United States Office ot Air Quality EPA-450/4-84-017
Environmental Protection Planning and Standards June 1984
Agency Research Triangle Park NC 27711
Ajr ~
Evaluation of
Complex Terrain
Air Quality
Simulation Models
-------
-,#
-------
EPA-450/4-84-017
Evaluation of Complex Terrain Air Quality
Simulation Models
by
David Wackier and Richard Londergan
TRC Environmental Consultants, Inc.
800 Connecticut Boulevard
East Hartford, CT06108
Contract No. 68-02-3514
U..S Environmental Protection Agency
F >g.:>n V. Library
2jO South Dearborn Street
Ch'cago, tsUncis 60604
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
June 1984
-------
DISCLAIMER
This report has been reviewed by the Office of Air Quality Planning
and Standards, U.S. Environmental Protection Agency, and approved for pub-
lication as received from TRC, Environmental Consultants, Inc. Approval
does not signify that the contents necessarily reflect the views and policies
of the U.S. Environmental Protection Agency, nor does mention of trade names
or commercial products constitute endorsement or recommendation for use.
Copies of this report are available from the National Technical Information
Service, 5285 Port Royal Road, Springfield, Virginia 22161.
U,S. Environmental Protection
-IT -
-------
TABLE OF CONTENTS
SECTION PAGE
1. INTRODUCTION 1
2. COMPLEX TERRAIN MODELS 3
Documentation 3
Technical Features 4
Model Input Data Requirements 4
Source Data 4
Receptor Date 11
Meteorological Data 11
3. DATA BASES FOR COMPLEX TERRAIN EVALUATION 13
Cinder Cone Butte Data Base 13
Source Data 16
Meteorological Data 16
Tracer Data 21
Data Selected for Model Input 21
Tracer Release Information 21
Ambient Tracer Concentrations 23
Meteorology 23
Westvaco-Luke Data Base 27
Source Data 27
Meteorology 27
Data Selected for Model Input 33
Source Information 33
Air Quality Data 34
Meteorology 34
4. STATISTICS APPROACH 43
Data Sets for Comparison of Observed and Predicted
Concentrations 43
Peak Concentrations 45
Comparisons of All Concentrations 46
Statistical Analysis of Model Performance 46
Statistical Measures for the Full Westvaco Data
Set 50
IMPACT Model: Analysis of Select Hours for
Westvaco 50
Statistical Measures for the Cinder Cone Butte
Data Set 53
-111-
-------
TABLE OF CONTENTS (Continued)
SECTION PAGE
5. MODEL PERFORMANCE RESULTS 57
Westvaco Full Year Results 57
Statistics for 2,5 Highest Values 57
Statistics for Highest Concentration at Each
Station 63
Statistics for Highest Concentrations by Event . 66
Statistics for All Concentrations Paired in
Time and Space 66
Westvaco - Impact Select Hour Results 74
Statistics for 25 Highest Values 74
Statistics for Highest Concentrations at Each
Station 76
Statistics for Highest Concentrations by Event . 76
Statistics for All Concentrations Paired in Time
and Space 30
Cinder Cone Butte Results 83
Statistics for 25 Highest Values 83
Statistics for Highest Concentrations by Event . 84
Statistics for All Comparisons Paired in Time and
Space 89
6. SUMMARY AND CONCLUSIONS 93
Summary of Results 93
REFERENCES 95
APPENDICES
A TEST RUN PACKAGE: EESCRIPTION OF MODELS "AS-RUN" FOR
COMPLEX TERRAIN MODEL EVALUATION
B STATISTICAL TABLES OF MODEL PERFORMANCE FOR WESTVACO
C STATISTICAL TABLE OF MODEL PERFORMANCE FOR CINCER
CONE BUTTE
-iv-
-------
LIST OF FIGURES
FIGURE PAGE
3-1 Cinder Cone Butte Field Experiment Layout 15
3-2 Cinder Cone Butte Vertical Cross Section Northwest (315°)
to Southeast (135°) 17
3-3 Cinder Cone Tracer Gas Sampler Locations 22
3-4 Map of the Study Area Surrounding the Westvaco Luke Mill . 28
3-5 Westvaco Vertical Cross Sections for radials of 135°,
170°, and 310°. The Westvaco Stack Height Along with
Monitor Heights and Distances from the Stack are
Superimposed 29
LIST OF TABLES
TABLE PAGE
2-1 Distinguishing Features of the Complex Terrain Models as
Run for the Current Evaluation 5
2-2 Composite of all Meteorological Parameters Expected by the
Complex Terrain Models to Exercise Various Models
Functions 11
3-1 Periods When Tracer Tests Were Conducted During the Cinder
Cone Butte Experiment 14
3-2 Units and Averaging Times Corresponding to Measured Variables
Reported in the Cinder Cone Butte Data Base 18
3-3 Cinder Cone Butte Tower Instrumentation and Measures ... 19
3-4 Summary of Cinder Cone Butte Meteorological Inputs to the
Complex Terrain Models 24
3-5 Units and Averaging Times Corresponding to Measured Variables
Reported in the Westvaco Data Base 30
3-6 Instrumentation and Parameters Measured on the Westvaco
Meteorological Towers 31
3-7 Primary Hourly Meteorological Inputs Included in the Westvaco
Modelers' Data Base as Compiled by H.E. Cramer Associates 35
3-8 Data Substitutions Used by H.E. Cramer Associates in
Developing Westvaco Hourly Meteorology Inputs 36
3-9 Summary of Westvaco Meteorological Inputs to the Complex
Terrain Models 40
-v-
-------
LIST OF TABLES
TABLE PAGE
4-1 Summary of Data Sets for Model Evaluation 44
4-2 Statistical Estimators and Basis for Confidence Limits on
Performance Measures 47
4-3 Performance Measures and Statistics Calculated for the
Westvaco Unpaired (25 Highest) Data Sets 51
4-4 Performance Measures and Statistics Calculated for Westvaco
Data Sets Paired in Time or Location 52
4-5 Performance Measures and Statistics Calculated for the
Cinder Cone Butte Unpaired (25 Highest) Data Sets ... 54
4-6 Performance Measures and Statistics Calculated for the
Cinder Cone Butte Data Sets Paired in Time or Location 55
5-1 Comparison of 25 Highest Observed and Predicted SOa
Concentration Values (UG/M**3) (Unpaired in Time or
Location) for the 1-Hour Averaging Period Westvaco
(1980/1981) 58
5-2 Comparison of 25 Highest Observed and Predicted SOz
Concentration Values (UG/M**3) (Unpaired in Time or
Location) For Various Data Sets Model: Complex I
for the 1-Hour Averaging Period Westvaco (1980/1981)
60
5-3 Comparison of 25 Highest Observed and Predicted SOz
Concentration Values (UG/M**3) (Unpaired in Time or
Location) For the 3-Hour Averaging Period Westvaco
(1980/1981) 61
5-4 Comparison of 25 Highest Observed and Predicted S02
Concentration Values (UG/M**3) (Unpaired in Time or
Location) For the 24-Hour Averaging Period Westvaco
(1980/1981) 62
5-5 Comparison of Highest Observed and Predicted S02
Concentration Values (UG/M**3) Paired by Station for the
1-Hour Averaging Period Westvaco (1980/1981) 64
5-6 Comparison of Second Highest Observed and Predicted SOz
Concentration Values (UG/M**3) Paired by Station for the
1-Hour Averaging Period Westvaco (1980/1981) 65
5-7 Comparison of Highest Observed and Predicted S02
Concentration Values (UG-/M**3) Event-by-Event (Paired in
Time) For the 1-Hour Averaging Period Westvaco
(1980/1981) 67
-VI-
-------
LIST OF TABLES (Continued)
TABLE PAGE
5-8 Comparison of All Observerd and Predicted S02 Concentration
Values (UG/M**3) Paired in Time and Location for the
1-Hour Averaging Period Summary Table (Part 1) Westvaco
(1980/1981) 68
5-9 Comparison of all Observed and Predicted SOz Concentration
Values (UG/M**3) Paired in Time and Location (For Various
Data Sets) Model: COMPLEX I for the 1-Hour Averaging
Period Westvaco (1980/1981) 71
5-10 Highest (H) and Highest, Second-High (HSH) 1-Hour
Concentrations for Westvaco with Associated Meteorology. 72
5-11 Highest (H) and Highest, Second-High (HSH) 3-Hour
and 24-Hour Concentrations for Westvaco Model Runs. . . 73
5-12 Comparison of 25 Highest Observed and Predicted S02
Concentration Values (UG/M**3) (Unpaired in Time or
Location) for the 1-Hour Averaging Period Westvaco
(1980/1981) Hours Selected for Impact Model Runs. . 75
5-13 Comparison of Highest Observed and Predicted S02
Concentration Values (UG/M**3) (Unpaired in Time or
Location) for the 1-Hour Averaging Period Westvaco
(1980/1981) Hours Selected for Impact Model Runs. . 77
5-14 Comparison of Second Highest Observed and Predicted SOz
Concentration Values (UG/M**3) Paired by Station for the
1-Hour Averaging Period Westvaco (1980/1981) Hours
Selected for Impact Model Runs 78
5-15 Comparison of Highest Observed and Predicted S02
Concentration Values (UG/M**3) Event-by-Event (Paired in
Time) For the 1-Hour Averaging Period Westvaco
(1980/1981) Hours Selected for Impact Model Runs . . 79
5-16 Comparison of All Observed and Predicted S02 Concentration
Values (UG/M**3) Paired in Time and Location for the
1-Hour Averaging Period Summary Table (Part 1) Westvaco
(1980/1981) Hours Selected for Impact Model Runs . . 81
5-17 Comparison of 25 Highest Observed and Predicted Relative
Concentration Values (10**(-6) S/M**3) (Unpaired in Time
or Location) For the 1-Hour Averaging Period Cinder
Cone Butte (1980) 85
-vi i-
-------
LIST OF TABLES (Continued)
TABLE PAGE
5-18 Comparison of 25 Highest Observed and Predicted Relative
Concentration Values (10**(-6) S/M**3) (Unpaired in Time
or Location) for Various Data Sets Model: COMPLEX I
For the 1-Hour Averaging Period Cinder Cone Butte
(1980) 86
5-19 Comparison of Highest Observed and Predicted Relative
Concentration Values (10**(-6) S/M**3) Event-by-Event
(Paired in Time) For the 1-Hour Averaging Period
Part 1 Cinder Cone Butte (1980) 87
5-20 Comparison of Highest Observed and Predicted Relative
Concentration Values (1C**(-6) S/M**3) Event-by-Event
(For Various Data Sets) Model: COMPLEX I For the
1-Hour Averaging Period Cinder Cone Butte (1980). . . 89
5-21 Comparison of All Observed and Predicted Relative
Concentration Values (lC**(-6) S/M**3) Paired in Time
and Location for the 1-Hour Averaging Period Summary
Table (Part 1) Cinder Cone Butte (1980) 91
-VI11-
-------
SECTION 1
INTRODUCTION
The Environmental Protection Agency (EPA) is currently involved in a
study to evaluate the performance of air quality dispersion models using
statistical measures recommended by the American Meteorological Society. It
was EPA's intent, as published in a notice in the March 1980 Federal
Register, to provide organizations the opportunity to submit dispersion
models for possible inclusion in the next revision of EPA's "Guideline on
Air Quality Models".1 EPA has undertaken a systematic evaluation of these
models to decide in an objective manner which models should be included in
the guideline and what recommendations should be made concerning the use of
these dispersion models for regulatory applications. Several categories of
models have been identified including models designed for complex terrain
situations. TRC, working under contract to EPA, has assembled the
aerometric data sets needed for model input and comparison, set up and run
the complex terrain models and produced statistics relating observed and
predicted air quality.
In September 1980 the American Meteorological Society (AMS), as a
professional organization with expertise in atmospheric dispersion,
organized a workshop (sponsored by EPA) to consider the issue of model
performance - evaluation. The 1980 workshop held at Woods Hole,
Massachusetts, produced a report entitled "Judging Air Quality Model
Performance."2 This report contains recommended statistical procedures
for comparing observed air quality with model predictions. The procedures
recommended by the Woods Hole workshop provided the basis for the
statistical comparisons presented in this report. TRC has performed similar
studies for EPA to evaluate eight rural models3'4 and six urban
models5. On the basis of these studies and subsequent comments by the AMS
reviewers, a trimmed-down list of statistical comparisons are provided for
the complex terrain model evaluation.
In Section 2 the eight complex terrain models are described. The models
include COMPLEX I, COMPLEX II, COMPLEX/PFM, 4141, PLUMES, RTDM, SHORTZ and
IMPACT. The distinguishing technical features of these models, as run for
the current evaluation, are described. Also, the procedures for
implementing and testing the models and the unique input data requirements
are presented.
-1-
-------
The above models have been evaluated with data obtained from two field
measurements programs which were carried out in complex terrain
environments. The Cinder Cone Butte tracer data base provides air quality
measurements with good spatial resolution (94 samplers) for a limited number
of study hours (104). The Westvaco data base comes from a rigorous
routine-measurements program one year of hourly data at 11 stations, for
this study designed for regulatory considerations. These data sets, along
with supplemental data, are described in Section 3.
In Section 4 the statistical approach is described. The sets of
observed and predicted concentration values have been paired in a variety of
ways to provide statistical model performance comparisons that reflect
either high concentration values or all concentration values, with and
without pairing according to time and space.
The results of this study are presented in Section 5. The tables of
statistical comparisons based on the performance measures recommended by the
AMS workshop are presented in this section for all eight models run with
both data bases.
Three appendices provide additional information. Appendix A is a copy
of the TRC document "Test Run Package: Description of the Models 'As Run'
for Complex Terrain Model Evaluation" which describes test run procedures,
model-by-model code modifications and listings of model input options
selected by the model developers for this evaluation. Appendices B and C
contain statistical tables for Westvaco and Cinder Cone Butte,
respectively. These tables provide statistical results by model for each
type of data comparison and for subsets by meteorology and source-receptor
geometry.
-------
SECTION 2
COMPLEX TERRAIN MODELS
The following eight complex terrain air quality models have been
evaluated by TRC using the performance measures recommended by the American
Meteorological Society:
COMPLEX I
COMPLEX II
COMPLEX/PFM
4141
PLUMES
RTDM
SHORTZ
IMPACT
Of these eight models, seven are based on the Gaussian plume assumptions,
while IMPACT is a numerical grid model. Specific methods for prescribing
plume rise, transport and dispersion differ from model to model, but all of
the models require similar basic user-supplied input data describing source
characteristics, receptor locations, and representative meteorology. IMPACT
generally needs more detailed meteorological input data than the Gaussian
models.
DOCUMENTATION
Computer code and documentation for each of the complex terrain models
are available to the public. COMPLEX I and COMPLEX II were developed by EPA
and are described as screening techiques for applications in complex terrain
environments 27. Currently no user guides exist for these models.
Documentation exists as part of the FORTRAN code and also in the MPTER
user's manual6 from which these two models were adapted. COMPLEX/PFM7
was developed for EPA by Environmental Research & Technology, Inc. (ERT) as
an adaptation of COMPLEX I with provisions for either COMPLEX I, COMPLEX II
or potential flow model (PFM) calculations. The model 3141/41413 was
developed by Enviroplan, originally as a modified version of CRSTER9, and
more recently as a modified version of MPTER. The 4141 option of the MPTER
version was employed in this study. PLUMES10 was developed by Pacific Gas
and Electric, and the Rough Terrain Diffusion Model (RTDM)11 was developed
by ERT. SHORTZ12 was developed by the H.E. Cramer Company. An updated
version of SHORTZ which includes an algorithm to account for vertical wind
direction shear13 was used in this study. Two versions of IMPACT
(Integrated Model for Plumes and Atmospheric Chemistry in Complex
Terrain)14'15 were submitted to EPA. The authors of both versions were
contacted and they agreed that very little difference between results from
the two versions was likely, at least for the purposes of this evaluation.
Therefore, the Fabrick and Haas version was selected for this evaluation.
-------
TECHNICAL FEATURES
Distinguishing features of the complex terrain models as run for the current
evaluation are listed in Table 2-1. The information listed in Table 2-1 is
presented by model and then by generalized modules including transport,
dispersion/stability, plume rise/terrain impaction and limits to vertical mixing.
These modules represent physical processes that the models are attempting to
simulate. It is not the intent here to fully describe each of the complex terrain
models, but rather to list briefly the primary technical features that distinguish
one model from another. In-depth technical discussions of each model can be
obtained from the appropriate model-aser guides. The reader is encouraged to
refer to the users manuals for technical details and references.
As part of the model evaluation process, test run packages were prepared and
supplied to the model developers for their review and concurrence. A description
of this procedure can be found in Appendix A which contains a copy of one of the
test run package documents16. This document also summarizes the model code
modifications made by TRC and describes the input options selected by the model
developers for each model and data base. Modifications to the models were needed
to adapt each model to the EPA UJ'IVAC computer, to adapt particular models to
accept the source-receptor inventories and to format the output of calculated
concentrations for input to the statistics system.
It is also noted that ERT (RTDM) and H. E. Cramer Co. (SHORTZ) previously had
the opportunity to test their models using at least portions of the data sets
selected for this evaluation. Both data sets were previously used by ERT, while
only the Westvaco data set was used by the H. E. Cramer Co. According to the
developer^, the models were not modified based on these evaluations. However,
these developers could select model options and model inputs to optimize model
performance, based on previous experience.
MODEL INPUT DATA REQUIREMENTS
All of the complex terrain models reguire basic user-supplied input data
describing source characteristics, receptor locations and representative
meteorology. Other model inputs control options for data input/output and
technical considerations.
Source Data
Each of the models requires that the fixed geographic and geometric
characteristics of each source be specified by the model user. The location is
generally specified in Cartesian or polar coordinates except for the IMPACT model
which requires horizontal source locations to be specified as central cell
positions within a Cartesian three-dimensional grid. The stack base elevation,
physical stack height, and stack gas exit diameter are fixed variables also
required by each of the models.
Pollutant emission rate, stack gas exit velocity and stack gas temperature are
generally needed by the complex terrain models in the calculation of plume rise
and ambient concentration. The temporal variation of these parameters is
available as one-hour averages in the Cinder Cone Butte and Westvaco data bases.
Plume rise was not a factor in the Cinder Cone Butte tracer study (passive
releases) and therefore stacK velocity and temperature are not available. Many of
the models had to be modified to accept hourly varying source data.
-4-
-------
TABLE 2-1 DISTINGUISHING FEATURES OF THE COMPLEX TERRAIN MODELS AS RUN FOR
THE CURRENT EVALUATION
COMPLEX I
Transport
Wind speed as input at release height
- Wind direction as input
Dispersion/Stability
Turner stability categories (class 7 treated as class 6)
- Gaussian vertical distribution using rural, (P-G) az
- 22.5 horizontal sector averaging
Buoyancy induced vertical dispersion
Plume Rise/Terrain Impaction
- Terrain adjustments = .5, .5, .5, .5, .0, .0 for Stability A-F
- Minimum terrain approach = 10m
- Briggs final plume rise, including momentum rise
Stack tip downwash for non-passive plumes
Limits to Vertical Mixing
Full reflection at ground and mixing height
- Uniform vertical mixing beyond where az - 1.6 x mixing height
COMPLEX II
Transport
- Wind speed as input at release height
- Wind direction as input
Dispersion/Stability
Turner stability categories (class 7 treated as class 6)
Bivariate Gaussian distribution (PGT ay and az)
Buoyancy induced dispersion
(Continued on next page)
-5-
-------
TABLE 2-1 (Continued)
Plume Rise/Terrain Impaction
- Terrain adjustments = .5, .5, .5, .5, .0, .0 for stability A-F
Minimum terrain approach = 10m
- Briggs final plume rise including momentum rise
Stack-tip downwash for non-passive plumes
- Linear concentrations drop-off with height above plume centerline
Limits to Vertical Mixing
Full reflection from ground and mixing height
Uniform vertical mixing beyond where az = 1.6 x mixing height
COMPLEX/PFM
Transport
- Wind speed as input at release height for COMPLEX I/II calculations
- Wind speed adjusted in potential flow model (PFM) calculations as a
function of streamline deformation
- Wind direction as input
Dispersion/Stability
Turner stability categories (cla.ss 7 treated as class 6)
- COMPLEX I (22.5 sector averaging) for D, E or F stability when
plume is below dividing streamline height
- COMPLEX II for A, B, or C stability
- PFM (adjusted PGT ay and az ) for D, E or F stability when
plume is above dividing streamline height
Buoyancy induced vertical dispersion
Plume Rise/Terrain Impaction
- COMPLEX I/II terrain adjustments = .5, .5, .5, .5, .0, .0 for
stability A-F
- COMPLEX I/II minimum terrain approach = 10m
PFM plume height reduced for deformed streamlines
Modified Briggs layered plume rise
Stack tip downwash for non-passave plumes
(Continued on next page)
-6-
-------
TABLE 2-1 (Continued)
Limits to Vertical Mixing
- Full reflection at ground and mixing height
- Uniform vertical mixing beyond where oz = 1.6 x mixing height
4141
Transport
- Wind speed as input at release height
- Wind direction as input
Dispersion/Stability
- Turner stability categories (class G treated as class F)
- Bivariate Gaussian distribution (PGT oz and time-enhanced
PGT ay)
- Buoyancy induced dispersion
Plume Rise/Terrain Impaction
- Terrain adjustments = .5, .5, .5, .5, .25, .25
- Briggs transitional plume rise
Limits to Vertical Mixing
- Full reflection at ground and mixing height
- Uniform vertical mixing beyond where az = 1.6 x mixing height
IMPACT
Transport
- Input wind speed and direction at multiple sites extrapolated and
interpolated to 3-dimensional grid cells
- Divergence-free wind field created
(Continued on next page)
-7-
-------
TABLE 2-1 (Continued)
Dispersion/Stability
Finite difference solution to diffusion equation
Diffusivities from DEPICT model using Smith's (empirical) formulations
Plume Rise/Terrain Impaction
Plume/terrain approach controlled by wind and diffusivity fields
- Briggs layered plume rise including penetration of stable layers
Limits to Vertical Mixing
- Temperature stratifications incorporated into wind and diffusion
fields
PLUMES
Transport
Wind speed as input at release height
Wind direction as input
Dispersion/Stability
Stability categories from horizontal turbulence intensity
and time of day (class A treated as class B; class G treated as class
F)
Bivariate Gaussian distribution (PGT ay and az)
Enhanced horizontal dispersion due to vertical wind directional shear
- Buoyancy induced dispersion
Plume Rise/Terrain Impaction
Conservative modification to one-half plume height concept
Briggs final plume rise with determination of stable layer penetration
Limits to Vertical Mixing
- Full reflection at ground and mixing height
Uniform vertical mixing beyond where az = 1.6 x mixing height
(Continued on next page)
-------
TABLE 2-1 (Continued)
RTDM
Transport
- Wind speed extrapolated from release height to plume height
- Wind direction as input
Dispersion/Stability
Stability categories from vertical turbulence intensity (a*)
measured at Westvaco
Bivariate Gaussian distribution (dispersion coefficients from
measured turbulence data)
Buoyancy induced dispersion
Enhanced horizontal dispersion for plumes rising through a shearing
wind
Plume Rise/Terrain Impaction
Terrain impingement for stable plumes below critical height
Half height correction for neutral or unstable conditions and stable
conditions when plume exceeds critical height
- Briggs transitional plume rise; hourly potential temperature
gradients for stable plume rise
Stack tip downwash for non-passive plumes
Limits to Vertical Mixing
Partial terrain reflection; full mixing lid reflection
Mixing height adjustment for plume path
- Unlimited mixing height for stable conditions
SHORTZ
Transport
- Wind speed extrapolated from release height to plume height
Wind direction as input
(Continued on next page)
-9-
-------
TABLE 2-1 (Continued)
Dispersion/Stability
Bivariate Gaussian distribution (Cramer dispersion coefficients from
measured turbulence data)
Cramer technique for enhanced horizontal dispersion due to vertical
wind direction shear
Buoyancy induced dispersion
Plume Rise/Terrain Impaction
Terrain impingement within the mixing layer
Modified Briggs final plume rise; hourly potential temperature
gradient for stable plume rise
Stack tip downwash for non-passive plumes
Limits to Vertical Mixing
Full reflection at ground and mixing height
- Uniform mixing beyond where reflection terms (i=3) exceed exp(-lO)
Mixing height constant above s;ea level, for determination of plume
penetration
- Minimum actual mixing depth of H = u x 200 meters (u = wind speed)
for Westvaco; Height where vertical intensity of turbulence drops
below 0.01 radians for Cinder Cone Butte.
-LO-
-------
Receptor Data
Each of the complex terrain models produces calculated concentrations at
multiple receptor locations. In all of the models except IMPACT, discrete
receptors at arbitrary locations are defined in Cartesian or polar
coordinates. The IMPACT model defines receptor locations internally as the
central cell position within each of the 3-dimensional grid "boxes." All of
the complex terrain models require receptor elevations above a local
reference plane.
Meteorological Data
Meteorological data are used by the models to calculate transport,
dispersion, plume rise and limited mixing between sources and receptors.
The complex terrain models expect a broad range of meteorological
parameters, as summarized in Table 2-2. The IMPACT model allows data from
one or multiple meteorological towers to be internally pre-processed into
3-dimensional fields for input to the grid model. The other complex terrain
models are exercised with meteorological data from one "representative"
station. The representative input data sets used in this evaluation consist
of a composite of parameters measured at more than one site. A detailed
description appears in Section 3.
TABLE 2-2
COMPOSITE OF ALL METEOROLOGICAL PARAMETERS EXPECTED BY THE
COMPLEX TERRAIN MODELS TO EXERCISE VARIOUS MODEL FUNCTIONS
Transport
Wind Speed
Wind Direction
Anemometer
Height
Power Law
Exponents
Model
Dispersion
(Stability)
P-G Stability
Ge or Iy
a* or Iz
Wind Direction
Shear
Function
Plume Rise
Temperature
Wind Speed
dT/dZ
P-G Stability
Limited Mixing
Mixing Height
Temperature and
Wind Speed for
Froude Number
and Critical
Height
-11-
-------
Some of the complex terrain models contain preprocessor programs that
must be exercised in order to obtain a complete set of model-consistent
meteorological input data. CRSMET, the CRSTER preprocessor, is used to
generate hourly Pasquill-Gifford stability categories from on-site wind
speeds and National Weather Service (NWS) cloud observations for input to
COMPLEX I, COMPLEX II, COMPLEX/PFM and 4141. RTDM uses this data for the
Cinder Cone Butte application. Westvaco mixing heights from CRSMET are used
by COMPLEX I, COMPLEX II and 4141. CONVRT, the preprocessor for PLUMES, is
used to generate stability class from horizontal turbulence measurements.
Westvaco mixing heights are also generated from CONVRT. METZ is the SHORTZ
preprocessor which is used to generate mixing heights for Westvaco. The
PROFILE preprocessor to COMPLEX/PFM is used with the Westvaco data set to
develop vertical profiles of temperature and wind speed which are
subsequently needed by the model for calculations of Froude number and
critical streamline height. TRC developed preprocessors for providing
profiles of meteorological data needed as input to the IMPACT model.
Description of specific model input data for both Westvaco and Cinder
Cone Butte are provided in Section 3.
-12-
-------
SECTION 3
DATA BASES FOR COMPLEX TERRAIN MODEL EVALUATION
The complex terrain models have been evaluated with data obtained from
two field measurements programs which were carried out in complex terrain
environments. The Cinder Cone Butte tracer data base provides air quality
measurements with good spatial resolution for a limited number of study
hours. The Westvaco data base, containing a small number of stations for an
extended period of continuous monitoring, results from a rigorous
routine-measurements program designed for developing a model to be applied
in a regulatory setting. Terrain at the Westvaco-Luke Mill is steep, uneven
and rugged; Cinder Cone Butte is a simple, isolated terrain feature.
Both data bases were originally obtained for what might be called
research objectives, or diagnostic model evaluation. As a result, there
existed an overabundance of meteorological data which was trimmed down to
enable operational evaluation of the models. Trimmed down or "modeler's
data bases" were recommended for use in this evaluation so that a common set
of input data could be used in as many models as possible. The intention
was to reduce uncertainties in model predictions resulting from minor
differences in input data, and hence allow relative differences between the
models to be evaluated strictly on the basis of technical merit. Of course,
model input data requirements do differ somewhat from model to model. These
model requirements were accounted for in the preparation of test run and
final input data sets as described in this section.
CINDER CONE BUTTE DATA BASE
The Cinder Cone Butte experiment represents the first major component of
the EPA-sponsored Complex Terrain Model Development Program.17 ! 8 • ; 9 The
broad objective of the experiment was to determine the behavior and impact
of an elevated plume in the vicinity of an isolated elevated-terrain
feature. During the period between October 16, 1980 and November 12, 1980,
18 multi-hour dual tracer gas experiments were conducted during primarily
stable atmospheric conditions. The periods when tracer tests were conducted
are listed in Table 3-1.
As can be seen from Figure 3-1, Cinder Cone Butte is a roughly
axisymetric, 100 meter high isolated hill. The hill is located in
southeastern Idaho about 50 m south-southeast of Boise. The surrounding
Snake River Basin is a broad, nearly level plain.
-13-
-------
TABLE 3-1
PERIODS WHEN TRACER TESTS WERE CONDUCTED
DURING THE CINDER CONE BUTTE EXPERIMENT17
Experiment
No.
201
202
203
204
205
206
207
208
209
210
211
213
214
215
216
217
218
1980
Date
10/16
10/17
10/20
10/21
10/23
10/24
10/25
10/27
10/28
10/30
10/31
11/04
11/05
11/06
11/09
11/10
11/12
Experiment
Hours (PST)
1700-2300
1700-2300
0000-0800
0000-0800
0000-0800
0000-0800
0000-0800
1700-0100
1700-0100
0000-0740
0000-0800
0000-0800
0200-1000
0000-0600
0000-0700
0200-1000
0200-1000
Hours
6
6
8
8
8
8
8
7
7
7
8
8
8
6
7
8
8
Typical
Stability
E
E
E-F
E-F
E
E
E-F
E-F
F
E-F
E-F
E-F
E-F
E-F
E-F
E-F
E
-------
\\
\\
Northwest
Pibal and Tetharsonde •
Location
WPL Acoustic •
Radar
Fm/Cvy Radar
' ; WPLLidar"0'
•WPL Acoustic Radar
10-m Toiwar
ERTCommind
Po«t
( i
\ ,
\ i,
Southeast Pibal and
Tethersonde Location
JO.. I
L. ^ : I
1 Mila
Figure 3-1. Cinder Cone Butte field exoeriment layout. A is the 150 m tower;
B is the 30 m tower; C, D, E, F are 10 m towers(Ref. No. 17).
-------
Ground-level measurements of two tracer gases, sulfur hexafluoride
(SFs) and freon (CF3Br), were conducted during testing periods. Each
test persisted for approximately eight hours with tracer gas releases
occurring during five or six of these hours. Of the 111 test hours, 104
were used in this model evaluation. Tracer measurements were accompanied by
an extensive collection of meteorological measurements taken on multiple
levels of several towers located in the area and on the butte itself.
Additionally, plume characteristics were inferred from photography and
remote sensing.
The Cinder Cone Butte data base, measured and compiled by Environmental
Research and Technology (ERT) under contract to EPA, contains most of the
parameters which are needed for the complex terrain model evaluation. The
parameters describe source emissions, atmospheric dispersion characteristics
and ambient measurements of tracer concentrations.
Source Data
The two tracer gases, SF6 and CF3Br, were released passively from
different levels utilizing a mob:le crane. The range of SFe release
heights and release distances relative to the butte is displayed in Figure
3-2. The crane release heights ranged from 15 to 57 meters above the local
terrain at the base of the butte. Tracer release distances from the butte
center ranged from 540 meters to 1420 meters. The mobility afforded by the
release system enabled tracer releases directly upwind of the butte,
producing a high number of successful hours per test. Although the
variability of gas flow for SF6 and CF3Br was monitored using separate
rotameters, the weight loss of the cylinders was used to determine the
emission rate of each tracer. The source data base compiled by ERT consists
of the vertical and horizontal crane Location (relative to the butte peak)
for each release period and average emission rate for each test hour (see
Table 3-2).
Meteorological Data
Six instrumented towers were used to measure local meteorology. These
included: a 150 meter tower approximately 2 kilometers north of Cinder Cone
Butte; a 30 meter tower at the summit of the butte; and four 10 meter towers
on the hill. The locations of the towers (A, B, C, D, E, and F) are shown
in Figure 3-1. The direct measurements and derived parameters obtained for
each of the towers are given in Table 3-3.
