Socioeconomic Clustering in Seismic Risk

Assessment of Urban Housing Stock

J. S. R. Prasad,a) Yogendra Singh,b) Amir M. Kaynia,c)
and Conrad Lindholmd)

A seismic risk assessment methodology based on socioeconomic

clustering of urban habitat is presented in this paper. In this methodology, the
city is divided into different housing clusters based on socioeconomic level of
occupants, representing reasonably uniform seismic risk. It makes an efficient
utilization of high resolution satellite data and stratified random sample survey
to develop the building stock database. Ten different classes of socioeconomic
clusters found in Indian cities are defined and 34 model building types (MBTs)
prevalent on the Indian subcontinent have been identified and compared with
the Medvedev-Sponheuer-Karnik (MSK) scale, European macroseismic scale
(EMS), parameterless scale of seismic intensity (PSI), and HAZUS
classifications. Lower and upper bound damage probability matrices (DPMs)
are estimated, based on the MSK and EMS intensity scales and experience
from past earthquakes in India. A case study of Dehradun, a city in the foothills
of Himalayas, is presented. The risk estimates using the estimated DPMs have
been compared with those obtained using the PSI scale. It has been observed
that poorer people are subjected to higher seismic risk, both in terms
of casualties and in terms of percent economic losses.
关DOI: 10.1193/1.3158547兴

INTRODUCTION
A need to develop methodologies for reliable seismic risk assessment is being ap-
preciated globally, so that effective risk mitigation policies can be developed. India is
one of the most seismically active countries of the world, and past earthquakes have ex-
posed the high seismic vulnerability of its housing stock, resulting in huge life and eco-
nomic losses. In recent years, steps involving both national and international initiatives
have been taken toward seismic risk assessment and mitigation. This paper presents a
reliable and cost effective methodology for seismic risk assessment of urban habitat, de-
veloped during a pilot study conducted for the Himalayan city of Dehradun, under the
Indo-Norwegian Programme of Institutional Cooperation.
In seismic risk assessment, estimation of earthquake hazard, structural vulnerability
and exposure of building stock are the three equally important components, out of

a)
Research student, Department of Earthquake Engineering, IIT Roorkee, Roorkee-247667, India
b)
Assoc. Professor, Department of Earthquake Engineering, IIT Roorkee, Roorkee-247667, India
c)
Deputy Division Director, Norwegian Geotechnical Institute, N-0806 Oslo, Norway
d)
Senior Research Geophysicist, NORSAR, P.O. Box 51, N-2027 Kjeller, Norway

619
Earthquake Spectra, Volume 25, No. 3, pages 619–641, August 2009; © 2009, Earthquake Engineering Research Institute
620 PRASAD ET AL.

which, the development of inventory databases is “the most difficult aspect of damage
prediction studies” (ATC-13 1985). In many seismic risk assessment studies, necessary
information on building stock is derived from available data sources (NIBS and FEMA
2006, IDNDR 1999). The use of high resolution remote sensing imageries and aerial
photographs, together with “Rapid Visual Screening” surveys of buildings, is also a
common method to generate database of building stock (Morales 2002, Hofstee and Is-
lam 2004). The Indian municipal and housing development enforcement system is quite
weak, and the relevant building stock information at city level is not readily available.
The information available from the Indian census is also inadequate. In a huge, densely
populated and developing country like India, a comprehensive building survey is very
costly and time consuming, and the required number of skilled people is also not
available.
Databases of socioeconomic parameters have also been used as sources of inference
for seismic risk distribution (Chen et al. 1997, Papadopoulos and Arvanitides 1996, Me-
noni et al. 1997, Cutter et al. 2003), in the absence of building stock inventories. It can
be argued that not only are the demographic and economic risk parameters dependent
upon the socioeconomic level of communities, but that the construction and mainte-
nance of buildings is also governed by the socioeconomic status of their occupants.
Therefore, it can be a reasonable approach to generalize and aggregate the various seis-
mic risk parameters at the level of housing clusters, representing more or less uniform
socioeconomic conditions.
Another bottleneck encountered in seismic risk assessment is the development of re-
liable vulnerability functions, in the absence of systematic damage data for the past
earthquakes. Use of the MSK and MMI intensity scales is common (IDNDR 1999,
BMPC 2006) in describing the expected damage scenarios. Recently, more objective in-
tensity scales, such as the parameterless scale of seismic intensity (PSI) (Spence et al.
1991) and European macroseismic scale (EMS) (Grunthal 1998) have been developed.
The properties of constituent materials and construction practices, and hence, the vul-
nerability of similar model building types (MBTs) can vary significantly from region to
region. HAZUS presents a rational approach to consider the effect of these parameters,
in terms of capacity curve of buildings (NIBS and FEMA 2006). Applicability of ca-
pacity curves prescribed by HAZUS, can be examined by comparing the Design Base
Shear Coefficient and ductility provisions of the concerned national codes with those of
UBC-94.

