A Proposed Rapid Visual Screening Procedure for Seismic Evaluation of RC-Frame Buildings in India
Sudhir K. Jain,a) M.EERI, Keya Mitra,b) Manish Kumar,c) d) M.EERI, and Mehul Shah

Poor performance of reinforced concrete (RC) frame buildings in India during past earthquakes has been a matter of serious concern. Hence, it becomes important to identify and strengthen the deficient buildings. When dealing with a large building stock, one needs evaluation methods for quick assessment of the seismic safety of existing buildings so that corrective retrofitting measures may be undertaken on the deficient buildings. This paper presents a review of some of the available methods for rapid visual screening (RVS) of RC-frame buildings and proposes a RVS method for RC-frame buildings in India based on systematic studies on damage data of the 2001 Bhuj earthquake. DOI: 10.1193/1.3456711 INTRODUCTION Massive damage caused by the 2001 Bhuj earthquake to modern RC-frame buildings in India has underlined the need for seismic evaluation of a huge stock of existing buildings. A number of seismic evaluation methods incorporating varying degrees of detail have been developed across the world. Rai (2005) reviews the different methods for seismic evaluation of existing buildings as developed in various countries. Most of the methods follow three level assessment procedures (or something quite similar to it) namely, (a) rapid visual screening (Tier 1 Evaluation), (b) preliminary assessment (or Tier 2 Evaluation), and (c) detailed evaluation (or Tier 3 Evaluation). Rapid visual screening (RVS) is a simple procedure for quick evaluation of a large building stock to prioritize the buildings for preliminary and detailed evaluations. It is usually based on walk down surveys requiring 15– 30 minutes on site for each building. RVS formats usually record the important components of seismic vulnerability and propose a scoring system that forms the basis for classifying buildings in different risk categories. Preliminary assessment techniques are employed to analyze the building performance when a more reliable assessment is required. This requires detailed information regarding the structural components, material properties and site conditions, and does not fall within the pur-

a)

Department of Civil Engineering, Indian Institute of Technology Gandhinagar, Chandkheda, Ahmedabad, India (skjain@iitk.ac.in) b) Department of Architecture, Bengal Engineering and Science University, Shibpur, Howrah, India (keyamitra@gmail.com) c) Department of Civil Engineering, Indian Institute of Technology Kanpur, Kanpur, India (manishkr@iitk.ac.in) d) School of Building Science and Technology, CEPT University, Ahmedabad, India (mrscept@yahoo.co.in)

709

Earthquake Spectra, Volume 26, No. 3, pages 709–729, August 2010; © 2010, Earthquake Engineering Research Institute

710

JAIN ET AL.

view of this paper. The in-depth evaluation through sophisticated structural analysis falls within the third category (detailed evaluation) of vulnerability assessment. This paper provides a brief overview of RVS procedures available in different countries and develops a rapid visual screening procedure for RC-frame buildings in India based on damage surveys of such buildings after the 2001 Bhuj earthquake. RAPID VISUAL SCREENING RVS prioritizes the buildings to be further studied so that technical and other resources could be most effectively utilized. The procedure involves a sidewalk survey without entering the building or entering the building for a short duration only. Typically, 15– 30 minutes are spent per building. RVS is useful particularly when the number of buildings to be evaluated is large, since properly trained nonengineers may collect data and then calculate scores. However, inclusion of only a few parameters may result in widely differing interpretations of the criteria by different individuals leading to inconsistent results.
RVS METHODOLOGIES IN THE UNITED STATES

A number of guidelines are available from the U.S. Federal Emergency Management Agency (FEMA) for seismic risk assessment and rehabilitation of buildings. These include FEMA 178 (1992), published in 1989 and revised in 1992, FEMA 310 (1998), developed as a revised version of FEMA 178 (1992), and FEMA 154 (2002), for rapid visual screening of buildings published originally in 1988 and revised in 2002. The FEMA 154 (2002) method (as given in its Appendix B) assigns a basic structural score based on the lateral force resisting system of the building. Performance modifiers are specified to take into account the effect of number of stories, plan and vertical irregularities, pre-code or post-benchmark code detailing, and soil type. This method has been adopted in several countries with some modifications in the data collection format or the values of performance modifiers. The Tier 1 screening phase of FEMA 310 (1998) has a set of evaluation statements for each building type, to identify the building performance level. Table 1 shows the basic scores and the modifiers assigned by FEMA 154 (2002) for moment resisting frame buildings. It shows a pre-code penalty for buildings designed and constructed before enforcement of the seismic codes. Similarly, a post-benchmark positive attribute is assigned to buildings constructed after the significant improvements in the code were implemented and enforced. The pre-code and post-benchmark modifiers have been given substantial weight when compared to the basic structural scores. Thus, there is a fair amount of reliance on the year of construction vis-à-vis the building codes in view of effective enforcement mechanisms prevalent in United States that makes it reasonable to assume that a building would fulfill the code requirements applicable at the time of construction. The FEMA procedure assigns a higher score for taller buildings. For instance, +0.4 is added to the score if building has four to seven stories and +0.6 is added if building has more than seven stories in high seismicity regions. The procedure addresses both vertical and horizontal irregularities, assigning a modifier of −1.5 and −0.5, respectively, in a high seis-

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

711

Table 1. Basic scores and modifiers for moment resisting frame buildings (FEMA 154 2002)
Low Seismicity Basic Score Mid Rise (4 to 7 stories) High Rise ( 7 stories) Vertical Irregularity Plan Irregularity Pre-Code Post-Benchmark Dense Soil Stiff Soil Soft Soil FINAL SCORE (S) 4.4 +0.4 +1.0 −1.5 −0.8 N/A +0.6 −0.6 −1.4 −2.0 Moderate Seismicity 3.0 +0.2 +0.5 −2.0 −0.5 −1.0 +1.2 −0.6 −1.0 −1.6 High Seismicity 2.5 +0.4 +0.6 −1.5 −0.5 −1.2 +1.4 −0.4 −0.6 −1.2

mic zone. This difference in the modifier value is recognition of the fact that vertical irregularities (such as soft stories) make a building far more vulnerable as compared to the plan irregularities. Further, vertical irregularities are easier to observe during sidewalk surveys than the irregularities in plan. FEMA considers the detrimental effects of soft story buildings or short columns as features subsumed within the vertical irregularity category. FEMA considers type of soil and assigns a modifier for dense soil, stiff soil, and soft soil (the basic score being for rock sites). The factors included in FEMA 154 (2002) were based on the HAZUS methodology.
TURKISH RVS METHODOLOGIES