Various atmospheric sounding devices were employed during tracer testing
periods. A tethersonde was operated usually at a location within 700 m of
the primary release point. An ascent-descent sequence conducted at a
minimum of once per hour generated profiles of temperature, pressure, wind
speed and direction to heights of at least 200 m above the local terrain.
When high wind speeds precluded tsthersonde operation, profiles were
obtained from minisonde flights. Hourly wind profiles were also derived
from pilot balloons (pibals). Additionally a frequency-modulated,
continuous-wave (FM/CW) radar, and two monostatic acoustic radars were
operated near the butte.
The meteorological tower data has been assembled by ERT and currently
resides on magnetic tape. Corrections were made by ERT to known errors in
the wind speed and wind direction measurements. Temperature corrections
were not made, given that the results of two independent audits were
-16-
-------
Qi
o
CO CO
< CO O
CO CO
C3
o
o
I
I
o
LO
(111)
X
1
tn
r sg g~i
°I/1U-
o
LO
CM
O
o
o
o
LO
o
o
LO
o
LO
CM
Ol
u
o c
0
LO
CM
o
• o
LT>
I
o
LO
o
o
o
LO
T 1
f:
c
o
U
cu
1/1
I/I
o
OJ
CU LO
•i-J ro
cu
C
o
i- rs
cu o
T3 ")
C
•>- o
o +->
CM
I
ro
CO
en
-------
tn cu
a e
S-l -H
£ H!
4-1 •-!
O OJ
01
CU _
4-J G
- 3 --1
>i G T3
r-l .H G
S-4 g CU
3 CU
O O CU G
£ i—I T) O
0) 4-1 G
T3 G -H
Q CU M
- e cu
CU rO 01
w 01 3
01 4H rH
cu O> O nJ
4J G -H O»
- 3 -^ Q) 4.) G
>. C T3 T3 G -H
i—I -H C O Q) rH
t-i g 0) g 3 CU-O
3 Oi C)1 g cu
O O cu G CD rd w
JC rH T3 O 01 Ol 3
CM
1
CO
w
J
s
H
W
W
J
3
M
1
Q
W
| i
C/3
,<£
O
H
O
2
M
Q
2
S
00
w
a
0
o
00
w
s
H i
H
U
2
M
O
OH
W
<*c
01
4J
"H
c
j — )
U
CU
4-1
CU
e
(0
S-i
m
O^
G
CJ1 0)
n3 6
t-l -H
1 cu H
Cu
ft]
fa
Ol
cu
4-1
3
6
in
-*
i
01
g
-a
cu
0)
cu
Ol
T3
G
•H
S
1
1— I
S-i
! 3
! 0
G
O
H
fO
4-J
G
cu
O
G
O
a
01
cu
4-1
3
G
g
in
01
a)
D
S-i
O^
cu
T3
G
O
•H
4-J
0
CU
S-l
•*-4
T3
TIJ
G
•H
^
Ol
CU
4-1
3
G
g
in
O
o
cu
S-i
3
4-1
a)
S-i
cu
cu
g
cu
4-1
01
0)
4-)
3
C
e
in
_*
i
G
•H
g
^1
i — I
G
0
•H
4-1
fO
•H
T3
rd
S-l
CU
cu
J_l
«
CO
u
01
Q)
4-1
3
g
in
01
C
<0
•H
rO
fO
S-i
o\°
4-1
O
£*i
4J
•H
M
C
cu
4J
G
•H
CU
o
G
cu
i-H
3
s^
3
4J
01
u
3
O
01]
^
• H!
G,
D!
01
Cn
-------
TABLE 3-3
CINDER CONE BUTTE TOWER INSTRUMENTATION AND MEASURES
Site
Instruments*
Direct Measures
Derived
Measures
Tower A
Level 0 (1 m)
Pyranometer
Net radiometer
Insolation
Net radiation
Level 1 {2 m)
Triaxial props
Cup and vane
RTD
U, V, W, IX, IY, IZ WS, WD
UX, VX SP, DR
T
Level 2 (10 m)
Triaxial props
Cup and vane
RTD
Fast bead thermistor
U, V, W, IX, IY, IZ WS, WD
UX, VX, U, 06 SP, DR
AT (10 m - 2 m) T
T,
-------
TABLE 3-3 (Continued)
CINDER CONE BUTTE TOWER INSTRUMENTATION AND MEASURES1
Site
Instruments*
Direct Measures
Derived
Measures
Tower B (Continued)
10 m
30 m
Towers C, D, E, F
2 m
10 m
Triaxial props
Cup and vane
RTD
Triaxial props
Cup and vane
RTD
Triaxial props
RTD
Triaxial props
Cup and vane
RTD
U, V, W, IX, IY, IZ WS, WD
UX, VX SP, DR
AT T
U, V, W, IX, IY, IZ WS, WD
UX VX SP, DR
AT T
U, V, W, IX, IY, IZ WS, WD
T
U, V, W, IX, IY, IZ WS, WD
UX, VX SP, DR
AT T
* All temperature sensors were mounted in aspirated radiation shields; an RTD
is a Resistance Thermometric Device.
U: westerly component of wind measured by east-west oriented propeller
V: southerly component of wind measured by north-south oriented propeller
W: vertical component of wind measured by vertically oriented propeller
SP: horizontal wind speed measured by cup anemometer
DR: horizontal wind direction measured by vane
o9: standard deviation of horizontal wind direction calculated from
vane output
UX: easterly component of wind calculated from the cup and vane outputs
VX: southerly component of wind calculated from the cup and vane outputs
WS: horizontal wind speed calculated from U and V
WD: horizontal wind direction calculated from U and V
IX: downwind intensity of turbulence
IY: crosswind intensity of turbulence (IY approximates cr9 for small
horizontal wind deviations)
IZ: vertical intensity of turbulence (IZ approximates a for small
vertical wind deviations)
T: temperature (resistance thermometric device)
aT: standard deviation of temperature
AT: temperature difference
-20-
-------
inconsistent. Turbulence intensity data were also left uncorrected although
errors in these data are known to exist due to the response characteristics
of the propeller sensors. Other identified errors that remain in the data
base are due to the effects of the wake of one instrument on another and the
effects of tower wakes on turbulence measurements and wind direction. Users
of this data set have been advised by ERT to give precedence where possible
to wind measurements from instruments that are more clearly out of wakes.
The meteorological tower data has been recorded on tape as five-minute
averages for the variables listed in Table 3-2. Data from pibal,
mini-sonde, and tethersonde flights are available on a separate magnetic
tape.
Tracer Data
Tracer samples were obtained with approximately 90 battery-operated
samplers which were sequentially operated for either 10 minute or 1 hour
periods. Figure 3-3 shows the locations of the 70 fixed samplers and the 10
movable samplers. The movable samplers were deployed either on the
northwest or southeast side of the hill, depending on the prevailing wind
direction. For a typical test, 60 of these 80 sites provided 1-hour average
samples and 20 were designed to obtain 10-minute average samples. An
additional 10 samplers were used: on masts for measuring plume reflection
from the ground; for measuring background concentrations; and as collocated
samplers for quality assurance purposes.
Bag samples were assayed for SFs and CF3Br concentrations using gas
chromatography. After all bags were analyzed, a data base consisting of
approximately 14,000 tracer concentration measurements representing the
entire experiment was assembled and recorded on magnetic tape. For each
10-minute and 1-hour assayed sample, the experiment number, sample
identification, sampling start time, sampling end time, SF6 concentration
and CF3Br concentration are stored on the data tape (see Table 3-2).
Data Selected for Model Input
A modeler's data base (MDB) was prepared by ERT from the archive of
Cinder Cone Butte data. This MDB contains hourly averages of tracer release
information, ambient tracer concentrations and meteorological parameters for
each of the 111 test hours in which tracer gas was released. Data needed by
the models and for the evaluation were selected from the MDB. Supplemental
data were also obtained to meet the needs of each of the complex terrain
models. These data are described below.
Tracer Release Information
The MDB contains tracer release information for 111 test hours
representing 17 different experiments. Freon gas was released along with
SFS for nine of the experiments. Only the SF6 releases were modeled in
this evaluation, since the SFS data had been shown to be of higher
quality. The tracer release information included:
-21-
-------
X
X
x
/x
/X
I
X
, \M.V c H4?*4^**^ *•
Fixed
Northwest Flows Only
Southeast Flows Only
0 20Q Meters
X-Tracer Release
Locations
x
X
m
x
if
•il
X
x;
•9
X X
^
&- —^
Fiqure "-"5. Tinder Cone Butte tracer gas sampler locations (Ref. No. 17
Contour intervals are 5 meters.
-22-
-------
average release rate for the period of each release
start and end time of each release
location of the crane (X, Y, Z, L) relative to a local reference
frame
Seven of the experiment hours were found to have release periods of less
than 40 minutes. These hours were then excluded from the evaluation, since
only hourly averages were to be modeled. The net result was that 104 test
hours were included in the evaluation.
The model results for Cinder Cone Butte have been evaluated on the basis
of relative concentration (X/Q), so emission rates were input to each model
as a fixed value of 1.0 g/s. The variable tracer emission rates provided in
the data base represent the average tracer release rate for the duration of
each release. Some of the experiment hours used in this study had release
periods of less than an hour (but greater than 40 minutes). Tracer release
rates for these hours were adjusted to hourly average rates. Hourly average
tracer release rates were then used to convert the hourly measured
concentrations to relative concentrations.
Ambient Tracer Concentrations
Hourly average SFS concentrations as measured at each of up to 94
sampler locations are included in the MDB. For the purposes of this
evaluation, the measured concentrations (in the units of parts per trillion)
were first converted to units of g/m3 (using hourly on-site temperature;
and pressure = 905 mb) , and then converted to relative concentration, X/Q
(using the release-time-adjusted hourly average tracer release rates).
Meteorology
Table 3-4 summarizes the meteorological inputs to the complex terrain
models for Cinder Cone Butte. The input data are primarily from the MDB,
but supplemental data were used to provide other required model input data.
The MDB contains hourly averages of measured and derived meteorological
parameters representative of tracer release height (which varied from
experiment to experiment). ERT used a "spline under tension" method of
interpolating tower measurements to release height. Wind speed and
direction are provided as both scalar and vector averages. Also provided
are hourly values of critical streamline height, Froude number (for the
layer between 2m and 150m), and scalar average wind speeds measured at the
10m level of Tower A.
Additional meteorological data were needed for input to some of the
models. Vertical profiles using on-site tower measurements of temperature
and wind speed were needed by the COMPLEX/PFM model to internally calculate
critical streamline height and Froude number. Hourly P-G stability
categories based on the Turner method (CRSTER preprocessor) were developed
from concurrent Boise National Weather Service cloud cover observations and
-23-
-------
i
<*
I J
ro <
O
W M
•J 52
OP O
g^
g
H
H
a)
0
0
a
OQ
Q
in
J
Oi
S
M
rH
Jg
CH
O4
CN
B
rH
X
O
rH r-H
1 1
3 ©
r-H rH
1 1
rH rH
1 1
•cl< CM
1 1
a ©
rH r-H
1 1
3 ©
rH r-H
1 1
a ©
r-H rH
1 1
!Z) ©
rH rH
1 1
D ©
'
^
OS
s
OS
'z
p«
s
r-H
rH
rH
rH rH
1 1
N O4
rH
isi fi
r-H
| Q£
M 2
rH
1 OS
N 2
rH
1 OS
N 2
rH
1
I/I
(N
rH
1
OQ
rH
1
r-H
1
w
r-H
1
OT
r-H
1
DO
M 1-H
i>) M
M M
gg
OS OS
2 2
OS OS
2 2
OS OS
2 2
tM
[ "] |py
r-H
rH 1
j Q
4J rH
(0 1 OS
0) J 2
5-1 01 4J
O 0) 4->
"4-1 01 3
0) (0 TO
01 01 0)
••H G rH CD
OS 0 0) G OS OS
•H OS O 2 2
0) 4-1 O
§rO 0)
rH > iH
rH 3 -H O
Oj U 01 T3
rH 01 G rH
O (0 (0 -H 1 OS
2 O Oi O J 2
rH t>]
1 1
J H
rH
^ §
rH
OS OS OS OS 1 PS
22^2 i 1 2
n
u
1)1
-pi
0)
3 n3
O« 5^
c n)
M a
d
o
•H
rH 4J
(P U
*^O G
O 3
01
4-1
G
0)
G ra • o
O 0) 4-> Ol
•H £ DC K
-PC/3 W
rO U t4
0)0) • O) S
0) S-i t4 4-1 rO
Ol'H -H 0) J
on Q Q e
0 SH
Tj "u TJ S QJ
G C G 0) S
•rl -H 'H G O
^ ^ ^
4-)
U
a
m
C
(0
u
H
a '
5-1 _Q
£>i 3 5n
4-> H 3
•H H
rH
•H N •
0) 5-1 5-1
4-> O 0)
00 X >
c
o
•iH
Ol
(-1
0)
Oi
Ol
•H
Q
0)
i-i T3
3 0)
(0 04
5J C/J
0) M
l^i rO ^o
e c >x
0) -H H
H S T3
(D
01
•H
OS
0)
rH
04
O>
•rH
>1 0)
4-J nC
•H
i-H O>
•H G
(0 X
4-> -H
00 S
Oi
G
X
•H
s
(U
4J
•H
g~
•H
,-4
-O/, _
0)
(-1
3
4-1
(0
*o
0
0)
Oi
0)
OiT)
0)
H
4-1
j^
O>
0)
prj
r-H
(0
O
• H
4-1
•rl
tj
O
G
•H
S
a
•rl
5-1
o
Ol
(U
Q
0)
TO
0)
01
£
(TJ
01
U-l
u
3
01
<4H
o
c c
'» '»
5J T3
0) T3
5-i t4 O S
(0 O 4J
rH -p O
m o' e 4J
U 0)
01 > O t3
m a)
•H -H O O
0) 0) U Oi
a>
cu a> a> 4~>
Q) a; 4-1
rH rH O >-•
Q)0)lJ 1
OJ
M)
a)
C
O
T3
0)
3
C
•H
U
C
O
U
Oi in Q N rH (N ro
S J EH X III
M Oi os OT a a tr>
T3
(U
0)
Oi
c
•H
s
-------
0)
3
C
• H
4J
G
O
a
^
1
m
S
H
ind direction from MDB
ed IMPACT grid
2 4J
ro
U 4J
o o
4-1 S-l
CJ
CU S-i
> o
4-1
_l_)
X! o
Cno
•H cn
0)
x: 01
3
0) G
01 -H
flj S
cu
rH rH
CU 1
OS ©
or each release
M-J
4J
x:
Cn
• H
0)
x;
cu
01
rd
cu
rH
cu
^
O
4-1
4J
CU
r/j
O
4-1
in
CO
cn
CO
CN
S,
o
CN
r-
rH
tn
rH
II
CU
*•
01
0)
3
i— i
rd
H
O
CU
r-l
"rd S3
G <
S-l ^-*
0)
-P r-l
G 1
M C/3
r vmethod (CRSTER preprocessor) using Boise
one Butte 10 m wind speed
cu U
G
tj S-l
3 Q)
H T3
G
e -H
0 0
4-1 T)
G
>i rd
O rd
0) ro
4-1 T3
rd
CJ T3
3
?rH
CU U
ONVRT preprocessor) using av/U-2 from
tal velocity turbulence at release height
o c
•~- 0
N
© -H
D S-i
o
e x:
o
S-i 01
4-1 "H
^1 ^
S-i 0
O
Cn Q)
CU S-l
4J CU
ro X!
u S
O m
CU §
release height
4
4-1
rd
^*«
CN
1
D
>.
O
• •
CQ
]gj
S
0
h-t
release height where awn is vertical
le with Horst correction for non-cosine
4-1 rd
rd O
01
CN CU
1 CJ
^ G
^ cu
i i— i
3 3
— S-l
3
4-1
pp >1
fis '''I
CJ
e o
O rH
S-l CU
JV| ^
cu
01
G
O
04
01
cu
S-l
Q
S
§
4-1
CU
S-i
1 i
rd
S-i
cu
CU
£ S-i
cu cu
4-1 5
o
-P 4-1
x:
O"> S
• H 0
0) S-l
x; 4-4
0) CU
O3 rH
rfl -H
CU 4-4
rH O
CU S-i
06 CU
01
G
rd
• H
•a
rO
U
^*»
rH
O
0
VI
H
M
4-1
O
0)
3
r— 1
rd
>
S-l
e cu
0 O
O 4-1
o
- cu
0 S-i
rH 0)
x:
4-1 S
rd
•a x:
cu cn
••H CU
fj-l J"Tj
1— 1
rd
O
•H
•H
S-l
O
V
*•- «*
S
0
in
r-l
I
e
CN
S-l
CU
G
0)
T3
§
4-1
0
01
CU
3
rH rH
rd I
> :=>
m -o
Q G
S rd
S >•
0 rH
S-i I
4-1 H
0) 4-1
> x:
$-1 -H
CU CU
p x:
T3
cu
S-i
•H
3
O1
cu
S-l
4-1
O
rH (SJ
I I
CD ©
CU
I
00
CN
C/2
rH CN
II
H H
rH CN
II
>J i-J
I
H
O
O
CJ
0)
S-l
c
•H
S
4-1
x:
Cn
•H
cu
S-i
0)
4-1
0)
§
e
cu
01
4-1
Q)
O
CU
w
rO
J
4-1
•H
X!
rO
4-1
W
0)
o
c
0)
$-1
3
H
0)
O
G
cu
S-l
3
H
CU
S-i
3
4-1
rO
S-i
0)
Oi
0)
H
4J
Cn
•H
0)
EC
•O
-25-
-------
10m wind speeds from the MDB. Except for SHORTZ, the effect of mixing
height (for these predominantly night time tracer tests) was precluded
through the use of a 10,000m mixing depth. SHORTZ requires mixing depth
defined as the height above which the vertical turbulence intensity drops
below 0.01. These heights were developed from the on-site tower data.
Plume rise calculations for Cinder Cone Butte were not needed due to the
passive nature of the tracer releases.
-26-
-------
WESTVACO-LUKE DATA BASE
Under an agreement between the Westvaco Corporation, the State of
Maryland, and the U.S. EPA20, ambient air quality and meteorological
measurements were carried out near the Luke Mill in western Maryland, from
December 1, 1979 through November 30, 1981. Data from these measurements
are intended to assist in the development of a rough terrain diffusion model
applicable in the Luke area.21'22 The complex topography of the area is
shown in Figure 3-4. Vertical cross sections of the terrain relative to the
Westvaco stack are presented in Figure 3-5 to give the reader a better feel
for the source-receptor geometry. As can be seen from this figure, all of
the monitors to the southeast of the mill are well above the top of the
stack. Most of the monitor distances from the stack range from 0.75 to 1.5
km. The only exception is the Stony Run monitor (Mo. 10) at 3.4 km
northeast.
The effective stack height (physical height plus plume rise) will vary
significantly depending on operating load and meteorology. For normal
operating loads, the effective stack height can be as low as 250-300 meters
for stable conditions, or strong wind, neutral conditions; and can exceed
1000 meters for light wind, unstable conditions, based on Briggs plume rise
equations.
Three types of data were measured in order to characterize SOz
emissions from the stack, atmospheric transport and dispersion, and ambient
SOz concentrations on the elevated terrain.
Source Data
The Luke Mill utilizes a 190 m stack to vent coal-fired emissions. Flue
gas SOz concentration and temperature were measured continuously by an
emissions monitor.
The source data base for the Westvaco-Luke stack includes sequential
hourly-averaged values of S02 emission rate, temperature, S02 concentra-
tion, and steam flow (monitored continuously at the plant). Table 3-5
presents a summary of the measured stack parameters, averaging times and
units of measure.
Meteorology
Three instrumented towers were used to measure meteorological parameters
(see Figure 3-3). The 100 m Beryl tower had instruments mounted at 10 m and
100 m; the 30 m Luke Hill tower was instrumented at the 10 m and 30 m
levels; and the 100 m Met Tower was instrumented at the 10 m, 50 m, and
100 m levels. The parameters measured at each tower are listed in Table
3-6. These include measurements of horizontal wind speed and direction,
vertical wind speed, intensity of turbulence in each of the three dimensions
with respect to the mean wind, vertical temperature gradient at various
levels and ambient temperature. Additionally: measurements of net
radiation were obtained with a radiometer at the base of the Met Tower; and
an acoustic sounder, operated near the Met Tower, provided mixing depth
values.
-27-
-------
Westernporl
TON^ RUN Monitor
BLOOMIN6TON Monitor
Bloomington 11
^Piedmont
WESTVACO Stacl)
Hampshire BERYL Tower
Key:
• S02 Site
Meteorology
<$ SOo and Met
I .5 o
H H H H H
Figure 3-4. Map of the study area surrounding the Westvaco Luke Mill.
Elevations are in feet above mean sea level (MSL) and the
contour interval is 500 feet (Ref. No. 21).
-28-
-------
UL
c
T3 tc
T3
ro
i/i
(^ 5-
r- C
ro +J
i- E
CD
•»-> r
u ro
00 CTi
bO -I—
O 0)
ro ro
u +->
a;
OJ
>
O
O Q-
O E
ro
>
o •(-> 01
I/I Q-
u
ro O)
> 3 i/i
4->
l/l OJ OJ
O) -C S-
3 I— ro
(1SW)
-------
Cn d)
2 H
Q
W
PS
P
ft£
w
2
O
H
3
n
P
2
O
a,
GO
w
a
0
o
C/2
w
s
M
H
O
3
M
a
£
w
c
0)
!-r-t
1X1
il
1
1
1
I
i
,
w
00 ;
m j
(
^
EH
1^
Q j
O
O -H
t^C rt
> C
EH -H
oo : o
w , c
S i-H
c
W s-
H a
4-
2 a
M 2
Q ;
w !
EH i
OS
o
Oi
w
i u
0)
4->
Q)
S
ro
ro
£n
G
•H
Cn a)
fO S
S-l -rH
d) H
£>
(<3^
Ol
n( 4J
1 "H
: G
|3
»
!
t-i
Q)
4->
O)
g
S-l
CLi
N
O
oo
S-l
3
O
XI
rH
01
cu
rH
•H
e
0
rH
|xj
'O
f 0)
0)
DJ
01
rrj
G
•r-l
!S
G
o
H
4J
fO
S-l
1 1
G
cu
U
G
O
U
|
J_4
3
O
x:
(_t
3
O
r-H
Ol
0)
0)
£_J
Cn
0)
T)
c
0
•H
4J
O
•H
01
G
a>
4-1
G
S-i
3
O
X!
r-H
CO
0)
4-J
0)
Cn
G
•H
•H
e
x:
4->
Qj
0)
T3
Q
OO
H
S?
•H
Cn a)
rO S
S-l -H
0) EH
S-j
3
5
O
3
O
CU
01 4->
N
O
00
y
0
rO
4->
01
5
O
rH
4-1
e
TO
Q)
4-)
01
-30-
-------
TABLE 3-6
INSTRUMENTATION AND PARAMETERS MEASURED ON THE WESTVACO METEOROLOGICAL TOWERS
Site
Instruments*
Parameters
Beryl Tower
Level 0 (10 m)
Level 1 (100 m)
Cup and vane
Propeller anemometer
Temperature probe
Cup and vane
Propeller anemometer
Delta temperature
sensor
SP, DR, IX, IY,
W, IZ
T
SP, DR, IX, IY,
W, IZ
AT(100 m - 10 m)
Luke Hill Tower
Level 0 (10 m)
Level 1 (30 m)
Cup and vane
Propeller anemometer
Delta temperature
sensor
Cup and vane
Propeller anemometer
Delta temperature
sensor
SP, DR, IX, IY,
W, IZ
AT (10 m - 2 m)
SP, DR, IX, IY,
W, IZ
AT (30 m - 10 m)
Met Tower
Level 0 (10 m)
Cup and vane
Propeller anemometer
Temperature probe
Delta temperature
sensor
SP, DR, IX, IY,
W, IZ
T
AT (10 m - 2 m)
* All temperature sensors were mounted in aspirated radiation shields.
(continued on next page)
-31-
-------
TABLE 3-6 (Continued)
Level 1 (50 m) Cup and vane SP, DR, IX, IV
Propeller W, IZ
Level 2 (100 m) Cup and vane SP, DR, IX, IY
Propeller W, IZ
Delta temperature
sensor AT (100 m - 10 m)
Key
W: vertical wind speed
SP: horizontal wind speed
DR: horizontal wind direction
IX: downwind intensity of turbulence
IY: crosswind intensity of turbulence
IZ: vertical intensity of turbulence
T: temperature
AT: temperature difference
-32-
-------
The meteorological data base contains sequential hourly-averaged values of
the parameters mentioned above. Table 3-5 indicates the units that
meteorological data have been reported in.
Ambient Air Quality Data
Monitors for measuring ambient SOz concentrations were established at
11 sites (see Figure 3-3). Eight of these sites were located within plant
property boundaries (from 800 to 1,500 m from the stack) on the high terrain
east and south of the mill. Ground-level elevations at these monitors
exceed the physical stack height. Two additional monitors, Luke Hill and
Stony Run, were located 900 m north-northwest and 3,300 m northeast of the
stack, respectively. Terrain elevations for these two monitors are nearly
level with stack top. The remaining monitor, Bloomington, located 1,500 m
northwest of the stack, is in a valley where the elevation is comparable to
stack base elevation.
The air quality data base reports the sequential hourly-average values
of the S02 concentrations measured at each of the 11 sites. Units
ascribed to each of the variables are given in Table 3-5.
Data Selected for Model Input
The hourly averaged measurements obtained during the Westvaco-Luke field
program were first reduced by ERT and then compiled by H.E. Cramer Company
into a modelers' data base (MDB) for input to the SHORTZ and LUMM*
models.23 Most of the model inputs were obtained from this MDB, however,
some additional data were compiled by TRC in order to meet all technical
requirements of the complex terrain models. The data inputs were restricted
to the second full year of measurements, December 1980 through November 1981.
Source Information
The source parameters, stack location (x, y, z), stack height and stack
diameter, were input as constant values to each of the models. The
following source parameters were input to the models on an hourly basis:
Q Hourly SOz emission rate
T Hourly stack gas exit temperature
VF Hourly stack gas volume flow rate, or
Vs Hourly stack gas exit velocity
In preparing the MDB, any missing values of Q, T, VF, or Vs were
substituted with the last reported value.
*The Luke Mill Model (LUMM) was developed specifically for the Westvaco-Luke
Mill site.
-33-
-------
Air Quality Data
Hourly S02 concentrations at each of the eleven monitors were recorded
in units of parts per trillion. Prior to the evaluation these
concentrations were converted to ug/m3 through the use of hourly on-site
temperatures and an average atmospheric pressure (from the U.S. Standard
Atmosphere) of 0.96 atmosphere. Receptor locations and elevations were
directly input to each of the models, although the horizontal coordinates
first had to be converted to source-oriented distance and direction for
input to RTDM and COMPLEX/PFM.
Model predictions for the Stony Run monitor (labeled receptor No. 10)
could not be made with the IMPACT model due to grid size constraints and
computer limitations. The other IMPACT receptors were located in the center
of the nearest grid cell.
Meteorology
The Westvaco MDB of meteorological data was developed by H.E. Cramer
Company23 to meet the input need? of the SHORTZ and LUMM models. The
specific parameters and their corresponding primary data sources are listed
in Table 3-7. In preparing the MDB, alternate sources of meteorological data
were preferentially ranked for use when data from the primary source were
unavailable or unreliable for ceirtain time periods. These data
substitutions are shown in Table 3-8.
The meteorological data provided in the MDB was supplemented with other
on-site tower data as well as some off-site data needed as input to the
models. Table 3-9 summarizes all of the Westvaco meteorological inputs to
the complex terrain models.
Data from off-site locations ware used to characterize stability
category for four of the models (COMPLEX I, COMPLEX II, COMPLEX/PFM, and
4141), mixing height for most of the models (except IMPACT) and, for
COMPLEX/PFM, profiles of wind speed and temperature. The CRSTER
preprocessor program was utilized with cloud cover data from the nearest
National Weather Service station (Morgantown, WV) in conjunction with
Westvaco 10 m wind speeds (averaged from the three towers) to categorize
stability. Mixing heights were developed from twice daily Pittsburgh mixing
heights as obtained from the National Climatic Center, and interpolated to
hourly values using the various model preprocessor programs. (The
interpolation scheme used to generate hourly mixing heights for COMPLEX/PFM
has been changed since the model was submitted for this study. ) The
Pittsburgh twice per day radiosonde data (TDF5600 tape) were employed along
with Westvaco tower data in the PROFILE program to generate hourly vertical
profiles of wind speed and temperature. These profile data are used by
COMPLEX/PFM to calculate Froude number and critical streamline height.
-34-
-------
TABLE 3-7
PRIMARY HOURLY METEOROLOGICAL INPUTS
INCLUDED IN THE WESTVACO MODELERS' DATA BASE
AS COMPILED BY H.E. CRAMER ASSOCIATES23
Input Parameter
Primary Source
Transport Wind Direction
Reference Level Wind Speed
Wind Profile Exponents
Vertical Potential
Temperature Gradient
Ambient Air Temperature
Lateral and Vertical
Turbulent Intensities
Mixing Depths
Stability Class1
Vertical Wind Direction Shear2
100 m level of Tower 1
30 m level of Tower 2
Based on speed difference between
upper levels of Tower 1 and 2
Based on temperature difference
between the 10 m level of Tower 2
and 100 m level of Tower 1
10 m level of Tower 2
30 m level of Tower 2
A constant value of 1000 m
From vertical turbulent intensity,
10 m level of Tower 2
Direction difference between upper
levels of Tower 1 and 2
Using the stability classification scheme suggested by EPA for a surface
roughness length of 15 centimeters: A < 0.2094 (stability class/
turbulent intensity in rad); B from 0.1746 to 0.2094; C from 0.1362 to
0.1745; D from 0.0874 to 0.1361; E from 0.0419 to 0.0873; and F < 0.0419.
Needed only for the modified version of SHORTZ submitted for evaluation.
-35-
-------
TABLE 3-8
DATA SUBSTITUTIONS USED BY H.E. CRAMER ASSOCIATES
IN DEVELOPING WESTVACO HOURLY METEOROLOGICAL INPUTS23
Input Parameter
Rank of
Parameter Source
Parameter Source
Transport Wind
Direction1
Reference Level
Wind Speed2
1
2
3
4
5
1
2
3
4
5
100
50
10
30
10
30
10
50
100
10
m
m
m
m
m
m
m
m
m
m
Level
Level
Level
Level
Level
Level
Level
Level
Level
Level
of
of
of
of
of
of
of
of
of
of
Tower
Tower
Tower
Tower
Tower
Tower
Tower
Tower
Tower
Tower
No.
No.
No.
No.
No.
No.
No.
No.
No.
No.
