United States
Environmental Protection
Agency
Research and Development
Environmental Monitoring Systems^
Laboratory
Las Vegas NV89114
EPA-600/S4-81-002 Apr.
Project Summary
Methodology for the
Design of an Optimum Air
Quality Monitoring Network
Mei-Kao Liu and Joel Avrin
A two-step objective method is
presented for determining the optimum
number and disposition of ambient air
quality stations in a monitoring net-
work. The method uses a data base
consisting of a comprehensive set of
simulated or measured air quality
patterns representative of the region
of interest. In the first step, the most
desirable monitoring locations are
identified and ranked. The minimum
number of required locations is deter-
mined in the second step through
eliminating redundancies among the
locations identified in the first step
with regard to spatial coverage over
the region of interest. Asa demonstra-
tion, the method is applied to the Las
Vegas Valley of Nevada for the pollu-
tant species carbon monoxide.
This Project Summary was devel-
oped by EPA's Environmental Moni-
toring Systems Laboratory, Las Vegas,
NV. to announce key findings of the
research project that is fully docu-
mented in a separate report of the
same title (see Project Report ordering
information at back).
Introduction
The Clean Air Act requires state and
local agencies to monitor ambient air
quality, primarily for documenting an
area's compliance with the National
Ambient Air Quality Standards. Addi-
tional monitoring may be needed to
satisfy secondary objectives such as
providing background or baseline con-
centrations. Currently, the determina-
tion of the number and location of
monitoring stations required in a net-
work is primarily based on subjective
considerations, semiquantitative rules
supported by experience, or limited use
of analytical tools such as simple Gaussian
models. Nontechnical considerations,
such as convenience and accessibility,
are usually the dominant factors in
selecting a specific monitoring location
within the area of interest. On the other
hand, because of the fluctuations in
pollutant emission rates and the turbu-
lent nature of the atmosphere, pollutant
concentration distributions are highly
variable, both in time and space. The
concentrations measured at any given
site depend on the emission patterns as
well as atmospheric conditions. The
design of an optimal monitoring net-
work, therefore, requires an a priori
knowledge of these concentration vari-
abilities.
Method
A two-step objective methodology for
the design of such an air quality moni-
toring network is proposed in this study.
The methodology uses a data base
consisting of a comprehensive set of
measured or simulated air quality pat-
terns representative of the region of
interest. The objective of the first step is
simply to ascertain the most favorable
locations for making air quality mea-
surements. To this goal, a concept
called the figure-of-merit (FOM), intro-
duced in an earlier study, is used to
facilitate the identification and ranking
-------
of potential monitoring sites.1 The
objective of the second step is to deter-
mine the network configuration by
deleting redundancies among the moni-
toring stations identified in the first
step. The spatial correlation coefficient
between pollutant concentrations at the
monitoring site and those at a neigh-
boring point is invoked as the parameter
for estimating the relevance that air
quality measured at one point has on
another point. By using the sphere of
influence, as determined by a minimum
or cutoff value in the correlation coeffi-
cient distributions, the monitoring
network, containing a minimum number
of monitoring stations, can be obtained.
The figure-of-merit is defined as the
sum over a comprehensive set of condi-
tions of the products of an air quality
index, either observed or predicted, and
the associated probability of occurrence.
Herein, the summation is performed
over all meteorological scenarios poten-
tially leading to high air pollutant con-
centrations. The FOM contains weight-
ing by the probabilities of occurrence of
scenarios to avoid situations related to
extremely rare events or periods. These
situations would not necessarily pro-
vide the best criteria for determining a
permanent or semipermanent site for a
monitoring network. In general, the air
quality index can be composite of several
pollutant concentrations, weighted by
the relative importance of the individual
species, if it is desirable to design a
multiple pollutant species monitoring
network. For simplicity, the air quality
index used here was the concentration
of a single pollutant species. Thus, for
any given location (x, y),
M
FOM(x, y) = I
k=1
/concentration at location \
I (x, y) under meteorological] .
\pattern k /
/probability of \
(meteorologicall
\pattern k j
(D
Although the concentration fields for
Equation (1) can be either observed or
predicted, the practice of using predicted
concentration fields was adopted for
this study; few, if any, regions have a
monitoring network in operation over a
sufficient time interval and of sufficient
density to yield the requisite concen-
tration distributions.
As is the case herein, the concentra-
tion distributions can be developed
using an air quality simulation model for
meteorological scenarios specified for
the region of interest. These distribu-
tions, when combined with the corre-
sponding probabilities of occurrence,
permit the evaluation of FOM. A selec-
tion of the most favorable air quality
monitoring sites can then be accom-
plished by ascertaining the noncontig-
uous peaks in the resultant FOM field
and ranking the locations according to
magnitude of FOM.
The approach for establishing the
minimum number of monitoring sta-
tions from those identified and ranked in
the first step involves the concept of the
sphere of influence (SOI) which for a
given monitoring station is determined
by the statistical properties of the
pollutant concentration distributions
used in the first step. Analogous to the
study of turbulence in an Eulerian
framework, a spatial correlation coeffi-
cient is introduced between the values
of pollutant concentrations at a potential
monitoring site and the corresponding
values at its neighboring points as a
function of radial distance away from
the station:
(2)
Vvar [c(So)]
var [C(SQ + As)]
where the C(SD) and c(s0 + As) can be
measured or predicted concentrations
at the points So and So + As and As =
[(Ax)2 + (Ay)2]172. The symbols cov and
var denote covariance and variance,
respectively.
