United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
xk'
i;
Research and Development
EPA/600/S3-88/006 Apr. 1988
Project Summary
EPA Complex Terrain Model
Development: Final Report
David G. Strimaitis, Robert J. Paine, Bruce A. Egan and Robert J. Yamartino
The Complex Terrain Model
Development (CTMD) project has met
Its original objectives of producing
an atmospheric dispersion model
appropriate for regulatory application
to elevated sources of air pollutants
located in complex terrain. The
model development effort has
focused on predicting con-
centrations during stable atmos-
pheric conditions.
The program, Initiated in June
1980, has involved the performance
of 4 major field experiments which
produced a wealth of data for model
development and verification
purposes. The first experiment, held
at Cinder Cone Butte (CCB) in Idaho,
involved the extensive use of a
mobile release system to provide a
high capture rate of ground-level
concentrations resulting from
elevated plumes flowing toward the
butte. The second experiment, at
Hogback Ridge (HBR) near
Farmington, New Mexico, featured a
very long ridge that provided a site
for testing the importance of terrain
aspect ratio on the flow dynamics.
The final field experiments were held
at the Tracy Power Plant (TIP) near
Reno, Nevada. This Full Scale Plume
Study (FSPS) provided a large-scale
test of the modeling concepts
developed. Data were also obtained
from a series of fluid modeling
studies performed at EPA's Fluid
Modeling Facility. These tests
provided confirmation of some of the
basic theoretical principles adopted
in the modeling effort and provided
Information on plume behavior as a
function of systematic changes in
terrain shapes, release heights and
distances to terrain objects.
The Complex Terrain Dispersion
Model (CTDM) is an advanced
Gaussian model that uses a flow
algorithm to provide terrain-induced
plume trajectory and deformation
information. CTDM is suitable for
regulatory use, but It requires
substantially more information on
terrain and local meteorology than
complex terrain screening models.
With simpler data bases, it
demonstrates degraded performance.
The model evaluation effort
concentrated first on the use of field
data collected within this program.
Subsequent tests were made with
two other data sets obtained from
SO2 monitoring networks near a large
paper mill and a large power plant,
both located in complex terrain.
Statistical performance results of
CTDM were compared with those of
other complex terrain models of
current regulatory interest or use.
The model evaluation demonstrates
that CTDM has superior performance
in the majority of tests and has
consistently good performance
among all the sites. The statistical
performance of CTDM In complex
terrain settings is shown to be
comparable to the performance of
EPA's current refined flat terrain
models in simple terrain settings.
This Project Summary was
developed by EPA's Atmospheric
Sciences Research Laboratory,
Research Triangle Park, NC, to
announce key findings of the research
project that is fully documented in a
separate report of the same title (see
-------
Project Report ordering information at
back).
Introduction
The CTMD program was initiated by
EPA in response to long-standing
controversies in the technical and
regulatory communities over the lack of
reliable methods for predicting air quality
concentrations in regions of mountainous
or complex terrain. Of particular concern
was the absence of a verified dispersion
model for predicting ambient air
concentrations during stable atmospheric
conditions, the conditions expected to
give rise to the highest short-term
concentrations.
The components of the CTMD
program and the progress made are
well-detailed and documented in a
number of reports. In particular, five
Milestone Reports were written during
the course of the program which
describe in detail the model development
efforts, the field experiments, the data
gathering and interpretation efforts, and
the model evaluation efforts performed
during the course of this program.
This report describes a refined model,
the Complex Terrain Dispersion Model
(CTDM) that was developed during the
CTMD program. Comparisons of
model.predictions with observations at
five complex terrain sites are presented
for CTDM and complex terrain screening
models. Results of a sensitivity analysis
of CTDM are discussed. Finally, the
limitations of CTDM, project conclusion
remarks, and recommendations for
further research are presented.
Field and Fluid Modeling
Program Results
Data collected at each of the field
experiments were archived for use in the
model development and evaluation
phases of this program and for future
use. To alleviate the difficulty others
have had trying to utilize raw data from
large field experiments, EPA has
developed Modelers' Data Archives
(MDA's) for each experiment. The MDA's
are subsets of the complete data sets
which are thought to be of most use to
those involved in dispersion model
development. The raw data have been
collected in a series of computer files
which are available on magnetic tape
from the Terrain Effects Branch of EPA.
A progression of understanding
emerged from these experiments.