1
1
1
2
2
2
1
1
1
2
Vertical Wind-
Direction Shear'
Direction difference between
100 m Level of Tower No. 1 and
30 m Level of Tower No. 2
Direction difference between
50 m Level of Tower No. 1 and
30 m Level of Tower No. 2
Direction difference between
10 m Level of Tower No. 1 and
30 m Level of Tower No. 2
Direction difference between
100 m Level of Tower No. 1 and
10 m Level of Tower No. 2
Direction difference between
50 m Level of Tower No. 1 and
10 m Level of Tower No. 2
(continued on next page)
-36-
-------
TABLE 3-8 (continued)
DATA SUBSTITUTIONS USED BY H.E. CRAMER ASSOCIATES
IN DEVELOPING WESTVACO HOURLY METEOROLOGICAL INPUTS2
Input Parameter
Rank of
Parameter Source
Parameter Source
Direction difference between
10 m Level of Tower No. 1 and
10 m Level of Tower No. 2
Direction difference between
100 m and 10 m levels of Tower
No. 1
Direction difference between
50 m and 10 m levels of Tower
No. 2
Wind-Profile
Exponent4
Based on speed difference between
100 m Level of Tower No. 1 and
30 m Level of Tower No. 2
Based on speed difference between
50 m Level of Tower No. 1 and
30 m Level of Tower No. 2
Based on speed difference between
10 m Level of Tower No. 1 and
30 m Level of Tower No. 2
Based on speed difference between
100 m Level of Tower No. 1 and
10 m Level of Tower No. 2
Based on speed difference between
50 m Level of Tower No. 1 and
10 m Level of Tower No. 2
(continued on next page)
-37-
-------
TABLE 3-8 (continued)
DATA SUBSTITUTIONS USED BY H.E. CRAMER ASSOCIATES
IN DEVELOPING WESTVACO HOURLY METEOROLOGICAL INPUTS23
Input Parameter
Rank of
Parameter Source
Parameter Source
Based on speed difference between
10 m Level of Tower No. 1 and
10 m Level of Tower No. 2
Based on speed difference between
100 m Level of Tower No. 1 and
10 m Level of Tower No. 1
Based on speed difference between
50 m Level of Tower No. 1 and
10 m Level of Tower Mo. 1
Vertical
Potential
Temperature
Gradient3
Ambient Air
Temperature
1
2
3
Based on temperature difference between
100 m Level of Tower No. 1 and
10 m Level of Tower No. 2
Based on temperature difference between
10 m Level of Tower No. 1 and
10 m Level of Tower No. 2
Based on temperature difference between
100 m Level of Tower No. 1 and
10 m Level of Tower No. 1
Based on temperature difference between
30 m Level of Tower No. 2 and
10 m Level of Tower Mo. 2
10 m Level of Tower No. 2
10 m Level of Tower No. 1
10 m Level of Beryl Tower
(continued on next page)
-38-
-------
TABLE 3-8 (continued)
DATA SUBSTITUTIONS USED BY H.E. CRAMER ASSOCIATES
IN DEVELOPING WESTVACO HOURLY METEOROLOGICAL INPUTS23
Input Parameter
Rank of
Parameter Source
Parameter Source
Lateral and
Vertical Turbulent
Intensities
1
2
3
4
5
30 m Level of Tower No. 2
10 m Level of Tower No. 1
50 m Level of Tower No. 1
100 m Level of Tower No. 1
10 m Level of Tower No. 2
Stability Class
1
2
3
4
5
10 m Level of Tower No. 2
10 m Level of Tower No. 1
30 m Level of Tower No. 2
50 m Level of Tower No. 1
100 m Level of Tower No. 1
When no non-variable wind direction was found, the hour was flagged by
setting the wind direction equal to 090 degrees and the mixing depth
equal to 1 meter.
Wind speeds above 0, but less than 1 meter per second, were set equal to
1 meter per second. When all of the wind speeds were calm, the hour was
flagged by setting the wind direction equal to 090 degrees and the mixing
depth equal to 1 meter.
When none of the data substitutions were possible, the wind-direction
shear was set equal to zero.
The wind-profile exponent was set equal to zero when the calculated
exponent was negative or if none of the data substitutions were
possible. The wind profile exponent was not allowed to exceed unity.
When none of the data substitutions were possible, the vertical potential
temperature gradient was set equal to the moist adiabatic value of 0.003
degrees Kelvin per meter.
When no turbulence measurements were available, the lateral and/or
vertical turbulent intensities substituted were climatological values for
the combination of season, wind speed and time-of-day categories.
-39-
-------
ro
W
J
PP
"H
CU
0
^
C*^
X
CO
Q
H
01
in
J
si
M '
1
j
rH 1
^ !
SI
s
Cn!
8!
i-HI
x;
ul
•*
i
£3
ro
1
|~~>
rH
1
N
D
rH
1
P
r-H
1
rH
1
D
rH
1
|""1
rH
i a
© 2
rH
i 01
© 2
rH
i a
© 2
•
n
rsi
>
M
OH
*
OH
2
a
2
OI
2
2
01
2
rH
N
M
rsi
N
M
01
01
2
PH
2
OH
2
01
2
2
rH rH
1 1
H P
rH in
1 1
H P
rH rH
1 1
H P
rH fSI
1 1
r-H r-H
1 1
H P
(N UO
1 1
H P
rH rH
1 1
H P
rH i-H
1 1
H P
rH
EH
Q
rH
H
Q
0£
2
2
01
2
OS
2
01
2
01
2
ro
2 *J
ro I-H
1 1
CO iJ
(N rsi
1 1
CO J
<*
co 2
i— 1 rH
1 1
CO J
r-H ^*
i i
CO J
rH rH
1 1
CO J
i-H rH
1 1
CO J
O« fX
55 5:3
rH (-H
F-t |
Q £)
(X ft
S ^
s 2:
01 01
2 2
vD rsi
1 1
P H
01 OS
2 2
a a
2 2
S-l
cu
4-1
CU
4J S
3 m
CXi Ui
C ro
M a
""O
0)
0)
QJ
CO
TI3
C
•H
^
C
o
•H
u
0)
•H
Q
T3
G
•H
S
(-1
m
0)
CO
.
£_l
•H
Q
T)
G
•H
|j£
4-1
{_!
CD
CD
e
o
e
Q)
r"
r?
Exponents
2
rd
(-1
CU
3
O
a
>i
4-J
•>H
rH
•rH
f}
OJ
4J
CO
^1
3
H
.
N
•H
u
o
a:
X3
!-i
3
EH
•
4J
^4
CU
>
0)
(-1
3
m
0)
a.
e
cu
H
T3
0)
0)
Oj
CO
•a
G
•H
^
CU
X
8"
Cl
o
M M Qi
0)
CU
4-1
X
0)
C
c
o
0)
c
•H
4J
c
o
o
(0
-p
CO
0)
l-i T3
3 CD
4-1 0)
a) QJ
!H CO
0)
CUTS
6 C
0) -H
H 2
O <* rH Oi « CO
in Q
0) TJ
TD G
O 3
S t<
4J
t-i
O
c
o
.1)
QJ
cn
0)
cn
as
0)
a
Cn
C
•rH
X
•H
S
0)
4-)
4->
Cn
a>
rt
u
S-l
O
C
0)
Cn
o>
rH
0>
-40-
-------
^•x.
*O
cu
3
G
•H
4-1
c
o
O
CT*
1
ro
w
cq
H
G|
oi
•H
•pl
cy
•H1
Ui
u
cn
cu
Ql
CU
— - CJ
m
Q T3 J^
S -H cn
-^ S-i S-i
Cn 3
CU 1 JQ —
cn T3 co f-i
fO -H 4-> O
£) e -P 03
•H CO
rg o CM Q)
4J4-1 O
<0 (N i-H CM X! O
T3 T3 1 1 1 4-1 U
0) CU CU CU -H Oi
cn 4-) ? cu
s-i ro cn Cn Cn s-i
CU r-4 G C C t3 &l
rH O -H -H -rH Q)
cu Gi cn co co cn S
•O S-i 3 3 3 S-J t
O cu cu Qi
e 4-1 tM CM CN] £ N^
Gill X
O -H isi N isi co W
CJ rH U
ro •* £ £ £ cu Cu
> co O O O > S
P rH S-l S-l S-l CU O
CO 0) 4-4 4-4 4-4 rH U
Q) > ^
S cu P P P s-i
rn x: x; x; cu w
Sen cn cn £ J
I— 1 -H -rH -H O H
S-l rH CU 0) CU 4-1 DM
4-1 m x: x: x: o
-H DC
•a - cu cu x rH a.
CU CO £ £ CJ (tt
cu s-i 3 3 nj x:
Ol Q) rH rH 4J £ Cn
co 3 CU Oj co O 3
O u O
T3 P O O O 4-1 S-i
c P 4-1 4-1 x;
H ro cn P
^ T3 T3 T3 t^
£ cu cu cu cu n)
Z O 4J 4-1 P CU 4->
4-1 S-i fO ffl f^ CU *Ti
x; 4-1 •— i r— i f— i cn t3
cn o o o
•H T3 CU CU Oj4-( CU
cu cu a) ro n) O T3
X! cu u s-i s-i G
CU 4-1 4-> P Q) O
^ co x x x r-H cn
o cu cu cu -H o
4-1 C rH i— 1 t— 1 O T3
C/) -H 1 1 1 s-i n3
- S S S S cu s-i
•H
S-i
Cn
•o
•H
£
o
4-1
t3
m
a
s
£
0
S-l
4-4
c
0
•H
p
o
cu
tj
•H
T3
TJ
G
u
;
p
x:
Cn
•rH
cu
x:
4-1
^4
a
CO
G
m
H
0)
4-1
fO
1— 1
0
Oi
S-i
0)
4-1
G
•rH
• s
cn
rH
cu
>
0)
r— 1
1 — 1
1— 1
nj
V
CO
S-l
cu
^
o
p
m
e
o
S-l
4-4
c
o
1 1
o
Q)
S-l
•H
T3
tlJ 1 — 1
G rH
•H Q)
S 0
o
4J
•O cu
c >
•H -H
S 4J
rfl
OlrH
O cu
4J !-i
J*
o
nj
4-1
cn
cn
4-1
G
0)
cn
Q)
S-l
cu
cu
S-i
t— 1
1
3
0)
CJ
G
•H
cn
^v
£
r-~
.
CTi
oo
r—|
XM*
s a
o
O 4-J
f->.
•N^ ^S*
r— t U
<2> 03
<^ 4_J
pq cn
p II
rS O
N P
e -o
O V. 4-1
S-l © CU
Cn 'O C/3
Q)
CO
(0
rQ
rsl
S-i
cu
s
o
H
£
03
.
CTv
r~
rH
II
<
N
N
^—
p
x;
Cn
•r-f
0)
x;
(%J
S_l
Q)
5
0
^ cu
CO
£ ro
J3
0
ro ^
CJ
II ro
4-1
rxi cn
H
a
H
£
O
S-i
4-4
•*^
CO
cu
3
r-H
ro
0)
>
•H
4-1
m
cn
cu
C
T3
CU
J>
Sj
cu
CO
1")
o
f.
-M
•H
S
pq
Q 4-J
S 2
-Q
£
0 rH
S-l 1
4-1 CU
^1 S-l
rH O
S-l 4-1
3
O cn
tiC r
o G
O -H
s-i s
CU
Q) g
S-l
cuo
r— 1
ex
W 0
H CJ
W rO
Of, >
U -P
-— co
cu
•0 2:
O
X3 *O
4-1 G
CU ro
£
fO
t^ -1_. *
0) 03
C "^3
i_i
D -0
EH 3
O
0 "o
S-i
4H >
^
^*i
S-l V
O G
Cn 3
cu o
4-1 4-1
m G
CJ (0
cn
1 0
Cu S
>
M
+^
&
D
Cn
C
•H
cn
3
*-^
S-i
O
cn
cn
0)
U
O
S-l
CU
cu
S-i
cu
in
w
£3
tjj
cu
^— *
EH
a
*2
0
u
s
0
4H
*>1
$-1
o
cn
co ro
S-i CJ
0)
S O
O 1
p pi
rom ERT)
temperatures
4-1
£
v O
CN S-l
4H
S-l
CU N
3 -a
o \
H ©
T3
4H
0 G
O
i— i
CU T3
> CU
CU CO
r-H fO
-Q
£
>1
O S-i
m o
Cn
G cu
O -P
(TJ
N U
M
— CJ
1
9- CU
D
4-1
Cn O
G
H Q)
CO rH
3 H
4-1
>i O
S-l S-i
o a
--- cn
OQ CU rH
S m o
CJ -H
£ P
O O s-i
(-4 1 CU
4-1 Q- >
CO
CU
O
p
CO
c
o
cu
Cn
rd
Oj
X
Q)
C
O
T3
CU
G
O
O
I I I I I
r— icM
II
©©
<— i(M
© ©
1
-------
cu
3
C
O
O
ro
W
pq
H
c.
o
S-l
CJ:
CO
CQ
Q
S
O
S-l
M-l
0)
O
C
cu
rH
3
,Q
S_j
3
1 1
rH
H
OS
W
o
t-J
M-l
0)
o
H
G
O
E
0
4-)
fO
CU
t-l
3
CO
rO
CU
E
cu
rj
G
0)
1— 1
3
Q
t-i
3
4-1
rH
rO
O
•H
[ 1
Ui
CU
>
all tower levels merged with Pittsburgh
E
O
U
M-l
W
CU
S-i
3
4->
rO
S-i
cu
CU
E
0)
1 1
M-l
O
Q CU
S ' — i
H
E MH
O 0
fa d,
ILE (COMPLEX/PFM preprocessor)
fl-j
0
OS
ft
c{
Cn
3
O
S-l
(~^
4J
ro
4-)
rO
CU
•a
G
o
CO
o
•H
TD
i-i
re gradient from MDB
3
4-J
rO
t-i
0)
CU
s
cu
4-1
rH
rO
•H
4-1
G
0)
4-1
0
rH
rO
O
•H
I i
S-i
CU
>
h Pittsburgh (Upper) and S-l
ssor) With Pittsburgh (Upper) and S-2
or) with Pittsburgh (Upper) and Westvaco
1 1
•H
5
S-i
0
CO
co
cu
u
0
S-i
cu
0)
cu
OS
H
VI
OS
u
s
0
fa
Q) cn
CJ W
O 0)
t-i O
cu o
0) S-i
t-i CU
Cu cu
£_4
in Q.I
W
JS N
5 EH
J OS
cu o
r/1
H —
> N
o w
0 S
E S
0 0
t-l t-l
fa fa
T3
CD
0
eprocessor) with Pittsburgh (Upper)
(^
cu
s
fa
x^
^3
pq
*-J
ft
s
o
o
CU W T3
01
•o
G
S
S
o
( — 1
t 1
M
fa
o
ft
£
0
S-i
fa
0)
J.4
•H
3
0)
t-l
4-1
o
rH CM rH (N]
N N II
I-H M H H
rH
-------
SECTION 4
STATISTICS APPROACH
The 1980 AMS Woods Hole workshop on model performance evaluation
recommended a comprehensive list of performance measures and statistics for
evaluating air quality models. The workshop recommended that performance
evaluations be based on comparisons of the full set of observed-predicted
data pairs, of the highest observed and predicted concentration per event
(e.g., 1, 3 or 24 hour time period) and of the highest N values (unpaired in
time or space). In addition, comparisons of observed and predicted
concentrations are to be carried out on data subsets representing individual
monitoring stations or selected meteorological conditions.
TRC and EPA reviewed the workshop report and formulated a statistical
approach for the rural model evaluations3 based on workshop
recommendations. The approach was modified for the urban model
evaluations5, primarily to reduce the volume of information by eliminating
redundant performance measures and statistics. Additional revisions as
appropriate to the complex terrain models were also made, and the
statistical approach followed for this evaluation is described below.
DATA SETS FOR COMPARISON OF OBSERVED AND PREDICTED CONCENTRATIONS
The data sets listed in Table 4-1 represent the different types of
comparisons recommended by the AMS workshop. In each instance, comparisons
were recommended for the basic 1-hour unit for model predictions and also
for 3-hour and 24-hour averaging times. The numbering scheme in the table
is derived from a summary prepared by William Cox of EPA of the data sets
and statistics recommended by the AMS workshop.
To compare observed and predicted air guality values on a common basis,
it is necessary to account for background concentration, i.e., contributions
to measured air guality from sources whose impact is not modeled. This
concern does not arise for the Cinder Cone Butte tracer study, since other
sources of SFS are non-existent. The effects of background in the
Westvaco monitored data were removed from measured S02 concentrations
before statistical comparisons were made between observed and predicted
concentrations. The uncertainty of plume transport in complex terrain poses
an uncertainty in attempting to define a method for the determination of
background concentrations. High observed concentrations in the Westvaco
network tended to occur with light and variable winds which can result in
-43-
-------
in re
O
h-
3
^
UJ
-i
UJ
a
y
ac
a
CO
i—
UJ
CO
I—
a
u.
o
"^m
at
^T*
2
ID
co
r»-
1
*
UJ
_J
ca
H-
CU —
3
re E
4J
T3
CU C
4-1 •»-
<_>
•r- X>
"O CU
cu v.
U -r- •
a. re ~
CX V)
•a 4->
C • OI
re c vi
o
CU 4-1 4->
> re re
U 4-1 T3
cu vi
1/1 •—
0 CU
cu •* o
i- 01
re i—
a. re re
E 4->
<_> re 4J
^_
|
CO
VI
c
0
re
u
4-> VI
C C
cu o
U VI
C -r-
o u
<_> re
i a.
<— E
< U
ai
•a
cu
i_
0) •<-
3 re
i— a.
re *-i ^-*
> VI
CU 4-1
CU Ol CU
U JZ CU
cu
o -r- re
2 vi
4->
CU C O
Ol >
•<- cu c
£ O
0) U -13
re cu •!-
Q. -o
E V- cu
O O i-
^^
p.
1
c
o
4-1
re
4-1
c vi
cu c
u o
C VI
o «-
<_) s-
re
S/ (-L
re S
Q. U
VI
cu
3
re
•o
cu
4-1
u
.^
*o
cu
1.
a.
•0
C
rt
*o
lo
cu
VI
o
cu
u
re
u
IM
i
CQ
j—*
C
O
re
u
o
0
c
I
! i
c
Ul
4V
C CU
•t- Ul
•a re
ai 4-1
u re
re
a. u -~
O I/I
• (t_ CU
•O 4J
o at •>-
•1- 4-1 VI
i- re
Q) *r* OI
Ok W C
a.~-
gj o t-
E i- o
*J Q. •»•
re c
c o
> o
OI-S CU
<«-
re cu
C-) .C
t- re 4->
<£ w'5
aj g wi
> £ 4)
t- J=
O -£Z Ol
-o re .c
o cu
cu
a)
oi re 4->
•1- CU •!-
CU CU C
1- JZ O
re 4J -i-
§4-1
^_ (^
O 4-1
0<4- VI
^wt
fNJ
1
VI
cu
3-00)
r- C E
re re •«-
> 4-*
T,i!x
CU •-• J3
4-1 4->
o -a
••- c c
•o i- re
cu
I.T3 ^
CX CU 4-1
u cu
c re
re a. re
4-1
"o * re
cu vi -o
> c
1- O 0)
cu •<- c
I/I 4J O
O 4J ""*
QJ O
IM r^ •**
re r— 4-1
cx re re
E u
u re®
M
1
CD
.
C
O
01 —
4-1 re
{j
4-1 o
re r—
s- c
0) *"
5^ "O
01 &-
jr ••-
4J re
Q.
a
<4_ C
O
C -i-
O 4-1 — .
•r- re CU
O VI •«-
•r- 4J
T3 0)
a) E 4-*
i- (3 o
.
x
re
•a
0
01
3
cfl
>
CU
>
t.
QJ
•g
1
X
re
c
cu
m
a.
o
re
CO
1
2
I/I
C
U- O X
O -f- .0
4-1
I/I •*- "O
4-1 -a c
cu c re
Ul O
-0 U "-
3 -0
I/I i— CU
re cu
u u a.
o •«- ui
u- oi
O "O
^t ^» ^
30-i-
"i32
* Q| "^3
^4J C
t E ?s
CD >\ rt
'^ >» jj -a
(/) "° r- O.
(O trt -^ O
•4-1 -O
§C rQ d)
Q* •*-* 6
CO CU^S
5
I
m
•a
Ol CU
c t-
c u re
•r- -t- O.
4J re c
4J c a.-^ •-
at ui ai ~a E
f- CU O CU -^
01 u re u 4-*
•r— CX CX 'r*
j= U s. a"" "cu
4-1 vi o - "o S
S 3 CU 0) V- 4J
r- E t- •«-
re > 4J re cx c
cu a. re
X ^O i > c • ••
0) C 3 C CU
cu 4-1 cu o o
4-1 f- 01 r- 4-1 O.
-o<»- r- re vi
O S- -f VI C
i— 1- CU O
m CU 3 «-
> Ul t/> r- 4J
3 .a re re
-a 4-> o o > o
CO Ul •*- C3
4.1 CU S- >• TO 1—
c> jz re -— * ui a>
•r" Ol > re CM 4J i-
•a i- <•» M O o
ai iz s- i — •<-
u. o < -o cu
Cl J= -r^ vi 1/1
E: X - JZ 0» vi
-a ui jz cu
cu cu cu oi cu on—
v.. £ > c i- ••- -a
ft) 4-1 U '^" n< f" t—
CL ai u a. re
E i- ui •— g -o 01
ci o ja re o c cu
<_) u- o Q. o re u
£} re
i_ i
-44-
•o -
CU ui
> 01
!_ 3
CUi—
1/1 re
a >
o
•o
^ cu
CN O
II t-
•~T3
Z 0)
Ik*
u.
0
c
O)r-
> re
H- 4->
01 O
4-1
re
•<
o*^
- C
cu o
•^ 4-J
4J Q.4J re
VI O —»
co za- o •
OI4-I
.i- Ul
-C CU
cu oi
U -t-
c8£
t-o
c
u re
2
^^
i
J^X
ja "^
Ul TJ
VI C C
4-1 o re
OJ 4-1, —
> -i- -O
cu -e cu
C cu
u- o cx
o o vi
-------
high measured concentrations at monitors located 180° upwind from the stack,
using the measured wind direction. It was assumed that the background
contribution is evenly distributed over the study area, and can be
represented by the lowest measured concentration in the network each hour.
Since observed concentrations of less than .005 parts per million (ppm) were
set to the minimum instrument detection level of .005 ppm, background
concentrations for these hours were set to .0025 ppm. The Luke Mill of
Westvaco is relatively isolated from other point sources of S02, so this
background method should be effective. In Table 4-1, and in the discussions
that follow, "observed value" denotes a measured concentration minus
background.
For many hours during the year at Westvaco, none of the monitoring
stations experienced significant observed or predicted S02 impact. These
hours of effectively zero observed and zero predicted impact are relatively
uninteresting for the evaluation of air quality models for regulatory
purposes. Including those hours in statistical analyses adds to the
computational burden and tends to dilute the model performance results from
hours with significant impact. Consequently, threshold values were imposed
to screen the data base for statistical analyses. If, for a given time
period, both the observed concentration and the predicted concentration at a
station were below the threshold, that data pair was excluded from further
analysis. A threshold value of 25 ug/m3 was used for 1-hour and 3-hour
averages, and a value of 5 ug/m3 was used for 24-hour averages.
Threshold checks were not imposed on the Cinder Cone Butte data.
Peak Concentrations
For peak concentrations, comparisons are made to determine model
performance both on an unpaired basis and for various pairings in time and
space. The first two items in Table 4-1 represent a comparison of the
highest observed and highest predicted concentrations, paired in time (A-l)
and paired in location (A-2). For the Westvaco data set, these two
comparisons provide quite different measures of performance since the number
of events is large (1 year represents 365 days or 8,760 hours) while there
are only 11 stations. Meanwhile for the Cinder Cone Butte data set, the
number of events is relatively small (104 hours) while the number of
stations (94) is relatively large. An additional (A-2) data set was added
for the complex terrain evaluation, representing the second-highest values
observed and predicted at each station.
Item A-3a represents a comparison of the highest observed concentration
values, regardless of time or space, and predicted values representing
different time and space pairing. Item A-3b is directly analogous to A-3a,
but starts from the highest predicted value. Results for data sets (A-3a)
and (A-3b) were relatively uninformative for the rural evaluation. These
sets were therefore dropped from subsequent evaluations.
Items A-4 and A-5 involve comparisons of the "N" highest observed and
predicted values, unpaired in time or space. The AMS workshop recommended
that such comparisons be based on the upper 2 to 5 percent of
concentrations, rather than on one or two extreme values. As an alternative
-45-
-------
to the percentile approach, TRC recommended using a small number (N=25)
which would more appropriately represent the set of highest observed and
predicted values, while still providing a statistical basis for establishing
confidence limits. On a percentage basis, 25 values represent roughly 7
percent of the 365 24-hour values in a year, about 1 percent of the 3-hour
values, and about 0.3 percent of the 1-hour values.
Air quality data often exhibit spatial and temporal correlation,
particularly over time periods of a few hours. For 1-hour and 3-hour
periods, the highest 25 values were screened to eliminate cases with two or
more high values from the same period, or with two consecutive high values
(Westvaco only) at the same location. This screening is intended to reduce
the effects of auto-correlation and to avoid double-counting a single
event. For 24-hour averaging periods, less correlation is expected, and
this screening was not included.
Comparisons of the highest 25 observed and predicted values were
performed for all stations combined (A-4a), for each station individually
(A-4b) and for subsets of events corresponding to selected source-receptor
geometry and to selected meteorological conditions (A-5). The subsets
selected for the evaluation of each data base are described in more detail
later in this section.
Comparisons of All Concentrations
In addition to peak concentration analyses, the AMS workshop recommended
that comparisons be made based upon all observed and predicted concentration
values. Table 4-1 lists three items of this type. Item B-l is the
comparison of observed and predicted values at a given monitoring station
(for all data pairs above the threshold values). Item B-3 represents
comparisons based on the set of values from all 11 stations combined. Item
B-4 represents subsets of B-3. The same criteria described for item A-5
above (for defining subsets of source-receptor geometry and meteorology)
were used to define subsets for comparisons of all concentrations.
STATISTICAL ANALYSIS OF MODEL PERFORMANCE
The AMS workshop report recommended two somewhat different lists of
performance measures for comparing model predictions with observed air
quality, one appropriate for data sets representing pairs of observed and
predicted values, the other appropriate for unpaired data sets. Paired data
sets provide a means for assessing how well a model predicts on an
event-by-event basis, while unpaired sets do not. Table 4-2 summarizes the
basic list of performance measures, and the statistical methods recommended
for establishing confidence limits on each measure. At the head of each
column (Paired and Unpaired) are listed the data sets from Table 4-1 to
which each list of measures and statistical methods has been applied.
-46-
-------
TABLE 4-2. STATISTICAL ESTIMATORS AND BASIS FOR CONFIDENCE
LIMITS ON PERFORMANCE MEASURES
Performance
Measure
Estimator
Bias
Noise/Scatter Variance
Gross
variability
Average
absolute
residual
Basis for Confidence Interval
Paired Comparison
(Sets A-l, A-2,
B-l, B-3, B-4)*
Average One sample "t," with
adjustment for serial
correlation
Median Wilcoxon match pair
Chi-squared test
on variance of
residuals
None
Unpaired Comparison
(Sets A-l, A-4, A-5,
B-l, B-3, B-4)
Two sample "t"
Mann-Whitney
None
F test on variance
ratio
Not applicable
Not applicable
Correlation Pearson Fisher "z"
correlation
coefficient
Frequency
distribution
comparison
Maximum
difference
between
two
cumulative
distribution
functions
Not Applicable
These sets refer to Table 4-1.
Not applicable
Kolmogorov-Smirnov
(K-S) test on
f (obs.) vs.
f (pred.)
-47-
-------
The data sets from item A-l (highest observed and predicted values for
each event) and from items B-l, B-3, and B-4 all represent observed and
predicted values paired in time. For these sets, statistical analyses based
on the residual (i.e., the differences between each pair of observed and
predicted values) are appropriate for measuring model performance: If the
time pairing for these data sets is ignored, however, it is also possible to
assess model performance (in aggregate) by comparing the features of the
composite set of all observed values to those of the predicted values.
Consequently, both paired and unpaired comparisons were recommended by the
AMS workshop for these data sets. Data sets representing comparisons of the
highest 25 values, regardless of time or space, provide no basis for paired
analysis. For these sets (A-4, A-5), only unpaired comparisons were
performed. Item A-2 represents comparison of the single highest observed
and predicted values from each of the N stations. Only the paired
comparison performance measures were computed for this case. Mo statistics
were computed for the single-value comparisons in item A-3.
For paired comparisons, as noted above, the performance measures are
based on an analysis of residuals. Model bias is indicated by the average
and/or the median residual, with a va,.ue of zero representing no bias. The
characteristic magnitude of the residuals is an indicator of the scatter
betv/een observed and predicted values on an event-by-event basis. Three
measures of noise or scatter were computed:
« Variance 1 V (d, - d)2
• Gross variability 1 v~^
N L-i
• Average absolute residual 1 ^~~* (dj
N
where di is the residual (observed minus predicted) for data pair i, d is
the average residual, and N is the number of data pairs. The correlation of
paired observed and predicted values is measured by the Pearson correlation
coefficient.
For unpaired comparisons, the list of performance measures is somewhat
shorter. Model bias is indicated by the difference between the average (or
median) observed value and the average (median) predicted value. A ratio' of
the variances of the observed and predicted values is provided to indicate
whether the distribution of values in the two data sets is comparable.
Similarly, the frequency distribution of observed values is compared with
that for predicted values.
-48-
-------
Standard statistical methods have been used to estimate confidence
limits for each of the performance measures. Discussion of the statistical
procedures may be found in most statistics textbooks. For parametric
procedures, the reader is referred to Snedecor and Cochran (1967), 24 while
for nonparametric procedures Hollander and Wolfe (1973)25 provide an
appropriate description.
For paired comparisons, the confidence interval on the average residual
can be estimated using the one-sample t test. This parametric test
incorporates the assumption that the residuals follow a normal distribution,
but for large N departures from normality are not critical. Serial
correlation can affect results significantly, however, since the number of
"independent events" will be overestimated and the calculated variance may
understate the magnitude of the actual random error component. The AMS
workshop recommended the adjustment of confidence limits for serial
correlation. A method described by Hirtzel and Quon (1981)2S has been
used to adjust the confidence interval from the one-sample t test. The
interval given by the standard one-sample t test is multiplied by the factor
[ (l+r)/(l-r)]l/z, where r is the lag-one autocorrelation coefficient of
the residuals.
An analogous nonparametric indicator of model bias is the median
residual. The statistical method for estimating a confidence interval on
the median residual is provided by the Wilcoxon matched-pairs test. Mo
straightforward method of adjusting the confidence intervals from the
Wilcoxon test for serial correlation has been identified.
A confidence interval for the variance of the residuals is calculated
using a chi-squared test. Mo adjustment was made for serial correlation.
No standard method is available for estimating confidence intervals for the
gross variability or average absolute deviation measures. For the Pearson
correlation coefficient, the Fisher z test provides a method of estimating
the confidence interval.
Comparison of two cumulative distribution functions is accomplished
using the Kolmogorov-Smirnov (K-S) test. For this test, the two
distribution functions are compared across the full range of concentration
(or residual) values, and the maximum frequency difference between the two
functions is identified.
For unpaired comparisons, two bias measures are computed. The average
of the observed values is compared with the average of the predicted
values. The confidence interval on the difference of the averages is
estimated with a two-sample t test. The median difference is also computed,
and the confidence interval is estimated using the Mann-Whitney
nonparametric test.
The variance of observed values is compared with the variance of
predicted values for unpaired data sets. The performance measure is tne
ratio of the variances; the F test provides confidence limits on the ratio.
-49-
-------
The frequency distribution comparison for unpaired data sets provides a
measure of the difference between the observed and predicted distribution
functions. The K-S test is again used to assess the statistical
significance of the maximum frequency difference.
Sta11stical Measures for the Full Westvaco Data Set
For Westvaco, the full data set represents hourly observed and predicted
concentrations at each receptor and hourly associated variables (for subset
analysis) for a one-year period of record. The specific performance
measures and statistics calculated for each of the unpaired and paired data
sets are summarized in Table 4-3 and 4-4. The notation for identifying data
sets corresponds to that employed in Table 4-1.
The routine monitoring network at Westvaco, with relatively few stations
and a very large number of events lends itself to an evaluation approach
focussed on peak values (unpaired in time or location), analysis by station,
and analysis for meteorological subsets (by stability and wind speed). The
added factor of terrain elevation is reflected in station-by-station
results. The performance evaluation considers 1-hour, 3-hour, and 24-hour
averaging times. In complex terrain, peak impacts are commonly thought to
be associated with stable conditions. Four stability categories, therefore,
have been selected: unstable (Class; A, B, and C); neutral (Class D);
slightly stable (Class E); and stable (Class F).
Table 4-3 indicates that the full set of estimators and confidence
interval calculations will be provided for the 25 highest values over all
stations and events (A-4a), but only a partial set of measures is provided
by station (A-4b) or for subsets by meteorology (A-5).
For the paired data sets (Table 4-4), the highest priority is placed on
comparisons of the highest value per station (A-2) and all events paired in
tune and location (B-3). The remaining data sets received a more- limited
analysis.
IMPACT Model: Analysis of Select Hours: for Westvaco
The IMPACT model runs with Westvaco data were limited to selected
periods in order to maintain reasonable computer costs. As previously
discussed, the primary basis for evaluating the models (except IMPACT) with
the Westvaco data is the set of performance statistics based on the full
year of Westvaco data. In order to provide some basis for comparing the
performance of the IMPACT model and the other complex terrain models,
performance statistics have been prepared for the other models based on the
same subset of hours selected for evaluation of the IMPACT model.
