Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Quantifying unreinforced masonry seismic risk and mitigation options in Victoria, BC Paxton, Brandon 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_february_paxton_brandon.pdf [ 14.5MB ]
Metadata
JSON: 24-1.0167639.json
JSON-LD: 24-1.0167639-ld.json
RDF/XML (Pretty): 24-1.0167639-rdf.xml
RDF/JSON: 24-1.0167639-rdf.json
Turtle: 24-1.0167639-turtle.txt
N-Triples: 24-1.0167639-rdf-ntriples.txt
Original Record: 24-1.0167639-source.json
Full Text
24-1.0167639-fulltext.txt
Citation
24-1.0167639.ris

Full Text

Quantifying Unreinforced Masonry Seismic Risk and Mitigation Options in Victoria, BC  by Brandon Paxton  B.A.Sc., The University of British Columbia, 2012  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Civil Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  December, 2014  © Brandon Paxton, 2014 ii  Abstract  Despite their well-known seismic vulnerability, unreinforced masonry (URM) buildings continue to be a leading source of loss of life and property damage in earthquakes around the world.  Victoria, British Columbia is a community with substantial seismic hazard that has yet to experience a damaging earthquake and, thus, has not seen widespread interest in mitigating seismic risks.  It is also a city with a substantial stock of URM buildings, constructed around the turn of the 20th century, reasonably similar in form to those devastated by the 2010/2011 Canterbury earthquake sequence in New Zealand.  To promote seismic upgrading of URM buildings in the region, cost-benefit analyses were undertaken specifically for Victoria, considering the seismic hazard, typical pedestrian and occupant exposure, building vulnerability, and local retrofit costs.  The loss estimates are underpinned by motion-damage relationships derived in this thesis from observed damage in past earthquakes in California and New Zealand. Upgrading measures considered range from parapet bracing to full seismic upgrades consistent with local practices.  Parapet bracing is shown to have favorable cost-benefit ratios – up to 4:1 for buildings on soft soils, indicating that the expected benefits outweigh the costs by a factor of four.  Partial retrofits (eg. tying walls back to all floors/roofs, plus parapet bracing) are shown to have favourable cost/benefit ratios for buildings on soft soils (up to 1.7:1).  Full upgrades are shown to have unfavorable cost-benefit ratios (maximum 0.7:1).  In all cases, public benefits (i.e. reduced casualties) represent a significant portion of the total benefits, which provides evidence for cost sharing among building owners and the public.  The motion versus structural damage relationships derived in this study are also used in developing proposed refinements to the FEMA 154 screening methodology. New, additional score modifiers (specific to URM) are presented, including score modifiers for various levels of earthquake strengthening.  A trial application on buildings in Victoria shows that the revised scoring system is able to more effectively differentiate amongst URM buildings, making it more suitable for URM-only surveys, such as are often implemented by communities interested in quantifying and mitigating URM seismic risk.  iii  Preface  A significant portion of the material in Chapter 3 was published in a paper for the 12th Canadian Masonry Symposium: Addressing URM Seismic Risk in Victoria Canada (Paxton, et al. 2013).  I was the lead author on this paper and was responsible for compiling the material on unreinforced masonry seismic risk mitigation in the United States.  Co-author, Dr. Ken Elwood, prepared the material on URM seismic performance in the Canterbury earthquakes and co-author, Mr. Steve Barber, prepared material on the seismic retrofit incentives for heritage buildings in Victoria.  The damage statistics analyzed in Chapter 4 were provided to the author by others, who were responsible for performing the damage surveys and compiling the resulting databases.  Databases for the Loma Prieta and Northridge earthquakes were graciously provided by Mr. Bret Lizundia of Rutherford + Chekene consulting engineers, who was responsible for collecting the data.  A database for the Canterbury earthquakes was graciously provided to the author by Dr. Jason Ingham of The University of Auckland. Dr. Ingham, Dr. Michael Griffith of the University of Adelaide, and their team of researchers were responsible for collecting the data. Section 4.7 reviews the work completed by those responsible for compiling the databases, while Section 4.8 contains the new work, undertaken by the author.  The results and conclusions presented herein are the product of academic research and only reflect the views of the author.  The results and conclusions do not necessarily reflect the views of UBC, the City of Victoria, or the Victoria Civic Heritage Trust.  Any individuals using the results or conclusions presented herein do so at their own risk and the authors take no responsibility for such applications.     iv  Table of Contents Abstract ............................................................................................................... ii Preface ................................................................................................................ iii Table of Contents ............................................................................................... iv List of Tables ....................................................................................................... x List of Figures ................................................................................................... xiii Acknowledgements .......................................................................................... xvii Dedication ....................................................................................................... xviii Chapter 1 - Introduction .................................................................................... 1 1.1 Background .............................................................................................................. 1 1.1.1 Project Partners & Structure.............................................................................. 1 1.1.2 Limitations ........................................................................................................... 2 1.1.3 About VCHT ....................................................................................................... 2 1.1.4 The Climate and Context of Seismic Risk and Mitigation Measures ............. 3 1.1.5 URM Seismic Risk in Victoria ............................................................................ 3 1.2 Purpose ..................................................................................................................... 5 1.3 Scope ......................................................................................................................... 5 1.4 Impetus & Objectives for The Study ..................................................................... 6 1.5 Organization of Thesis ............................................................................................. 7 Chapter 2 - Characterizing URM Bearing Wall Buildings ............................. 9 2.1 Purpose and Scope ................................................................................................... 9 2.2 Defining Typical Unreinforced Masonry Buildings ............................................... 9 2.3 Structural Model .................................................................................................... 13 2.4 Components of a Typical URM Building ............................................................ 15 2.4.1 Unreinforced Masonry Walls ............................................................................ 15 2.4.2 Interior Framing ................................................................................................ 17 2.4.3 Floor and Roof Diaphragms ............................................................................. 17 2.4.4 Wall to Floor/Roof Connection ........................................................................ 19 2.4.5 Building Appurtenances .................................................................................... 20 2.5 Failure Modes ......................................................................................................... 22 2.5.1 Parapet, Gable, and Chimney Failure ............................................................. 23 2.5.2 Cornice Failures ................................................................................................. 24 2.5.3 Veneer Failure .................................................................................................... 24 2.5.4 Out-of-Plane Wall Failure ................................................................................ 25 2.5.5 In-Plane Wall Failure ........................................................................................ 26 2.5.6 Anchor Failure ................................................................................................... 27 2.5.7 Diaphragm Failure ............................................................................................. 27  v  2.6 Retrofitting Measures ............................................................................................ 28 2.6.1 Fall Hazards Mitigation .................................................................................... 28 2.6.2 Partial Rehabilitation ........................................................................................ 34 2.6.3 Comprehensive Rehabilitation .......................................................................... 38 2.7 Summary and Conclusions .................................................................................... 44 Chapter 3 - URM Seismic Risk Mitigation Programs ................................... 45 3.1 Purpose and Scope ................................................................................................. 45 3.2 The Impetus for URM Seismic Risk Mitigation .................................................. 45 3.3 United States .......................................................................................................... 46 3.3.1 California ............................................................................................................ 46 3.3.2 Washington ........................................................................................................ 56 3.3.3 Oregon ................................................................................................................ 57 3.4 New Zealand ........................................................................................................... 58 3.5 Canada .................................................................................................................... 60 3.5.1 Vancouver .......................................................................................................... 60 3.5.2 Victoria ............................................................................................................... 61 3.6 Comparing Victoria to Other Jurisdictions ......................................................... 66 3.7 Additional Considerations ..................................................................................... 67 3.7.1 Retrofit Design Standards ................................................................................. 67 3.7.2 Performance-Based Design................................................................................ 67 3.7.3 Enhanced Design and Construction Supervision ............................................ 68 3.7.4 Other Building Upgrades .................................................................................. 68 3.8 Conclusions ............................................................................................................. 68 Chapter 4 - Quantifying Vulnerability Through Observed Damage ............ 69 4.1 Purpose and Scope ................................................................................................. 69 4.2 The Role of Damage Observations in Engineering and Policy-Making ............ 69 4.3 Review of Systematic Damage Assessment Methods.......................................... 70 4.3.1 Wailes and Horner ............................................................................................. 71 4.3.2 ATC-13 ............................................................................................................... 71 4.3.3 ATC-20 ............................................................................................................... 75 4.4 Collecting Damage Data ....................................................................................... 76 4.4.1 Statistics from Post-Earthquake Safety Evaluations ...................................... 77 4.4.2 Statistics from Ad-Hoc Damage Surveys ......................................................... 78 4.5 Ground Motion Intensity Measurements ............................................................. 79 4.5.1 Intensity Measurements .................................................................................... 79 4.5.2 Estimating the Intensity Measurement at a Site ............................................ 83 4.6 Developing DPMs & Fragility Curves from Damage Statistics ......................... 88  vi  4.6.1 Processing The Damage Data .......................................................................... 88 4.6.2 Curve Fitting ..................................................................................................... 89 4.7 Literature Review and Summary of Available Data........................................... 92 4.7.1 Whittier 1987 Earthquake ................................................................................ 93 4.7.2 Loma Prieta 1989 Earthquake .......................................................................... 96 4.7.3 Northridge 1994 Earthquake........................................................................... 101 4.7.4 Canterbury 2010/2011 Earthquake Swarm ................................................... 106 4.8 New Damage Statistics Results .......................................................................... 115 4.8.1 Whittier 1987 Earthquake .............................................................................. 115 4.8.2 Loma Prieta 1989 Earthquake ........................................................................ 118 4.8.3 Northridge 1994 Earthquake........................................................................... 127 4.8.4 Canterbury 2010/2011 Earthquake Swarm ................................................... 128 4.8.5 Comparison of Results from North America and New Zealand .................. 129 4.8.6 Comparison of Results to Published Sources ................................................ 134 4.9 Motion-Damage Relationships for Victoria ....................................................... 137 4.9.1 Relative Vulnerability of Victoria URM Buildings ....................................... 138 4.9.2 Building Specific Damageability Effects ........................................................ 139 4.9.3 Defining Fragility Curves for Victoria ........................................................... 140 4.10 Summary .............................................................................................................. 145 4.11 Conclusions ........................................................................................................... 147 4.11.1 General Conclusions .................................................................................... 147 4.11.2 Conclusions for Victoria .............................................................................. 147 4.11.3 Future Research Opportunities .................................................................. 148 Chapter 5 - Cost-Benefit Analysis for URM Seismic Rehabilitation .......... 149 5.1 Purpose and Scope ............................................................................................... 149 5.2 Background .......................................................................................................... 149 5.3 Types of Costs and Benefits to be Considered .................................................. 150 5.4 Literature Review for Cost-Benefit Analyses .................................................... 152 5.4.1 San Francisco Study ........................................................................................ 152 5.4.2 Seattle Study .................................................................................................... 157 5.4.3 Reinforced Concrete Upgrading Study .......................................................... 160 5.4.4 New Zealand Studies ....................................................................................... 161 5.5 Development of Loss Estimate Methodology .................................................... 163 5.5.1 Seismic Hazard Model ..................................................................................... 165 5.5.2 Exposed Building Assets ................................................................................. 165 5.5.3 Motion-Damage Relationships ........................................................................ 166 5.5.4 Downtime ......................................................................................................... 176  vii  5.5.5 Occupant & Pedestrian Exposure .................................................................. 179 5.5.6 Casualty Rates ................................................................................................. 180 5.5.7 Scenario versus “Expected” Losses ................................................................. 187 5.5.8 Economic Parameters ...................................................................................... 190 5.6 Quantifying The Losses ....................................................................................... 191 5.6.1 Building Damage Losses ................................................................................. 192 5.6.2 Downtime-Related Losses ............................................................................... 192 5.6.3 Casualty-Related Losses .................................................................................. 193 5.6.4 Total Expected Losses ..................................................................................... 193 5.6.5 Other Losses Not Considered ......................................................................... 195 5.7 Expected Benefits of Strengthening ................................................................... 195 5.8 Quantifying the Costs of Seismic Upgrading ..................................................... 198 5.8.1 Parapet Bracing ............................................................................................... 198 5.8.2 Partial Retrofitting .......................................................................................... 199 5.8.3 Full Retrofiting ................................................................................................ 200 5.8.4 Other Costs Not Considered ........................................................................... 201 5.9 Cost-Benefit Analysis .......................................................................................... 202 5.9.1 Overall Cost-Benefit ........................................................................................ 202 5.9.2 Owner-Only Cost-Benefit ................................................................................ 203 5.9.3 Additional Considerations ............................................................................... 204 5.10 Sensitivity Analyses ............................................................................................. 207 5.10.1 Building Replacement Value ...................................................................... 207 5.10.2 Cost of Upgrades ......................................................................................... 208 5.10.3 Discount Rates ............................................................................................. 208 5.10.4 Time Horizon ............................................................................................... 209 5.10.5 Structural Vulnerability .............................................................................. 209 5.10.6 Length of Street Front Exposure ............................................................... 210 5.10.7 Value of a Statistical Life ............................................................................ 210 5.11 Limitations of Cost-Benefit Analysis & Expected Costs .................................. 211 5.12 Summary .............................................................................................................. 212 5.13 Conclusions ........................................................................................................... 214 5.13.1 General Conclusions .................................................................................... 214 5.13.2 Conclusions For Victoria............................................................................. 214 5.14 Future Research Opportunities .......................................................................... 215 5.14.1 Damage to Nonstructural Components ..................................................... 215 5.14.2 Downtime Estimates ................................................................................... 216 5.14.3 Improved Casualty Estimation .................................................................. 216  viii  5.14.4 Complete Database of URM Buildings in Victoria .................................. 216 5.14.5 Scenario Loss Estimates .............................................................................. 216 5.14.6 Improved Risk Assessment ......................................................................... 217 Chapter 6 - Assessing and Prioritizing URM Seismic Risk ......................... 218 6.1 Purpose and Scope ............................................................................................... 218 6.2 Existing Methodologies ........................................................................................ 218 6.2.1 California URM Prioritization Methods ........................................................ 218 6.2.2 Seattle ............................................................................................................... 221 6.2.3 FEMA 154 ........................................................................................................ 222 6.2.4 NRC Screening Methodology.......................................................................... 223 6.2.5 New Zealand IEP ............................................................................................. 228 6.2.6 Summary of Existing Procedures ................................................................... 231 6.3 Improvements to Methodology ........................................................................... 231 6.3.1 Selected Methodology for Adaptation/Improvement ................................... 231 6.3.2 Improvements .................................................................................................. 232 6.4 Summary .............................................................................................................. 236 6.5 Conclusions ........................................................................................................... 237 6.5.1 General Conclusions ........................................................................................ 237 6.5.2 Conclusions for Victoria .................................................................................. 237 6.6 Future Research Opportunities .......................................................................... 238 Chapter 7 - Inventory and Screening of Victoria's URM Buildings ........... 239 7.1 Purpose and Scope ............................................................................................... 239 7.2 Data Collection .................................................................................................... 239 7.2.1 Existing Survey ................................................................................................ 239 7.2.2 Pilot Survey of Buildings ................................................................................ 243 7.3 Conclusions ........................................................................................................... 249 7.3.1 General Conclusions ........................................................................................ 249 7.3.2 Conclusions for Victoria .................................................................................. 249 Chapter 8 - Summary and Conclusions ........................................................ 250 8.1 URM Buildings and Risk Mitigation Efforts in Victoria, BC .......................... 250 8.2 Quantifying Building Vulnerability Through Observed Damage .................... 251 8.2.1 General Conclusions ........................................................................................ 252 8.2.2 Conclusions for Victoria, BC .......................................................................... 252 8.3 Cost-Benefit Analysis for URM Seismic Rehabilitation in Victoria ................ 253 8.3.1 General Conclusions ........................................................................................ 253 8.3.2 Conclusions for Victoria, BC .......................................................................... 254 8.4 Methodologies for Assessing and Prioritizing URM Seismic Risk ................... 254  ix  8.5 Inventory and Screening of Victoria’s URM Buildings .................................... 255 8.6 Objectives Achieved............................................................................................. 255 8.7 Contributions ....................................................................................................... 256 8.8 Future Research Opportunities .......................................................................... 257 Bibliography .................................................................................................... 258  Appendix A - Costs for URM Seismic Rehabilitation in Victoria, BC ...... 268 Appendix B - Observed Damage Data .......................................................... 292 Appendix C - Fragility Curves ....................................................................... 297 Appendix D - Seismic Hazard Data ............................................................... 318 Appendix E - Cost-Benefit Analysis Calculations ........................................ 327 Appendix F - Data Collection Forms ............................................................ 340  x  List of Tables  Table 3.1 – Division 88 Compliance Timelines .................................................................... 47 Table 3.2 – Los Angeles URM Ordinance Compliance Rates ............................................ 48 Table 3.3 – Ordinance 225-92 Compliance Timelines ......................................................... 49 Table 3.4 – Measures Included in Various Strengthening Schemes ................................... 50 Table 3.5 – San Francisco URM Ordinance Compliance Rates ......................................... 52 Table 3.6 – Palo Alto URM Ordinance Compliance Rates ................................................ 53 Table 3.7 – California URM Loss Reduction Program Statistics ....................................... 55 Table 3.8 – California URM Risk Mitigation Statistics by Program Type ....................... 55 Table 3.9 – Proposed Seattle Ordinance Compliance Timelines ........................................ 57 Table 3.10 – Comparison of URM Seismic Risk Mitigation by Jurisdiction ..................... 66 Table 4.1 – Wailes and Horner Damage Scale ..................................................................... 