Observations from CCB demonstrated
the validity of the dividing streamline
concept or critical height, Hc, in
simulating the flow fields during stable
atmospheric conditions. Tracers and
smoke released directly upwind of the
butte and above Hc were generally
observed to flow up and over the butte,
in accordance with theory, while plumes
released below this height were generally
observed to pass around to the side of
the butte. At HBR, the dividing streamline
concept also was shown to be
applicable, although the flow behavior
below this height was different from that
observed at CCB. In particular, at HBR,
the lower portion of the flow was
"blocked" behaving as a relatively
stagnant flow with correspondingly low
wind speeds. The largest concentrations
observed at HBR occurred for releases
below Hc and the magnitude of the
concentrations (normalized by release
rate) were much larger than those
observed at CCB.or subsequently at the
FSPS. Observations from the FSPS at
the Tracy site showed both kinds of
behavior for releases below Hc. There
were portions of the flow which, when
they encountered high terrain away from
the valley walls, became relatively
stagnant as observed at HBR. Plumes
embedded in flows having a primary
down-valley component and encoun-
tering terrain obstacles protruding from
the valley side walls, on the other hand,
exhibited an ability to lift up and over the
terrain or readily pass around the sides
as seen at CCB.
An integral part of the CTMD program
from the beginning was .the efforts
undertaken at EPA's Fluid Modeling
Facility (FMF) at Research Triangle Park,
NC. Theoretical aspects of the
phenomena associated with interactions
of stable atmospheric flows with terrain
obstacles suggested that scaled-down,
fluid modeling experiments could be
used to investigate many of the fluid
mechanical issues. Experiments at the
FMF included simulations with models of
CCB and HBR and for conditions
corresponding to some of the field
experiments. Verification of the dividing
streamline concept was a central focus
for experiments that included testing of
the effects of changes in release height
and terrain shape. Another series of
experiments addressed the effects on
maximum surface concentrations of
sources of different heights being placed
upwind and downwind of simply-shaped
terrain obstacles.
Overview of CTDM
CTDM is a point-source Gaussian
plume dispersion model designed to
estimate hourly-averaged concen-
trations of plume material at receptors
near an isolated hill or near a well-
defined segment of an array of hills.
When a hill is present, CTDM accounts
for how the changed flow alters the way
in which plume material can reach the
surface. Obviously, the path of the plume
can change as the flow spreads over or
around the hill so that there is a shift in
the relative position of a receptor and the
center of the plume. The rate at which
the material diffuses toward the surface is
also changed, and the center of the
plume is allowed to impinge on the
surface of the hill. As a result, peak
concentrations expected on terrain are
increased beyond those concentrations
that would have been expected for the
same meteorological conditions on level
terrain.
In the absence of stratification, all
streamlines in the flow pass over a hill as
modeled by CTDM. The centerline of a
plume in this flow follows the streamline
that passes through the source of thai
plume. As the plume grows in the vertical
and horizontal directions (in the plane
perpendicular to the flow), plume materia
diffuses across adjacent streamlines
eventually reaching the set of streamlines
that marks the surface of the terrain
Distortions in the flow which are inducec
by the hill change the position anc
relative spacing of the streamlines frorr
their initial distribution over level terrain
and therefore change the shape of the
plume as it passes over the hill. Ir
CTDM, the plume stretches in th«
horizontal as it passes over the crest of«
simple three-dimensional hill and thi;
stretching produces a wide
concentration footprint over the hill. Ir
addition, spacing between streamlines ii
reduced in the vertical and expanded ii
the horizontal, while the speed of the flov
increases over the crest. These change:
tend to increase the diffusion in thi
vertical and reduce it somewhat in th<
horizontal, thereby affecting thi
magnitude of the predictei
concentrations on the hill.
The nature of the flow change
dramatically when the flow is very stabl
stratified. A two-layer structure develop
in which the flow in the lower laye
primarily deflects around the hill, whil
the flow in the upper layer travels ove
the top of the hill. A critical height H
defines the boundary of these two layer
in CTDM. In the layer above HC, th
approach flow has sufficient kineti
energy to transport a fluid parcel up an
over the hill against the density gradier
of the ambient stratification. In the lay*
below Hc, the approach flow ha
-------
insufficient kinetic energy to push the
parcel over the hill, so that the flow below
Hc is restricted to lie in a nearly
horizontal plane, allowing little motion in
the vertical. Consequently, plume
material below Hc travels along and
around the terrain rather than over it.
