Fifth Symposium on the Urban Environment, American Meteorological Society,	PB2004-106739
Vancouver, BC, Canada 23-28 August 2004	EPA/600/A-04/077
9.1	URBAN MORPHOLOGICAL ANALYSIS FOR MESOSCALE METEOROLOGICAL
AND DISPERSION MODELING APPLICATIONS: CURRENT ISSUES
Steven J. Burian1*, Michael J. Brown2, Jason Ching3, Mang Lung Cheuk4,
May Yuan4, WooSuk Han1, and Andrew T. McKinnon1
department of Civil & Environmental Engineering, University of Utah, Salt Lake City, Utah
2Energy and Environmental Analysis Group, LANL, Los Alamos, NM
3ASMD, ARL, NOAA, on assignment to U.S. Environmental Protection Agency
4Department of Geography, The University of Oklahoma, Norman, OK
1. INTRODUCTION
Accurate predictions of air quality and atmospheric
dispersion at high spatial resolution rely on high fidelity
predictions of mesoscale meteorological fields that
govern transport and turbulence in urban areas.
However, mesoscale meteorological models do not
have the spatial resolution to directly simulate the fluid
dynamics and thermodynamics in and around buildings
and other urban structures that have been shown to
modify micro- and mesoscale flow fields (e.g., see
review by Bornstein 1987). Mesoscale models therefore
have been adapted using numerous approaches to
incorporate urban effects into the simulations (e.g., see
reviews by Brown 2000 and Bornstein and Craig 2002).
One approach is to introduce urban canopy
parameterizations to approximate the drag, turbulence
production, heating, and radiation attenuation induced
by sub-grid scale buildings and urban surface covers
(Brown 2000). Preliminary results of mesoscale
meteorological and air quality simulations for Houston
(Dupont et al. 2004) demonstrated the importance of
introducing urban canopy parameterizations to produce
results with high spatial resolution that accentuates
variability, highlights important differences, and
identifies critical areas. Although urban canopy
parameterizations may not be applicable to all
meteorological and dispersion models, they have been
successfully introduced and demonstrated in many of
the current operational and research mode mesoscale
models, e.g., COAMPS (Holt et al. 2002), HOTMAC
(Brown and Williams 1998), MM5 (e.g., Otte and Lacser
2001; Lacser and Otte 2002; Dupont et al. 2004), and
RAMS (Rozoff et al. 2003).
The primary consequence of implementing an
urban parameterization in a mesoscale meteorological
model is the need to characterize the urban terrain in
greater detail. In general, urban terrain characterization
for mesoscale modeling may be described as the
process of collecting datasets of urban surface cover
physical properties (e.g., albedo, emissivity) and
morphology (i.e., ground elevation, building and tree
height and geometry characteristics) and then
processing the data to compute physical cover and
morphological parameters. Many of the surface cover
and morphological parameters required for mesoscale
* Corresponding author address'. Steve Burian, 122 S.
Central Campus Dr., Suite 104, Salt Lake City, UT
84112; E-mail: burian@eng.utah.edu
meteorological models are also needed by atmospheric
dispersion models. Thus, most of the discussion below
is relevant to both types of modeling.
In this paper, the term urban morphological analysis
will be used to define the component of urban terrain
characterization concerned with the morphological
parameters. Furthermore, the focus will be building
morphological parameters; therefore, the term urban
morphological analysis will refer exclusively to the task
of inventorying, computing or estimating building
morphological parameters. Several approaches to
perform urban morphological analysis exist; however, all
have in common three types of practice issues related
to the uncertainty of (1) data, (2) parameter definitions
and calculation methods, and (3) extrapolation
techniques. The objective of this paper is to describe
the state-of-the-practice of urban morphological analysis
by reviewing the primary approaches presented in the
literature and outlining and commenting on key aspects
of the three types of practice issues listed above.
2. BACKGROUND
As described above, the urbanization of numerical
models has introduced the problem of defining the
urban canopy with a set of representative geometric,
radiation, thermodynamic, and surface cover
parameters. These urban canopy parameters (UCPs)
defined broadly include aerodynamic roughness
properties (e.g., roughness length), building height
characteristics (e.g., mean height, standard deviation,
histograms), building geometry characteristics (e.g.,
height-to-width ratio, wall-to-plan area ratio, complete
aspect ratio), building volume characteristics (e.g.
building plan and frontal area densities), radiation
trapping parameters (e.g., sky view factor), surface
cover properties (e.g., impervious surfaces, albedo),
surface material properties (e.g., heat storage capacity,
emissivity), vegetation type, height and geometry, and
more. The types and attributes of datasets required for
urban terrain characterization depends on the
processes being simulated and the spatial and temporal
scales of interest (Grimmond and Souch 1994). But in
general characteristics at the micro-scale (10 2 to 103 m)
must be determined and aggregated or averaged to the
grid cell size of the model. Representing this level of
detail in mesoscale simulations is needed to accentuate
the spatial heterogeneities of simulated surface
temperatures, increase the value of turbulent kinetic
1

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
energy production, and increase the planetary boundary
layer height to better represent urban areas (e.g.,
Dupont et al. 2004; Holt et al. 2002).
A handful of researchers over the years have
pioneered the work on obtaining surface cover and
morphological parameters for cities at the micro to
neighborhood scale (102 to 104 m) (e.g., Ellefsen
1990/1991; Theurer 1993,1999). Grimmond and Souch
(1994) were among the first researchers to present a
geographic information system (GlS)-based technique
for representing surface cover and morphological
characteristics of the urban terrain for urban climate
studies. Petersen and Parce (1994) and Petersen and
Cochran (1998) presented software (ROUGH) for
estimating the geometric parameters of buildings and
structures in urban and industrial sites. Cionco and
Ellefsen (1998) and Ellefsen and Cionco (2002) updated
Ellefsen's (1990/1991) morphological inventorying
procedure using a 100 m X 100 m grid cell size (and
then a 50 m X 50 m cell size) for use in a high resolution
wind flow model and included more characteristics of
urban canopy elements in the database.
