OCR error (C:\Conversion\JobRoot\00000CUY\tiff\20015OIH.tif): Unspecified error
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U S Environmental
Protection Agency, have been grouped into nine series These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields
The nine series are
1 Environmental Health Effects Research
2 Environmental Protection Technology
3 Ecological Research
4 Environmental Monitoring
5 Socioeconomic Environmental Studies
6 Scientific and Technical Assessment Reports (STAR)
7 Interagency Energy-Environment Research and Development
8 "Special" Reports
9 Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL MONITORING series
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161
-------
EPA-600/4-80-013a
February 1980
EVALUATION OF THE REAL-TIME
AIR-QUALITY MODEL
USING THE RAPS DATA BASE
Volume 1. Overview
by
Ronald E. Ruff
Atmospheric Science Center
SRI International
Menlo Park, California 94025
Contract No. 68-02-2770
Project Officer
John S. Irwin
Meteorology and Assessment Division
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection Agency, nor does men-
tion of trade names or commercial products constitute endorsement or recom-
mendation for use.
11
-------
ABSTRACT
The theory and programming of statistical tests for evaluating the
Real-Time Air-Quality Model (RAM) using the Regional Air Pollution Study
(RAPS) data base are fully documented in four report volumes. Moreover,
the tests are generally applicable to other model evaluation problems.
Volume 1 presents an overview of the tests, displays, software and appli-
cation of the resulting statistical package. This report was submitted
in fulfillment of Contract No. 68-02-2770 by SRI International under the
sponsorship of the U.S. Environmental Protection Agency. This report
covers a period from 1 October 1977 to 1 April 1979, and work was comple-
ted as of 1 April 1979.
111
-------
CONTENTS
Abstract iii
Figures vi
Tables vi
1. Introduction 1
Background 1
Objectives 2
2. Overview 4
General approach 4
Accomplishments 6
3. Technical Summary 7
Preview 7
Selection of statistical tests 7
Statistical and display software 10
Evaluation of the RAM 11
References 13
Appendix
A. Description of the test data base 14
-------
FIGURES
Number Page
1 Project flow chart 5
A-l Vertical profile of Millstone Unit 2 and NUSCO
meteorological tower, looking 344° 17
A-2 Instrumentation layout at Millstone station 18
A-3 Comparison between observed normalized
centerline concentrations and those predicted
by the split-H model for S^ and S2 stability
classification criteria 21
TABLES
Number Page
A-l Summary of Millstone SF, Tracer Data 19
vi
-------
SECTION 1
INTRODUCTION
BACKGROUND
Until recently, the ability to provide objective evaluations of air
quality models has suffered for two reasons. In the first place, adequate
test data bases have been scarce, and secondly, the selected statistical
procedures have often been inadequate. The first problem has been alle-
viated somewhat through the acquisition of the extensive Regional Air
Pollution Study (RAPS) data base. The primary objective of the RAPS was
to provide a data base useful in evaluating, upgrading, and developing
air quality models.
With the availability of the comprehensive RAPS data base in mind,
EPA retained SRI International to develop a statistical package that
would overcome the limitations of currently used statistical procedures.
While EPA sought a generally applicable statistical package, they recog-
nized the need to test the package, and for this purpose chose the Real-
l ^
Time Air-Quality Model (RAM) along with the RAPS data base.
RAM
The RAM was developed by EPA several years ago as a tool for calcu-
lating SO^ surface concentrations of 1-hour and 24-hour averaging times.
These can be used to form estimates of SO^ concentrations for 3-hour,
monthly, seasonal, and annual averaging times. This Gaussian-plume
multiple-source (point and area) model requires hourly input data for
emissions and meteorological parameters from a single site. Hence, the
model output consists of an SO. concentration prediction for each hour.
k
References are listed at the end of this report,
-------
RAPS Data Base
The RAPS was conducted in St. Louis because that geographic area
already had a fairly good historic air quality data base and was one of
the least complicated urban areas to model. (St. Louis has fairly level
terrain and is free of pollutant intrusion from other large urban areas.)
The heart of the RAPS is the 25-station air-monitoring network that col-
lects data on the criteria pollutants (including SO ) and meteorological
parameters (wind, ambient temperature, dew point temperature, and temper-
ature gradient). In addition, numerous special measurements were made
at the monitoring sites, other surface locations, and aloft. Hence, the
data base is rich not only in the SO measurements needed to evaluate
model performance, but also in the emission and meteorological measure-
ments needed to understand the limitations of the RAM input parameters.