72 Table 4.2 – ATC-13 Damage States ..................................................................................... 74 Table 4.3 – Example Damage Probability Matrix............................................................... 74 Table 4.4 – ATC-20 Placards ................................................................................................ 76 Table 4.5 – Whittier DPM – Unstrengthened ..................................................................... 95 Table 4.6 – Whittier DPM – Partially Strengthened .......................................................... 95 Table 4.7 – Whittier DPM – Fully Strengthened ................................................................ 95 Table 4.8 – Loma Prieta DPM - Unstrengthened ............................................................... 98 Table 4.9 – Soil Type vs. MDF ........................................................................................... 100 Table 4.10 – Wall Height vs. MDF..................................................................................... 100 Table 4.11 – Occupancy vs. MDF ...................................................................................... 100 Table 4.12 – Storeys vs. MDF ............................................................................................. 101 Table 4.13 – Northridge DPM - Strengthened................................................................... 103 Table 4.14 – Number of Storeys vs. MDF.......................................................................... 104 Table 4.15 – Presence of Basement vs. MDF .................................................................... 104 Table 4.16 – Horizontal Aspect Ratio vs. MDF ................................................................ 104 Table 4.17 – Vertical Aspect Ratio vs. MDF ..................................................................... 105 Table 4.18 – New Zealand URM Typologies ..................................................................... 112 Table 4.19 – Average Characteristics for San Francisco (2005 buildings) ....................... 123 Table 4.20 – Avg. Characteristics for Christchurch .......................................................... 124 Table 4.21 – Victoria Buildings vs. Other Regions ............................................................ 138 Table 4.22 – Typology MDF Adjustment Factors ............................................................ 140 Table 5.1 – Costs & Benefits for URM Seismic Upgrades ................................................ 150 Table 5.2 – Fatality Rates Used in San Francisco Study ................................................. 155 Table 5.3 – Square-Foot Costs for Retrofits in San Francisco Study Adjusted to 2014  xi  Canadian Dollars .................................................................................................................. 156 Table 5.4 – HAZUS Recovery Times .................................................................................. 176 Table 5.5 – HAZUS Loss of Function Times ..................................................................... 179 Table 5.6 – Pedestrian Densities [persons/1000ft] ............................................................. 180 Table 5.7 – HAZUS URM Fatality Rates .......................................................................... 181 Table 5.8 – Rutherford and Chekene URM Fatality Rates .............................................. 181 Table 5.9 – HAZUS URM Fatality Rates .......................................................................... 182 Table 5.10 – Rutherford & Chekene (1990) Pedestrian Fatality Rates ........................... 182 Table 5.11 – Pedestrian Fatality Rates Used in This Study ............................................ 183 Table 5.12 – Damage and Pedestrian Fatality Rates for Unretrofitted Buildings .......... 183 Table 5.13 – Damage and Pedestrian Fatality Rates for Braced Parapet Buildings ..... 184 Table 5.14 – Damage and Pedestrian Fatality Rates for Partial Retrofit Buildings ...... 184 Table 5.15 – Damage and Pedestrian Fatality Rates for Full Retrofit Buildings ........... 184 Table 5.16 – Building Damage Losses ................................................................................ 192 Table 5.17 – Tenant Relocation Expenses ......................................................................... 192 Table 5.18 – Lost Rental Income ........................................................................................ 192 Table 5.19 – Indoor Casualty Losses .................................................................................. 193 Table 5.20 – Outdoor Casualty Losses ............................................................................... 193 Table 5.21 – Present Value of Expected Benefits for Seismic Strengthening .................. 195 Table 5.22 – Costs for Parapet Bracing ............................................................................. 198 Table 5.23 – Costs for Partial Retrofits .............................................................................. 199 Table 5.24 – Costs for Full Retrofits .................................................................................. 201 Table 5.25 – Overall Benefit/Costs Results ....................................................................... 203 Table 5.26 – Owner-Only Benefit/Costs Results ............................................................... 203 Table 5.27 – Sensitivity Results for Building Replacement Value ................................... 207 Table 5.28 – Sensitivity Results for Costs of Upgrades ..................................................... 208 Table 5.29 – Sensitivity Results for Discount Rates ......................................................... 208 Table 5.30 – Sensitivity Results for Time Horizon ............................................................ 209 Table 5.31 – Sensitivity Results for Structural Vulnerability ........................................... 209 Table 5.32 – Sensitivity Results for Streetfront Exposure ................................................ 210 Table 5.33 – Sensitivity Results for VSL ............................................................................ 210 Table 6.1 – Los Angeles Prioritization ................................................................................ 219 Table 6.2 – San Francisco Prioritization ............................................................................ 220 Table 6.3 – Sonoma County Prioritization ........................................................................ 221 Table 6.4 – Seattle Prioritization ........................................................................................ 221 Table 6.5 – New Zealand URM IEP Points vs. %NBS .................................................... 230 Table 6.6 – Calculated BSH by Strengthening Level ........................................................ 233  xii  Table 6.7 – Final Score Modifiers for Strengthening ......................................................... 233 Table 6.8 – Score Modifiers for Occupant Density ............................................................ 234 Table 6.9 – Score Modifiers for Pedestrian Density........................................................... 235 Table 6.10 – Score Modifiers for Building Typology ......................................................... 236   xiii  List of Figures  Figure 1.1 – The Cascadian Subduction Zone ....................................................................... 4 Figure 1.2 – Seismic History of The Cascadia Region ........................................................... 5 Figure 2.1 – Typical URM Building and Components ....................................................... 10 Figure 2.2 – San Francisco URM Prototypes ...................................................................... 11 Figure 2.3 – Example Buildings in Victoria, BC ................................................................. 12 Figure 2.4 – URM Building Seismic Load Path .................................................................. 13 Figure 2.5 – Dynamic Model with Rigid Diaphragm & Flexible Walls ............................. 14 Figure 2.6 – Dynamic Models with Flexible Diaphragm, Walls May be Rigid ................ 14 Figure 2.7 – Courses and Wythes of Unreinforced Masonry .............................................. 15 Figure 2.8 – Stretcher and Header Courses of Unreinforced Masonry ............................... 16 Figure 2.9 – URM Building with a Veneer Wythe (Left) and Veneer Ties (Right) ........ 16 Figure 2.10 – Light Wood (Top Left), Heavy Timber (Top Right), Structural Steel (Bottom Left), and Reinforced Concrete (Bottom Right) Framing .................................. 17 Figure 2.11 – Straight Sheathing(Left), Diagonal Sheathing (Top Right), Solid Laminated Dimensional Lumber (Bottom Right) ............................................................... 18 Figure 2.12 – Floor Framing at Ledge (Left) and Pocket (Right) ..................................... 19 Figure 2.13 – Floor-to-Wall Anchors Viewed from Inside (Left) & Outside (Right) ....... 19 Figure 2.14 – Masonry Parapets Viewed From Street (Top Left) and Rooftop (Top Right); Corbelled Parapet (Bottom Left); Concrete Parapet (Bottom Right) ................. 20 Figure 2.15 – Example Photos of a Typical Lightweight Cornice ...................................... 21 Figure 2.16 – Example Photos of Chimneys & Pilasters .................................................... 22 Figure 2.17 – Example Photos of Gables ............................................................................. 22 Figure 2.18 – Failure of Parapets (Left), Gables (Right) and Chimneys (Bottom) ......... 23 Figure 2.19 – Cornice Failure ................................................................................................ 24 Figure 2.20 – Veneer Failure ................................................................................................. 24 Figure 2.21 – Illustration of Out-of-Plane Wall Failure ...................................................... 25 Figure 2.22 – Photo of Out-of-Plane Wall Failure .............................................................. 25 Figure 2.23 – Illustration of In-Plane Shear Failure ............................................................ 26 Figure 2.24 – Photo of In-Plane Shear Failure .................................................................... 26 Figure 2.25 – Photo of Anchorage Failures .......................................................................... 27 Figure 2.26 – Photo of Excessive Diaphragm Deflections ................................................... 28 Figure 2.27 – Typical Parapet Bracing Detail ..................................................................... 29 Figure 2.28 – Parapet Bracing Phases – Sheathing/Blocking Installed (Top Left), Roof Anchors Installed (Top Right), Braces Installed (Top Right), .......................................... 30 Figure 2.29 – In-Place Shear Testing .................................................................................... 31  xiv  Figure 2.30 – Retrofit Veneer Anchors ................................................................................. 32 Figure 2.31 – Miscellaneous Fall Hazard Retrofit ................................................................ 33 Figure 2.32 – NRC URM Seismic Strengthening Matrix .................................................... 34 Figure 2.33 – Tension Anchors .............................................................................................. 35 Figure 2.34 – Shear Anchor ................................................................................................... 36 Figure 2.35 – OoP Bracing, Intermediate Brace (left) and Strongback (right) ................ 37 Figure 2.36 – Photos of Intermediate Braces (left) and Strongbacks (Right) ................... 38 Figure 2.37 – Seismic Performance Levels ............................................................................ 39 Figure 2.38 – Illustration of URM Retrofitted with Steel CBF.......................................... 41 Figure 2.39 – Example Photo of URM Retrofitted with Steel CBF .................................. 42 Figure 2.40 – Illustration of Plywood Overlay Retrofit ....................................................... 43 Figure 2.41 – Diaphragm Strengthening............................................................................... 44 Figure 3.1 – Design Spectra vs. Chch February 2011 Response Spectra ........................... 59 Figure 3.2 – Damage vs. %NBS for 125 URM Buildings in Christchurch ........................ 60 Figure 3.3 – Seismic Evaluation and Retrofit Practices in Victoria ................................... 63 Figure 3.4 – Victoria and “Old Town” ................................................................................. 64 Figure 3.5 – Example TIP-Retrofitted Buildings................................................................. 65 Figure 4.1 – URM Placard Statistics Comparison from Northridge Earthquake ............. 77 Figure 4.2 – Damage Ratio Comparison from Coalinga Earthquake................................. 78 Figure 4.3 – Spatial Variation in Spectral Acceleration ...................................................... 82 Figure 4.4 – Spatial Correlation Between Ground Motion Parameters ............................. 85 Figure 4.5 – Map of Statistically-Derived PGA for Christchurch Earthquake ................. 85 Figure 4.6 – Number of Buildings vs. PGA for Canterbury Database (Initial) ................ 86 Figure 4.7 – Percentage of Buildings vs. PGA for Canterbury Database (Final) ............ 87 Figure 4.8 – Example Construction of Fragility Curve ....................................................... 89 Figure 4.9 – Beta vs. Lognormal Distribution for MDF of Canterbury Buildings ........... 92 Figure 4.10 – Instrumental Instensity Map – Whittier Narrows Earthquake ................... 93 Figure 4.11 – Instrumental Instensity Map – Loma Prieta Earthquake ............................ 96 Figure 4.12 – Level 1 Damage Data for Loma Prieta Earthquake ..................................... 98 Figure 4.13 – Instrumental Instensity Map – Northridge Earthquake ............................ 101 Figure 4.14 – MMI Breakdown for Northridge Earthquake ............................................. 102 Figure 4.15 – Instrumental Instensity Map – September (Darfield) Earthquake ........... 107 Figure 4.16 – Instrumental Instensity Map – February (Christchurch) Earthquake ..... 107 Figure 4.17 – Placard Statistics For September (Left) and February (Right) ................ 110 Figure 4.18 – ATC-13 Damage Statistics For September and February ........................ 110 Figure 4.19 – Damage vs. Number of Storeys .................................................................... 113 Figure 4.20 – Damage for Row vs. Isolated (Left) and Middle vs. End (Right) ............. 113  xv  Figure 4.21 – Damage by Building Typology .................................................................... 114 Figure 4.22 – Whittier PGV vs. Sa(1) Regression (left) and USGS Shakemaps (right) . 116 Figure 4.23 – MDF vs. Sa(1) For Whittier Earthquake ................................................... 117 Figure 4.24 – MDF vs. Sa(1) For Loma Prieta Earthquake ............................................. 119 Figure 4.25 – Selected San Francisco Buildings with Soils Conditions ............................ 120 Figure 4.26 – MMI Map of San Francisco (From USGS, 1989) ....................................... 120 Figure 4.27 – San Francisco vs. Christchurch NZ Typology ............................................ 122 Figure 4.28 – Isolated versus Row Buildings for San Francisco and Canterbury ........... 123 Figure 4.29 – Damage to San Francisco Buildings by Typology ..................................... 125 Figure 4.30 – Middle/End/Corner Buildings for San Francisco and Canterbury........... 126 Figure 4.31 – Damage to San Francisco Buildings by End/Middle/Corner ................... 127 Figure 4.32 – MDF vs. Sa(1) For Northridge Earthquake ................................................ 128 Figure 4.33 – MDF vs. Sa(1) For Canterbury Earthquakes .............................................. 129 Figure 4.34 – Comparison of Unretroffited Buildings ........................................................ 130 Figure 4.35 – Comparison of Braced Parapet Buildings ................................................... 132 Figure 4.36 – Comparison of Partially Retrofitted Buildings ........................................... 132 Figure 4.37 – Comparison of Fully Retrofitted Buildings ................................................. 133 Figure 4.38 – Published vs. Observed Damage for Unretrofitted Buildings .................... 134 Figure 4.39 – Published vs. Observed Damage for Fully Retrofitted Buildings ............. 135 Figure 4.40 – Published vs. Observed Damage for Unretrofitted Buildings .................... 137 Figure 4.41 – Basic Structural MDF vs. Sa(1) Curves (Best Estimate)........................... 141 Figure 4.42 – Basic Structural MDF vs. Sa(1) Curves (Upper Bound) ........................... 142 Figure 4.43 – Basic Structural MDF vs. Sa(1) Curves (Lower Bound) ........................... 142 Figure 4.44 – HAZUS Struct. Fragility for Unretrofitted Type D/F Victoria ................ 143 Figure 4.45 – Comparison of Damage State Fragilities by Retrofit Type ....................... 145 Figure 5.1 – Occupancy for San Francisco (left) vs. Victoria (right) URMs .................. 153 Figure 5.2 – Motion-Damage Relationships Used in San Francisco Study ..................... 154 Figure 5.3 – Components for Seattle (left) and a More Common Assumption (right) .. 159 Figure 5.4 – Flow Chart of Loss Estimation Process used in this Study......................... 164 Figure 5.5 – Component Breakdown used in this Study .................................................. 165 Figure 5.6 – MDF vs. Sa(1) Relationships (Best Estimate) .............................................. 166 Figure 5.7 – Structural Fragility Curves for Unretroffited Buildings ............................... 167 Figure 5.8 – Structural Fragility Curves for Fully Retroffited Buildings ........................ 168 Figure 5.9 – Drift-Sensitive NSC Fragility Curves for Unretrofitted Buildings .............. 169 Figure 5.10 – Drift-Sensitive NSC Fragility Curves for Retrofitted Buildings ................ 169 Figure 5.11 – Acceleration-Sensitive NSC Fragility for Unretrofitted Buildings ............ 171 Figure 5.12 – Acceleration-Sensitive NSC Fragility for Retrofitted Buildings ................ 172  xvi  Figure 5.13 – Contents Fragility for Unretrofitted Buildings ........................................... 173 Figure 5.14 – Contents Fragility for Retrofitted Buildings ............................................... 174 Figure 5.15 – Overall Motion-Damage Relationships used in This Study ....................... 175 Figure 5.16 – Initial and One-Year Cordon Extents in Christchurch CBD .................... 177 Figure 5.17 – Building Repair Time Distribution vs. Damage ......................................... 178 Figure 5.18 – Building Recovery Time Estimates Used in This Study ........................... 178 Figure 5.19 – Hazard Curve Discretization (Top) and Damage Curve (Bottom) .......... 189 Figure 5.20 – Total Expected Losses for Victoria Site Class C (Top) and E (Bottom). 194 Figure 5.21 – Relative Benefits by Strengthening Level ................................................... 196 Figure 5.22 – Breakdown of Loss Reductions by Strengthening Level (Site Class C) ... 197 Figure 5.23 – Full Retrofit Costs from Victoria Projects .................................................. 201 Figure 5.24 – Example Risk-Averse Utility Functions ...................................................... 205 Figure 5.25 – Effect of Insurance for Braced-Parapet Buildings (Site Class C) .............. 206 Figure 5.26 – Distribution of Benefits from Strengthening ............................................... 212 Figure 6.1 – High Seismicity Screening Form for FEMA 154, 2nd Edition ...................... 224 Figure 6.2 –Screening Form From NRC93 (2nd of 2 pages) .............................................. 227 Figure 6.3 – New Zealand IEP Procedure Flow Chart ..................................................... 228 Figure 6.4 – New Zealand Alternate IEP for URM Buildings.......................................... 229 Figure 7.1 – 1989 Survey Extents ....................................................................................... 240 Figure 7.2 – Breakdown of Buildings by Construction Type (329 Buildings) ................ 241 Figure 7.3 – Victoria URMs by Decade of Construction (254 Buildings) ....................... 242 Figure 7.4 – Victoria URMs by Number of Storeys (258 Buildings) ............................... 242 Figure 7.5 – Pilot Survey Area, Victoria, and “Old Town” ............................................. 243 Figure 7.6 – “Exterior Access” Data Collection Sheet ...................................................... 245 Figure 7.7 – Comparison of NZ Typology .......................................................................... 247 Figure 7.8 – Retrofit Rates in Victoria (81 Buildings) ...................................................... 248 Figure 7.9 – FEMA 154 Scores, Original and Modified (81 Bldgs).................................. 248      xvii  Acknowledgements  This research was funded jointly by the Victoria Civic Heritage Trust (VCHT) and the Natural Sciences and Engineering Research Council of Canada (NSERC) through an Industrial Postgraduate Scholarship (IPS).  I am grateful to all parties involved in this program, who had confidence in my abilities and granted me this opportunity.  In particular, I would like to thank the VCHT board of directors and the members of the Seismic Upgrading Strategy Subcomittee, including: Steve Barber, Stephanie Blazey, Gary Braun, Bruce Johnson, John Knappett, Robert Law, Pamella Madoff, Murray Miller, Tom Moore, Chris Ryzuk, Ian Sutherland, Avy Woo, and Catherine Umland.  I would especially like to thank Catherine Umland, whose passion and conviction for protecting the people and heritage of Victoria motivated me to complete this study.   I owe another very special thank-you to my supervisor, Dr. Ken Elwood.  He has shown me not only how to be an analytical, yet practical engineer, but also the importance of interfacing with the broader issues of society.  Never in my time at UBC (graduate or undergraduate) have I seen Ken give anything less than 100% for his students.  I would also like to thank my second reader, Dr. Jason Ingham.  His openness to, and genuine interest in, collaboration is a quality that all researchers and design professionals would do well to adopt. Of course, his insights were also highly invaluable, being directly responsible for a significant portion of the work that underpinned my research.  I owe a sincere thank-you to Bret Lizundia of Rutherford + Chekene, who shared damage data from his past works on the Loma Prieta and Northridge earthquakes. His work heavily inspired my desire to quantify costs and benefits of URM seismic upgrading using empirical data.  I would also like to thank the following UBC students from the Civil Engineering and Geography departments: Daniel Waine, Greg Eng, Helena Trajic, Keith Hand, Lena Zhu, Andrew Tesarowski, Hayley Kiernan, Anson Lau, and Hannah Guo.  These individuals were responsible for collecting much of the data for the building survey presented in Chapter 7.  Finally, I would like to thank my present and former classmates at UBC for making the journey enjoyable during the good times and comically memorable during the not-so-good times.  xviii  Dedication           I dedicate this thesis to my long-time girlfriend, Brandi Fredrickson, who without hesitation encouraged me to pursue graduate studies.  Completing these studies has affected both of our lives and may this thesis be a reminder to me of how fortunate I am to have such a selfless and caring person with whom to share my life.              1 Chapter 1  Introduction Unreinforced masonry (URM) was a common form of building construction in many cities around the turn of the 20th century throughout North America, and abroad.  Despite their relatively well-known poor seismic behavior, URM buildings continue to be a major source of loss of life and property damage in earthquakes, and seismic risk mitigation programs are routinely met with substantial resistance.  In virtually every case, risk mitigation measures were spurred on by emotional and political responses to losses in subject or nearby communities.  The overarching purpose of this thesis is to provide rational evidence and decision-support tools promoting URM seismic risk mitigation and, thus, public safety.  Particular focus is made on Victoria in application of the results obtained herein. However, the methodologies are applicable – and hopefully of interest – on a more general scale.  The following subsections provide some of the pertinent background information, define the scope and objectives of this study, and provide an outline of the general topics addressed herein.    1.1 Background Where present, URM buildings affect nearly every member of a community, whether it is simply an office building one walks past, an apartment building one lives in, or a coffee shop one frequents.  Moreover, seismic risk mitigation is a complex socioeconomic issue to which each community must find its own suitable solution. The following is a brief summary of the background information behind the initiation and execution of this study.  1.1.1 Project Partners & Structure This study was funded through a Natural Sciences and Engineering Research Council (NSERC) Industrial Postgraduate Scholarship (IPS).  Funding was jointly provided by NSERC and the “Industry Sponsor,” the Victoria Civic Heritage Trust (VCHT).  A third partner to the study was the consulting engineering company, Read Jones Christoffersen Ltd. (RJC), with whom the author is employed.  RJC provided office space for the author and both VCHT and RJC provided oversight and guidance for the Chapter 1 – Introduction 2 study.  Finally, the study was supervised by Dr. Ken Elwood, Professor at the University of British Columbia in Vancouver.    Other parties also provided significant assistance throughout the process of this study, including undergraduate students at The University of British Columbia, the City of Victoria’s GIS department, researchers from New Zealand, and prominent engineers from California.  Further details are provided in the Preface and Acknowledgement sections.  1.1.2 Limitations The purpose of this study was not to identify or assess individual buildings in Victoria as vulnerable and, as such, the results are not fit for that purpose.  Rather, the purpose was to develop methodologies that could be used to promote and implement URM seismic risk mitigation on a municipal basis in Victoria and in other interested communities.  The results and conclusions presented herein are the product of academic research and only reflect the views of the author.  The results and conclusions do not necessarily reflect the views of UBC, the City of Victoria, or the Victoria Civic Heritage Trust.  Any individuals using the results or conclusions presented herein do so at their own risk and the authors take no responsibility for such applications.  Individuals of varying expertise and experience were involved throughout the study and the results presented herein are based on a number of highly variable factors.  The results are a general indication only and are not a replacement for a building-specific study by suitably qualified professionals.  1.1.3 About VCHT The Victoria Civic Heritage Trust is a non-profit society in BC and a federally registered charity, whose mission statement is to “work in co-operation with government and heritage agencies to develop, administer, and financially support programs that preserve, promote, interpret, and enhance the cultural and natural heritage resources of the City of Victoria and its environs.”   The VCHT was created in 1989 by the City of Victoria as a civic vehicle to administer heritage incentive programs and initiatives on behalf of the city.  