Grimmond and Oke (1999) reviewed several
methods to define aerodynamic characteristics of urban
areas using morphometric approaches. The work
compared several methods to determine the roughness
length, displacement height, depth of roughness
sublayer and aerodynamic conductance based on
measures of building and tree morphology. GIS was
developed for 11 sites in seven North American cities
and were used to characterize the morphological
characteristics of the terrain and, using the
morphometric equations, the aerodynamic parameters.
With recent advancements in data collection and
management, digital 3D building and tree datasets have
been developed for many locations in the U.S. providing
an available data pool for automated and semi-
automated analyses to compute UCPs. Computer
software products including GIS and image processing
tools have also been enhanced and now large areas
covered by 3D digital building and tree datasets can be
analyzed automatically to extract morphological
information (e.g., building height and geometry
characteristics, roughness length). Several researchers
have developed automated and semi-automated
computational procedures to process the 3D building
and tree data to obtain UCPs. Ratti and Richens
(1999), for example, built upon the initial effort of
Richens (1997) to implement efficient urban terrain
analysis algorithms in an image processing framework
built within the MATLAB software package. Ratti et al.
(2002) used the image processing approach to compute
building plan and frontal area densities, distribution of
heights, standard deviation, aerodynamic roughness
length, and sky view factor for three European cities
(London, Toulouse, and Berlin) and two U.S. cities (Salt
Lake City and Los Angeles). The results illustrated the
roughness length differences between European and
U.S. cities.
Burian et al. (2002) presented an approach using
GIS to process 3D building datasets to compute building
height characteristics (mean, standard deviation, plan-
area-weighted mean, histograms), plan area density,
frontal area density, wall-to-plan area ratio, complete
aspect ratio, height-to-width ratio, roughness length, and
displacement height. The UCPs were calculated for
each grid cell in predefined grid meshes and the
average values for each land use type were also
determined. This automated GIS approach was used to
compute UCPs for Los Angeles, Phoenix, Salt Lake
City, Portland, Albuquerque, Oklahoma City, Seattle,
and Houston. Comprehensive reports are available for
each city (e.g., Burian et al. 2002; visit
www.civil.utah.edu/~burian for copies of the reports).
The GIS approach has recently been expanded to
include analysis of 3D vegetation, other 2D GIS
datasets (e.g., roads) and multi-spectral imagery to
compute an expanded set of parameters including
surface cover fractions, impervious surfaces, sky view
factor, predominant street orientation, and more (Burian
et al. 2003). The processing capability continues to be
enhanced and is currently available as a graphical user
interface tool using a VBA macro for the ESRI ArcGIS
software package.
Long et al. (2002) developed and tested the DFMap
software to process vector building and vegetation data
(BDTopo) available from the French National
Geographic Institute (IGN). With DFMap, a user can
select a cell size and wind direction to compute a series
of morphometric and aerodynamic roughness
parameters. Long (2003) used the DFMap software to
compute morphological statistics and define urban land
use/cover types using an unsupervised k-means
analysis. The analysis tools and approach were tested
using data for the city of Marseille. Long et al. (2003)
extended the DFMap application by incorporating the
analysis of multi-spectral and panchromatic imagery in
an attempt to improve the definition of urban surface
cover.
Urban morphological approaches have evolved in
less than two decades from detailed inventorying using
aerial photographs and extensive field surveys to
computationally intensive processing of integrated 2D
and 3D GIS and multi-spectral imagery datasets. The
users of such data have also expanded to include
federal agencies (e.g., LANL, LLNL (e.g., Chin et al.
2000), DTRA (Pace 2002), U.S. Environmental
Protection Agency (e.g., Ching et al. 2002; Dupont et al.
2004), U.S. Army and Naval Research Laboratories
(e.g., Cionco and Luces 2002; Holt et al. 2002)),
university research centers (e.g., University of Houston
Institute of Multidimensional Air Quality Studies
(IMAQS) (Daewon Byun, personal communication)),
university researchers (e.g., Rozoff et al. 2003), and
private consultants (Haider Taha and Robert Bornstein,
personal communication). During this time, the basic
concept of defining UCPs for given homogeneous areas
(e.g., land use, land cover, terrain zone) or model grid
cells has remained the same, but new developments
have improved the methods used to calcualte the UCPs.
Even with the advancements, the need for datasets
covering large areas and extensive data management
and processing requirements limit the ability to derive
gridded parameter datasets for entire mesoscale model
2

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
domains. Further developments in data collection and
processing are needed. In addition, there is a need to
further refine urban morphological analysis methods by
standardizing processes and reducing the uncertainty of
methods. The following section outlines several issues
that are currently of interest to those conducting urban
morphological analyses for mesoscale meteorological or
atmospheric dispersion modeling applications.
Resolving these issues will contribute to improvements
of analysis methods and likely increase the level of
accuracy of calculated UCPs.
3. MORPHOLOGICAL ANALYSIS ISSUES
Urban morphological analysis methods have
matured during the past two decades, but there still
remain several operational issues to be resolved. Three
primary types of issues facing those deriving UCPs are:
(1)	Data Issues. What data are available, what
are the available levels of resolution, accuracy
and detail, and what levels of resolution,
accuracy and detail are necessary for UCP
calculations?
