O)
CL 3
o cr
O- O)
it- Li_
O
-o
-------
U3
VO
ID
1C
CO
C\J
r--
100
90
80
70
2:
o
i — i
1 60
O.
O
Q-
-i 50
•3.
O
U- 40
o
«—
z
UJ ~n
<_> ou
C£
yj
o.
20
10
0
MM*
, ',•.*.•,' " * ]
••<<•: ilis
• :•••:•:•:•:: !]
,:x:>fv:::->M*«
1 X
•
Bffl-::';;:r::
i i :ii:a £ i
CA. standar
d
P IS I OF DAYS STANDARD EXCEEDED
p < 6
'.:'. 6 £ p < 1 2
01
ro
Figure 16. Changes in Population Exposure to NC^ During Five 2-Year
Periods in the Los Angeles Region [3].
-------
63
REFERENCES
1. Harvard Laboratory for Computer Graphics, "User's Reference Manual
for Synagraphic Computer Mapping 'SYMAP1 Version V," Harvard
University, Cambridge, MA, 1968.
2. Horie, Y., and A. C. Stern, "Analysis of Population Exposure to
Air Pollution in New York-New Jersey-Connecticut Tri-State Region,"
U.S. EPA/OAQPS Publication No. EPA-450/3-76-027, University of
North Carolina, Chapel Hill, NC, March 1976.
3. Horie,Y., ejt a_]_. , "Population Exposure to Oxidants and Nitrogen
Dioxide in Los Angeles: Volumes I, II, and III," U.S. EPA/OAQPS
Publication No. EPA-450/3-77-004a, b, and c, Technology Service
Corporation, SantG Monica, CA, January 1977.
4. U.S. Erivironmenta i Protection Agency, "National Air Quality and
Emissions Trends Report, 1975," U.S. EPA/OAQPS Publication No.
EPA-450/1-76-002, Monitoring and Data Analysis Division, Office
of Air Quality Planning and Standards, Research Triangle Park,
NC, November 1976.
5. Zupan, J.M., The Distribution of Air Quality in the New York
Region, Resources of the Future, Inc., Washington, DC, 1973.
6. McHarg, I. L., Design with Nature, Published for the American Museum
of Natural History, Doubleday/Natural History Press, Garden City,
NY, 1971.
7. Anderson, J. A. et al., "Correlation between Air Pollution and
Socio Economic Factors in Los Angeles County," submitted to
Urban Ecology, Department of Chemistry and Energy Center, University
of California, San Diego, CA, January 1977.
8. Brian, J.L., Berry, et al. The Social Burdens of Environmental
Pollution: A Comparative Metropolitan Data Source, Department of
Geography, University of Chicago, 1976.
9. Istvan,Takacs and G. Bradford Shea. Estimations of Human Population-
at-Risk to Existing Levels of Air Quality, EPA Contract No. 68-01-2820,
Enviro Control, Inc., Rockville, Maryland, February 1975.
10. TRW Systems Group, "Air Quality Display Model," prepared for Department
of Health, Education and Welfare, National Air Pollution Control
Administration, Washington, DC, Contract No. Ph-22-68-60, and
Available from NTIS, Springfield, VA, 22151 as PB-189-194, 1969.
-------
64
11. Busse, A.D., and J. R. Zimmerman, "User's Guide for the Climatological
Dispersion Model," U.S. EPA Publication No. EPA-R4-73-024, National
Environmental Research Center, U.S. EPA, Research Triangle Park, NC,
December 1973.
12. Ludwig, F. L., W. B. Johnson, A. E. Moon, and R. L. Mancuso,
"A Practical Multipurpose Diffusion Model for Carbon Monoxide,"
Stanford Research Institute, Menlo Park, CA, Contracts CAPA-3-68
and CPA 22-69-64, 1970.
13. Zimmerman, J. R., and R. S. Thompson, "User's Guide for HIWAY, A
Highway Air Pollution Model," U.S. EPA Publication No. EPA-650/
4-74-008, National Environmental Research Center, U.S. EPA,
Research Triangle Park, NC, February 1975.
'.4. Fran!,, N. H., W. F. nunt, ;-., and W. H. Cox, "Population Exposure:
An Indicator of Air Quality Improvement," Paper #77-44.2 presented
at the 70th Annual Meeting of the Air Pollution Control Association,
Toronto, Canada, U.S. Environmental Protection Agency, Research
Triangle Park, NC, June 1977.
•3- Tri-State Regional Planning Commission, Re g iona1 Profi1e--Regiona1
Employment 1970, Vol. II, No.6, New York, NY, December 1973.
16. Wolaver, T. G., "The Dsstribution of Natural and Anthropogenic
Elements and Compounds in Precipitation Across the U.S.: Theory
and Quantitative Models," Botany Department, University of
North Carolina, Chapel Hill, NC, October 1972. Copies available
through Division of Ecological Research, U.S. EPA/NERC, Research
Triangle Park, NC.
17. Cline, A. K., "Curve Fitting Using Splines Under Tension,"
Atmospheric Technology, National Center for Atmospheric Research,
No. 3, pp 60-65, September 1973.
18. Wright, T. J., "Utility Plotting Programs," Atmospheric Techno1ogy,
National Center for Atmospheric Research, No. 3, pp 51-57, September
1973.
19. Bureau of the Census, "Number of Inhabitants," PC(1)-A Series,
U.S. Department of Commerce, Washington, DC, May 1972.
20. Bureau of the Census, "Census Tracts," PHC(l) Series, li.S. Department
of Commerce, Washington, DC, May 1972.
-------
65
APPENDIX A. USE OF COORDINATE TRANSFORMATION FOR
LOCATING MONITORING STATIONS ON A SKELTON MAP
When the study region is large, a mathematical method of locating
monitoring stations on a skeleton map of the study region often works
more effectively and more accurately than heuristic methods, such as
those based on the street address of station site or on knowledge •/ •=•
relative location of one station to another. The mathematical method
utilizes the UTM coordinates given in the SAROAD format and applies a co-
ordinate transformation on the UTM coordinates of an individual station.
When the mathematical method is used, the first step is to locate
the two most familiar stations on the skeleton map. The location of
other stations on the map is then determined by a coordinate transfor-
mation of their UTM coordinates into the coordinate^ used in the map.
The map coordinates (x, y) of a station whose UTM coordinates are (p, q)
are given by the following equations:
x = m(p - pj) + x]
y = m(q - q,) + y,
and the slope, m, is given by
m = - , (A-2)
(P1-P2)y2 - (q1-q2)x2
where (x,,y,) and (x2,y2) are the map coordinates of the two known stations
and (p,,q,) and (p2,q2) are their UTM coordinates.
-------
ENVIRONMENTAL PROTECTION AGENCY
Technical Publications Branch
Office of Administration
Research Triangle Park, North Carolina 27711
OFFICIAL BUSINESS
AN EQUAL OPPORTUNITY EMPLOYER
POSTAGE AND FEES PAID
ENVIRONMENTAL PROTECTION AGENCY
EPA - 335
SPECIAL FOURTH-CLASR RATE
BOOK
Return this sheet if you do NOT wish to receive this material
or if change of address is needed [ ]. (Indicate change, including
ZIP code.)
PUBLICATION NO. EPA-450/2-77-024a
(OAQPS NO. 1.2-083)
------- |