At the behest of Metropolitan Municipality of Istanbul (BU-ITU-METU-YTU 2003), the Earthquake Master Plan for Istanbul was developed by a consortium of four leading Turkish universities in two teams: (1) Middle Eastern Technical University (METU) and Istanbul Technical University (ITU), and (2) Bogazici University (BU) and Yildiz Technical University (YTU). Multistage building assessment procedures were developed with three stages of assessment, namely, (a) first stage, or rapid visual assessment from the street; (b) second stage, requiring access to a building; and (c) third stage, requiring detailed computational assessment procedures. The RVS method developed by BU and ITU for preliminary prioritization of buildings is based on the ratio of roof displacement capacity to roof displacement demand determined for life safety performance criteria and collapse prevention performance criteria. The procedure seems to be fairly complicated and requires in-depth knowledge on the subject to develop the results. The RVS procedure developed by the METU has since been revised and reformu˘ lated by Sucuoglu et al. (2007) on the basis of data consisting of 454 three to six story RC-frame buildings surveyed after the 1999 Düzce earthquake and classified in four damage grades. This method assigns a “Basic Score” to different RC-frame buildings depending on the number of stories (up to 7) and the seismic zone. Since the METU

712

JAIN ET AL.

Table 2. Basic scores and vulnerability scores for reinforced concrete frame buildings ˘ (Sucuoglu et al. 2007)
Initial Performance Scores No. of Stories 3 4 5 and 6 Vulnerability Coefficient

60

PGV 80 (cm/s)
80 73 64

40

PGV 60 (cm/s)
107 91 76

20

PGV 40 (cm/s)
138 115 92

Soft Story 23 22 24

Apparent Quality 9 15 23

Heavy Overhangs 23 30 33

method was developed as part of the Earthquake Master Plan of Istanbul, it uses seismic zones based on the expected range of peak ground velocity (PGV) in the area under consideration as obtained from the microzonation map of Istanbul. Parameters corresponding to short columns, pounding and topographical effects were included in the original METU method (BU-ITU-METU-YTU 2003) but were ˘ dropped in the 2007 version (Sucuoglu et al. 2007). As the soil conditions were uniform and topography was flat in the district surveyed, these parameters were also dropped. The pounding effect was not observable in the Düzce database and hence it was excluded from the set of parameters. The short column parameter may have been dropped considering that it is not easy to observe from the street. The parameters chosen were number of stories, presence of a soft story, presence of heavy overhangs, and apparent building quality. Experience of past Turkish earthquakes suggested a very strong relationship between the number of stories and the degree of damage sustained and similar pattern was observed in the Düzce database as well. Soft stories were considered to be present in the situations where ground stories had fewer partition walls compared to the upper stories (and hence had lesser stiffness) or had taller clearances leading to further irregularities. Floor areas in the upper floors cantilevering out from the exterior column axes were considered as heavy overhangs. Apparent quality took into account material and workmanship quality and level of maintenance of a building to arrive at a three point classifica˘ tion system of good, moderate and poor. Tables 2 and 3 give the values as per Sucuoglu et al. (2007) for RC buildings. Unlike FEMA 154 (2002), which considers taller buildings to be less vulnerable, METU scores are lower for the taller buildings indicating

˘ Table 3. Vulnerability score modifiers for concrete buildings (Sucuoglu et al. 2007)
Vulnerability Feature Soft Story Apparent Quality Heavy Overhang Vulnerability Score Modifiers Does not exist= 0; Exists= −1 Good= 1, Moderate= 0, Poor= −1 Does not exist= 0; Exists= −1

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

713

higher vulnerability. This is based on statistical analysis of performance of buildings in 1999 Düzce earthquake wherein taller buildings suffered more damage. The final seismic performance score for a building is calculated using the formula:

Performance Score = BS +

VSM

VS

1

where VSM represents the vulnerability score modifiers, and VS represents the vulnerability score that is multiplied with VSM to obtain the actual modifier which is added to the basic score (BS). A lower performance score implies greater seismic vulnerability.
OTHER RVS METHODOLOGIES

The method suggested by National Research Council, Canada (NRCC 1993) is based on a Seismic Priority Index which accounts for structural as well as nonstructural factors including soil conditions, building occupancy, building importance, falling hazards to life safety, a factor based on occupied density, and the duration of occupancy. The Japanese procedure (JPDPA 2001) is based on Seismic Index Is for total earthquake resisting capacity of a story which is estimated as the product of a basic seismic index based on strength and ductility indices, an irregularity index, and a time index. The evaluation is based on very few parameters and lacks clarity regarding ranking of buildings based on a scoring system. In the method proposed by Hassan and Sozen (1997) for RC-frame buildings in Turkey, priority for remedial action is expressed in terms of a Priority Index obtained by adding wall and column indices. Wall index is obtained by normalizing the total area of shear walls and infill walls with the total floor area of the building. Similarly, column index is obtained by normalizing the total column area with the total floor area. Thus, the method is primarily based on two parameters, the total wall area and the total column area besides total floor area. Also, it has been assumed here that the seismic demand is reasonably uniform as are the quality and type of construction. The New Zealand code (NZSEE 2006) recommends a two-stage seismic performance evaluation of buildings. The initial evaluation procedure (IEP) involves making an initial assessment of performance of existing buildings against the standard required for a new building, known as percentage new building standard (%NBS; assessed structural performance of the building, taking into consideration all reasonably available information, compared with requirements for a new building expressed as a percentage). A%NBS of 33 or less means that the building is potentially earthquake prone according to the Building Act and a more detailed evaluation is required for the same. The process requires the expertise of earthquake engineers to yield quality results. A fuzzy logic based RVS procedure was developed in Greece (Demartinos and Dritsos 2006) for the categorization of buildings into five different damage grades in the event of a future earthquake. The method was developed based on information on 102 buildings affected by the Athens earthquake of 1999. The fuzzy logic-based RVS (FLRVSP) proposed a probabilistic reasoning method that treats the structural properties of