Adjusting receptor positions while
keeping the trajectory of the plume a
straight line simplifies the mathematics of
CTDM a great deal. Rather than keeping
track of the actual boundary of a hill and
the deformed trajectory of each of the
segments of the plume, concentrations
are computed at receptor points in a
coordinate system along the plume
trajectory. The resulting equations are
only slightly more complicated than
those for flat terrain. A similar adjustment
of receptor positions is employed for
receptors above Hc. When the deflection
of each streamline is removed, and the
distortion in the plume is scaled out, an
equivalent plume-receptor geometry is
obtained.
Many of the concepts contained in
CTDM are not present in complex terrain
screening models currently in use for
regulatory assessments. Partitioning of
plume material about Hc in the vertical
and about the stagnation streamline in
the horizontal is unique to CTDM. This
partitioning is fundamental to describing
the transport of plume. material in the
flow field around hills. Furthermore, the
treatment of the effect of the hill on the
dispersion process for material above Hc>
avoids the use of the plume height
correction factor found in other models.
This factor is typically applied as a
function of stability and receptor height
only, and it leads to an inconsistent
treatment of reflection of plume material
from the lower boundary. In the
screening models, the height of the
plume above the ground is constant all of
the way from the source to a receptor,
but this height varies from receptor to
receptor. Hence, adjacent receptors at
unequal terrain elevations are modeled
with two very different plumes. If an
impingement computation is invoked, this
treatment produces a concentration
equal to twice that at the center of the
plume in the absence of terrain. In
CTDM, the impingement concentration is
equal to that at the center of the plume.
The method used to specify the rate
of plume growth also differs from the
other models. Both oy and oz functions
depend on the turbulence intensity,
rather than stability class. In the case of
oz, the function describing the rate of
growth with time also depends on the
scale of the mixing processes, which
depends on the elevation of the plume
above the surface, and on the
stratification and turbulence near this
elevation. In contrast, the other models
incorporate a fixed rate-of-growth
function for each stability class, and do
not contain the influence of processes at
plume height.
CTDM Evaluation Analysis
CTDM and other complex terrain
models used for rural applications,
COMPLEX I and RTDM, were evaluated
at five sites. These sites included the
three CTMD sites (CCB, HBR. FSPS) as
well as two other sites with conventional
SO2 data for a one-year period (the
Westvaco Lake paper mill and the
Widows Creek steam generating station).
A large part of the CTMD data base
(CCB, HBR, FSPS) was used in the
development of CTDM, so the overall
results for CTDM at these three sites do
not represent an unbiased test of CTDM
versus COMPLEX I and RTDM.
A series of statistical tests were run on
the models for data sets both paired or
unpaired in time and/or space. These
tests examined the models'
overprediction or underprediction bias as
well as the root-mean-square (RMS)
error, and the percentage of predictions
within a factor of two observations.
Another aspect of the evaluation
analysis involved an examination of the
spatial distribution and magnitude of
CTDM concentrations for each hour at
the three CTMD tracer sites. CTDM
performance in the LIFT and WRAP
components was assessed by examining
the behavior of hourly patterns of
predicted and observed concentrations.
Conclusions
The Complex Terrain Model
Development program objectives have
been met. The Complex Terrain
Dispersion Model, CTDM, is the primary
product of the effort. This model displays
considerable improvement over the
models that EPA has been using in
regulatory practice, especially on an
event-by-event basis. It also shows
improved performance over RTDM, a
model EPA is adopting as a third- level
screening model and which benefited
from the early findings of the CTMD
program on the importance of the
dividing streamline concept to
understanding stable flows.
CTDM is an improved and versatile
refined air quality model for use with
elevated point sources in high terrain
settings during stable conditions. Its
improvements over the screening models
currently used in complex terrain
applications can be attributed to several
factors:
its ability to use observed vertical
profiles of meteorological data (rather
than just one level) to obtain plume
height estimates of these variables;
computation of plume dispersion
parameters, oy and oz, directly from
turbulence measurements rather than
indirectly from discrete stability
classes.
Despite these advances, CTDM still
contains several limiting assumptions:
Its framework is a steady-state
Gaussian model. It is not designed for
extreme light-wind conditions with
highly variable wind directions.
The mathematical depiction of terrain
shapes is simplified from actual
shapes.