Through annual capital and operating funds from the City, the VCHT supports research, conservation, and rehabilitation efforts for legally protected commercial, industrial, and institutional heritage buildings in Victoria, through cash and tax incentives.  The VCHT board of directors is comprised of professionals with specific Chapter 1 – Introduction 3 knowledge of heritage buildings and construction: developers, contractors, city planners, professional engineers, and architects.  More recently, VCHT has become interested in developing new incentive programs to address the need for partial upgrading of URM buildings (i.e. those that will not undergo full seismic upgrading due to a re-development or change of occupancy).  VCHT’s interest in initiating this study stemmed from the need to develop substantiating material on the benefits of partial seismic upgrades such as parapet bracing.  1.1.4 The Climate and Context of Seismic Risk and Mitigation Measures Various communities around the world have made strides in mitigating URM seismic risk.  However, many more have yet to make virtually any progress whatsoever.    As shown in Chapter 3, efforts in Victoria are lacking compared to many other jurisdictions facing similar seismic hazards.  While certain parties are acutely aware of the risk, such as emergency managers and heritage planners, there has been no political movement to speak of in terms of seismic risk mitigation.  In general, the problem has yet to be openly and specifically addressed on the public stage.  This is in stark contrast to areas in California where seismic risk mitigation has been an important political topic for decades (and continues to be one).  More recently, efforts in the pacific northwest of the USA have started to develop, including the City of Seattle and the state of Oregon (see Chapter 3).  1.1.5 URM Seismic Risk in Victoria Victoria lies in the Cascadia Subduction Zone (CSZ), which is a somewhat special seismic setting.  Figure 1.1 shows the CSZ.  Many areas, such as California, are susceptible to one main type of earthquake, known as “crustal” earthquakes.  In the CSZ, there are two more potential types of earthquakes: “subcrustal” and “subduction” earthquakes.  The two former types tend to be small to moderate in size, but happen more frequently (and possibly close to any given area).  The latter is much larger (they are sometimes referred to as “mega-thrust” earthquakes); they occur much less frequently, but affect larger areas with high intensity, long duration shaking.  Crustal and Subcrustal earthquakes are more commonly observed around the world.  An example of a crustal earthquake is the 2011 Christchurch earthquake (in New Zealand) and an example of a subcrustal earthquake is the 2001 Nisqually earthquake in Chapter 1 – Introduction 4 Washington State.  An example of a subduction earthquake is the 2011 Tohoku earthquake in Japan.                    Victoria has not experienced a damaging earthquake in its history.  However, this is merely good fortune.  Potentially damaging earthquakes have occurred all around Victoria over the past 100 years, and Victoria is considered to have one of the highest seismic hazards in Canada (NRC 2010).  Figure 1.2 shows some significant earthquakes that have occurred in the general region.   To put matters in statistical terms, consider the following earthquake probabilities (Onur and Seeman 2004):  The probability of a “structurally damaging” (MMI ≥ 7) crustal or subcrustal earthquake for Victoria in the next 50 years is 21%  The probability of a “non-structurally damaging” (MMI ≥ 6) crustal or subcrustal earthquake for Victoria in the next 50 years is 53%  The probability of a M9 mega-thrust earthquake (which is expected to produce MMI=7 shaking in Victoria as per USGS, 2011a) in the next 50 years is 11%    Figure 1.1 – The Cascadian Subduction Zone  From: http://commons.bcit.ca/civil/students/earthquakes/unit1_02.htm  Note that “non-structurally damaging” refers to a level of shaking at which most existing structures would have damage to only “nonceiling, contents).  As will be seen in this study, however, unstrengthened unreinforced masonry buildings will likely experience some structural damage at this level.                    1.2 Purpose The author wishes to note that this thesishas two primary purposes: as with every other academic thesis, this study aims to contribute new knowledge to its field; existing information for practical uses by the industry sponsor.are highly valuable in improving public safety. 1.3 Scope This study was one that touched on many different aspectsthe author’s specific field of expertise.  Aimpacts of retrofit ordinances, value of a statistical life) could easily merit a much more detailed treatmentisolating one element and perfectingFigure 1.2 –Source: http://www.quaketrip.com/wpChapter 1 – Introduction -structural” items (eg. partitions, windows,  is perhaps different from most otheradditionally, it also has the purpose of gathering   Both purposes, however,  , some of which fell beyond ny one aspect (for example, socioeconomic seismic motion-damage relationships, or the monetary  it is valuable in academia, such an effort would not  Seismic History of The Cascadia Region  -content/uploads/2011/01/CascadiaSubductionZone2.jpgVictoria 5  s in that it . Although  Chapter 1 – Introduction 6 have been nearly as useful to the industry sponsor.  In general, therefore, efforts were only concentrated on a given aspect to a point where it was felt that further efforts would not significantly change the conclusions reached in this specific study.  However, a careful effort has been made to document the rationale and decisions so that future researchers may readily understand them and modify/improve upon the work contained herein.  There are many types of unreinforced masonry buildings around the world.  This thesis focuses on clay brick URM bearing wall buildings that were typically constructed in the late 19th and early 20th centuries on the west coast of North America.  Ultimately, we are interested in the performance of Victoria’s buildings, which are quite consistent with those in many other locations on the west coast of North America.  URM buildings in New Zealand are also studied and prove to be somewhat similar.    1.4 Impetus & Objectives for The Study Unreinforced masonry buildings have proven to be a leading source of loss of life and property damage again and again in earthquakes around the world.  Unfortunately, the high economic and social costs of seismic rehabilitation1 (and associated public opposition) frequently hamstring risk mitigation efforts.  The 2010/2011 Canterbury (New Zealand) earthquakes have once again raised awareness regarding the seismic vulnerability of unreinforced masonry buildings.  The highly-publicized losses from the February 2011 Christchurch earthquake included:  39 deaths attributed to URM buildings (Canterbury Earthquakes Royal Commission 2012)  The partial or full demolition of 242 of 252 heritage buildings within the city, most of which were of URM construction (CERA 2014).  While damage was severe, heritage agencies reported that the swift demolition process eliminated more buildings than was necessary (The Press 2011)  The widespread closure of the city’s central business district, a significant portion of which remained closed for over one year as officials assessed and demolished unsafe buildings, most of which were of URM construction                                        1 Throughout this this thesis, the terms “upgrading,” “strengthening,” “rehabilitation,” and “retrofitting” are variously used, all of which are found in the literature and in industry to some degree.  For the purposes of this thesis, all of the above terms can be taken as synonyms. Chapter 1 – Introduction 7 These losses show the need for effective risk mitigation strategies and this lesson has reverberated strongly among many concerned parties in Victoria, BC.  Victoria is markedly similar to Christchurch in many ways, including British colonial history, historic masonry construction, population size and demographics, importance of tourism to the economy and, most importantly, seismic hazard.  This study was initiated by the industry sponsor, the Victoria Civic Heritage Trust as part of a multi-pronged approach to URM seismic risk mitigation, including retrofit incentives, public outreach, and research.  The high-level objectives identified for this study were as follows: 1) To gain an improved understanding of the seismic risk due to unreinforced masonry buildings in “Old Town” (an area of Victoria’s downtown core) 2) To develop material for educating stakeholders about the risks 3) To develop a rational and scientific basis for future work by VCHT in encouraging seismic upgrading through existing and new incentive programs 4) To develop a rational and scientific basis for future work by city officials in developing a URM seismic risk mitigation program 5) To develop methodologies that can be extended to other communities  These objectives will be revisited in the conclusions section (Chapter 8) to discuss how they were fulfilled by this study.  1.5 Organization of Thesis Because the purpose of this thesis is to fulfill several needs of the industry sponsor and because unreinforced masonry buildings pose a complex socioeconomic problem, this thesis is quite broad in nature.  Some chapters are intended to educate those unfamiliar with URM buildings and how their risks are typically mitigated, while other chapters are highly detailed and technical in quantifying the seismic performance of URM buildings and the benefits of strengthening.  The outline for this thesis is as follows:  Chapter 2 – Characterizing URM Bearing Wall Buildings: defines the typical buildings that are the subject of this study, and their components.  Typical construction practices are discussed, as are the common deficiencies and strengthening measures.  Chapter 3 – URM Seismic Risk Mitigation Programs: reviews the efforts in several regions throughout North America and abroad and compares them to the efforts Chapter 1 – Introduction 8 to date in Victoria (and southwestern BC in general).  Retrofit ordinances, incentives, and financing schemes are discussed.  Chapter 4 – Quantifying Building Vulnerability: past earthquakes in California and New Zealand have provided valuable insight into the seismic performance of both strengthened and unstrengthened buildings.  To quantify the performance in an objective manner, we turn to statistical analyses.  This chapter reviews the available statistical data on seismic damage to URM buildings and makes use of previously collected raw data to generate new results, specifically for the purpose of this study.  The final product is motion-structural damage relationships for URM buildings of various strengthening levels (including partial strengthening, such as parapet bracing).  Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation: making use of the motion-structural damage relationships derived in Chapter 4, other existing relationships for non-structural components, cost analyses for seismic strengthening, and many other inputs, a cost-benefit analysis is performed for seismic strengthening of a prototypical URM building in Victoria and a sensitivity analysis is conducted.  Strengthening levels considered include parapet bracing, tension ties for all floors, and comprehensive seismic rehabilitation.   Chapter 6 – Assessing and Prioritizing URM Seismic Risk: reviews several different seismic screening methodologies in the literature.  Merits and flaws specific to the assessment of URM buildings are discussed.  Ultimately, modifications to an existing procedure (FEMA 154) are provided to render the assessment better suited to the subject URM buildings.  Chapter 7 – Inventory and Screening of Victoria’s URM Buildings: reviews the available data on URM buildings in Victoria and presents a variety of summary statistics, characterizing the building stock.  The need for the City of Victoria to commission a new, complete inventory of its URM buildings is highlighted and a potential methodology (including the newly modified assessment procedure) is presented.  Summary statistics for a pilot survey of 81 buildings in Victoria are presented.  Chapter 8 – Summary and Conclusions: aims to succinctly summarize the results of the study.  9 Chapter 2  Characterizing URM Bearing Wall Buildings 2.1 Purpose and Scope In this chapter, the buildings that will be the focus of our study are defined.  This will include an introduction for those unfamiliar with URM2 buildings, a discussion on their seismic behavior, a review of the commonly observed failure modes, and common measures for seismic rehabilitation.    2.2 Defining Typical Unreinforced Masonry Buildings URM buildings are present around the world and come in a great variety of forms.  In this study, however, the focus is primarily on the type of URM building that is most typical on the west coast of North America.  The word ‘typical’ is used loosely here to indicate that the structural materials and assemblies are most commonly observed.  This is an effective approach because there is relatively little variation in URM construction in this region, as it was sparsely populated before the early to mid-19th century (thus what small structures existed were built of wood) and URM construction fell out of favour in the early to mid-20th century as its seismic vulnerability became apparent.    Figure 2.1 shows some common components of a URM building.  It should be noted that the building shown in the figure is not itself “typical” – it is rather large and complex so as to include all the possible components (and in some cases, alternate components, such as components #4 and 5).  As will be shown later in this chapter, smaller two and three storey buildings, often constructed in rows immediately adjacent one another, are more common in many areas, including Victoria.  The essential characteristics of a typical west coast URM building are:  Bearing walls constructed from clay bricks and lime-based mortars, composed of multiple wythes  Stone rubble foundations                                       2 URM buildings other than clay brick bearing wall buildings are rare in Victoria and this study focuses essentially solely on clay brick bearing wall buildings.  Throughout this thesis, the term “URM” refers specifically to clay brick URM bearing wall buildings, unless otherwise noted. Chapter 2 – Characterizing URM Bearing Wall Buildings 10  Timber floor/roof structures, composed of board sheathing and rough sawn lumber joists (occasionally supported on heavy timber interior framing)  Open fronts (i.e. little to no wall) are common for row buildings                       Based on the literature (ABK 1981, Lizundia, Dong and Holmes 1993, Rutherford & Chekene 1997) and discussion with practicing engineers and researchers, it is understood that this type of URM construction is common throughout western North America.  Of course, there are many variations in structural form fitting this broad description and the structural form of a URM building is often quite indicative of its original use.  Rutherford and Chekene (1990) defined 15 prototype URM buildings for San Francisco, U.S.A. based largely on original use.  The illustration from the document is reproduced here, as Figure 2.2.  This essentially shows the variety of URM buildings that is commonly encountered on the west coast of North America, including Victoria. All would fit the aforementioned broad description.  Figure 2.3 shows photos of a few example buildings in Victoria.  Figure 2.1 – Typical URM Building and Components  Modified From: FEMA, 2002a  Chapter 2 – Characterizing URM Bearing Wall Buildings 11                                         Figure 2.2 – San Francisco URM Prototypes From: Rutherford and Chekene, 1990  Chapter 2 – Characterizing URM Bearing Wall Buildings 12     Figure 2.3 – Examples of URM Buildings in Victoria, BC  Chapter 2 – Characterizing URM Bearing Wall Buildings 13 2.3 Structural Model While seemingly simple in form, the seismic behavior of URM buildings is substantially different from typical modern buildings and, perhaps, more complex.  In modern buildings constructed of concrete or steel, a greater proportion of the mass of the structure is located at the floor levels, due to the presence of floors slabs.  These floor slabs, which act as diaphragms, often possess much more flexural stiffness than do the vertical elements of the seismic force resisting elements (eg. moment frames, shear walls, braced frames).  As such, it is typical in seismic design and analysis of new buildings to represent the dynamic behavior of the building with a “lumped mass” model, in which all dynamic response amplification is assumed to occur in the vertical components of the SFRS.  The diaphragms are assumed to be rigid and thus experience no further dynamic amplifications.  This is the basis of the seismic design provisions for new buildings in modern codes such as the National Building Code of Canada (NRC 2010), the International Building Code (ICC 2012) and the New Zealand Standard NZS 1170.5-2004 (NZS 2004).  Many URM buildings, however, are not accurately represented by this model.  They tend to have very light and flexible wood diaphragms, typically consisting of plank sheathing (which is even more flexible than modern blocked plywood diaphragms) and the exterior walls are quite heavy and stiff.  The result is that URM buildings are essentially the opposite of most modern buildings in terms of their distribution of mass and stiffness: the “side walls” (also known as “end walls” as per Figure 2.4) are often assumed to be extremely stiff and, thus, experience negligible dynamic amplification; the diaphragms are more flexible and experience significant dynamic response amplification.    Figure 2.4 – URM Building Seismic Load Path  (From Bruneau, 1994)  Chapter 2 –This is the basis of many URM seismic assessment procedures, such as Evaluation Guidelines for Existing BuildingsBuilding Code (ICC 2012a).  structural models.                          Provided sufficient connections exist, the seismic load path is idealized as such:1) The ground motion excites the URM end walls2) The end walls transmit the motion to the wood diaphragms3) The diaphragms transmit the motion 4) The head walls respond dynamically in their outMany unretrofitted buildingsseismic load path and this is the greatest seismic vulnerabiliSeveral post-earthquake reconnaissance reports have cited a lack of sufficient connection between diaphragms and walls as a common reason for damage and collapses 1988, Bruneau 1990, LATF 1994, Ingham and Griffith 2011b)Figure 2.5 – Dynamic Model with Rigid Diaphragm & Flexible Walls Figure 2.6 – Dynamic Models with Flexible Diaphragm (Walls M  Characterizing URM Bearing Wall Buildings Canada’s Seismic  (NRC 1992) and the International Existing Figure 2.5 and Figure 2.6 below illustrate the two different   (and provide restraint) to the head walls-of-plane direction  lack sufficient connection at one or more points in the ty of URM. From  CCMPA, 2009 ay be Rigid )From: CCMPA, 2009 14    buildings.  (Deppe    Chapter 2 – Characterizing URM Bearing Wall Buildings 15 2.4 Components of a Typical URM Building Having defined the form of a typical URM building and formulated a structural model, some of the more common components will now be examined.  2.4.1 Unreinforced Masonry Walls URM walls are composed of some form of unit block (eg. clay bricks, stones, concrete block) and, typically, a mortar (although dry stacked masonry also exists).  In this study, we restrict our focus to URM buildings with walls constructed of clay bricks and mortar.  Bricks used in building construction were typically fired at local plants and mortar composition was either sand-lime or a blend of lime and portland cement with sand, dependent upon the era of construction: before the turn of the 20th century, sand-lime mortars were used almost exclusively for URM building construction (ASTM 2003).  Note that the primary difference between sand-lime and sand-lime-cement mortars is that the latter cures quicker and achieves higher strengths.  While the higher strength may be desirable from a seismic point of view, there are a variety of reasons why repointing (a term for replacing the mortar) should use a similar mortar to that of the original construction.  Among these, is the fact that a strong, stiffer mortar may attract too much stress which could damage the adjacent masonry (ASTM 2003); such mortars also tend to be less permeable, which could lead to increased water transmission through the bricks, which is undesirable.  The walls are constructed by laying the bricks atop one another in a bed of mortar.  Each vertical layer of bricks is called a “course” and the mortar between the courses is called the “bed joint.”  Mortar is also applied between the ends of the bricks in what is called a “head joint.”  Structural walls are generally several layers (also known as “wythes”) wide.  Between these wythes, is another mortar joint called a “collar joint.”  Figure 2.7 shows courses and wythes within a wall.           Figure 2.7 – Courses and Wythes of Unreinforced Masonry From: http://www.tpub.com/engbas/7-32.htm  Chapter 2 – Characterizing URM Bearing Wall Buildings 16 The bricks can be laid in various orientations.  The typical orientation (shown in Figure 2.7) is a known as a “stretcher” orientation.  In a multi-wythe wall, both wythes are laid primarily in the stretcher orientation and the wythes are tied together by laying a course bricks in the “header” position every five or six courses.  Figure 2.8 shows stretcher and header courses.         Walls may also have an additional wythe that is not tied in with header courses and may have an air gap.  This is known as a “veneer wythe” – they are considered non-structural and are only connected to the structural wall by metal ties at some regular spacing.  In modern buildings, the ties are designed to transfer the seismic loads of the veneer back to the structural wall.  In URM buildings, however, the ties are generally quite minimal and may be severely deteriorated due to corrosion; thus, they are often quite vulnerable to seismic damage and pose a significant fall hazard.  Veneer wythes can be identified by the lack of header courses.  For URM buildings, veneer wythes are typically an architectural feature that is only present on main facades (if at all).  Figure 2.9 show a photo of a building with and without veneer wythes as well as a photo of veneer ties exposed at an area of excavated bricks.  Note the absence of header courses in the veneer wythe.              Figure 2.8 – Stretcher and Header Courses of Unreinforced Masonry From: http://www.tpub.com/engbas/7-32.htm  Figure 2.9 – URM Building with a Veneer Wythe (Left) and Veneer Ties (Right) Photo credit: Read Jones Christoffersen Ltd.  Chapter 2 – Characterizing URM Bearing Wall Buildings 17 2.4.2 Interior Framing Wood is typically the material of choice for interior framing in URM buildings.  Depending on the size and original use of the building, this may include light framing such as studs and joists (similar to a modern building), or it may include large, heavy timber post and beam type construction.  The latter is typically found in larger structures that originally served as warehouses or some other industrial purpose.  Occasionally, interior framing will also consist of structural steel or reinforced concrete beams and/or columns.  Figure 2.10 shows various interior framing elements. Of course, interior walls constructed of URM or other archaic materials are possible.                          2.4.3 Floor and Roof Diaphragms Wood framing is by far the most common type of floor and roof construction for URM buildings on the west coast of North America.  It is acknowledged that other types are occasionally encountered, but this study will focus almost exclusively on flexible wood diaphragm type URM buildings. Figure 2.10 – Light Wood (Top Left), Heavy Timber (Top Right), Structural Steel (Bottom Left), and Reinforced Concrete (Bottom Right) Framing Photo credit: Read Jones Christoffersen Ltd. Chapter 2 – Characterizing URM Bearing Wall Buildings 18 Wood floor and roof structures in URM buildings are typically comprised of some form of sheathing material atop rough sawn wood joists, spaced at 12”-24” on centre.  Original sheathing material is often flat boards, about 6” wide and ½” to 1” thick, laid with small gaps between each board.  The boards may be laid perpendicular to the joists (“straight sheathing”), or at an angle (“diagonal sheathing”) and are typically fastened with two nails at every second or third joist.  The orientation is significant because it affects the strength and stiffness of the diaphragm.  Occasionally, two layers of sheathing may be present.  Joist size varies depending on the span, typically falling between 2”x6” and 3”x12.”  Panel sheathing such as plywood may be encountered, typically installed as part of past renovations.  In buildings with heavy timber framing, floors may be constructed of rough sawn lumber laid immediately next to one another to form a solid surface.  Figure 2.11 provides photos of various floor construction types.                          Figure 2.11 – Straight Sheathing (Left), Diagonal Sheathing (Top Right), Solid Laminated Dimensional Lumber (Bottom Right) Photo credit: Read Jones Christoffersen Ltd. Chapter 2 – Characterizing URM Bearing Wall Buildings 19 2.4.4 Wall to Floor/Roof Connection Another key component of diaphragms is their connection to the URM walls.  The two most common conditions are for joists/beams to sit on ledges (created by reducing the wall thickness by one wythe) or in pockets formed into the wall lay-up.  In both cases, the floor framing may or may not be anchored to the wall to resist lateral loading (eg. from wind and earthquake) as part of the original construction; the frequency of anchorage varies by region, presumably due to local construction practices of the day.  In any case, the anchorage (if present at all) is generally not capable of resisting the loads imposed by any modern seismic codes.  Figures 2.12a shows ledge and pocket details and Figure 2.12b shows ties from original construction.                             Figure 2.12a – Floor Framing at a Ledge (Left) and a Pocket (Right) Figure 2.12b – Floor-to-Wall Anchors Viewed from Inside (Left) & Outside (Right) Photo credit: Read Jones Christoffersen Ltd. Chapter 2 – Characterizing URM Bearing Wall Buildings 20 2.4.5 Building Appurtenances Common building appurtenances of interest for URM buildings include parapets, cornices, corbels, and chimneys.  For the purposes of this study, we are essentially interested in any component that poses a fall hazard.  The following sections discuss the individual components.  2.4.5.1 Parapets Parapets are the portions of URM exterior walls that extend above the roof.  They serve many functions, but were originally provided to prevent fire from spreading from rooftop to rooftop (Rutherford & Chekene 1997).  As URM construction became more popular in the early 1900’s, parapets became an important architectural feature and character defining element of a building, and often were often large and ornately decorated, particularly on the main facades.  Height varies from just a few inches above the roof surface to several feet and the parapet may project straight up from the wall below or be corbelled outwards in a decorative fashion.  Occasionally, concrete parapets will be found on URM buildings.  Figure 2.13 shows examples photos of parapets.                      Figure 2.13 – Masonry Parapets Viewed From Street (Top Left) and Rooftop (Top Right); Corbelled Parapet (Bottom Left); Concrete Parapet (Bottom Right) Chapter 2 – Characterizing URM Bearing Wall Buildings 21 2.4.5.2 Cornices Cornices are often constructed from lightweight materials such as wood and sheet metal, although terra cotta (which is heavy) is also common.  Lightweight cornices obviously pose a significantly lower fall hazard, but the consequences of a large, heavy cornice falling to the streets can be similar to those of a parapet.   