714

JAIN ET AL.

a building in a holistic way and gives a score that represents possible damage in the event of major earthquakes producing ground accelerations equivalent to the values provided by the relevant codes. There have been some efforts in India towards developing rapid visual screening methods. Sinha and Goyal (2004) have proposed a methodology for rapid visual screening of 10 different types of buildings. They have used a scoring system similar to that of FEMA 154 (2002). An Appendix in Draft IS 13935 (2004) suggests a format for rapid visual screening of masonry buildings. However, this format specifies the “damageability” of buildings in different zones based on the construction materials alone. For instance, all “unreinforced brick masonry buildings in lime mortar” are defined as Building Type B+ and placed in the same class of damageability. DAMAGE DATA OF INDIAN BUILDINGS The Bhuj earthquake of 26 January 2001, in the western state of Gujarat in India impacted urban constructions at a scale unprecedented in any other recent earthquake in the Indian subcontinent. While maximum damage was observed in the Kachchh area, Ahmedabad city (Seismic Zone III), located around 250 km away from the epicenter experienced shaking of Intensity VII on the Modified Mercalli Intensity (MMI) scale, and witnessed large scale damages and collapse of about 130 RC-frame buildings, leading to many fatalities (e.g., Jain et al. 2001). Damages observed in the RC-frame buildings include those due to open ground stories, short columns, irregular configurations, torsional irregularities, pounding effects etc. An important feature of the RC buildings in Ahmedabad was a highly irregular pattern of column placement, leading to lack of frame action of the structural system (Murty et al. 2002). Damage of Ahmedabad buildings was studied as it is the only sample available of a recent Indian earthquake that exposed the deficiencies in the construction typologies representative of urban India, in terms of structural systems, architectural designs, building materials, construction milieu, and regulatory framework. A team from Centre for Environmental Planning and Technology (CEPT) University, Ahmedabad surveyed 6670 buildings of Ahmedabad during the three month period following the earthquake and assigned damage grades (G0: no damage to G5: collapse), to RC-frame buildings, load bearing masonry buildings, and load bearing wooden frame buildings. A smaller subset of 3,720 load bearing masonry and RC-frame buildings from this survey is currently available with CEPT University. In 2007–2008, a group of students from CEPT University short listed 101 RC-frame buildings out of these 3,720 buildings, based on location in the city. A study was conducted by them on these 101 buildings where they collected detailed information on different vulnerability parameters as well as damages sustained. The study included site visits, as well as review of architectural and/or structural drawings. The information recorded for each structure included (1) general vulnerabilities such as torsion, maintenance, horizontal and vertical irregularities, hollow plinth, soft stories, stub columns, short columns, and diaphragm irregularities, and (2) detailed damage data. This sample was made available for the present study and will be referred as Phase I Sample here onwards.

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

715

Table 4. Vulnerability parameters and scores
Parameter Title Basement Number of Stories Maintenance Staircase asymmetry with respect to plan Re-entrant corners Open Story Stub Columns Short Columns Score

x0 x1 x2 x3 x4 x5 x6 x7

Absent= 0, Present= 1 N 5 = 0, N 5 = 1. Good= 0, Moderate= 0.5, Poor= 1 Absent= 0, Present= 1 Absent= 0, Present= 1 Absent= 0, Present= 1 Absent= 0, Present= 1 Absent= 0, Present= 1

To verify the observations based on Phase I Sample, and to make sure that these were not peculiar to the buildings that were selected in Phase I, another 169 RC-frame buildings (chosen from data set of 3,720 buildings referred to earlier) were surveyed by the authors with the help of students of CEPT University during a field visit in June 2008. This sample will be referred to as the Phase II Sample. The survey looked at some easy-to-observe vulnerability features in these 169 buildings. It was a sidewalk survey during which a two-member team spent about ten minutes in each building premise, observing the vulnerability parameters from the ground floor but without going into individual apartments or to the roof of the buildings. The Phase I Sample consisted of 30, 24, 46, and 1 buildings in damage groups G1, G2, G3, and G4, respectively, as per post-earthquake survey by CEPT in 2001. Here G1 is slight nonstructural damage, G2 is slight structural damage, G3 is moderate structural damage, and G4 is severe structural damage; more details of damage classification are available in Jain et al. (2002). These numbers were 44, 59, 65, and 1 for the Phase II Sample. None of the surveyed buildings belonged to G0 or G5 damage categories, which represent no damage, and total collapse, respectively. For analysis purpose, a third sample of 270 buildings was considered, comprising of all the buildings from Phase I and Phase II. Henceforth, this sample will be referred to as the Combined Sample. It has been assumed that all the building sites experienced a seismic shaking intensity of VII on the MMI scale as was reported for the city of Ahmedabad. VARIABLE SELECTION Overall, eight parameters were considered as listed in Table 4 along with the vulnerability scores. Generally when a parameter is present in a building, corresponding score has been assigned as 1, and 0 otherwise, except for the cases of number of stories and maintenance which are discussed elsewhere in the paper. For the statistical studies, arbitrarily chosen values of observed performance scores (OPS) of 100, 85, 70, 50, 25, and 0 were assigned to the buildings belonging to damage groups G0, G1, G2, G3, G4 and G5, respectively. These are somewhat similar to the values of 100, 80, 50, and 0 for no damage, light damage, moderate damage, and severe damage or collapse, respectively, ˘ used by Sucuoglu et al. (2007).

716

JAIN ET AL.

