Click for next page ( 27


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



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 26
4 Building Damage and Losses CLASSIFICATION OF BUILDINGS For loss estimation purposes, the buildings within a region are put into a number of categories according to a construction classifi- cation system. This is the starting point in the vulnerability analysis process, as shown in Figure 4-1. The primary consideration in developing a classification scheme is differences in the resistance of various buildings to damage during ground shaking. Some of the factors taken into account are the type of structural system, the materials of construction, the size of the building, and the degree to which structural features limiting damage have been provided during design and construction. The age of a building is sometimes used as an indirect indicator of seismic design level in areas where seismic codes have been adopted, and it can indicate typical construction practice in a given region. In the planning stages for a study, the steps of selecting a clas- sification system, developing methods to prepare the inventory, and assembling motion-damage information are all interdependent. That is, the choice of a classification system depends on the availability of information for the inventory and the effort that can be put into carrying out the inventory. The availability of data relating motion and damage for various kinds of construction is also limited, and this similarly restricts the classification options. 26

OCR for page 26
27 This 14-story reinforced-concrete apartment building experienced extensive damage to the spandrel beams during the 1964 Great Alaska earthquake (M 8.3-8.6~. Its twin in another location in Anchorage was similarly damaged. Structures that have adequate strength to resist moderate shaking may not be able to withstand strong ground shaking. Photo courtesy of G. Hou~r~cr.

OCR for page 26
28 L Facility Classification System / Inventory _ \ \ Vulnerability ~ 1 1 - \ Motion-Damage-Loss | - FIGURE 4-1 Structure of the vulnerability analysis portion of an earthquake loss estimate study for buildings, lifelines, facilities with essential emergency roles, and facilities with potentional for large loss. The most commonly used classification system in the United States for estimation of earthquake Toss is that developed by AIger- missen and Steinbrugge (1984~. As shown in Table 4-1, this scheme has 21 categories, determined primarily by the type of information readily available to property insurance companies. A more recent classification system used in the ATC-13 study (Applied Technology Council, 1985) has over 40 categories, with height emphasized as a factor. Both of these systems have been heavily dependent on the work of experts in California. For loss studies elsewhere in the United States, these basic schemes should be reviewed and possibly modified and simplified to take into account local construction variations and problems of assembling an adequate inventory. For example, in the study of six cities in the Midwestern United States (Allen and Hoshall et al., 1985), only eight building construction categories were used. INVENTORY Preparation of the inventory is usually the most time-consuming and costly aspect of a loss study. It is also often the most frustrating,

OCR for page 26
29 TABLE 4-1 Construction Classes Used in the ISO and NOAA/USGS Methods Building Class Brief Description of Building Subclasses 1B 2A 2B 3A 3B 3C 3D 4A 4B 4C lA-1 Wood-frame and stuccoed frame dwellings regardless of area and height 1A-2 Wood-frame and stuccoed frame buildings, other than dwellings not exceeding three stories in height or 3,000 square feet in ground floor area Wood-frame and stuccoed frame structures not exceeding three stories in height regardless of area Wood-frame and stuccoed frame buildings not qualifying under class 1A One-story, all metal; floor area less than 20,000 square feet All metal buildings not under 2A Steel frame, superior damage control features Steel frame, ordinary damage control features Steel frame, intermediate damage control features (between 3A and 3B) Steel frame, floors and roofs not concrete Reinforced concrete, superior damage control features Reinforced concrete, ordinary damage control features Reinforced concrete, intermediate damage control features (between 4A and 4B) Reinforced concrete, precast reinforced concrete, lift elate Reinforced concrete, floors and roofs not concrete Mixed construction, small buildings and dwellings Mixed construction, superior damage control features Mixed construction, ordinary damage control features Mixed construction, intermediate damage control features Mixed construction, unreinforced masonry Buildings specifically designed to be earthquake resistant 4D 4E 5A 5B 5C ED 5E 6 SOURCE: Algermissen and Steinbrugge (1984~. since in principle it is possible to develop a perfect inventory, but in practice compromises must always be made. Time and budget constraints lead to shortcuts and extrapolations, but evaluation of building seismic performance necessarily involves the use of reliable building data not obtainable by shortcut methods. Facility inventories can be maintained and later used both for updating initial loss estimates and in determining follow-up Toss estimates for facilities or geographic areas or for other purposes within a study region. Therefore, the pane] is persuaded that it is wiser in the long run to compile systematically an inventory that is as accurate as possible under the circumstances and resources available. Guidelines for this approach to the inventory are suggested in Working Paper D.