Figure 2.14 provides example photos of cornices.                    2.4.5.3 Chimneys/Pilasters Chimneys and pilasters are common elements on URM buildings that are constructed of URM and pose significant fall hazards to life-safety.  Because all four sides are exposed to weather, there is often a greater fraction of the mortar in a deteriorated condition.  For chimneys, the portion projecting from the roof may be just one wythe thick and the result is that they are highly vulnerable to earthquakes.  Pilasters may appear similar to chimneys, but are generally solid sections.  It should also be noted that these elements are not always located at the perimeter of a building and, thus, may not be readily visible from street level.  Figure 2.15 provides example photos of chimneys and pilasters.   Figure 2.14 – Example Photos of a Typical Lightweight Cornice Chapter 2 – Characterizing URM Bearing Wall Buildings 22            2.4.5.4 Gables Buildings with pitched roofs often employ gable ends, resulting in an increased wall height.  There is also typically little dead load or restraint from the roof, as the roof framing spans parallel to the gable.  As a result, gables are also highly vulnerable elements.  Although considered part of the structural wall (and thus not a true appurtenance), it is identified in this section because it poses a fall hazard similar to the other elements herein.  Figure 2.16 shows examples photos of gable ends.              2.5 Failure Modes Because URM buildings were constructed long before the implementation of seismic design provisions and, in some cases, building codes whatsoever, there are several possible failure modes under seismic loading that were not addressed in their design and construction.  The following sections present the commonly observed failure modes. It draws primarily on the work of others (Rutherford & Chekene 1990, Ingham and Griffith 2011a, Ingham and Griffith 2011b) and is included as part of the practical information Figure 2.15 – Example Photos of Chimneys & Pilasters Figure 2.16 – Example Photos of Gables Chapter 2 –gathering for the industry sponsor for the purposes of educating various stakeholders about the risk.  Also, becausetechnical personnel, discussion on the failure modes is quite limited.  Readers interested in a more technical analysis should refer to the original  2.5.1 Parapet, Gable, and ChimneyParapets, gables, and chimneys components of URM buildings1993, Rutherford & Chekene 1997)passed the roof is in a cantilever configuration and is subject to dynamic loading from its own inertia and the diaphragm.  that exceeds the tensile capacity of the masonry, a crack will occur and the parapet will begin to rock.  Depending on the remainder of the excitation, the parover or remain dynamically stable and eventually come to rest.chimneys is similar.  All such failures pose lifeand sidewalks, thereby hampering rescue efforts.gable, and chimney failures.                      Figure 2.17 – Failure of Parapets (Left), Gables (Right) and Chimneys (Bottom) Characterizing URM Bearing Wall Buildings  the primary purpose of this section is to educate nondocuments and related literature. Failure are widely accepted to be among the most vulnerable  (Deppe 1988, Bruneau 1990, Lizundia, Dong and Holmes .  In the case of parapet failure, the wall If the flexural load on the cantilever results in a stressapet may topple   The failure of gables -safety threats and may also block roads  Figure 2.17 provides photo  From: Ingham & Griffith, 2011a 23 - projecting  and s of parapet,  Chapter 2 –2.5.2 Cornice Failures Cornices fail due to insufficient anchorage to the building face, either because of the original construction or deterioration of the masonry at the anchorage.  are constructed of lightweight material, the life-safety threat is diminished.                  2.5.3 Veneer Failure As mentioned in Section 2.4.1URM buildings often consists of either corrugated metal tiesmiscellaneous steel ties.  The seismic resistance of the ties is further reduced due to low strength, high-lime content mortar typically used in URM buildings.  provides a photo of a veneer failure.   Characterizing URM Bearing Wall Buildings Mand thus the seismic demands are reduced and Heavier cornices pose hazards similar to parapets., anchorage of veneers to the structural backing wythes , long nails  Figure 2.18 – Cornice Failure From: Rutherford and Chekene, 1990  Figure 2.19 – Veneer Failure From: Ingham and Griffith, 2011b  24 any cornices  in , or other Figure 2.19 Chapter 2 – Characterizing URM Bearing Wall Buildings 25 2.5.4 Out-of-Plane Wall Failure Under shaking of increased intensity or duration, entire walls may topple outwards.  This is due to either a lack of anchorage or an excessive flexural stress in the wall as it bends between floors.  The life-safety hazard to pedestrians is increased due to the increased volume of debris falling outwards.  Often, the interior walls and columns are sufficient to support the floors and the remainder of the structure is left standing, while the wall lies on the ground in ruins.  As pointed out by Ingham and Griffith (2011b), the hazard is greater to the public on the street than the occupants of building.  Figures 2.20 and 2.21 provide an illustration and a photo of an out-of-plane (OoP) failure.                    Figure 2.20 – Illustration of Out-of-Plane Wall Failure From: Rutherford and Chekene, 1990  Figure 2.21 – Photo of Out-of-Plane Wall Failure From: Ingham and Griffith, 2011b  Chapter 2 – Characterizing URM Bearing Wall Buildings 26 2.5.5 In-Plane Wall Failure Significant damage can also result from in-plane seismic actions.  The most commonly observed in-plane failure is an ‘X’ crack shear failure; other in-plane failures include bed joint sliding shear and toe crushing.  In general, in-plane failures pose less of a life safety threat.  Figures 2.22 and 2.23 provide an illustration and a photo, respectively.  Note that although the life-safety threat is considered lower for such failures, it is still significant, especially if the damage is widespread.  The building shown in the photo below was ultimately demolished due to safety concerns.                              Figure 2.22 – Illustration of In-Plane Shear Failure From: Rutherford and Chekene, 1990  Figure 2.23 – Photo of In-Plane Shear Failure From: Ingham & Griffith, 2011b  Chapter 2 – Characterizing URM Bearing Wall Buildings 27 2.5.6 Anchor Failure A variety of anchorage failures are possible, depending on the details of the connection.  For through-bolted anchors, this could include yielding/rupture of the steel components, punching shear failure of the masonry wall around the plate, or failure of the wood joist/blocking to which the anchor is attached.  For adhesive-type anchors, there is the additional potential failure mode of pullout/breakout of the anchor and the surrounding masonry.  Figure 2.24 provides photos of anchorage failures.                       2.5.7 Diaphragm Failure Recall that URM floor and roof diaphragms are typically constructed of board sheathing (straight or diagonal) which is even more flexible than modern wood diaphragms.  Given the low stiffness of the diaphragms, diaphragm “failure” is most commonly associated with excessive deflections, which can lead to collapse of the out-of-plane walls (as the walls lean in/outwards with the deflection).  Figure 2.25 shows an aerial view of a building that suffered excessive diaphragm deflections.  Observe that the majority of the debris has fallen in the direction of the diaphragm deflection (i.e. into the building on Figure 2.24 – Photo of Anchorage Failures From: Ingham & Griffith, 2011b  Chapter 2 – Characterizing URM Bearing Wall Buildings 28 one side and away from the building on the other).  Note that seismic assessment standards typically limit the aspect ratio of diaphragm to 3 or 4 to 1.              2.6 Retrofitting Measures Having identified the various manners in which URM buildings tend to fail, we now proceed with identifying manners in which the weaknesses are commonly addressed.  Again, this section draws almost exclusively on the work of others (Rutherford & Chekene 1990, 1997, FEMA 2006, Ingham and Griffith 2011b) and the goal is to provide a level of detail suitable for non-technical personnel.  The discussion focuses on the various levels of strengthening that are commonly employed.  A limited presentation is made on specific details.  Refer to the aforementioned documents for further details.  For the purposes of this study, three levels of strengthening were identified: 1) Fall hazards mitigation (eg. parapet bracing) 2) Partial seismic rehabilitation 3) Comprehensive seismic rehabilitation  These are general categories that are refined further for engineering purposes. The following sections discuss these three levels of strengthening.   2.6.1 Fall Hazards Mitigation As shown in section 2.5, parapets, cornices, chimneys, gables, decorative elements, other appurtenances (awnings, mechanical equipment) are potential fall hazards during a seismic event.  They tend to be the most vulnerable components and fail at the lowest intensities of shaking.  Moreover, they are isolated items and often require only exterior access for strengthening work.  As such, they are generally considered the first step in Figure 2.25 – Photo of Excessive Diaphragm Deflections Source: Canterbury Maps (http://www.canterburymaps.govt.nz/, retrieved May 2014) Chapter 2 – Characterizing URM Bearing Wall Buildings 29 mitigating URM seismic risk.  Indeed, some cities in the United States have had parapet bracing ordinances in place since the middle of the 20th century.    2.6.1.1 Parapets, Chimneys, and Gables Typical Details Parapets are the most commonly addressed fall hazard item for URM buildings.  The typical solution is to brace the parapets with structural steel and provide roof-wall anchors.  Another solution is to replace the parapet with a reinforced concrete cap beam.  Figure 2.26 shows a typical parapet bracing detail.                     The benefits of parapet bracing are twofold:  The fall hazard is substantially reduced (although not eliminated)  The wall is now secured at the roof and base, which reduces the likelihood of complete OoP wall collapse, particularly for 1-2 storey buildings  As will be seen in Chapter 4, parapet bracing can significantly reduce the overall damage to the building for low intensity shaking.  However, it should be remembered that the remainder of the seismic load path within the building is often incomplete. Figure 2.26 – Typical Parapet Bracing Detail From: FEMA, 2006  Chapter 2 – Characterizing URM Bearing Wall Buildings 30 Cost & Disruption One great merit of parapet bracing is its relatively low cost.  A typical value3 is about $300/lin. ft. for thru-bolted connections (see Appendix A for typical costs).  As stated by Rutherford and Chekene (1990), there is little disruption to building occupants.  For a typically-sized building (say about 30 feet of parapets along a streetfront), the work can be completed within about a week.  Noise-generating work would include drilling/coring of the parapets (for short periods) and operation of power tools for removing and installing lumber.  If desired, the work can be completed almost exclusively from the exterior, although this necessitates removal of all roofing and sheathing within 4-6 feet of the parapet.  Alternatively, the scope of re-roofing can be minimized at the cost of increased interior access.  Figure 2.27 provides photos at various stages of a typical parapet bracing project.  Note the bottom left picture in which a “pitch pocket” has been provided to seal the roof penetration.  This is no longer typical practice as improved methods have been devised.                                                            3 Throughout the body of this thesis, costs are presented in 2014 Canadian Dollars (CAD). Figure 2.27 – Parapet Bracing Phases – Sheathing/Blocking Installed (Top Left), Roof Anchors Installed (Top Right), Braces Installed (Top Right),  Finished Roof (Bottom Right) Source: http://www.structuralrenovations.com Chapter 2 – Characterizing URM Bearing Wall Buildings 31 Note that if parapets are sufficiently short, the angled bracing may not be required.  As per the Canadian (NRC 1992) and American (ICC 2012a) standards, parapets of height-thickness ratios of 1.5:1 or less need not be braced (although roof anchors are still required).  This limit is relaxed to 2.5:1 and 4:1 for moderate and low seismic zones, respectively.  Chimneys can be treated in a similar fashion, although the work is more localized.  Gable walls are essentially walls of increased height.  The top portion of the wall should be braced in a method similar to the above detail for parapets.   Anchors are typically also provided at the base of the gable.  Items Sometimes Overlooked Maintaining the integrity of the roof is an important aspect that is sometimes overlooked.  Owners should confirm that the consultants/contractors involved are knowledgeable in this area and that roofing is not simply “left to others” by both parties.  Another key issue sometimes overlooked is the condition of the masonry.  In most post-earthquake reconnaissance reports, deteriorated/poor quality mortar is correlated with parapet (and other URM) collapses.  There are several methods of assessing the condition of the mortar, from a simple scratch test to (destructive) in-place shear testing.  Such testing may represent a slightly or modestly increased cost (say a few hundred to a thousand dollars), but is vital to ensuring the strengthening meets its intended objective.  Figure 2.28 provides photos of in-place shear testing, which involves determining the lateral load necessary to initiate shear failure in the bed joints.  The process and interpretation of results is commonly discussed in the aforementioned retrofit standards for URM.            Figure 2.28 – In-Place Shear Testing Photo credit: Read Jones Christoffersen Ltd. Chapter 2 – Characterizing URM Bearing Wall Buildings 32 2.6.1.2 Veneers Typical Details Veneer courses should be identified and addressed as part of any strengthening (including parapet bracing).  Few life safety benefits will be realized if the wall remains standing but sheds its veneer onto the street.  Some retrofit standards (ICC 2012a) accept existing corrugated metal ties (as shown in Figure 2.9), provided they meet certain requirements.  Of course, the difficulty lies in confirming the presence and condition of these ties, which requires removal of sections of the veneer.  This is valuable on larger projects where retrofitting the veneer of an entire building would be extremely costly.  For small areas such as a single parapet, it may be more economical to simply provide new ties.    Retrofit ties are typically helical anchors screwed into the backing wythes or steel dowels grouted-in with adhesive.  Figure 2.29 shows a typical detail for epoxy veneer anchors.  Note that the ties provided in the mortar joints can be hidden by repointing.                   Cost & Disruption Veneer anchors are somewhat costly to install, largely due to the cost of access (see Appendix A for costs); a typical unit cost for epoxy dowels is about $20/sq.ft. of wall surface.   Figure 2.29 – Retrofit Veneer Anchors From: FEMA, 2006 Chapter 2 – Characterizing URM Bearing Wall Buildings 33 For a small area or a 1-storey building a moveable scaffold platform will suffice.  For a larger area, however, the entire building elevation may need to be scaffolded for a period of weeks to months.  Interior access can be avoided, but regular drilling of the masonry will produce noise and vibrations that are disruptive to the occupants.   Items Sometimes Overlooked Veneer anchors are typically quite sensitive to installation workmanship.  It is important that the installers are familiar with the requirements specific to the product being used (manufacturers often provide field training for free) and quality control/assurance programs should be implemented.    Some products also specify a minimum quality mortar which should be confirmed by testing.  Note that the exterior of a building may have been repointed with a shallow layer of harder, modern mortar; this must not be erroneously taken as being representative of the entire wall.   2.6.1.3 Other Fall Hazards Typical Details Other fall hazards include lightweight cornices, and other decorative elements on the building face.  The fall hazards are mitigated by providing anchorage into the backing wythes.  Figure 2.30 shows a typical detail.                Figure 2.30 – Miscellaneous Fall Hazard Retrofit From: Rutherford & Chekene, 1990 Chapter 2 – Characterizing URM Bearing Wall Buildings 34 Cost & Disruption Cost and disruption are similar to that for veneer anchorage, although the work is somewhat more targeted.    2.6.2 Partial Rehabilitation Partial rehabilitation is the next step beyond fall hazard mitigation and is specifically geared towards improving the overall behavior of the building and reducing the potential for collapses.  In this study, we define partial rehabilitation to include, as a minimum, tension anchors at all floors and roofs (in addition to parapet bracing).  It may also include shear anchors, and out-of-plane strengthening for slender URM walls.  However, partial rehabilitation measures do not include new elements to increase the in-plane resistance on the walls.    For Canadian practice, the most relevant design standard is the Guidelines for Seismic Evaluation of Existing Buildings, as published by the National Research Council of Canada in 1992 (NRC 1992).  This standard explicitly recognizes partial rehabilitation by providing a matrix of measures to be included as a function of the seismic hazard, as shown in Figure 2.31.    Figure 2.31 – NRC URM Seismic Strengthening Matrix From: NRC 1992 Chapter 2 – Characterizing URM Bearing Wall Buildings 35 Note that the seismic zones are not compatible with the current edition of the National Building Code of Canada, but effectively areas of low seismic hazard are required to perform little to no strengthening and the required scope increases with increasing seismic hazard.  Other examples of partial rehabilitation measures have been incorporated into seismic retrofit ordinances for various cities in the United States, such as San Francisco, Los Angeles, and Seattle; see Chapter 3 for further discussion.  The sections below discuss the individual measures, typical of partial rehabilitation.  2.6.2.1 Tension Anchors Typical Details Tension anchors may be either through-bolted or adhesive types and can be installed with access from above or below the floor.  The purpose of the tension anchors is to make the OoP walls span between the floors, substantially reducing the potential for collapse.  Figure 2.32 shows typical details for tension anchors installed from below and above the floor.               Cost & Disruption Tension anchors are a reasonably low cost item, with a typical value of about $10/sq.ft. of floor area for thru-bolted anchors (see Appendix A).  Note that this is intended to be applied to projects in Victoria and is in 2014 Canadian dollars.  Of course, this cost is highly sensitive to the finishes, and here it is assumed that expensive finishes such as tile floors need not be removed/replaced.  Adhesive anchors tend to be more expensive because the epoxy (or other adhesive) is costly and because the tensile capacity of the anchors is lower (thus more anchors are needed). Figure 2.32 – Tension Anchors From: FEMA, 2006 Chapter 2 – Characterizing URM Bearing Wall Buildings 36 The disruption is moderate, as interior access is required. As stated by Rutherford and Chekene (1990), noise and dust would be generated by drilling of the masonry.  Building contents within about 4-6 feet of the wall will need to be relocated in many instances.  However, this would only be for a short period of time – perhaps a day or two, as the work proceeds from room to room.  For residential buildings, this type of work is ideally suited to be completed during tenant turnover.  2.6.2.2 Shear Anchors Typical Details Shear anchors must involve some form of grout or adhesive.  Grout is a less costly material, but installation is more taxing, because it must be mixed properly and the substrate must be properly prepared (eg. saturated so as not to absorb water from the grout). According to Dizhur et al. (2013), the seismic performance is similar.  In practice, grouted anchors are somewhat antiquated and adhesive anchors are the standard.  Note that they can serve as tension anchors if properly installed (eg. drilled at an angle, with hold-down and blocking for tension).  Figure 2.33 shows a typical shear anchor.                 Cost & Disruption Cost & disruption considerations are similar to those for tension anchors.  An important decision for the owner performing partial rehabilitation is whether to install anchors for just tension, or for both shear and tension.  Tension anchors can simply be thru-bolted and so are less costly.  However, the combined cost of installing tension anchors and Figure 2.33 – Shear Anchor From: FEMA, 2006 Chapter 2 – Characterizing URM Bearing Wall Buildings 37 subsequently also installing shear anchors (as part of a more comprehensive upgrade) would obviously exceed the cost of simply having installed combination anchors to begin with.   2.6.2.3 Out-of-Plane Strengthening Typical Details With tension and shear anchors in place, the walls span vertically between the floors under seismic loading.  However, they can crack, rock, and subsequently collapse if the shaking is of sufficient intensity and duration.  Out-of-plane strengthening addresses this failure mode.  The most common solution is structural steel “strongbacks” installed at regular spacing along the wall, which act as a splint to help prevent the out-of-plane cracking and rocking.  Another common solution is an “intermediate brace” which helps to reduce the unsupported height of the wall.  Figure 2.34 provides a detail for each of these solutions and Figure 2.35 provides photos.  Other solutions could include shotcrete overlays and reinforced/post-tensioned centercores (see FEMA 2006).                    Figure 2.34 – OoP Bracing, Intermediate Brace (left) and Strongback (right) From: FEMA, 2006 Chapter 2 – Characterizing URM Bearing Wall Buildings 38                    Cost & Disruption Providing out-of-plane strengthening (in addition to parapet bracing, and shear and tension anchors) represents a significant increase in cost.  A typical unit cost (in Victoria in 2014 dollars) is about $18-25/.ft. (see Appendix A).  Disruption is also significantly increased: as large structural steel components are involved, the contractor will require somewhere to store them on site.  Removal of wall finishes may also be necessary.   2.6.3 Comprehensive Rehabilitation Comprehensive rehabilitation includes all the measures from partial rehabilitation (which effectively address OoP collapse of URM walls), but also addresses in-plane demands on walls and generally ensures a complete seismic load patch.  This often involves introducing new seismic resisting systems, such as structural steel frames or reinforced concrete shear walls.  Several design standards/guidelines can be applied.  These include the following:  Seismic Evaluation Guidelines for Existing Buildings (NRC 1992)  The International Existing Buildings Code (ICC 2012a)  ASCE 41-13: Seismic Evaluation and Retrofit of Existing Buildings (ASCE 2013) Figure 2.35 – Photos of Intermediate Braces (Left) and Strongbacks (Right) Photo credit: Read Jones Christoffersen Ltd. Chapter 2 –Where existing URM walls are insufficient, traditional design practice is to provide a new seismic force resisting system to ignore the resistance and deformation compatibilityVictoria, 70% of current code is commonly specified by the local building authorities for seismic upgrading of URM buildings. In the United States, "performanceincreasingly common in seismic evaluatiauthorities begin to specify these standards.  force level, these standards set performance in displacements, accelerations,retrofit is designed to ensure these  Such standards have several scientific merits (eg. percent of code).  Perhaps the most important mseveral performance levels that could be targeted and thus require a dialogue with the owner and other stakeholdersFigure 2.36 below shows a spectrum of possible performance levels from ASCE 41.can be seen, performance goals could range from "very little damage" to "barely standing."                  Figure  Characterizing URM Bearing Wall Buildings - designed to some fraction of current code forces  of the in-plane URM wa -based" standards such as ASCE 41 are becoming on and upgrading of buildings, as the building Rather than specifying a "percent code" criteria for individual components ( or some other indicator of damage/safety) criteria are met. over the traditional, prescriptive approacherit is that they explicitly provide  as to the expected seismic performance of the2.36 – Seismic Performance Levels Modified From: ASCE, 2013 39 - and lls.  In measured and then the es  building.   As Chapter 2 – Characterizing URM Bearing Wall Buildings 40 Note that setting a performance goal also involves specifying under what earthquake intensity a given level of performance should be achieved (eg. under a code-level earthquake).  Traditional design practice for retrofitting could be considered as targeting the “life safety” performance objective at the specified design force demands (eg. 70% of current code), since this is the objective of codes for new buildings. In the engineering community, it is well-known that such designs make no attempt to prevent a building from being damaged beyond economical repair as a result of a design-level earthquake.  As noted by Rutherford and Chekene (1997), achieving higher performance levels such as Immediate Occupancy is very difficult for URM buildings and would likely necessitate nearly reconstructing the entire structure.  Nonetheless, it is important to explicitly specify the expected performance, as many building owners have spent hundreds of thousands or even millions of dollars on earthquake strengthening, only to have the building demolished after an earthquake - not because of safety concerns but because repairing the damage was not economically feasible.  In general, a comprehensive rehabilitation will necessitate providing a largely new seismic force resisting system at some locations.  Regardless of the desired level of performance, many of the new elements introduced will be similar in form (although the size, location, number, and detailing may differ).    2.6.3.1 New Seismic Force Resisting Elements Typical Details New seismic force resisting elements are employed at open fronts of row buildings (where URM walls are essentially absent) and where existing URM walls have insufficient resistance.  For particularly elongated buildings, new elements may also be required in the interior (rather than just along exterior walls).  Where existing URM walls are insufficient, traditional practice is to design the new elements to some fraction of current code forces and ignore the resistance (and deformation compatibility) of the URM walls.  In Victoria, steel concentrically braced frames are by far the most common design solution and so we focus on those herein.  Many other elements such as reinforced concrete/masonry shear walls or steel/concrete moment frames are possible.  Rutherford and Chekene (1990) provides a summary of all the common elements.  Figure 2.37 shows an example illustration of steel braced frames in a URM retrofit applications and Figure 2.38 provides an example photo.  Chapter 2 –Cost & Disruption The cost to complete this type of retrofit (in Victoria, in 2014 dollars) is about $40/sq.ft(see Appendix A), which includes all elements of a comprehensive retrofitthe cost of the seismic retrofit work. renovations are often combined with the strengthening.                         