Table 5. Result of different variable selection techniques employed on the combined sample
R 2/ Adjusted R2
Suggested Independent Variables Forward Selection Backward Elimination Stepwise Selection Akaike’s Information Criteria Bayesian Information Criteria

x0 x1 x2 x3 x4 x5 x7

x0 x1 x2 x4 x7

x0 x1 x2 x4 x7

x0 x1 x2 x4 x7

x0 x1 x2 x4 x5 x7

x0 x1 x2 x4 x7

Various variable selection techniques (e.g., Montgomery et al. 2003) were employed on the Combined Sample to identify statistically significant vulnerability parameters, considering OPS to be the dependent variable. As can be seen from Table 5, vulnerability parameters, x0, x1, x2, x4, and x7 are commonly suggested by all the techniques, while x3 is suggested only by the R2/Adjusted R2 method, and x5 is suggested by R2/Adjusted R2 and the Akaike’s Information Criteria (AIC) methods. x6 is not suggested by any of the methods. The parameter open story x5 is a known factor behind poor seismic performance in RC-frame buildings, and is easy to observe through RVS. Hence it was decided to include x5 in the set of independent variables for regression analysis in addition to x0, x1, x2, x4, and x7. To see if multicollinearity was present among the vulnerability parameters considered for the analysis, variance inflation factors (VIF; e.g., Montgomery et al. 2003) were calculated for each of the independent variables. It was found that highest of those values was 1.18 (much less than 5) indicating that multicollinearity does not exist among the parameters. Besides, it was found that adding new cases to the Sample did not change the parameter estimates drastically which would be expected if multicollinearity was present. VULNERABILITY PARAMETERS The number of buildings with different vulnerability parameters for Phase I, Phase II, and Combined Samples has been shown in Table 6. Table 7 shows the percentage of buildings with a particular feature which suffered low (G1 or G2) or high (G3 or G4) damage for the three samples. For example, in the first row for the Combined Sample, it is seen that 32% of the buildings with basements sustained high damage (G3 or G4), while 45% of those without basement suffered such damage, implying that the buildings with basement performed better. Table 8 shows the correlation coefficients and the partial correlation coefficients between OPS and the scores for different vulnerability parameters. These values show significant positive or negative correlation between the OPS and these vulnerability parameters. However, to address the concern that in some cases the value of correlation

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

717

Table 6. Number of buildings with different characteristics for the three samples
Buildings with ↓ Total Number of Buildings → Phase I 101 27 69 63 76 94 26 Phase II 169 36 145 129 45 144 81 Combined 270 63 214 192 121 238 107

Basement Present Number of Stories 5 Moderate or Poor Maintenance Re-entrant Corners Present Open Story Present Short Columns Present

coefficient is somewhat low, 1,000 bootstrap samples (e.g., Montgomery et al. 2003) randomly extracted with replacement from the Combined Sample were analyzed. A bootstrap sample thus obtained may have one or more buildings more than once, while some of the buildings may not be there in the sample in such a way that the sample size is same as that of the Combined Sample. Figure 1 shows the histogram of correlation coefficients between OPS and different vulnerability scores. It is clear from the figure that the correlation coefficients between the different vulnerability scores and OPS show a consistent trend (either positive or negative) even though the individual values may be small.

Table 7. Percentage of buildings with different characteristics experiencing low (G1 or G2) and high (G3 or G4) damage for the three samples
Phase I Low Damage (G1/G2) With Basement Without Basement Number of Stories 5 Number of Stories 5 With Good Maintenance With Moderate or Poor Maintenance With Reentrant Corners Without Reentrant Corners With Open Stories Without Open Stories With Short Columns Without Short Columns 60 52 42 78 83 34 53 54 55 28 36 60 High Damage (G3/G4) 40 48 58 22 17 66 47 46 45 71 64 40 Phase II Low Damage (G1/G2) 74 57 57 87 63 55 40 68 58 80 52 69 High Damage (G3/G4) 26 43 43 13 37 45 60 32 42 20 48 31 Combined Low Damage (G1/G2) 68 55 52 82 67 42 48 66 57 69 47 65 High Damage (G3/G4) 32 45 48 18 33 58 52 34 43 31 53 35

718

JAIN ET AL.

Table 8. Correlation (and partial correlation) coefficients between observed performance score and different vulnerability scores for the three samples
Correlation Coefficients Between Observed Damage and ↓ Basement Score Story Score Maintenance Score Re-entrant Corner Score Open Story Score Short Column Score Phase I 0.05 (0.09) 0.30 (0.20) −0.47 −0.47 −0.04 −0.15 0.12 (0.04) −0.27 −0.31 Phase II 0.16 0.10 −0.06 −0.26 −0.15 −0.10 (0.17) (0.03) −0.09 −0.28 −0.11 −0.06 Combined 0.11 0.18 −0.26 −0.17 −0.06 −0.15 (0.14) (0.13) −0.25 −0.22 −0.09 −0.16

150 (a)

150 (b)

150 (c)

Bootstrap Samples

Bootstrap Samples

100

100

Bootstrap Samples −0.2 0 0.2 Correlation Coefficient 0.4

100

50

50

50

0 −0.4

−0.2 0 0.2 Correlation Coefficient

0.4

0 −0.4

0 −0.4

−0.2 0 0.2 Correlation Coefficient

0.4

150 (d)

150 (e)

150 (f)

Bootstrap Samples

Bootstrap Samples

100

100

Bootstrap Samples −0.2 0 0.2 Correlation Coefficient 0.4

100

50

50

50

0 −0.4

−0.2 0 0.2 Correlation Coefficient

0.4

0 −0.4

0 −0.4

−0.2 0 0.2 Correlation Coefficient

0.4

Figure 1. For 1,000 bootstrap samples drawn from the Combined Sample histogram of correlation coefficients between observed performance score and (a) basement score, (b) story score, (c) maintenance score, (d) reentrant corner score, (e) open story score, and (f) short column score.