OCR for page 26
30 There are three interrelates] factors to consider at the outset of a project: the content of the inventory, the process of assembling the information, and the manner in which the data are to be recorded or stored. Content of the inventory What information concerning buildings is required? The basic minimum data are: Geographic location; Category of seismic resistance; Economic value of the building; Number of occupants, at different times of day; ancI Type of occupancy of the building (e.g., housing, commercial, or essential facility). Seisimic resistance must be derived from information on such characteristics as construction class, age, height, and so on. The meaning of economic value may differ according to the purpose of the Toss study, as discussed below. Other information, such as the function of the building (e.g., office or light manufacturing), may also be desired. A key problem is the degree of disaggregation or aggregation of this information. At one extreme, the inventory may list only the total economic value and total number of occupants aggregated for all buildings in a given construction class within some geographical area. At the other extreme each building might be listed separately and then aggregated for purposes of predicting Tosses. Obviously this question is strongly related to how the inventory is to be compiled and how the information is to be recorded. Another key question is the smallest geographical area to be used. As discussed in the section on user needs, it should be possible to disaggregate losses to any local political unit, which in the case of a large city may mean wards, precincts, or districts. Census tracts or postal zip codes also are convenient minimum geographical units, but if used they may require localized modifications to make the tract or zip code data correspond to other boundary lines. There are a number of possible definitions for economic value, and the choice depends primarily on the purpose of the loss estima- tion study. Cash value and replacement cost have both been used. For most studies, it seems appropriate to use replacement cost.

OCR for page 26
31 Carrying Out the Inventory The inventory process is a matter of assembling and using avail- able sources of information, carrying out some amount of onsite inspection, and applying some judgment. Census data are valuable, particularly for housing, and generally some local records are avail- able from, for example, planning departments and assessors' offices. The most difficult information to pin down is the seismic resistance or construction class. Here is where the experience of local engineers, building officials, and architects, combined with judgment, have to play a major role. Field sampling is also useful to define typical local construction patterns. It might seem ideal to develop a listing of all individual build- ings, but this seldom is feasible. While some data files, such as those maintained by assessors, are typically compiled for individual properties, they are unlikely to contain adequate information for as- signing seismic resistance. Moreover, for loss estimation purposes it is quite satisfactory to have crude data for the more seismically resistant buildings. Attention should be concentrated on developing a reasonably good inventory of the seismically suspicious buildings of high vulnerability that will incur the bulk of the serious damage (Arnold and Eisner, 1984~. Onsite surveys to identify and enumerate these buildings are vital to a satisfactory loss estimate. One example of a seismucally suspicious construction class is unreinforced masonry, which is often concentrated in recognizable districts. ATC-13 describes three methods for assembling an inventory, ranging from situations where detailed information is available in lo- cal files to cases where very few data are available. For the common latter situations, a method for abstracting an inventory from socioe- conom~c data is described. The panel feels that extensive field studies would be necessary to validate this approach, and that the varieties of situations to be encountered make success unlikely. The pane! believes that corresponding sums of money spent on direct observa- tion of buildings to discern specific seismic performance indicators would yield more useful results. There appears to be only a weak correlation between socioeconomic characteristics, such as number Of employees and the Standard Industrial Classification number in- dicating economic sector, and construction characteristics relevant to earthquake loss estunation. While a convenient data file, such socioeconomic information is not particularly relevant to the task of producing an inventory of facilities according to construction classes.