(2)	Definition and Calculation Issues. What are
the current UCP calculation methods and what
are the effects of ambiguity and uncertainties in
parameter definitions and calculation methods
on UCP values and mesoscale meteorological
and dispersion simulation results?
(3)	Extrapolation Issues. What are	the most
accurate means to extrapolate	UCPs to
mesoscale meteorological and	dispersion
model domains?
The salient points associated with the three types of
issues listed above are described in the following three
subsections.
3.1 Data Issues
3.1.1 What Data are Available?
The building elevation data used to define the
urban morphology is usually obtained by:
•	Analyzing aerial photographs to estimate
building data (time consuming, feasible only
for very small areas)
•	Performing ground surveys (time consuming,
feasible for very small areas; elevation can be
more accurately obtained than by analysis of
aerial photographs)
•	Analyzing stereographic images (can be time
consuming to perform manual digitization;
elevation can be accurately estimated)
•	Analyzing airborne LIDAR data (requires
managing and processing massive datasets;
newness of LIDAR technology presents
several problems; elevation measurements can
be obtained with vertical and horizontal
accuracies of 15 cm RMSE)
•	Obtaining/purchasing datasets from
government or municipal agencies (may be
outdated or lack information necessary to
compute all UCPs)
•	Purchasing datasets from commercial vendors
(potentially high cost; enhanced data products
can be delivered)
The data are obtained in two basic forms: raster
and vector. Raster data are usually square-grid cells
with one elevation attribute per cell. Vector data use
one or more polygons to represent the building footprint
and rooftop. Vector data can represent the geometry
and details with higher precision than raster data. In
addition, vector data can represent multiple building
layers and can contain multiple attributes per polygon
(e.g., height, roof pitch, color, material). Building
polygon data in vector form can be obtained from
municipal governments and commercial vendors. Data
products range from the raw stereographic pairs or
airborne LIDAR data to the finished end products of
building polygon vector or full-feature raster digital
elevation model (DEM) datasets. Other options for
obtaining building data are federal government
agencies. For example, a potential major source of
building morphology data in the U.S. is the so-called
"120 cities" project. This project involves a consortium
of federal agencies to collect and prepare airborne
LIDAR databases for cities for domestic preparedness
applications. Unfortunately, the current availability of
the data and the ultimate distribution policy is uncertain.
3.1.2 How Accurate are the Data?
A critical question to be asked of all available
building datasets is the accuracy. The primary errors
possible include:
1.	Buildings may currently exist, but are not
included in the dataset
2.	Buildings may not currently exist, but are
included in the dataset
3.	Several individual buildings may be
represented as a single building
4.	A single building may be represented as
several individual buildings
5.	Building outline may not accurately represent
the actual building shape
6.	Location of building polygon or parts of the
polygon may be inaccurate
7.	Building heights may be inaccurate
The causes of these data errors could be operator error
during extraction, poor quality assurance/quality control
(QA/QC) procedures, limitations of building feature
extraction methods, changes of morphology during data
collection, and the age of the building dataset (which
may cause the data to not represent the morphology for
the time period of interest). To illustrate several of the
common building morphological dataset errors, Figure 1
3

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
was created showing red outlines representing building
footprints extracted using an automated LIDAR DEM
processing technique overlaid onto an aerial photo. Of
the potential errors listed above, numbers 3, 4, 5, and 6
can be directly observed in the figure. Although not
observable the other three errors were also contained in
this sample dataset.
Figure 1. Sample residential area in Salt Lake City.
Red outlines represent a vector dataset derived through
an automated analysis of a full-feature DEM.
Airborne LIDAR has emerged as one of the primary
sources for the development of 3D urban models
(Priestnall et al. 2000). in most cases, both horizontal
and vertical resolutions of LIDAR data can be down to
one meter. Consequently, the high resolution may be
misunderstood as high accuracy. It should be
understood that LIDAR data processing is a nontrivial
task. Three general approaches have been developed
for urban 3D data analysis using LIDAR data,
categorized as using an edge operator, mathematical
morphology, and height bins, respectively (Zhou et al.
2004). In practice, LIDAR data processing may involve
all three approaches. To assess possible LIDAR data
inaccuracies a comparison was made between building
heights derived from LIDAR data for Oklahoma City and
field observations. The LIDAR data was acquired in late
October 2001 by an Optech Airborne Laser Terrain
Mapper (ALTM) 2033 sensor with a differential Global
Positioning System. The Joint Precision Strike
Demonstration (JPSD) Program Office of the U.S. Army
processed the raw LIDAR data and provided data of
LIDAR estimated building heights in downtown
Oklahoma City. Preliminary findings identified four
categories of significant differences between building
heights estimated from LIDAR data and those from field
observations. Significant differences were defined as
those greater than 5 meters, which represents at least
one floor of the structure. The four types of differences
are briefly described below; more details and examples
will be included in the presentation.
The first LIDAR inaccuracy found from the
Oklahoma City analysis is common to all data collection
and building extraction approaches. It involves the sate
of the building morphology changing from the time of
data collection. Buildings that were recorded in the
dataset may have been demolished or new buildings
may have been built in areas where buildings previously
did not exist. The second type of LIDAR inaccuracy
was observed at locations where buildings were
constructed of highly reflective materials. Figure 2
shows two buildings that had heights under-estimated
by the LIDAR data by more than 5 m. The building
shown on the left side of the figure is a greenhouse with
a curved roof-top. A significant portion of the south end
of the building has an under-estimated height. The
second example is a buiiding nearly completely glass-
covered on the south side. The end of the building
covered by glass is noted to have a height difference of
greater than 5 m. The third type of difference noted was
associated with buildings having complex or narrow
structures, e.g., passages, gaps, platforms, extensions.