OBJECTIVES
The overall objective of this research is the implementation of a
computer-based statistical package useful in both evaluating model per-
formance and gaining insight into the causes of poor performance. Spe-
cific goals of the delivered statistical package (software and reports)
include the following:
• Figures of merit (or scores) that describe the absolute accuracy,
temporal distribution, and some spatial features of the compared
parameters (observed versus predicted concentrations).
• Techniques that enable one to relate such figures of merit to
the model input parameters.
• Auxiliary displays that allow one to visualize the significance
of the figures of merit.
• Software compatible to the EPA Univac 1110 and readily applicable
to the evaluation of the RAM using the RAPS data base.
• Complete documentation that allows one to adapt the package to
any air quality model evaluation problem, with a special emphasis
on the RAM evaluation.
Because the complete RAPS data base was not ready for use in time,
the actual application of the developed package to the RAM evaluation
was not the immediate goal of this research. A special test data base
-------
described later was used to check the software package. However, appli-
cation to the RAM evaluation is the immediate goal of follow-through
research.
-------
SECTION 2
OVERVIEW
GENERAL APPROACH
The main project work elements are identified in the flow chart of
Figure 1. Basically, the general approach was formulated in the proposal
but was modified slightly because essential parts of the RAPS data base
were not immediately available. The RAPS air-quality data were still
undergoing quality assurance checks, and the emissions data base was not
ready for input to the RAM. Consequently, it was necessary to select a
different test data base.
2
The test data base was acquired in a recent SRI tracer study.
This data base is summarized in Appendix A. While it would have been
more efficient to work with a segment of the RAPS data base, the test
data base offered the following advantages:
• It had previously been analyzed, so some of the results from our
newly programmed techniques could be corroborated.
• A limited but well-defined set of meteorological model input
parameters is part of the data base, so model performance could
be evaluated as a function of these parameters.
Concurrent with the selection of the test data base, many statis-
tical techniques were considered for inclusion in a recommended statis-
tical evaluation package. Essentially, all recommended tests and aux-
iliary displays were implemented in six distinct computer programs which
were tested and documented.
As shown in Figure 1, a second phase consists of adaptation of the
statistical package to the RAM evaluation problem, using the RAPS data
base.
-------
FORMULATE
GENERAL APPROACH
SELECT
TEST DATA BASE
REVIEW AND RANK
EVALUATION TECHNIQUES
SELECT
EVALUATION TECHNIQUES
ADAPT AND PROGRAM
INSTALL AND TEST
DOCUMENT
FIRST PHASE
SECOND PHASE
ADAPT SOFTWARE FOR
RAPS DATA BASE
APPLY SOFTWARE PACKAGE
TO RAM EVALUATION
DOCUMENT
Figure 1. Project flow chart.
-------
ACCOMPLISHMENTS
Six statistical programs were developed, installed on the EPA Univac
1110, and successfully tested. In general, the combined programs meet the
the objectives outlined in Section 1. The techniques and programs are
documented in this four-volume final report.
-------
SECTION 3
TECHNICAL SUMMARY
PREVIEW
Six statistical programs have been developed for the purposes of
model evaluation. A distinction is made between a "final-evaluation"
statistic and an "intermediate-evaluation" statistic. The first program--
accuracy score--evaluates the performance of the model, and as a final
evaluation statistic is considered the bottom-line computation. The
remaining five tests are intermediate-evaluation statistics that are meant
to assist one in the diagnosis of the causes of poor model performance.
Most of the programs contain auxiliary display software that allows
the user to visualize the results as an aid to determining the meaning
of the numerical output. Each program was installed and tested on the
EPA Univac 1110 computer system in Research Triangle Park, North Carolina.
A Tektronix 4014 terminal was used for graphic outputs.
The following subsections summarize the statistical tests, the
rationale behind their selection, the auxiliary displays, the software,
and evaluation procedures, which are described more comprehensively in
Volumes 2 and 4 of this report.
SELECTION OF STATISTICAL TESTS
In Volume 2 we consider a number of statistical tests for the pur-
poses of final evaluation and intermediate evaluation. (It is noted that
there is often some interchangeability between tests used for both
purposes.)
