Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 130
Working Paper D
Inventory of Facilities
This paper addresses the inventory problem. It limits the term
inventory to the task of listing man-made facilities and their at-
tributes, rather than the parallel task of producing an inventory of
the attributes of different soil types or other geologic data, which is
a seismic hazard analysis task. Structures other than buildings, such
as lifeline facilities, can be inventoried similarly as for buildings, al-
though the information sources and data collection techniques vary.
Most major lifeline facilities are already inventoried to some extent
by their owners, and it is the more difficult problem of conducting
an inventory of buildings that is the focus of this paper.
The number of buildings and other structures in the study area
of most large-scale loss estimation studies is great. The earthquakes
selected as the planning basis for large-scale loss est~rnation studies
can be strong enough to shake 5,000 or more square miles, and the
study area of most interest often contains a population of several
million. Pre-existing files or data bases do not contain the amount or
quality of information that is desired for the purpose of estimating
earthquake losses. Inventories used for earthquake loss estimation
purposes must be developed in a highly selective manner because
this is the most time-consuming and costly step in the loss estimation
process. Thus, the inventory task is often a matter of using the data
that can be collected and organized within the budget allotted, rather
than developing the ideal inventory.
130
OCR for page 131
131
The losses of concern may be facilities damaged or rendered dys-
functional, dollar losses to facilities or dollar value of lost production,
casualties, or homelessness. The kind of loss information sought is
a determinant of the kinds of inventory information needed in the
analysis. Hence, the types of loss to be estimated must be specifically
defined prior to selecting an inventory method.
Theoretically speaking, a unique and all-purpose inventory might
be created, but its contents would include so many descriptors and
other items of information that it would not be feasible to assemble.
Moreover, given the lack of understanding about motion and damage
or ground failure and damage relationships and the prospect that
this understanding will improve in tune, the chances of anticipating
all relevant inventory data today for some future use are, indeed,
slim. Efforts to create an exhaustive inventory of information about
facilities would be a misguided, ineffective effort.
HIERARCHY OF DATA
Based on recent loss estimations prepared by different methods
and people, the best hierarchy of data items seems to be:
~ Facility location (addresses are preferred for buildings and
structures, but they are often listed only by zip code or census tract;
the census tract or other appropriate zone is used for linear or area-
wide facilities);
Type of structure;
Materialts) of construction (for the load-carry~ng system);
Height (for buildings);
Floor area (for buildings);
Date constructed;
Value (market or replacement value adjusted to a selected
base year);
Use of facility (occupancy or social function); and
· Number of people in facility at different times of day and
season.
Many other data might be added that, based on present knowI-
edge of earthquake effects, could improve the accuracy of loss esti-
mates. Among these (not necessarily in order of unportance) are:
Type of foundation system;
Configuration of facility (in plan and in elevation or section);
Special-damage control features of facility;
OCR for page 132
132
Code under which facility was constructed; and
~ Nonstructural features or contents with special fire or haz-
ardous materials characteristics.
Nevertheless, no loss estimates are possible without certain ba-
sic information about facilities. Further, some kinds of inventory
information are common for all loss estunation methods, that is, fa-
cility location, construction classification, occupancy data (number
of occupants and type of occupancy or use), and facility property
value. According to current procedures, this information is assem-
bled (inventoried) by (a) field observation or sampling, (b) review
of other previously assembled records for a given community, or (c)
extrapolation from conveniently available records to the end-form
construction data desired, such as by inferring floor area from num-
ber of employees or land-use acreage figures, degree of earthquake
resistance incorporated in the design from date of construction, or
value from floor area.
When other economic losses are to be estimated, additional in-
ventory information is needed, especially the facility's economic use,
or "social function" in the terminology of ATC-13 (Applied Tech-
nology Council, 1985), and facility contents. Economic relationships
that are not a part of the inventory also must be modeled.
Essential inventory information can be assembled in several ways,
and no single inventory method can be recoin-mended. However,
two major alternatives discussed in this paper are (~) the NOAA-
USGS method of field observation, coupled with input from local
building experts, lancI-use patterns, and census data, and (2) the
FEMA/ATC-13 method, which would use existing detailed construc-
tion class inventories where available but in practice would generally
rely on extrapolations from economic data to impute almost all con-
struction characteristics.
The most important attributes of facilities other than buildings
seem to be unique to the type of facility and are not addressed herein.
For example, while underground pipelines can be treated in a parallel
manner to buildings in classification systems, the general headings
have very different meanings. The size of a pipeline would probably
mean the diameter of a pipe, while for a building it would mean the
square footage or height. The types of materials used for pipes are
also different from buildings.
OCR for page 133
133
DISCUSSION OF ESSENTIAL DATA
The following subsections discuss three aspects of the data re-
quired for essentially every study.
Construction CIasees
The most frequently used approach to developing an inven-
tory of building construction characteristics Is the construction class
method. Once the facilities are described in terms of their location
and construction class, and after construction classes are tied to
motion-damage-Ioss relationships, this overall vulnerability analysis
can be combined with the seismic hazard analysis to predict damage.