Based on benchmark computer costs, it was estimated that approximately
500 hours could be simulated with the IMPACT model. Selection of this many
hours allowed the consideration of 21-hour as well as 3-hour and 1-hour
averaging periods. However, the number of 24-hour periods, restricted to
about twenty is marginal from a statistical standpoint. The selection of
twenty 24-hour periods did ensure that a large number of 1-hour and 3-hour
-50-
-------
C/3
a
0)
u
O M
H
4J O
0)
O «—
fl C
m a) M
•H Sj
•a o) o
0) 4J ^*
S 4-1 S-
cn
0) CO
o cn
C rt
0) t-j
t-i CD
Q) >
4-4 <
'^W
•H 4-1
Q O
T3
CU 0)
Cn 4-1
to o
t( .H
CO 13
> cu
< !-•
a.
cu cu
CT >
m ij
u cu
cu 01
> J3
<: o
o
01 01
C 4-1
O C
•H CO
4J > ~
nj co nj
4-1 «3«
01 r-\ I
rH <
i-l (0 •*"
1—I
<
01
•^ 4-1
C C
O CU
•H > -.
4J CU J3
ca
>,
PQ
4J
0)
OT -H -->
4J T3 in
0) C I
03 O <
£> U «-
-51-
cu
01
o
C
0)
-a
o
CJ
o
».
PQ
< Q pa t,
CO 01 01 01
01 01 01 01
nJ nJ fO (^
^-1 ^H i—I i—I
O O O O
01
O
S-l
en
01
01
m
r-t
a
03
4J
00
0)
-a
r-1
o
01
4-1
0)
01
XI
OQ
01
04
o
$-1
en
•a
(V
a
T3
01
u
1
rsi
01
0)
S
•H
cn
0)
-------
Q£
o
fa
Q
W
§
D 2
< H
O <
CO O
U J
M
H OS
CO O
M
H W
H M
CO H
Q «3
CO W
W OS
OS M
E-3 i^«
CO ft
w co
Jj-j
W CO
O
2 /<
o Q
&4 O
OS O
U <
Q. ^^
H
CO
• w
1
W
«
H
-a
Q» G
M-l U O
O -H 0) -H
a) 6 4->
01 a«"-t re -s
4J 4J O ^*
0) 01 O 1
03 4-1 C i-H OQ
.Q C -H —
30) -O
CO > C
0) (t)
0)
e
01 -r-1
-U 4J
G -C G -H
0) U O G r-l
> nl -H -H |
cu i
-U >H J3 C
01 4-> O 'H
0) (TJ T3 -H (N
^ 4-1 O) 4-1 1
cn 01 ti (0 <
-H -H O ^*
ffi u m 0
0) O4 f~4
04
4J C
4-> C -H
01 0) •-••
0» > 13 0) i-H
X 0) 0) g 1
CT S-i ••-• <<
-H U -H 4J *-*
X 0) (0
O4 Ol
**> **> **^
•^ '•^ ^*v
"-*
M
*
O
**"> **> *"*>
M
•
o
**"> "*"> **">
^
M
•
O
*""> ^> "*^
0)
U
03 T3 G
4-10) 0)
G > u
0) S-i 0)
> 0) 44
0) 01 4-1
4-4 O TJ
O
0) 0)
ti o* CT>
a) m a)
J3 U U
S 0) 0)
3 > >
S < <
^
X
^
""•^
^
O4
/\
O
^
c
o
•H
4-1
U
m
b
X X
X
X
X X
X
X
O
o
o
M
6
o
o
•~> x x
o
X X
X
01
0)
•r-l
o
G
(0
Q4
0)
U
o
01
O
01
4-1
G
0)
•H
o
•H
4-1
4^
0)
O
O
C
o
nl
•—i
0)
O
O
G f^
O S
03 (4
ij m
(T] 0)
CU Q4
a. co
G
o
01
•H
U
(TJ
04
§
o
0)
(0
•rH
S-l
m
O 0)
C U
Q) G
3 0)
O) O)
!-4 4-1
4-1 4-1
• H
§ ^
£
•H
X
s
(fl
!-l
0)
4J
G
a>
O
O)
T5
• ^
4-4
G
O
O
M
CJ
-------
periods were modeled. Not all these periods, however, involve significant
observed and predicted impact at monitor locations. The following selection
criteria were followed:
1. The days with the six highest observed concentrations at each of
the 10 monitors to be modeled with the IMPACT model were
identified.
2. 20 days (480 hours) were randomly selected from (1).
3. The 3-hour periods with the six highest observed concentrations at
each of the 10 monitors were identified as in (1).
4. 20 3-hour periods were randomly selected from (3).
Model results for the hours selected in (2) and (4) above were then
analyzed. Implementation of the above criteria resulted in a data set
containing 480 hours. Performance statistics for this limited data set
include only a portion of the measures listed in Tables 4-3 and 4-4. For
unpaired (25 highest) analysis, the all stations/all events case was
examined for the 1-hour averaging period. For data sets paired in time or
location (Table 4-4), statistics for A-l, A-2, B-l, and B-3 were generated
for the 1-hour average, but subsets of events (B-4) were not considered.
Statistical Measures for the Cinder Cone Butte Data Set
The tracer experiments at Cinder Cone Butte represent a different type
of data set for evaluating model performance. The number of concentration
measurements per hour is much greater, and the number of events much fewer,
than those from a long-term, continuous monitoring program. The greater
spatial density of measurements make comparisons between observed and
predicted values event-by-event (paired in time) more informative for a
tracer data set. Analyses for individual hours and/or individual tests are
also feasible. By contrast, the number of monitors and the use of movable
arrays make it difficult to perform station-by-station analyses.
The evaluation for Cinder Cone Butte shifts the emphasis toward
event-by-event analysis. The Cinder Cone Butte data sets representing the
25 highest observed and predicted values (unpaired) received similar
treatment to that for Westvaco, as indicated by Table 4-5. No analysis by
station has been attempted. Subset analysis included station groups defined
by receptor terrain elevation (relative to release height). Receptors were
grouped into four categories: below release height; at release height
(within 10 meters); between 10 and 30 meters above release height; and more
than 30 meters above release height.
The paired data sets analysis for Cinder Cone Butte, summarized in Table
4-6, does not include any analyses by station. The highest-by-event data
set (A-l) has been analyzed for the full set of paired performance measures,
as was the "all pairs" (B-3) data set.
-53-
-------
w
ac
H
g
Cn
Q
W C/Q
H H
< W
i— ] C/2
P
o <
J H
**C "^
O Q
M -^
O H
M C/3
H W
M O
< X
H
tn in
(N
Q ~-
Q
i
|
|
*
^-^
01 i
a) a) ; M
O tn i
cm o
Q) !-l : ^-
iJ 0) ^>
CU >
4-1 i< i
4-1 i
•r-( 4-1 !
Q O |
,
0) Q)
Cn 4-1
(X) U
SJ -H i •->
a) TU
> Q)
!
i rO ti i
fcj Q) ! %
CO 01
> X3 i
^->.
^1
a c — •
S-i O IT)
01 O C -H I
4J 4J -H 4-> <;
CU Oi m fXJ ^*
01 CU U >
xi o tj cu
3 0) CU rH
V3 U 4-1 Q)
m
u
cu
4-1
c
cu
o
fi
cu
•a
4-4
c
o
o
o
*
>1
1—I
o
o
CU
fO
a>
-54-
-------
W
X
H
a
o
ptl *^A
o
Q M
W H
H ff.
38
o
o
co a
O M
M H
H
CO S
M M
(
XI G -H
3 CU TD
CO > G
cu rO
G
fO -H T3 G
-P C O ^
fO T3 fO -H ro
tj CU -PI
U d) ro PQ
.-H -H £ 0 ^
i — 1 03 -H O
T3 CU OJ 03
J5 cu cu e en r-i
Cn 1-4 • "~( ^Q |
• H !H -H J-l 3 <
ffi cu o) en —
Pi PI
-P G
-P G -H
W CU '~~
0) > TD O) i-H
J3 O) CU 6 1
tn s-i -H
0)
m
0
!H
Q)
XI
s
0)
^>
l_l
cu
en
XI
O
cu
tji
0)
S-i
0)
£>
1^4
MH
•tH
•a
cu
CJi
03
S-i
0)
£>
H
ft)
Qj
S
O
CJ
cu
u
G
03
H
S-i
0)
55
^3
CP
0)
1-1
<4H
e
-3
e
rH
X
03
S
•11
0
c
3)
4)
'-J-t
t| |
r-1
T3
T)
cu
o
o
o
-------
Since the tracer releases for the Cinder Cone Butte experiments were
generally shorter than three hours in duration (less than one hour in many
instances), it was not feasible to evaluate model performance for averaging
times as long as three hours. Analyses were limited to one-hour averaging
periods.
For the Cinder Cone Butte data base, analysis of the unpaired 25 highest
values for subsets of events was not attempted. Subset analysis was instead
performed for the "highest-by-event" paired data set. The recommendation to
replace "highest 25" subsets with "highest paired in time" subsets for
Cinder Cone Butte reflects two considerations. First, the total number of
events is small, and some subsets may contain fewer than 25 hours. Second,
the sampler density provides relatively good estimates of peak values for
all hours, and the experimental periods were generally selected to provide
high impact at receptors.
Paired data subsets for Cinder Cone Butte (for 1-hour periods) were
defined by stability group, wind speed, release hight and release distance.
Stability group wind speed categories for Cinder Cone Butte are the same as
those for Westvaco. Three release height categories (relative to the base
elevation of 945 meters) were used: below 16 meters; between 16 and 26
meters; and above 26 meters. Two release distance categories were used:
less than or greater than 900 meters from the release point to the top of
the butte.
-56-
-------
SECTION 5
MODEL PERFORMANCE RESULTS
Statistics comparing observed and predicted concentrations have been
generated for each of the eight complex terrain models and two data bases.
The results are presented by data set for the Westvaco full year model runs,
Westvaco-IMPACT select hours, and Cinder Cone Butte tracer tests. Each data
set is organized into four types of tables providing statistics for 25
highest values, highest concentrations by station (except for Cinder Cone
Butte), highest concentrations by event and comparisons of all observed and
predicted concentrations paired in space and time. Tables of statistical
subsets by meteorology and source-receptor geometry are provided in
Appendices B and C.
WESTVACO FULL YEAR RESULTS
The full year of Westvaco data was run for all of the models except
IMPACT. Statistical measures were produced for three averaging times
(1-hour, 3-hour, and 24 hours) for each of the seven complex terrain models.
Statistics for 25 Highest Values
Statistics for the set of 25 highest observed and 25 highest predicted
1-hour average SOz concentrations are presented for each model in Table
5-1. The first two columns of results are simply the average of the 25
highest observed values and the average of the 25 highest predicted values
for each data set. The first performance measure, presented in column three
is the difference between the two averages. A positive value implies model
underprediction. In parentheses under the calculated differences are 95
percent confidence intervals, determined by using the two-sample Students' t
test. These results show that all seven of the complex terrain models
overpredicted the 25 highest values at the 95 percent confidence level. The
largest overprediction, by a factor of 20, is by COMPLEX II, and the
smallest overprediction, by a factor of 1.6, is by RTDM.
The second performance measure is the median difference (313th highest
value) between all 625 possible pairings of the 25 highest observed and
predicted concentrations. The 95 percent confidence interval is determined
with the nonparametric Mann-Whitney test. Results for the median difference
are very similar to those for the difference of averages.
-57-
-------
U C'
c> t/> I
«r tx. a.
-< a ^
or a c"
«t 5 cl
i* O O
-> 00
O O
• o
to o
• o
C3 O
• o
LxJ O
U bj
y u. oi
u- o
•o o
I fO
•-• Kl
IX
I
Ci]
i> ? <: ^
* *- CE:
— ^oo
UJ UJ I
ll > (/>
^i r\i
t tr.
O f-
«t u
O > U
< a; D
O
U
O
I
-58-
-------
The third performance measure is the variance comparison. The variance
of the 25 highest observed values was divided by the variance of the 25
highest predictions. The F test was used to calculate the 95 percent
confidence levels for these comparisons. Results indicate that, for all of
the models, the scatter of 25 highest model predicted concentrations is much
larger than the scatter of 25 highest observations.
The last performance measure presented in Table 5-1 is the frequency
distribution comparison. The cumulative distribution function f(C)
represents the fraction of the data set (in this case, the fraction of 25
data points) with concentration values less than or equal to C. The value
presented in this column is the largest absolute difference between the
observed and predicted distribution functions (for the same concentration
value) obtained when the two functions are compared for all concentration
values. The value given in parentheses is the maximum difference which is
significantly different from zero, at a 95 percent confidence level, as
given by the Kolmogorov-Smirnov (K-S) test. This confidence interval is a
function of the number of cases. The value is, therefore, the same (0.385)
for all models, since the number of cases is always 25. The results for the
comparison of maximum frequency differences (1.00 for six of the seven
models) indicate there is no overlap between the distributions of 25 highest
observations and 25 highest predictions.
Table 5-2 is presented to exemplify how comparisons of the 25 highest
observations and 25 highest predictions selected by monitoring station and
for various meteorological subsets reveal more detailed aspects of model
performance. Results for the COMPLEX I model are depicted in Table 5-2,
while results for all of the models are presented in Appendix B.
Comparisons of median difference and frequency distributions have been
eliminated from the subset tables since they don't provide a great deal of
additional information.
While reviewing Table 5-2, the reader should notice from Figure 3-3 that
the largest overpredictions by COMPLEX I occur at the close-in monitors
(stations 1, 3, 4, and 6) on the ridge southwest of the stack; while
underpredictions occur at the two monitors (stations 2 and 11) located
northwest of the plant. Overpredictions of the highest 25 concentrations
occur on average for all wind speed categories and for stable conditions.
Neutral and unstable conditions, however, result in underpredictions of the
highest 25 concentrations by COMPLEX I. The average of the 25 highest
predictions for stability D is only 8 |ag/m3, while the observed average
is 1517 ug/mj, resulting in an extremely large variance ratio for this
subset. For COMPLEX I (and some other models as well) this result is
probably due to the half-height terrain treatment (lifting the plume over
terrain) combined with small values of the vertical dispersion coefficients
for stability D.
Comparisons of the highest 25 observed and predicted concentrations for
data sets of 3-hour and 24-hour averaging periods are presented in Table 5-3
and 5-4. Subset tables for 3-hour and 24-hour averaging periods are
-59-
-------
u1 r> o
O GO LJ
Z <—* Ct,
< or Q.
•-« «s *•*
tr a to
«r s: cr
> o o
vOO «-• K>
eg o eg
c. a- cs
fO O «-• CM
tv in *-i c-
o
UJ
tr *«•
Q *~i **
* 2
C * O
«t X. t-
O D U t-'
LJ.J »" ri ui
UJ CO ~
u LJ a
LU «r a;
oc a: a.
^- o
C u ~
o
tr
«•
CM
c\ic\jf\jOf^rr*Of\ic\j i
I if) I*- lO
M C i .Jl
_J CO •-'I
LJ UJ or- jj n
> 5: i
o
UJ UJ
<: o ;
uj n <
a. -H
t-i a
1 y s: «*
CQ
u o o
o •- ^ or
HH < «3
X rr- O >
t~ U-
in !£. O. Ct
C\, UJ *-• O
U < U-
U. 2- Q-
O O S
O Z>
O £\J
CO O
•-H C/3
C
I
UJ UJ
o :>
CC J
L f
> cr
« o
o
o
O
UI CO •
OOO3OOOOOOO
t^ as. >»
O t/5 O E.
_J ID >-
O O • •
oc iT
O »- •
~ y> co co to
co co co c/"> to
-60-
-------
X fJ li.
-•
o o:
. H- a
tH O
I O
cr
sD
UJ t/> •-
CJ Ut O
O O
' ^_' O O
• i-t X UJ >
IL! C/1
> a:
-a o
s:
o
o
I:
u
a.
-61-
-------
ae ~ CD
C. hO «•* O
* ^ —i
L» * O a!
P- 5- —. i,j
*I ^s. i— a.
f~~ "^ o t_r *-«
..' — o p* n
> _j »-* n-
Ll_ 2 X V"
O O 2
(. T) CT
(J X
s: a: -H
u o
ui o o
O tf, I.I
=» o o
U w
IU O
U U
vC K", *H
X I Cvi
CC CO
I CO -»
: ui n
u *: Jl
r-- u. *•
LiJ uj
O Z> UJ
•s cr o
> m
•=: o
IP
•O
.
O
u
O
I
-62-
-------
presented in Appendix B. Results for 3-hour and 24-hour averages are
similar to the results for the 1-hour averages. For 24-hour averages, RTDM
predicted the average of the 25 highest values with no significant bias, at
the 95 percent confidence interval.
Statistics for Highest Concentrations at Each Station
In Table 5-5 performance statistics are presented which compare the
maximum concentration values observed and predicted during the year at each
monitoring station (a total of 11 observed and 11 predicted values). This
table illustrates results for the 1-hour averages, and seven models.
Similar comparisons for second highest observed and predicted concentrations
are presented in Table 5-6. Caution should be exercised when interpreting
the meaning of some of the statistics for these rather small data sets.
The statistics provided in these tables compare observed and predicted
values for 11 data pairs as shown in the first column. The next two columns
present the average of the 11 observed concentrations and the average
difference between observed and predicted values. The 95 percent confidence
interval is given in parentheses, as calculated with a one-sample t test.
As with the 25 highest concentrations, these results indicate overprediction
(negative average differences), especially for COMPLEX II. At the 95
percent confidence level, the RTDM overpredictions are not significant.
(The confidence interval is almost as large as the average observed value.)
The fourth column in these tables displays the fraction of positive
residuals. This performance measure indicates the fraction of
observed-predicted data pairs for which the observed concentration is larger
than the predicted concentration. The results indicate overprediction at
from 6 of 11 stations (for RTDM) to all 11 stations (COMPLEX/PFM).
The next three performance measures provide estimates of scatter, or
characteristic discrepancies. They include the standard deviation of
residuals (differences) with 95 percent confidence limits calculated from an
F test; root mean square error; and average absolute residual. RTDM has the
smallest values for all three measures, and COMPLEX II has the largest.
The Pearson correlation of observed and predicted concentration pairs
and the nonparametric Spearman correlation of ranked sets of observed and
predicted concentrations provide indications of the spatial correlation of
the maximum concentration values at each station. Results from Table 5-5
and 5-6 show Pearson coefficients that range from 0.54 (4141 and SHORTZ) to
0.85 (COMPLEX II).
The last column in Tables 5-5 and 5-6 (variance comparison) presents the
ratio of observed variance divided by the predicted variance, with 95
percent confidence bounds in parentheses as calculated by an F test. The
1-hour variance comparisons for highest and second highest by station
results are significantly less than unity for all seven models, reflecting
the large magnitude and range of predicted values.
-63-
-------
5- • U
ac a: u_
•J P' Ul
bi o o
a tj u
t/j
o • u
t/^ iv LL
or cr tiJ
£C
r, r^ t.i
o o-
m
!
m
^- ^*
O
?=• h-
O <3
tr O
_i U
i- u H- z> a
O O i-* Q /N
cr o t/"1 x
U, tk UJ O
a «-
f^> rH CM
L
X
in
i
in m
I rO
a
tr
^L »•
I X
•a: > o
a: a: _J
(T
tT
-64-
-------
i* ri
U.I O liJ
y c
o —
I? •
O • U.
t/) or u_
UI O *-H
> CO I/)
«T fT"l Ul
^D
(
m
CQ
n"
UJ
rr
hi +*
i t- O
1 LJ
a. -3
O >
It t—
I— >
CO UJ
•3- rH
(V »•
» OJ -
fL O O '
y- z* i
o o
LJ LJ I
t\j
c :
O O «-t Q A
a: u_ t— oi
s: o ^r -
-65-
-------
Tables of statistics for 3-hour and 24-hour averages of highest and
second highest observed and predicted concentrations paired by station are
presented in Appendix B. The results for the longer averaging periods are
very similar to those for the 1-hour averages.
Statistics for Highest Concentrations by Event
Another data set consists of the highest observed and predicted
concentrations over the monitoring network for each sampling period, paired
in time (i.e., one pair of values for each 1-hour, 3-hour or 24-hour
sampling period). Results for the 1-hour averaging period are presented in
Table 5-7. While the data sets discussed up to this point contained
relatively few points, event-by-event comparisons for a full year involve
much larger volumes of data (i.e., a large "N"), as shown in the first
column of this table. The numbers of events is different for each model,
because the number of predicted values above the threshold values of
25|ag/m is different. The performance measures and confidence intervals
presented in this table have been discussed previously for Tables 5-1 and
5-5.
The average differences displayed ;.n Table 5-7 indicate that six of the
seven models tend to overpredict. The largest overprediction is by COMPLEX
II. The average difference predicted by 4141 is not significant at the 95
percent confidence level. RTDM underpredicted by 40 percent.
The standard deviation of residuals is an indicator of the range of
residual values encountered for each model. The smallest standard deviation
was obtained for RTDM, and the largest for COMPLEX II. Comparisons of
observed and predicted frequency distributions of concentration values
ignore any time pairing between observed and predicted values. Frequency
differences were significantly different from zero for all of the models.
The smallest frequency difference was obtained for SHORTZ (0.247), while
four of the models gave frequency differences between 0.77 and 0.82.
Tables for 3-hour and 24-hour average highest concentrations by event
are provided in Appendix B. The results are generally quite similar to the
results for 1-hour values. All of the models except RTDM overpredict, on
average, but the differences are not significant for 4141.
Statistics for All Concentrations Paired in Time and Space
The largest data sets considered in this evaluation represent all
concentration values paired in time and location. Results for the 1-hour
data sets are presented in Table 5-8 (Parts 1 and 2). Due to computer
work-space limitations, the size of the data sets for 1-hour values was too
large to calculate the maximum difference between observed and predicted
frequency distributions.
On average, three of the models overpredicted and four of the models
underpredicted. All of the over-and underpredictions are significant at the
95 percent confidence level. The Largest average overprediction is by
COMPLEX I and COMPLEX II, and the largest average underprediction is by
RTDM. The smallest average difference is by PLUMES.
-56-
-------
PQ
Qi O 3)
- *-
CJ O
U! r- LU
> IL -r
cr >
'1 O
-67-
-------
03
I
LO
P3
<
H
a _i S uj ** oj
j <*<*•> (TN
t/! > «T I i I wH
if. UJ _J —'
o z* 5" or en
O •- D —. K- O t- O
_J >— X
LL UJ LJ £.
2 LJ or x r>
*—i LJ o_ or
c. c
^ (/
>•- '-5 LJ
<5 u ' U.
>- a: u
<: « D
n t- •—
0 u.
o
O
Ul UJ
t1^ •> 1.1
<. a: =)
-68-
-------
^7 O
Ui —1 Uj
<_j cT cr
in * in
o —
a or L_
GO a: uj
n o
o o
D-
-. i-« Q /\
K V) >-t l/i
CJ O Vi O3
«i Q- LU O
ct: a: •-•
-------
Values for standard deviation of residuals, root mean square error and
average absolute residual are larger than average observed concentrations
for all seven models. The largest values for all three measures occur with
COMPLEX II. The smallest values of the standard deviation and root mean
square error were obtained for RTDM. The smallest average absolute residual
was obtained for 4141.
More than 50 percent of all residuals are positive, indicating that
while all of the models tend to overpredict in terms of the average
differences, there are more hours with underprediction than overprediction.
COMPLEX/PFM underpredicted 96 percent of the values, while SHORTZ under-
predicted 70 percent of the values.
As shown in previous studies, predicted concentrations correlate poorly
with concentrations observed at the same time and place. Pearson
correlations range from 0.008 (COMPLEX I and COMPLEX/PFM) to 0.079 (RTDM).
Spearman correlations range from -0.138 (SHORTZ) to 0.151 (COMPLEX/PFM).
Variance ratios are consistently less than 0.1 (except for 0.844 for RTDM)
and significantly different from unity.
Similar results are found for the 3-hour and 24-hour average statistics
for all concentrations paired in time and space (Appendix B). The same
models over- and underpredict as for the 1-hour averages. Correlation
coefficients improve for the 24-hour averages, but remain quite low ranging
from 0.1 (COMPLEX/PFM) to 0.38 (RTDM).
Table 5-9 is presented here to exemplify the results for data subsets of
observed and predicted concentrations paired in time and space. Subsets are
presented by station and for various meteorological conditions. Subset
tables for each of the models are presented in Appendix B. Table 5-9 shows
how COMPLEX I produces a mixture of over- and underpredictions, with all
overpredictions at the close-in receptors southeast of the stack.
Underpredictions are noted at more distant receptors and receptors located
in different directions from the plant. Overpredictions occur for all wind
speed categories and for stable (E and F) conditions. Neutral and unstable
hours produced underpredictions, on average.
Highest and Highest, Second-High Values
In many regulatory applications, model predictions of the highest or
highest, second-high concentrations ars of interest. Observed and predicted
highest and highest, second-high 1-hour concentrations are presented in
Table 5-10. These values clearly show an overprediction by all of the
models with the largest 1-hour overprediction by COMPLEX II (nearly a factor
of 20 for the highest prediction) and the smallest 1-hour overprediction by
RTDM (a factor of just under two for both values). Table 5-11 shows similar
results for 3-hour and 24-hour averages with the largest overpredictions
again by COMPLEX II and smallest overpredictions by RTDM.
-70-
-------
O fO
UJ «
O£ * Z
Q- X O
^ I- —
Q O I- (f!
Z 3 «T I-
< •* U ui
o in
o i/> _j
ID
;»
LO to >
£Q UJ l/l
i i i o 2* s: s
s ;;rl
O _l
a
o
00
O'
IX — > UJ rt >-
-
O Z C. 2.
Lu LJ C
z u a: u.
O 2 «-i —
W O «l
•-i O 0.
a:
-i cj
0. O
: ui
r
o z <
a: o n
i « o
a u.
o
UJ U. (/)
> u. 01
13 > LJ
1 OL 2
a: w —'
UJ VI <
> CD >
< O
a u. e
s. c u.
^ »^ K)
cu
c\j
10
i
ff\
i
ZZZZ27ZZZZS
ooooooooooo
* -T vfl
«• « eg
~ -. «• IT
-1 K) I
I I
U. * O C.
a, a: K- U O
-71-
-------
TABLE 5-10 - HIGHEST (H) AND HIGHEST, SECOND-HIGH (HSH)
1-HOUR CONCENTRATIONS FOR WESTVACO WITH
ASSOCIATED METEOROLOGY
1-HOUR AVERAGES
Model
OBSERVED
COMPLEX I
COMPLEX II
4141
RTDM
PLUME5
COMPLEX/ PFM
SHORTZ
Concentration
(ug/m3)
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
2570
2344
23678
21063
50705
46260
18714
18020
5073
4312
26630
16946
34281
20960
24200
21810
Receptor
Number
4
7
1
1
6
1
9
9
6
1
7
1
4
1
6
6
Stability
Category*
4
6
6
6
6
6
6
6
5
5
6
4
4
6
6*
6*
Wind
Speed(m/s)*
2.9
3.1
1.2
1.3
1.6
2.0
2.1
1.8
1.0
1.5
1.0
4.3
1.0
1.5
3.8
3.3
* The SHORTZ model uses intensity of turbulence (ly and Iz) to define
dispersion. The stability categories presented here were the ones used to
define meteorological subsets.
-72-
-------
TABLE 5-11 - HIGHEST (H) AND HIGHEST, SECOND-HIGH (HSH)
3-HOUR AND 24-HOUR CONCENTRATIONS FOR WESTVACO
FULL YEAR MODEL RUNS
3-HOUR
24-HOUR
Model
OBSERVED
COMPLEX I
COMPLEX II
4141
RTDM
PLUME 5
COMPLEX/ PFM
SHORTZ
Concentration
(ug/m3)
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
H
HSH
2066
1509
17973
16827
26960
25537
10088
6238
2564
1954
15268
11901
14007
11485
10751
7605
Receptor
Number
7
6
1
1
6
6
9
9
6
6
1
6
1
4
3
6
Concentration
(ug/m3)
487
401
4647
4102
8854
6338
2026
1260
766
596
8450
3843
3024
2949
2227
1811
Receptor
Number
1
6
1
1
1
1
9
5
6
6
6
6
6
6
3
6
-73-
-------
WESTVACO - IMPACT SELECT HOUR RESULTS
Performance statistics have been prepared for all models for the subset
of hours selected from the Westvaco data set for IMPACT model runs. Using
the criteria discussed in Section 4, a total of 20 days representing 480
hours were selected. The days selected for this analysis all contained high
observed concentrations at one or more monitors, while at the same time
other monitors recorded zero concentrations.
The 3-dimensional grid which was constructed for the IMPACT runs was
constrained, due to computer limitations and model resolution requirements,
to exclude the most distant monitor at Stony Run (receptor No. 10 in the
full Westvaco model runs). Therefore the statistical comparisons for each
model are based on data sets representing predictions made at 10 receptors
for 480 hours (4800 receptor-hours/model).
Statistical comparisons were produced for the 1-hour, 3-hour and 24-hour
averaging periods, except for the 25 highest data sets which do not contain
24-hour averages. Since only twenty 24-hour periods were analyzed, the
highest by event data sets provide similar information for this averaging
period.
Subsets of events by station or for various meteorological conditions
are not presented for the Westvaco-IMPACT selected hour analysis.
Although the early test run package: for IMPACT was approved by the model
developer, the results suggest that the model did not operate properly for
all prediction runs. In his review of the draft report, the model developer
(Alan Fabrick) commented "the predicted concentrations are so large that
they could not have occurred if the model was running correctly. For some
reason the model's numerical algorithm for simulating advection and
diffusion went unstable for a few hours; of Westvaco simulations." The model
input data for these periods of high concentrations have been reviewed along
with the model code, however, to date, the specific technical problem has
not been identified.
Statistics for 25 Highest Values
Table 5-12 presents statistics for the comparison of 25 highest observed
and predicted S02 concentrations for 1-hour averages. The performance
measures and confidence intervals are the same as those described for Table
5-1.
The largest overpredictions, as depicted by the difference of averages,
are by COMPLEX II and IMPACT. The smallest overpredictions were obtained
for 4141 and RTDM. The overpredictions are significant at a 95 percent
confidence level for all models except RTDM.
Overpredictions are similarly indicated by the comparison of median
differences, with two exceptions. The median differences predicted by 4141
and RTDM are not significant at the 95 percent confidence level. The IMPACT
model has a much improved (but still poor) performance for this measure
indicating that the average is affected by extreme overpredictions for a few
hours.
-74-
-------
0
UJ
t~
( J
1-4
O 00
u) 2:
C£ «•"» O 13
Q fo -. o a:
+ *'•—>
C- + O cr" _I
•^_ y — i u| uj
_i •— * en h-
fy C/l O «-« LJ
(-\l UJ UJ ttT ^ "X «l
r-H o: u o a: o a.
1 cc _a LJ ot> s:
10 > y- «r —i
t- •-» *- or
W oo 2 t- a: o
1— 1 Ul O 13 O U.
CQ X t-, 2: o u
u
T- ry r~i t-« *— H-
>— LJ 00 CJ
in ^r u^ uj uj
fV LJ HI LJ 3 _J
0 -J X UJ
L_ ?- 0_ r- 00
0 0 ^
U =) 0- V)
y — o cc
r~j oj u ^3
00 C
*— W T
err
o
LJ
I>1
>- t )
y i_> ^
•-H D Ul
si or »— «
U C~i
2 ^
UJ O O
O 0" Ul
H —• a:
O o
LJ w
* — .
UJ O
2 O LJ
O
a: «-» _j
U 1 £~) «3
> UJ >
*J Q
CL
O
O > UJ
c>jj-«-H
>o o oj •-* fO ^-i»~(c;
0 t-! CM (T. '-J 0 C->
• •••••••
IP * CM* r-< * u3* in* in* ^n * cs»
«-i a- CM TO
- o h- rn CD *
in ^ I-H* «•* CM* •a-* -a-* s-oCL
oOr-.h-_joxs:
LJLj
-------
Again, predicted variances are large relative to observed variances.
The variance ratio for RTDM (0.096) is highest, while the lowest value
r,'",'-, wit?) IMPACT (0.000).
)•.' ' /.• • • ii-"i j on:= itvt-oJ Utdt. theife is vfety little overlap between them for most
of the models. Differences between the observed and predicted distributions
are shown to be significantly different from zero at the 95 percent
confidence level for all models except RTDM.
Results for the 25 highest 3-hour averages (Appendix B) are quite
similar to those for the 1-hour values. The main exception is that 4141
underpredicts the median difference for 3-hour values.
Statistics for Highest Concentrations at Each Station
Comparison of the sets of highest concentrations by station for the
1-hour values can be seen in Table 5-13. Results for second highest values
are presented in Table 5-14. For these comparisons, each data set consists
of only 10 data pairs.