The LIDAR-derived building heights contained both
under- and over-estimations in the vicinity of such
complex building elements. And the fourth difference
between the LIDAR-derived heights and field measured
heights was noted in the vicinity of vegetation adjacent
to buildings. Trees adjacent to buildings that overhang
the building rooftop will cause the LIDAR-derived
building height to be over-estimated because the LIDAR
does not differentiate between objects.
south
Downtown Oklahoma City
Figure 2. Examples of association between building
materials and under-estimated building heights based
on LIDAR data.
While four categories of differences between LIDAR
estimates and field observations have been presented
here, these differences represent preliminary findings
and are by no means exhaustive. Current work focuses
on trying to understand the physical causes of the
observed uncertainties associated with the LIDAR-
derived building heights. It is also expected that other
uncertainties tied to building morphological
characteristics will be noted. inaccurate data is
prevalent but the effect of data inaccuracies on UCP
4

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
values and numerical model results is being studied.
Results will be included in the presentation.
3.1.3 What are the Available Levels of Data Detail?
The detail of the data used to compute the UCPs
may also influence the resulting UCP value. In terms of
detail, the necessity to represent precise building
geometry and ancillary attached structures is the
primary issue. For example, a building dataset might
represent the change in building geometry with height or
include rooftop structures. The alternative to this level
of detail is to simply represent each building with a
single polygon that has a uniform shape with height. An
example of the difference in detail for a small section of
downtown Salt Lake City is shown in Figure 3. The top
part of the figure shows a CAD dataset with a very high
level of detail included, while the bottom part of the
figure shows a GIS shapefile with a relatively low level
of detail. Visual comparison of the top and bottom parts
of the figure suggests that the low-level of detail
captures the general form of the location. Therefore,
average UCPs for a 1-km model grid cell size are likely
to not be affected by approximating each building as a
uniform polygon with a single height. This needs to be
quantitatively confirmed.
One question to consider if a uniform polygon with
single height attribute is to be used to calculate the UCP
is the choice of average or maximum height as the
attribute used in the calculation. For some UCPs (e.g.,
sky view factor), rooftop structures are probably not
important and in areas with tall buildings, the rooftop
structures are also probably negligible. In these cases,
the choice of average or maximum building height will
probably not impact the UCP value significantly. In
residential land uses however the use of average or
maximum height may alter the UCP values. For
example, consider Figure 1, which contained an aerial
photograph of a residential block in downtown Salt Lake
City consisting of predominantly single-story, single-
family homes with basements and pitched roofs. The
buildings were extracted from LIDAR data using an
automated approach and the heights were determined
in two ways: (1) finding the average height of raster cells
inside each building polygon boundary and (2) selecting
the maximum raster height inside each building polygon
boundary. Using the mean height for each building, the
mean and standard deviation of building height for the
entire block were calculated to be 3.8 m and 0.7 m,
respectively. However, using the maximum height of
the rooftop, the mean and standard deviation were
found to be 5.5 m and 1.5 m. These differences are
significant for the UCP value (-45% and 100%
differences), but the significance of the magnitude of
differences is uncertain for mesoscale meteorological or
dispersion model results considering that the UCP value
will be aggregated with other data for a grid cell size on
the order of ~1 km2. Further analysis is needed.
Figure 3. Comparison between a high-detail building
dataset (top) and a coarse dataset (bottom) for smal
section of downtown Salt Lake City. The Heber-Wells
Building is identified in both for frame of reference.
3.2 UCP Definition and Calculation Issues
A second concern facing those deriving UCPs
involves the decisions of what are the precise UCP
definitions, what calculation methods and tools to
choose, what formats of input data to expect, and what
outputs need to be produced. The following questions
are pertinent to these concerns.
3.2.1 What is the Effect of Ambiguous UCP Definitions
on Calculated Values?
The definition of the UCP may be ambiguous and
thus not account for all potential real-world building
arrangements. For these circumstances calculation of
the UCP will be subjective because the individual will
need to decide whether to ignore the unforeseen
situation, implement a simplified calculation approach,
or develop a modified calculation approach accounting
for the encountered circumstance. The potential impact
of UCP definition ambiguity on UCP values will be
considered herein using mean building height. The
presentation will contain the assessment of other UCPs
including height-to-width ratio and frontal area index.
Mean building height is seemingly a well-defined
5

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
UCP. However, there are several ways in which the
mean can be computed. The most common are: simple
average, weighting by plan area (top or bottom or
average), or weighting by frontal area. Based on
analysis of real building data the values can be
significantly different depending on the definition
chosen. For example, consider Table 1 which lists the
computed mean building heights using a simple average
and weighted by plan area for the downtown core areas
of eight cities in the U.S. The plan area weighted mean
building height ranges from being approximately equal
to the simple mean building height to being more than
40% greater. What these differences mean in terms of
mesoscale simulation results must be determined.
Table 1. Comparison of mean building heights for
downtown core areas of U.S. cities.

Mean
Mean

Building
Building
City
Height (m) -
Height (m) -

Simple
Plan Area

Average
Weighted
Albuquerque, NM
15.2
24.6
Houston, TX
22.8
36.0
Los Angeles, CA
45.0
44.1
Oklahoma City, OK
19.4
25.4
Phoenix, AZ
17.2
21.2
Portland, OR
18.1
18.5
Salt Lake City, UT
23.6
41.5
Seattle, WA
21.1
31.9
3.2.2	What is the Effect of Ambiguous UCP Calculation
Methods?