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

719

150

150

n=3
Bootstrap Samples 100 Bootstrap Samples 100

n=4

50

50

0 −0.2

−0.1

0 0.1 0.2 Correlation Coefficient

0.3

0.4

0 −0.2

−0.1

0 0.1 0.2 Correlation Coefficient

0.3

0.4

150

150

n=5
Bootstrap Samples 100 Bootstrap Samples 100

n=6

50

50

0 −0.2

−0.1

0 0.1 0.2 Correlation Coefficient

0.3

0.4

0 −0.2

−0.1

0 0.1 0.2 Correlation Coefficient

0.3

0.4

Figure 2. Histogram of correlation coefficients between observed performance score and story score (=0 if number of stories n, and 1 otherwise) obtained using 1,000 bootstrap samples drawn from the Combined Sample.

In the following sections, each of the selected vulnerability parameters is briefly discussed.
NUMBER OF STORIES

It was felt that the parameter number of stories can be treated as an ordinal variable (0 or 1) rather than a metric variable (2, 3, 4,…). The buildings of the Combined Sample were divided into two groups, one with more than number of stories, n, (Story Score = 1) and other with less than or equal to n stories (Story Score= 0). Correlation coefficient between this score and the OPS came out as 0.085, 0.101, 0.180, and 0.152 as n was set as 3, 4, 5, and 6, respectively. Further, 1,000 bootstrap samples randomly drawn from the Combined Sample were analyzed. Figure 2 shows the histogram of correlation coefficients between OPS and story score for n ranging from 3 to 6 wherein the highest correlation between story score and OPS is observed when n is 5. When all buildings with more than 5 stories are considered, the correlation coefficient between OPS and number of stories (6, 7, 8,…) is obtained as −0.03. Similarly,

720

JAIN ET AL.

when all buildings with equal to or less than 5 stories are considered, the correlation coefficient between OPS and number of stories (1, 2, 3, 4, 5) is obtained as +0.04. These values of correlation coefficient are rather low, implying that all buildings with more than 5 stories can be considered as one group while those with less than or equal to 5 stories as another group. From Table 7 it can be seen that 18% of the buildings with number of stories more than 5, suffered higher damage (G3 or G4), while 48% of the buildings with number of stories equal to or less than 5, suffered higher damage (G3 or G4), clearly indicating that the performance of taller buildings was better during the earthquake. This is in line with FEMA 154 (2002), but at variance with METU method. Better performance of taller buildings may have been because of higher level of concern among the structural engineers and construction agencies for structural safety in such buildings, leading to better quality control in design and construction. It may however be mentioned that the requirements and enforcement of the codes were same in the city regardless of the building height.
BASEMENT

From Table 7, it is clear that the buildings without basement suffered higher level of damage as compared to the buildings with basement. That is, the presence of a basement is beneficial for the seismic performance of a building. This may be because buildings without basement tend to have significantly larger story height in the lowest story from structural view point. For instance, in the absence of a plinth beam, the height from top of the foundation to the soffit of the first floor beams is much higher than the column heights for the upper stories. Also, buildings with a basement tend to have raft foundation and reinforced concrete walls all around the basement boundary. These give such buildings stronger and stiffer foundations, as compared to buildings without basement which may be supported on isolated footings. Finally, buildings with a basement may tend to receive better engineering inputs both in design and in construction, as compared to those without basement.
MAINTENANCE

It is well known that the quality of construction has a significant impact on the seismic performance of a building. However, it is not possible during the rapid visual assessment to directly ascertain how good the quality might have been at the time of construction. But, by observing the current state of the building and its maintenance, an inference can be made about the original quality of construction. For instance, a group of buildings showing higher levels of distress currently (in the form of seepages, spalling of concrete, corrosion of steel, etc.) is likely to have received poorer quality of construction originally, as compared to the buildings that presently show better conditions. Of course, the age of a building will also affect the level of spalling and corrosion; however, that is not included here. Hence, during the two surveys, level of maintenance in a building was observed by looking for signs of leakages and corrosion. A limitation however is that these two surveys were conducted seven years after the earthquake. Some

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

721

poorly constructed buildings may have received much better maintenance after the earthquake, and it is possible that some buildings that look well maintained today were in fact poorly constructed. In Phase I survey, the buildings were classified as having good or poor maintenance, while in Phase II survey the buildings were classified as having good, moderate or poor maintenance. Table 7 indicates that the buildings with moderate or poor maintenance suffered higher grades of damage. This was more clearly evident from the Phase I Sample where 66% of the poorly maintained buildings suffered higher grades of damage (G3 or G4) while 34% of these buildings suffered lower damage (G1 or G2), whereas for well maintained buildings these numbers were 17% (G3 or G4) and 83% (G1 or G2), respectively. Therefore it is clear that seismic performance of a building is significantly affected by the quality of construction, which may be inferred from the current level of maintenance. This implies that the poorly or moderately maintained (low quality of construction) buildings suffered higher level of damage as compared to the buildings that were well maintained.
RE-ENTRANT CORNERS

Buildings with horizontal offsets (that is, re-entrant corners) are known to be more vulnerable seismically. A re-entrant corner was considered to be present when an interior corner was created in the plan footprint in such a way that the length of a wing projection beyond a re-entrant corner was greater than 15% of the length of the plan dimension in that direction; this is in line with the definition for re-entrant corners adopted by many building codes. This was determined from architectural drawings and field measurements, in Phase I and by eye estimation in Phase II.
OPEN STORIES

Open story was considered to be present when open parking was provided over the entire plan or over a part of the plan of the building in its ground story. For purposes of analysis, both full plan and partial plan open stories have been considered to be having the open story vulnerability. Of the 101 buildings in Phase I Sample, only 7 buildings did not have open stories, and hence this Sample was not considered appropriate for statistical analysis in terms of performance with or without open story. Among the 169 buildings in the Phase II Sample, 25 buildings did not have open stories. Thus, the Combined Sample of 270 buildings containing 238 buildings with and 32 buildings without open stories was considered for analysis.
SHORT COLUMNS

Columns that are confined along their length by masonry walls to accommodate openings are rendered “short” and therefore not free to deflect over their entire length. Short columns were identified in Phase I Sample, either through visual inspection of the building premises or through a review of the architectural and structural drawings. In the

722

JAIN ET AL.