Results indicate overprediction by all eight models for the highest
values, and by all models except RTDM for the second highest values.
However, for four models, the overpredictions are not significant at the 95
percent confidence level. The largest overpredictions occurred with the
IMPACT model.
Results for the fraction of positive residuals indicate that IMPACT
overpredicted the highest and second highest values at all 10 stations.
PLUMES underpredicted second highest values at 7 stations. Measures of
scatter were largest for IMPACT and smallest for RTDM.
The variance comparison indicates that the variance of predicted values
is significantly larger than the varia.nce of observed values for all eight
models.
Statistics for the 3-hour average and 24—hour average highest
concentrations at each station are given in Appendix B. The results are
similar to those for 1-hour values.
Statistics for Highest Concentrations by Event
Statistics for the comparison of highest observed and predicted
concentration values event-by-event (paired in time) are provided in Table
5-15 for the 1-hour values.
The number of events (418-463), representing the number of hours
analyzed for the 1-hour data sets, is less than the number of hours modeled
(480) primarily as the result of screening for threshold values. Results
for the average difference indicate that all of the models except 4141 and
RTDM tend to overpredict the highest values each hour. However, three of
the models neither over- nor underpredict significantly at the 95 percent
confidence level. Large overpredictions (factor of three times observed
values) occur for COMPLEX II, while the average underprediction for RTDM is
acout 50 percent of the average observed value.
-76-
-------
PQ
K> O OL
* O
c~ U! to «or —> o
Ui *— !3 rH S"
T () ** (V or
O M ID fT> t—
f— f-i CO O ^ ( )
X UJ Ul T *x «l
rr "3 <-j L*.
U_ LL -J ^H Q- E
fi
CA tr
oi or u.
a z co
>/)
C/l LJ UJ
O a:
u- CO D
> _J LJ
D. ui O
a — •
L.
«I L_
U O
O UJ uJ
> 00 >
in
^-t OD
t ^
—• i in
-77-
-------
w
5
H
0
c
1— 1
ry
UJ
Q
O tO
"") O Ct
t- * o
C^ I.J 1*3 *3 *» C~i
»- i- :D — < s:
T ( i — ' cr cr
o o to o *-H o
c, 'rV' ri ~ ^
ej a. _i .^ oc ;c
uJ «a cr, •—
V/^ f~J > T-<
;? Ui — • or
U_ UJ
0" or rr* h ^ f—
*— I.J t— t/"> U
CL. *-o ^ 2 uj LU
<. a" ui o 3 -J
a o u —t ui
r z H- to
O 0 LJ - c
C" O X
C
UJ
cr
1-4
c^1 to
^ m ui
«J Of
«
Uj UJ cr
*r cr
^- 3 ac
o rs 1. 1
o i/>
or
*
C3 Z CO
Ct f~J _J
*3" «— I CO
CO Ul UJ
O CC.
r*
£L UJ C/J LJ
O > -J UJ
— i—
-------
" UJ U-
£C U-
U_ »-H
o: o =5
-«
O U_
O
m
i—i
i
UJ « O
ct 10 «. o
Q. * LJ HH
•* 5" a:
o s: -. ui
:> to
ct; til
LU n a_ cr
> U- nri
I
X i- I
i; >-
u. LJ 0.1 u-
O CJ I T
^T h- h-
^ O 2
o o ui a:
^ > o
LJ oo
> cr
Ml
ro
LJ
c,
s:
o
X
CO
-79-
-------
Standard deviations of residuals indicate the largest scatter for IMPACT
(20 times average observed value), with the smallest scatter by RTDM.
Maximum frequency differences indicate a distinct difference among the
models. For four models (COMPLEX I.. COMPLEX II, 4141, and COMPLEX/PFM)
there is little overlap between observed and predicted distributions, and
the maximum difference is close to 1. The lowest value, for IMPACT, is
0.385.
Tables of 3-hour and 24-hour comparisons of the highest concentrations
by event are displayed in Appendix B. The results are quite similar to the
1-hour comparisons. The average observed value drops rather slowly with
increasing averaging time (338-349 ug/m3 for 3-hour periods; 261-269
ug/mj for 24-hour periods).
Statistics for All Concentrations Paired in Time and Space
Table 5-16 (Parts 1 and 2) presents the comparison of all observed and
predicted 1-hour concentration values paired in time and location for the
Westvaco IMPACT select hours. The total number of events (2476-2595)
implies that roughly half of the hourly observed-predicted pairs (4800
total) passed the tests for threshold and/or missing data. Average observed
values are quite high (about 200 ug/m3) due to the nature of the data
selection criteria.
Results for the average differences show considerable variability among
the models. From the 95 percent confidence intervals, one model
overpredicts significantly (IMPACT), four models underpredict significantly
(4141, RTDM, PLUMES and COMPLEX/PFM), and three models show no significant
tendency to over- or underpredict (SHORTZ, COMPLEX I and COMPLEX II). The
magnitude of the average differences represents from 12-71 percent of the
average observed values, except for IMPACT (343 percent). The prediction
biases indicated by these results should be interpreted with caution, since
the selection criteria favored hours with high observed concentrations.
Results for the full data set (for all models except IMPACT) are more
reliable for judging bias, because they are not subject to this limitation.
Values for the standard deviatior. of residuals, root mean square error
and average absolute residual all exceed the average observed values. The
largest measures of scatter occur for IMPACT, followed by COMPLEX II; while
the smallest values occur for RTDM, and also 4141 and PLUMES.
Maximum frequency differences and fractions of positive residuals are
all quite large.
Correlations of observed and predicted concentrations are extremely low,
and negative in many cases. Variance ratios indicate variances for
predicted values are much greater than variances for observed values.
Results for the 3-hour and 24-hour concentrations paired in time and
space can be found in Appendix B. The results are generally similar to the
results for 1-hour values.
-30-
-------
o m
UJ *
n (i t— »••
o
<-% O
vD
i—<
I
PQ
Z3 o tr n i
tt _i ^ uj ** '
tkJ
J •-« t- O
_J (- X
O L_
(_)
—( t— c/o
t^- o a:
O LJ C£
« o-r (X
£t Ul I
UJ li tf>
> u cc
«r ^H o
> re >
-------
=- « a.
«3 O' X.
*-«
*— h- rr, t-
C
O
U
\D
t—i
I
in
hJ
PQ
c~5 ui _i *- a
o n -*: cf a> o
O »-i 15 -3 O U.
_J »-* K- C H- O
_J K X .
O .2 O
UI LJ LJ
^ O (T X
CO O «3
« rr >- •—
_» f)
cj O w CL
-------
CINDER CONE BUTTE RESULTS
All eight complex terrain models were run with the full Cinder Cone
Butte data set, consisting of 104 hours of SF6 tracer and meteorological
data. Ambient tracer samples were observed at up to 94 receptor locations.
Statistical performance measures were generated for the 1-hour average
values, only. Results are not presented for highest concentrations by
station. Otherwise, the performance measures and confidence intervals
presented for Cinder Cone Butte are the same as the ones described
previously for Westvaco. Slightly different subsets by meteorology and
source-receptor geometry were selected for the Cinder Cone Butte analysis.
All observed and predicted values and the corresponding performance measures
are for relative concentrations (i.e., mass concentration per unit of
emission rate, X/Q)• No screening was performed for threshold values of
observed and/or predicted concentrations.
Cinder Cone Butte results for the IMPACT model did not show any evidence of
the instability suspected for Westvaco. However, the model developer's
comments to the draft report indicated that the eddy diffusivity algorithm
in the IMPACT model is not appropriate for the grid resolution (50m
horizontal; 10m vertical) used for the Cinder Cone Butte model runs. (This
issue was not raised when the test package was reviewed).
Statistics for 25 Highest Values
Statistics for the comparison of 25 highest observed and predicted data
sets are given in Table 5-17. From the difference of averages this table
shows that six of the eight models overpredict on average, and that these
differences between observed and predicted averages are non-zero at a 95
percent confidence level. The IMPACT model average underprediction is also
significant at the 95 percent confidence level. Only the RTDM model shows
no significant bias. Results for median difference are similar to the
results for difference of averages.
Variance ratios are below 0.5 except for COMPLEX I and IMPACT. The
confidence interval for COMPLEX I indicates that no significant difference
exists between the variance of COMPLEX I predictions and the variance of
observations at a 95 percent level of confidence.
Observed and predicted frequency distributions differ significantly for
all of the models except RTDM and SHORTZ.
It should be noted that interpretation of the 25 highest observed and
predicted concentrations for Cinder Cone Butte is not quite as simple as for
Westvaco. This is because, in addition to the impairing in space and time,
the experiments included changes in source-receptor geometries since a
mobile crane was used for the releases. One group of subsets based on
source-receptor geometries was selected for investigation with the 25
highest Cinder Cone Butte data sets. Table 5-18 is presented here to
exemplify the subset results for COMPLEX I. Additional subset tables are
provided for all models in Appendix C. As shown in Table 5-18, the four
subsets selected for the 25 highest comparisons are based on receptor
height. The intention was to investigate whether model performance varied
with receptor height. No pronounced differences in performance were
identified.
-83-
-------
Statistics for Highest Concentrations by Event
Comparisons of highest observed and predicted concentrations
event-by-event for Cinder Cone Butte are found in Table 5-19 (Parts 1 and
2). This table is identical in form to the Westvaco tables for the full
data sets paired in space and time. For this tracer data set, no threshold
screening was performed. The number of events and average observed values
are identified for all models.
Results for the average difference indicate overprediction by all of the
models except IMPACT, which underpredicted by an average of 50 percent, and
RTDM which exhibited no significant bias. The largest overprediction (by a
factor of 3.6) occurred with COMPLEX II.
Measures of variability between observed and predicted concentrations
(standard deviation of residuals, root mean square error and average
absolute residual) are largest for COMPLEX II and smallest for IMPACT and
RTDM.
The predicted frequency distributions are all significantly different
from the observed distributions for all models except RTDM and SHQRTZ. The
largest frequency difference occurs for COMPLEX II.
From the fraction of positive residuals, COMPLEX II overpredicted for 74
percent of the highest concentrations by event, while IMPACT overpredicted
for only 28 percent. The best performance for this measure was by SHORTZ
which overpredicted 52 percent of the events.
Correlation coefficients for Cinder Cone Butte show some improvement
over Westvaco, but remain fairly low. Pearson coefficients range from 0.26
(SHORTZ) to 0.60 (RTDM), while Spearman coefficients range from 0.32
(COMPLEX II) to 0.51 (RTDM).
The variance ratios are significantly different from unity for all the
models, with the variance of predictions larger than the variance of
observations for all models except IMPACT.
Table 5-20 is presented here to exemplify, for COMPLEX I, the evaluation
of model performance for various subsets of source-hill characteristics and
meteorological conditions. Similar tables for each model are presented in
Appendix C. Six of the subsets are based on two release distance categories
(less or greater than 900 m from source to butte top) and three release
height categories. Wind speed and stability categories are also evaluated.
The number of events in some categories; is quite small.
For COMPLEX I overpredictions occurred for five of the six
distance/height categories, and for low wind speeds and stable conditions.
Underpredictions, on average, occurred for the higher wind speeds and
non-stable conditions.
-------
r~-
I
a -^
UJ (/"
o n u
UI ,-H O
t/J U O o o
(_) w
H- ui
O rv
bJ UJ
cr a: LL
Uj uJ I
or •-. _j
LJ f, en >
-------
I
l-O
W
J
PQ
— . r
O ^
C. o (-) t-
Ul r-4 O Ul
or
uj (/5 a: <
o. u. c; i-
ct 3 «
O _l Ul Ci
- > 1-1 (/)
C/1 »— 3
UJ 2 _^ O
o — — a.
a
o
IS UJ
•" >
JC — •
« h
CL a
r. _i
O UJ
o a:
a. <» it
c
o or UJ
o D s
o o
.. I u
-i i
i.i — a
c u-1
c c
I b. 2-
UJ O C-.
O M UJ
2 — 0:
< tr a.
•« «« -v
ae a. w
< r a.
> o o
u u o
z a u.
u U l>
< o:
a.
o > ui
< a: 3
(TV — —. K)
>o m «• «•
LJ Ul
K X
Z UJ
O CO
9 Ul
Ul IX
o
>- o
s: >-
O 3 o • .
H- o *-* o • o
d. (/! I « CT K>
UJ V I —, A
(J O
-86-
-------
-H G
O h- »-•
O IL.
O
o ?-
•—• X
d CO
l 5~ rr
^ * t- CL
^ &
o —, M ^
^ o tt
or w i.i -i
U.. L.. Q- CX
CJ3 UI I^
•j o: Q.
or LJ l
1.1 It 0"
*«»
cr:
t— CXL - > i
X ci i uj
t- >- C-
U. ^ Cl L^ ^
O I
LJ 00 <1
> cr, >
•=r O
-------
-o
(U
c
o
CL t/j ..) «I U'
u. L.1 D' U *—
S/J Z) t-t U OJ i—
rri _j — c=.
*- or < Uj
co p' t— 13 a z:
^ c ^ c o
T i- ^J X U
o uj i :r
c; •— t~
Z* i=" 7*
o o i.i cr
LO LJ > Cj
•—< Ui U_
cr u
G- •—
r: t—
o «
O _J
Cfl 0£ LJ
O O
bl U
or u.
o o
O U
a: _j o
> c/^
-------
o •- ~
tn * ( )
Q£ _l —
rt O
<.' U
CO
O Z _J
Q£ O *t
< 1-1 3
Q I- Cl
«1 -^ t/>
UJ IX
c
O K1
UJ *
O l/>
LJ I
(X (/J
c/) ^
UJ O
T. 1-1
U. Z I
o uj
O L2
i— — «t uJ
ac >. c. i—
Q. - _l I
2T li t t-i Qi
UJ O '-ij
• > o a
1 UJ E UI 2
X —
»— LJ
<_) O
UJ Z UJ
O UJ O!
< n- a
Cr- ui I
UJ 11 O
» b m
o > i
r~
Z Ul
o
( ) >-
ru
H-
M i-
X I.I
(f O
>—
A i: SL
f- V II /\ 1— \/ II /\
t/3TIX OOXXX
Q* CNJ C\J —• */) C^
f\J f\J OJ Kj CVI fj
I- a:
o on •
u. ^ in i/) >•
LJ E ^* i—
i O JC -•
-89-
-------
Statistics for All Comparisons Paired in Time and Space
Statistics for the full set of paired observed and predicted
concentrations are presented in Table 5-21. These data sets have the
largest populations (3836 data pairs) of any group for the Cinder Cone Butte
data base.
As with the high-by-event data group, average differences for all
concentrations indicate overprediction by all of the models except IMPACT,
which underpredicts by about 50 percent, and RTDM, which exhibits no
significant bias. The largest average overpredictions are by 4141 (by a
factor of over two).
Measures of variability between observed and predicted concentrations
are largest for COMPLEX II, and smallest for IMPACT and RTDM. Frequency
distributions of observed and predicted values are significantly different
(at a 95 percent confidence level) for all eight models.
Correlation coefficients for Cinder Cone Butte model results are
substantially better than for Westvaco results for all concentrations.
Pearson coefficients range from 0.22 (PLUMES) to 0.43 (RTDM), while Spearman
coefficients range from 0.33 (COMPLEX I) to 0.45 (SHORTZ).
The variance ratio for RTDM was not significantly different from unity
(at a 95 percent level of confidence). For the other models, the variance
of predictions was significantly larger than the variance of observations,
although IMPACT, the opposite relationship was true.
-90-
-------
x: r? LJ
or
CM
I
w
nJ
FQ
en « o
ill «J- r-.
oi t 2 o-
n —• o uj ~»
* •-< CL o
n * t— —. tr
P* n U [ t ft- Li »~
0" 32 ^ UJ •»• f-
t.i _J «J > -1
C1
LJ ~.
O O
3 UJ 0-
tr O,
Of. >
QJ LJ
.-i l—. ^-4 or iv
L. cr <-, LI
O t- O 21 O
zr L.I L_J sr *:
^ LJ :r x n »-.
O CJ •-' >— C/l O
c/^ I?1 <:
•-"OQ.CC
rv C) O
c •-«
(_• i-
O
CJ
o
X
00
-91-
-------
o o
o o
O
O
o o
o o
T3
__.._, IY C. K-
UJ —J
5~ »: oj u
c i- o >- o
-J Q
(- —
w
,J
PQ
<
H
C3 ^-1 ^ rr- ff
u. a' <^ u_'
o H- £_, i. o
T' UJ CT T "1 •-.
O CJ •— t- ^7 (_'
£/> 2 «
•-« o c_ or
a:
^ o
-
u
o
-92-
-------
SECTION 6
SUMMARY AND CONCLUSIONS
The performance evaluation of the complex terrain models has produced an
imposing array of statistical measures to compare observed and predicted
concentration values. The principal objective of this project is to produce
performance statistics so that EPA and a group of reviewers may judge the
relative merits of different models. In this report, the results have been
discussed and explained, but no attempt has been made to compare the
performance of one model versus another. Many of the model developers, upon
reviewing this report, indicated the desire to see more detailed depictions
of the results such as scatter plots of observed and predicted
concentrations, histograms, cumulative frequency plots, isopleth analyses
and time series displays. Graphical displays can be useful in exploring
possible causes of poor model performance and are particularly desirable in
diagnostic model evaluations. One of the difficulties encountered in the
presentation of operational evaluation statistics is selecting meaningful
graphical or tabular displays with limited report space. An abundance of
useful information remains to be extracted from the results of this study
and it is hoped that further analyses are pursued in the future. The
conclusions and recommendations presented below are concerned with model
evaluation methods and with the performance of the models as a group.
The complex terrain models were evaluated using two data bases
representing different terrain settings and experimental approaches. The
Westvaco data set consisted of one year of measurements at eleven SOz
monitoring stations in the rugged terrain of western Maryland and
northeastern West Virginia, for a buoyant tall-stack release. The Cinder
Cone Butte experiments were conducted for 104 hours using a non-buoyant
tracer release, with impacts measured from a 94-station sampling grid on an
isolated small hill.
SUMMARY OF RESULTS
The results discussed in Section 5, plus those in Appendices B and C,
contain a wealth of information concerning the performance of each of the
eight complex terrain models. Distinct differences in performance are
evident among the models. The patterns of results changed between the two
data sets and, to a lesser extent, with averaging time (for Westvaco). A
few key results are highlight below.
-93-
-------
Westvaco. For Westvaco, seven of the models overpredicted the 25
highest concentration values for 1, 3, and 24-hour averaging times, by
factors ranging from 2 to 20. RTDM predicted with less bias than the
other models for all three averaging times. (The IMPACT model was
evaluated only for selected hours from Westvaco.) The COMPLEX II model
and the IMPACT model gave the largest overproductions. COMPLEX I also
overpredicted the average of the 25 highest 1-hour values by almost a
factor of 10.
Cinder Cone Butte. Six of the eight models overpredicted the 25 highest
1-hour values. IMPACT underpredicted, and RTDM predicted with no
significant bias. COMPLEX II again gave the largest overprediction,
roughly a factor of 4 times observed.
Thus, COMPLEX II showed the most consistent and pronounced tendency to
overpredict peak concentrations; RTDM showed the least bias for estimating
peak 1-hour values; and IMPACT showed the greatest inconsistency between the
two data sets.
Model performance results for the two data sets showed several striking
differences:
• The models showed a much greater tendency to overpredict peak
1-hour concentrations for the Westvaco data set than for Cinder
Cone Butte.
• Comparisons between predicted and observed concentrations, paired
in time and location, showed smaller discrepancies and higher
correlation for Cinder Cone Butte than for Westvaco.
• For Westvaco, model performance was very different for stable and
neutral conditions (for most of the models). For Cinder Cone
Butte, model performance was generally similar for both stability
categories.
These differences point to the importance of the source characteristics
and the local terrain setting (as well as other design factors) for model
performance in complex terrain.
The Westvaco data set permitted model performance to be evaluated by
monitoring station and for several averaging times. From these analyses,
the following conclusions could be drawn:
• Distinct differences in model performance were found between those
monitors within 2 km of the plant and those at greater distances.
Overprediction was more pronounced at monitors close to the plant.
• Results for 1-hour and 3-hour averages were quite similar. For
24-hour averages, however, distinct differences in model
performance were found for estimating peak concentrations.
-•94-
-------
REFERENCES
1. United States Environment Protection Agency, 1978. Guideline On Air
Quality Models. EPA-450/2-78-027. OAQPS, Research Triangle Park, NC.
2. Fox, D.G., 1981. Juding Air Quality Model Performance (A Summary of the
AMS Workshop on Dispersion Model Performance, Woods Hole, MA, 8-11
September 1980). Bull. Am. Meteorol. Soc., 62, 599-609.
3. Londergan, R.J., D.H., Minott, D.J. Wackter, T.M. Kincaid and D.M.
Bonitata, 1982. Evaluation of Rural Air Quality Simulation Models.
Prepared for EPA by TRC Environmental Consultants, EPA-450/4-83-003,
OAQPS, Research Triange Park, NC.
4. Minott, D.H., R.J. Londergan, W.M. Cox, and J.A. Tikvart, 1982.
Comparative Performance Evaluations of MPTER and Alternative Rural
Models. Presented at the 75the Annual Meeting of the Air Pollution
Control Association, Mew Orleans, LA.
5. Londergan, R.J., D.H. Minott, D.J. Wackter and R.R. Fizz, 1983.
Evaluation of Urban Air Quality Simulation Models. Prepared for EPA by
TRC Environmental Consultants, EPA-450/4-83-020, OAQPS, Research
Triangle Park, NC.
6. Pierce, T.D. and D. B. Turner, 1980. User's Guide for MPTER.
EPA-600/8-80-016, U.S. Environmental Protection Agency, Research
Triangle Park, NC.
7. Stnmaitis, D.G. , J.S. Scire and A. Bass, 1982. User's Guide for
COMPLEX/PFM Air Quality Model. EPA-600/8-83-015, Environmental
Protection Agency, Research Triangle Park, NC.
3. Enviroplan, Inc., 1981. User's Manual for Enviroplan's Model 3141 and
Model 4141. Enviroplan, Inc., West Orange, NJ.
9. United States Environmental Protection Agency, 1977. User's Manual for
Single Source (CRSTER) Model. EPA-450/2-77-013, OAQPS, Research
Triangle Park, NC.
10. Pacific Gas and Electric, 1981. User's Manual for Pacific Gas and
Electric PLUMES Model. Pacific Gas and Electric, San Francisco, CA.
11. Environmental Research & Technology, Inc., 1982. User's Guide for the
Rough Terrain Diffusion Model (RTDM, Rev. 3.00). ERT Report Mo.
M 2209-585. Environmental Research & Technology, Inc., 3oncord, MA.
-95-
-------
12. Bjorklund, J.R., and J.F. Bowers, 1982. User's Instructions for the
SHORTZ and LOMGZ Computer Programs, Volumes 1 and 2. EPA 903/9-82-004,
U.S. Environmental Protection Agency, Research Triangle Park, NC.
13. Cramer, H.E., et al. , 1972. Development of Dosage Models and Concepts.
Final Report under Contract DAAD 09-67-C-OO 20 (R) with the U.S. Army,
Dessert Test Center Report DTC-TR-72-609, Fort Douglas, UT.
14. Fabrick, A.J. and P.J. Haas, 1980. User Guide to IMPACT: An Integrated
Model for Plumes and Atmospheric Chemistry in Complex Terrain. Radian
Corporation, Austin, TX.
15. Tran, K.T., R.C. Sklarew, 1979. User Guide To IMPACT: An Integrated
Model For Plumes And Atmospheric Chemistry In Complex Terrain. Form &
Substance, Inc., Westlake Village, CA.
16. Wackter, D.J., 1983. Test Run Package: Description of Models "As-Run"
for Complex Terrain Model Evaluation. Prepared for EPA by TRC
Environmental Consultants under Contract 68-02-3514, W.A. 27, OAQPS,
Research Triangle Park, NC.
17. Lavery, T.F., A. Bass, D.G. Stnma.itis, A. Venkatrom, B.R. Greene, P.J.
Drivas and B.A. Egan, 1982. EPA Complex Terrain Model Development:
First Milestone Report - 1981. EPA-600/3-82-036, Environmental
Protection Agency, Research Triangle Park, NC.
18. Strimaitis, D.G., A. Venkatrom, B.R. Greene, S. Hanna, S. Hesler, T.F.
Lavery, A. Bass and B.A. Egan, 1983. EPA Complex Terrain Model
Development: Second Milestone Report - 1982. EPA-600/3-83-015
Environmental Protection Agency, Research Triangle Park, NC.
19. Truppi, L.E., and G.C. Holzworth, 1983. EPA Complex Terrain Model
Development: Description of a Computer Data Base from Small Hill
Impaction Study Mo. 1, Cinder Cone Butte, Idaho. Environmental Sciences
Research Laboratory, Research Triangle Park, NC.
20. Maryland State Department of Health and Mental Hygiene, 1979. Westvaco
Corporation Amended Consent Order.
21. Cramer, H.E., 1981. Westvaco-Luke, Maryland Monitoring Program: Data
Analysis and Dispersion Model Evaluation (First Two Quarters). H.E.
Cramer Company, Inc., Salt Lake City, UT.
22. Hanna, S. , C. Vaudo, A. Curreri, J. Beebe, B. Egan, and J. Mahoney,
1982. Diffusion Model Development and Evaluation, and Emission
Limitations at the Westvaco Luke Mill. Document PA439 prepared for the
Westvaco Corporation by Environmental Research & Technology, Inc.,
Concord, MA.
-96-
-------
23. Cramer, H.E., 1982. Portocol for the Evaluation of the SHORTZ and LUMM
Dispersion Models Using the Westvaco Data Set. H.E. Cramer Company,
Inc., Salt Lake City, UT.
24. Snedecor, G.W. and W.G. Cochran, 1967. Statistical Methods, 6the
Edition. Iowa State University Press, Ames, Iowa.
25. Hollander, M. and R.A. Wolfe, 1973. Nonparametric Statistical Methods.
John Wiley and Sons, New York, NY.
26. Hirtzel, C.S. and J.E. Quon, 1981. Estimating Precision of
Autocorrelated Air Quality Measurements. Summary of Proceedings
Environmetrics 81, 200-201.
27. United States Environmental Protection Agency, 1981. Regional Workshops
on Air Quality Modeling: A Summary Report. EPA-450/4-82-015,
EPA/OAQPS, Research Triangle Park, NC.
-97-
-------
APPENDIX A
TEST RUN PACKAGE: DESCRIPTION OF MODELS "AS-RUN"
FOR COMPLEX TERRAIN MODEL EVALUATION
-------
TEST RUN PACKAGE:
DESCRIPTION OF MODELS "AS-RUN"
FOR COMPLEX TERRAIN MODEL EVALUATION
Environmental
Consultants, Inc
TRC Project No. 2164-R81
David Wackter
Project Manager
September, 1983
800 Connecticut Blvd.
East Hartford, CT 06108
(203) 289-8631
-------
TABLE OF CONTENTS
SECTION PAGE
1.0 INTRODUCTION 1
2.0 COMPLEX-I AND COMPLEX-II 3
2.1 Technical Modifications to COMPLEX-I and COMPLEX-II 3
2.2 COMPLEX-I and COMPLEX-II: Input Options and
Variables for Cinder Cone Butte 4
2.3 COMPLEX-I and COMPLEX-II: Input Options and
Variables for Westvaco 4
2.4 TRC Changes to COMPLEX-I for Cinder Cone Butte . . 5
2.5 TRC Changes to COMPLEX-I for Westvaco 5
2.6 TRC Changes to COMPLEX-II for Cinder Cone Butte . . 6
2.7 TRC Changes to COMPLEX-II for Westvaco 6
3.0 PLUMES 7
3.1 Technical Modifications to PLUMES 7
3.2 PLUMES: Input Options and Variables 8
3.3 TRC Changes to PLUMES Code for Cinder Cone Butte . 9
3.4 TRC Changes to PLUMES Code for Westvaco 10
4.0 RTDM 11
4.1 Technical Modifications to RTDM 11
4.2 RTDM: Input Options and Variables for Cinder Cone
Butte 12
4.3 RTDM: Input Options and Variables for Westvaco . . 13
4.4 TRC Changes to RTDM for Cinder Cone Butte 14
4.5 TRC Changes to RTDM for Westvaco 15
5.0 SHORTZ 16
5.1 Technical Modifications to SHORTZ 16
5.2 SHORTZ: Input Options and Variables for Cinder Cone
Butte 16
5.3 SHORTZ: Input Options and Variables for Westvaco . 17
5.4 TRC Changes to SHORTZ for Cinder Cone Butte .... 18
5.5 TRC Changes to SHORTZ for Westvaco 18
6.0 4141 19
6.1 Technical Modifications to 4141 19
6.2 4141: Input Options and Variables for Cinder Cone
Butte 19
6.3 4141: Input Options and Variables for Westvaco . . 20
6.4 TRC Changes to 4141 for Cinder Cone Butte 21
6.5 TRC Changes to 414L for Westvaco 22
7.0 COMPLEX/PPM 23
7.1 Technical Modifications to COMPLEX/PFM 23
7.2 COMPLEX/PPM: Input Options and Variables for
Cinder Cone Butte 24
7.3 COMPLEX/PPM: Input Options and Variables for
Westvaco 24
7.4 TRC Changes to COMPLEX/PFM for Cinder Cone Butte . 25
7.5 TRC Changes to COMPLEX/PFM for Westvaco 26
-ii-
-------
TABLE OF CONTENTS
(continued)
SECTION PAGE
8.0 IMPACT 27
8.1 Technical Modifications to IMPACT 27
8.2 IMPACT: Input Options and Variables for Cinder
Cone Butte 28
8.3 IMPACT: Input Options and Variables for Westvaco . 29
8.4 TRC Changes to IMPACT (Version 1 from Radian) for
Cinder Cone Butte 30
8.5 TRC Changes to IMPACT (Version 1 from Radian) for
Westvaco 30
-iii-
-------
1.0 INTRODUCTION
EPA has contracted with TRC to evaluate the performance of complex terrain
air quality simulation models using performance measures recommended by the
American Meteorological Society. Eight models are to be evaluated:
COMPLEX-I, COMPLEX-II, PLUMES, RTDM, SHORTZ, 4141, COMPLEX/PFM, and IMPACT.
Prrior to running the complex terrain models for evaluation, it is desireable
to confirm that the models have been implemented in accordance with the
expectations of the model developers. To accomplish this, test-run packages
were prepared and are being supplied to the model developers for their formal
review and concurrence. The package supplied to each model developer contains
the following information:
• Descriptions of the complex terrain model evaluation data bases
(Cinder Cone Butte and Westvaco);
• Summary of model-code modifications;
• Summary of input options;
• Test-run data (listings of all input and output data) for the
model developer's particular model;
• Complete listing of the model code "as run,' (for the model
developer's particular model) to enable the model developer to
confirm the code line-by-line.
Also provided as part fo che test case package are three other relevant
documents:
• "Data Archiving Recommendations for Complex Terrain Model
Evaluations" (TRC, November 1982).
• "Addendum to: Data Archiving Recommendations for Complex Terrain
Model Evalutions (Response to Comments from Model Developers)"
(TRC, July 1983).
• "Statistical Evaluation for Complex Terrain Models" (TRC, June
1983).
-------
This document summarizes the model code modifications made by TRC and
input options selected by the model developers for each model and data base.
Modifications to the models were needed for three basic reasons:
• To adapt the model to the EPA UNIVAC computer.
• To adopt particular models to accept the source-receptor
inventories.
• To format calculated concentrations for input to the statistics
system.
Detailed summaries of line-by-line changes made by TRC to each model's
computer code are also described in this document.
Computer code listings for the models "as run,* plus the test run input
and output data listings are supplied separately.
-2-
-------
2.0 COMPLEX-I AND COMPLEX-II
2.1 Technical Modifications to COMPLEX-I and COMPLEX-II
TRC altered COMPLEX-I and COMPLEX-II to accept input data from the model
input file on Unit 18 rather than Unit 5. Statements were added to facilitate
writing calculated concentrations to a work file for future statistical
analysis. These changes were made for both the Westvaco and Cinder Cone Butte
data bases.
To accommodate the Westvaco data base, TRC modified COMPLEX-I and
COMPLEX-II in three areas. The models were altered to accept hourly input of
source exit velocity and exit temperature. TRC made changes to circumvent
problems that could be caused by the Westvaco data starting in one calendar
year and ending in the next calendar year. Code was added to check for hours
with missing stability during which no concentrations were calculated.