In addition to the ambiguities associated with UCP
definitions, there are ambiguities associated with UCP
calculation methods. Most notable is the many
equations available to compute the aerodynamic
roughness parameters (roughness length and
displacement height), with confusing guidance on
relative applicability and accuracy. Grimmond and Oke
(1999) compared many of the methods to compute the
aerodynamic roughness parameters and could not
conclusively identify an order of preference, although
several of the more intuitively accurate methods were
highlighted as recommended. Additional studies must
be conducted to investigate the differences between the
results produced by the various approaches for real
cities and how the differences affect mesoscale
meteorological and dispersion model results.
3.2.3	What are the Advantages and Disadvantages of
UCP Derivation Approaches?
Several approaches have been presented in the
literature to derive UCPs (see Background section
above):
1.	Conduct a field survey to measure building
geometry and height information and record
construction materials.
2.	Analyze high-resolution aerial photographs to
estimate building geometry and height
information and record construction materials.
3.	Perform automated or semi-automated image
analysis of remotely sensed data to compute
the building information.
4.	Perform automated or semi-automated
analysis of full-feature digital elevation models
(DEM) or vector representations of buildings
using GIS or image processing software.
Conducting a field survey can provide a high level of
detail, yet it is time consuming/labor intensive and
requires being in the location of the buildings. It is
feasible for analysis of very small areas of cities.
Analysis of high-resolution aerial photographs is also
time consuming/labor intensive, but it does not require
the analyst to be present at the building location. It is
however similar to field surveys in that it is only
applicable for small areas of cities. Automated or semi-
automated analysis approaches are the most time
efficient and once the codes are written to perform the
automated calculations, labor requirements are limited
(computer time is the only requirement). However,
removing the trained analyst from the interpretation and
measurement of building characteristics eliminates
subjectivity at the cost of reducing quality assurance
and quality control and oversight. In addition, simplified
approximations to the calculations may be necessary to
execute an automated approach and this may further
reduce the accuracy of the results.
Despite the reduction in accuracy potentially
caused by using an automated analysis approach, it is
the only way to derive UCP coverage for large areas in
a reasonable amount of time. Besides being able to
process large areas, the image and GIS processing
approaches have other advantages including:
•	Spatial data in GIS compatible formats (e.g.,
roads) is readily available for incorporation into
the analysis and for map making.
•	The quality and availability of building
morphological data is continuously improving
and the automated approaches are specifically
designed to work with these datasets.
•	GIS and image processing software continues
to improve and computer processing speed
continues to be increased, all of which
enhances the ability of automated approaches
to provide more accurate UCP values more
efficiently.
Some questions may arise regarding the use of image
processing versus GIS software. The image processing
software may provide more efficient computation than
the GIS, but with the speed of today's computers the
difference in time is negligible compared to the time to
be invested in the modeling and analysis activities.
6

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
3.2.4	How Important Are Trees for Accurate UCP
Values?
Because of their relative importance in downtown
areas, building morphological data have been collected
in cities more often than tree morphological data. Yet
tree morphology can play an important role in the overall
morphological characteristics of an urban area,
especially residential zones and other areas with
significant vegetation elements. Including trees is
recommended by Grimmond and Oke (1999) in order to
better represent the overall roughness of the urban
surface. However, due to the difficulty to obtain
sufficient tree morphological information for large areas,
the importance of sparse coverage of urban trees
remains relatively uncertain. Further study is needed.
3.2.5	How Sensitive are UCP Values to Data
Resolution?
Building datasets typically are vector products
(perhaps derived from raster data, e.g., LIDAR), but the
computation of many UCPs is most easily performed on
raster data. For such circumstances, the vector data
must be converted into a raster dataset with a selected
horizontal resolution. Higher resolution data should
produce a more accurate UCP value, but at the cost of
higher computational requirements. A study is needed
to determine the methods of rasterization and how using
coarser cell sizes will affect the building data and then
how raster cell size affects processing time, accuracy of
the UCP value, and the ultimate mesoscale
meteorological or dispersion model results.
3.2.6	Is a Consistent Data Model/Format Necessary?
Currently, a standard data model for urban
morphological input data or output format does not exist.
Although some agencies have identified the need for
standardization of the collection of building and
morphological data (e.g., DTRA, John Pace, personal
communication), consensus has not been established
by those performing urban morphological analysis in
support of mesoscale modeling. In terms of the input
data used to compute UCPs, having a standard data
model would promote data sharing and the reuse of
developed software products and methodologies for
urban morphological analysis. Further, reaching
consensus on base data requirements would encourage
data collectors to obtain information needed by all users
thus making the collected data more valuable. In terms
of output data, a consistent format may be unnecessary
because the advancements in software products has
enabled numerous forms of building morphological data
to be accessed and processed with a variety of
computational tools. For example, data in CAD form
can be accessed in GIS and processed and vice-versa.
The use of proprietary data forms (e.g., ESRI shapefile)
may limit the applicability of some software products,
but conversion tools are generally available. More
consideration is needed to develop data standards that
meet the needs of the widest section of the mesoscale
meteorological and atmospheric dispersion model user
community.
3.3 Extrapolation Issues
3.3.1 What is the Appropriate Size Area for UCP
Analysis to Obtain Meaningful Building Statistics?
Time or budget constraints often limit the size of the
area that can be morphometrically analyzed. Use of a
small analysis area to compute mean parameters as a
function of land use/cover types (or other homogeneous
units) can lead to errors when extrapolating to larger
areas of the modeling domain outside of building data
coverage (Burian et al. 2003). The heterogeneity of
urban terrain may cause the resulting land use-specific
mean UCPs to vary depending on the size of the area
included in the calculation.