Final Evaluation
Various loss functions were implemented as a final-evaluation sta-
tistic. These were termed Accuracy Scores. The Accuracy Score program
is applied to a data base of observed and predicted concentrations. As
-------
the difference between observation and prediction decreases, so does the
Accuracy Score. Hence, Accuracy Score is like a golf score—the lower,
the better.
The following eight Accuracy Score tests are recommended:
• Average absolute error
• Average squared error
• Percent incorrect predictions above an absolute error threshold
• Percent incorrect prediction above a constant percentage error
threshold
• Symmetric high-low loss function
• Asymmetric high-low loss function
• User-supplied loss matrix
• Location of maximum concentration.
%
The first seven accuracy scores are concerned with errors in pre-
diction both for a given location and for the composite at all locations.
The eighth accuracy score is concerned with losses due to an error in
predicting the location of the maximum observed concentration.
For the first two loss functions, calculation of the accuracy score
is straightforward from the objective comparison of observed and predicted
concentrations. The third through fifth tests require entry of a user-
supplied threshold value. For cases when agreement between observation
and prediction is within the threshold value, a loss of zero is assigned;
otherwise, the loss is 1. The sixth and seventh tests again use the
threshold concept, but assign subjectively derived loss values supplied
by the user. (Hence, in these cases, the user can determine the types
of disagreement that are most important and weight them accordingly.)
The eighth test also uses user-supplied loss values that typically are
proportional to the distance between the location of the observed maximum
and the location of the predicted maximum.
-------
The output of each test is a score that is the summation of scores
of each individual comparison. In addition, confidence limits for the
accuracy score are computed. As a software option, the user can display
the accuracy scores on a geographical map. Then the spatial trends in
prediction accuracy can be assessed.
Intermediate Evaluation
A number of statistical tests were considered for the purpose of
intermediate evaluation. (As noted in Volume 2, there are special cases
where these tests can also be used as a final evaluation tool.) The
following tests and displays are recommended and included as part of the
software package:
• Scatter plot
• Linear regression with confidence and prediction bands
• Pearson's correlation coefficient
• Chi-square test and frequency distribution displays
• Interstation error correlation
• Multiple regression of error residuals
• Analysis of residual time series.
In the software package, as discussed later, some of the above tests
were combined to make computer execution more efficient. In the follow-
ing paragraphs, the tests are considered separately.
A simple scatterplot, also known as a scattergram, is one of the
most useful evaluation tools because it displays a case-by-case compari-
son of all data (i.e., observed versus predicted concentrations). A
regression analysis can be added to compute and display (on the scatter-
plot) a least-squares fit of the data. Confidence and probability bands
can be calculated and added to the display to demonstrate the statistical
significance of the regression (least squares) line and the probability
of a new test point falling outside the bands, respectively. The
Pearson's correlation coefficient is calculated as a measure of how
linear the comparison is.
-------
The chi square test and accompanying frequency distribution display
are a measure of how well the predicted frequency distribution agrees
with the observed frequency distribution. This test is also a useful
final evaluation statistic when the distribution itself is important and
not the point-by-point comparison.
The interstation error correlation tests examines the correlation
between station errors. (Error is the difference between observed and
predicted concentrations.) This test is useful for identifying geo-
graphical (or spatial) bias in the model comparison process. Correlation
of the error (residuals) with the model input parameters is computed by
the multiple regression of error residuals test. The primary purpose of
this test is to diagnose which of the input parameters most influence the
model's poor performance. The analysis of residual time series test
examines the errors for any cyclical behavior. It determines whether
errors occur on a periodic basis and what the time period is.
STATISTICAL AND DISPLAY SOFTWARE
The statistical tests and associated displays mentioned in this sec-
tion were programmed, debugged, and tested. The Pearson's correlation
coefficient and linear regression tests (with scatterplot showing confi-
dance and prediction bands) were combined into one test named "bivariate
regression and correlation". Thus, the six individual programs have test
names as follows:
• Accuracy Score
• Residual Time Series
• Chi-square Goodness of Fit
• Bivariate Regression and Correlation
• Interstation Error Correlation
• Multiple Regression of Error Residuals.
The first five programs were written in the FORTRAN V programming lan-
3
guage. The last test was written in the specialized SPSS language.