SOCIOECONOMIC CLUSTERING
India is a country of wide socioeconomic disparity. Cities undergo a natural division
into different clusters/neighborhoods based on socioeconomic levels of residents. As the
city grows with time, such clusters develop naturally. In the planned cities, also, the
housing clusters are arranged according to the income levels. Not only the size of the
dwelling, but also, the material and technology, used in a building, depend on the eco-
nomic status of the occupants. Hence, it can be argued that the seismic risk of building
stock depends to a large extent, on the socioeconomic level of the occupants, together
with the historical development and prevailing construction practices.
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 621

Two types of development patterns are prevailing in India. In the traditional pattern,
independent houses are built by individuals for their own use but are sometimes sub-
rented to two or three other families. These houses are generally constructed using local
skill and technology without the involvement of any engineer. Group housing or multi-
family (attached homes/apartments), multi-storyed buildings, are a relatively recent
trend. Generally, families of equal economic status live in a group housing complex con-
sisting of several multi-story building blocks. These constructions are built by profes-
sional builders on contract or developed and promoted by big real estate companies.
Sometimes, government agencies are also developing and constructing such complexes.
Services of trained architects and engineers are employed in design and construction of
these buildings. The designs/floor plans are required to be passed by local administra-
tion. However, proper enforcement of earthquake resistant provisions is generally lack-
ing. Further, mixed occupancy is another characteristic of Indian urban habitat. Com-
mercial, and sometimes even small manufacturing units, are dispersed within primarily
residential clusters. Based on the concentration of the commercial activity, these clusters
can be classified as residential or commercial clusters. Tables 1–3 describe the typical
characteristics of various socioeconomic clusters, and Figure 1 shows typical examples
of residential and commercial clusters.

INDIAN MODEL BUILDING TYPES

In India, the construction of building stock is dominated by the locally available ma-
terials, traditional constructional practices, topography of the area, and income level of
the owner. The implementation of the codes is rarely observed in the housing construc-
tion. Hence, the evaluation of lateral load resisting capacity of existing buildings is a
difficult exercise. Further, data on strength characteristics of existing building types are
not available. A practical approach is to classify the existing buildings into Model Build-
ing Types of having reasonably uniform vulnerability characteristics, based on their per-
formance during past earthquakes and understanding of the primary features of vertical
and horizontal framing systems and the number of stories in the buildings.
In this study, 34 model building types, representing existing buildings of the Indian
subcontinent, have been identified as shown in Table 4. The existing constructions have
been broadly classified into three classes, according to vertical framing system (as de-
fined in the case of MSK and EMS scale classifications), given that this plays the most
important role in earthquake resistance. These are (i) adobe and random rubble masonry
construction, (ii) masonry wall construction using rectangular units, and (iii) framed
structures. However, there is a possibility of substantial difference in the lateral resis-
tance of the buildings, within the same class, due to the existence of different types of
units, mortar, and roof/ floors. Hence, each class is subdivided based on the possible
combination of units, mortar and roof/ floor types. The role of the roof/ floor system is
crucial in developing integral action of the vertical framing elements and these contrib-
ute the most of the seismic mass of the building. Roofs/ floors of six different types have
been identified and different commonly used combinations with vertical frame systems
have been considered.
622 PRASAD ET AL.

Table 1. Characteristics of different socioeconomic clusters in independent housing

Cluster Label Characteristics

Slum RS Inhabited by people with uncertain

means of livelihood, such as construction
laborers, rickshawand cart pullers, daily
wagers, and low-paid unorganized industry
workers, etc. Slums aregenerally situated in
low-lying areas along drains/riverbeds.
Buildings vary from temporaryshacks to
permanent, but poorly maintained, structures.
Generally single—or sometimes double story
construction with very simple floor plans and
elevations. A variety of low cost materials
are used. Adobe and un-plastered brick
masonry walls are common. Flooring is
generally in mud and/or burnt clay bricks.
Population density in these localities is
extremely high. Most slums lack roads,
clean water, electricity, sanitation and other
basic services. Obviously, seismic aspects
have no consideration in the construction.
Low Income RL Occupied by low paid government and
private sector workers, generally having
a regular sourceof income. The majority of
dwellings are double storied, built by
employing local mason without any use of
machinery. Walls and roofs are made of
poor to average quality of materials.
Masonry walls unplastered from outside,
are commonly seen. Flooring is generally
in cement concrete or unpolished stone.
Buildings are generally built continuously
without any gap and population density in
these clusters is very high. Electricity and
water are generally provided.

The natural period of vibration and base shear of the buildings depend strongly on
their height. Hence, the number of stories is also an important parameter to be consid-
ered in the vulnerability classification. In the proposed classification, the number of sto-
ries for load bearing wall buildings has been considered as per prevailing practice and
for framed structures, HAZUS classification has been followed (NIBS and FEMA
2006).

PREPARATION OF BUILDING STOCK INVENTORY

The first step in the inventory preparation is identification and mapping of various
socioeconomic clusters in the city. For this purpose, information from all the available
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 623

Table 1. (cont.)

Cluster Label Characteristics

Middle Income RM Occupied by small/middle businessmen,

shopkeepers, government servants, middle level
workers in commercial and industrial establishments,
etc. Economically this is a wider strata of society
which is further classified as lower middle (RML) and
upper middle (RMU) income class. Nowadays, these
localities are usually much better planned with wider
roads and approaches. However, old localities are
quite crowded with very narrow lanes. These have
simple to highly complex floor plans and elevations.
Walls are made of relatively good quality material.
Burnt clay bricks with cement mortar, plastered on
both sides are the most commonly used. Reinforced
concrete is also common for columns, beam and
floor/roof slabs. These have attached as well as
detached buildings. Generally more than one family
(owner and tenants) reside in one building. Population
density in these localities is also quite high.
High Income RH Occupied by the economically affluent class,
consisting of big businessmen, industry owners,
upper level executives and government officials,
ministers and established professionals. These
houses are better planned in terms of space utilization
and aesthetics, have lavish interior and exterior
architectural finishings and fittings. These buildings
have complex floor plans and elevations. Generally a
small family, with attached/separate small dwellings
for servants and their families, is residing in a large
detached house. Population density is generally low
with enough open/green spaces.

sources—e.g., topographical maps, high resolution satellite imageries, input from local
people, builders and various government and nongovernmental agencies—is to be uti-
lized. Then, the building stock inventory can be developed at cluster level, using the high
resolution satellite imagery and a sample survey. In the present study, the georeferenced
IKONOS satellite imagery has been used in ArcGIS to estimate the number of buildings
and total plinth (or footprint) area of the buildings in each cluster. It is a desktop exercise
that can be performed with reasonable effort and accuracy, except in crowded areas with
continuous construction of buildings separated by narrow paved lanes. In such areas, the
information from satellite imageries has been augmented by ground survey. Some re-
searchers (Mansouri 2006, Mueller et al. 2006, Toshiro et al. 2000) have used satellite
data to estimate the building heights, also, but the technique is not reliable for low-rise
624 PRASAD ET AL.