To investigate this question an analysis of a large
650,000 building dataset for Houston was performed.
The downtown core area of the city was delineated and
a set of UCPs were calculated for the delineated area
and the mean value was determined for each land use
type. Then the boundary of analysis was increased
incrementally and the UCPs re-calculated (see Table 2
for a summary of the size characteristics of the
incremental analysis zones). The trends of the UCP
values as the boundary of the analysis zone increased
were quantified. The choice of the initial analysis zone
to be the downtown core area may influence the results
of this analysis. But, the selection of the downtown core
area is appropriate for this preliminary analysis because
in most cases the downtown will be of interest from a
modeling perspective (and will have unique
morphological characteristics). To eliminate this
subjectivity, the analysis will be repeated using a series
of randomly selected initial analysis zones and the
results will be reported in the presentation.
Based on results using the downtown core area as
the center of the analysis zones, Figure 4 shows the
trend of the mean building height for several urban land
use types as the size of analysis area increases. The
mean building height in Residential and Industrial land
uses was found to not be significantly sensitive to the
size of the analysis zone, but the Commercial &
Services values do change significantly as the size of
the analysis area increases. In the initial analysis area
extent encompassing the downtown core area, the
buildings are predominantly high-rise and the land use
is predominantly Commercial & Services. As the
analysis extent increases the character of the
Commercial & Services land use changes from high-rise
to shopping malls, strip malls, and other forms with
much shorter building heights (and less variable
heights) than the downtown core area. Thus, the
observed change to mean building height of the
Commercial & Services land use is understandable.
However, Residential and Industrial land uses are
usually not prevalent in the downtown core area and as
the analysis extent increases the building heights
typically do not become significantly smaller. This
observation is also noted in Figure 4.
7

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
Table 2. Characteristics of analysis zones.
Zone
Area
(km2)
Number of
Buildings
Building Density
(# /km2)
1
3.4
489
144
2
7.5
1539
205
3
13.1
5126
391
4
27.0
15888
588
5
53.3
46058
864
6
105.9
99868
943
7
207.0
180746
873
8
371.4
278266
749
9
722.1
434285
601
Figure 5 suggests that building plan area fraction is
sensitive to analysis area for all three land use types.
The results indicate that the plan area fraction
decreases as the analysis zone increases, reflecting a
decreasing building density with distance from the
downtown core area. This is most likely due to including
areas of lower density housing and industrial parks
normally found on the outskirts of urban areas. Analysis
of the behavior of other UCPs as the analysis zone size
increases will be reported in the presentation.
3.3.2 Is a Standard Urban LULC Classification Needed?
Land use and land cover have served for many
years as surrogate data layers to define surface
parameters in mesoscale meteorological and dispersion
models. The approach involves defining model
parameters for LULC classes based on analyzed
samples of the use/cover or literature values
representative of the use/cover (e.g., a roughness
length for urban land use). The LULC dataset then
serves as the extrapolation medium to parameterize the
entire model domain using the established parameter
values. This approach has been necessary because
datasets describing these model parameters or datasets
that they are derived from (e.g., 3D building databases)
were not available or could not be efficiently analyzed
for areas covering the extent of mesoscale model
domains.
-~-Commercial & Services
-¦-Residential
-A-Industrial
Overall
2 10
0 100 200 300 400 500 600 700 800
Analysis Area (km2)
Figure 4. Mean building height as a function of analysis
area size.
0.28
0.24
« 0.20
0.16
Commercial & Services
Residential
-a- Industrial
-x-Overall
0.12
0 100 200 300 400 500 600 700 800
Analysis Area (km2)
Figure 5. Building plan area fraction as a function of
analysis area size.
The available LULC datasets themselves, however,
may potentially have limitations. For example, the U.S.
Geological Survey (USGS) LULC dataset has been
commonly used and is freely available in nationally
consistent form for the U.S. But, the primary source of
data for the USGS LULC dataset were NASA high-
altitude aerial photographs and National High-Altitude
Photography (NHAP) program photographs collected
mostly during the 1970s. These data are outdated for
representing current conditions in areas at the
urbanizing fringe of cities, especially those that have
experienced massive growth during the past 30 years.
Moreover, the urban land use categories in the
Anderson classification scheme used by the USGS
(Anderson et al. 1976) are not based on morphological
characteristics. A potential source of more updated
LULC data is the National Land Cover Dataset (NLCD)
(Vogelmann et al. 1998). However, the NLCD is based
on semi-automated classification of remote sensing data
with some incorporation of ancillary data; consequently,
the definition of urban land use types is limited to a
small number because of the heterogeneous surface
properties (Vogelmann et al. 1998). The four urban land
use types represented in the NLCD (Low Intensity
Residential, High Intensity Residential, Commercial/
Industrial/Transportation, and Urban Vegetation) in most
cases will include a wide range of urban surface
fractions (e.g., paved areas, rooftops, landscaped
areas, bare soil, etc.) with different reflective properties
within each class, which will confuse classification
algorithms and potentially cause misclassification.
Moreover, the aggregation of Commercial, Industrial,
and Transportation land uses into a single land use
category will mask the heterogeneity of the urban
surface because the building morphological
characteristics of these three individual land uses are
not similar (e.g., see building statistics in Burian et al.
2002 or Grimmond and Oke 1999).