10
-------
Auxiliary programs were written or converted to complement the sta-
tistical tests. Several SPSS procedures were written to sort, retrieve,
and display selected portions of the data base. Also, to assist the
evaluator in visualizing differences between the observed and predicted
concentrations, programs to display the frequency distributions on a
logarithm-probability axis are included in the package.
The software has some special features unique to the EPA Univac 1110
computer system and EPA-supported software for the Tektronix 4014 graphics
terminal. However, the programs are generally compatible with software
supported on most major computers and are therefore easily convertible.
The software is fully documented in Volume 3 of this final report.
EVALUATION OF THE RAM
The general method of applying the statistical and display programs
remains essentially the same regardless of the specific air-quality model
being tested. To adapt the programs to the RAM evaluation, only minor
software changes must be made.
Evaluation of a model requires the following steps:
• Preparation and analysis of the data base
• Selection of the specific data set of interest
• Selection and application of the statistical test
• Interpretation of the test results.
These steps are repeated in an iterative fashion. As test results are
analyzed, the evaluator tries another test and so forth until full value
has been garnered from the statistical tests.
As noted in Volume 4 of this report, when evaluating the RAM, one
or more specific indicators of poor performance are looked for. These
include:
• General bias occurs when the differences in observed and pre-
dicted concentrations are mostly of the same magnitude and sign.
Several of the tests will detect this condition.
• Spatial bias is similar to general bias, differing in that the
patterns are examined for individual monitoring stations. The
11
-------
same tests for general bias are applicable here, but they are
applied to each site. In addition, the interstation error corre-
lation test is useful in addressing error residuals and then
spatial pattern.
• Temporal (cyclic) bias is detectable by the residual time series
test. Temporal bias is indicated when the errors in misprediction
are periodic.
• Randomness is indicated when poor agreement exists (between
observed and predicted concentrations) and there is no apparent
trend. This is best displayed in the bivariate regression and
correlation program—visually with the scatterplot and numeri-
cally with a low Pearson's correlation coefficient.
• Input-parameter dependence exists when poor performance can be
traced to one or more of the input parameters. The multiple
regressions of error residuals test is useful for this purpose.
Other tests, applied to subsets of the data base, can also be
effective in detecting input parameter dependence.
The tests are able to detect all of the above indicators of poor model
performance. Guidelines in their applications are presented in Volume
4.
12
-------
REFERENCES
1. Turner, D. B. and J. H. Novak, "Users Guide for RAM," EPA-600/8-78-016
a & b, Environmental Protection Agency Publication, (2 Volumes)
(1978).
2. Johnson, W. B., E. Shelar, R. E. Ruff, H. B. Singh, and L. Salas,
"Gas Tracer Study of Roof-Vent Effluent Diffusion at Millstone
Nuclear Power Station," Final Report, Project 3588, SRI Interna-
tional, Menlo Park, California (1975).
3. N. H. Nie, C. H. Hull, J. G. Jenkins, K. Steinbrenner, D. H. Bent,
Statistical Package for the Social Sciences (McGraw-Hill Book Co.,
New York, New York, 1975).
13
-------
APPENDIX A
DESCRIPTION OF THE TEST DATA BASE
OVERVIEW
During the fall, 1974, a series of diffusion tests was conducted by
SRI International (then Stanford Research Institute) at Millstone Nuclear
Power Station in southeastern Connecticut under sponsorship of the Atomic
Industrial Forum (AIF) and the Northeast Utilities Service Company
(NUSCO). This field study was planned by the AIF as part of its program
for evaluating roof-vent effluent diffusion from reactor and turbine
buildings.
The goal of the overall program was to develop and validate a more
realistic general model than was currently available for calculating
annual-average concentrations resulting from effluent diffusion from roof
vents. The limited objectives of the Millstone field study were to:
• Collect data sufficient to describe the diffusion of effluents
from the Millstone reactor and turbine building roof vents, espe-
cially under conditions of low wind speed.
• Use these data to develop and validate a site-specific diffusion
model for calculating annual-average concentrations.
• Evaluate the suitability of the planned experimental design for
an overall test program.