Flow interactions among different
terrain features are not explicitly
accounted for.
Meteorological data can be input to
the model for only one x,y location.
Flow deformation is treated with
linearized equations of motion for
steady-state Boussinesq flow, with
higher order terms neglected. These
assumptions are not valid for
applications involving steep terrain
(greater than about 15°) or strongly
stable flow (Froude number less than
1) if the LIFT module is used.
CTDM can be used for regulatory
applications involving a long series (e.g.,
a full year) of model simulations. Several
of its limitations are related to the desire
to keep the computer execution time
reasonable.
An operational limitation of the current
version of CTDM is that it provides
concentration estimates only for stable
hours. For multi-hour concentration
averages involving nonstable conditions,
a second model must be run to augment
the CTDM predictions. CTDM also
presents operational challenges to the
user. Detailed terrain and meteorological
data must be provided. "Isolated" terrain
elements need to be defined, and this
task can be complicated by
superimposed and/or interconnected
features. The considerable demands for
meteorological input, while necessary,
represent a significant increase over
those for current models that use a single
level of data.
The CTDM user must be careful in
obtaining the proper meteorological data
for input to the model. CTDM can be
very sensitive to errors in wind direction,
for example. Plume oy and oz calculations
-------
are critically dependent upon on-site
turbulence measurements at plume
height. The evaluation results show thet
use of low-level or poor vertical
resolution measurements will degrade
the performence of CTDM on an event-
by-event basis. The use of tall towers or
doppler acoustic sounders will be
necessary to obtain representative wind
and turbulence data. The capability for
accurate remote temperature sensing is
still being developed, but representative
AT measurements are essential for
obtaining accurate concentration
estimates. Such measurements can be
obtained from two levels on a tall tower
or two separate (but electronically linked)
shorter towers (one on a hill) if
instruments are placed well away from
the ground (e.g., 50 meters or higher) on
each tower.
It is useful to compare CTDM's
normalized mean square error values
with those of EPA refined models as
listed in Appendix A of the Guideline on
Air Quality Models (Revised), 1986.
CRSTER has been tested at tracer sites
in Illinois (flat site) and Tennessee
(moderately hilly site). These
experiments, sponsored by the Electric
Power Research Institute, featured
several weeks of data collection at a
network of 150-200 tracer samples. ISC
was tested with tracer data bases
collected by the American Gas
Association (AGA) at two natural gas
compressor stations. The comparison
shows that CTDM's performance at the
CTMD tracer sites is comparable to
those of EPA-designated refined
models in similar test environments.
CTDM, while showing good
performance at the evaluation sites, also
exhibits an overprediction tendency at
most sites tested; this is important for
regulators who are interested in
protecting air quality through the use of
analytical modeling techniques. The
most serious underprediction result, at
Hogback Ridge (CFsBr), is associated
with mobile crane tracer releases close
to the ridge, while using meteorological
data from the main tower farther from the
ridge. This supports the concept that the
location as well as the vertical resolution
of the meteorological data must be
designed with care for CTDM use.
The data analyses performed during
this program effort support the concept
that there are inherent limits to our ability
to predict measured or observed air
quality concentrations. Improvements to
models, such as those accomplished in
this effect, establish confidence that a
model is properly accounting for the
physical phenomena involved, and is
therefore "fair" in its application to
different situations. It is especially
noteworthy in this regard that CTDM
consistently performed well with all of the
data sets used, in contrast to the other
models tested. Nevertheless, the effort
has not resulted in a "breakthrough" in
reducing statistical uncertainty
associated with individual predictions
versus observations. The use of a high
resolution profile of meteorology
measurements with height resulted in
improvements to CTDM's performance. It
is apparent from case-study analyses
that further model performance
improvements would emerge from an
increase in the information on horizontal
as well as vertical variations in
meteorological data.
&U. S.GOVWNMBff PRINTING OFFICE: 1988/548-158/67100
-------
-------
David G. Strimaitis and Robert J. Yamartino are with Sigma Research
Corporation, Lexington, MA 02173; Robert J. Paine and Bruce A. Egan are
with ERT, Inc., Concord, MA 01742.
Peter L Finkelsteln is the EPA Project Officer (see below).
The complete report, entitled "EPA Complex Terrain Model Development: Final
Report," (Order No. P8 86-1621101 AS; Cost: $38.95, subject to change)
will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S3-88/006
0000329 PS
------- |