Table 3-1 presented an example of a typical construction class
system (see Chapter 33. Developed by the Insurance Services Office
(ISO), this scheme has been widely used for insurance as well as
noninsurance loss estimation purposes. Once the difficulties of prop-
erly counting buildings and assigning them to the appropriate class
are overcome, relationships between shaking intensity and resultant
damage are used to project damage (see Working Paper E).
The degree of approximation present in this approach Is typical
of earthquake loss estimation studies. It is very expensive to col-
lect precise data about construction characteristics, and these data
are not already tabulated in inventories prepared for nonseismic
purposes. Although this scheme may seem to categorize the building
stock rather coarsely, it is usually more than precise enough to match
the accuracy of the inventory work.
The extreme case of what might be called a detailed inventory is
the information an engineer collects concerning materials properties
and geometric data on each structural member and connection in a
building for the purposes of new design or an evaluation of a build-
ing's earthquake vulnerability. This "inventory" is then subjected
to detailed load and capacity calculations to design an adequately
strong and stiff structure or to see if the existing building is ade-
quately earthquake resistant. Even when an inventory of this detail
is collected for a single building, the estimation of earthquake damage
that would result from a specified earthquake is still an approxima-
tion. Thus, while it is true that the better the inventory the better
the accuracy of the resulting loss estimates, it is also true that even
with a perfect inventory there would still be a large amount of am
proclamation inherent in the process of estimating the losses that
might occur in future earthquakes.
OCR for page 134
134
A discussion of the extent of building stock inventory information
already available for earthquake loss estimation purposes will be
found In work conducted at Cornell University (Jones et al., 1986)
and by the Association of Bay Area Governments (ABAG) (Perkins
et al., 1986~. A rough estimate of the field work required in the
ABAG project to survey commercial or industrial areas is that about
five census tracts per day can be Windshields surveyed from a slowly
moving auto with a two-person team.
The study by Gauchat and Schodek (1984) is innovative for its
use of aerial photo analysis, although it restricted itself to housing be-
cause the construction characteristics of housing are easier to observe
in this way; commercial and industrial building construction charac-
teristics are more varied and less easily observed from the exterior.
In a study of Los Angeles County earthquake losses by Scawthorn
and Gates (1983), except for construction data on high-r~se and unre-
inforced masonry buildings, inferences were used to convert land-use
maps showing acreage of various uses into 13 construction classes
and into building areas. A committee of engineers, building officials,
and realtors was relied on for these extrapolations.
The NOAA-USGS studies also capitalized on existing files con-
cerning high-rise or other special categories of buildings, used census
data to inventory most of the housing, and relied on field sampling
of commercial-industrial areas coupled with land-use maps and local
engineering knowledge of typical construction patterns.
Occupancy
When life safety impacts of an earthquake are to be estimated, as
is almost always the case except for insurance or other property loss
studies, the number of occupants in buildings must be estimated.
Once the damage to a class of construction is estimated, the per-
centage of occupants or passersby who would be slightly or seriously
injured, or killed, is estimated. This allows for the number of persons
to be multiplied by this ratio to produce estimated casualties. An-
other approach used instead of or in combination with this method is
to apply a casualty ratio to the overall population of an urban area.
The number of people who would be outside of buildings must
be estimated because for some classes of construction, notably un-
reinforced brick buildings, the collapse of at least some brickwork
off the outside the building to the sidewalk or other exterior area is
more likely than complete collapse.
OCR for page 135
135
The time of day must be taken into account. In many areas of
the United States, people work, shop, and engage in other daytime
activities in buildings that are on average more hazardous than the
residences where they spend the night. E~tunating losses for different
tunes of day is typical of loss studies for this reason. Fortunately,
census data, planning department studies or economic data, and
reliable inferences relating the number of occupants to land-use or
building area data (Jones, et al., 1986) are usually available. This is
not as difficult an inventory task or as prone to error as the listing of
buildings according to construction classes.
Another aspect of occupancy or use that must be collected for
some studies is the type of occupancy or function of the building. For
estimating the ability of emergency response agencies to experience
an earthquake and yet be able to provide essential services, most
loss studies pay particular attention to hospitab. In terms of the
overall medical system in the area, the medical roles of other facili-
ties, including ambulance garages, wholesale pharmaceutical supply
locations, ~d blood banks, must also be properly inventoried. For
estunat~ng economic losses, an estunate of the economic activity
occurring in buildings must be made.
The designation of type of use for facilities with essential emer-
gency response functions (e.g., fire stations and hospitals) is almost
always easily available from government agencies or other sources.
Since these more essential facilities can be listed quickly, it is possi-
ble to segregate them and address their inventory and analysis tasks
differently. Detailed, facility-specific techniques are more costly, but
relatively few essential facilities exist (and in some cases only the
most essential among this small population need detailed attention).
The greater cost is also justifiable on the grounds that the vuInera-
bility analyses for these buildings should be more accurate because
these facilities are more important for emergency planning and to
some extent for hazard reduction purposes.