To collect the necessary data, diffusion tests were conducted,
during which sulfur hexafluoride (SFg) tracer gas was released from the
reactor building main vent and dibromodifluoromethane (Freon-12B2) from
three of the ten turbine building vents. Air samples were collected by
sequential multiple-bag samplers at 44 locations on three concentric
88-deg arcs at distances of 350, 800, and 1500 m from the SF, release
point. The air samples were analyzed by gas chromatography to obtain
downwind concentrations of the tracer gases.
14
-------
A visible oil-fog tracer was also injected into the reactor building
vent, and the resulting plume was photographed by two orthogonally posi-
tioned time-lapse cameras.
Supplementary special observations included three-component wind
measurements at three locations above the reactor and turbine building
roofs, and vent efflux speed and temperature excess on both buildings.
A complete set of meteorological measurements at heights up to 137 m
above the ground was also available at the site from the existing NUSCO
instrumented tower.
Data from a total of 36 hour-average runs were available. These data
were compared with the predictions of current models endorsed by the U.S.
Nuclear Regulatory Commission (NRC). In addition, the data were used
to develop and validate a realistic annual-average model, the "Split-H"
model.
The "Split-H" model takes account of the observed relationship
between the degree of plume entrainment and the velocity ratio (Ue/U),
where U is the vent efflux velocity and U is the wind velocity. The
name "Split-H" was given because the model divides concentrations at a
receptor into both a wake entrainment component coming from ground level
(H = 0), and a liftoff component coming from the effective vent height
(H = H ). The respective weighting of these components is taken to be a
function of U /U.
MILLSTONE SITE CHARACTERISTICS PLANT CONFIGURATION
AND INSTRUMENT LAYOUT
The Millstone Nuclear Power Station is located in Waterford,
Connecticut, on the tip of a small peninsula that extends southward into
Long Island Sound. The land rises gradually from the shore to reach a
maximum elevation of 100 ft (30 m) above mean sea level along the northern
edge of the test area. Approximately 30 percent of the test area is
*
Formerly the U.S. Atomic Energy Commission (AEC)
15
-------
wooded, composed mostly of deciduous trees. The remaining area has been
cleared and consists of brush and pasture land.
Millstone Station is currently composed of two electric generating
units, Units 1 and 2. At the time of the field study, Unit 2 was still
under construction, but all the external structures had been completed.
In addition, excavations had begun for a third unit. Testing was con-
fined to the reactor and turbine building roof vents of Unit 2. Meteo-
rological measurements at five different levels were available from the
137-m NUSCO tower located 500 m SSE of the Unit 2 reactor building vent
(see Figure A-l).
The original experimental design contained in the General Test Plan
called for a 180-deg array of bag samplers on at least two arcs. This
layout had to be modified for the Millstone study because of the con-
straints imposed by the surrounding waters of Long Island Sound. The
instrumentation layout that was adopted during the field tests consisted
of 45 samplers located at 40 stations arranged on three arcs extending
out to 1500 m, plus an upwind background station, as depicted in Figure
A-2. Five samplers were elevated at a height of 25 m by means of towers
uniformly spaced along the outer arc.
TEST RESULTS
Table A-l presents the relevant data from each of the 36 analyzed
SF.. runs. As indicated, the Pasquill stability classes (A through G)
o
were derived by using temperature differences (AT) between two sets of
heights. The scheme, called S1, uses the observed AT between the 137-m
and 10-m levels, while the S,, scheme uses the AT between 43 m and 10 m.
The concentrations (X) were normalized with respect to wind speed (U) and
source strength (Q). Other variables in the table are wind direction (Q )
and efflux velocity (U ).
16
-------
CO
O)
c
-------
N
MILLSTONE
NUCLEAR
POWER
STATION
238.5 deg
: 150.8 deg
LEGEND
+ Air Samplers at 1 m Height
Air Samplers at 1 m and 25 m Heights
Site Boundary
A NUSCO 137-m Meteorological Tower
^^B Time-Lapse Cameras
O£l SRI Sample Analysis cacility
SA-3588-3
Figure A-2. Instrumentation layout at Millstone station.