As briefly noted, the USGS and NLCD datasets
have potential drawbacks for UCP definition and
extrapolation to mesoscale meteorological and
dispersion model domains. From a meteorological or
8

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
dispersion modeling perspective, numerous researchers
have devised more detailed LULC classification
schemes to overcome the potential deficiencies of the
aforementioned national datasets and provide more
accuracy for their analyses. Ellefsen (1990/1991) for
example, derived an urban LULC dataset for urban
climate and meteorological modeling applications. His
classification system included 17 urban terrain zones
(UTZ) that were homogeneous units from a building
morphological perspective. Each zone was defined to
have a distinctive mix of function, development age,
street pattern, lot configuration, and type of construction
and density and height of buildings. A primary
difference between UTZ categories is whether the
buildings are attached or detached. UTZs with
detached structures were further subdivided by the
closeness of the structures.
Using a morphological-based classification
approach similar to Ellefsen, Theurer (1993, 1999)
identified a set of nine typical building arrangements for
cities in Germany. Another classification approach
incorporating morphological characteristics was
introduced by Grimmond and Souch (1994). They used
a 36-category urban land use classification scheme to
represent urban terrain in a GIS-based surface
characterization methodology. The first level of land
use categories included traditional single and apartment
residential, commercial/industrial, institutional,
transportation, vacant, vegetated, impervious, and
water. Sub-categorization was based on building height
and density and cover fractions.
Burian et al. (2002) also used a two-tier
classification approach, but their scheme was based on
the Andersen Level 2 urban land use categories as the
first tier. Morphology was then incorporated to
subdivide traditional USGS urban land use classes into
morphologically-based sub-categories. For example,
the Residential land use category was subdivided into
low-density or high-density based on a chosen threshold
building density level. Long et al. (2003) used two
automated approaches to classify land use, one based
on morphological parameters and the other based on
analysis of multi-spectral imagery (20-m resolution
SPOT). Originally, a 7-class scheme based on
morphological parameters was tested, but was
expanded to 10-classes after high parameter variability
was noted in each class. The morphological
classification identified urban land use types well, but in
general did not define land cover well. The imagery
approach did give accurate information about the cover
type, but was unable to accurately identify urban
elements and their combinations to form the urban
terrain classes.
This sampling of urban LULC classification
approaches indicates that morphologically-based
approaches are available and can be used in urban
morphological analysis and may have application in
mesoscale modeling. The question then becomes the
need for standardization of classification and the
creation of repeatable, objective, automated approaches
to derive a spatially-consistent LULC dataset.
Standardization would benefit morphological analysis
and may benefit the mesoscale meteorological modeling
community, but perhaps would decrease the value of
previously collected data and derived relationships
unless the new land use/cover categories could be
chosen to be consistent with categories from other
classification approaches (or categories from other
classification approaches can be generalized into the
categories of the new classification approach). The
popularity of the USGS LULC dataset with the
mesoscale modeling community might also influence
the selection of classification categories. One possible
solution is to use morphological characteristics (e.g.,
building density, mean building height, vegetative cover,
impervious surface fraction) to subdivide a generalized
first level of land use/cover, but this would require a
nationally consistent morphological database.
4.	SUMMARY
This paper reviewed the historical advancements to
urban morphological analysis approaches and
discussed the details of several practice issues. The
issues discussed are not considered exhaustive, but do
provide a sampling of current practice inconsistencies
and uncertainties that could potentially reduce the
accuracy of urban canopy parameters and mesoscale
model results. Further analysis of these issues and
others will be included in the presentation.
Acknowledgements
This work was partially supported by the Department of
Homeland Security Chemical and Biological
Countermeasures Program. The United States
Environmental Protection Agency through its Office of
Research and Development also partially funded and
collaborated in the work described here. It has been
subjected to Agency review and approved for
publication. The authors wish to thank Tanya Otte for
insightful comments that helped to improve the
manuscript.
5.	REFERENCES
Anderson, J. R., E. E. Hardy, J. T. Roach, R. E. Witner,
1976: A land use and land cover classification system
for use with remote sensor data. USGS Prof. Paper
964.
Bornstein, R. D., 1987: Mean diurnal circulation and
thermodynamic evolution of urban boundary layers.
Modeling the Urban Boundary Layer, AMS, Boston,
MA.
Bornstein, R. D. and K. J. Craig, 2002: Survey of the
history of the urbanization of numerical mesoscale
models. Preprints, Fourth Symposium on the Urban
Environment, AMS, Norfolk, VA.
Brown, M. J., 2000: Urban parameterizations for
mesoscale meteorological models. Mesoscale
Atmospheric Dispersion. Ed., Z. Boybeyi., pp. 193-
255.
Brown, M. J. and M. D. Williams, 1998: An urban
canopy parameterization for mesoscale
9

-------
Fifth Symposium on the Urban Environment, American Meteorological Society,
Vancouver, BC, Canada 23-28 August 2004
meteorological models, Preprints, Second Urban
Environment Conference, AMS, Albuquerque, NM.
Burian, S. J., M. J. Brown, and S. P. Linger, 2002:
Morphological analysis using 3D building databases:
Los Angeles, California. LA-UR-02-0781. Los Alamos
National Laboratory, 66 pp.
Burian, S. J., W. Han, S. P. Velegubantla, S. R. K.
Maddula, 2003: Development of gridded fields of
urban canopy parameters for Models-3/CMAQ/MM5.
U.S. EPA Internal Report, RTP, North Carolina.
Chin, H. N., M. J. Leach, and M. J. Brown, 2000: A
sensitivity study of the urban effect on a regional-
scale model: An idealized case. Preprints, Third
Symposium on the Urban Environment, Davis, CA.
Ching, J., T. L. Otte, S. Dupont, S. Burian, and A.