When COMPLEX-I and COMPLEX-II were tested with the Cinder Cone Butte data
base, one technical modification was needed. The models were altered so that
only the source with a source number (1-111) equal to the hour (1-111) being
modeled has an impact on the calculated concentrations. This modification,
consistent with the input emissions inventory, was needed because a single
emission point was moved each hour in the Cinder Cone Butte study.
-3-
-------
2.2 COMPLEX-I and COMPLEX-II; Input Options and Variables for Cinder
Cone Butte
Variable Name
IOPT(1)
I OPT ( 2 )
IOPT(3)
I OPT ( 4 )
IOPT(25)
HANE
PL
CONTER
ZMIN
Input Value
1
1
1
1
1
0.90
0. ,0. ,0. ,0. ,0. ,0.
0.5,0.5,0.5,0.5,0. ,0.
10.
Description
Use terrain adjustments.
No stack downwash.
No gradual plume rise.
Calculate initial plume size.
Use complex terrain option.
Anemometer height in meters.
Wind profile power law
exponents.
Terrain adjustment factors.
Distance limit for plume
HAFL
0.
centerline from ground.
No pollutant loss.
2.3 COMPLEX-I and COMPLEX-II; Input Options and Variables for Westvaco
Variable Name input Value
IOPT(1) 1
IOPT(2) 0
IOPT(3) 1
IOPT(4) 1
IOPT(25) 1
HANE 189.7
PL 0. ,0.,0.,0.,0.,0.
CONTER
ZMIN
HAFL
Description
Use terrain adjustments.
Use stack downwash.
No gradual plume rise.
Calculate initial plume size.
Use complex terrain option.
Anemometer height in meters.
Wind profile power law
exponents.
0.5,0.5,0.5,0.5,0.,0. Terrain adjustment factors,
10.
0.
Distance limit for
centerline from ground.
No pollutant loss.
plume
..4-
-------
2.4 TRC Changes to COMPLEX-I for Cinder Cone Butte
Line Number Description of Modification
1-2, 134-137 Comments.
358-360 Dimension TRC variables.
377-381 Define work file.
451-455, 466-475 Initialize I/O units and hour
counter. Check data base ID.
736-740 Do not read met station
identifiers.
945-948 Increment the TRC hour counter.
1061-1064, 1568-1570 Transfer TRC hour counter to
subroutine PTR.
1065-1074 Write to hourly work file.
1617-1619 Ignore sources other than the
one which corresponds to the
hour of simulation.
1702-1703 Set distance to final plume rise
equal to zero.
1734 Allow for stack temperature
equal to ambient.
2.5 TRC Changes to COMPLEX-I for Westvaco
Line Number Description of Modification
1-3, 135-138, 472 Comments.
359-362 Dimension TRC variables.
379-384 Define work file.
454-462, 474-485 Initialize I/O units. Check
data base ID.
900-905, 910-912, 917-919, Changes to accommodate data from
924-931, 934-936, 1084-1087, two calendar years.
1482-1483
978-987 Flag missing stability data.
1088-1101, 1722-1726 Read in hourly source data.
1108-1118 Write to work file.
-5-
-------
2.6 TRC Changes to COMPLEX-II for Cinder Cone Butte
Line Number Description of Modification
1-2, 132-135 Comments.
356-358 Dimension TRC variables.
375-379 Define work file.
449-454, 465-474 Initialize I/O units and hour
counter. Check data base ID.
735-749 Do not read met station
identifiers.
943-946 Increment the TRC hour counter.
1059-1062, 1565-1567 Transfer TRC hour counter to
subroutine PTR.
1063-1071 Write to hourly work file.
1614-1616 Ignore sources other than the
one which corresponds to the
hour of simulation.
1699-1700 Set distance to final plume rise
equal to zero.
1731 Allow for stack temperature
equal to ambient.
2.7 TRC changes to COMPLEX-II for Westvaco
Line Number Description of Modification
1-3, 133-136, 465 Comments.
357-359 Dimension TRC variables.
376-381 Define work file.
451-455, 467-478 Initialize I/O units. Check
data base ID.
893-897, 902-904, 909-911, Changes to accommodate data from
921-923, 926-928, 1077-1079, two calendar years.
1472-1473
970-980 Flag missing stability data.
1080-1091, 1712-1716 Read in hourly source data.
1098-1108 Write to work file.
-6-
-------
3.0 PLUMES
3.1 Technical Modifications to PLUMES
TRC added code to PLUMES to write calculated concentrations to a work file
for future statistical analysis. The model was altered to allow input from a
disk file rather than cards. The meteorological data input unit has been set
to 11. To reduce computer core requirements, receptor arrays dimensioned by
500 were reduced to the number of receptors in each respective data base.
For the Westvaco data base only, TRC modified PLUMES to accept hourly
values of emission rate, stack exit velocity, and stack exit temperature.
Several changes were made to adapt PLUMES to the Cinder Cone Butte data
base. Code was added to skip the reading of station identifiers on the disk
file containing meteorological data and to read the meteorological data one
hour at a time. The DO loops on days and hours were merged into a single loop
to handle the non-sequential nature of the Cinder Cone Butte experiment
hours. Daily and annual average output were skipped. Plume rise was set
equal to zero. TRC also modified the model so that only the source with a
source number equal to the consecutive hour number has an impact on calculated
concentrations (See Section 2.1).
-7-
-------
3.2 PLUMES: Input Options and Variables
Variable Name
Input Value
Description
CONVRT (PLUMES
ISTAT
MST
DTHDZ
THICK
SIGMAF
LAT
LONG
ZONE
NCCOFF
PLUMES :
IUR
BKGRD
I GRID
ICIRC
IATOB
I PLUME
ISGFLG
MODFLG
Westvaco
preprocessor) :
2
1
0.01
800.
1
39.5
79.3
5
0
1
l.E-30
0
0
1
0
1
1
CCB
2
1
0.01
NA
1
43.0
115.5
7
NA
1
l.E-30
0
0
1
0
1
1
Stability classified by aA.
Modify unstable stability at night
as a function of wind speed.
Default value for change of
potential temperature with height
through stable layer.
Default value for the thickness of
stable layer.
Default multiplier for sigma value.
Latitude of surface station.
longitude of surface station.
Standard time zone.
NCC mixing height data used.
RURAL1 mixing heights used.
Background concentration in wg/m3.
Do not use receptor grid.
Do not generate receptors usi:
radial rings.
Changes Class A stability to Class B
No hourly plume rise input.
"nitial plume expansion allowed.
Pasquill modification to t
WINDHT
MSLFLG
crosswind spread of plumes due to
vertical wind directional shear
allowed.
189.7
10.
NA
Wind speed
(meters).
measurement
height
Mixing heights are above ground level.
—. ft—»
-------
3.3 TRC Changes to PLUMES Code for Cinder Cone Butte
General: Receptor array arguments reduced from 500 to 94 to reduce
core requirements. The number of point source locations
was raised from 10 to 111, while the number of release
heights per location was reduced from 15 to 1. One source
per hour of simulation. Mixing height set to 9999 meters.
Line Number
1-13
26-28, 523-524
47-60
84
117-119, 138-139
145-146, 1402-1403
150-151
178-180, 480-481, 485-486
521-522
552-554
628-636, 723, 728-730
651-652
656-672
694-714
766-769, 857
Description of Modification
Comments.
Define TRC COMMON block.
Initialize I/O units. Define
work file. Check data base ID.
Change loop on sources from 10
to 111.
Skip section which reads station
identifiers from meteorlogical
data file.
Change maximum number of sources
allowed.
Change maximum number of heights
per source.
Change write statement.
Dimension TRC variables.
Reduce maximum number of
receptors allowed from 500 to 94.
Change the day and hour loops
since CCB data is not in 24 hour
groups.
Change unit number for input of
meteorological data.
Read in the meteorological data,
one hour at a time.
Change write statement and
format for output of
meteorological data.
Separate the loops on source
location and release h
-9-
-------
-3 Tin: changes to PLUMES Code for Cinder Cone Butte
(Continued)
!'*»,*'/1ii"Mi'lt o^ Modification
//U--//J, &8y-89J Ignore sources other than the
one which corresponds to the
hour of simulation.
1211-1220 Write to the work file.
1222-1226 Skip output of daily and annual
averages.
1855-1857 Set plume rise to zero.
3.4 TRC Changes to PLUMES Code for Westvaco
General: Receptor array arguments reduced from 500 to 11 to reduce
core requirements.
Line Number Description of Modification
1-5 Comments.
18-20, 515-517 Define TRC common block.
39-60 Initialize I/O units. Define
work file. Check data base ID.
119-123, 638-642 Change unit number for input of
meteorological data.
511-514 Dimension TRC variables.
545-549 Change maximum number of
receptors allowed from 500 to 11.
691-700 Read and print the hourly point
source data.
1167-1173 Write to the work file.
-10-
-------
4.0 RTDM
4.1 Technical Modifications to RTDM
TRC made general and data base specific modifications to RTDM. For both
the Westvaco and Cinder Cone Butte data bases, code was added to write
calculated concentrations to a work file, and to read model input data on Unit
18 rather than Unit 5. Meteorological data is read from Unit 10 for Cinder
Cone Butte, and Units 10 and 11 for Westvaco, instead of Unit 7. For both
data bases, assignment of the PR005 parameter has been fixed to properly
correspond to wind profile exponents, not terrain factors.
Modifications specific to the Westvaco data base include reading hourly
source data from Unit 15, reading meteorological station identifiers from
Unit 10, and checking for hours with missing stability. Concentrations are
not calculated for the hours with missing stability.
For the Cinder Cone Butte data base, RTDM was modified to set plume rise
and wind profile exponents equal to zero, to set anemometer height equal to
release height, and to allow hours which are out of sequence. TRC modified
RTDM so that only one source contributes to the calculated concentration in
any given hour (See Section 2.1).
-11-
-------
4.2 RTDM; input Options and Variables for Cinder Cone Butte
Variable Name Input Value
ZWIND1 Release height
ZWIND2 Not used
IDILUT 0
EXPON
ICOEF
IPPP
I BUOY
IALPHA
IDMX
ITRANS
TERCOR
RVPTG
IHVPTG
ISHEAR
IEPS
IREFL
IHORIZ
0.,0.,0.,0.,0.,0.
1
3.162
0.5/0.5,0.5,
0.5,0.5,0.5
0.02, 0.035
ITIPD
IY
IZ
IRVPTG
0
1
1
0
Description
Anemometer height (m)
Wind speed at level 1 is used
for plume rise and transport
calculations.
Wind speed profile power law
exponents.
ASME (1979) stability-dependent
dispersion parameters.
No partial plume penetration.
Use buoyancy-enhanced dispersion.
Unlimited mixing height in
stable conditions.
Use transitional plume rise.
Plume path correction factors.
Default VPTG for stabilities 5
and 6.
No stack-tip downwash.
User-supplied ry.
User-supplied lz.
Default VPTG for plume rise
calculations.
User-supplied VPTG for
calculations.
Wind direction shear is not used
in Oy computation.
No hourly wind profile exponents.
Use partial reflection algorithm.
Off-centerline horizontal dis-
tribution function.
IEMIS
Use constant emission rate.
-------
4.3 RTDM; Input Options and Variables for Westvaco
Variable Name Input Value
ZWIND1 30.
ZWIND2
IDILUT
ZA
EXPON
ICOEF
IPPP
I BUOY
IALPHA
IDMX
ITRANS
TERCOR
RVPTG
IHVPTG
ISHEAR
Not used
0
179.6
0.,0.,0.,0.,0.,0.
1
3.162
0.5,0.5,0.5,
0.5,0.5,0.5
0.02, 0.035
ITIPD
IY
IZ
IRVPTG
1
1
1
1
Description
Anemometer height (m) above ZA,
for plume rise.
Anemometer height for transport.
Wind speed at level 1
extrapolated to stack top for
plume rise calculations and to
plume height for transport
calculations.
Height above stack base where
the wind profile originates.
Wind speed profile power law
exponents.
ASME (1979) stability-dependent
dispersion parameters.
No partial plume penetration.
Use buoyancy-enhanced dispersion.
Unlimited mixing
stable conditions.
height in
Use transitional plume rise.
Plume path correction factors.
Default VPTG for stabilities 5
and 6.
Use stack-tip downwash.
User-supplied Iy.
User-supplied Iz.
User-supplied VPTG for plume
rise calculations.
User-supplied VPTG for HCrit
calculations.
Wind direction shear is used in
O computation.
-13-
-------
4.3 RTDM; input Options and Variables for Westvaco
(Continued)
Variable Name Input Value
I EPS
IREFL
IHORIZ
IEMIS
Description
User-supplied hourly
profile exponents.
wind
Use partial reflection algorithm.
Off-centerline horizontal dis-
tribution function.
User-supplied hourly emission
rate.
4.4 TRC Changes to RTDM for Cinder Cone Butte
Line Number
1-4
21
22-32
33-41
474-475
1177-1178
1365-1367, 1695-1696
1436-1448, 1452-1453
1463-1464
1510-1514, 1546-1547,
1549-1550, 1583-1586
1697-1698
1712-1717
1738-1734
1759, 1763
1836, 1850-1856
Description of Modification
Comments.
Define work file.
Check data base ID.
Read and print the experiment
hours being modeled.
PR005 should read wind profile
exponents, not terrain factors.
Change requested by ERT.
Define TRC common block.
Read meteorological data.
Allow hours which are out of
sequence.
Change output formats.
Dimension TRC variables.
Allow source contribution from
only one source per hour.
Set wind profile exponents equal
to zeco and wind measurement
height equal to release height.
Set plume rise equal to zero.
Write fco the work file.
-14-
-------
.5 TRC Changes to RTDM for Westvaco
Line Number
1-4
21
22-34
35-36
471-472
1175-1176
1365-1374, 1736-1739
2891
1443-1475, 1479,
1485-1486
1480-1484, 1742-1748,
2897-2902
1491-1507
1610-1618, 1624-1627
1871-1874, 1887-1893
Description of Modification
Comments.
Define work file.
Check data base ID.
Read station identifiers from
meteorological data file.
PR005 should be reading wind
profile exponents, not terrain
factors.
Change requested by ERT.
Define TRC common block.
Dimension TRC variables.
Read meteorological data from
two files.
Flag hours with missing
stability.
Read point source data file.
Change error message formats.
Write to work file.
-15-
-------
5.0 SHORTZ
5. 1 Technical Modifications to SHORTZ
The SHORTZ model was modified to accept input data from a disk file, and
to write calculated concentrations to a work file for subsequent statistical
analysis. For the Westvaco data base run, an hour counter and an alternate
output format for the time period in question were added. Modifications
specific to the Cinder cone Butte data base include setting plume rise equal
to zero, adding an array to hold calculated concentrations, and allowing the
maximum number of hours in a case to equal 111.
5.2 SHORTZ: Input Options and Variables for Cinder Cone Butte
Variable Name
ISW(7)
ISW(9)
ISWU7)
G
ZR
GAMMA1
GAMMA2
XRY
DECAY
HA
Input Value
1
0
9.80
9.99
0.60
0.66
50.
0.
99.9
Description
Terrain elevation data are
input.
Wind speed is not terrain
following.
Rural option.
Acceleration
(m/S2).
of
gravity
Wind speed measurement height
(m).
Entrainment coefficient for
unstable atmosphere.
Entrainment coefficient for
stable atmosphere.
Distance (m) over which
rectilinear expansion occurs
downwind of source.
No pollutant loss.
Elevation (m) of
weather station.
base of
-16-
-------
5.3 SHORTZ: Input Options and Variables for Westvaco
Variable Name
ISW(7)
ISW(9)
ISWU7)
G
ZR
GAMMA1
GAMMA2
XRY
DECAY
HA
Input Value
1
0
0
9.80
30.0
0.60
0.66
50.
0.
467.6
Description
elevation
Terrain
input.
Wind speed
following.
Rural option.
Acceleration
(m/s2).
data are
is not terrain
of
gravity
Wind speed measurement height
(m).
Entrainment coefficient for
unstable atmosphere.
Entrainment coefficient for
stable atmosphere.
Distance (m) over which
rectilinear expansion occurs
downwind of source.
No pollutant loss.
Elevation (m) of
weather station.
base of
-17-
-------
5.4 TRC Changes to SHORTZ for Cinder Cone Butte
Line Number
2-14, 118-135
26-55, 656-684
86-94
98-114
136-171
202-210
232
1194-1199
1211
1483-1488
1797-1802
1812-1834
5.5 TRC Changes to SHORTZ for Westvaco
Line Number
2-14, 118-135
26-55, 656-684, 2043-2048
86-94, 154
98-114
136-171
202-210
1795-1817
1875-1876
2120-2125
Description of Modification
Comments.
Define TRC COMMON block EVAL.
Initialize I/O units.
Define work file.
Check data base ID.
Set TRC variable NMON»NXXYY.
Set MKQ-111, maximum number of
hours.
Zero the TRCONC array each hour.
Let maximum number of hours =
111.
Set plume rise equal to zero.
Put calculated concentrations
into array TRCONC.
Write to the work file.
Description of Modification
Comments.
Define TRC COMMON block EVAL.
Initialize I/O units.
Define work file.
Check data base ID.
Set TRC variable NMON-NXXYY.
Write to work file.
Set hour counter IHRTRC.
Change the output hour format.
-18-
-------
6.0 4141
6.1 Technical Modifications to 4141
Modifications to 4141 are the same as for COMPLEX-I and COMPLEX-II.
6.2 4141: Input Options and Variables for Cinder Cone Butte
Variable Name
MODEL
lOPT(l)
IOPT(2)
IOPT(3)
IOPT(4)
HANE
PL
HAFL
Input Value
4141
1
1
0
1
0.9
0. ,0. ,0.,0.,0.,0.
0.
Description
Select 4141 Model Option.
Sets CONTER = 0.5,0.5,0.5,0.5,
0.25,0.25.
Sets IOPT(4) = 1.
Sets IOPT(1) * 1.
Use terrain adjustments.
No stack downwash.
Gradual plume rise.
Calculate initial plume size.
Anemometer height in meters.
Wind speed profile power law
exponents.
No pollutant loss.
-19-
-------
6.3 4141: Input Options and Variables for Westvaco
Variable Name
MODEL
IOPT(1)
IOPT(2)
IOPT(3)
IOPT(4)
HANE
PL
HAFL
Input Value
4141
1
1
0
1
189.7
0.,0.,0.,0.,0.,0.
0.
Description
Select 4141 Model Option.
Sets COMTEK - 0.5,0.5,0.5,0.5,
0.25,0.25.
Sets IOPT{4) * 1.
Sets IOPT(1) « 1.
Use terrain adjustments.
No stack downwash.
Gradual plume rise.
Calculate initial plume size.
Anemometer height in meters.
Wind speed profile power law
exponents.
No pollutant loss.
-20-
-------
6.4 TRC Changes to 4141 for Cinder Cone Butte
Line Number
1-2, 82-85, 185-188
304-306
325-329
404-409, 421-430
689-692
899-902
1015-1018, 1597-1599
1019-1028
1645-1647
1729-1731
1761-1762
Description of Modification
Comments.
Dimension TRC variables.
Define work file.
Initialize I/O units and hour
counter. Check data base ID.
Do not read
identifiers.
met
station
Increment the TRC hour counter.
Transfer TRC hour counter to
subroutine PTR.
Write to hourly work file.
Ignore sources other than the
one which corresponds to the
hour of simulation.
Set distance to final plume rise
equal to zero.
Allow for stack
equal to ambient.
temperature
-21-
-------
6.5 TRC Changes to 4141 for Westvaco
Line Number
1-3, 83-86, 186-189
305-307
326-331
406-410, 423-434
848-853, 858-860, 865-867,
872-879, 882-884, 1033-1035
926-936
1036-1047
1054-1064
Description of Modification
Comments.
Dimension TRC variables.
Define work file.
Initialize I/O units. check
data base ID.
Changes to accommodate data
from two calendar years.
Flag missing stability data.
Read in hourly source data.
Write to work file.
-22-
-------
7.0 COMPLEX/PFM
7.1 Technical Modifications to COMPLEX/PFM
The technical modifications to COMPLEX/PFM consist of the same changes made
to COMPLEX-I and COMPLEX-II, plus several alterations specific to COMPLEX/PFM.
For both the Westvaco and Cinder Cone Butte data bases, COMPLEX/PFM was modified
to read receptor data from a unique disk file. Also, array sizes were reduced
in accordance with data base requirements in order to reduce the need for computer
core storage.
Some modifications were needed only for the Cinder Cone Butte data base.
These include reading the potentially non-sequential list of experiment hours to
be modeled; reading hourly values of critical streamline height (Hcrit)and Froude
number from a disk file; and accounting for the absence of vertical wind and
temperature profiles in the Cinder Cone Butte input data set.
-23-
-------
7.2 COMPLEX/PFM; Input Options and Variables for Cinder Cone Butte
Variable Name
IOPT(1)
IOPT(2)
IOPT(3)
IOPT(4)
IOPT(25)
IOPT(26)
HANE
PL
CONTER
ZMIN
HAFL
Input Value
1
1
1
1
1
1
0.90
0.,0.,0.,0.,0. ,0.
0.5,0.5,0.5,0.5,0.,0.
10.
0.
Description
Use terrain adjustments.
No stack downwash.
No gradual plume rise.
Calculate initial plume size.
Use complex terrain option.
Long-term PFM option.
Anemometer height in meters.
Wind profile power law exponents.
Terrain adjustment factors.
Distance limit for plume centerline
from ground.
No pollutant loss.
7.3 COMPLEX/PFM; Input Options and Variables for Westvaco
Variable Name
IOPT(1)
IOPT(2)
IOPT(3)
IOPT(4)
IOPT(25)
IOPT(26)
HANE
PL
CONTER
ZMIN
HAFL
Input Value Description
1 Use terrain adjustments.
0 Use stack downwash.
1 No gradual plume rise.
1 Calculate initial plume size.
1 Use complex terrain option.
1 Long-term PFM option.
189.7 Anemometer height in meters.
.10,.15,.20,.25,.25,.25 Wind profile power law exponents.
0.5,0.5,0.5,0.5,0.,0. Terrain adjustment factors.
10. Distance limit for plume centerline
0.
from ground.
No pollutant loss.
-2:4-
-------
7.4 TRC Changes to COMPLEX/PFM for Cinder Cone Butte
Line Number
1-18, 183-186, 314-316, 373-374,
429-430, 602, 2322, 4141-4142,
5960-5961.
453-455, 2323-2327, 5495-5497
458-460
484-489
569-573, 4172-4175
582-587
604-625
920-923, 1258-1264, 1298-1299
1141-1147
1318-1328
1887
2423-2426
5549-5552, 5581-5584, 5856-5858
5607-5612
5714-5716, 5747-5748
6005-6010
Description of Modification
Comments.
TRC common block definition .
Dimension TRC variables.
Define work file.
Change the maximum number of
receptors from 180 to 99 to reduce
core requirements.
I/O device initialization.
Read and verify data base and work
file identifiers. Read in the
experiment hours to be modeled.
Modifications to account for the
absence of wind and temperature
profiles.
Read Hcrit and Froude number from
TRC disk file.
Write calculated concentrations to
work file.
Write format change.
t
Print Hcrit and Froude number.
Do not call subroutines which calculate
Hcrit and Froude number.
Allow source contributions from only
one source per hour.
Allow for ambient temperature identical
to stack temperature.
Change format and input unit of
statements which read receptor data.
-25-
-------
7.5 TRC Changes to COMPLEX/PFM for Westvaco
Line Number Description of Modification
1-11, 176-179, 307-309, 366-367 Comments.
422-423, 592, 5933-5934
448-450 Dimension TRC variables.
474-479 Define files.
559-563, 4166-4169 Change maximum number of receptors
allowed from 180 to 15 to reduce
computer core requirements.
572-577 Device initializations.
594-605 Read and verify data base and Work
file identifiers.
1064-1068, 1073-1075, 1080-1082, Changes to accommodate data from
1087-1094, 1098-1100, 1297-1298 twd calendar years.
1142-1152 Check for missing stability. Set
calculated concentration to missing.
1299-1311, 5659-5663 Read hourly point source data.
1318-1329 Write calculated concentrations to
the work file.
5978-5983 Change format of statements which
read receptor data.
-26-
-------
8.0 IMPACT
8.1 Technical Modifications to IMPACT
TRC inserted additional codes within specific sections of the IMPACT
model to produce the following two results:
i) Identify and write to the output work file those 1-hour average
surface level concentrations for calls corresponding to monitor
sites. These changes were included for both the Westvaco and
Cinder Cone Butte versions of the model.
ii) Redimension arrays in the COMMON block TREFOR to accommodate the
number of cells utilized in the X-, Y-, and Z- directions for
each data base. In the case of Westvaco, the number of cells are
13, 15, and 20, respectively, with corresponding cell dimensions
of 200, 200, and 60.96m. In the case of Cinder Cone Butte, the
number of cells are 36, 45 and jWi" respectively, with correspond-
ing cell dimensions of 50m, SOmfand 10m.
*6
The IMPACT model allows for a maximum of 40 cells in the "X-" direction.
The actual grid developed by TRC for Cinder Cone Butte contains 45 cells in the
East-West direction. In order to avoid additional code revisions,the grid was
rotated 90° counter-clockwise. There are now 36 cells in the X- direction
(north-south) and 45 cells in the Y- direction (east-west).
Another modification to the IMPACT model was required for the Cinder Cone
Butte application. The minimum time step, DTMIN (specified in a Data statement
a£~
located in subroutine DIFFUS), was reduced from 3.6 seconds to J^T seconds. This
change allows the model to calculate a time step appropriate for the small grid
spacing defined for Cinder Cone Butte.
-27-
-------
8.2 IMPACT: Input Options and Variables for Cinder Cone Butte
Variable Name
DX
DY
DZ
NX
NY
NZ
IDOWND
IDOCEM
IDOPLM
IDODIF
IDOBAK
NUMHRS
IDOPLT
IDOPRN
IDOCAL
HRSAUG
IDOSUR
Input Value
50.
50.
jw? 7.r
36
45
>r »
-------
8.3 IMPACT: Input Options and Variables for Westvaco
Variable Name
DX
DY
DZ
NX
NY
NZ
IDOWND
IDOLEM
IDOPLM
IDODIF
IDOBAK
NUMHRS
IDOPLT
IDOPRN
IDOCAL
HRSAUG
IDOSUR
Input Value
200.
200.
60.96
13
15
20
1
1
0
3
1
25
0
1
0
1
1
Description
E-W cell size in m.
N-S cell size in m.
Vertical cell size in m.
Number of grid cells in x-direction
Number of grid cells in y-direction
Number of grid cells in z-direction
WEST wind model
1 tracer effluent
Briggs' '74 Plume Rise
DEPICT algorithm diffusivities
User specified background set to 0.0
Number of hours to be modeled
No contour plots
Printer edit every hour for test run
No CALCOMP plots
Hourly printout for test run
Print surface values only
-29-
-------
8.4 TRC Changes to IMPACT (Version 1 from Radian) for Cinder Cone Butte
Line Number
1-18
113-130, 386-402, 898-915, 998-1014,
1325-1341, 1735-1751, 1975-1991,
2180-2196, 2569-2585, 2813-2829,
2962-2978, 3069-3085, 3145-3161
131-265
2197-2224
2230-2239
2481-2514
Description of Modification
Comments
Common TREFR1, TREFR2
Comments, TRC COMMON verify
input files, load I, J of receptors
TRC COMMON
DTMIN, minimum time step, set to 1.0 sec.
Write to work file
8.5 TRC Changes to IMPACT (Verion 1 from Radian) for Westvaco
Line Number
1-18
113-124, 381-392, 888-900, 983-994,
1305-1316, 1710-1721, 1945-1956,
2145-2152, 2531-2542, 2770-2781,
2914-2925, 3016-3027, 3087-3098
125-260
1969-1984
2153-2186
2443-2476
Description of Modification
Comments
Common TREFOR
Comments, TRC COMMON, verify
input files, load I, J of receptors
Set unset variable
TR£ COMMON
Write to work file
-30-
-------
APPENDIX B
STATISTICAL TABLES OF MODEL PERFORMANCE FOR WESTVACO
Table Page
Westvaco Comparison of 25 Highest, 1 Hour B-l
Westvaco Comparison of 25 Highest, 3 Hour B-9
Westvaco Comparison of 25 Highest, 24 Hour B-17
Westvaco Comparison of Highest by Station B-25
Westvaco Comparison of Second Highest by Station B-28
Westvaco Comparison of Highest by Event B-31
Westvaco Comparison of All Events Paired in Space and Time B-34
Westvaco-IMPACT Hours Comparison of 25 Highest B-61
Westvaco IMPACT Hours Comparison of Highest by Station B-63
Westvaco IMPACT Hours Comparison of Second Highest B-66
Westvaco IMPACT Hours Comparison of Highest by Event B-69
Westvaco IMPACT Hours Comparison of All Events Paired
in Space and Time B-72
-------
*-« O UJ
X C2 U.
o o
O -O
rD o
• o
C£ •*. C.
0. w"< * C
u. en
u. o
vD «•
in F-
t-H K) r-« <
m
(Tv
to I
O ^H
«-H f\J
I o
UJ «/t ^*
CJ UJ O
UJ UJ t
U > CO
•- O
Q U ^
._) IT-
*-H •&
I IT)
•— x
UI
CO
ir> i: a: u^
OJ UJ •-• '_J 31
u ^ x
U- -*7 CL t-
O O 2*
LJ co
> CTi
*» O
a.
in
cr
B-l
-------
c
o
c: * D
^ a: -•
_ .
c r= o •-
c -a uj n _j > IT-
> s: a. ~ D o a:
UJ O O U Z- O
1 -« ? i-i O (_>
O *- —• (r •• X «I
«- < « _J I >
, u^ — C 21
U <. U,
o o
LJ 00 *-
U UJ O
u. «r ce
a: a; a.
UJ LJ I
u, >
u -a m
C u —
o
UJ >
cc >
«l O
^J CT f\J C~> _t
a-
c-
CM
vD »
^i O^
I «•
ro *•>* K)
C\J O CNJ
r- -i in
in co Lf>
O
U
O
_J
o
o
LJ t/> •
ui >* m (si
or ^
to os.
ifi »-
o » •
2" <\; 'fi ifj
*-* •
"3T V C\J A
H^ O f-t (
CV IT ^ :
* • •
?*• ro
o- cr .-, --.
r- cr ^- »_•
vo m r- in
tn to (/)
-------
Uj O Ct
CJ C/) UJ
Z »-H cr
*i a: CL
•-» o o
UJ C
O UJ
Lu UJ I
u. > to
U- «tf m
>-. O
C. U- —
o
Q_ fO ~
« 2
0*0
O
t-H
ct
ixJ
UJ UJ (X .
t/j Z) O *— >. a: C.
CG _J cr
t- t— L^l Z **
w 2 t- n o o;
UJ O O U ZD C
O t- i-4 £K •• X
X iv CD > UJ r-* ^-
h- U- il' en >
«s o
r-- rr> KI m ^ a.'forr'a)r-^Hmt\jo^c\j*Hc\im*-*
fO OJ C\J rO »-*
ooooooouooo
t/.' I/1 I/) (.0
r- «-< LT
in GO to
o o • •
cc :z to m IT
o ^- «
C\J h- X r-.
Q rH ^ ^£
»-. on c/i c-, t"i
OJ 00 CO CO 1,0
B-3
-------
o :D (_> »- o —i
LI •— O LJ y ^ aj
[X CO CL O ~-t
uj uJ cc. cr
> s: CL «! «-*
h- >— co s: w
co a t- 3 c er
u_ o o o o o
X —i 2* -^ O O
,
U1 2 a. a; o u*
c\ ui i— o x LJ at
u < u_ x
LL. 2 D. H-
o o ^
on-
c^ o
o: CL c/j
o o
UJ •"'
LJ UJ D
^ O LJ
LJ «a a:
a: a; c.
L.I UJ I
U. > «
u^ — _j
LJ o «i
> LJ >
UJ
«i cr ^
fO O
tr ^-t
«-» i
tr
I
) ON r- vft o in o- i-(
M a c • i
*-« «-f *-» <
KJ K> »-i i
•-< (N- »•
if> c\j oj r-
A, f»- O CV
i\i r-- u, r
cr i-» *-^
tT IT! -H v
•-« CO CO l/l
OD CO tO (/) t
B-4
-------
UJ
(X •*•
a. K> —
* 2
O*O
^ -r -*
•a x. H-
CD «a oo
O Z3 O h-
UJ <- O uJ
a
o
Q.
o •-*
or. (/i co .-t
U-l UJ Q^ < ^ X.
t/j ^ C H- ttl *—'
CO -J ^ UJ CO
o «j u^ c: > c^
> s: t-* « ^H
co 2 h- D ^H tr:
uj o o «- r> o
X — 2 ~ O U
o H- •-. a: •• x <
_j i >
I iv
• U.J «"H t—
a:
o
Lt. O C
O 00 LJ
a a: CL
UJ UJ I
I-H O
o u **
o
L-t UJ
CD >
in co «-» .