18
-------
M-
C
O
4J
03
f*l
*O
1
0
Cr
3
1
O
z
0
o
»n
n
o
o
X
o
("•»
II
*-
«?r
o
o "")_,
« "
0
o
"^if4
o
0 '"^j
.3"
M
O
u
o '"^,
*™
u
M
0
C/l
V)
Q M
^
0 <"
0
u
00 ^
s <
3
a"
«5
= j
n
8
"oi
*-H«
h V
CO js
W H
«^
O ~*
Li
c «
51
2
t4 ^J -4 ,-1
o -* o -S'^ro^T^j'O
P^ ^ r-« 1-4 •-«
i
o>«r,«r,«-«10-C^--00-000«,^CO^C.<«>^^^OOCO«>
•yrfy?-y???-? ^>
(OOJeooo sf st in ir,o-<--oooo»<»1o,-j_j__<.-JCT^n<-imuM^coooOO.-.c.J22"*'* > >
r-l «-f ^-< (M Cl •-( CJ
---------„«„„„„**«»*„„
•ri S
E £
§~ Vi
*w
"la
<-> u c o
11
41 <]
a A-
E -M c c
41 -O -D
oo e o o
-------
VALIDATION OF THE "SPLIT-H" MODEL
Figure A-3 compares the measured normalized-centerline tracer con-
centrations with those predicted by the "Split-H" model for the two dif-
ferent stability classification schemes, S.. and S_. (The test data base
for the RAM evaluation uses only the S version. The distinction among
the radial distances--350 m, 800 m, and 1500 m--serves as an analog to a
distinction among monitoring sites—labeled 1, 2, and 3 in the test data
base.)
20
-------
1000
500
in
'o 200
100
tr
a.
20
10
I '
t 350 m
A 800 m
O 1500 m
10
20
50 100
200
500 1000
MEASURED XU/Q — 10 6 m"2
(a) S.
1000
500
(N
'E
(O
'o 200
I
^ 100
I-
y
Q
Ui
oc
50
10
10 20 50 100 200
MEASURED XU/Q — 10~6 m~
500 1000
SA-3588-25
Figure A-3. Comparison between observed normalized centerline
concentrations and those predicted by the spltt-H model
for S.j and S2 stability classification criteria.
21
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/4-80-013a
2.
4. TITLE AND SUBTITLE
EVALUATION OF THE REAL-TIME AIR-QUALITY MODEL USING
THE RAPS DATA BASE
Volume 1. Overview
7 AUTHOR(S)
Ronald E. Ruff
9. PERFORMING ORGANIZATION f
SRI International
333 Ravenswood Avenue
Menlo Park, California 94025
JAME AND ADDRESS
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory — RTP, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
3. RECIPIENT'S ACCESSION NO.
5 REPORT DATE
February 1980
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO
Final Report
SRI Project 6868
10. PROGRAM ELEMENT NO.
1AA603 AA-26 (FY-77)
11. CONTRACT/GRANT NO.
68-02-2770
13. TYPE OF REPORT AND PERIOD COVERED
FINAL 8/77-4/79
14. SPONSORING AGENCY CODE
EPA/600/09
15 SUPPLEMENTARY NOTES
16. ABSTRACT
The theory and programming of statistical tests for evaluating the Real-Time Air-Quality Model (RAM) using the
Regional Air Pollution Study (RAPS) data base are fully documented in four report volumes. Moreover, the tests
are generally applicable to other model evaluation problems. Volume 1 presents an overview of the tests, displays.
software, and application of the resulting statistical package.
17. KEY WORDS AND DOCUMENT ANALYSIS
a DESCRIPTORS
* Air pollution
* Mathematical models
* Evaluation
* Tests
* Computer systems programs
* Statistical tests
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
b. IDENTIFIERS/OPEN ENDED TERMS
Real-Time Air-Quality Model
Regional Air Pollution Study
Data Base
19 SECURITY CLASS (This Report)
UNCLASSIFIED
20 SECURITY CLASS (This page)
UNCLASSIFIED
c. COSATI 1 icld/Ciroup
13B
12A
14B
09B
21 NO. OF PAGES
28
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
22
-------
c
CD
m c
3 3
5 *
o a.
3 y
(T> CU
2 <»
0) <"
C
O
3
M
~l
•3-?f
- "1 O
•y
* *>
.*&%
•^b
&i
a D
i A
1.S-.
— »•
OJO
T -o
2 O 3
a
il
Sg?
O 3 rn
I B
NJ
en
a)
O
•3
3
CD
3
(T
U)
o
rr
m > ID
~D co r
> m S
-n -o
3 Q. Q)
(B 3
3 Q.
CJ
------- |