Figure 2.37 – Illustration of URM Retrofitted with Steel CBF Characterizing URM Bearing Wall Buildings .  This is just Because of the cost and disruption, architectural   From: Rutherford and Chekene, 1990 41 .  Chapter 2 –             The work involved with this level of strengthRutherford and Chekene (1990)areas would need to be turned over to the contractor for several weeks at a time.  The logistics of actually getting the steel into the buildings and erected can be challenging and often the most effective way to do so is to cut an access hole in the roof and make use of a mobile crane.  Holes in floors also must be cut to accommodate the co In general, steel moment frames involve areinforced concrete or masonry shear walls eliminate the need for cranes and steel members, but the work generates much more dust.most appealing from a disruption and constructability point of view, but their capacity is limited and thus they are most commonly employed in the interiors of buildings to reduce excessive diaphragm spans. 2.6.3.2 Diaphragm StrengtheningTypical Details The flexible wood diaphragmtheir dynamic response.  Depending on the diaphragm stiffness, the accelerations at a floor can be increased or decreased and the outis also heavily affected by diaphragm flexibility properties of the diaphragms and outdiaphragms may be found to have insufficient strength or stiffness. The typical method of strengthening diaphragms is by installing plywood, either atop the existing floor sheathing or to the underside of the joists.  Steel sheet metal straps calso be used to strengthen existing wood diaphragms, or the diaphragm could effectively Figure 2.38 – Example Photo of URM Retrofitted with Steel CBF Characterizing URM Bearing Wall Buildings ening can be quite disruptive. As stated by , workers are often on site for several months and large  similar level of disruption.  Of course,   Wood shear walls are li   s common to URM buildings have a profound impact on -of-plane rocking response of URM walls (Penner 2013).  Depending on the -of-plane walls to which they are connected,  Photo credit: Read Jones Christoffersen Ltd. 42 lumns.  unwieldy kely the an  Chapter 2 – Characterizing URM Bearing Wall Buildings 43 be replaced by installing new structural steel horizontal bracing.  Finally, a concrete overlay can be applied. Wilson (2012) discusses in detail the seismic behavior and the strengthening of wood diaphragms in URM buildings.  Figures 2.39 and 2.40 show design details and provide an example photo, respectively, for board sheathing retrofitted with a plywood overlay. An unblocked retrofit solution is shown, but blocking can be provided in cases where additional strength and stiffness are required.  It should be noted that certain Canadian and American design standards (NRC 1992, ICC 2012a) are limited to flexible diaphragm buildings, and so design solutions that render the diaphragm rigid are often not preferable.                               Figure 2.39 – Illustration of Plywood Overlay Retrofit From: FEMA, 2006 Chapter 2 –                Cost & Disruption Diaphragm strengthening is typically performed only as part of a comprehensive rehabilitation.  Much of the cost associated with diaphragm removal and reinstatement of finishes and other building contents.  As such, a cost for diaphragm strengthening alone is not provided.   The level of disruption associated ceiling finishes must be removed and building contents must be relocated.common practice to design retrofits so as to avoid the need for diaphragm strengthening.  Similarly, it is noted that many URM seismic evaluation/strengthening methodincluding the Canadian guideline buildings in the highest seismic hazard zones. 2.7 Summary and ConclusionsIn this chapter, an introduction to unreinforced masonry (URM) The form and components of a “typical” building on the west coast of North America were reviewed, and buildings in Victoria, BC were found to be similar.idealization and dynamic behavior of such buildings was presetheir relatively rigid walls and different from that assumed for most modern buildings in common seismic deficiencies and typical corresponding retrofit solutions were presented.   Figure  Characterizing URM Bearing Wall Buildings strengthening is with diaphragm strengthening is high,   (NRC 1992), only require diaphragm evaluation  buildings was provided.    The structural nted and it was noted that flexible diaphragms result in seismic behavior that is very new building codes.2.40 – Diaphragm Strengthening From: Wilson, 2012 44  due to since floor or As such, it is ologies,  for  Finally,  45 Chapter 3  URM Seismic Risk Mitigation Programs 3.1 Purpose and Scope Despite their well-known seismic vulnerability, URM buildings continue to be a leading (structural) source of loss of life and property damage in earthquakes and mitigation programs are routinely met with substantial public and private resistance, primarily due to the cost of seismic upgrading.  The purpose of this chapter is to review the historical impetus for URM seismic risk mitigation, and the programs that have been implemented.  The efforts to date in Victoria, BC will be compared to those abroad.  3.2 The Impetus for URM Seismic Risk Mitigation Historically, observations of poor seismic performance have been the primary driving force behind URM seismic risk mitigation.  Two interestingly similar examples include the 1931 Hawke’s Bay earthquake in New Zealand and the 1933 Long Beach earthquake in California, USA: unreinforced masonry construction was effectively banned in both of these locations, in direct response to the resulting losses.    The following statistics provide a sampling of the seismic performance of URM buildings in various significant historical earthquakes.  Examples of both poor performance of unstrengthened buildings and good performance of strengthened buildings are provided.  1925 Santa Barbara: 40% of unstrengthened buildings suffered severe damage or collapsed (FEMA 2009)  1933 Long Beach: 20% of unstrengthened buildings suffered severe damage or collapsed (FEMA 2009)  1971 San Fernando: 49 deaths caused by collapse of URM buildings at Veteran’s Administration Hospital (Stover and Coffman 1993)  1983 Coalinga: 60% of unstrengthened buildings suffered severe damage or collapsed (FEMA 2009)  1989 Loma Prieta: in regions of Modified Mercalli Intensity VIII (generally within 50km of the epicenter), 40% of unstrengthened buildings were demolished; additionally, 9 deaths were attributed to URM (Lizundia, Dong and Holmes 1993) Chapter 3 – URM Seismic Risk Mitigation Programs 46  1994 Northridge: of about 7000 URM buildings in the Los Angeles area (which experienced mostly MMI VII shaking), only about 600 were unstrengthened at the time of the earthquake; no fatalities due to URM buildings were recorded (Bruneau 1995)  2009 L’Aquila: 39 deaths attributed to URM, representing approximately 20% of the total deaths toll. The remainder were due to high population density in a few reinforced concrete buildings that collapsed (Alexander and Magni 2013)  2011 Christchurch: 39 deaths attributed to URM (Canterbury Earthquakes Royal Commission 2012); 82% of URM buildings red-tagged and only 1% green tagged (Ingham and Griffith 2011b); no public access permitted to many areas of central business district for over a year  While tragic, these losses have played a vital role in the advancement of URM seismic risk mitigation policies.  The following sections review the seismic risk mitigation efforts to date by various jurisdictions (eg. city, state, province, or country) facing similar seismic risks as Victoria, BC.  Seismic retrofit ordinances and incentive programs are discussed.  3.3 United States The United States has been at the forefront of seismic risk mitigation, especially in California and (more recently) throughout the Pacific Northwest.  This section summarizes seismic upgrading policies and financial aid/incentive programs for selected jurisdictions.  Note that this is not an exhaustive presentation and additional information is available in the literature.    3.3.1 California This section reviews the seismic risk mitigation efforts in the cities of Los Angeles, San Francisco, and Palo Alto.  The focus is on strengthening of the general existing stock, through risk mitigation ordinances.  Measures such as the 1933 Field Act, 1939 Garrison Act, and the 1967/1968 Greene Acts were significant steps for the seismic risk mitigation of new and existing school buildings (many of which were of URM construction).  Rutherford and Chekene (1997) provides a succinct summary.  3.3.1.1 Los Angeles The City of Los Angeles was the first local government to pass a retroactive URM seismic ordinance, in the form of its 1949 parapet correction ordinance (Rutherford & Chekene 1997).  In 1981, the City adopted an ordinance for comprehensive seismic Chapter 3 – URM Seismic Risk Mitigation Programs 47 strengthening, which is now known as Division 88 (City of Los Angeles 1985).  It covered all URM bearing wall buildings (i.e. infill buildings were not included), except one and two family dwellings and apartments with four or less units (Rutherford & Chekene 1997), which totalled over 8000 buildings.  Varying timelines for compliance were established based on a “rating classification.”  The rating classification is based on building function and occupant load, as shown in Table 3.1.    As expected, higher priority buildings had shorter timelines for compliance: essential buildings were to be strengthened within three years of their owners being served notice from the City that their building fell within the scope of the ordinance.  Lower priority buildings had the option of extending the deadline (as shown in Table 3.1) for full compliance by performing partial retrofits.  Partial retrofits included parapet bracing and tension anchors.  Table 3.1 – Division 88 Compliance Timelines Rating Classification Definition Full Compliance  (no partial retrofit) Full Compliance  (with partial retrofit) Essential Building (Highest Priority) Medical/Emergency  Services Centers 3 years 3 years High-Risk Building >100 Occupants 3 years 3.25 years Medium-Risk Building All Others 3 years 4-6 years (dep. on occupant load) Low-Risk Building (Lowest Priority) <20 Occupants 3 years 7 years  The City of Los Angeles did not provide any significant incentives for the general building stock.  However, the City’s Community Development Department (CDD) provided low-interest loans to cover project costs for residential and mixed-use buildings.  Statistics on the total number of buildings covered were not available, but there were over 1500 residential or mixed-use URM buildings affected by Division 88.  Additionally, the City’s Rent Stabilization Division controlled rent increases (Comerio 1989).  It should be noted that in order to be eligible for the financing, buildings also had to receive basic fire safety upgrades such as sprinklers and egress equipment.  While low-interest financing is a useful option, Comerio (1989) notes a few pitfalls that were encountered: Chapter 3 – URM Seismic Risk Mitigation Programs 48  The funds were intended only for seismic upgrading and the minimally required architectural and fire safety work, but several building owners took advantage of the loans to complete other work; thus, a strict control system was needed.  Changes in the work during construction were reviewed by the Building and Safety Department, which tended to slow construction; it was recommended that measures be put in place to expedite this process.  By 1989, 8% of the 1500+ residential/mixed-use buildings had been demolished and another 9% was in danger of demolition due to non-compliance with the ordinance; it was recommended that the City implement some type of demolition control, including requirements that an owner at least obtain and submit cost estimates for strengthening and that the owner meet with the City to discuss funding options.  Many of the buildings housed low-income tenants, who already spent an above average portion of their income on rent; it was recommended that rent increases be limited to $100/month or less (existing rental rates were $400-500/month).  While ultimately quite effective, the ordinance was fiercely contested and was debated in political arenas for over eight years (1973-1981).  Detractors argued (rightfully so to at least some extent) that the ordinance would place pressure on poor, marginalized citizens of the city through displacement and increased rent (Alesch and Petak 1986).  With regard to the compliance rate of the program, the California Seismic Safety Commission (CSSC) provided the following figures in 2006.  Table 3.2 – Los Angeles URM Ordinance Compliance Rates Total URMs Historic URMs % Strengthened % Demolished % Non-Compliant 9211 255 67% 21% 12%  With regard to the seismic performance of the retrofits, the 1994 Northridge earthquake provided a significant test.  Shaking in Los Angeles was mostly of MMI VII (PGA=0.15-0.35g).  At this time, only about 600 of 8200 URM buildings had yet to be mitigated (Bruneau 1995).  Bruneau noted that for unstrengthened buildings, out-of-plane failures were numerous. However, no lives were lost since the earthquake occurred at 4:30am, when the streets were largely empty.  The majority of strengthened buildings survived undamaged, although about 200 of 6000 suffered moderate to severe damage.  As noted by Bruneau (1995), this performance actually exceeded prior expectations (judgmentally) established by a panel of experts (EERI 1994). Chapter 3 – URM Seismic Risk Mitigation Programs 49 3.3.1.2 San Francisco In 1976, the City/County of San Francisco enacted its Parapet Safety Program, which required owners to retain a structural engineer to provide a seismic assessment of the parapets of their building.  The ordinance applied to all pre-1949 URM buildings that posed fall hazards to public sidewalks or occupied spaces (Bonneville and Cocke 1991) and bracing requirements were as per the engineer’s assessment, to a prescribed force level.  In 1986, Senate Bill 547 (commonly known as “the URM law”) was passed, which required jurisdictions in Seismic Zone 4, the highest seismic zone, to (CSSC 2006):  1) Identify all existing unreinforced masonry buildings by the end of 1989 2) Develop a risk mitigation program 3) Report on progress to the California Seismic Safety Commission   The law allowed each jurisdiction to develop its own mitigation program, which could range from simply informing owners that their buildings appeared to be of URM construction to mandatory comprehensive seismic upgrading of all subject buildings.  See 3.3.1.4 for further discussion on the URM law.  Like many communities, San Francisco opted to employ a mandatory strengthening program.  In 1992, it passed ordinance 225-92, which mandated strengthening/ abatement of approximately 2000 identified URM buildings.  Similar to the Los Angeles ordinance, various timelines for compliance were established based on levels of risk.  Table 3.3 shows the selected compliance deadlines.  Table 3.3 – Ordinance 225-92 Compliance Timelines Risk Level Definition Apply for  Building Permit Construction Complete Level 1 Group A occ. (300+ occupants) Group E occupancies 4+ storey buildings on poor soil 2.0 years 3.5 years Level 2 Non-Level 1 buildings located poor soils (in certain high-density areas, such as downtown) 2.5 years 5.0 years Level 3 Non-Level 1 buildings located on poor soil in other areas (i.e. lower density) 8.0 years 11.0 years Level 4 All other URM buildings 10.0 years 13.0 years Chapter 3 – URM Seismic Risk Mitigation Programs 50 Strengthening for most buildings was essentially to be in conformance with the 1991 Uniform Code for Building Conservation (ICBO 1991), which was largely based on an extensive testing and research program completed in the 1980’s by three prominent engineering firms in California (ABK 1984).  Extensive discussion is provided elsewhere on the technical basis (Rutherford & Chekene 1990, 1997, Bruneau 1994), but here it is simply mentioned that the methodology was created to reduce the required structural interventions and thus is thought to provide a reduced level of safety as compared to new buildings codes of the day.    Ordinance 225-92 also contained a relaxation for certain residential and commercial buildings: eligible buildings were able to meet the requirements of the ordinance by performing a “bolts-plus” retrofit, which called for just diaphragm-to-wall connections (shear and tension) and out-of-plane bracing for walls exceeding height-to-thickness limits, as specified in the UCBC (SFBIC 2013).  Table 3.4 shows the various measures included in the various strengthening schemes discussed thus far.  Table 3.4 – Measures Included in Various Strengthening Schemes Modified After: Bonnevile and Cocke, 1991 Strengthening Measure Parapet Ordinance Bolts-Plus UCBC Code for New Buildings Parapet Bracing X1 X X X Roof-to-Wall Tension Anchors X1 X X X Diaphragm-to-Wall Anchors (Shear and Tension)  X X X Out-of-Plane Wall Bracing  X X X2 In-plane Wall Strengthening   X X2 Diaphragm Strengthening   X X2 Veneer Ties  X3 X3 X3 Other Rooftop Fall Hazards (eg. cornices, chimneys)   X3 X3 1 Required only at locations where parapets pose a public fall hazard (eg. walkways, entrances, locations overlooking buildings below 2 More stringent requirements than those of the Uniform Code for Building Conservation (UCBC) 3 Not historically well-enforced in some jurisdictions  Chapter 3 – URM Seismic Risk Mitigation Programs 51 The full provisions of the UCBC are considered to provide a lower level of life-safety than new building codes and the “bolts-plus” provisions must obviously achieve less.  Nonetheless, the pressing socioeconomic issues associated with mandatory strengthening led to the bolts-plus provisions (FEMA 1998).   It should be noted that, in order to qualify for the bolts-plus procedure, buildings were required to meet a number of criteria, including:  The building does not contain occupancies of group A (assembly) with > 300 persons, group E (education), group H (hazardous), or group I (industrial)  Mortar shear strength (from in place shear tests) ≥ 30psi  Wood diaphragms at all levels above the base of the building  Maximum of 6 storeys  The building does not have various irregularities, listed below o Soft/weak storey o In-plane discontinuity (of walls) o Diaphragm discontinuity o Out-of-plane offsets  Minimum of two lines of lateral force resisting elements in each direction (i.e. open front buildings do not qualify); solid wall must comprise at least 40% of the walls length to be considered a line of resistance  The building has or will be provided with crosswalls at a spacing not exceeding 40 feet on center  Owners could either correct deficiencies and rehabilitate to bolts-plus or implement a UCBC-compliant retrofit scheme.  These requirements are relatively restrictive: in a review of 66 red-tagged San Francisco buildings after the Loma Prieta earthquake, Bonneville and Cocke (1991) estimated that only 35 would have qualified for a bolts-plus retrofit.  In terms of incentives, low-interest loans were made available through a $350 million general obligation bond, approved by a public vote.  $150 million of this was devoted to affordable housing, while the remainder was available for any building (City of San Francisco n.d.).  With regard to the compliance rate of the program, the California Seismic Safety Commission (CSSC) provided the following figures in 2006.  Chapter 3 – URM Seismic Risk Mitigation Programs 52 Table 3.5 – San Francisco URM Ordinance Compliance Rates Total URMs Historic URMs % Strengthened % Demolished % Non-Compliant 1976 516 78% 8% 14%  In comparing these figures to those of Los Angeles, we can see that San Francisco achieved a similar compliance rate, but appears to have had more success in preventing demolitions.  This is likely due to a number of factors, including the following:  The ordinance allowed a relaxation for certain buildings  Design standards and construction methodologies were more refined, resulting in lower costs to upgrade  Loans were made available to all buildings, not just those of residential/mixed-use occupancies  A much higher fraction of San Francisco’s buildings (26%) were considered historic than those of Los Angeles (3%) (CSSC 2006)  San Francisco was likely able to take advantage of lessons learned in Los Angeles in terms of demolition control  Although mandatory strengthening measures were not yet in place, the 1989 Loma Prieta earthquake provided a test of the parapet strengthening measures.  Based on damage observed in their post-earthquake safety evaluations, Bonneville and Cocke (1991, 1992) assessed the effectiveness of the parapet bracing.  Their observations were as follows:  66 red-tagged URM buildings were included in the study  50 of these had parapet bracing – none experienced collapse of parapets or out-of-plane walls  16 of these did not have parapet bracing – 3 suffered partial collapse, one of which caused 5 deaths (Wiggins, Breall and Reitherman 1994)  Based on the extents and types of failures observed, Bonneville and Cocke also projected the performance that bolts-plus retrofits would have provided and concluded that “no collapse would have occurred if all UMBs had undergone the equivalent of a bolts-plus strengthening.”  Of course, it was also noted that the good performance of these limited strengthening measures was largely because shaking was of only moderate intensity.    Chapter 3 – URM Seismic Risk Mitigation Programs 53 3.3.1.3 Palo Alto The City of Palo Alto implemented a different, but reasonably successful, URM seismic risk mitigation program.  After a failed attempt to establish a mandatory program in 1982 – and the Coalinga earthquake of 1983 – the City created a “Seismic Hazard Committee.”   The committee represented a variety of stakeholders and was tasked with developing an acceptable risk mitigation program.  In 1986, the City passed an ordinance that entailed the following (FEMA 1998, Gibson Economics 2014):  Three buildings types were included o Unreinforced masonry o Pre-1935 non-URM with 100+ occupants o Pre-1976 non-URM with 300+ occupants  Building owners were required to engage a structural engineer to conduct a seismic evaluation of the building, specifying the necessary upgrades  Seismic evaluations were submitted to the City and owners were required to inform building occupants that the reports were available for their review  Within one year of filing the report, owners were required to submit a letter to the City indicating their intentions to address the building  Strengthening remained voluntary and incentives were made available, including FAR (floor/lot area ratio) increases of up to 25%.  However, the incentives were not widely used (Gibson Economics 2014).  The primary driving factor was the public and occupant awareness created by the publicly available engineering assessments. For example, the level of awareness was so high that seismic improvements were able to be marketed by building owners.  Additionally, some tenants agreed to help finance upgrade costs and other voluntarily agreed to vacate the space during construction and return upon completion (FEMA 1998).  Table 3.6 provides the compliance statistics as of 2006.  Table 3.6 – Palo Alto URM Ordinance Compliance Rates Total URMs Historic URMs % Strengthened % Demolished % Non-Compliant 47 4 43% 21% 36%  The compliance rate (strengthened + demolished) is much higher than other voluntary programs which average 24% (CSSC 2006).  Aside from the program itself, however, there were a number of factors which likely played a role, the three most significant being the following:  Chapter 3 – URM Seismic Risk Mitigation Programs 54  The relatively small size of the community (1990 population of 55,000), which facilitated community involvement and generation of support for the program  The relatively small size of the URM building stock, leading to smaller overall costs  The fact that the community was rather affluent, being home to Stanford University, many political figures, and wealthy high-technology professionals (FEMA 1998)  Nonetheless, this program is an example of policy-making that was carefully crafted to generate public support rather than opposition, making it successful from both an engineering and a political view.  3.3.1.4 Effectiveness of Mandatory versus Voluntary Strengthening Having reviewed a few successful programs in detail, we now take a step back and examine the broader question of the relative effectiveness of mandatory versus voluntary strengthening for URM buildings, as it was experienced in the State of California.  In 1986, the State of California passed Senate Bill 547, more commonly known as the “URM Law” (California Legislature 1986).  The law applied to approximately 26,000 URM buildings in areas of highest seismic hazard under the California Building Code (based on the 1985 Uniform Building Code at the time the law was passed) and required 365 affected local governments to: inventory URM buildings, establish loss reduction programs, and report on progress (CSSC 2006).  The law recommended, but did not require, that local governments include mandatory strengthening in their loss reduction programs.  “Voluntary strengthening” and “notification-only” programs also met the requirements of the law.   Mandatory strengthening programs generally required comprehensive upgrading for in-plane and out-of-plane seismic demands.  The most commonly used standard was Appendix Chapter 1 of the Uniform Code for Building Conservation (ICBO 1991).  Other mandatory programs, such as San Francisco’s, required only partial retrofits for some buildings to limit economic impacts on the community.  Voluntary strengthening programs typically encouraged comprehensive seismic upgrading, similar to the mandatory requirements noted above.    Chapter 3 – URM Seismic Risk Mitigation Programs 55 Notification-only programs typically included only a letter from the local authority having jurisdiction to buildings owners, stating that their building is of URM construction and is a potential seismic risk.  Of the 365 affected local governments, 283 were found to have URM buildings in their jurisdiction.  The majority of these adopted mandatory strengthening programs.  Table 3.7 summarizes the loss reduction program types established and affected number of URM buildings as of 2006 (CSSC 2006):  Table 3.7 – California URM Loss Reduction Program Statistics After: Paxton, Elwood, Barber & Umland, 2013  Program Type # Jurisdictions % Jurisdictions # URMs % URMs Mandatory Strengthening 134 47% 19,043 73% Voluntary Strengthening 39 14% 1,269 5% Notification-Only 46 16% 1,487 6% Other1 41 15% 3,737 14% No Program Established 23 8% 409 2% Total 283 100% 25,945 100% 1 Combinations and variants of the above noted programs  As of 2006, approximately 55% of the affected URM buildings had been retrofitted and 15% had been demolished, for an overall mitigation rate of 70%.  Table 3.8 shows a breakdown of the results for each program type.  The statistics clearly show that mandatory programs are much more effective at mitigating seismic risks.  Table 3.8 – California URM Risk Mitigation Statistics by Program Type After: Paxton, Elwood, Barber & Umland, 2013   Mandatory Voluntary Notification Only Other No Program Total No. of  Jurisdictions 134 39 46 41 23 283 No. of  URM Bldgs 19,043 1,269 1,487 3,737 409 25,945 % Retrofitted 70% 16% 7% 15% 4% 55% % Demolished 17% 8% 6% 11% 27% 15% % Mitigated 87% 24% 13% 26% 31% 70%   Chapter 3 – URM Seismic Risk Mitigation Programs 56 3.3.2 Washington URM seismic risk mitigation has also been a topic of interest in Washington State, due in large part to property damage and deaths caused by earthquakes in 1949, 1965, and 2001.  Efforts in Seattle and Tacoma are reviewed.  3.3.2.1 Seattle In Seattle, an inventory survey of URM buildings was recently completed by the Department of Planning and Development (City of Seattle 2012).  This survey built on past surveys completed for the DPD (Hoover 1992, Reid Middleton 2007).  Approximately 800 buildings were identified, 10-15% of which appeared to have received some degree of seismic upgrading.  