Table 2. Characteristics of different socioeconomic clusters in group housing

Cluster Label Characteristics

Low Income GL Generally constructed by the government for

economically weaker sections of the society.
Three—to four-storied or sometimes single-
storied URM construction with regular floor
plans and elevations. Walls are made of average
quality brick/concrete block masonry in cement
mortar. RC slab is common for floors/roofs.
Flooring is generally in cement concrete. Usu-
ally, the quality of workmanship and mainte-
nance is not adequate and there are visible signs
of deterioration in buildings. Population density
is quite high.
Middle Income GM These are planned by architects and have very
complex floor plans and elevations due to a
high level of competition among the builders.
Medium to high-rise RC frame structures with
URM infills, sometimes deigned for earthquake
forces but detailing and execution is not as per
national codes. Machinery is used in construc-
tion, but lower strength and low durability
concrete is common. These buildings generally
represent a poorly constructed structure hidden
inside lavish interior/exterior finishes, floorings
and fittings. Open ground story for vehicle
parking is common feature in most of these
buildings. Irregular framing system is also
commonly observed. Lift, emergency power sys-
tem and fire fighting equipment are provided.
High GH These are generally occupied by highly paid
Income professionals and big businessmen. These have
good aesthetics and large floor area per
household. Structurally, these have similar
characteristics as GM. Lift, emergency power
system, and fire fighting equipment are
provided.

buildings in densely populated areas (Miura and Midorikawa 2006). It is also possible to
identify some of the types of roofs using the satellite imagery, but it is not possible to
identify the wall material.
For identifying the building structural types and estimating the economic and demo-
graphic parameters, a building-to-building ground survey is ideally required. This is a
field exercise and requires a huge amount of resources and time to perform for all the
buildings in a populous city. Therefore, a survey of randomly selected buildings (about
10% samples) within the chosen representative clusters only, has been performed. Care
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 625

Table 3. Characteristics of different commercial clusters

Cluster Label Characteristics

Low CL Scattered commercial activities within a pre-

Commercial dominantly residential area generally catering to
the daily needs of the local population. The aver-
age value of contents in such units has been
estimated to be about 100% of building replace-
ment cost. These are front open structures on
ground floor of residential buildings.
Medium CM Shops off the main road in the prominent market
Commercial places of the city and along the main road
in not so prominent market places. The average
value of contents in such units has been
estimated to be about 200% of building replace-
ment cost. Generally, these consist of
continuous rows of front open shops along the
streets and storage spaces for other high
commercial activity shops. These are generally
single story or ground floor/basement of a
residential building.
High CH Road side shops in the crowded prominent market
Commercial places of the city, where a high value of
mercantile is stored in a small area. The average
value of contents in such units is generally
more than 400% of building replacement cost.
These shops may consist of single or double
story front open rows of shops or multistory
complexes housing several shops. Recently big
malls and stores are also coming up in big cities.

is required in selection of representative clusters, as there may be significant variations,

even between clusters of the same socioeconomic level. In the present study, the repre-
sentative clusters have been identified considering the socioeconomic level of occupants,
population, and commercial activity density, and age of construction. Using the satellite
data and the survey, the following average parameters have been estimated for each rep-
resentative cluster:
• Relative percentages of different MBTs present in the cluster
• Average plinth area of the buildings
• Average floor area of buildings
• Average number of stories per building
• Average number of households per building
• Average number of people per household during day and night times
• Average replacement cost of the buildings of each MBT
• Average value of contents per household
626 PRASAD ET AL.

Figure 1. Typical views of different residential and commercial clusters: (a) slum cluster; (b)
low-income cluster; (c) lower-middle-income cluster; (d) upper-middle-income cluster; (e)
high-income dwelling; (f) low-income group housing cluster; (g) middle-income group housing
cluster; (h) high-income group housing cluster; (i) low commercial activity within a residential
cluster; (j) medium commercial activity cluster; (k) high commercial activity cluster.

These parameters estimated for representative clusters have been used for other similar
clusters, along with the building footprint areas obtained from the satellite imagery.