OCR for page 26
32 Recording the Inventory There are several reasons for collecting the inventory data in a format consistent with computerization. At a m~nunum the data should be stored in such a way that losses from several different earthquakes can be evaluated. It is desirable that data be retained so that updated loss estimates can be made In the future. Finally, in- formation in an inventory is potentially valuable for entirely different purposes, such as economic development planning and city planning. It is vital to include meetings with various potential users of inventory information at the beginning of a loss estimate study. Such discussions will indicate how much effort is justified in obtaining and formatting the inventory so that it can be accessed and used by various governmental agencies. A key question is whether there is the will and the means to maintain the inventory in an updated condition. Where a significant long-term effort appears warranted, use can be made of some impressive digital mapping technology well along in its development by USGS and others (Alexander, 1987; Brabb, 1985; Schulz et al., 1983~. Role for a National Data Base Creation and maintenance of a complete nationwide data base on the construction characteristics of all buildings is an unpracti- cal idea. However, some incremental, less geographically complete projects, or efforts limited to simplified construction classifications, may be feasible and desirable and should be investigated. Modest improvements in the compilation of ciata might include: . Comparing classification schemes so that future Toss studies collect and organize their data in a format similar to either the ATC- 13 or NOAA-USGS construction classes, or to some new scheme. . Suggesting data that could be reliably collected at virtually no additional cost by the U.S. Bureau of the Census. Noting the height of a building (e.g., placing it in one of three or four ranges of height in terms of numbers of stories) may be such a possibility. ~ Investigating the potential of using the FEMA Multihazard Vulnerability Survey method (FEMA, 1985) in connection with large- scale earthquake loss estimation rather than for the field survey of individual essential emergency operation facilities and life support systems, which was the initial purpose for devising this multihazard survey method. Field sampling of buildings previously surveyed by

OCR for page 26
33 this method and easy access by earthquake loss estimators to Mul- tihazard Survey data computerized by FEMA, are promising ideas. The applicability of the data collection and analysis components of the FEMA Multihazard Vulnerability Survey method (which in- cludes wind and flood hazards in its scope as well as earthquakes, depending on the site's location) should be evaluated in the context of loss estimation. MOTION-DAMAGE RELATIONSHIPS Identifying the relationship between the intensity of ground shak- ing and the damage experienced by a group of generally similar struc- tures, or a construction class, is essential to vulnerability analyses. One intensity/da~nage relationship is needed for each type of facility in the classification system. There are several ways in which this relationship may be ex- pressed and evaluated. Additional discussion appears in Working Paper E. Use of Mean Values The most common method for presenting the relationship be- t~veen ground shaking and darnage is by a loss ratio curve. Typical curves, developed some years ago by Steinbrugge et al. (1984) for the Insurance Services Office (ISO), are shown in Figure 4-2. The curves truncate at MM! OX because of the interpretation by ISO of the MMI scale: intensities above OX were taken to represent ground failures rather than ground shaking. (The classes of construction are those in Table 4-1.) Percent loss, also called mean damage ratio or mean damage factor, is the cost of damage expressed as a percentage of replacement value. This is a mean value for a large population of buildings of a given class. Relationships of this form are particularly useful when only the expected value of the dollar cost of damage is evaluated in a loss study. Irlfo~mation About Distribution of Damage For some purposes, knowing only the mean level of damage is inadequate. For example, serious casualties and injuries are usually related to extreme damage experienced by a minority of buildings.

OCR for page 26
34 Splh-level houses that were deAclent in earthquake resistance collapsed durlog the 1971 San Fernando, C~ll~rnla earthquake (~ 6.6~. ~lLbulk houses in the area supplied, experlenclug only cracks in plater. Compton Boulevard between Alameda and Second streets ~llowlng the March 10' 1933, Long Be=h e~hqu~ke (~ 6.2]. So many ages coped that the street w~ completely blocked ~ bricks. Tbe poor pe~rm~nce of these buildings led to chants in the bulldlog code It problbRed the construction of unreln~rced-brlck bulldlugs.