Table 9. Parameter estimates and their standard errors obtained from 1,000 multiply imputed samples with different number of missing cases nadded of G0, G4, and G5 each added to the combined sample using parametric regression method nadded 0 10 20 30 40 50

A 76.4± 3.1 81.8± 4.8 84.8± 5.8 87.0± 6.3 88.3± 6.9 89.1± 7.4

x0 4.8± 2.1 7.9± 3.4 9.6± 4.1 11.0± 4.4 11.7± 4.8 12.4± 5.1

x1 4.7± 2.2 7.6± 3.6 9.4± 4.5 10.5± 5.0 11.5± 5.5 12.1± 5.6

x2 −9.3± 2.2 −15.3± 3.3 −18.9± 4.0 −21.1± 4.4 −22.8± 4.7 −24.1± 5.0

x4 −6.3± 1.8 −10.2± 2.7 −12.7± 3.3 −14.4± 3.8 −15.5± 3.9 −16.4± 4.2

x5 −3.8± 2.7 −6.3± 4.4 −7.5± 5.4 −8.7± 6.0 −9.3± 6.5 −9.7± 7.0

x7 −4.7± 1.8 −7.6± 2.8 −9.5± 3.4 −10.7± 3.7 −11.4± 4.1 −11.9± 4.2

Phase II survey short columns were marked as present if they were visible from outside the building. This was a limitation in Phase II survey because it was possible that short columns were present elsewhere in the building but not visible from the ground floor. The negative effects of short columns on seismic performance are very clear in Phase I (Table 7) where the recording of short columns was more accurate for reasons discussed earlier. The criteria for assigning damage grade to RC-frame buildings, used after the 2001 earthquake, laid a lot of emphasis on damage to RC columns and the nature of cracks in the columns. Since the short column effect leads to damage in the columns, this may explain a rather strong influence of this parameter in the damage grade. PROPOSED MODEL
REGRESSION ANALYSIS

A statistical analysis was performed on the Combined Sample. The following multiple linear regression equation is fitted to the Ahmedabad damage database for predicting the expected performance scores (EPS):

EPS = A + C0x0 + C1x1 + C2x2 + C4x4 + C5x5 + C7x7

2

Here, the vulnerability parameters, xi, are as defined in Table 4. Since buildings belonging to G0, G4, and G5 categories were not adequately represented in the Sample, and proportion of buildings for the damage groups was not known for the population, some fixed number of new cases belonging to each of these damage categories were added to the Sample and multiple imputation (MI) analysis (e.g., Rubin 1987) was carried out employing parametric regression method using SAS software. Parameter estimates Ci shown in Table 9 have been obtained by averaging the parameter estimates from 1,000 imputed samples. Corresponding standard errors account for variation in a sample as well as variation across imputed samples. Table 9 indicates that parameter estimates do not change much with the number of added cases to the Sample. Hence, it was felt that a representative rounded-off coefficient corresponding to each independent variable as well as the constant term could be used for the model. Thus, following model to predict the EPS is proposed:

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

723

120

100

Number of Bootstrap Samples

80

60

40

20

0 0

5

10

15

20

25

30

Average Absolute Percentage Error

Figure 3. Histogram of average absolute percentage error for the model obtained by using 1,000 bootstrap samples drawn from the Combined Sample.

EPS = 85 + 10x0 + 10x1 − 20x2 − 10x4 − 10x5 − 10x7

3

where xi are as defined in Table 4. Standard error of regression (mean-square residual) for the model is 15.84 for the Combined Sample. To determine the percentage average absolute error in the prediction of EPS and to get a sense of variability of the same, 1,000 bootstrap samples drawn from the Combined Sample were analyzed. Figure 3 shows the histogram of percentage average absolute error obtained from these samples. It can be seen from the figure that the mean of the average absolute percentage error is about 20%, while the minimum and maximum are about 17% and 25%, respectively. In order to get an idea as to how well the model predicts vulnerability, buildings of the Sample were classified based on their EPS. An EPS above 77.5 was considered as G1, between 60 and 77.5 as G2, between 37.5 and 60 as G3, and less than 37.5 was considered as belonging to G4 damage group. Table 10 shows the distribution of buildings based on their observed damage category vis-à-vis their predicted damage category. If after classification a building lies in its original group (as obtained from the OPS), it is considered to be correctly classified, and otherwise, the classification is deemed as incorrect. Further, in case of incorrect classification, if a building grouped as G2 based on OPS, is placed in G3 EPS group, or in G1 EPS group, the situation is defined as first level of incorrectness, characterized by a shift of one level (e.g., G1 to G2, G2 to G1, G2 to G3, G3 to G2, etc.). Table 10 shows that 124 (46%) of the 270 buildings have been correctly classified and 114 (42%) buildings are in the first level of incorrectness group. In addition to the 124 buildings which were identified correctly with their group, there were another 102 (38%) buildings which were classified to be in the groups with lower

724

JAIN ET AL.

Table 10. Distribution of buildings in different damage groups based on observed and expected performance scores
Observed Damage Category G1 G2 G3 G4 Predicted Damage Category Total No. of Buildings 74 83 111 2 G1 EPS 77.5 6 8 1 1

77.5

G2 EPS 45 47 34 0

60

60

G3 EPS 37.5 23 27 70 0

G4 EPS 37.5 0 1 6 1

performance scores. This indicates that only 16% buildings have not been classified conservatively. An interesting observation from Table 10 is that for more vulnerable buildings, the rate of correct prediction is higher. However, in many cases, performance of less vulnerable buildings is underestimated. This may not be a very serious limitation of the method since primary goal of rapid visual assessment is to identify buildings that need attention on highest priority. Considering the building performance is being assessed by very crude visual inspection, potentially by non-experts, without entering the building and without looking at the building plans or structural drawings, the method is quite acceptable and can serve as an effective tool in prioritizing a large building stock for seismic safety assessment.
CONSIDERATION FOR NONRESIDENTIAL BUILDINGS