VULNERABILITY ASSESSMENT
Different approaches have been used for vulnerability assessment in different parts
of the world. Expert opinion (ATC-13 1985, ATC-21 1988) and past earthquake expe-
rience (Porro and Schraft 1989, Spence et al. 1991, Orisini 1999, Pujades et al. 2000,
 SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK
Table 4. Proposed classification of model building types in Indian subcontinent

Most likely Vulnerability class as per existing

Description of Indian model building types classifications

S. Roof/Floor
No. Label Wall/Framing type Typea Stories MSK EMS PSI HAZUS

Adobe and Random Rubble Masonry

1 AM1 Rammed mud/sun-dried bricks/ R1, R2 1-2 A A AA1 Not defined
2 AM2 rubble stone in mud mortar R3 1-2 AR1
3 AL1 Rubble stone in lime-surkhi R1, R2 1-2 AR1
4 AL2 mortar R3, R4 1-2
5 AL3 R5 1-2
6 AC1 Rubble stone in cement mortar R1, R2 1-2
7 AC2 R3, R4 1-2
8 AC3 R5, R6 1-2
Masonry consisting of Rectangular Units
9 MM1 Burnt clay brick/ rectangular R1, R2 1-2 A A AA1 Not defined
10 MM2 stone in mud mortar R3, R4 1-2
11 MM3 R5 1-2
12 ML1 Burnt clay brick/ rectangular R1, R2 1-2 B B BB1, BD1
13 ML2 stone in lime-surkhi mortar R3, R4 1-2
14 ML3 R5, R6 1-2 BB2
15 MC1 Burnt clay brick/ rectangular R1,R2 1-2 BB1, BD1,
16 MC2 stone/ concrete blocks in cement R3, R4 1-2 BC1
17 MC3L mortar R5, R6 1-2 C BB2
18 MC3M 3+
19 ME1L Burnt clay brick/ rectangular R5, R6 1-2 Not Not Not defined
stone/ concrete blocks in cement defined defined
mortar and provided with seismic
20 ME1M bands and vertical reinforcement 3+
at corners and jambs

627
 628
Table 4. (cont.)

Most likely Vulnerability class as per existing

Description of Indian model building types classifications

S. Roof/Floor
No. Label Wall/Framing type Typea Stories MSK EMS PSI HAZUS

Framed Structures
21 RC1L RC frame/shear wall with URM R6 1-3 C C CC1 C3L-Precode
infills—constructed without any
22 RC1M consideration for earthquake 4-7 C3M-precode
forces
23 RC2L RC frame/shear wall with URM 1-3 Not D Not defined C3L (Pre-Code/
infills—earthquake forces defined Low-Code)
considered in design but detailing
24 RC2M of reinforcement and execution 4-7 C3M
not as per earthquake resistant
25 RC2H guidelines (Low-Code/ 8+ C3H
Moderate-Code)
26 RC3L RC frame/shear wall with URM 1-3 E Not C3L (Pre-Code/
infills-designed, detailed and defined/ Low-Code/
executed as per earthquake Not Moderate
resistant guidelines defined/ Code)
27 RC3M 4-7 DC-UBC-2 C3M
28 RC3H (Low-Code/Moderate-Code/ 8+ C3H
High Code)
29 ST1L Steel moment frames with URM 1-3 E Not defined S5L Pre-Code/
infills Low-Code/
30 ST1M (Low-Code/ Moderate-Code/ 4-7 S5M Moderate-
High Code) Code

PRASAD ET AL.
31 ST1H 8+ S5H
 SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK
Table 4. (cont.)

Most likely Vulnerability class as per existing

Description of Indian model building types classifications

S. Roof/Floor
No. Label Wall/Framing type Typea Stories MSK EMS PSI HAZUS

32 ST2L Steel braced frames 1-3 S2L Pre-Code/

(Low-Code/ Low-Code/
33 ST2M Moderate-Code/ 4-7 S2M Moderate
High Code) Code
34 ST3H 8+ S2H
a
Roof/Floor types: R1—Heavy sloping roofs-stones/burnt clay tiles/thatch on sloping rafters; R2—Heavy Flat flexible heavy roof-wooden planks, stone/burnt clay
tiles supported on wooden/steel joists with thick mud overlay; R3—Light sloping roofs—corrugated asbestos cement or GI sheets on sloping rafters without cross
bracing; R4—Trussed roof with light weight sheeting (without cross bracing); R5—Trussed/hipped roof with light weight sheeting (with cross bracing); R6—Flat
rigid reinforced concrete or reinforced masonry slab

629
630 PRASAD ET AL.

Coburn and Spence 2002, Hengjian et al. 2003) are commonly used. In HAZUS, fragil-
ity curves have been developed for generic MBTs in the United States, based on push-
over analysis of low-rise buildings with assumed uncertainties in expected hazard, struc-
tural capacity, and damage state thresholds (NIBS and FEMA 2006). Direct computation
of vulnerability functions, considering uncertainties in material strengths and hazard, has
also been performed by several researchers (Erberik and Elnashai 2003, Porter et al.
2001).
There is insufficient data available from past earthquakes for Indian buildings to be
able to develop the empirical damage functions. Analytical or hybrid fragility curves are
also not available for the Indian MBTs. In the present study, the characteristics of Indian
MBTs have been correlated (Table 4) with the classifications available in the MSK
(Medvedev et al. 1965), EMS-98 (Grunthal 1998), and PSI (Spence et al. 1991) scales of
intensity and HAZUS. The probability ranges of 10% to 20%, 15% to 55%, and 55% to
100% for the descriptive terms “few,” “many,” and “most,” respectively, based on EMS-
98, have been used to define the lower and upper bound damage probability matrices
(DPMs, Table 5) for the identified MBTs. While defining upper bound damage matrices,
first, the upper bound probability of the most severe grade of damage has been assigned,
and then the probabilities of less severe grades have been adjusted to keep the sum
within 100%. Arya (2003) has suggested some modifications in damage distribution cor-
responding to Intensities VI-IX, based on his experience of past earthquake damage sur-
veys in India. He has assigned slightly higher damage expectancy to Indian buildings
and has proposed a broader damage distribution, covering lower damage grades, as com-
pared to MSK and EMS-98. These modifications have also been incorporated in the es-
timated DPMs.
A comparison of the structural systems for various Indian MBTs, and those de-
scribed in HAZUS (NIBS and FEMA 2006), suggests that none of the fragility functions
given in HAZUS is directly applicable for the Indian adobe and masonry buildings. In
order to examine the applicability of the HAZUS vulnerability function for the Indian
framed buildings, design base shear coefficients of the IS:1893 (2002) at 0.2 sec and
1 sec periods have been compared with those of UBC 1994/97 for reinforced concrete
(RC) frames with OMRF, IMRF, and SMRF types of construction, as shown in Table 6.
Per IS:1893, the design base shear coefficients, ␣D, for rock/hard soil, is given as