OCR for page 26
35 30 25 - a) ~ 20 a) CL - cn CO o 10 5 / _ / BE // ~.3/~: / 3C 4A DB / i// - - ~ '-''ZU-A- V Vl Vll Vll IX MM INTENSITY FIGURE 4-2 Loss ratio versus Modified Mercalli Intensity (mean damage ratio curves). Designations on curves refer to Table 4-1 construction classes. Source: Algermissen and Steinbrugge (1984~. One method for expressing the distribution of damage is a dam- age probability matrix (DPM) (Table (2).t The spectrum of damage, iIn Table 4-2, the original source (ATC-13) used MMI levels XI and XII to represent increasingly severe shaking severities beyond MMI X. As noted earlier, confusion results when this is not explicitly stated, because a literal reading of XI and XII indicates ground failure and at XII ~total" damage. In Table 4-2 the DPM has been truncated at MMI X to avoid different portrayals of MMI when definitions for MMI XI and XII may not be clear to the reader.

OCR for page 26
36 TABLE 4-2 A Damage Probability Matrix Form D amage Central Factor Damage Probability of Damage (in Percent) Range Factor by MMI and Damage State Damage State (percent) (percent) VI VII VIII IX X 1--None 0 0.0 95.0 49.0 30 14 3 2--Slight 0-1 0.5 3.0 38.0 40 30 10 3--Light 1-10 5.0 1.5 8.0 16 24 30 4--Moderate 10-30 20.0 0.4 2.0 8 16 26 5--Hea~ry 30-60 45.0 0.1 1.5 3 10 18 6--Major 60-100 80.0 -- 1.0 2 4 10 7--Destroyed 100 100.0 -- 0.5 1 2 3 NOTE: These definitions are used as a guideline: 1--None: no damage. 2--Slight: limited localized minor damage not requiring repair. 3--Light: significant localized damage of some components generally not . . . requlrlng repair. 4--Moderate: significant localized damage of many components warranting repair. 5--Heavy: extensive damage requiring major repairs. 6--Major: major widespread damage that may result in the facility being razed. 7--Destroyed: total destruction of the majority of the facility. aExample values are listed. S OURCE: Applied Technology Council ( 1985~. from none to total, is divided into damage states, each of which is described both by words and by a range of damage ratios. For each intensity of ground shaking, numbers in a column give the fractions of buildings experiencing different damage states; the numbers in each column sum to unity. Fragility curves (Figure ~3) provide essentially the same infor- mation as does a DPM, but in graphical rather than tabular form. Each curve gives, as a function of the intensity of ground shaking, the probability that the indicated damage state is squalled or exceeded. While the curves shown in Figure 4-3 are only for one construction class (wood frame), the general form is typical. The steeper the slope of a curve, the less the variability in expected performance for that damage state. The steep slope of low-damage curves 1 and 2 implies

OCR for page 26
37 that it is relatively easy to predict that this class will have only slight structural damage or only nonstructural damage at low intensities. DPMs and fragility curves provide the same information in dif- ferent formats. Thus, the choice between DPMs and fragility curves is a matter of style and precedent. The DPM originated in connection with loss est~rnates for buildings. Use of fragility curves developed in studies of the performance of mechanical equipment and have been applied in seisrn~c risk studies for facilities such as nuclear power plants. It is important to note that mean loss ratios may be cal- culated from the information in DPMs or fragility curves, but the reverse is not true; information about the distribution of damage about a mean cannot be inferred from a mean loss ratio curve. Evaluating Motion-Damage Relationships The loss ratio curves in Figure ~2 were constructed, employ- ing considerable judgment, using loss data gathered during various earthquakes, principally those occurring in California and a few other western states, along with data from foreign earthquakes where con- struction has been compatible. Actual data of this type are most complete for wood-frame dwellings (these data do not appear in Fig- ure 4-2), and more judgment has been required to construct curves applicable to other buildings. In a few cases, DPMs have been constructed using data from actual earthquakes, tempered with judgment. A recent report com- piled data on earthquake damage from a variety of sources (Thief and Zsutty, 1987) and indicates the usefulness of hard data about past performance in studies that attempt to estimate future performance. However, for many types of buildings, and especially for those in areas that have experienced few if any damaging earthquakes, actual data are either very sparse or nonexistent. For such buildings, it is necessary to rely on expert opinion to develop loss ratio curves, DPMs, or fragility curves. A systematized Delphi method approach was used to synthesize diverse expert opinions into the family of DPMs found in the AT~13 study. The pane! examined the method used to develop these DPMs and considered the credibility of the results (see Working Paper E). Concern was expressed that the ATC-13 DPMs underestimated the dispersion in the damage because zero probabilities were assigned in each column to damage states away from the predominant damage state. However, in the ATC-13 method, each matrix is meant to