To determine if there is a difference in vulnerability between residential and nonresidential buildings of Ahmedabad, data on 2566 RC-frame buildings (a subset of 3,720 building data collected by CEPT University in 2001) that had complete information on building usage x8 was analyzed and another regression analysis was performed between building usage score (=0 for residential buildings, and 1 otherwise) and OPS. The expression for performance score is obtained as

Performance Score = 77.3 + 3.2x8

4

The positive performance point obtained for nonresidential buildings may be explained by the fact that they tend to be symmetrical with more regular structural grids compared to residential buildings. Moreover, in case of nonresidential buildings, the tendency to provide smaller column width to ensure equal thicknesses of walls and columns for better interior layouts is not so pronounced. Thus, it is proposed to award +5 performance points to a building, if it is nonresidential.
CONSIDERATIONS FOR SEISMIC ZONE AND SOIL TYPE

The basic score of 85 for RC-frame buildings was assigned on the basis of the earthquake damage studies in Ahmedabad city (Equation 3), which is located in Zone III as

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

725

per the Seismic Zone Map of India, and is expected to experience a seismic shaking intensity of VII on the Medvedev Sponheuer Karnik (MSK) scale (IS 1893 2002). As per Indian seismic code expected shaking intensities in Zones II, IV, and V are VI, VIII, and IX and above, respectively. The code also classifies the damage to buildings into five categories, wherein Grade 1 corresponds to slight damage while Grade 5 damage represents total collapse. This classification of damage is similar to the one adopted in the paper (e.g., G1 is slight nonstructural damage, G5 is total collapse, etc.). A well built concrete structure is expected to undergo a damage of Grade 1 in the event of shaking of intensity VII on MSK scale, while Grade 5 damage is expected if shaking intensity is XI or above. This indicates that with the change in shaking intensity of one scale, a structure is likely to experience a damage one grade higher. Hence a modifier, representative of one grade of damage is proposed to be added to the Basic Score for a lower Seismic Zone and subtracted for a higher Seismic Zone. Thus, for a city in Zone II, the proposed Basic Performance Score would be 100 85+ 15= 100 while it would be 60 85− 15 . = 60 for a city located in Zone IV and 55 85− 15− 15= 55 for a city located in Zone V Since the surveys do not include information about the geotechnical conditions at the building sites, all the buildings of the sample surveys are assumed to have medium soil conditions. However, in many instances of rapid visual assessment work on buildings, it may be known if the building is located on soft, medium, or hard soil. In general, shaking intensity may increase by 1 and 2 points, respectively, for medium and soft soils with respect to rock (e.g., Reiter 1990). Hence, it is proposed that for rocky and for soft soil conditions, modifiers be applied representative of one grade of damage. That is, for rocky sites, the base score would receive an additional 15 points while for soft soils a penalty of 15 points would be imposed. ˘ The METU method (Sucuoglu et al. 2007) shows the basic scores for different values of PGV, which takes into account the seismic zone as well as the soil effects. The range of values, e.g., 80 to 138 for three-story buildings, indicates that the above choices for variations in seismic zone and soil type are reasonable. Thus, the proposed method for rapid visual assessment of RC-frame buildings in different seismic zones and on different soil types may be represented by Equation 1; the values of Basic Score, VS, and VSM are given in Tables 11a and 11b. SUMMARY AND CONCLUSIONS The identification of seismically vulnerable buildings and neighborhoods is a necessary first step in developing effective disaster mitigation programs for the community. Even though such assessment tools exist in other seismic countries such as U.S.A and Turkey, these are not applicable to Indian building typologies. Hence a need has long been felt to develop a methodology for rapid visual assessment of a large building stock that can be applied to Indian buildings. Since Ahmedabad was the only Indian city to have been significantly impacted during a recent earthquake (Bhuj 2001) from where data collection was possible, a sample survey of buildings was carried out in Ahmedabad on a representative sample of 270 RC-frame buildings. These buildings had been assigned different grades of damage in

726

Table 11a. Proposed method for the rapid visual assessment of RC-frame buildings in India
BASIC SCORES Soil Type Soft Medium Rock Seismic Zone Zone II 85 100 115 Zone III 70 85 100 Zone IV 55 70 85 Zone V 40 55 70 Basement Vulnerability Scores for different vulnerability parameters (VS) Number of stories Re-entrant Corners Open Story Short Column Nonresidential Use

Maintenance

+10 +10 +10

+10 +10 +10

−20 −20 −20

−10 −10 −10

−10 −10 −10

−10 −10 −10

+5 +5 +5

JAIN ET AL.

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

727

Table 11b. Vulnerability score modifiers
Non residential use Basement Number of stories Maintenance Re-entrant corners Open Story Short Column

Yes= 1, No= 0 Present= 1, Not present= 0 5 = 1, 5 = 0 Poor= 1, Moderate= 0.5, Good= 0 Yes= 1, No= 0 Yes= 1, No= 0 Yes= 1, No= 0