ZI
␣D = 2.5 共for T = 0.2 sec兲 共1兲
2R

ZI
␣D = 共for T = 1 sec兲 共2兲
2R
where
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 631

Table 5. Lower- and upper-bound damage probabilities of proposed MBTs

Damage Probability (%)

MSK/ Lower Bound Damage Scenario Upper Bound Damage Scenario
Proposed EMS
MBT Intensity Gr1 Gr2 Gr3 Gr4 Gr5 Gr1 Gr2 Gr3 Gr4 Gr5

AM1, AM2, VI 15 10 0 0 0 55 20 0 0 0
AL1, AL2, VII 18 17 55 10 0 0 0 80 20 0
AL3, AC1 VIII 0 18 17 55 10 0 0 0 80 20
AC2, AC3, IX 0 0 43 42 15 0 0 23 22 55
MM1, MM2 X 0 0 0 45 55 0 0 0 0 100
MM3 XI 0 0 0 0 100 0 0 0 0 100
XII 0 0 0 0 100 0 0 0 0 100
ML1, ML2, VI 15 10 0 0 0 55 20 0 0 0
ML3, MC1, II 75 15 10 0 0 25 55 20 0 0
MC2 VIII 0 35 55 10 0 0 0 80 20 0
IX 0 0 75 15 10 0 0 25 55 20
X 0 0 43 42 15 0 0 23 22 55
XI 0 0 23 22 55 0 0 0 0 100
XII 0 0 0 0 100 0 0 0 0 100
MC3L, VI 10 0 0 0 0 20 0 0 0 0
MC3M, VII 52 10 0 0 0 67 20 0 0 0
RC1L, VIII 35 55 10 0 0 0 80 20 0 0
RC1M IX 0 75 15 10 0 0 25 55 20 0
X 0 0 75 15 10 0 0 25 55 20
XI 0 0 30 55 15 0 0 0 45 55
XII 0 0 0 0 100 0 0 0 0 100
RC2L, RC2M, VII 10 0 0 0 0 20 0 0 0 0
RC2H VIII 0 10 0 0 0 0 20 0 0 0
IX 75 15 10 0 0 25 55 20 0 0
X 0 75 15 10 0 0 25 55 20 0
XI 0 0 75 15 10 0 0 25 55 20
XII 0 0 0 0 55 0 0 0 0 100
RC3L, RC3M, IX 45 10 0 0 0 40 20 0 0 0
RC3H, ST1L, X 75 15 10 0 0 25 55 20 0 0
ST1M, ST1H, XI 0 75 15 10 0 0 25 55 20 0
ST2L, ST2M, XII 0 0 0 0 55 0 0 0 0 100
ST3H

Z the zone factor, defined as the effective peak ground acceleration for maxi-
mum considered earthquake,
I importance factor (considered equal to unity for comparison), and
R response reduction factor (3 for RC-OMRF, and 5 for RC-SMRF).
Similarly, per UBC-94/97:
Table 6. Comparison of design base shear coefficients, ␣D, per Indian Code and UBC 1994/97

632
IS 1893 (2002) UBC 1994/97

Zones II III IV V 1 2A 2B 3 4
Zone factors 0.10 0.16 0.24 0.36 0.075 0.15 0.20 0.30 0.40
Seismic Intensity VI VII VIII 艌IX V, VI VII VII 艌VII —
Expected damage Low Moderate Severe Very Minor Moderate Major Major
Severe
␣D at OMRF 0.042 0.066 0.100 0.150 0.057 0.107 0.142 0.214 0.285
0.2 sec
IMRF — — — — 0.036 0.068 0.090 0.136 0.182
period
SMRF 0.025 0.040 0.060 0.090 0.023 0.044 0.058 0.080 0.118
␣D at OMRF 0.016 0.026 0.04 0.060 0.023 0.043 0.057 0.086 0.114
1 sec
IMRF — — — — 0.014 0.027 0.036 0.054 0.073
period
SMRF 0.010 0.016 0.024 0.036 0.009 0.018 0.024 0.035 0.047
Proposed/HAZUS Ordinary Low- Moderate-Code High- Low-code Moderate-Code High-Code
classification of seismic (good) Code Code
design level quality of
construction
Inferior No- Low-Code Moderate- Pre-Code Low -Code Moderate-
quality of Code Code Code
construction

PRASAD ET AL.
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 633

I
␣D = 2.5CA 共for T = 0.2 sec兲 共3兲
R

I
␣D = CV 共for T = 1 sec兲 共4兲
R
where
CA , Cv seismic coefficients (=Z for soil profile type SB)
I importance factor (considered equal to unity for comparison), and
R structural system coefficient for overstrength and ductility (3.5 for RC-
OMRF, 5.5 for RC-IMRF, and 8.5 for RC-SMRF).
It is interesting to note that the value of R for OMRF and SMRF construction as per
IS:1893 correspond to the OMRF and IMRF, respectively, of UBC. It can be observed
from Table 6 that the design base shear coefficients for Zones 1 and 2B of UBC 1994/
1997 are comparable with those for Zones III and V, respectively, of IS 1893. Hence,
damage functions for pre-code, low-code, and moderate-code design levels of HAZUS
can be used for the corresponding MBTs of Indian construction with low-code,
moderate-code, and high-code design levels, respectively, as defined in the table. In In-
dia, a large number of buildings are being constructed, even today, without following any
earthquake code. Therefore, the HAZUS nomenclature of pre-code has been replaced by
no-code.