OCR for page 26
\ \ LLJ In to J o 1 LL CE ill LD cn 1 LL LL C) o \ CE z o 1 UJ IS cn 1 1 1 \ xJ - 0 ~ - 1 1 CO ~ \ \ \ C')\ - \ \ \ \ \ \] 1 1 1 1 Cot Cot . . . . 31V ~ S Kiowa Jo ~1llOVEOUd 38 - z O ~ CD Z LL LD ~ Z LL Z LL] o G CD A CO Go - V v v ._ - _ O O g - ._ be cB o - CO or

OCR for page 26
39 apply for average California design and construction, and the AT~13 report provides a method for combining adjacent columns in a DPM ~ . .. . ~ . . ~ to renect the dispersion Introduced when good, average, and above- average construction are lumped together. The pane} recommends the development of new DPMs that incorporate this range of different qualities of construction. For common building types, loss ratio curves calculated from the DPMs in AT~13 are very close to the corresponding curves developed by the ISO. For less common buildings (e.g., tilt-up wall construction) for which there are only limited data, the differences in loss ratios expressed by the ISO and ATC-13 methods are within the range of uncertainty in the data. The best use of the ATC413 DPMs, in the panel's view, is for building types for which there are no ISO curves. Both the ISO loss ratio curves and the ATC-13 DPMs are in- tended primarily for use in California. The question then is: How should motion-damage relationships be developed for use in loss em timates for other areas? One answer lies in using expert opinion to modify the California-based information for the types of buildings found in the area to be studied. Analysis of some selected build- ings can assist by indicating the general level of seismic resistance of generic examples of building types in relation to the resistance of the buildings forming the data base. A Look to the Fature It is clear that there are major gaps and uncertainties in the state of the art for evaluating damage from an earthquake. Improvements in this situation can come about only by systematically collecting data from actual earthquakes. More effort should be devoted to this purpose, not only for earthquakes in the United States but also for earthquakes in other countries. In all such future studies, the distribution of damage should be documented not just the mean loss ratios, and not just by documenting interesting or dramatic individual failures in a reconnaissance overview. There has been an effort to develop and use empirical relations connecting damage directly to magnitude and distance from an earth- quake (Steinbrugge et al., 1984~. This approach bypasses the need to evaluate the intensity of ground shaking at sites, and avoids di~cul- ties in using MMI. Initial efforts to establish such relations are under way using data from earthquakes in California. This is an interesting

OCR for page 26
40 idea and should be pursued, but there are obvious limitations and difficulties. First, different relations will be necessary for different soil and topographic conditions. Second and more important, dif- ferent relations will be required for different regions of the country according to variations in attenuation of motion with distance. LOSSES ASSOCIATED WITH BUILDINGS One form of loss the cost of repair has already been discussed in the previous section. The total cost of repair may be obtained by simple summations, such as: (dollar value in each category) x MAR, all building categories or (average dollar value) x (number of buildings) x MERE, all building categories where MART is the loss ratio (or mean damage ratio) for the intensity of the scenario earthquake. Such summations are made for subareas of constant intensity and are then combined. Considering uncertainties that will inevitably exist in the inven- tory and the additional uncertainties in motion-damage relations, the accuracy of the estimated loss for a given scenario is not great. A prudent claim would-be accuracy to within a factor of 1.5 for the aggregation of singI+family, wood-frame California dwellings, 3 for commercial, industrial, and institutional buildings, and an order of magnitude (factor of 10) for an area with no recent earthquake history.2 However, even such uncertain estimates are still very useful for hazard reduction efforts and emergency planning. 600. 2 these expressions of uncertainty indicate the panel's judgment as to the accuracy with which losses can be estimated. A precise statement about the meaning of these ranges is not possible with the present state of the art, but the following example indicates a reasonable interpretation: Statement: "Uncertain by a factor of 3.~ Interpretation: Best estimate, 1,000; high estimate, 1,800; and low estimate,