Recognizing the risks associated with unretrofitted URM buildings, the City of Seattle passed ordinances requiring retrofitting of all URM buildings in 1973; however, the ordinances were repealed a couple of years later due to public opposition and administative difficulties (FEMA 1998). Currently, comprehensive seismic upgrading is only triggered by changes of use or occupancy, similar to Victoria.  However, Section 3401.8 (“Unsafe Building Appendages”) of the Seattle Building Code states that “Parapet walls, cornices...that are in a deteriorated condition or are otherwise unable to sustain the design loads...are hereby designated as unsafe building appendages” and “shall be abated in accordance with Section 102 (City of Seattle 2009).”  This is essentially a narrowing of focus of the general “unsafe condition” clause (in this case Section 102) that is present in most building codes.  Unfortunately enforcement of this clause has reportedly been limited (Rogers, et al. 1998).  More recently, development of URM risk mitigation policies has again gained traction in Seattle: recent draft documents indicate that comprehensive mandatory upgrading is being considered, with relaxations to partial upgrading requirements for certain buildings (similar to San Francisco) (City of Seattle 2012b).  Table 3.9 shows selected compliance deadlines.  Other timelines were set for obtaining an engineering assessment and permit approval.  Sanctions for non-compliance may include the following:  Quarterly fines ($500 at assessment stage, $1000 at permit stage, $45,000 for full compliance deadline)  Public posting of non-compliance (online or on site)  Freeze on new permits for the building  Denial of incentives  Abatement of the property by the City Chapter 3 – URM Seismic Risk Mitigation Programs 57 Table 3.9 – Proposed Seattle Ordinance Compliance Timelines Rating Classification Definition Apply for  Building Permit  Complete Construction Critical Risk (Highest Priority) Emergency Services, Shelters, Schools 1 year 7 years High Risk 4+ storeys on poor soil, 100+ occupants 2 years 10 years Medium Risk All others 3 years 13 years  Several incentives were identified by City personnel for consideration, including the following (City of Seattle 2014):  Federal grants: available for public/non-profit owned buildings from the Federal Emergency Management Agency  General obligation bonds: voter-approved municipal bonds for a city-administered retrofit funding program  Levies: voter-approved increase in money collected from each property owner for a City-administered retrofit funding program  Tax abatement: reduction/elimination of property taxes for a designated period of time; owners to use funds for retrofitting  Transfer of Development Rights: allowing owner of buildings in a designated area to sell developable air space above the building to other developers, who could then increase the density of their developments  Federal Tax Credit: 10% (20% for national historic buildings) tax credit for construction costs of seismic retrofits  3.3.2.2 Tacoma In 1965, just months after Olympia earthquake, the City of Tacoma adopted an ordinance for parapet strengthening.  The ordinance specifically identified parapets as hazardous building appendages and made it possible for the city to require abatement (Rogers, et al. 1998).  More recently, with the adoption of the International Building Code (ICC 2012) and International Existing Building Code (ICC 2012a), parapet bracing is mandatory when more than 25% of the roof area is re-roofed.    3.3.3 Oregon In Portland, a publicly available URM survey was completed in the early 1990’s (City of Portland 2001).  However, the information has not been verified or updated by the City.  Chapter 3 – URM Seismic Risk Mitigation Programs 58 In 2007, a state-wide seismic needs assessment for public buildings was completed (Lewis 2007). This survey included all types of structures, rather than solely URM.  Approximately 1800 URM buildings were identified.   Similar to many areas, comprehensive seismic upgrading is currently only required as part of a change of use or occupancy or other significant renovation, although the trigger system in the IEBC (International Existing Building Code) is somewhat detailed.  The retrofit rate in Portland is thought to be similar to Seattle (Powell 2011), although statistics are not available.  However, the Portland City Code does have a retroactive parapet ordinance requiring an engineering assessment and remediation when roof repairs or replacement are undertaken (City of Portland 2011).    3.4 New Zealand In New Zealand, the traditional ‘change of occupancy/significant renovation’ triggered seismic upgrades are required.  The country also has legislation addressing “earthquake prone” buildings (EPB’s).  Note that the definition of an EPB is one that “will have its ultimate capacity exceeded in a moderate earthquake” and if it were to collapse “would be likely to cause injury or death.”  (Government of New Zealand 2014).  A moderate earthquake is defined as having one-third the intensity (and same duration) as that specified for new buildings.  Before discussing practices in New Zealand, it should be noted that the nation-wide provisions of the Building Act are under review and could change drastically.  Currently, however, each Territorial Authority (a level of local government, sometimes individual cities) is responsible for developing its own EPB policy and is afforded considerable flexibility.  Territorial Authorities could either actively identify EPB’s and require strengthening, or identify buildings as owners apply for building consents for other construction work.  Note that the method of determining the ultimate capacity is not explicitly specified, but the typical methods used in practice are as specified in New Zealand’s national guidelines for seismic assessment (NZSEE 2006).  An “Initial Evaluation Procedure” (IEP) is used for the screening and identification. The IEP for URM buildings is discussed in more detail in Section 6.2.5.  Strengthening is simply required to achieve a capacity such that the building is no longer earthquake-prone (i.e. 34% of current code).  However, the New Zealand Society for Chapter 3Earthquake Engineering (NZSEE) recommends strengthening to 67%NBS2006).  This viewpoint is relatively consistent with seismic upgrading practices in most countries whereby, once strengthening is triggered, it ishigher level than just above the trigger level. With regard to performance of earthquakecounterparts, the Christchurch earthquake of February 2011 provided a significant testBefore discussing the performance was well in excess of even theFigure 3.1 compares the average response spectra from the earthquake with the design spectra for Christchurch and Victoria.                 As expected given the high intensity of shaking, damage was severe.  in the Christchurch Central Business District (CBD) were surveyed.  Of these, the %NBS value was estimated for illustrates the observed performance.performance with increased strengthening.  Although buildings in the 33%NBS range appeared not to perform particularly well, it should be remembered that the intensity of the earthquake was 150-200%NBS over much of the response spectrum.  Buildings strengthened to the 67-100%NBS range performed well.                                     4 “NBS” is an abbreviation for “New Building Standard” (i.e. current code for new buildings)Figure 3.1 – Design Spectra vs. February 2011 Response From: Paxton, Elwood, Barber, and Umland, 2013 – URM Seismic Risk Mitigation Programs  customary to strengthen to a  -prone buildings and their strengthened it should be noted that the intensity of the earthquake  design standards in place at the time of the earthquake   370 URM buildings 125 buildings (Ingham and Griffith 2011b) The results clearly show the trend of impro   Spectra 59 4 (NZSEE . .  .  Figure 3.2 ved  Chapter 3              3.5 Canada Having looked abroad for a sampling of seismic risk mitigation practices, now turn to efforts in Canada.  are discussed for Vancouver and for Victoria 3.5.1 Vancouver The City of Vancouver does not have an inventory of URM buildings.  However, it has incorporated a comprehensive section on existing buildings into its building code, the Vancouver Building Bylaw, or VBBL Buildings” of the VBBL and its Appendix includes triggers clearly indicating when seismic upgrading is required and to what extent.  In addition to the traditional triggers of change of use/occupancy and “significant improvements,”scenarios triggered by repairs, reconstruction, as well as “minor” and “major” renovations and additions.    For renovations and certain changes to lower risk occupancies, seismic requirements are limited to, at most, securing nonadditions (i.e. those that impose additional gravity loads/seismic weight to the existing structure), comprehensive seismic upgrading to at least 75% of current code demands would be required if the sucapacity is less than 60% of number of retrofits achieved under this trigger  Figure 3.2 – Damage vs. %NBS for 125 URM Buildings in ChristchurchFrom: Paxton, Elwood, Barber, and Umland, 2013 – URM Seismic Risk Mitigation Programs the focus willRisk mitigation efforts and seismic upgrading practice. (City of Vancouver 2012): “Part 10  Part 10 defines upgrade -structural items and fall hazards.  For major vertical bject building’s seismic evaluation shows that the current current code.  There are no statistics available for the -based system.  60  s – Existing  Chapter 3 – URM Seismic Risk Mitigation Programs 61 3.5.2 Victoria Having summarized seismic risk mitigation in many other jurisdictions, we will now focus on Victoria and provide a more detailed account, including the building code, and heritage incentive programs for seismic upgrading.  3.5.2.1 Building Standards The City of Victoria falls under the requirements of the British Columbia Building Code (BCBC), which is largely based on the model code, the National Building Code of Canada (NRC 2010).  Part 1 of Division A specifies the application of the code, including but not limited to, the following activities (as per Clause, 1.1.1.1):  A change of occupancy of any building  An alteration of any building (note: an alteration is defined as “a change or extension to any matter or thing or to any occupancy…”)  An addition to any building  Reconstruction of any building that has been damaged by fire or earthquake or other cause  The correction of any unsafe condition in or about any building  An alteration, rehabilitation, or change of occupancy of heritage buildings  Clause 1.1.1.2 (“Application to Existing Buildings”) also specifies the following: “Where a building is altered, rehabilitated, renovated, or repaired, or there is a change in occupancy, the level of life-safety and building performance shall not be decreased below a level that already exists (see Appendix A).”  Although the aforementioned Appendix A has no legal effect (as stated in Cl. 1.1.3.1), it contains a variety of relaxations that apply to heritage buildings, mostly concerned with fire safety and accessibility.  However it also gives reference to the NBCC 2010 User’s Guide (NRC 2011), Commentary L on application to existing buildings.   Commentary L is relevant to this study because, although it has no legal effect, it prescribes minimum force levels (through reduced load factors) for the evaluation of existing buildings. For earthquake loads, a factor of 0.6 is specified. It goes on to state that the reduced load factors are “based on maintaining the level of life safety [but not other objectives such as economy] implied by Part 4.” Finally, it notes that this force level “should be considered suitable as a triggering criterion for seismic upgrading…” but that “…for design of upgrading, the load factor should be increased, preferably to the NBC value.”  For unreinforced masonry buildings, Commentary L also refers to the Chapter 3 – URM Seismic Risk Mitigation Programs 62 NRC “Guidelines for Seismic Evaluation of Existing Buildings,” (NRC 1992) in particular to Appendix A of that document (not to be confused with Appendix A of the building code, as discussed above), which is the Canadian equivalent of the URM provisions in the previously mentioned Uniform Code for Building Conservation, from the United States.  This represents the extent of code provisions and formal guidance on seismic upgrading of existing buildings.  As can be seen, the triggers are quite vague and enforcement relies on the judgment of the building officials who enforce the code.  3.5.2.2 Local Enforcement The aforementioned requirements of the BCBC and recommendations of Commentary L are essentially reflective of the local practice in Victoria for triggering of seismic upgrades: when a building undergoes a change of occupancy or a significant (undefined) renovation, the City requires that a seismic assessment be performed to 60% of current code forces; if the resistance of the structure is less than this threshold, upgrading is required, typically to a level of 60-75% of current code.  The exact upgrade force level is not specified, and may vary from building to building.  For unreinforced masonry buildings, the issue is somewhat complicated by the fact that the aforementioned Appendix A of the seismic evaluation guidelines was developed to be compatible with the 1990 version of the NBCC and, thus, is not entirely compatible with the current code for new buildings.  Unlike the City of Vancouver, Victoria does not have well-defined formal triggers or target force levels for upgrades and, thus, requirements are essentially left to the discretion of the building officials.  However, the intent of the enforcement is similar and Victoria can look to the Vancouver triggers as a guideline.  Figure 3.3 provides a summary of the seismic triggering/upgrade practice in Victoria.   Chapter 3 – URM Seismic Risk Mitigation Programs 63                      While the aforementioned practice is typical in assessing requirements for overall building seismic upgrades, it is not translated directly in terms of partial upgrades, such as restraint of non-structural components.  Seismic restraint is typically required for new buildings and new components of renovations, but there are no triggers for partial upgrading of the remaining components.   As a final note on triggers, Victoria does not make use of the “correction of any unsafe condition” clause to require seismic upgrading, even of publicly hazardous non-structural components such as unreinforced masonry parapets and veneers.  This is consistent with general practice in the Pacific Northwest.  As previously discussed, however, some jurisdictions in the U.S. have adopted mandatory retroactive ordinances requiring bracing of parapets; others have adopted triggered ordinances, whereby the parapets of URM buildings must be assessed (and potentially upgraded) when the building is re-roofed.  3.5.2.3 Heritage Incentives In the 1970’s, the City began a sophisticated heritage program to conserve its important historic buildings in a core area of downtown known as “Old Town.”   Figure 3.4 shows Victoria as well as the extents of Old Town. The city also established an arms-length, non-profit organization known as the Victoria Civic Heritage Trust (VCHT) in 1989 to deliver programs of financial assistance to the owners of protected heritage buildings in the downtown core. The vast majority of these buildings are URM buildings. BCBC 2012 Triggers: -Change of Occupancy -Alteration -Addition -Reconstruction Unofficial guidance on triggers from VBBL Part 10 Seismic Eval -General: 60% BCBC -URM: NRC Guidelines No seismic upgrade required Upgrade: -General: 60-75% BCBC -URM: NRC Guidelines  OK Not OK Figure 3.3 – Seismic Evaluation and Retrofit Practices in Victoria Chapter 3 – URM Seismic Risk Mitigation Programs 64                    Since 1990, the VCHT has operated the “Building Incentive Program” (BIP), which offers cash grants on a 50% matching fund basis up to $50,000 to owners of legally protected heritage commercial, industrial, and institutional buildings in Victoria (Victoria 2012a).  The purpose of the BIP grant is to promote the conservation and rehabilitation of the buildings; it does not specifically require seismic rehabilitation work, although structural and seismic upgrading work is an eligible cost when allocating grants.    VCHT also assists the City with the administration of the “Tax Incentive Program” (TIP).  Started in 1998, this program provides up to a 10-year tax exemption to building owners to assist in the creation of new residential units on the upper floors of underutilized heritage buildings or the substantial renovation of non-residential heritage buildings in the downtown core.  The objective is to help offset the high cost of the seismic upgrading.  Considered a major success, the program has stimulated the creation of 660 residential units and attracted over $225 million in private sector investment in 31 seismically upgraded and rehabilitated heritage buildings (City of Victoria 2014).  These include former warehouses, industrial buildings, hotels, a department store, and a hospital.  Figure 3.5 shows example photos of the types of buildings addressed under the TIP. OLD TOWN Figure 3.4 – Victoria and “Old Town” After: Paxton, Elwood, Barber, and Umland, 2013 Chapter 3 – URM Seismic Risk Mitigation Programs 65              Victoria has approximately 200 heritage buildings in the downtown core; this includes both “heritage-designated” buildings (legally protected by municipal bylaws and subject to City Council requirements in the case of alterations) and “heritage-registered” buildings (unprotected, but registered for monitoring and potential future designation) (City of Victoria 2014).  While both the BIP and TIP programs have stimulated a considerable amount of seismic upgrading, a significant challenge for Victoria is the remaining unquantified number (likely a few hundred) of unreinforced masonry buildings which continue to exist with limited occupancy or operate with successful retail businesses on the ground floor and no incentive to undergo a change of use or significant renovation.  Provisions for voluntary partial upgades are likely the only manner in which such buildings will be addressed in a timely manner.  Without an inventory of these buildings, the problem cannot be quantified, let alone addressed.   In an effort to promote voluntary seismic upgrading of parapets and facades, the VCHT has developed a new “Parapet Incentive Program” (PIP) for launch in early 2015.  The program’s structure is similar to the BIP, except that funds are marked solely for seismic upgrading of parapets and facades (i.e. wall anchorage).  The program is currently in a trial phase for a target area.  In order to better understand the costs and benefits of such a program, an effort was made in this study to quantify the seismic performance improvement and expected costs/benefits on a probabilistic basis, as presented in Chapter 4 and Chapter 5, respectively.  Figure 3.5 – Example TIP-Retrofitted Buildings From: Paxton, Elwood, Barber, and Umland, 2013 Chapter 3 – URM Seismic Risk Mitigation Programs 66 3.6 Comparing Victoria to Other Jurisdictions Table 3.10 provides a summary of URM seismic risk mitigation measures throughout the Pacific Northwest.  Note that for Seattle the measures have not been finalized, but the intent is to show the general trends in activity.  Only selected jurisdictions in California are shown here, but there is a great variety among the 283 affected jurisdictions with URM buildings.   Overall, the efforts to identify and mitigate URM seismic risk in Victoria appear to be lacking in comparison to other regions of the Pacific Northwest.  While the incentives for heritage buildings have been successful in promoting re-development, and thus comprehensive seismic upgrading, a complete inventory of URM buildings in Victoria has not been performed and risk mitigation measures are not in place to effectively address occupied buildings, nor are there currently provisions for partial seismic upgrading of buildings to address the most pressing and easily corrected issues of parapet and facade seismic restraint.  Table 3.10 – Comparison of URM Seismic Risk Mitigation by Jurisdiction Jurisdiction URM Inventory Retrofit Requirements URM Mitigation Rate Parapet Ordinance Los Angeles, CA Yes Mandatory 88% Mandatory (enacted 1949) San Francisco, CA Yes Mandatory 86% Mandatory (enacted 1976) Palo Alto, CA Yes Mandatory 64% Triggered (Reroofing) Seattle, WA Yes Triggered Proposed Mandatory 10-15% Mandatory (enacted1973) Tacoma, WA No Triggered (Detailed System) Unknown Triggered (Reroofing) Portland, OR Yes Triggered (Detailed System) Unknown Triggered (Reroofing) Vancouver, BC No Triggered (Detailed System) Unknown None Victoria, BC No Triggered Unknown None    Chapter 3 – URM Seismic Risk Mitigation Programs 67 3.7 Additional Considerations 3.7.1 Retrofit Design Standards Good policy-making is the foundation of a successful URM seismic risk mitigation program.  To ensure the retrofits perform as intended, however, appropriate technical standards must be in place.  Because the focus of this chapter is not on technical standards or seismic performance, the discussion herein is limited to identification of a few issues that may also come into play in policy making.  Firstly, an appropriate design standard must be specified.  In the United States, the International Existing Building Code (ICC 2012a) is the most commonly specified standard. Another possible standard is ASCE 41 (ASCE 2013).  Both of these documents are actively maintained with updates published every 3-5 years.  In Canada, the Guidelines for Seismic Evaluation of Existing Buildings (NRC 1992) is the most recent and relevant standard.  This document was derived from American documents that were the predecessors to the IEBC and ASCE 41.  Unfortunately, it is not actively maintained and no updates have been made since its initial publication.  Although the NRC guidelines contain much of the content that is still found in the American standards, much new earthquake engineering knowledge has come to light since its publication.  For example, the Northridge earthquake showed deficiencies with URM veneer retrofits, mostly due to a lack of proper enforcement and field quality control.  The 2010/2011 Canterbury earthquakes showed poor performance of adhesive wall anchors for URM buildings, again largely due to quality control issues.  Another issue is that the NRC guidelines were developed to accompany the 1990 NBCC; since then, new editions have been issued in 1995, 2005, and 2010 (with 2015 issues coming soon).  Because of significant changes to the NBCC since 1990, the guidelines are not highly compatible with current code for new buildings. This can be problematic because building authorities often specify performance in terms of a fraction of current code (herein abbreviated as “%code”).  3.7.2 Performance-Based Design An alternative to the aforementioned force-based specification is performance-based design, as provided for in the ASCE 41 standard (recall Figure 2.36).  Although a displacement-based approach for a building with largely unknown (and brittle) material properties is likely not as fruitful as it could be for a modern building, this type of standard at least promotes a discussion on performance.  On one hand, a retrofit using public money should perhaps use no more resources than are necessary to achieve Chapter 3 – URM Seismic Risk Mitigation Programs 68 minimally acceptable life-safety goals (i.e. “collapse prevention” performance).  On the other hand, heritage societies and city planners should be completely aware that retrofits often do nothing to actively ensure that a building will not need to be demolished after a design level earthquake.  3.7.3 Enhanced Design and Construction Supervision Errors in design (eg. undetected veneer wythes or poor mortar) and quality control issues (eg. adhesive anchor installation) have been the most frequent issues linked to failures of retrofitted URM buildings (Bruneau 1995, Rutherford & Chekene 1997).  Moreover, Rutherford and Chekene (1997) showed that the marginal costs for increased design and construction supervision measures are quite small, relative to other improvement measures.  Because of these issues, it would seem logical to devote additional resources to enhanced design and construction supervision.  Several such positive example programs are in existence.  California’s Field Act should certainly be considered successful, as “not one public school building constructed under the Field Act has collapsed nor has anyone died in earthquakes (CSSC 2007).”    3.7.4 Other Building Upgrades Another question to be answered in developing a URM seismic risk mitigation program is the extent to which other building deficiencies, such as accessibility and fire safety should be addressed.  Generally, mandatory upgrade programs have excluded accessibility upgrades.  In the case of Los Angeles, minimal fire safety upgrades, to the “Dorothy Mae” Ordinance, were completed at a reported cost of $50,000 per building, on average (Comerio 1989).  Other jurisdictions have typically excluded fire safety upgrades.  3.8 Conclusions Based on the review of California’s statistics, it is concluded that mandatory programs are much more effective at mitigating URM seismic risk than are voluntary (or other passive) programs.  However, there are a number of socioeconomic issues to be considered and it is essential that any ordinance must have substantial input from the stakeholders within the community.  Based on the facts that Victoria does not have an inventory of URM buildings, does not have ordinances requiring parapet upgrades, and only requires comprehensive upgrades as part of a change of use/occupancy, it is concluded that URM seismic risk mitigation measures in Victoria are lacking compared to other jurisdictions.  The same may be said of Vancouver or southwestern BC in general.  69 Chapter 4  Quantifying Building Vulnerability Through Observed Damage Statistics 4.1 Purpose and Scope In Chapter 3, URM seismic risk mitigation efforts of several communities were reviewed.  It was found that, in most cases, the decision to implement these programs has been based primarily on anecdotal evidence and emotional/political response to disastrous earthquakes.  In order to develop a more rational basis for such decisions, it is first necessary to quantify the seismic performance of such buildings in a systematic way.  The improvements can then be weighed against the costs of strengthening and a decision made regarding the appropriate action.   In this chapter, a statistical basis for the performance of unreinforced masonry buildings is developed, including the effects of strengthening and various base building characteristics.  These results will subsequently be used in Chapter 5 to perform a cost-benefit analysis for URM seismic strengthening and in Chapter 6 as a basis for developing a seismic screening system for URM buildings.  4.2 The Role of Damage Observations in Engineering and Policy-Making Observations from earthquakes have long played an important role in the advancement of earthquake engineering and in policymaking.  The 1906 San Francisco earthquake led to the formation of the Structural Association of San Francisco, which later became the Structural Engineers Association of California (MCEER 2008).  In response to the 1925 Santa Barbara earthquake, the first seismic provisions in the United States were published in the 1927 Uniform Building Code (Risk Management Solutions 2006).  The 1933 Long Beach earthquake resulted in immediate passage of the Field Act, which prohibited the use of unreinforced masonry construction for new schools (Rutherford & Chekene 1997) and effectively marked the end of unreinforced masonry construction in many locations through changes to building codes (Risk Management Solutions 2006). Another significant outcome of the Long Beach earthquake was the first comprehensive, systematic damage survey in the United States, including damage scales for various building types and statistical break downs (Wailes and Horner 1933).   