In California, for example, there are (in about 1985) 520 hospi-
tals, 433 essential communications facilities or emergency operating
centers, and 441 police or sheriff stations (Office of Emergency Ser-
vices, 1986~. There is a greater number of fire stations (3,155), but
many of these are small in size or significance and are generally one
of the easiest of the essential emergency function buildings to field
survey. ~ the set c safety study for the general plan of the cities
of El Cerrito, Richmond, and San Pablo in the San Francisco Bay
Area (Cities of El Cerrito. Richmond, and San Pablo, 1973), every
OCR for page 136
136
fire station in the three cities was enumerated according to address
and location on a seismic hazard map, and the type of framing of
walls and floor or roof was noted; this was a minor aspect of the
overall project and only a small effort was devoted to it.
These different types of inventory data that relate to the con-
struction class and the various occupancy-related information items
are not centrally collected by any agency or organization, and their
availability can vary from one local jurisdiction or region to the next.
Skiff inventory development is largely a matter of carefully ex-
tracting the useful but inexpensive data from pre-existing sources,
such as local planning or assessor's departments, or from field survey
work and then moving on to a completely different source to obtain
other information to fill gaps.
Facility Location
Typically seismic hazard maps of ground failure or ground shak-
ing are only available on a relatively coarse scale. Either census tracts
or zip codes often provide a more detailed scale than is required to
match the detail of the seismic hazard mapping. Where detailed geo-
logic maps showing the distribution of soft soil or high ground-water
areas are not available, and where the seismic sources are relatively
distant rather than located within the study area, facilities some-
tunes need not be located more accurately than by general district
of a city, or even by city, for the purposes of that particular study.
Because refinements in the geologic data base or changes in the
analysis of seismic sources may occur and because the inventory may
be useful for nonseismic purposes, it is always desirable to locate
facilities according to a scale at least as fine as zip codes or census
tracts, unless especialRy rapid and inexpensive studies are to be at-
tempted. Since Bureau of the Census data include an enumeration
of one- to four-family dwellings, dwelling lomes are generally esti-
mated from an inventory that is already conveniently broken down
into census tracts, block groups, and blocks.
Census tract boundaries are redrawn periodically by the Bureau
of the Census, and zip codes are also rearranged by the Post Office,
which create an updating problem. While not a major problem,
census tract, zip code, and political jurisdiction boundaries must
also be reconciled; a census tract, for example, may extend into more
than one municipality.
Disaggregating the inventory down to a small geographic level is
OCR for page 137
137
a goal sought by users, but they also face problems of confidential-
ity or controversy if specific facilities are identified. In the seismic
safety study for San Francisco's general plan (UEtS/Blume and As-
sociates, 1974), a building-specific inventory of the larger seismically
hazardous or suspicious buildings of the city and county was pro-
duced, based on a rapid technique using county assessors' data and a
walk-by of each major building by an experienced engineer. This de-
tailed and potentially very useful information—the detail that users
often request was also very controversial and never made public.
According to the engineer in charge of the study and the head of
the planning department, the information was withheld at the direc-
tion of the city government out of fear of lawsuits. The head of the
building department at that tune advised in a memo that publicizing
the list would do no good and would cause "panic, accusations, etc."
(Finefrock, 1980~.
On the other hand, failing to disclose information about hazards
may increase liability exposure, so this issue of the specificity of an
inventory should be considered with legal advice. It ~ also true
that earthquake hazard inventories required by state or local law,
as distinct from inventories compiled in loss estimation studies, have
withstood legal tests over more than a decade.
Another approach to defining location is to use an arbitrary
grid or rectangular cell system. The I-hectare cell (about 2.5 acres)
system used by ABAG in a recent earthquake loss inventory project
was found to be generally adequate. In Japan, a grid is often used
to map both seismic hazards and building inventories using similar
small-scale ceils.
Where local government assessors' files contain construction-
related or other useful information, the assessor's parcel can be used
as the basic mapping unit. Assessor's parcels conform to land owner-
ship patterns, which are usually much finer-scaled in urban areas than
zip codes or census tracts, or even census blocks. Census tract, zip
code, arbitrary grid, and assessor's parcel boundaries are unrelated
to each other, although with extra cost they can be cross-referenced.
Geographic information systems using digitized maps provide
several advantages once their initial cost is paid and funding for their
maintenance is assuredly. Changes in seismic hazard zones or contours
can be easily accommodated. Changes in the facility inventory, once
the new information is collected, can be included inexpensively in
new calculations of loss. In addition, the mathematical manipulation
of units within geographic areas (such as calculating the number
OCR for page 138
138
of dwellings located where the intensity is estimated at a certain
level) can be easily accommodated. A recent conference devoted
to geographic information systems indicates the range of possible
natural hazard as well as other applications (American Society for
Photogrammetry and Remote Sensing, 1987~.
Another great advantage of computerized approaches is in deal-
ing with problems where various combinations of layers of informa-
tion on the map must be compared. A study of regional southern
California earthquake response issues (Haney, in progress) is digitiz-
ing pre-existing information, some of which is related to lifelines, from
a California Division of Mines and Geology (CDMG) report (Davis
et al., 1982a). Broadcast coverage areas for Emergency Broadcast
System stations can be compared with the CDMG study's projection
of intensities and with the languages of residents as determined from
census data, for example. No files on building structures are being
added to the data base, although there are plans to use the AT~13
method in its present form for this purpose.