the immediate aftermath of the earthquake in 2001. The findings were used to understand the significance of the different vulnerability parameters by looking at the distribution of buildings with each of these vulnerability parameters across the different grades of damage. The vulnerability parameters considered were general, broad based, and easily observable from a sidewalk survey. A set of six vulnerability parameters are used in the proposed method: presence of basements, number of stories, apparent quality of maintenance, re-entrant corners, open stories, and short columns. In addition, performance scores are assigned for building usage (residential versus nonresidential), seismic zone, and soil type. A statistical analysis has been performed to develop Expected Performance Score (EPS) for buildings based on the rapid visual surveys undertaken in Ahmedabad. It accounts for the fact that the surveyed samples did not represent all the damage groups adequately, by doing multiple imputation analysis employing parametric regression method. Correctness of fit between the EPS obtained by the proposed method, as compared to their OPS, was determined to check the level of correctness of the method. It was found that for the Combined Sample, the method has predicted the damage category correctly in 46% of the buildings and within one level of incorrectness for the 88% buildings. The histogram of average absolute percentage error obtained using 1,000 bootstrap samples drawn from the Combined Sample indicates that the errors range from 17% to 25%, with a mean error of about 20%. The proposed method is based on limited data from damages in one Indian city on one building typology. This needs to be updated as well as tested as more data becomes available. Also, similar method needs to be developed for other prominent building typologies, e.g., the unreinforced masonry constructions. ACKNOWLEDGMENTS The authors are thankful to the anonymous reviewers for very valuable review comments on an earlier draft of this paper which helped them significantly revise the work. Riddhi Sheth, Bhairab Patel, and Rajesh Patel, students of the CEPT University, are acknowledged for the data they collected for the Phase I Sample and for assistance in carrying out the surveys of the Phase II Sample. Raushan Kumar Singh, a student of IIT Kanpur, and K.I. Praseeda, former Senior Project Associate at IIT Kanpur, helped in

728

JAIN ET AL.

data analysis and preparation of the manuscript. The research was financially supported by the Poonam and Prabhu Goel Foundation at IIT Kanpur and the Indian National Academy of Engineering. REFERENCES
˘ Bogaziçi University, Istanbul Technical University, Middle East Technical University, and Yildiz Technical University (BU-ITU-METU-YTU), 2003. Earthquake Master Plan for Istanbul, published by the Metropolitan Municipality of Istanbul, Planning and Construction Directorate, and Geotechnical and Earthquake Investigation Department, Turkey. Demartinos, K., and Dristos, S., 2006. First-level pre-earthquake assessment of buildings using fuzzy logic, Earthquake Spectra 22, 865–885. Federal Emergency Management Agency (FEMA 154), 1988. Rapid Visual Screening of Buildings for Potential Seismic Hazards: A Handbook (FEMA 154), Washington, D.C. Federal Emergency Management Agency (FEMA 178), 1992. NEHRP Handbook for the Seismic Evaluation of Existing Buildings (FEMA 178), Washington, D.C. Federal Emergency Management Agency (FEMA 31), 1998. Handbook for the Seismic Evaluation of Buildings—A Prestandard (FEMA 310), Washington, D.C. Federal Emergency Management Agency (FEMA 154), 2002. Rapid Visual Screening of Buildings for Potential Seismic Hazards: A Handbook (FEMA 154), 2nd edition, Washington, D.C. Hassan, A. F., and Sozen, M. A., 1997. Seismic vulnerability assessment of low-rise buildings in regions with infrequent earthquakes, ACI Struct. J. 94, 31–39. Bureau of Indian Standards (IS 13935), 2004. Draft Revision of Guidelines for Repair, Restoration, Condition Assessment and Seismic Strengthening of Masonry Buildings, under discussion in Bureau of Indian Standards, New Delhi. Bureau of Indian Standards (IS 1893), 2002. Part I: Criteria for Earthquake Resistant Design of Structures-General provisions and Buildings. Bureau of Indian Standards, New Delhi. Jain, S. K., Lettis, W. R., Ballantyne, D., Chaubey, S. K., Dayal, U., Goel, R., Goyal, A., Hengesh, J., Malhotra, P., Murty, C. V R., Narula, P. L., Saikia, C. K., Singh, M. P., and . Vatsa, K., 2001. Preliminary observations on the origin and effects of the January 26, 2001 Bhuj (Gujarat, India) earthquake, EERI Special Earthquake Report, EERI Newsletter, 35(4). Jain, S. K., Murty, C. V. R., Jaiswal, A., Shah, D., Mehta, V M., Shah, R. J., and Desai, N. S., . 2002. Post-earthquake handling of buildings: 2001 Bhuj, India, earthquake reconnaissance report, Earthquake Spectra 18 Supplement A, 297–317. Japan Building Disaster Prevention Association (JPDPA), 2001. Seismic Evaluation and Retrofit, Japan. Montgomery, D. C., Peck, E. A., and Vining, G. G., 2003. Introduction to Linear Regression Analysis, Wiley India Pvt. Ltd. New Delhi. Murty, C. V. R., Goel, R. K., and Goyal, A., 2002. Reinforced concrete structures: 2001 Bhuj, India, earthquake reconnaissance report, Earthquake Spectra 18 Supplement A, 149–185. National Research Council Canada (NRCC), 1993. Manual for Screening of Buildings for Seismic Investigation by Institute for Research in Construction, Ottawa. New Zealand society for Earthquake Engineering (NZSEE), 2006. Assessment and Improvement of the Structural Performance of Buildings in Earthquakes, Recommendations of a NZSEE Study Group on Earthquake Risk Buildings, June 2006, New Zealand.

PROPOSED RAPID VISUAL SCREENING PROCEDURE FOR SEISMIC EVALUATION

729

Rai, D. C., 2005. Review of Documents on Seismic Evaluation of Existing Buildings, IITKGSDMA-EQ03-V1.0, National Information Center of Earthquake Engineering, Indian Institute of Technology Kanpur, India, Available at www.nicee.org/IITK-GSDMA_Codes.php, last accessed 23 November 2009. Reiter, L., 1990. Earthquake Hazard Analysis: Issues and Insights, Columbia University Press, New York, 254 pp. Rubin, D. B., 1987. Multiple Imputations for Nonresponse in Surveys, John Wiley & Sons, New York. Sinha, R., and Goyal, A., 2004. A National Policy for Seismic Vulnerability Assessment of Buildings and Procedure for Rapid Visual Screening of Buildings for Potential Seismic Vulnerability, Department of Civil Engineering, Indian Institute of Technology Bombay, India, Available at www.civil.iitb.ac.in/~rsinha/Vulnerability_Assessment.pdf, last accessed 23 November 2009. ˘ Sucuoglu, H., Yazgan, U., and Yakut, A., 2007. A screening procedure for seismic risk assessment in urban building stocks. Earthquake Spectra 23, 441–458.

(Received 23 February 2009; accepted 16 December 2009