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 70 Reconnaissance by researchers, building officials, and engineers from past earthquakes has provided a wealth of information on the seismic performance of unreinforced masonry buildings (and structures in general) and led to many important developments.  However, much of the reconnaissance has been anecdotal in nature.  Fortunately, systematic damage surveys and statistical analyses of performance have become more common over the past few decades, particularly with the increasing availability of strong motion recordings throughout the late 1980’s and 1990’s.  Such studies have been invaluable because they have permitted the quantification of performance results and, ultimately, more rational and objective conclusions about seismic performance and risk mitigation alternatives.    In this chapter various damage assessment methodologies encountered in this study are reviewed, as are the various sources of damage statistics.  Subsequently, some of the more pertinent unreinforced masonry damage surveys in the literature are presented and, finally, further work by the author using various existing databases is presented.  As will be seen, a reasonably substantial body of work on this topic was completed throughout the 1980’s and 1990’s.  The impetus for the work stemmed largely from the URM seismic risk mitigation programs that were being enacted throughout California at the time, such as the 1986 “URM law” (California Legislature 1986) and Los Angeles’ Division 88 (City of Los Angeles 1985).  Building owners, engineers, and city officials alike were interested in quantifying the benefits achieved through these expensive mitigation strategies.  Since this time, however, interest has waned as the affected communities have completed their mitigation programs.  Little information has been collected on URM buildings in British Columbia.  Therefore, a thorough review of existing information is needed to capture the range of possible performance of British Columbia URM buildings. The collection presented herein is certainly not exhaustive, but is believed to represent a sample that sufficiently supports the conclusions reached.  4.3 Review of Systematic Damage Assessment Methods As might be expected based on the number of different interested parties and possible uses of the results, there are many different damage assessment methods and/or rating scales.  The relevant scales for this study include those from Wailes and Horner (1933), ATC-13 (ATC 1985), ATC-20 (ATC 1989), and HAZUS (FEMA 2012).  There are several other similar damage scales in the literature, but the focus herein is on those which see use in the damage surveys encountered in this study.  In all cases, the damage scales separate the continuum of possible seismic damage into discrete “damage states” which are linked to tangible damage descriptions or repair needs (eg. “Parapets fell, Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 71 separation of veneer from backing” or “extensive structural damage requiring repair…”).  In some cases, the damage states are also related to loss values.  4.3.1 Wailes and Horner As aforementioned, this scale was developed in conjunction with an assessment of damage from the 1933 Long Beach earthquake.  The scale uses five damage states (A through E), each associated with a description and example photograph for each type of structure.  Table 4.1 provides the damage state descriptions and illustrative photos for URM buildings.  Note that available reproductions of the original photos are low-quality and so more recent, higher-quality photos are provided.  The photos are from various sources, as noted, and were selected to be similar to the original photos. Rutherford & Chekene (1997) provides the original photos.   4.3.2 ATC-13 The ATC-13 report was developed by the Applied Technology Council and published in 1985 (ATC 1985).  The main purpose behind the study was to develop a loss estimation methodology that could be applied on a regional scale throughout California. Before discussing its use in damage surveys, we will discuss its background and theory as a loss estimation tool to better understand what the collected damage statistics should represent.  This is a worthwhile endeavour, since the majority of the damage statistics in this study use the ATC-13 damage scale.  ATC-13 was considered a landmark study in earthquake loss estimation and much research effort was subsequently expended throughout North America to adapt it to other regions (Blanquera 1999, Onur 2001, Thibert 2008).  Until approximately the year 2000, ATC-13 was the most commonly used methodology in North America.  Even today, its roots are embedded in newer methodologies, such as HAZUS (FEMA 2012), which have become more popular. The main difference between the ATC-13 and HAZUS methodologies is that ATC-13 was based primarily on expert opinion, while HAZUS takes a more analytical approach and is based on nonlinear static analysis of prototypical buildings.    The ATC-13 methodology accounts for various forms of losses, including: 1) Direct losses (losses due to damage caused by earthquake effects, eg. repair costs) 2) Indirect losses (loss of functionality, eg. downtime)  3) Social losses (eg. casualties)  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 72 Table 4.1 – Wailes and Horner Damage Scale Damage State Description  A Undamaged or small cracks only  (from Ingham & Griffith, 2011)   B Parapets fell, separation of veneer from backing  (from Ingham & Griffith, 2011)   C Major damage to less than 50% of bearing walls  (http://reidmiddleton.wordpress.com)   D Major damage to more than 50% of bearing walls  (from Ingham & Griffith, 2011)   E Total loss; collapse or severe damage requiring demolition (http://www.newswire.co.nz/2011/02/day-4-the-search-for-survivors-continues/)   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 73 The losses are predicted as a function of the earthquake intensity (i.e. the hazard) and the expected response of the building to these demands (i.e. the vulnerability).  Therefore, the methodology is a form of risk analysis.  As will be seen herein, compiling and interpreting earthquake damage statistics is a different but related problem.    At the time of the development of ATC-13, ground motion recordings were much more scarce than today. Due to this lack of availability of quantitative engineering measurements of earthquake intensity in real earthquakes, it was decided to use the Modified Mercalli Intensity (MMI) scale as the measure of earthquake intensity (ATC 1985). MMI contour maps have long been routinely developed by geological/seismological agencies such as the United States Geological Survey based on “felt reports” from earthquakes (i.e. witness accounts of shaking and damage). Details on the MMI scale are commonly available in earthquake engineering literature and are not presented here.  Various building prototypes (78 in total for ATC-13) are defined to represent the vulnerability.  The classifications are general in nature (eg. “High-Rise Concrete Shear Wall Building” or “Low-Rise Unreinforced Masonry Building”) and serve as a baseline indicator of the expected losses for a given building; the focus of the classification is on the seismic force resisting system.  Unfortunately, these classifications do not account for several particularities of structures such as irregularities of the seismic force resisting system or the era of design and construction.  While these classifications have proven effective on a regional scale, a more detailed assessment is often warranted when examining a specific subset of the building stock, such as strictly unreinforced masonry buildings.  In ATC-13, damage is discretized into seven Damage States ranging from “None” to “Destroyed.”  Each damage state is associated with a given damage factor5 range (Damage Factor = Repair Cost/Building Replacement Value) and a Central Damage Factor (CDF), which is usually taken as the arithmetic mean Damage Factor for that range, although some have instead advocated the use of the geometric mean (Wiggins, Breall and Reitherman 1994). Results are presented in what are called Damage Probability Matrices (DPMs).  Table 4.2 provides the damage states and Table 4.3 is an example of such a DPM.  This procedure also lends well to processing observed statistics.  However, in ATC-13 the DPMs were developed using expert opinion: more specifically the Delphi survey method was employed.  This method was employed due to a lack of observational data and is discussed in detail in the original ATC-13 document.  Since this time much data has                                       5 The terms “damage ratio” and “damage factor” are both used in the literature.  In this thesis, they can be taken as synonyms Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 74 become available and DPMs based on observed data have been produced in the literature, such as will be reviewed and undertaken herein.   Table 4.2 – ATC-13 Damage States Damage State Description Damage Factor Range Central Damage Factor 1 None No Damage 0% 0.0 2 Slight Limited localized minor damage not requiring repair 0-1% 0.5% 3 Light Significant localized damage of some components generally not requiring repair 1-10% 5.5% 4 Moderate Significant localized damage of many components warranting repair 10-30% 20% 5 Heavy Extensive damage requiring major repairs 30-60% 45% 6 Major Major widespread damage that may result in the facility being razed, demolished, or repaired 60-100% 80% 7 Destroyed Total destruction of the majority of the facility 100% 100%  Table 4.3 – Example Damage Probability Matrix  (Modified from ATC-13) Damage State Damage Factor Range CDF Probability of Being in Damage State  For A Given MMI [%] VI VII VIII IX X XI XII 1 None 0% 0.0 95 49 30 14 3 1 0.4 2 Slight 0-1% 0.5% 3 38 40 30 10 3 0.6 3 Light 1-10% 5.5% 1.5 8 16 24 30 10 1 4 Moderate 10-30% 20% 0.4 2 8 16 26 30 3 5 Heavy 30-60% 45% 0.1 1.5 3 10 18 30 18 6 Major 60-100% 80% - 1 2 4 10 18 39 7 Destroyed 100% 100% - 0.5 1 2 3 8 38 MDF = Σ[P(DSi)*CDFi] =  0.2% 2.9% 6.6% 14.3% 25.9% 42.4% 78.0% Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 75 In examining Table 4.2 it can be seen that the Damage Factor Range increases with each damage state, which represents the increasing uncertainty in the losses as damage increases.  For instance, the range of possible repair costs for a building with only “significant localized damaged, generally not requiring repair” must be small (for example some painting or plaster repairs).  Conversely if “extensive damage requiring major repairs” is present, then the scope of potential repair work is much more unclear.    In examining Table 4.3 it is worth noting that the DPM provides a framework that explicitly considers the uncertainty in the relationship between ground motion intensity (such as MMI) and damage. For example, it can be seen that if a subject building experiences ground shaking of intensity of MMI VIII, there is a 40% probability that the building will be in the “Slight” damage state, a 16% chance of being in a “Light” damage state, and so on.  Another way of expressing the probability would be to say that there is an 86% probability that the building will fall into a damage state of “Light” or lower (i.e. 40%+30%+16%).  One can also readily compute the standard deviation of the MDF and thus establish reasonable upper and lower bounds for the expected damage.    4.3.3 ATC-20 ATC-20 (ATC 1989) is a post-earthquake safety evaluation method commonly employed by communities throughout North America and abroad.  The methodology is designed to facilitate rapid assessment of the safety of buildings for occupants and passersby.  It involves a somewhat cursory review of the building, which may or may not involve interior access.  The primary end result is a ‘placard’ that is posted at the building, indicating the level of permitted access.  Three different placards are possible, as shown in Table 4.4.    Although the methodology is not specifically intended for damage assessment, the safety of a building for occupancy is obviously closely related to the damage and many researchers have used tagging data in damage surveys. Section 4.4.1 provides a discussion of the associated merits and drawbacks of this form of data.    Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 76 Table 4.4 – ATC-20 Placards Placard Description  Green “Inspected” The building was reviewed and no access restrictions were imposed  Yellow “Restricted Use” Access is only for possession retrieval and access to portions of the building may be prohibited  Red “Unsafe” No access is permitted   4.4 Collecting Damage Data As previously noted, the most useful form of damage survey is one in which observations are systematically collected and, thus, the observed performance can be objectively quantified and statistically analyzed for various hypotheses of interest, which entails going beyond the commonly completed reconnaissance papers with case studies or anecdotal evidence. It entails collecting a representative sample of all structures for the given population of interest (in this case, unreinforced masonry bearing wall buildings), not just those that performed noticeably well or poorly. As a preface to the damage statistics presented later in this chapter, it is worth discussing the various sources.  Damage statistics are typically collected in two types of surveys:  Post-earthquake safety evaluations performed for the local building authority  Ad-hoc surveys for research purposes Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 77 4.4.1 Statistics from Post-Earthquake Safety Evaluations Damage statistics are often collected during post-earthquake safety evaluations, but the primary focus in this case is the safety of potential building occupants and passersby.  There are often two outcomes: an ATC-20 placard indicating the level of access permitted for the building and a damage percentage, representing the cost of repairs as a fraction of the building replacement value.  The ATC-13 damage states (see Section 4.3.2) are quite commonly used to describe the anticipated cost of repairs. However, there is significant uncertainty in the results for the following reasons: 1) The background and experience of those completing the assessments is variable.  Therefore, the accuracy of any individual assessment could be affected 2) Many assessors are usually involved.  Therefore there can be inconsistencies in the interpretations of damage.  As noted by Lizundia (1993), “one assessor may look at a cracked parapet and call it a 1% loss, assuming that repointing the mortar cracks would be necessary.  Another may see the same cracks, imagine that a complete parapet retrofit will be required, and call it a 10% loss” 3) The level of detail in the assessment can vary.  For example, the ATC-20 methodology offers both “Rapid” and “Detailed” assessments.  In the former of the two, the building is generally not entered.  Placard statistics for URM buildings in the Northridge earthquake for initial inspections vs. reinspections are compared in Figure 4.1 (Rutherford & Chekene 1997).  Obviously a rapid assessment (from the exterior) would produce less reliable estimates of the damage, as the interior structural and nonstructural components are not inspected              Figure 4.1 – URM Placard Statistics Comparison from Northridge Earthquake (Reproduced From Rutherford & Chekene, 1997)  48216866541137480100200300400500600Green Yellow RedNumber of BuildingsAssigned PlacardFirst Inspection First Re-InspectionChapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 78 4) There can be sampling bias, because buildings that are lightly damage are less likely to be assessed  4.4.2 Statistics from Ad-Hoc Damage Surveys Academics, building officials, and practicing engineers may perform damage surveys specifically for the purpose of compiling damage statistics and quantifying seismic performance/losses.  These studies often are more detailed, but may not cover as large a fraction of the total building population as resources can be more limited and the assessments are more in-depth.  A related merit is that there is less chance of inconsistencies between assessors, as there are often fewer individuals and greater communication between them.  The level of detail in these studies can also vary, yielding significantly different results.  Figure 4.2 compares damage ratios for two different surveys from the 1983 Coalinga earthquake: as noted by Lizundia (1993), the study by Shah et al. (1984) consisted of a one-day walk-through by a team of engineering students, while a study by Reitherman et al. (1984) was sponsored by the National Science Foundation and included a walk-through, materials testing, a photo review process, and discussion between at least two engineers before a damage level was assigned.  As shown in the figure, the mean damage ratio of the former was 50%, compared to 35% for the latter.                     Figure 4.2 – Damage Ratio Comparison from Coalinga Earthquake (Reproduced From Lizundia, 1993)  3%16%21%60%0%8%18%75%0%10%20%30%40%50%60%70%80%90%100%0-10% 10-30% 30-60% 60-100%% Buildings in Damage StateDamage StateReitherman et al, 1984 Shah et al, 1984(MDF=50%)(MDF=35%)Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 79 It should be noted that some researchers (Ingham and Griffith 2011b) have combined the “none” and “slight” damage states in their surveys, as it may be difficult to ascertain that there is indeed no damage whatsoever.  At the opposite end of the damage scale, one could argue that it is also difficult to distinguish between the “Major” and “Destroyed” damage states.  One case in point is noted by Lizundia (1991): in this damage survey of URM buildings in the Loma Prieta earthquake of 1989, a portion of the data came from post-earthquake safety evaluations completed by San Francisco building officials.  16 of 1923 unretrofitted URM buildings were originally assigned damage states of “Major” or “Destroyed” (exact figures not disclosed) but as noted by Lizundia “the follow-up survey indicated that none of these 16 buildings were actually destroyed and a better estimate of the damage they suffered [was] approximately 60% of their replacement costs.”  Such sources of error and uncertainty in assigning damage states is best addressed by careful re-examination of the data and, when possible, comparing the results of more than one damage rating method.  There is clearly substantial uncertainty regardless of the chosen evaluation method.  Nonetheless damage statistics play an important role in assessing and improving earthquake performance of structures, as they can account for effects not adequately captured by analytical models.  4.5 Ground Motion Intensity Measurements With the damage data collected, one could simply compile overall damage statistics for the entire sample and get a sense of how the structures performed.  However, results are often much more useful when sorted by the intensity of seismic demand that each specific building experienced.  Since ground motion is only measured at a few select locations throughout the region, one must estimate the ground motion experienced at each site.  This section discusses the various ground motion intensity measurements (IM’s) that were used in this study and manners of estimating an IM at a given site.  4.5.1 Intensity Measurements Identifying ground motion parameters that are good indicators of damage has itself received a fair amount of attention in the literature.  This section discusses the IM’s that were encountered in this study.   4.5.1.1 Modified Mercalli Intensity The Modified Mercalli Intensity (MMI) scale was traditionally used as the intensity measurement in compiling damage statistics, as it was the only commonly available Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 80 measurement (ATC 1985).  The validity of MMI as an indicator of damage has been questioned by some (Rutherford & Chekene 1997, Onur 2001, Thibert 2008).  The basis of this criticism is that MMI – which is partially measured in terms of damage (eg. “…parapets will topple…”) – is used as a predictor of damage.  Another issue is that the MMI scale perhaps does not provide enough “resolution” in the demand: for example, MMI IX is typically associated with peak ground accelerations of 0.65-1.24g (USGS 2011).  If a jump in damage occurs in this range, there is significant uncertainty as to the seismic demand at which the jump actually occurred.    More recently the Instrumental Intensity (IMM) scale was developed by Wald et al. (1999) for automated generation of shakemaps by the United States Geological Survey.  The scale was developed through a regression analysis of MMI versus PGA and PGV for several California earthquakes, including all of those discussed in this report.  The regression equations are as follows:  IMM=3.66*log(PGA) – 1.66, IMM ≤ 7 (4-1)  IMM=3.47*log(PGV) + 2.35, IMM > 7 (4-2)  Although the use of IMM offers a somewhat less subjective measurement, it must be remembered that it is still closely related to the observed damage (especially for the subject earthquakes that were included in the regression analysis) and thus the circularity issue would still apply when used as an intensity measure for damage surveys.  4.5.1.2 Peak Ground Response Values The use of peak ground acceleration (PGA) has been investigated in some studies relating damage to ground motion (Lizundia, Dong and Holmes 1993, Rutherford & Chekene 1997, King, et al. 2005).  The merits of using PGA include its simplicity, lack of period-dependence, and the fact that associated hazard values are readily available; this latter item is important when the results are to be used in loss estimates, as is the case in this study.  However it has drawbacks in that PGA is not, strictly speaking, associated with soils amplification factors (NEHRP 1994). Applying short period amplification factors seems reasonable and has been performed in some cases, such as HAZUS (FEMA 2012). Another drawback is that PGA is sensitive to high-frequency ground motions, which may not impact the structure significantly (King, et al. 2005).  Peak ground velocity (PGV) has been investigated by others (Lizundia, Dong and Holmes 1993, King, et al. 2005) and was shown to be a reasonably good indicator of Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 81 damage.  Again, the fact that PGV is not period-dependent would be considered a merit.  However, most modern seismic hazard analyses do not include PGV values and, thus, it would be necessary to first perform a seismic hazard analysis before applying the results to a loss estimate.  4.5.1.3 Spectral Response Values The use of an elastic spectral response value has merit in that it can represent more accurately the response of the building.  The use of five percent (5%) damped spectra is typical.  Both spectral acceleration (Sa) and spectral velocity (Sv) have been shown to correlate reasonably well with damage (Lizundia, Dong and Holmes 1993, King, et al. 2005).  Spectral displacement (Sd) has also been shown to be a good indicator of damage and is the primary intensity measurement in HAZUS (FEMA 2012).  Of course, the period-dependence can be a potential drawback if an inappropriate period is selected and there is a large spike at this point of the response spectrum for the subject ground motion. Lizundia et al. (1993) made use of a building code period expression in an attempt to address the issue of period selection. Although some improved correlation was noted, the expression used was based on the number of storeys, which is fundamentally flawed for typical URM buildings with flexible diaphragms.  Moreover, many buildings may not be adequately connected so as to respond as a single unit. Using multiple periods (i.e. different periods for different buildings) also presents a complication when performing loss estimates.    Given the aforementioned issues and the considerable uncertainty in the remainder of the process, it is the author's opinion that the choice of period for spectral values should be governed largely by simplicity. For example, representative short (T=0.2sec) or long (T=1.0sec) periods could be used, both of which can be directly associated with seismic hazard values that underlie current Canadian codes (Adams and Halchuk 2003).  The use of a readily available spectral value eliminates the added effort of performing a probabilistic seismic hazard analysis.  Of the two candidate periods, Sa(1sec) is thought to be the more representative of URM buildings and is commonly used in a variety of URM seismic assessment procedures of both diaphragms (ICC 2012a) and out-of-plane wall response (Derakhshan 2011, ICC 2012a, Penner 2013).  The more pronounced effect of soils on Sa(1) than Sa(0.2) is also consistent with the observed profound effect of soils on the performance of URM buildings, as will be subsequently demonstrated in this chapter. Another reason for using Sa(1) instead of Sa(0.2) (or PGA) is the reduced spatial variability.  Figure 4.3 from Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Turner et al. (2010) provides an example plot of short period (0.2 second) and long period (1.0 second) spectral accelerations for recorded values versus values at a distance of 500 feet.  This assumes similar soil conditions, but there is alredispersion, particularly for the short period.             4.5.1.4 Cumulative Absolute VelocityCumulative absolute velocity the acceleration time history, as shown below: CAV = ∫|a(t)|dt  Its use as a damage indicatorand Herraiz 1997, Moon, et al. 2014)is modified to include some threshold value of accelerationcontribution to damage.  The motivation for the use of CAV is that it captures intensity and duration, as well as a critical value of acceleratiothe case of multiple earthquakesshaking.   While CAV represents a fairly robust and accurate IM, it has drawbacks in its easuse. There is much less literature equations and the effects of soilsstructural design and cannot be relative to, for example, current seismic design standards. Figure 4.3 – Spatial Variation in Spectral Acceleration Statisticsady significant   (CAV) is defined as the integral of the absolute value of   has been investigated in various studies (Cabanas, Benito , with good results.  Typically, the above definition , below which there is no n below which there is no effect, it can be used to capture the cumulative ground regarding CAV, including ground motion prediction .  It is also not a meaningful parameter in terms of readily communicated to decision-makers     (From Turner et al., 2010)  82 (4-3) .  In e of  in terms Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 83 4.5.1.5 Selection of Intensity Measure for This Study As can be seen, there is a great variety of intensity measures.  Those mentioned here provide only those most relevant to this study.  The aforementioned works by Lizundia et al. (1993) and King et al. (2005) provide a more in-depth treatment of the subject.  For the purposes of this study, a single intensity measure must be selected as the basis for all damage statistics and loss estimates.    MMI or instrumental intensity has a long history of use in damage surveys, but would require a seismic hazard analysis for application in loss estimates; there is also the aforementioned issue of circularity in using MMI as an indicator of damage.  For these reasons, it was decided to not use MMI or IMM as the intensity measure.  CAV is the most appealing from theoretical and forensic standpoints, but introduces significant additional effort (and uncertainty).   While it is certainly possible to overcome the aforementioned difficulties in using CAV, the purpose of this study was not to develop a motion-damage relationship using the best possible IM, but rather to use the results of a reasonable motion-damage relationship to address a broader host of issues, such as quantifying the economic and societal impacts of URM risk mitigation. Due to the breadth of topics addressed in this study, it was decided to not pursue using CAV as the intensity measure.  