Two disadvantages of computerized systems are the initial costs
of establishing the system and the costs of maintaining the system.
The first-year cost of establishing a Regional Information Manage-
ment System in southern California using the earthquake loss esti-
mation method applied to a pilot project area in San Bernardino
County was estunated at about $1 million (Schulz et al., 1983),
although other nonseism~c benefits were postulated.
The work in southern California that is jointly funded by FEMA
and the state of California (Haney, in progress) and three recent
projects illustrate this evolving approach: digitizing of several dif-
ferent types of seismic hazard and facility data for Sugar House
quadrangle in Utah by the USGS Rocky Mountain Mapping Center
(Alexander, 1987~; digitizing of seismic hazard maps for San Mateo
County, California (Brabb, 1985~; and digitizing of a small study area
in San Bernardino County, California (Schulz et al., 1983~. None of
these projects deal very specifically with the problem of enumerating
buildings in terms of construction characteristics, which is by far the
single biggest inventory problem in the earthquake loss estimation
field. This is not what computerized approaches do best. Manipula-
tion of already collected information, rather than data collection, is
the strong point of the computer-aided inventory approach.
Portions of the USGS map system for the United States, the
familiar topography maps produced at scales as fine as 1:24,000, are
now digitized and the remainder of the USGS maps will eventually
OCR for page 139
139
be converted to this format, allowing for various types of digitized
data to be related directly without having to convert via paper maps.
The U.S. Bureau of the Census will digitize the results of the 1990
census (Marx, 1986~; future earthquake loss studies that tie into
a geocoded information system may benefit more than at present.
Many local organizations, such as utility companies, planning depart-
ments, emergency services departments, and others are investigating
the potential of combining resources to produce multipurpose maps.
SUGGESTED SOURC1:S OF INVENTORY INFORMATION
Guidelines are suggested here for preparing rapidly an inventory
of facilities when the preferred ideal inventory cannot be done for
an earthquake loss estimation study. A number of ways have been
used or proposed. These have typically been uniquely tailored for
a particular type of loss study in a particular area. The techniques
suggested or followed in preparing such inventories have been shaped
not only by the kinds of data needed for the particular study, but
also by the kinds of information readily available in the particular
area. An additional element of expert judgment from persons familiar
with the study locality has been an important part of these inventory
techniques, because it typically has been necessary to infer needed
end-form data from other types of information.
Inventories that are less than the ideal type have advantages
as well as limitations that must be recognized In the beginning.
Foremost among the advantages are: in general, they are less costly
to prepare, and they typically can be completed in less tune than an
ideal inventory would take.
Foremost among the disadvantages are: more sophisticated ex-
pert knowledge must be employed in extrapolating essential data
from available raw data, and they are less accurate than more de-
ta~led inventories and these inaccuracies carry over to the loss esti-
mates. Poor-quality input information leads to poor output results.
Rarely have earthquake loss estimation studies quantified their un-
certainties, so a study with less accurate inventory, and thus less
accurate loss estimates, may appear to be as valid as a more accurate
study, but this is not the case.
Owing to the diverse types and forms of readily available data
about facilities in a study area, a ste~by-step procedure cannot be
suggested for preparing an inventory, nor can it be suggested that
OCR for page 140
140
one source of data is better than another. The process to be followed
depends on several factors, among them:
Financial resources available for the study;
~ Type of loss study, which establishes the type of end-data
needed for the inventory and which relates to the geographic scale,
kinds of facilities and losses, and time frame as discussed earlier; and
~ Kinds of existing data (i.e., what kinds of data have been com-
piled on, for example, schools, dwellings, publicly owned buildings,
and high-rise buildings).
From earthquake loss estimate studies prepared by others and
from examination of basic elements of loss estimation methods, some
general guidelines for an inventory procedure can be inferred. First,
the end-form of the inventory data for the particular loss study must
be established, which in most cases consists of:
~ Numbers of facilities of various types that are located in
specified zones (e.g., blocks and census tracte), in short, a count of
facilities.
Classification of facilities according to the classes in the
motion-damage relationships to be used in the analysis phase.
Value of facilities, normalized to some base year.
. Occupancy information, since casualty loss estimates are in-
cluded in many studies.
E unction or use classification, if economic sector loss estimates
are to be prepared and if essential emergency response facilities are
to be identified.
Second, the inventory must be built at least partly from ex-
isting data sources. Inventories created from field observation are
much more costly than those based on reuse of existing data. More-
over, some of the end-form data can be extrapolated with reasonable
accuracy from existing data sources, especially when someone knowI-
edgeable about the study area is utilized. The degree of extrapolation
that is acceptable is a significant msue In this regard and relates to
the required overall accuracy of the result from the user's viewpoint.
Following is a brief list, with some discussion of the existing data
sources most often used for preparing earthquake loss estimation
inventories.