OCR for page 26
41 The estimation of other types of losses-casualties and homeless- ness-is more complex and difficult. Casualties Of all the losses to be estimated, deaths and injuries are perhaps the most important to governmental organizations. Protection of life is a primary function of government and a prime incentive for undertaking hazard reduction. Estimates of casualties are desired for different times of day-typically mid-day, at night, and perhaps at a commuting hour-and sometimes for different seasons of the year. Unfortunately, the ability to predict casualties is not as good as In the case of property loss. Data on which rational, systematic esti- mates can be made are very sparse. The early NOAA-USGS studies generally used historical rates of casualties per unit of populaton for wood-frame dwellings and estimated rates for other types of con- struction, or used city-wide casualty rates from previous earthquakes applied to the population as a whole, adjusted up or down based on changes in construction practice. These estimates were In effect crude extrapolations of the limited data available, primarily from California earthquakes. A method specifically intended to estimate life safety risk factors for most of the ISO construction classes was devised by McClure et al. (1979) and applied to the problem of prioritizing engineering studies for buildings owned by the State of California. More recently (e.g., in the ATC-13 project) the tendency has been to relate casualties to levels of damage. For example, Table 4-3 gives casualty rates tied to the damage states described in Table (2. These rates are then multiplied by the estimated numbers of people in buildings of varying classes. This information is based on limited data plus considerable judg- ment. This does represent a rational approach to estunating casual- ties, and the pane} recommends use of this method combined with careful judgment and comparison with historical data, where com- parable cases pertain. It is essential that it be used with a DPM that reflects the considerable dispersion of damage among buildings of any one type, and the recommendations in ATC-13 for noting variations in construction quality should be followed. .

OCR for page 26
42 TABLE 4-3 Injury and Death Rates in Relation to Damagea- Central D amage Damage Factor Fraction Injured State (percent) Minor Fraction Serious Dead 1 0.0 0 0 0 2 0.5 3/100,000 1/250,000 1/1,000,000 3 5.0 3/10,000 1/25,000 1/100,000 4 20.0 3/1,000 1/2,000 1/10,000 5 45.0 3/100 1/250 1/1,000 6 80.0 3/10 1/25 1/100 7 100.0 2/5 2/5 1/5 aEstimates are for all types of construction except light steel construction and wood-frame construction. For light steel construction and wood-frame construction, multiply all numerators by 0.1. SOURCE: Applied Technology Council (1985~. It is evident that estimates of casualties will be very crude and uncertain, and this uncertainty should be represented by, for exam- ple, giving ranges of estimates, along with providing the best estimate figures. Homele~nese Estimates for the number of people requiring shelter by public agencies are also important for planning postdisaster operations. It is even more difficult to make such estimates, partly because data are scarce and partly because potential need ~ a function of weather conditions and the ability and inclination of the population to find their own shelter, such as with friends and relatives. The NOAA-USGS studies used a 50 percent dwelling damage ra- tio as an indicator of the need for alternative shelter. The most com- plete effort at systematic estimation of homelessness is by Gulliver (1986), who suggested a 20 percent damage ratio as the threshold point past which homelessness results. Clearly, great judgment is re- quired when estimating homelessness, and any estimate will involve a high level of uncertainty.

OCR for page 26
43 Estimates of casualties and homelessness should be regarded as having an order of magnitude (factor of 10) uncertainty, although it is possible to provide a tighter range of estunates when a study is restricted to a few well-understood classes of construction. These obviously are both matters for which far more data from actual earthquakes are required to advance the state of the art.

OCR for page 26
44 An example of the effects of landslides and debris flows triggered by the March 5, 1987, earthquakes (M 6.1 and 6.9) along the eastern flank of the Andes in north-central Ecuador. Destruction can be seen of the Trans-Ecuadorian oil pipeline (indicated by arrows) and adjacent highway by a debris flow issuing from a minor tributary of the Coca River. Photo courtesy of R. L. Schemer.