RISK CALCULATION
For a given damage state, the casualty rate depends on the type of construction and
on a number of secondary phenomena, including the total extent of damage caused by
the earthquake. HAZUS has prescribed indoor and outdoor casual rates for various dam-
age states in different MBTs. Coburn and Spence (2002) have proposed estimates of ca-
sual rates in collapsed masonry and RC buildings. However, they have not considered
the distribution of casual rates for different damage states. Further, their approach over-
estimates the deaths, by a large extent, as compared to the experiences in past earth-
quakes in India. The data available for Indian earthquakes is not sufficient and system-
atic to estimate the casual rates, therefore, in the present study, the expected life-loss and
injuries, and direct economic loss have been estimated using the procedure of HAZUS.
The casual rates, prescribed in HAZUS for unreinforced masonry buildings, have been
used, which predict the total life-loss and injuries in the range of observations in past
Indian earthquakes.
In the masonry building construction prevalent in India, the contribution of nonstruc-
tural components in the cost of building is rather low. Further, data on vulnerability and
the relative cost of nonstructural components is not available. Therefore, in the present
study, the loss due to building damage has been considered jointly for structural and
nonstructural components. The loss ratios for different damage states have been consid-
ered as given in HAZUS. The replacement cost per square meter of building floor area,
has been estimated through a builder survey in the city. The construction cost varies con-
siderably with the type of finishes, flooring and services, which is governed by the in-
634 PRASAD ET AL.

Figure 2. Dehradun City: (a) ward map and (b) identified socioeconomic clusters.

come level of the occupants. The contents value depends largely on the occupancy class
and socioeconomic level of occupants. A survey has been conducted to assess the value
of contents, in the ratio of building replacement cost, in residential and commercial oc-
cupancies.

CASE STUDY: DEHRADUN

The proposed approach has been applied for the Himalayan city of Dehradun. It is an
old city in northern India, having administrative and strategic prominence since the Brit-
ish era. The city houses a number of national academies, institutes and research centers.
In 2000, after the formation of the new state of Uttarakhand, the city became the capital
of the state, which resulted in rapid and uncontrolled growth of the city. Proximity to the
seismically active Himalayan region, rapid uncontrolled growth, sharply polarized so-
cioeconomic development, high population density, housing stock ranging from very old
traditional adobe/stone buildings to new masonry and RC buildings, generally without
the implementation of seismic codes, poses a severe earthquake risk to the city.
The city is divided into 60 municipal (census) wards (Figure 2a; Census of India
2001). Each ward covers a geographical area varying between 0.1 km2 and 3.2 km2 and
has a population varying between 2,810 and 11,852. A municipal ward generally con-
sists of several socioeconomic clusters ranging from slums (RS) to high income groups
(RH) and is too heterogeneous a unit for aggregation of seismic risk. A comprehensive
reconnaissance survey of the city together with the maps published by Survey of India
and IKONOS satellite data has been used to map the different socioeconomic clusters in
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 635

Table 7. Comparison of developed database with available census data

Dehradun municipal census

wards Number of households Population
Ward No. Name of the ward From survey From census From survey From census

1 Rispana 1756 1641 9042 8746

6 Raita mandi 1887 1796 10645 10114
11 Gandhi gram 1950 1819 10725 10046
26 Azabpur Dhanda, 2131 2157 10865 10130
Defense Colony
35 Nehru colony old 1525 1499 7472 7231
39 Doonvihar, Jakhan 1710 1744 7905 8291
47 Prempur Mafi 721 693 3606 3408

the city. Figure 2b shows the 254 identified clusters. These clusters were identified based
on homogeneity in construction quality, population density, cost of construction, and
value of contents, etc. Out of the identified clusters, 47 representative clusters were cho-
sen for sample survey. The survey was conducted by a team of students from the local
polytechnic school, trained for this purpose. The commercial clusters in Dehradun, typi-
cal of all Indian cities, consist of open-front (weak story) buildings (Figure 1). In view
of well documented poor performance of such buildings in past earthquakes (e.g., the
1971 San Fernando, 1989 Loma Prieta, and 1994 Northridge, and 1999 Chi-Chi earth-
quakes), these have been downgraded to the next highest vulnerability class. Using the
information obtained from the survey and the satellite data, a building stock database
has been developed.
The reliability of the developed inventory has been cross-checked with the ward wise
data available from the Census of India (2001). The contribution of different clusters (or
parts thereof) in a ward has been accumulated and compared with the census data. The
survey was conducted during May and June 2005. The city has undergone expansion and
significant increase in population after becoming the capital of Uttarakhand and the rate
of expansion is increasing with time. However, this expansion is mainly on the outskirts
of the city, without much effect on older clusters. Table 7 compares the number of
households and night time populations obtained from the survey with those reported in
the census, for seven relatively stable wards, where significant changes have not taken
place during 2000–2005. It can be seen that the data from the two independent processes
matches fairly for these wards, indicating the reliability of the database.
The Indian seismic zonation map (IS:1893 2002) divides the country into four seis-
mic zones with the associated MSK intensity as shown in Table 6. According to this zo-
nation map, Dehradun lies in seismic zone IV. Accordingly, the seismic risk scenario of
the city has been evaluated for MSK intensity VIII. Life-loss and injuries per thousand
people, and direct economic loss per square meter of built-up area and as a percentage of
the total economic value (cost of building replacement plus value of content) of the clus-
ter, have been calculated. Table 8 shows these estimates for typical clusters, using the
636 PRASAD ET AL.