1. For housing:
~ U.S. census information. These data, giving dwelling
unit counts, occupancy numbers, and relatively precise locations, are
OCR for page 148
148
.
dwellings and for population distribution, the best source of data was
found to be the United States Census data. The Census provider
information of the numbers and geographical distribution of dwellings
according to census tract. Census tracts are a convenient unit since
the number of one to four family dwellings in each tract seldom exceeds
2000 units in the Salt Lake area. The most accurate cost estimates
for housing were obtained from boards of realtors or realtor associa-
tions which compile frequent (usually monthly) summaries of actual
dwelling sales. Aerial photos and appropriate sampling techniques
were used to develop the construction characteristics of dwellings
since there is a great difference in vulnerability between wood frame
and other types of housing construction. Studies (Steinbrugge and
others, 1969) have shown that the number of brick, concrete block
and related types of construction used for dwellings in, for example,
California is small (less than a few percent). It was found that brick,
concrete block and related construction types made up about 60 per-
cent of dwellings in the Salt Lake City area. A detailed inventory
of buildings by classes of construction other than dwellings was un-
dertaken by the H.C. Hugh Company of Salt Lake City for the U.S.
Geological Survey. The development of the inventory was supervised
by K.V. Steinbrugge. Air photos and drive-by inspection of buildings
in-each census tract were conducted. Construction type was noted
and the dimensions of the buildings were obtained either from the air
photos or from actual measurements. Replacement cost per unit area
for the various classes of construction was estimated by a professional
building inspector in Salt Lake City with long experience in the re-
gion. It is believed that the inventory obtained in Salt Lake City is
extremely accurate for the purposes of an earthquake loss study and
that the errors in the estimation of ground motion are likely to be
much larger than the inventory errors in this particular study.
In contrast, the inventory method for the San Etrancmco Bay
Area was based on building information extrapolated from census
data (dwellings) and modified fire insurance property values (other
buildings). AIgermissen and Ste~nbrugge (1984) give the following
description of the inventory method in this case.
Data on dwellings was obtained in the same manner as described in
the Salt Lake City study i.e., from census data and summaries of
real estate transactions. For buildings other than dwellings a novel ap-
proach was used. Quoting from Steinbrugge and others (1981~: `'The
initial data were fire insurance property values by county for north-
ern California and an assumed 8.3 magnitude earthquake on the San
Andreas fault. These values included dwellings, commercial buildings,
manufacturing plants, warehouses, offices, and all other fire-insured
properties. These property values were increased to include non-
insured private property as well as increased to include under-insured
property. Adjustments were made on a judgement basis to include
the value of Federal, State of California, and local governments-owned
buildings. Intensities from the NOAA report's isoseismal maps were
OCR for page 149
149
converted into loss {actors, or the percent loss based on an im-
personal definition basis. These percentages were multiplied by the
property values to obtain the total impersonal 1088 by county in the
study area, then summed to obtain the total aggregate loss. In this
process, values were adjusted to compensate for inflation to 1980.
Building contents for the aforementioned San Andreas earthquake
were analyzed in a similar manner to derive the total contents aggre-
gate loss.
A strong point of this NOAA-USGS inventory approach is its
balancing of accuracy versus detail pushing the available data as
far as appropriate and then stopping short of making further assump-
tions that would be necessary to obtain more detailed estimates. The
expertise used in these studies appears to be appropriate to the task:
While earthquake engineering experts were employed, the expertise
of real estate, building inspection, insurance, or other local sources of
knowledge concerning the distribution of classes of construction was
also utilized. A weak point in the method is that complete documen-
tation of the technique—complete enough for others to replicate or
test the technique in an updating study of the same area or to apply
it elsewhere—is lacking. Since the experience of a few key individuals
has been heavily relied on in these studies, documentation may be
inherently difficult, and to some extent it would be more a matter
of teaching an art rather than specifying the precise steps that could
be mechanically followed.
The 1?EMA/ATC-13 Inventory Method
The method for estunat~ng losses from earthquakes described in
the AT~13 report (Applied Technology Council, 1985) was designed
to provide Formation on damage, casualties, and immediate func-
tional loss to be combined with an economic model for predicting
economic losses, that is, direct building and structure lopes, loss
of equipment, production losses, losses to infrastructures such as
utilities and transportation systems, and losses due to interrupted
business. To serve its original intended purpose, the inventory and
loss estimates had to be compatible with the economic sectors to
be used in the interindustry input-output model. Accordingly, this
method ~ comprehensive in the inventory it seeks. Forty classes of
building construction and 38 nonbudding structure ciames are de-
fined, and each facility must also be defined In terms of one of 35
occupancies or asocial functions.
OCR for page 150
150
The broader nature of the ATC-13 inventory makes it only par-
tially comparable with the NOAA-USGS method, and only the por-
tion of the ATC-13 inventory method that dead with buildings is
discussed here. It should be noted that the breadth of the AT~13
method which encompasses lifelines, industrial structures, ground
failures, and functional losses in a quantitative manner ~ one of its
significant accomplishments.
The FEMA/ATC-13 inventory method aims at compiling loca-
tions and quantitative measures for ad facilities plus descriptors of
the construction that allow classification for use In estimating dam-
age. Facility values are also needed, as is information about each
facility's economic use for input into an economic mode} that begins
with damage and reduction In functional levels and then forecasts
longer-term econorn~c impacts. The portion of the loss study that
inventories the information and analyzes it to produce estimates of
immediate losses is called FEDLOSS by FEMA in its automated
form. The portion of the loss study that would employ an economic
input-output mode} to est~ate longer-term econorn~c losses ~ called
FElMS (FEMA Earthquake hnpacts Modeling System).