J tr, ^> «jj a-
•-< CNJ o c\j
ooooooooooo
co oo v, co oo i
C\< rO £T
OL rr •-»
vJ3 K) r-.
f- --< lO
in ^x; 10
o
_j
o
a o-
x — a K-
a LT * • 1^1
— ex •
I r— 'X t-»
^ *- -x
*—• LO tO (/^
CD 00 ^0 C/1 i
B-5
-------
d D U M- (!*•-•
l.i — O U i 2T eC
> _J Ol I-H CT-
or (/) («>«-»
i/ 12 c j *- or tr
QU -j ^. in .jj cc
c cr>
> 7- r -*-
LJ bJ O
C£ a: a.
LJ UJ I
IA. > CO
U_ UI >
cr >
«* o
ortoccfT1 ^/iin-* cMtO1
o a> CD t.> c:, o o .-i o i
m
f-* fO
I r\j
« CVJ C\J Cv) CM r-t
tr
J)
0-
VD K> —i
r- *H tn
in cc ID
o
u
o
-J
o
uj >N in co
c. 5 v.
i" or
in H-
o
j u i IT
CNJ ^. vT
ID Od »O I
) CM N-
*-> I
o
a:
o
I-H W W) I/O
-------
o z> u H-
l.i ~ O l.l
,
CC* _J
O
h- L_
OJ LJ »-* C
O «1 Lu
U~ n Q.
O O 2
u :D
C CM
CO C
t-t CO
^ «-• a:
O O
LJ <-
LJ CO *•>
(_> Lij a
UJ UJ I
U- *I CL
»-. O
O U. *-
o
or •-» _j
UJ CJ <
> UJ >
a
LJ UJ
O > LJ
tO (/"J LO
CD tO tO CO CO
-J -J -J _J
O '-J '_> LJ
B-7
-------
•-" DC
a; G.
o »
• fO
<—i KJ c\. ocu
f— CSJ
•-• cy t
O UJ Q
Z O Ul
i*-1 co
U, <5 CT
•^ C
O U *
a-u in •»-» o a1 j~* c •»• o o c_ o
1T> t f^-IKimfOO-lfl*-«K) I r-(
•-« •> oo m «• •-» o^ <-« i i
—1 in ( lll't
a: •-
Q- fo ~
* ^
Q * O
2 s: •—
< X. K-
O _»
CC i^/
UJ UJ
r>
00
CD
OO t—i 0^
a: <* «* >^
v-j t— a <—
^ NJ uj oo
o «i u1 t~. »— > cr
> E: cc « ^H
t- >-. oo o —
M ?> i- ^> -z cz
Ui O C CO ^J O
X -» ? —' O U
tj) ^ i—. y * • T <[
1-1 <:
X rv O > uJ •-( t—
t- Ul C> V".
U"< ^ tr ct o uj
f\. LJ "—• O 51 UJ 3
O «i a. I
U. 2 O. K-
c o z
u 3 a:
Of •-• _l
UJ Q «S
> U) >
«i a:
OJ «• O
aj c\j a^
\£ »-i r-4
I •-. I
t
O
2
O
o
a:
r- a c r- n
ooooooooooo
o
_J
o
•n co
*s.
o i
*-i 00 V) 00 00
OJ CO 00 00 00
«J.
-------
o o
o w
fO vf)
O *-H
• c:
o
UJ
o: ^ O
CL K- « o
* 5* *-•
n * o or
« «•»
UJ O
2T O UJ
_J i-« fT\
rr c/i O *~i
UJ l»J Cf «G •*•«.
co 15 o cc e.
Q_> _l UJ QJ
r «-' 1,1 •> T*
> 5: -a r-t
UJ t/1 **
o ui a
.
•-• c-
D li »-•
O
w.-' i-i C>J I
fn Hi
i cc
in i-
i a.
to <; a: i*j
C\J UJ »-« UJ I«
(_) < X
U_ IT CL »-
O O *"
o
UJ UJ
o t— LJ
ui >
«• cr
a
ui to cr >
-------
UJ O CT
CJ GO L1
zr •— ct
<: a: a
•-» o o
u *-
CL **•
a r--> «
UJ (/>»*•
O Ut CJ
cr a: Q-
Ll UJ 1
a u. -
o
OL O
r- 1—i
—• in
I <*-
a ry —« tv> r- in «-n
t*- in cu C"' CM t t t
titii
o r> o *- o
lit *- O lit ^
> _j on •—•
oo z- t- D o err
u. O O U ZJ C
X *« 2 *-* C U
ts i- ~ or •• X -a
I—
T rv (Ti > UJ »o ^~
^~ LJ O l/l
u- i; LCI Lt o LJ
o u- — o a. LJ :*
u < u. x
U. ? Q. t-
O O 2
CJ D £H
2- - C
O CJ w_
V O
»-. V)
< U 2
rf >— _j
LJ o
-------
ce ~*
CL fo ^
o rD u t-
hi «- O UJ
o; c/i
uj uj ce XT
»- i-i cx;
Ui O O
o H- i-« a:
T (T D >
in 2 u> a:
cv uj »-* c
o a <_
UJ UJ CXi
_J > (^
a. o
c o
• • X
LJ tO f-
C- en
O U,
S. UJ JE
UJ C O
U C/1 UJ
(X CL CO
•a ar Oj
> o o
LJ CO -"v
<_) LJ t_J
.
»-• o
O IL ~
< O O
a: •— -J
UJ c «r
> cr. '
-------
;? -v
i 0 C
«t o: o.
*-t o c
o c\j c\i 3-0 .-( r- -
UJ C/3 **
O UJ C.
Z O UJ
UJ (/•
u- «r en
-» o
C: u- —
o
U3
vjj
•x a, r- c\, o r-
r- if' m ff- ro o KJ
II
CL
Lr «Z I
Cl O O I
u i "*• o i. j 5~ ** cr
> _J (/] LL. •-« O^
O- V: G. »1 —«
»,'. ^> C •- X cr t-
£Lj _J <. uJ UJ CC
V, z. t~ Z) o a:
LJ O O U D O
o
UJ UJ
ut >
CC
a.
X V C > UJ -O k—
t- U' O L'
ir> z. u: a. o i—
cv L_ i— o a: LJ ^
u «a u, x
u. «' j. t~
o o :
<_j r
O CM
t/J C
en
o
Q
UJ UJ
C > U
< a: Z3
nt: d j
s:
o
OOOOOOt^OOOO
/ t/) l/J (/) C1", (/> (/-- V) CO (/)(/)
B-12
-------
a: *•*
O- PT> *•*
Cj * O
a r> o »-
ui ~ o ut
> ^> co
c- z;
CO _J
LJ O O
o i— >~t a:
*-• I*J
K- UJ Cj
if) i; a. a: O
o o
CJ *-
UJ CO *^
LJ UJ O
2: o uj
UJ < a
a: a: CL
bJ UJ I
LL > CO
u UJ
^ cc ^
c -J -j
ro
o ••
* j;
PJ K>
-------
c
L.
O_ —>
«i in i
o u_ :
U. C O O. O O
X H- .? t-i O U
•-» <
x cr o > LJ to h-
>— UJ C- V^
in ^. a: ce c- -J
C,' L. t— O ZT LJ 3
O -3 U. X
O (M
t/3 O
M-. 00
> o o
t > *-
UJ 00 —
O U." O
^ tD LJ
LJ «X O-
o: cr a
UJ U, I
u_ > t^
u -a cr
o u *-
a: —• _J
UJ O «
<\J --* JO ^
•-( o r*- r-i a, CNJ
10 •-< *o in r- r«-
>^ fr> h- ^n or
- ro ou to (\j c\j
oc-ooooooooo
t/j LO (/i l/J 0" CO LO t/' CO 00
B-14
-------
Q
O
LJ O OCKOO
I 1-1 Z —i O O
o *- »-< o; • • X
T or O > lil « K-
L") *1 CH ^ O Uv
(J = o:
r- — o
o IM u.
oo c
o
u
H *v
U» C1 r*
> o o
LJ •-
to *•*
UJ o
CD LJ
a: or
UJ Ul
U- >
UJ LJ
O > LJ
< a: r>
QC J J
L t' -.
i o co cr CD OD vU «-< (r
cj cj K) i_- «-< c\j ir- in
in t-i to
K)
00
CO
C\J
t
CM I
i
«.' c\. r\j in vfi P) o c m <\ •-»
r-* ^H •—< »-« »H C\J T-4
a t-* cc \n
otofor-
r- *-i in 10
OOOCOOOOOCJO
J t/) CO t/5 C/) CO t/1 0" C1"1 t/) t
B-15
-------
a: +*
0- NO «
O
o
3C.
uJ
a
0"^ J "4 r-si uj cc
o «a u • o *~ > c^
> 5; er: — V?
(X
i:
o
Ul
-------
£X *•» Q
CL fo ~. o
Ct * O n:
?r s: h- i.j
(1 <3 **•
i.i — O ^ nri
tY 'SI f> ^-*
0"J ^ t— tr
UJ O Z>
C1 H- H-. X
(\- UJ «-i l*J
— 13 UJ
X C3 U.
cc -.
U_ O
or Q, CO
O O
* A
iu a
O UJ
o a- i
UJ UJ GO
5 U_ CD
U. O
O LL w
O
CM sC
I CO
f\i *>
r-l O
I X
O
X
B-17
-------
o
o
CO Z* »— ;3 O t£
UJ O O U :D O
n: •- z "i o u
o t- •— a: *• X «t
.-. *T UJ * K-
t- U CJ CV C"
u'i £. a. a: o MJ
oj LJ «—i o £ u- i*
CJ *3 LL. 31
L_ r? CL h-
o o ^
O f J
C/, O
I—( t/)
i*J O Ci
O CO UJ
2! •— or
o o
o ~
LJ co ^
o LJ o
^ e uj
UJ CO
u, < m
UJ Ci <
;> uj ;
-------
> _j oo *— tr>
(X. CO •-> LO -H
o Z5 O i— x cjf c-
co —i «a: uj LJ CG
t- *i ui O _J => ^
> 5: a.<«-.
t- »- co 5- —
to 2- *- ID o o:
UJ O O O 13 O
x -* ^ — oo
O »- •-• a: •• x
T rr O > LJ * »-
»~ LLJ O CM t/5
\T' '£. CL U. O uJ
C\J LJ *- O Z LJ 3
O -4 LL. X
U- P" O. h-
O O 2
O Q
t/) LJ
a: a. co
«x 3: a)
> o o
LJ (/> ~
o LJ a
C. I-
-
LJ LJ
> LJ
LJ LJ
O > LJ
cr. :>
o
m »H cr KJ .3- -
O«-H<"2«-«CV
rO o *-t o en
en -a- ff^ i
C\J »-H I
If Co
JCT'a-L
i i
> «-« j- rr. r\i
c *x.*-i
CM «-H<-H»HC\J«-H»-<«-*
ooooooooooo
B-19
-------
a: •->
0- K, ^
* ^
CL * O
or: \j~< Q- t" •-(
UJ UJ ttT ^ "V fT>
c^ 2 >- o o a:
UJ O O O — O
X - 2 - CO
O *- »-, (T •' X <
X (V Q > U! 3- t-
tJ -3 U. X
O C 2
CJ ID CL.
^ *~ C
O f\. U.
(/} O
•-. CO
ac
tu to *•*
o t— a
ui a «ff
> LJ >
U-' UJ
tD >
<< QC
> cn >
-------
Of -*
a. »o *»
z s: i-«
•0. X. >-
tD _J C/"-
CC IS>
uj uj o: *i
o
o
a: c
UJ CO
uJ O O
o i- —- c:
X rr o :> uj
»- U1 O i
U) 2 Ct it O
OJ UJ H- O E I
cj «ac u_
U. ^ Q. I
O O Z
O ^ i
O CO I
00 O
3) o
o o
X CO U!
cc CL co
o o
o UJ o
2 O UJ
UJ (/>
U. UJ >
cr.
UJ UJ
O >
-------
cc «
D- K, ^
cr "*
to ^ >- 3 _j a:
LJ O O CL Z3 O
x »-» ^: —« cu
O h- i-t QC •• I <
X er- n > LJ a- >-
H LJ Q f\j t/i
LO 2 LC. U. O UJ
C\J LJ •-" O i: LJ 3
o •« u, X
L- ^ CL h-
o o s
2 -• o:
« or a
or D. t/?
o o
CJ »
UJ l/j <*
U LJ O
LJ < a:
a: a: a.
D U. »-
O
o
j UJ
»~
U
_
O > LJ
«i a: r>
L. t/
> or
) in rr« r- .-H oo *•« t-< if)
ooooooooooo
B-22
-------
uj o n
O W LJ
z >-« a
« or a.
x: cu
o o
CJ w
o: a: o.
U.1 LJ I
U- i.i «t- >-
t- L- O t\- f
u~. Z uC tt ^ UJ
(\J UJ -« O £1 LJ J
CJ <£ U. X
u_ s?1 a. t—
002
UJ ID QL
2 »•- C
O OJ U-
31
O
CD >
O
ooocooooooo
B-23
-------
UJ O Q
<_> t/1 U-l
O O
o —
i CD «-t CM -H in C,
^ C UJ
uj CO
U. «3 ff
"H O
O Lu w
m i ^ i
CL ^ *•» O
* 2: «-H
£_ * o £jr
^ a: ^ uj
«4 V t- D.
O «
O — O •— O -M
Ui — O UJ ?• GC
uJ u_> Ct- «tf «!**>»
i/ i; c_ t- o- c
LL _J «I fsj U- CO
C «J l,| O *- > O
> £ CC «I •-»
*- •— CO O —
c^ i: (~ z> x a:
: QC £t o u»
j *— o s: LJ 3
I 1 U_ X
O O 2-
LJ D
V) O
•— t/)
ui UJ
O > UJ
C'OOOOOOOCOO
t/1 CT (,0 0" t/1 O- t/1 CO CO CO t/,
B--24
-------
O V. >
T O — rtf CC
T LJ I,J X -v
ft -1 *_
L_ a _i t-t co
O -d tTi
a > rH
o *i z x
O^ O K- O
1 IX- LJ O
O O *—
o
Ul
a:
( ) GO re-
y • u
a: ec u-
LJ
OC _J Q
L> O *-*
OH 00
Ct O _l
o >- u- r>
2" ci
W LJ UJ
C > _) LJ
»-« *-i *r a:
U O >-» O A
«f OO *—< t/)
LJ Ld
> U-
< C/1
U- *— a:
O
-------
u- a. —i *"> ~
r f -r
r^ •> ~+
r.* ^ L , —
CJ (-0
t/> UJ UJ
O o;
«£ uj t/> CJ
O > — J UJ
HH t-H L.
-------
I/ D
•3 tr
B-27
-------
z o
LJ O LJ
t) co or
y •— o.
•tf »v /\
or o; u.
« cc LJ
LJ o o
CL o o
to of u.
Q' CC UJ
< o o
1.1 «t t j
CL
Q
O
UJ UJ _l
O I- «
o: _i Q
Ul O «-«
:» to to
c/i «"
Q. O I
O
: Z X
O i- O
j i— o: «x
•i O >
nr- li >—
i- to
I Z Z —
u/ C i
c >-
t^ m
CJ 2 to
cr o —i
«i ~ <
o >- u, rs
z «r o o
< M M
*- > to
CO LJ LJ
£ UJ to O
O > _l U
** ^^ ^ £C
H- U. I— 3 0.
«J O "-" O /N
« M i-i
K O V) IE
u. n_ LJ o
to CM •-< »-
\s\ ^H cr cu
i -a
K1
c*» c\*» cr» CM«*
^cr CID tnr »o r-
ir cc "< LI
l r- rj cc
CM i «
I l
CC
CM *
c^ in
in CD
CNJ 00
C\J
I
CM
• 00 •
o o cr
-i l ro
CM oo
CC vC
OC f-
Jl O
U.' C
O uj LJ
I LI <
• to >
: m
o
UJ
O
cc.
u «i
cr u_ »- cr
z o n «i
2
LJ
O
o
o
i:
a.
r
o
-j
a.
r
o
u
m
ac
a.
r
o
u
o
I
B-28
-------
H-. <3 CO
iV Q. m
* fl
* or
iii LJ or
y a; o
r~\ o
(^ O -H
UJ X V.
z:
o
, U O
i U >-•
o i- u r>
s -J UJ
CL o co c:
LL. CL Ul O
C —
3- O -« —«
B-29
-------
C",
0
»-4
or
UJ
CL
o
ro O
t— * * CL
'•i y ui
X 0 *v =•
i*! UJ i£ «* «•*
t~ h- -3 -H
X ( ) *^ tY* a"
f-t C3 fT
r> n (^ o — *
?• Ui UJ X "X
i j tt Z) <=-
o u, _j «• oo
ui «M
^ U( »"
u. - rr: «3
O > «J O >
to o^ n* it ^
•— • L-J t— CO
DC UO ^ il t*J
o f-
c/:
C\J
rr >-
(/. C
C
UJ
CO CT;
i rr lii
«i a:
«t
UJ UJ ££
K CC O
^ ^
t- D OL
O C3 Uj
O C/)
0"
*
Q ^ 00
rr n _J
t/3
VJ UI UI
O or
^^
^ LJ UO O
O > _J UJ
t— « i— i «i a£
t- U. »- D CL
CJ O *^ O /\
<£ tO 1-1 CO
O-' O t/» fTi
u n_ iii o
c, *-
*
UJ
Uj CJ
t? 2
u.
ZD
EC 3; — 1
U- U, -a
r«-
on
oj r>-
i-H «•
i T-C
OJ
i
vD
1C
C'J
r-»
i— (
i— t
x:
UJ
_j
0.
5-
0
O
o
_H
(=
N-\ •
o
tn «
• <-4
r^
U>
•
CNJ
CTv
-.
i—*
X
UJ
_l
Q.
s:
o
i_>
rr.
t*^
Co
ifi •
f*.
o »
• <_J
(\l
c->
m
»O
IO
f—
tf
o
tf>
10
r-
fO
*
.-*.
•sT
N-
OJ
^~
00
fO »
o:
\D
OJ
«v
vD
to
^
rr\
f-H
CC
K) »
OJ ^
i as
-a-
i
01
lO
OJ
t-t
I—t
rH
jj-
f-H
fl-
or
in
C*
rrs *
00 .-»
m ••
• CM
^t-
*-i
•
^
•-4
1
in
in
CM
t-i
t— l
z:
Q
H-
X
lO
K)
CT
•
0
O v
• rO
el
C-
•
[0
(T>
fO
K
en
»o
CM
^
t~i
IT)
ID
rr.
•sr *H
»H
rH »
t-H CO
r-
r^-
<«.
ji
ro
^
ro
^_,
._«
to
to *
"«0 K>
i cc
ro
,_!
1
O)
in
rj
fH
t-H
in
UJ
sr
^
_j
CL
K>
xH
C5
f^- •
i— <
C3 »
• in
r>-»
o
*
a:
ai
03
r*.
>H
ON
tn
*o
v/1
ITx
M
til
rH
«-
Jl <-<
O
a: »
K-|
tj.'
in
tan
r-
OJ
*-.
^*
K)
|
cc
r*» *•
if) r:
1 t\i
f-i
i-H
'
O)
in
CO
*-4
F-<
E
n
V
•x
UJ
_J
a
y
o
u
rr,
— .
_t
r\i •
10
n «
• n\
^T
0
•
r-4
f^
cv
a
f->
C3
O3
h">
^
C
i«l
vD
\i •
fO K^1
1 *
r-
t
vfl
U"J
OJ
f*
T-4
NJ
h-
a:
o
X
to
L
X
B-30
-------
* _J
O P» f O -H
o — »-• r* cr
*-, O
in —
ji u. r1
y O LU
B-31
-------
*J U1 L-
1- Cf li
Q ^
CC O
«3 t—i
LJ
UJ — O
o^ MI ** ^
a * u» »-
* a- c^
o y- —• i.i
2 -^ *- a.
LJ »•»
C_) Cl
UJ P» UJ
or u» i
I
* (T I, I K-
*? cr. n
a: uj —t
> GJ ,
<3 C
C-
5
O
o
C. U. "
=: o
U-
CL
o
X
B-32
-------
n ii-
o
to i ^
^ M a
-J (V Q
a: LJ i
«~i O
o —
_J (-1 _J O.
X
*—
2
Lt i.i r:
> <-T
o
o
B-33
-------
rtr o r;
• u m
o **
r-( I oo -o
Kl LP,
o ZT y cr en
!~ Lu LJ L
^ LJ «: X r>
D
ut UJ
B-34
-------
rr1 >-< «-H (V
U h- «;
O il O £.
u- L.) L^J s;
^ o ct: X ZJ
o ;?• •--* H- LI
C^ O o «
-------
c, c- fo rsj co ro o
•--•a- ceo- tr-*-
> u. c~
•a —• O
.
u. 31 O
>s .-, -»
c o t~ co
r: ID <=• H-
c
_j _ t- .-, O u
_j t~ o: •' X «r
o
u_> U.
O >
> CO >
-------
o to
LJ *
Z5 o f- >s. a. <-
o: _i z «a. uj uj cc
UJ «I «l O I > CT^
co > CL «r --H
CD UJ CO Z~ ^
O -? s: D o o:
o —• o LJ 13 o
_J •-« t- >-< O O
^T LJ tt O i
L.J uj o z: uj
z u a: u, x
o z: "-• ** >-
CQ o «i
—• LJ a. o:
c; o
•« cy u_
a o
o
O ;? a:
CO Ul
c u.
tr CL
Lu I
=> U, Q-
«: *-« o
> LJ
a: !"D
UJ CO «l
> 00 >
C-J i-i rO Ki OC
UJ •
s: -t
»—' i-. IT.1 vT
B-37
-------
o
Q
cc * z:
Q. s: o
. i-. « a.
O O »- IA>
2- ID Q. «J -i
CD LJ CO 5" *-
O 2 s: 2 o o:
o — o o ID o
_j •-, H- »-i O O
_j t- a: •• x «
•c <=. 2^ *i _) t >
U, *~ O C^r
o z ci a a —j
LJ u c E: u. 3t
^ u a; u, x
o ^ —> *- »-
VD O
« ~ n
o H- •-•
or
t^ u,
u ^
u o
a u-> i
Ut ti- t/1
uJ
<* a: n
a: u1 _t
D U_
r c
vO oo r\j
^H OJ fO
u, IT.
U~l IT.
O
z
o
CJ
u
d
O £L
CL
to I
o a.
ooooooooooo
o
_j • •
O Cj f J U L"
O •— V C\J A
*a C- U- u
-J -J — I
O (_> U
B-38
-------
a z o
, •-* — L.
n ti H- t/j
^ Z)
n oo _i
uj
o 2 z: r)
o •- o
—i •>
LL t—
o ^ Ci a;
UJ UJ O
z o a: u.
o z- »-< ~
fJ O «i
«-• o a
a:
«s cj
a o
Z 00
•-, en
O -H
tr
•t < .-I
•3- **
ri a:
-3- o o
G (_)
.. X
Ltl ^( t~
tr
O 2* u_ or
«i ^ c
o —
ID > LJ
o: >
a. r-
v co r-
^ —« *
If K
o
2
o
-------
o z -a
or o =>
« —• o
o i- >-
cr- «- *-« cr
•a rr o-
cr LJ I
^ a: v
vfl t^
o
o
r* 2 «• »- o f-<
.
a: L ? J-
v > 3: cc >
0' a a c\j
a C1
o
h- rO tO
«• f\) K)
CM O J3
a. iT r-
r-i CT ^H OC
c
o
o
o
O
a:
ooc-coooooo o
_
t- U O O U
B-40
-------
CE * 2
CL y o
a.
o •
o c/>
> ^
I s: ui
O 2. c: a: O
LJ LJ O 5T U.
2 u a: b- x
O 2 t-H W f-
to o «a
t-i U U. CC
cc c
«3 CM U.
0- O
a 2-
a: o
LLJ »-.
o o
o
a:
< a: a
or LJ i
ui u> on
> U- ff
Lu
m >
-t o
cr IT o r\i cr-
r- CNJ •-< c-_ t
a: ^ c. ^H
OOC'OOOOOOOO
(M r- vD
t\i ru *-*
cc cc
I t
a
z
o
LJ
U.J 1 -X.
CL 05.
CO IT t—
O — v '
CO IT
rH I
B-41
-------
O
UJ
«j a: =
> CL >
«t C
_
5 C L
z: >
—< x
o^ CT^
vc a
co io f- j cr r- c j
' IT in •-« *o
ooooooooooo
\i) 0s «•
CM PO rH
o
o
o e/i •
u- >. u, t,"
ii- a x
o. o i,
to in H-
cr r\j
<-• <\'
c\j uo ^H r-
O e^;
ix
O X
*3 D U U
i V) CO CO 00
B-42
-------
o ro o
uJ * •-«
ty* * ? (f
Q. *- O UJ
o t*; i— »•» *»
Z D
I.J rUJ « «Z X.
O. _J Z UJ ** CO
U! « ^ >• (Tv
CO > «X IU *-i
m uj «j »—
o 2T E tr en
O »-4 Z) < O
_J «-< h- O *- »_/
_J t- X «I
ar »-. Ki o- K-
U. t- -a to
o ^ a St u>
LJ LJ U.' E S
I/; O n-
X O LJ
•c UJ L-
£ a u-
U. *-*
O
O ?•
o; o
O u_
O
u a
UJ 2 LJ
c LJ tr
u. a.'
O
III
fit UJ _l
LJ (/} <
> m >
•« o
•-< 10 -H
CC I
> *H ON
in
in
in
i
U.
a.
o
CL
Q.
C
O
^1
CL
O
O
I
V)
B-43
-------
~ "I uJ
•v. —« a.
7* O n (.^ n- n c_
Lt _J ^. b.J — X
o zr 5: a- or.
•_ -« D - LJ
j-' U U E*
LJ es: x ~
o o
u u
G or uj
O C
O l )
-------
-i i-1 a
Q H- HH
Z <: to
< •— LJ
h- > tr
00 U1
C U.
o
(J O
Z LJ
< a:
a: LJ
o m
LJ *
or + Z
. •-• M. a.
Z, O I- ,
tr _i Z •
O 2T S- O
C •-• O
O Z CJ ir
LJ Li O
Z D OC Lu
O Z •"• *••
t/) O . CL C.
uj LJ a
_J > O-
Q. 0 r-t
O CC
o r> o
o o
•- X
LJ 10 t-
o v~-
> OJ >
B-45
-------
c. to
UJ «
CL. * 2:
2. 5" O
C, C *- CO
2" z: «? t-
O ^H
(~> On _j i—• o px
LJ LJ < —' rs c, t— >. a. t.
CC _I Z fT'
tO .-> 0- «J r-t
i ^J oo S. —
o z n r> o o:
O — O LJ Z> O
— . n t^
O ^ t_ O. O uj
LJ LJ o z: LJ ;s
S O Q. U. X
o ^ —. — H-
(/; O <•
•-. u G. a:
G. C
s:
> a:
i uj
O L
O
u o
LJ 2 LJ
c, LJ cr
u. c:
« *-• o
o **
or LJ
uJ co
•o r* -H
-------
C_J K}
LJ *
or * Z"
a. s: o
v. •—
n L*i K-
z — «
o
uJ LJ
O
o
Ul
^. Q_
a.. LJ
C 2- y
O •-»
a o
s: w
o
> i/i i^ »-i cr
i a o f-
. i— *. a. t-
: CT-
C. O <-•
7 o or
O O Z3 O
f» O U
a .. x
< —« lij
H- > or
c/i u:
o u.
o
UJ Z! LJ
o ui a:
CD >
« c
f in --« «•
*-4 C- in r-i
in r^. c- a-
i- o-
to -*
«• -o
in
cc
B-47
-------
a: O 3
u_ cr
CJ
o
tr. « s
a. E o
r> -
*- O U!
o t/>
> Z1 i- >-
££ _J 2 -^
it) *-<
.•D L.1 y -a-
O if E ID —
O — c »
EC t_
UJ CC
z> o
o u
I I.I I
U. I— O
O ^ CJ u- O
Uj L_ o i: i
2 u (r u.
i/> o tfl (/)
B-48
-------
CO
_J
0 2 «
a: o 3
-i n
a i- >-i
"Z «a CO
a:
O. «
D
O
u
a. uj I
1.1 u, co
> U. CC.
1 i-i O
o ••
LD "
"v ~ ~> CL
O O H- CO
z r> -
«J ^ U UJ
o co t-i cr\
a v; _j o ^
LJ UJ < z; o H- ti. f->
a: -J ^ «t ID LJ cc
UJ «a <. C. Uj > 0-
O Z- 21 O _j LL
a —« o a. D o
_j „, ^_ _ O o
_J »~ CC •• X «a
<
CSC •-< > UJ hO H-
o i; LJ a. o Uj
LJ LJ O ST UJ 31
2 u a: u. T
O 2T —. — H-
oc o «t
^ O CL Ct
Q; O
< UJ
« a; z>
01 UJ _l
uj co a
> CD >
t/)cototo^ooo
B-49
-------
CJ
•- — Ci
O w> O
UJ * •—•
OL * 2 cr:
a x: o uj
N. ^ .-. Q.
Cj O H- 1/5 *^
Z^ i— h- Oc. L.
Ct. _J ^ 51 «T *-
X U_ CO C. *-
o Z Z — »- a:
j *-• c a r> o
_j -4 >— —• o LJ
_j *•* G^ •• X ^
a: — > ui M-> »-
o z —• —
K-. LJ Q.
C- O
s: i/j
a:
00 I; 1
c. u
o LJ a:
< nc CL
a: U i
UJ L. (/>
> li- cr
< H-. o
o w
t-t C\J O f\J
I CVJ »-i *-i
t I I
B-50
-------
o K;
ui *
cc * a1
a. i: o
O CD »— to
« — O UJ
O 00
O 00 _l
UJ UJ <
o. _l
Lu <1
o
o
o:
UJ
a.
: NI uj a,
a: «r ^H
CD UJ t/1 O *»
o z s: n r ce
o i~i o to D o
_j ^^ H- *— o o
_i i- a •• x «
« « z « _i i >
rt' »-- > bl 10 t-
O 2 >- >•
00 O
D u-
O
(J Q
> u. cr
•a cc
a- UJ
Ul c/>
> CD
T-I CNJ K)
B-51
-------
-^ — c
(S; > <* 1,1 —i
T UJ _J —
o 2 s: fr m
O ^ 3 - >
*r r-. ^j- ^ »_
u »- ^v
-------
CJ
i-t *•» CJ
OK) O
UJ * •—
ff * ^ CL
CL 51 O uJ
oo a: uj
o o
o u
o z
^ t-i LJ
O H- *-«
LO -< ID <5 O
J •— (- O *- CJ
_j t- X - >
nr .—i a- tr t—
u h- cj <[ 1/1
O e o z i,j
Ui UJ UJ E I«
o y >-> t- is)
oo o _J UJ
^ i— ->*-• a j\
K CO •—" 0,
O O t/) CO
-------
o
or
uj
CL
e -H
a. s: o
x. <—< f^
O C"> H- C"
;2 Z> «3 t-
p
l^ > C. *i »-•
on LJ c-' 5: —
o ? r- Z3 o a:
o —• c u D o
_J «-« t- >-i O U
_J I- 01 •• X <
<
K *- > 1.1 -3- ^~
U K- C C- l/>
o £: w- u. c u-
i^ UJ o a: ui 3
z o a L x
O Z *~ — h-
C/i
CJ
< r\.
a o
(C
o
a 2" <
CC O ID
> a:
u,
c u.
u o
Ul ^ UJ
co uj a:
«i n c.
a. t.j t
UJ U CO
"
«r a:
a: LJ
> QJ
«3 O
B-54
-------
to
—I
O Z 1
&. o rs
« —i O
O I— 1-1
2 < V
or
CO 14.