Table 8. Loss estimates for typical socioeconomic clusters at MSK Intensity VIII

Life-loss per 1000 Injuries per Direct economic losses Direct economic
people 1000 people (INR/ m2 of built area) losses (%)

Cluster LB UB PSI LB UB PSI LB UB PSI LB UB PSI

RS 11 21 17 97 160 94 1091 1495 782 43 60 32

RL 1.57 3.1 5.5 28 45 39 902 1335 740 22 32 19
RML 0.04 0.05 3 13 20 24 1500 2288 1385 16 24 15
RMU 0.03 0.04 1.97 9 14 19 1230 1930 1329 11 18 13
RH 0.02 0.03 1.97 8.5 13 18 1413 2268 1578 10 17 12
CL 0.09 0.11 6 26 41 44 2767 4050 1926 30 45 22
CM 1.6 3.13 7.5 37 59 51 5736 8278 4319 29 42 21
CH 3.11 6.15 9 47 76 59 9091 12954 6328 28 40 20

lower-bound (LB) and upper-bound (UB) vulnerability matrices and the PSI scale. The
similar estimates for the whole city are shown in Table 9. It can be observed from Tables
8 and 9 that the PSI scale estimates life-loss higher, and injuries close to the upper
bound estimates, but it predicts economic loss lower than the lower bound estimates. An-
other interesting observation from Table 8 is that although the economic loss per square
meter increases with the economic level of the clusters, the life-loss and percent eco-
nomic loss show the reverse trend. The economically weaker sections of society are sub-
jected not only to higher life-loss but also to a higher percent economic loss. Figure 3
shows the distribution of life-loss, economic loss per square meter 共m2兲 of built area,
and percent economic loss, for the city, using PSI scale estimates. Again, the distribution
of expected economic loss per square meter is opposite to that of life-loss and percent
economic loss.
Figure 4 shows the effect of the variation of seismic intensity on the estimated risk
parameters in different socioeconomic clusters. It can be observed that for low and mod-
erate seismic intensities, the risk estimates predicted using the PSI scale are either in
between the lower and upper bond estimates, or these are close to the lower bound es-
timates of the MSK and EMS scales. In the case of high seismic intensities, the MSK

Table 9. Citywide risk assessment for MSK intensity VIII

Lower bound-Upper bound

Estimates PSI Estimates
Estimated
Parameter Night time Day time Night time Day time

Life-loss 555–1090 360–705 1,964 1,428

Injuries 10,931–16,895 7,388–11,411 15,098 11,549
Direct Economic 13.22–18.94 billion 12.25 billion
Losses (in INR)
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 637

Figure 3. Distribution of expected life and economic losses in the city for MSK Intensity VIII.

and EMS scales saturate and lower and upper bound estimates become equal for all the
clusters. This represents the subjectivity in damage observation, particularly in the case
of severe devastation. It is also inherent in the descriptive quantification of the damage,
which represents a very narrow band distribution of damage. However, the PSI scale
gives reasonable risk estimates even at higher seismic intensities.

CONCLUSIONS
A cost-effective methodology based on socioeconomic clustering has been presented
for quick seismic-risk assessment of populous cities lacking comprehensive building
stock databases. It allows for an efficient use of high-resolution satellite data and strati-
fied random sample survey. The methodology has been demonstrated for an Indian city
638 PRASAD ET AL.

Figure 4. Effect of variation of intensity on expected loss in typical clusters, using lower bound,
upper bound, and PSI estimates.
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 639

of Dehradun, located at the foothills of Himalaya. The reliability of the database devel-
oped using the proposed methodology has been established by comparing the demo-
graphic data with that obtained from the Indian census.
Various MBTs on the Indian subcontinent have been identified and compared with
the available classifications. Lower-bound and upper-bound DPMs have been estimated
for the identified MBTs, based on the MSK and EMS98 intensity scales, and experience
from past earthquakes in India. The seismic risk estimates, using the MSK-EMS and PSI
intensity scales have been compared. At low and high seismic intensities, the estimates
using the PSI scale are more reasonable. In the case of high seismic intensities, lower
and upper bound estimates using MSK-EMS scales tend to be equal, for all the clusters,
indicating the inherent subjectivity in damage observation, in case of severe devasta-
tions. It has also been observed that the poorer sections of the urban population are
prone to higher seismic risk, both in terms of expected life-loss, and percent economic
loss.

ACKNOWLEDGMENTS
The study was supported by the Indo-Norwegian Programme of Institutional Coop-
eration (INPIC) and the government of India’s Ministry of Human Resource Develop-
men. Help was received from the government of Uttarakhand’s Ministry of Disaster
Management in procuring the satellite data. Dr. Steven Gibbons provided help in editing
the language. The help received from “Technology and Research Network,” a Dehradun-
based welfare society, Professor D.K. Paul, Mr. R.K. Mukerji, and Mr. N.V.V. Raghavan,
in conducting the housing stock survey, is also gratefully acknowledged. The authors are
indebted to the anonymous reviewers whose thorough review and constructive sugges-
tions have significantly improved the paper.