The AT~13 report states its preferred source of inventory data
as pre-existing inventories of facilities containing the required con-
struction class detail, but because even less demanding classification
systems cannot be supported by data that have already been col-
lected, this preference will in most cases be unfulfilled. This hoped
for pre-ex~sting inventory is called a Level ~ inventory. A Level 2
inventory, the one necessary in most cases, wiD be described below.
A Level 3 inventory ~ sunply a complete synthesis of an inventory
based only on overall population data, such as by a~um~ng both the
number and construction types of all buildings in a city on the basis
of its population.
In the Level 2 approach, the location and descriptors of con-
struction are obtained by extrapolation from a variety of economic
and census data. The sources for these data are discussed in ATC-13,
and are described in detail in the FEMA Data Base Catalog (FEMA,
1985b), which lists the many different computer data files acquired
by FEMA from other agencies, through marketing or economic anal-
ysis services, or in some cases from within the FEMA organization.
These data bases have been accumulated and funded primarily as a
function of the civil defense program of FEMA and its predecessor
agencies and have been used in nuclear war loss estimations. Corre-
lations between facility and use classification were developed in the
OCR for page 151
151
ATC-13 project to allow for the transformation of economic data into
construction data. The relationships imputed In the ATC-13 study
were developed only in the context of California.
~ some ways, facility classifications of the ATC-13 method are
similar to those of the USGS method, but are more detailed. The
ATC-13 method has almost two tones as many classes of construc-
tion as the NOAA-USGS method (40 versus 2l, comparing building
classes only), and each individual facility must also be amigned one
of 35 use categories. There are, however, some buildings whose
construction would be more precisely defined by the NOAA-USGS
inventory (or ISO) scheme, such as a steel moment-resisting (rigid
frame, or rigidly connected joints) building with flexible diaphragms
(or floors acting to resist lateral forces).
Clearly more information is required to construct an inventory
for the 40 ATC-13 classes of facilities than for 21 classes. Given any
comparable inventory budget, the accuracy of the assignment of a
facility to its proper class in the ATC-13 method would usually be
less than in the NOAA-USGS method.
The greatest advantage of the ATC-13 method ~ that it is very
powerful: it can assemble a very large and detailed inventory inex-
pensively by using already computerized socioeconomic data. This
is also its biggest disadvantage compared to methods that use ac-
tual inventory data obtained from or checked by fieldwork with less
extrapolation.
The large amount of extrapolation and reliance on rules of thumb
developed by combining the opinions of earthquake engineering ex-
perts can be seen from a typical example of how the inventory method
would operate. First, the ATC-13 method would probably start with
the number of employees who work at a commercial or industrial
business. (For some small number of industrial facilities, the FEMA
data bases may contain construction data and thus make the Level
2 extrapolations unnecessary. The number of ones to four-family
dwellings can be obtained directly from census data.) One of a few
data bases, such as the Census Bureau's Manufacturing Establish-
ments by Industry Sequence, would be used in which the known
information (excluding economic data on value of goods produced,
for example) is simply number of employees and the location by zip
code, along with the detailed (four-digit) Standard Industrial CIas-
sification (SIC) code that defines the type of economic activity. The
precision of the location is sometimes but not usually an issue. For
OCR for page 152
152
example, the zip code location listed for a supermarket company in
a city wall lump Al of the employees at the headquarters' zip code.
These data number of employees, type of economic activity,
and location- are the only data known directly for the facility in
most cases, and the remainder of the necessary data ~ synthetic. As
ATC-13 notes,
The FEMA Manufacturing Establishment File, the Wholesale Trade
Establishment File, and other business establishment/company files
do not include either the size, location, or structural characteristics of
facilities. This information must be estimated based on economic data
such as the number of employees or annual production amounts. . .
. Few if any existing facility databases or the inventories synthesized
using Level 2 and 3 procedures contain sufficient information to
allow the accurate determination of Earthquake Engineering Facility
Classifications.
The second step in the AT~13 inventory method Is to relate
the number of employees to the building size, according to estimat-
ing factors for different occupancies. These relationships are gener-
ally drawn from transportation studies, especially those of Caltrans
(California's highway department). In the ABAG inventory method
(Perkins et al., 1986), similar relationships were used to estimate
building square footage, using instead Federal Highway Administra-
tion data. This extrapolation is more accurate than those of the
other steps and is not a major source of error. As noted in the
work of Jones et al. (1986), stable and reliable relationships exist
for square footage per person estimating factors, although a curious
effect of this relationship ~ that an inventory would show buildings
swelling and shrinking in size as fluctuations in the economy cause
the number of employees in a building to rise or fan.
The third step ~ to divide up the buildings, known at this
point only in terrors of location and, by extrapolation from number
of employees, the size, into construction classes. The height of the
approximately 3,000 high rises in California can be known from files
specific to high rises assembled by the Council on Tall Buildings of
Lehigh University. For the majority of buildings that remain, they
can be divided into mid-rise and low-rise categories based on rules
of thumb developed by a process of asking earthquake engineering
experts their opinions.