Statistically, the correlation coeffi-
cient measures the linear association
between c(So) and c(s0 + As), namely, the
concentrations measured at the
monitoring sites and those at its
neighboring points. This coefficient,
lying between -1 and +1, thus by itself
furnishes an ideal dimensionless tool
for determining the SOI. The spatial
correlation coefficient is expected to
initially decrease from one as the
distance increases. Consequently, the
SOI can be delineated on the basis of a
predetermined minimum or "cutoff"
value for the correlation coefficient.
The square of the correlation coeffic-
ient represents the fraction of the
variance of one variable explained by
the other variable.2 One hundred times
this value then can be considered as the
percentage of concentration variations
explained by concentration variationsat
a potential monitoring site. In practice,
values of pollutant concentrations are
generally taken from a finite sample of a
larger population. Thus, after a value for
the minimum desired amount of
variance explained is selected, the
minimum acceptable value of the
population correlation coefficient can
be computed using the above-mentioned
relationship. Procedures and tables
developed by F. N. David can in turn be
used with the minimum acceptable
value for this coefficient to yield a cutoff
value for the sample correlation
coefficient for a specific sample size and
confidence level.3
Once the cutoff value for the sample
correlation coefficient is determined, a
SOI can be subsequently developed for
each of the ranked potential monitoring
sites. The determination of the mini-
mum number of monitoring stations
required can then be carried out by
deleting lower-ranked stations whose
SOI overlap the SOI of higher-ranked
stations and whose SOI provide non-
overlapping coverage of less than some
fixed percentage of the coverage of the
higher-ranked stations.
Application
The siting methodology was applied
as a demonstration to the Las Vegas
Valley of Nevada (see Figure 1) for the
pollutant species carbon monoxide. The
simulated concentration fields were
obtained by exercising the SAI Atmos-
pheric Pollution Simulation Model for
six different meteorological scenarios
for Las Vegas.1 Nineteen stations cover-
ing concentration maxima and three
stations covering background concen-
trations in rural areas were considered.
For example, a 10-station network,
among which 3 stations are for back-
ground concentrations, was selected for
a desired minimum detection capability
of 50 percent of concentration fluctua-
tions (95 percent of the time) and hence
a cutoff sample correlation coefficient of
0.8. The network configuration and joint
areal coverage (shaded areas) are shown
in Figure 2. The fraction of total area
covered by each station and the cumu-
lative areal coverage, beginning with
the highest-ranked station, are alsc
tabulated in this figure. Stations iden-
tified by T, U, or V are the backgrounc
stations.
Networks with fewer stations than ir
the above network would be selected il
smaller minimum detection capabilities
-------
Sheep Range
' NLV Airport /
Flamingo
Tropicana
A, McCarran Airport. ^
Blue Diamond
of concentration fluctuations are deemed
acceptable, and vice versa. Background
stations could, of course, be deleted for
networks with the sole objective of
discerning peak concentrations.
References
1. McElroy, J. L. et al. (1978), "Carbon
Monoxide Monitoring Network
Design Methodology," EPA-600/4-
78-053, Environmental Monitoring
Systems Laboratory, U.S. Environ-
mental Protection Agency, Las
Vegas, Nevada.
2. Ezekiel, M. (1941), Methods of
Correlation Analysis, John Wiley
and Sons, Inc., London, England.
3. David, F. N. (1938), Tables of the
Ordinates and Probability Integral of
the Distribution of the Correlation
Coefficient in Small Samples, The
Biometrika Office, Cambridge Univer-
sity Press, Cambridge, England.
Figure 1. Map of the Las Vegas Valley.
-------
/L/. i i iii r r i i TV i i i i i i r~l
60.
50.
40.
30.-
20.
10.
0.
0.0723
0.0353
0.04S5
0.01SS
0.0122
0.0211
0.0104
0.0182
0.0336
0.0119
0.0723
0.0976
0.1432
0.1586
0.1708
0.1920
0.2024
0.2205
0.2542
0.26(1
0.
Figure 2.
10.
20.
30.
40.
An optimum air quality monitoring network based on a cut-off sample
correlation coefficient of 0.8 — Las Vegas Valley.
-------
Mei-Kao Liu and Joel Avrin are with Systems Applications. Inc., San Rafael. CA
94903.
James L. McElroy is the EPA Project Officer (see below).
The complete report, entitled "Methodology for the Design.of an Optimum Air
Quality Monitoring Network," (Order No. PB 81-171 191; Cost: $8.00. subject
to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Monitoring Systems Laboratory
U. S. Environmental Protection Agency
Las Vegas. NV89114
» U.S GOVERNMENT PRINTING OFFICE. 1961 -757-012/7051
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Postage and
Fees Paid
Environmental
Protection
Agency
EPA 335
Official Business
Penalty for Private Use $300
PS 0000329
U S fcNVIR PROTECTION A
------- |