REFERENCES
Applied Technology Council (ATC-13), 1985. Earthquake Damage Evaluation Data for Cali-
fornia, Report ATC-13, Redwood City, California.
Applied Technology Council (ATC-21), 1988. Rapid Visual Screening of Buildings for Potential
Seismic Hazards: A Handbook, Report ATC-21, Applied Technology Council, Earthquake
Hazard Reduction Series 41. Washington, D.C.
Arya, A. S., 2003. Rapid Visual Screening of Buildings in Various Seismic Zones in India, Ca-
pacity Building Advisor, GOI-UNDP (DRM), New Delhi.
Building Materials and Promotion Council (BMPC), 2006. Vulnerability Atlas of India, Minis-
try of Urban Development, Government of India, New Delhi.
Bureau of Indian Standards (IS:1893), 2002. Criteria for Earthquake Resistant Design of Struc-
tures (Part-1), New Delhi.
Census of India, 2001. Office of the Registrar General & Census Commissioner, India.
Chen, Q., Chen, Y., Liu, J., and Chen, L., 1997. Quick and approximate estimation of earth-
quake loss based on macroscopic index of exposure and population distribution, Natural
Hazards 15, 215–229.
Coburn, A., and Spence, R., 2002. Earthquake protection, John Wiley & Sons Ltd., 2nd Edi-
tion.
640 PRASAD ET AL.

Cutter, S. L., Boruff, B. J., and Shirley, W. L., 2003. Social vulnerability to environmental haz-
ards, Soc. Sci. Q. 84, 242–261.
Erberik, M. A., and Elnashai, A. S., 2003. Loss estimation analysis of flat-slab structures, Nat.
Hazards Rev. 7, 26–37.
Grunthal, G., 1998. European Macroseismic Scale, Center Europeen de Geodynamique at de
Seismologie, Luxembourg. Vol. 15.
Hengjian, L., Kohiyama, M., Horie, K., Maki, N., Hayashi, H., and Tanaka, S., 2003. Building
damage and casualties after an earthquake, Natural Hazards 29, 387–403.
Hofstee, P., and Islam, M., 2004. Disaggregation of census districts: better population informa-
tion for urban risk management, in The 25th Asian Conference on Remote Sensing, Chiang
Mai, Thailand, 1206–1211.
Mansouri, B., 2006. Remote sensing and GIS applications for urban risk and disaster manage-
ment for Tehran, in The 4th International Workshop on Remote Sensing for Disaster Re-
sponse, Cambridge, UK, 25–26th September 2006.
Medvedev, S., Sponheuer, W., and Karnik, V., 1965. Medvedev-Sponheuer-Karnik (MSK) in-
tensity scale, Quoted from ATC -13 1985.
Menoni, S., Petrini, V., and Zonno, G., 1997. Seismic risk evaluation through integrated use of
geographical information systems and artificial intelligence techniques, in Proceedings of
the Review Meeting on Seismic Risk in the European Union, Brussels, Belgium, Vol. II, 117–
49.
Miura, H., and Midorikawa, S., 2006. Updating GIS building inventory data using high-
resolution satellite images for earthquake damage assessment: application to metro Manila,
Philippines, Earthquake Spectra 22, 151–168.
Morales, A. L. M., 2002. A case study of earthquake risk assessment in Cartago, Costa Rica,
Research Thesis, International Institute for Geo-information Science and Earth Observation
(ITC), Enschede, The Netherlands.
Mueller, M., Segl, K., Heiden, U., and Kaufmann, H., 2006. Potential of high-resolution satel-
lite data in the context of vulnerability of buildings, Natural Hazards 38, 247–258.
National Building Institute and the Federal Emergency Management Agency (NIBS and
FEMA), 2006. HAZUS-MH MR2 Technical Manual, Washington, D.C.
Orisini, G., 1999. A model for building vulnerability assessment using the parameter less scale
of seismic intensity (PSI), Earthquake Spectra 15, 463–483.
Papadopoulos, G. A., and Arvanitides, A., 1996. Earthquake risk assessment in Greece, in
Earthquake Hazard and Risk, V. Schenk (editor). The Netherlands, Kluwer Academic Pub-
lishers, 221–229.
Porro, B., and Schraft, A., 1989. Investigation of insured earthquake damage, Natural Hazards
2, 173–184.
Porter, K. A., Kiremidjian, A. S., and Legrue, J. S., 2001. Assembly-based vulnerability of
buildings and its use in performance evaluation, Earthquake Spectra 17, 291–312.
Pujades, L. G., Canas, J. A., Mena U., Espinoza. F., Alfaro, A., and Caselles, J., 2000. Seismic
risk evaluation in Barcelona, Spain, in Proceedings of the Twelfth World Conference on
Earthquake Engineering, Auckland, New Zealand.
SOCIOECONOMIC CLUSTERING IN SEISMIC RISK ASSESSMENT OF URBAN HOUSING STOCK 641

Spence, R. J. S., Coburn, A. W., Sakai, S., and Pomonis, A., 1991. A parameterless scale of
seismic intensity for use in seismic risk analysis and vulnerability assessment, Earthquake
Blast and Impact: Measurement and Effects of Vibration, SECED, Elsevier Applied Science,
Amsterdam.
Toshiro, S., Sotaro, T., and Hideki, H., 2000. Estimation and discussion of elevation data cal-
culated by IKONOS stereo images, in Proceedings of the Japanese Conference on Remote
Sensing, 283–284.
Uniform Building Code (UBC), 1994/1997. Uniform building code, International Conference
of Building Officials, Whittier, California.
United Nations International Decade for Natural Disaster Reduction (IDNDR), 1999. Risk as-
sessment tools for diagnosis of urban areas against seismic disasters (RADIUS), http://
geohaz.org/radius/.
(Received 22 January 2008; accepted 17 December 2008兲