In this third step of developing a synthetic construction class
distribution, the other basic task is to assign a construction class
(e.g., reinforced masonry shear wall with moment-resisting frame,
reinforced masonry shear wall without moment-resisting frame) to
OCR for page 153
153
each facility. This was done by obtaining collective expert opin-
ion from the engineers involved in the ATC-13 project, assigning a
certain percentage of the buildings in each use category to each of
the construction classes. In each use category (e.g., s~gle-family
dwelling) the fractions for low-r~se wood frame, low-rme reinforced or
unreinforced mansonry, and so on sum to 100 percent.
The result ~ the end-form data: construction class (and high-
, mid-, or low-r~se subclass designation), floor areas, use, and zip
codes for ad buildings in the study area. Steps one and two involve
relatively noncontroversial extrapolations common to many loss esti-
mation methods. It ~ the third step, where the inventory variable of
central importance construction class is synthesized on the basis
of opinion, that involves untested relationships. Essentially, the con-
struction class inventory ~ synthesized knowing only the number of
employees, the zip code of the business, and the economic function.
Comparison of the NOAA-USGS and FEMA/ATC-13
Inventory Methods
A full application of the ATC-13 method has not yet been re-
ported in the literature. The NOAA-USGS method is a general
method that can be extracted from the reports of its application, for
example, the large-scale NOAA-USGS loss study of San Franc~sco.
ATC-13 is a report that describes its method very specifically, but
there ~ no loss estimation study or actual application to refer to
as a concrete case. This makes a comparison of the two inventory
methods difficult. Also, the two methods were devised for different
purposes.
Although the comparative information given in Table D-1 on
the type of end-form data implies that inventories would be much
the same for both methods, this is not precisely true. Somewhat
different characteristics are used for classifying the facilities In the
two methods, and this affects the details for each. However, the
striking difference in the two methods ~ not the data they seek,
but how they assemble them. While both methods use judgment
in the inventory process, the ATC-13 method is more reliant on
judgment. The application of judgment in the ATC-13 method is,
however, generally more apparent, In that it would be easier for other
investigators to rely on the published description of the method, reuse
it, and replicate the results obtained by others.
Descriptors used in the classification process for each method
OCR for page 154
154
TAl3LE D-1 Comparison of ATC-13 and NOAA-USGS Inventory Data
FEMA/ATC-13
NOAA-USGS
Seventeen basic building construction
classes with subclasses for low,
mid-, or high-rise heights, and
combinations of systems (such as
shear wall and frame); 40 classes
total
Primary categories
Wood frame
Light metal
Unreinforced masonry
Reinforced-concrete shear wall
Reinforced-masonry shear wall
Braced steel frame
Moment-resisting steel frame
Moment-resisting concrete frame
Precast concrete
Long span
Tilt-up
Mobile home
Descriptorsa
Structural material
Framing system
Floor area
Height
Ductility
Economic use, social function
Thirty-five classes that are cross-
referenced to the broader range
of SIC classes; each facility
inventoried is assigned a class
Nine basic building construction
classes, with subclasses for
size and degree of earthquake-
resistant design; 21 classes
total
Primary categories
Wood frame
Light metal
Unreinforced masonry
Reinforced-concrete shear wall
Reinforced-masonry shear wall
Steel frame
Concrete frame
Precast concrete
Tilt-up
Descriptorea
Structural material
Framing system
Floor area
Height
Earthquake-resistant design
Economic use, social function
Collected for some essential
facilities (e g., hospitale)
but not collected for each
building
aAll of these descriptors are not necessarily inventoried for all
classes.
vary in detail in some cases but would be identical for the two meth-
ods in other cases, for example, construction material or height. The
lists shown in Table D-1 are in a different form than they appear
in either method and are organized more generically to allow for
comparisons. For example, the NOAA-USGS approach contains a
class for mixed construction (different was and diaphragm mate-
rial) that includes buildings with wood roof and floors with walls
OCR for page 155
155
of tilt-up, reinforced masonry (brick or block) or poured-in-place,
reinforced-concrete construction. Variations in earthquake-resistant
quality ratings can result in these buildings then being assigned to
different classes (Insurance Services Office, 1977~. In Table D-l, these
variations on the mixed NOAA-USGS class of construction are listed
as separate classes to allow for closer comparison with ATC-13.
Inventories for the two methods contain much of the same type
of information, although the broader purpose of the AT~13 method
(economic loss estunation) leads it to develop two additional detailed
sets of information, one on economic function and the other on
lifelines and nonbuilding structures.
The PEPPER Study In~rento~ Method
The method for estunating earthquake lodes used ~ the PEP-
PER (Pre-Earthquake Planning for Post-Earthquake Rebuilding)
study (Spangle, 1984) relied on automated data already collected by
the planning department of the City of Los Angeles. No new field
surveys were conducted, partly because of budget limitations and
partly to try to test the usefuIne" of this large data base, which
had been assembled from assessors' tax records and other sources.