Lacser, 2002: Urban morphology for Houston to drive
Models-3/CMAQ at neighborhood scales. Preprints,
Fourth Symposium on the Urban Environment, AMS,
Norfolk, Virginia.
Cionco, R. M. and R. Ellefsen, 1998: High resolution
urban morphology data for urban wind flow modeling.
Atm. Env., 32(1), 7-17.
Cionco, R. M. and S. Luces, 2002: Coupled mesoscale-
microscale model to compute neighborhood scale
wind fields. Preprints, Fourth Symposium on the
Urban Environment, AMS, Norfolk, VA.
Dupont, S., J. Ching, and S. Burian, 2004: Introduction
of urban canopy parameterization into MM5 to
simulate urban meteorology at neighborhood scale.
Preprints, Symposium on Planning, Nowcasting, and
Forecasting in the Urban Zone, 84th AMS Annual
Meeting, Seattle, WA.
Ellefsen, R., 1990/1991: Mapping and measuring
buildings in the canopy boundary layer in ten U.S.
cities. Energy and Buildings, 15-16, 1025-1049.
Ellefsen, R. A. and R. M. Cionco, 2002: A method for
inventorying urban morphology. Preprints, Fourth
Symposium on the Urban Environment, AMS, Norfolk,
VA.
Grimmond, C. S. B. and T. Oke, 1999: Aerodynamic
properties of urban areas derived from analysis of
surface form. J. Appl. Met., 38, 1262-1292.
Grimmond, C. S. B. and C. Souch, 1994: Surface
description for urban climate studies: A GIS based
methodology. Geocarto Int., 9, 47-59.
Hanna, S. R. and R. E. Britter, 2002: Wind flow and
vapor cloud dispersion at industrial and urban sites.
American Institute of Chemical Engineers, New York,
NY.
Holt, T. R., S. Chin, M. Leach, and G. Sugiyama, 2002:
Sensitivity of mesoscale real-data simulations to an
urban canopy parameterization. Preprints, Fourth
Symposium on the Urban Environment, AMS, Norfolk,
VA.
Lacser, A. and T. L. Otte, 2002: Implementation of an
urban canopy parameterization in MM5. Preprints,
Fourth Symposium on the Urban Environment, AMS,
Norfolk, VA.
Long, N., 2003: Analyses morphologiques et
aerodynamiques du tissue urbain - Application a la
climatologie urbaine de Marseille pendant la
campagne ESCOMPTE. Thesis, Ecole Centrale de
Nantes, Nantes, France.
Long, N., S. Kermadi, C. Kergomard, P. G. Mestayer,
and A. Trebouet, 2003: Urban cover modes and
thermodynamic parameters from urban database and
satellite data: A comparison for Marseille during
ESCOMPTE, Proceedings, Fifth Int. Conf. Urban
Climate, Lodz, Poland, September 2003.
Long, N., Mestayer, P. G., and C. Kergomard, 2002:
Development of a software to describe the city
morphology and to compute aerodynamic parameters
from an urban database. Preprints, Fourth
Symposium on the Urban Environment, AMS, Norfolk,
Virginia.
Otte, T. L. and A. Lacser, 2001: Implementation of an
urban canopy parameterization in MM5 for meso-
gamma-scale air quality modeling application.
Preprints, Ninth Conference on Mesoscale
Processes, AMS, Ft. Lauderdale, FL, 78-81.
Pace, J. C. and R. J. Babarsky, 2002: DTRA urban
terrain database project. Preprints, Fourth
Symposium on the Urban Environment, AMS, Norfolk,
VA.
Petersen, R. L. and B. C. Cochran, 1998: Development
and testing of methods for estimating surface
roughness length at refineries and other industrial
facilities. CCP Project 95-1214, CPP, Ft. Collins, CO.
Petersen, R. L. and D. K. Parce, 1994: Development
and testing of methods for estimating surface
roughness length at refineries. CCP Project 92-0890,
CPP, Ft. Collins, CO.
Priestnall, G., J. Jaafar and A. Duncan, 2000: Extracting
urban features from lidar digital surface models.
Computers, Environment and Urban Systems, 24(2),
65-78.
Ratti, C. and P. Richens, 1999: Urban texture analysis
with image processing techniques, Proceedings of the
CAADFutures99 Conference, Atlanta, GA.
Ratti, C., S. Di Sabatino, R. Britter, M. Brown, F. Caton,
and S. Burian, 2002: Analysis of 3-D urban databases
with respect to pollution dispersion for a number of
European and American cities. Water, Air, and Soil
Pollution: Focus, 2, 459-469.
Richens, P., 1997: Image processing for urban scale
environmental modeling. Proceedings of the Int. Conf.
on Building Simulation '97, Prague, 163-171.
Rozoff, C. M., W. R. Cotton, and J. O. Adegoke, 2003:
Simulation of St. Louis, Missouri, land use impacts on
thunderstorms. J. Appl. Met., 42, 716-738.
Theurer, W., 1993: Ausbreitung bodennaher
Emissionen in komplexen Bebauungen. Dissertation,
IHW, Universitat Karlsruhe.
Theurer, W., 1999: Typical building arrangements for
urban air pollution modeling. Atm. Env., 33, 4057-
4066.
Vogelmann, J. E., T. L. Sohl, P. V. Campbell, and D. M.
Shaw, 1998: Regional land cover characterization
using Landsat Thematic Mapper data and ancillary
data sources. Environmental Monitoring and
Assessment, 51, 415-428.
Zhou, G., C. Song, J. Simmers, P. Cheng, 2004: Urban
3d GIS from lidar and digital aerial images.
Computers & Geosciences, 30(4), 345-353.
10

-------