C. l~
r-- ~«
ffv
u
Q Kl
UJ «
- 10
2 3 « >-
*t *^ O UJ
O
O
UJ *~
<-> Cl
U 2 LJ
ID LJ cr
« or a.
a: iu t
UJ Lu VI
> u. or
« *- o
o «^
t-l ^ •-
cn CD CM
I K1
O ? ^ Z)
D 1-1 C
X ft t-
UJ UJ OC'
_J > 0-
a *> -i
s: ~
O K
<_> 2 CJ
O O
: _J I >
O > UJ
«s or n
o: uj _j
uj to o: >
OOOOOOOOOCJO
B-55
-------
cr: o D
a:
CO LI
C u.
O
LJ **
U C
LJ
t-
LJ
a m
LJ «
CL * 2
a. si o
o
o
u. tr
CL f> ^
a: -J 2 v
V- > Q_
Cf t-i > Li «»• >-
L- t- c. ex tr
o 2 ii a o uj
LJ LJ O IE L 3
2 LJ a: u. x
O ? -H « H-
to o
Z L.)
If) <-» 00
ooooooooooo
B-56
-------
Z « CO
•a M UJ
>- > ct
GO Ui
O U
O
O Ul £t
L- tr:
« i-. o
o •*
^-i C,'
«S >P» O Ltl 5* CT
O CO !-• ffN
O t/1 _J CO f-4
UJ U' «S «I *v
•> D c i H- ce c
CZ _J 5* «-< « tH
ic L- co -a «-
o 2 y z> *H a:
O •— O -*• D o
_j »-, j_ ^-, O LJ
_J I- C£ •• X *I
rr" i—< > LU UJ
•a or 3
or ui _J
u co *i
JJ rt -H
t-i CM K)
-I
<
B-57
-------
a: o ^
a «- c
o i- •-
i: *r i/-
a
C/l L-
O U
O
uj Z* UJ
O u.1 a:
« o: a
> ii. c
«i — o
o •-
O fO O
UJ * •—
or * <_• en
u. s: o LU
-x *« ~ o_
,2 z> <: »- o ^
ii i
: co
O C/J
> :D c t- u. *.
fl£ _J i2 a
CT > y «* ^
cr.
a
t. L-
O > UJ
O 2 3T Z3
o -. c
J a:
r> o
O o
W N1 > 1, I mcc ^H^KJ co to *-» cc
cC'HflOOf^cy^cwfO^'in CM
^ m \s r*- o_ cr
OOOOOOOOOCJO
B-58
-------
CO
_(
C 2 «
o; c s
« •- o
Q I- ~
i£ «r V.
O ro
Ul «
Qi * ^
a. z ^
O ID *- l/l
z r> _j
UJ UJ It O!
a —• O
c, «-
o z s:
o •- c
•a « z «
of »—< >
u t—
O Z 0 QL
Ul UJ O
z U o: u
O Z l-< «.
t^ O **
M (J i.
o:
Q. O
C3
E a «-i
i c —
i t— a:
a: O O
o u
• • I «
_i i >
iii -sr t—
C- (Sj ^
o u.
o
U LJ
O > UJ
<[ a: =i
a: ul -J
UJ OO «I
> en >
« o
UJ
-Q J.
I C
B-59
-------
or
t/l UJ
Ci LJ
O
Q K-l
UJ «
U <••
(/. >
CO
o
o
o
or
o-
c- «-<
3 an
•« tr
o *-»
-
i UJ CC
> rr
o **
O "-, O (/I 13 O
IJ K c: .. x «
n < 2 « _l l >
O Cj
LJ rr UJ
(.0 uj a:
«i a: a
or UJ I
UJ u (/^
> u cr
«i « c
o —
in «-<
CM
o -=i c- a: o »
u^ LU o n L.' :
^ t_» a: u. I
O ^ —• ~ h-
C/3 O «
•-H o CL a:
: (/j
i ^
i u_ u
c u
CNJfO^-lDsflf^SuO'C r-»
O O O O O O
O O O O
<1 C/D CO C/) 00 CO
-------
UJ O O
o en U!
2 -. cr:
> o o
u —
UJ C
Z O LJ
•-» ui n
O (X I
1.1 L,J CO
U_ O
Q
UI I/] *-
O U) f~\
5* (1 1.1
-I U 13
a: »-. _i
bl O «3
> LJ >
<- D
CL
Q
UJ LJ
> en >
< o
i ft 10
--4 CM
t 0.
IP
*-t
I
I fO
00
cc
^
I
t
-------
" O ul
X C3 li.
<: UJ u.
£ or >-
u. n
T 0 I, I
^ C- O
• o tO
~. I
e*
3
J
a.
B-62
-------
o
o
X t i ~ ry
•—• <~i t/i o
X t.J ul X
^. O
*-t z
or
tr t-
_J «-« en SI
cc tjj v- cr
O L.
o
v-
r- >—
o <*
C
X
t/j CO
_j uj
»- »-(
-------
LJ O LJ
r- « n
n m
5T O
a: or u.
«r o" bi
UJ O O
oo a: u.
o: or uj
«i o o
«r rr ui
O
f/1 Ut C5 w rt' CT
LT •-< D a H-
•— O t/1 O •—. ( J
T UJ Ui TT "X.
2? -J w
-i «-> «r a:
h- U. H- 3 CL
O C *—• Cl A
«I tO t-4 tO
tr o v, ec
a •*•
in
i
L.
-------
UJ O UJ
l > •
x o — K i
o —. r> i
»— o t^ o .
X Ul UJ X '
CJ. ^ *
a. o_ _i -a- <
<~. < o
ft ft- rt' 11
s: uj i—
o i/j ^ ^ i
O O HH
^? >—
o
CO OI
o
UJ
or
-^ CC
o
o u.
UJ
t— (—
f; U
UJ uJ
3 -1
^0
ty
c
O »~ It 3
2- ^ O Q
^ UJ CO Q
O => —I UJ
•-« 1-1 «t OL
t— u. K- 13 CL
LJ O »-• O /\
CO >
s: o •« —t
o.
y
O
B-65
-------
tiJ O O
D- <_) LJ
Ct U_
cr: UJ
O O
o
c
t£ _J O
L,J O »-»
-. r»
<-H s:
T ( i — or or
O i—i c/) O t—I
21 lii l.l T X.
C, rr O
LJ 0 1 *H a.
o > <: o :
c^ rr" «v i
i *- cr
O
O L_
O
t"
00 1,1 UJ
a ae
o > -J LJ
*-» •-»
-------
o
o
21 3-
-* »-• ID
ro O a:
H- * o —« c)
i X x. Z .
ry a m
«j 31 O
-» n —
LJ
ui O o
n o u
o • u,
c/) cr u_
or a uj
i O •-«
(/> UT
(D LJ
o ^z. c/o
o: cr _i
c, »- u. o
2 U.
-------
=r o
UJ O UJ
t) W iv
r* *- a
0.1 111
• * • 1
or • r-
: o o
. u u
1>J C HH
> 00 (/J
«3 IT) lil
K"1 O
* «3
* or
o ri i/- o —* c)
^ i.i UJ X "v «J
on o > »-i
:=" L j ^ or1
^ «I O Q
tH » OJ
ct. ^> ^ ^ i
<: n. LJ O
G. O O t-t
c >-
V C
CJ O ^^ O /N
-------
u, -- o ;r
cc K> « o ct
CL + UJ •—
* 5- C£ -J
Z^ -X. t— Q. O
« O
to o *- n o u.
«_J ^-* ^Z O O
r »- ui x I > LJ
X »~
t/i
~2- O S
O t_) U! QC
t/-' > O
i— CM lu U-
c. c
>C CJ UJ
u. m
cc >
-------
X C3 UJ
*: &• u.
u. •-«
n
a: o ^
< M a
o t— »-•
* ?" a: _i
a s: —« i.i ui
:E* -N. ^- a. o
n ,H o ^ y
r\ -> »-* -^ a-
U) H-! cr t-
^> to o ti ^-i o
tr uj uj
L- UJ CT U-
C O I T
o o
j ^ LJ
> U- fT
-------
« _J
D 2 a
K O =>
« i— O
D I- —
Z t VI
<. •-< UJ
t- > o;
• i-i *O l\J
, -x O OC
Q. O
—• n
' O t-t 5T
U Q
UJ "Z. LJ
CD UJ IT
o Uu
o o
X « O
U- UJ CD
O CJ I
^ c
o c3
— » r\t LJ t
a. o
o
UJ LJ
i-i
*t ££ ^
cr uj _j
LJ 00 <.
CD U_ Z
Z O U
Q
O
o
o
o
X
B-71
-------
c? + 5T o: -J
a r n ui 1.1
v HH a. O
n t*> *— ^* ** O
;? r3 «i o •-( -H sr
t l1"- rr rH t)
ui id ra n a r» o a_
a: _j ;? uj *» 03 3:
CD U-< _J *^ CL
a z s: cc cc o
O t-t D - > UJ
U- I
O
«I CO
s: u;
^ o a: X -D
O 2 i-c 1- I/)
Crt O «I
*-« o a. ts;
^)
c
i
*- tJ O
~> ^ ^
y. ui ui
•-. -» o^
x a ui
-i o
c >- —i
O U-
o
o o
UJ Z UJ
o uj a:
« nr a
or LJ t
UJ h (/»
> u. cn
« M O
n —
> ro >
« o
Hi u. ^
z: o ut
in
cc
Q.
o
r>
_i
Q.
O
O
o
I
UJ
a
-------
O fO O
UJ * *-*
, CD »- ^- ~ rt
O • > ui
^- r-. or *- *-
o r»
co o
nr X ID
•— H- co
(V n rn
~ »-*
00 «I CJ
(T* _l »-i
*I UI U
LJ Ci U-
tL £K UJ
o o
U CJ
UJ O *-
> (/i (/i
00 Cl
> _l UI
^ •-* <1 CK
Oh- r> Q.
•-• ^ O A
I— CO i-i CO
LJ O co a:
«i o. UJ o
a! a: •»
a
s:
B-73
-------
i-t — 0
o K-J o
LJ «
Of *
2* OL _J
O uJ LJ
«-• CL C.
*— *•» .«, o
«T O .- -rH 5-
t i r* cr
r> i-» t- rr K-
_J t^ Qt ^- U
uj — oc s:
> rr *-»
a:
a: en o
• D < o u.
o 2
L-
z U
o ^
co o
*-< LJ
tr
^* c\j
C C
77- t/-
o
(J
LM LJ £ :
a: X :3
•-" t— LO
« UJ U
£ cc: L.
U, •—
a: o Z3
o a
CD
i cr
a: LJ i
b! U, C/3
> U- m
-------
o ;?
—• LJ
LJ LJ U,
Q. nr u
c.\
LJ
h-
U
•-« *~> o
Q rO O
LJ * >-i
or * ?* a
a. sr o uj
*v t-. O.
O O t—
c;/ d
0) s:
rr *-*
a; _j o
LI o •-•
o z a: or on
-J t- X
•a -a Z I >-
nr «- »o jv
o
o L.
2 u (X X T5
O 2 i-i f— 01
CO O
-• HH n A
t- w *-» c^
O O to DO
< Q. LJ O
a: o: *-
0
vO
B-75
-------
* ^* CL _l
y n u uj
X^ KH Q_ O
,-> >- « ^* O
r> • n i—'
^ - > UJ
> 0. cc
O
o o
LJ 3 LJ
^-. o
n —
a:, u-
s: o
in <-> r-i n
a.
E:
o
X
o
a:
UJ
Q.
lO
O-
B-76
-------
O Kl O Of
U! + —»
or + 2* cr —i
CL Fi o uj uj
V. HI 0. O
;=* 13 «J o cv »H s~
O •—i h- C3^ k—
o oo _i t^ ce f-t o
LJ li) <; -a NS. z? c~i »r n o u.
Q; _i 2 10 •*• tc s:
C/1 > -t 13 CC 00
ce
^ o
- 00 *-« CO
c; o oo oj
-------
APPENDIX C
STATISTICAL TABLE OF MODEL PERFORMANCE FOR CINDER CONE BUTTE
Table Page
Cinder Cone Butte Comparison of 25 Highest
Cinder Cone Butte Comparison of Highest by Event
Cinder Cone Butte Comparison of All Events Paired in
Space and Time C-20
-------
o
. « o
cr
CL
(V (1
bJ OT or U O (T t—
x o ^: o o
ti H- ^H X O
ri: LJ
jj t— tii
c« z *-»
U. (_) _
02"^
T ^
y
ID
,_,
X
UI
CJ
2"
<*
Of
«•
^
- tJ
C 1 ;*•
27 ui
Z> UJ
C L.
U O
O 0
O" UJ
»-» cr
oc Q.
c_ to
y en
0 0
* 0-t
uj n
O 1. 1
^ a:
1,1 O
IY" i
UJ I/1
u_ nn
U- 0
0
*
Ul O
UJ
Ql ^
O
C\i
Kl
•
•
••
C\J
•
a:
,_!
1
*
«• »
• o
•
in
,_«
r-4
•
ID »
• t-l
•
C\t
CT
|
*
vn ••
• rj
•
-------
— a:
a: a
ro ^
if) •
o o
i «-
o ~
uJ *o
>— *
o *
C.
C
UJ •* C£
K 0- Ci.
LJ L: I
U. >
c? O k-
-• Obi
Q_ C_,
o:
o« (- r> o or L.J
u. z o o :D s
x c zr ^ c? o
CJ ^ t-t & . . X O
•-i h- < -J I
x « rt > in »-« or
Q. U O L'
u" K- o: a- O c-
Cv S *^ O £ L ^~
UJ 13
uj C -a
> LJ >
« o:
0.
5" -I
O LJ
:> n-
< o
1- I
< o
_l «
LJ UJ
tZ I
2 UJ
O C/>
> LJ
LJ o:
O
t- O
X
-------
U f
i-+ y
o -v
UJ — « C1 U.I
CL <_•
a'
o cr
2T
X O ^ —• CO
O *-« f- (X •• X l_>
»- t- UI «—' CC
C. U1 C_ Uj
o z>
^_- c- —
L^r O D
^ •— a:
H «I X.
or a w
O O
,-, o
c u. —
^- a in CM
C-3
-------
r> -x
a. — ~»
o :-"
c. t c
p- ^. ^
<» * t—
-« «: f
O .1 t_» t-
UJ •-» O ui
>*-•_' c/1
,_ I/ iX <
OJ ^
C_ _J
-,o *~ z; o or LJ
i,' ir o o 15 r-
x c" r —. o c
cT •— •— QL • • ^1 <_-
„ li- L L.
i."- *~ i— U_ O Li
O 2.' 7!
O 3 a:
^ w «- C
O u_
c^ U.
£ —
7* — a:
•a cr: a.
•> C O
L . <-
•— O
o u —
"^ ^.J J
L C «
> rr >
—
-------
Ct C- c--1
o o
•a- co •« .-H
o. r- o ir>
rn CM —i — «J CO
a:
uJ co a: — to -a-
CO t— O r-«
u,1 2 O a-
x o 22 t~i
o *-< •-* or ••
— t- UJ
£1 U- O
IT/ f- cc. u: o
CV 2 — O £
Uj *I U.
U. vj G-
o 2: :?
o o
o
c-" u*
*-" >
o o
X O
•«* U Z5
a: ^ _j
< a:
D.
t- X
< e>
UJ QC
_j
U, U'
O =>
t- O
C-5
-------
CJ •*
0 *V
o t o
^ -. t-t
O
O
l.i O O
£_> t-0 uJ
c: C. G"1
^ s: tv
> o o
u uj c;
ir o LJ
a: a c.
UJ U.I 1
-
O LL ^
v t- i> _i or ui
I O Z — "" O O
«-« t- -3 _J I
T LI t-i or
cr" L- C- Li
if. >- u- a. C C,
r, 2- ^ o 5_ L_ r
uJ *t U- T" •—
U. U U. ^ U
O 2 Z
o 13 a:
?- u «- c-
^ _
« t-
Cl «t
3~ _J
O Uj
o a.
o
I LJ
> LJ
> rr >
LJ
LJ a:
O •
i-H C3
•o
o
t- O
5 v-
5. i:
• 2L
o • •
f-t O • O
I —t :: c,
V I 1-1 A
C-6
-------
LJ O C-
LJ CO UJ
or. c.
o
LJ c/
C x.
LJ CO
a:
a. *•. *•»
J> z
LJ I O
o ^-> o t—
UJ «~* O U.
> <- _) c-n
a:
LJ LO tr - O O
o »— •— nr *• x o
a: ij_
u" t- a: a_
C^ ^ KH O
LJ
•a. a:
i: u
u i. ,
> IT
ai to ro cc
vD Lf TJ £T^
^-t c- \£ ri
LJ LU
IX X
o
t— o
C-7
-------
o *
>-> y
C v.
* « to
! C ' LJ t-
I —» O Ut
C _| 1.1 o t- > S
•an a: *J tr
t- > — « c
f* H- :s i or u:
H O 2 ^ 00
c& i— —. rr .. x o
~- i- <: ^ t
x «r o > i,; " cc
cr k- C Uj
u*i t— LL a. o (—i
(X 2 K_ O 2. L_ r
UJ o o
cj «-
UJ (/> —
CJ U- O
a. cc. o.
LJ UJ I
t- O
Q U —
o
LJ UJ
IL t~ L.
< CJ 13
or t— _j
uj o UJ >
UJ UJ
U > b-
cr >
< o
t- r\i •-<
a- IT
K C K,
f 0 O'
I- I
•a a
K X
2 UJ
O CO
uj
ui or
_j
LJ L
O
a. cc
O 3
I- O
a. c'.
I r-H C K,
V I —» A
C-8
-------
LL *- ^
^ ;*
O t O
D en CJ I—
LJ .-• O LJ
j co or
L. O
O
c
a. o
tr (r>
flO
C, _
«-« oo
oo t- o
LJ zr o
i o ^r >—
c^ *— «-H cr:
LJ =• Zi
«r «r Oj
C^
s: o^ UJ
c o
•• X U
_J t
lu »-• or
c u
I «
f- U
C/l LJ
O LI
o a
CJ t/i LJ
2 H- r»:
C' O
o LJ C:
^ t.^ LJ
LJ -i a:
a: a: a
Ll b-1 I
I-H O
CJ U- **
o
U-, LJ
tl* t-
> L
LJ
(T- —» t
f) IT.
LU LJ
a: X
-I
UJ L
C-9
-------
S" L-J Ul
*-. —» «V
X C* Ui
s: a: u.
U- "—i
o
iri fr%
^7 CO
fr O 13
«3 •— C
D t- t-H
> li_ CC
7
T
— Cf
CC en >
-------
O y>
UJ «
t~ *
O S~
C. f>
LJ O
iV *» -- O
I 5" Q'
z « *- a.
«s *
02-0
UJ O Z. O
T i-i ^ X
O t- > t
i' D I
u. z: c. u
O IL, I ^T
u »— t-
^ ^ 7*
O O l.l ^
s: t-
o
-------
7- ~
"v t/j
'./> H-
C. vT
I CT
O
-H o C-
LL Z7 UJ 2. UJ ^
O u_ I 'Z. —
o >- ^— ^J
O O I
Ct- U1 LL-
«I > ^>
or
o *- H-,
f/l «T t J
a: _j •—
< -H
O H-
u. •*
UJ 2 Ul
o uj ce
*I r^ Q
IT LI I
JT
A -•
(I /\ h- V M A
2- X X X T-
t C.^ f/> 1/1
•—• '-'J t/'1 trt
C-12
-------
a: ^,
o O
-a. >-H t/;
I— > ^J
I
> LL on
CC ?
l- O O
X re u r
i- j> O
u. r* L, t,
O UJ (
D >-
p- e x
O O I
C. LJ uJ
>
a. '-« j
L. c^ *J
> i^. :>
T O
^i vc
o
-------
Lj «
LJ bj C
01 — 00 o
C"i --• t
Ld —
t— O O cr LJ!
v *: u i_j ra 2*
n o —• O C;
X — • • T i_)
tl r- >- _) 1
»-l - t-H
*3 —i (/
f- :> .jj
^ U! IT
T U ?•
r^ o uj
13 >
IT r,i ^r
rO .-H
< I
v s r:
x
iO »— 1 r/1 C/l t/1
_t • • "— ' ir G" co
O )— i\/r\jA H-
C--14
-------
0 5
c (/;
La
a: -*
Q. \O
I
O —
^ *
-a *
u_ o
to o
I ~ 13 *-« —
O -H QT UJ
O t— *-"
t/i L-
< cr z:
ex u/ _J
Li V) +
u. :: -^ t- u i_ i
u. or: —
t/l |(-1 >— _J (^ L/1
• • •—' \f -J~
C-
-------
> C)
O_ OT i—•
u.
C o: LT
O Ul I
o >-
DC UJ
o o
Ki ro v.r
r- c- ^
f*- 0- fO
C iT O
>— f—
5: A T:
^ri r M
C\, ^H C-
r\ r- o»
I i I
O CM
r,, ex
0.1 t-« >o i
< D 00 •
C-16
-------
Cl ^
a: o
Lo U.
ce « c^
a o
—' n
o —
> .X C
H- O f- :£ UJ
c--, r- u a: Z- ^
u O — t- o
X « •• T u
ts t- t— _J I
> -
O LJ I
o >-
2 2 CD
LJ 00 C- >
«-" O
t— V II /\
(/) X. i X
vD
O
II /\
x x
C-17
-------
^
111 +
•— V, O"
CJ L' t—
£T ~ C.O O
O_ u <—*
1 <4 cr.
O — - t— i,J —«
^ * <: CL c_.
*r * O cr
r O^
O --. (S r* —i
LJ »- — I-H *-
•> r; c?
CS '^ *~ *- i> ~>
o <* ff ^ cr
> LiT C
h- O X .X uJ
C/5 ^ U (^ ^ **
— — . *• I O
0 h- »- J 1
— t «.- ^ l.| _- ^V
z a- L. r j.,
t- > O O
u ^ L- 3: uj s:
O UJ 1 X •"•
W >• t- v-r
~ r* m
oof a:
(," LJ »- **-
•— L' U
cr ^ i.j
>
•X —« 'J
(T «:•
t > .1
a
^ t—
0 7
?» i— i.i
O H- 1—
I/I u. cr
u-
^ r£ ^
D* liJ _J
UJ £^ -5
> .'I'. >
«3 O
^ cn
LU t—
O"i 1L T9
p- o ui
13 >
^ LJ
-
-I
v£
n
^D
If
rv
•
*
a
t-
c\,
i
s£i
rj
-
t—
T h I
^ 2
T
o vj.1 -a-
* oo
H- V II A
00 X X T
V II A
T I X
.* r- o
I-H rH J-
I I I
i \/ C\' A *— i
C-18
-------
Ill -»
a. o
•=> o o
tr (/j •-• « uJ
u^ U. Q CE. t-
f) Z> «J *— LJ h-
Ji _l > LJ > ")
o u Q.
t- o s: o: LJ
O CJ
O liJ I I f-
o >- •-»
LJ or
Q. ££ k
CJ »- O
-f -H Ul
oo LJ a:
o
*-t O
n —•
*- V
c/> X
h- V II
1/3 T X
U CM (V
I.S. "J .>
rH ^ "-J
i-~i CM <~J
—i \/ c\j A
C-19
-------
y u cj
=> :=» r-
y ui iu
tt C 3
-a -* o
o *
u r
*~ "^
l_J t/1
i- O
o *•» o
U vT i-t
a: i ^ T
a — n ui
« —, CL
C^J
rr
c-. «3 L5
—< ( ) ^ ^-i
_J 'T Cf
I C'" U' J
LJ
LJ ^
O O
LJ 3? Lu
co UJ rr
u_ en
- O =. O
O C) •— t—
t/1 ^ <
—' o Q- or
r >
tl- i-
uJ
r
CL
O
C-20
-------
V L-1
l.i O Ui
u ir. rr
f —. *i_
•cr or x.
—i ~J
LjJ
o s: f
T' O S
3T I— *-i
n- iff u
LJ! LL. LL
n cr ii
o 2T
p- •-* LJ
O t— »-*
iv > •—i
-J LJ u_
CO CO
t CC LI
o t— c.
^ L.' LJ
{_• ^, ^ CO O
?• C
*a tr
C.
o
o t- D a
•—!•-» O A
t- IS) t~< f)
DO CO C,
«x a LJ o
or a: —•
CL
o
O
C-21
-------
— O UJ «.
*.-.,". 0. <=>
7 O o nr UJ
;; •— o o r> r*
: •— tr •• I u
<: -H
i rr:
o
01
«a -• O
D *- «-t
O UJ OT
u a,
^ •— o
*" ^-1 -D r\j
I/JTIXX tnixx
vn *o ^
r- C\i CM
o co •
!»• "- u
u1 :•
D
U v V +
-V (- L UJ L.
»— CO
) CO
C-22
-------
»-t -. IX O
** O C/l '-' <~
I/) —
^ ^ «. UJ LJ H-
_J •« O .J > 13
> LJ 00 S"
,Tt-
ir C. L1
h- i— JL CJ uJ
2- U.' O £- L.I .-?
UJ or u. X 1-1
u •— •— t- o
<" a:
C-l U.
a: ut _J
u» c/> <
r -
UJ LJ
> : A >
* • uj •
«-« j w :r •-(
V II A
T X X
t— V II /\
01 T J. X
A »- LJ (
C-23
-------
n CD
"N. ..! C.T :
*T U.' O ST LJ ^
LJ nr u, x •-»
> 0"
*=. C
m >r
i .-(
t^ tO
x r x
C E. -^
t-. M j-i IT
^ •
'-«V(\i/\
I— OLJCJ
C-24
-------
»•%
*o
+
C; +
UJ 5^
O (S*
>- c.
o -> o
G- i :* u:
« *-» *"» Q_ c-
Cj * t~ (/i cf
*r o U_ LJ H- U. t—
CC O ^ «l L*J f—
u, _j «t O :> ™j
i/i *x »— t *i fin
C > M C^ ^r
o sr z> r-* a; uj
^r — o i - 1 — < rr-
u u' L_ ^.
O r~ ^J QL O CJ
*• UJ O El U ^
2" ^U tt U_ X •— i
O ^J r^ w t- O
oo ;?* «i
r-i o a a.
*r < ) r
«. il
Ci. it.
51 ^
"5 -H
C H
<=-
_J
c£
uj or
u. nr.
** o —
o
i*. (ji
O > UJ
a: >
•J O
a. v
u >—
0 11. -*
r c u
~> >
2; u •
t—
10
t/1
•a.
t—
«3
O
c. m tT1 r\j h- »o r-
10 c\j TJ in ^-i f*") »-*
a; (\j tJ- in <~- cr f^-
1 I I f\j it
i
ui u L (Ti a it . i- .
r-i
^ uT; Oi C" a. t-t -if
WJ ID *H \£i _D f-» -^
O — < J" "\J rO H« T\
h t-
x a. n
u. r~ a.
o • •
rz H- • \£ • vH
r\(
>— (J) C,' C_
co —i tr> c DO
ex v ; r / : r
21 r"
LuU. UJ* • uJ • •
COCO LJJ1»U1 (_)>JJ«\i»
-
*t iTi
fw r^ ^-* m ..
3- r— ^~
^ i ,") \j-r-
C ^ »- ^ r-
-*
^
t—
c
CL
O C'. ^2
0 -v. O
i- u;
_J Ll
U' 1 !'"-_' C- >- *
— liJ> *-. *~C-LL^
c: u. or •-
O CO tD t— _J f^ l/.1 k/
_J • • *-• CO LO L/
c t .\ •: L j <-. " -'
j: ? •
-------
c —
i,, ,f
O LJ •— — H- <_J
^- o i- c;
O" ( J C~
•A U
o *- »-.
2 -t CO
LJ LJ
CO l/l
H- V It A
coxxx
t- V H /\
ooxrx
CV. -3 ^^
r*- fo ^
vj a n
r r~
_J W CO
•-H CO CO
C-26
-------
C LJ C/ D
5- ID t- a; uj
r *-« o or =» r»
;, r- i— O O
o c, *— *-
or" O Z)
« »-H O
tf U^ -J
UJ i/- -«
-------
— «
f,-\
*
;*
-V
(,
o
— o
t Z. a.
— 3 i.i -»
* i-* **> Ct O
* •— t/1 C
r-i
—« ( ) tii ;* ,-<
— o to — —
c/*1 *a «J UJ
"D 5 -x rv UJ H-
—t -d L_, t~ > "i
<. a: -j t/1 C-
5T O X n; lu
^7 — i r1 LO z) ;^
C H- —« ^J O
•- QC -. T- LJ
*— ir ^> t . 1 .-^ &"
^ LT L.
t~ CJ Ui C- CJ
a: u. c E u. z:
tj — — t- U
r- i
o a. rr
L. r~>
u
L_:
J
K
-=r
_f
LJ
CO
_J
jr n ri
*j t-i O
0 »- *~
0
UJ *s
LJ Cj
u.1 ~ UJ
to UJ or
u. n
c£ -N 0
("^ *—
Q
CJ > UJ
•J, O
0. fc".
O J- "»
: c L
r> >
•2 Lu
l_
i«
_/l
'i
£
n
>o m vc ;• »o r: wi
CM *— ' i— < in «~* f-'j
«-» <,' *-* fO v/1 •-< IT)
1 1^1
1
^J
vi 1/1 o a «x _.-» *
"o ^ *dr "v ^0 r^ jJ
K ^>
•^ > r
LJ 5. s:
o • •
— t— • O • \i.v
r\j
H- O O O
l/i ~H CT^ C ff1 O
-H L»J t— h-
C> 3. V ^. f /\ S T
s- r
UJUJ UJ» • uJ • *
COl/3 (_)vO*>JJ L)i^»-n
J » ^
pi— ^ •- «-
3!
CJ
t—
0
L' i-L
j: c/' 33
O X O
E: cc
_J O
<•- n (/^ • o
L. 1. X. tf ^ V »
—• u, : v. *- u u. u
C5 a_ on ^
O XtDf— _Jt/lCOoO
_J • * •—. ^j y< V
' j r t v 1 u ; 4* i »•, o •• *.
U- ;-J • *I _l — 1 _J
o *-« v c\. /\ i- LJ o o
*... ^£ k."
UJ • *
r t-. nj
C-28
-------
a
o
«-« CJ I.I
<*> O CO
_! *-* (X. »• X
I. I •—<
LJ_ u; C
O i— £_- ti O
CL H <1
or O ID
«1 ^ Q
> u cc
*T »— O
r ~
U.' CO
> x
IT1 (\i KJ
•-H ^-t C\J
c r: ^-.
t- V li /V
CO J_ X X
t- v it r
I/ X T X
C-29
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing}
. REPORT NO.
EPA-450/4-84-017
3. RECIPIENT'S ACCESSION NO
4. TITLE AND SUBTITLE
Evaluation of Complex Terrain Air Quality
Simulation Models
REPORT DATE
June 1984
6 PERFORMING ORGANIZATION CODE
. AUTHOR(S)
David J. Wackter & Richard J. Londergan
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
TRC Environmental Consultants
800 Connecticut Boulevard
East Hartford, CT 06108
10. PROGRAM ELEMENT NO.
11 CONTRACT/GRANT NO
68-02-3514
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
OAQPS, MDAD, SRAB (MD-14)
Research Triangle Park, N.C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
EPA-450/4-84-017
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This report summarizes the results of a comprehensive evaluation of eight
air quality models applicable to complex terrain. Seven of the models are
"Gaussian" and one is "numerical." The models are evaluated with data obtained
from two field measurements programs. The Cinder Cone Butte data base is for
tracers released upwind of a dense sampler network for a limited number of hours.
The Westvaco data base contains a year of routine hourly S02 measurements for an
11 station network. The report includes numerous tabulations of each model's per-
formance in terms of statistical measures of performance recommended by the American
Meteorological Society.
The purpose of the report is two-fold. First, it serves to document for the
models considered, and similar models, their relative performance. Second, it
provides the basis for a peer scientific review of the models. To stay within the
spirit of this latter purpose, the report is limited to a factual presentation of
information and performance statistics. No attempt is made to interpret the sta-
tistics or to provide direction to the reader, lest reviewers might be biased.
17.
a.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Air Pollution
Mathematical modeling
Meteorology
Power Plants
Sulfur Dioxide
Statistical Measures
Performance Evaluation
b.IDENTIFIERS/OPEN ENDED TERMS C. COSATI 1 Icid/Group
Air Quality Impact
Assessment
New Source Review
18. DISTRIBUTION STATEMENT
Release to public
19 SECURITY CLASS (Tins Report)
Unclassified
21 NO OF PAGES
244
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
-------
U.S. Environmental Protection Agenpjj
Region V, Library
230 South Dearborn Street
Illinois 60604 p
------- |