As partial checks on the accuracy of this comprehensive data base
of about ~ million buildings, files containing information specific to
building construction characteristics were consulted. An accurate
inventory of pre-1934 (preseism~c code) unreinforced masonry build-
ings was already in existence because of the city's retroactive seismic
ordinance, and the characteristics of high-rme buildings were tabu-
lated in a real estate survey. Census data on population and housing
from the 1980 census were used, along with a 1974 city study.
Buildings were (~) aggregated in planning areas of the city, and
(2) classified according to type of construction ~ five classes: steel,
concrete, masonry, wood, and special. Use was classified according
to four classes: residential (with three subcIames), commercial, in-
dustrial, and other. No other details appear in the report to suggest
the way in which buildings were allocated to each class. As noted
in the study's engineering report, The inventory of structures . . .
is probably the least reliable component of the various factors that
determine the damage pattern" (Degenkolb, 1984~.
The building classification method might be described as an
adjusted NOAA-USGS method. The PEPPER method adjusted
the basic ISO or NOAA-USGS construction classification system
OCR for page 156
156
because the available data were not that finely subdivided. This
also had implications for the analysis task, because the hybrid or
combined construction classes of the PEPPER inventory had to be
analyzed using hybrid motion-damage relationships. The beginning
form of the data In the city planning department's data base did not
differentiate high rises according to their type of enclosure system
(e.g., curtain wall, poured-in-place concrete). Inferences based on
year of construction (e.g., assuming that post-1960 high rises were
predorn~nantly of curtain wall exterior) were used.
One point made by this study is that even if a very large com-
puterized file of buildings exists, this does not necessarily mean that
the data are detailed or accurate. Lack of detail is evident from the
fact that all steel buildings, or aD concrete buildings, for example,
were lumped together In one class. This level of detail is a common
constraint in the use of assessors' or local planning department data.
The accuracy of the inventory was also limited and was related to the
fact that this data base was assembled for nonseism~c, nonengineer-
ing purposes. An example of a major type of inaccuracy concealed in
the data bee was that high-rise buildings were sometunes described
as having wood-frame structures. Another problem was that this
data base was not current because the cost of updating it had been
considered too high by the planning department a few years after it
had been created.
POSTEARTlIQUA1lE STUDIES OF LOSS
Related to the pre-earthquake inventory problem ~ the task of
postearthquake inventory of damage by class of construction, loca-
tion, ground conditions, and intensity or measured ground motion.
Although all loss estunation investigators bemoan the fact that there
are not more historical loss data available, there are few ongoing ef-
forts outside of the insurance industry to collect this type of data after
earthquakes occur. As pointed out in the Earthquake Engineering
Research Institute's guide to postearthquake investigation (Earth-
quake Engineering Research Institute, 1977), und~naged as well as
damaged buildings should be tabulated. Statistical techniques pro-
vide many tools for analyzing damage data, and these are explained
in the guide in a special section. However, most earthquake recon-
naissance reports or detailed studies do not comprehensively report
damage or loss data, but rather concentrate on the more unique or
instructive individual cases of damage.
OCR for page 157
157
Because the types of pre-earthquake inventory data and con-
struction classes that are generally used are known prior to initiating
postearthquake investigations, damage data could be collected effi-
ciently, on a sampling basis where necessary, to try to fill gaps in
historical loss data. Although in theory systematic studies of build-
ing damage could result in complete data for estimating purposes, in
practice this is not so. Construction innovation will always be ahead
of recorded earthquake experience. Earthquakes in Chile and Mexico
in 1985 tested the building construction methods in use in these coun-
tries of the 1950s, 1960s, and 1970s. There are no data, however,
on the performance, under moderate to severe ground motion, of
tall welded perimeter tube structures, modern m~d-rme steel-braced
frame structures, or large t~vo-story tilt-up concrete structures that
are common in many parts of the United States.
SUMMARY
In theory, a perfect inventory can be created. However, it will
never be achieved because of cost and tune constraints. Therefore,
ways of obtaining the most useful, imperfect inventory are being
studied. The attempt to start from an econorn~cally based inven-
tory, as in ATOLL, is not advised. Although the final output is
intended to be economic, economic loss can only be estunated on the
basis of an estimate of earthquake damage. Earthquake damage can
only be estimated accurately when building construction data are
directly sought. Converting economic data into construction cIassifi-
cation data ~ not recommended because this can greatly reduce the
accuracy of the inventory.
If the focus is to be on building damage, then the inventory
should focus on vulnerable or ~seisrn~ca]ly suspicious" buildings. Pro-
cedures that provide an initial screening by low-cost means, leading
to a more detailed survey to provide accuracy, make more sense
than an attempt to develop a complete inventory from which the
hazardous buildings must be selected.
Facilities with a potential for large loss, or with essential emer-
gency functions, should be inventoried on a case-by-case, field survey
basis.
The insurance industry, particularly in California, has much in-
formation both on building damage and on building inventory. This
information is generally unobtainable due to industry's confiden-
tiality requirements and competitiveness, although the California
OCR for page 158
158
Department of insurance obtain this Information, aggregated by
geographic zone and clam of construction, on an annual basis. 0~
tanning some of this information would benefit national or regional
interests, solve some of the data problems of earthquake damage
estimating, and yet-preserve such proprietary information as the
industry deems necessary.
Representative terms from entire chapter:
loss estimation