Peak ground acceleration and spectral acceleration values offer the greatest ease of use, since they are routinely mapped for earthquakes. The intermediate results of this study (eg. fragility curves) will also be most useful to others if they are in terms of one of these parameters because such values can be readily extracted from building codes.  Of these two, spectral acceleration is thought to be more representative of building response and damage.  It is noted that Penner (2013) found Sa(1) to be the preferred IM for URM out-of-plane assessment.  Ultimately, it was decided to use spectral acceleration at a period of one second, Sa(1), as the intensity measure for use in this study.    4.5.2 Estimating the Intensity Measurement at a Site Having selected a given intensity measure, a value must somehow be assigned to each building.  Several methods have been employed in the literature, including the following:  Closest station  Interpolation between stations  Statistically-derived values   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 84 4.5.2.1 Closest Station If the recorded ground motion is sufficiently close to the building, it can be assumed that the ground motion is the same at both locations.  However, the distance must be quite small: a limit of 1000 feet has been used in past studies (ATC 2001, King, et al. 2005). While this method is convenient to implement, it potentially necessitates the exclusion of many buildings from an overall sample.  For example, King et al. made use of the Rutherford and Chekene (1997) damage statistics for rehabilitated URM buildings in the Northridge earthquake, but King et al. reported that only 50 of about 800 buildings qualified for use in the study.  This is clearly not a sufficient sample size to produce statistically meaningful results, especially after filtering for various criteria (eg. type of retrofit, storey height, etc.).  The use of closest station values are likely more appropriate for a building-specific forensic-type study.  4.5.2.2 Interpolation Between Stations Interpolation between recording stations has been employed in some studies (Lizundia, Dong and Holmes 1993, Rutherford & Chekene 1997).  Of course, one could question the accuracy of the interpolation on the basis of the earlier discussion of spatial variability; this would likely be a crucial issue in a forensic-type study of a specific building, but it should be recalled that in this study, several hundred buildings are typically included in a bin and the results focus primarily on the mean value which is captured quite well given the high number of samples.  Furthermore, in an effort to address issues with spatial variability, soil conditions, and proximity to the epicenter, Lizundia et al. (1993) took the following steps:  Interpolation between stations and buildings using ground motion prediction equations rather than, for example, linear interpolation  Interpolation only between stations with soil conditions similar to the buildings  Weighted interpolation between stations based on proximity (for example Lizundia weighted the values based on the square of the distances)  4.5.2.3 Statistically-Derived Values A more rigorous method of “interpolating” is to derive values based on conditional probability theory.  This process combines theoretical values (based on ground motion prediction equations) with the measured values to generate expected values (and, if desired, a distribution) for the ground motion parameter in question.  In short, the method yields results that approach the measured ground motion parameter at the stations and that approach the calculated value as the distance from stations increases.  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 85 The basis for the gradient is the spatial correlation as a function of distance. Figure 4.4 shows such a function from the work of Goda and Hong (2008).  Bradley and Hughes (2013) outline the theory in detail and completed such an exercise for the Peak Ground Accelerations during the various earthquakes of the Canterbury earthquake sequence. Figure 4.5 reproduces the map for the February earthquake from the original paper.                                  Figure 4.5 – Map of Statistically-Derived PGA for Christchurch Earthquake (From: Bradley and Hughes 2013) Figure 4.4 – Spatial Correlation Between Ground Motion Parameters (From: Goda and Hong 2008)  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 86 This is obviously a more scientifically justified method of estimating the IM at the subject buildings than the interpolation method described in the previous section.  However, one key weakness of this method is that a single site class must be assumed in the ground motion prediction equation.  Depending on the regional geology, this may or may not be a reasonable limitation.  4.5.2.4 Selection of Method for Estimating IM at Sites in This Study Similar to the choice of IM, it was desired to select a method of estimating that would be reasonably simple for all the available data and yet sufficiently accurate so as not to invalidate the conclusions reached (e.g. whether or not certain retrofit measures are economically justified for various stakeholders).  To this end, the results of the three aforementioned methods were investigated for the Canterbury database (as presented in Section 4.7.4 and 4.8.4).  Figures 4.6 and 4.7 provide histograms of the number of buildings versus PGA.  Note that in Figure 4.7, some bins have been aggregated and others excluded where there are insufficient samples to generate the damage probability matrices and motion-damage relationships (as presented in Section 4.8).  Note that even for the first histogram, the IM’s estimated at each site have been rounded to the nearest .05g.                    Figure 4.6 – Number of Buildings vs. PGA for Canterbury Database (Initial) 0501001502002503003504000.05 0.25 0.45 0.65 0.85 1.05 1.25 1.45Number of BuildingsPGA [g]Closest StationWeighted InterpolationConditional Median (Bradley, 2013)Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 87    The ‘Closest Station’ estimate clearly yields quite different results from the other two.  However, the ‘Weighted Interpolation’ and ‘Conditional Median’ results are quite comparable, especially after consolidating the bins; in examining Figure 4.7, there appears to be about a 10% discrepancy between these two methods and some of this is due to the binning process itself.  Another source of error is that the initial ground motion parameters may not come from the same source and could be subject to minor differences arising from processing of the accelerograms, which is especially true for PGA.    One noticeable difference between the Weighted Interpolation and the Conditional Median results is that there is much less variation in the latter (see Figure 4.6), which seems to contradict the earlier discussion on spatial variability.  One possible explanation is the underlying assumption of a single representative soil condition for the generation of the PGA map.  In a forensic-type analysis of a specific building, only the most accurate of estimates would be appropriate.  The data used herein, however, contains hundreds (if not thousands) of samples and we are primarily interested in the average damage/ground motions (as will be seen in Section 4.8). As such, it was concluded that the Weighted 0%10%20%30%40%50%60%70%80%90%100%0.05 0.2 0.35 0.5 0.65 0.8 0.95Fraction of BuildingsPGA [g]Closest StationWeighted InterpolationConditional Median (Bradley, 2013)Figure 4.7 – Percentage of Buildings vs. PGA for Canterbury Database (Final) Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 88 Interpolation method provided an acceptably accurate estimate for the purposes of this study. The decision to use the Weighted Interpolation method (in lieu of the more rigorous Conditional Median method) represented a significant simplification, as it eliminated the need to generate conditional median maps in terms of Sa(1) for all of the databases.  4.6 Developing DPMs & Fragility Curves from Damage Statistics With the damage data collected and the ground motions assigned to each building, the next step is to compile the statistics in a manner that shows the relationship between ground motions and damage.  The two most common forms are damage probability matrices (DPMs) and fragility curves.  This section describes how the data was processed into the necessary forms and how the fragility curves were fit to the data.  4.6.1 Processing The Damage Data This section discusses the process in terms of the ATC-13 damage states, but the methodology is valid regardless of the chosen damage scale.  Refer to the previously shown Table 4.3 as an example DPM (note that the actual numbers of buildings are not shown – only the percentages). The DPM is constructed in the following manner: 1) The earthquake intensities are discretized into bins (eg. 0.3g < PGA < 0.4g); note: if using MMI, the “discretization” would simply be by MMI Level 2) The damage measure is also discretized into bins.  In this case, the discretization is already completed by the data collection method, in the form of the ATC-13 Damage States 3) The total number of buildings in each intensity/damage bin is calculated 4) For each intensity level, the number of buildings (#bldgs) in each damage state is then normalized by the total number of buildings in this intensity bin, giving a percentage of buildings (%bldgs) in each damage state for each intensity 5) Calculate the Mean Damage Factor (MDF) by multiplying the %bldgs in each damage state by its associated Central Damage Factor (CDF) 6) Finally, an optional step is to compute the sample standard deviation of the MDF at each intensity by taking the second moment of each damage state about the MDF (i.e. SMDF = Σ[(CDFDS=i*#BldgsDS=I -MDF)2]/Σ[#BldgsDS=I -1] This level of processing alone is sufficient to make several observations; for example, one could note that the MDF at a given intensity is much lower for a set of modern steel buildings than for a set of older URM buildings (i.e. the modern buildings suffered less damage, although the difference in replacement value also plays a role).  Further processing is then performed to obtain the desired fragility curves. Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 89 The fragility curve for each damage state is obtained in the following manner and is illustrated in Figure 4.8, below.  It is the cumulative probability distribution curve given in the lower right hand corner of the figure, and the process of translating the data is as described below:   1) For the given damage state, the fraction (or percent) of buildings equaling or exceeding the DS is calculated for each intensity level (in this case, PGA); the calculated values represent the conditional probability of being in this DS or greater, given the intensity level 2) The process is repeated for each damage state and results are plotted on a graph of PGA versus Probability of Reaching/Exceeding the given DS; note: the PGA to be plotted is the mean for the bin (eg. for the bin of .3g < PGA <.4g, the results are plotted at .35g) 3) The data is then fitted with an appropriate cumulative probability distribution. This may be accomplished in several ways and is discussed next                   4.6.2 Curve Fitting With the data translated into the appropriate form, the next step is to fit the cumulative probability distribution.  Various probability distributions have been used in the literature: in ATC-13, the beta distribution was used, while others have used the lognormal distribution (King, et al. 2005).  It was noted that the primary reason for Figure 4.8 – Example Construction of Fragility Curve (From King 2005)  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 90 using the beta distribution in ATC-13 was the fact that the lognormal distribution did not fit the data well at the lower damage states.    There are also many ways of fitting the chosen distribution to the data.  Some common methods are as follows:  Moment Matching – In this methodology, k number moments (i.e. mean, variance, skewness, kurtosis) of the distribution are matched to the sample data, where k is the number of unknown distribution parameters. For the lognormal distribution, there are two unknown parameters (µ,σ), so typically the first two moments are matched.  Other moments could be matched instead  Moment Optimization – Exact matching of selected moments as noted above will not necessarily give appropriate matches for the remaining (unmatched) moments.  This method aims to address this by instead minimizing the total mismatch of all the moments and is therefore an optimization method.  A common objective function is argmin{(m-µ)2 + (s2-σ 2)2 +…} (where m,s= sample estimates; µ,σ= distribution parameters)  Nonlinear Regression – In regression analysis, we seek to minimize the errors between the observed data and the predicted values of the dependent variable (i.e. the Y variable, or in this case the cumulative probability from the lower bound to X).  The most common method is the “least squares" method, in which the objective is simply to minimize the sum of the errors (i.e. the difference between the data and the fitted curve) squared  Maximum Likelihood Estimation (MLE) – In MLE, we seek to obtain the parameter values, which make observed data the most probable (i.e. the likelihood function is maximized). This method is not reviewed here except to note that, for the beta distribution, the method is quite sensitive to the upper and lower bounds and thus is not recommended for curve-fitting of the beta distribution (Simaan and Halpin 1994) For this study it was decided to do fitting by nonlinear regression.  One of the major advantages of using nonlinear regression is that the individual terms in the objective function (i.e. the sum of errors to be minimized) can be weighted.  As the number of buildings within a bin could vary from less than fifty to over a thousand, it was decided to weight the fitting by the number of buildings in the bin.  The weighting factor was defined as follows:  WFi=MDFi/SEi (4-4)   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 91 Where: WF ≡ Weighting Factor for bin ‘i’ MDFi ≡ Mean Damage Factor for the bin ‘i’ SEi ≡ Standard Error of the Mean for bin ‘i’   Note that the standard error of the mean is defined as the sample standard deviation divided by the square root of the number of measurements in the sample.  The theory is discussed in textbooks on elementary statistics (Navidi 2010), but the essence is that for larger samples of the same population, the variation in the mean from sample to sample is quite small even if the distribution of the population is quite wide.    The MDF is included in the weighting as a method of normalizing the variation, similar to a coefficient of variation.  It should be noted that, in some instances, the weighting was manually adjusted where the data appeared unreliable or where there were very few buildings, all with mostly the same damage rating (which would give a SE near zero and therefore a large weighting factor).  In these cases, the weighting was reduced to zero, or some nominally small value so as not to affect the results – such data points will be highlighted in the subsequent results, presented in Section 4.8.  As aforementioned, both the beta distribution and the lognormal distribution were candidates for the fitting.  Background on the distributions is commonly available in textbooks on statistics (Navidi 2010), but the major difference is that the lognormal distribution is defined by two parameters (for example the mean and standard deviation), while the beta distribution is defined by four parameters, which offers more flexibility in the shape of the distribution.  The beta distribution is also defined over a finite range: for example, one could prescribe the PGA at which 0% and 100% of the structures reach a certain damage state.  In this study, it was decided to prescribe a minimum of Sa(1)=0g and to leave the maximum value unconstrained.    Both the lognormal and beta distributions were fitted and the results were compared.  Figure 4.9 provides an example of the resulting curves for the MDF versus Sa(1) from the Canterbury buildings (see Section 4.8.4).  For each data set, it was found that there was little difference between the two distributions in the region of the data, which is intuitive.  In the regions beyond the data, the lognormal distribution consistently fell below the beta distribution, presumably because the shape of the beta distribution was less constrained.  As the higher values of the beta distribution were more consistent with physical insight (i.e. at an Sa(1) of 5.0g one would expect all unretrofitted URM buildings to be destroyed) it was decided to use this distribution in the fitting.  For Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 92 practical purposes, however, the choice between the Beta and Lognormal distribution is of minor consequence because the probability of occurrence of such extreme shaking is low enough so as to not affect the resulting loss estimates.                    4.7 Literature Review and Summary of Available Data A significant body of work on URM damage statistics was completed during the 1980’s and 1990’s as a result of the major earthquakes in California at the time.  More recently, a significant amount of data on URM buildings was collected from the 2010/2011 Canterbury earthquakes.  Because there is no similar data on western Canadian URM buildings, these sources are thought to be the best available data.  The available data and results were in various forms and, as such, a significant amount of effort was devoted to reviewing these results and working with the raw data to produce additional results that were of interest to this study.  This section reviews the work that has been completed by others. The reviews are kept reasonably brief and the reader may refer to the original documents (Deppe 1988, Lizundia, Dong and Holmes 1993, Wiggins, Breall and Reitherman 1994, Rutherford & Chekene 1997, Ingham and Griffith 2011a, 2011b) for more information.  Section 4.8 presents the new results that were prepared by the author.  Both sections are organized by the subject earthquakes. 0%10%20%30%40%50%60%70%80%90%100%0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Mean Damage FactorSa(1) [g]UNRET-BETA DISTRIBUTIONUNRET-LOGNORMALFigure 4.9 – Beta vs. Lognormal Distribution for MDF of Canterbury Buildings Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 93 4.7.1 Whittier 1987 Earthquake The Whittier Narrows earthquake occurred on October 1, 1987.  It was a ML=5.9 event and the epicenter was about 15km east of downtown Los Angeles.  Figure 4.10 shows a map of instrumental intensity, produced by the United States Geological Survey.                    4.7.1.1 Summary of Damage Survey Two URM damage surveys were performed: one by Deppe (1988) and another by Wiggins et al. (1994).  The survey by Deppe covered 2431 of the 7300 URM buildings in Los Angeles, but the damage assessment was limited to: 1) Undamaged 2) Damaged, including nonstructural, but entirely functioning (not vacated) 3) Vacated (wholly or partially) Deppe estimated that approximately 1100 buildings had been fully strengthened.  Rutherford and Chekene (1997) later estimated this number to be only approximately 800.  Buildings were sorted into those that were strengthened (to the city’s mandatory retrofit standard, Division 88, 1985) and those that were unstrengthened or partially strengthened (eg. tension ties only).  More detailed information is provided by Deppe (1988), but the following key observations summarize the findings: Figure 4.10 – Instrumental Instensity Map – Whittier Narrows Earthquake Source: http://earthquake.usgs.gov/earthquakes/shakemap/sc/shake/Whittier_Narrows  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 94  Unstrengthened/partially strengthened buildings were 50% more likely to have observable damage; this includes any damage ranging from non-structural cracking in, say, plaster ceilings, to partial collapse  Unstrengthened buildings were 170% more likely to have been wholly or partially vacated (presumably due to damage)  Residential buildings were approximately 150% more likely to have been partially or wholly vacated  The second database, by Wiggins (1994), was more detailed.  It was based on ATC-13 damage levels and MMI was the intensity measurement.  However, the intensity assigned to each building was based on Wiggins’ own attenuation equation rather than measured values as shown in the map above.  Wiggins accounted for soils in his assessment, but only on a regional basis by assigning soil types using zip codes and his own map of regional soil types.  Table 4.5, Table 4.6, and Table 4.7, show the resulting breakdown by MMI and ATC-13 damage state for unstrengthened, partially strengthened, and fully strengthened buildings (to Los Angeles requirements), respectively.  As can be seen, the vast majority (over 95%) of the buildings fell into the MMI VII range.  However, the data set is valuable because it has a reasonable number of buildings in the partially strengthened state, which was found to be rare in the literature that was reviewed.  This data will be further processed in Section 4.8.1.   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 95  Table 4.5 – Whittier DPM – Unstrengthened  (From Wiggins, 1994) Damage State Geom. DF Number of Buildings in MMI 5.5-6.0 6.0-6.5 6.5-7.0 7.0-7.5 7.5-8.0 8.0-8.5 1 None (0%) 0% 9 29 18 213 800 0 2 Slight (0-1%) 0.5% 1 1 0 45 66 0 3 Light (1-10%) 3.2% 0 3 1 52 230 0 4 Moderate (10-30%) 17.3% 0 0 2 29 122 0 5 Heavy (30-60%) 42.4% 0 0 1 2 26 0 6 Major (60-100%) 77.5% 0 0 1 2 6 0 7 Destroyed (100%) 100% 0 0 0 0 0 0 MDF= .06% .031% 6.86% 2.71% 3.56% --   Table 4.6 – Whittier DPM – Partially Strengthened  (From Wiggins, 1994) Damage State Geom. DF Number of Buildings in MMI 5.5-6.0 6.0-6.5 6.5-7.0 7.0-7.5 7.5-8.0 8.0-8.5 1 None (0%) 0% 1 1 4 39 132 1 2 Slight (0-1%) 0.5% 0 0 0 6 21 0 3 Light (1-10%) 3.2% 2 1 1 9 57 0 4 Moderate (10-30%) 17.3% 0 0 0 5 29 0 5 Heavy (30-60%) 42.4% 0 0 1 0 2 0 6 Major (60-100%) 77.5% 0 0 0 0 0 0 7 Destroyed (100%) 100% 0 0 0 0 0 0 MDF= 2.11% 1.59% 7.69% 2.01 3.23% 0.00%   Table 4.7 – Whittier DPM – Fully Strengthened  (From Wiggins, 1994) Damage State Geom. DF Number of Buildings in MMI 5.5-6.0 6.0-6.5 6.5-7.0 7.0-7.5 7.5-8.0 8.0-8.5 1 None (0%) 0% 2 6 9 98 256 0 2 Slight (0-1%) 0.5% 0 0 0 6 11 1 3 Light (1-10%) 3.2% 0 0 1 11 53 0 4 Moderate (10-30%) 17.3% 0 0 0 0 16 0 5 Heavy (30-60%) 42.4% 0 0 0 2 4 0 6 Major (60-100%) 77.5% 0 0 0 0 0 0 7 Destroyed (100%) 100% 0 0 0 0 0 0 MDF= 0.00% 0.00% 0.33% 1.06% 1.83% 0.5%   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 96 4.7.2 Loma Prieta 1989 Earthquake The Loma Prieta earthquake occurred on October 17, 1989.  It was a MW=6.9 event and the epicenter was about 14km northeast of Santa Cruz.  Figure 4.11 shows a map of instrumental intensity, produced by the United States Geological Survey.                        4.7.2.1 Summary of Damage Survey Lizundia (1991) compiled damage statistics from various cities throughout California, stretching approximately from Salinas (in the south) to Oakland (in the north). The survey essentially covered all areas of MMI VII and greater.  Lizundia et al. (1993) later conducted various statistical analyses on the damage statistics.  The different levels of data are noted as follows: 1) Level 1 data includes the most buildings (4824 throughout 113 cities); however, it is limited to a “general damage status,” which is the number of buildings that were damaged, vacated, or demolished 2) Level 2 data includes 2356 buildings throughout nine cities; it includes ATC-13 damage states, basic building information such as the number of storeys and occupancy, and soils information Figure 4.11 – Instrumental Instensity Map – Loma Prieta Earthquake Source: http://earthquake.usgs.gov/earthquakes/shakemap/nc/shake/LomaPrieta/ Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 97 3) Level 3 data covers only buildings in San Francisco, including 896 buildings; it includes more detailed information on cracking patterns, veneer failures, corner damage, and more The vast majority of the buildings in this survey were unstrengthened, as most communities had not yet implemented mandatory retrofit ordinances.  Strengthening status statistics from the Level 2 database are as follows:  2110 buildings are noted as unstrengthened  32 buildings are noted as partially strengthened  29 buildings are noted as fully strengthened  1310 buildings have unknown strengthening statuses  The analyses undertaken by Lizundia et al. (1993) focused primarily on unstrengthened buildings and, therefore, excluded buildings in the latter 3 categories.  Note that an additional 1085 buildings from San Francisco were excluded by Lizundia et al (1993) because these buildings had received parapet strengthening.  For this study, the risk reduction afforded by bracing parapets was of great interest, as the industry sponsor (VCHT, as introduced in Chapter 1) was in the process of developing a new incentive program for parapet bracing at the time of this study.  Analysis of this data, and more, was completed by the author and is presented in Section 4.8.  4.7.2.2 Level 1 Data The level 1 data is not directly meaningful in terms of quantifying damage, as it does not provide numeric estimates of loss/damage.  However, it is useful because it provides results that are easily appreciated in terms of their impact on a community.  Figure 4.12 is reproduced directly from Lizundia et al. (1993) and shows the fraction of buildings that were be damaged, vacated, and demolished.  Although not noted in the original document, it appears that the figures did not include data from San Francisco or Oakland. This is presumably because of the large number of buildings (together, they represent approximately 44% of the 4800+ building) and because the soils were known to be quite variable.  The key point of the figure is that there is a large jump in damage at MMI VIII.  It should be noted that the MMI VIII sample represents just 86 buildings in total, while the MMI VII and VI samples represent 1108 and 1499, respectively.  Nonetheless, one can imagine the impact on the communities that experienced MMI VIII, such as Santa Cruz, Los Gatos, or Watsonville.   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 98                 4.7.2.3 Level 2 Data (ATC-13 Damage Statistics) The level 2 data comprised much of Lizundia’s work and resulted in damage probability matrices, based on ATC-13 damage states and MMI.  Unlike the Whittier data (see Section 4.7.1), the Level 2 Loma Prieta data was binned into whole-MMI ranges, rather than half-MMI ranges (such as 6.0-6.5).  Although the earthquake produced a relatively high level of shaking intensity near the epicenter, much of the data again fell into the lower MMI ranges, with 91% of the data in the VI-VII range; only 8% (179 buildings) fell into the VIII range, and 1% (27 buildings) fell into the IX range.  The resulting DPM from Lizundia et al. (1993) is reproduced here as Table 4.8, except that the MDF is recalculated using the geometric DF rather than the CDF.  Table 4.8 – Loma Prieta DPM - Unstrengthened  (Modified From Lizundia et al., 1993) Damage State Geom. DF Number of Buildings in MMI VI VII VIII IX 1 None (0%) 0% 409 1199 82 0 2 Slight (0-1%) 0.5% 98 141 20 6 3 Light (1-10%) 3.2% 53 1111 29 14 4 Moderate (10-30%) 17.3% 24 34 12 3 5 Heavy (30-60%) 42.4% 5 40 33 3 6 Major (60-100%) 77.5% 3 8 2 1 7 Destroyed (100%) 100% 0 0 1 0 MDF= 1.82% 2.17% 10.98% 11.27% Figure 4.12 – Level 1 Damage Data for Loma Prieta Earthquake (Reproduced from Lizundia et al., 1993) Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 99 Although this data set is also somewhat limited to lower levels of shaking intensity, it is still quite a bit more diverse in ground motion intensity than was the data from the Whittier earthquake.  Since 95% of the Whittier data fell into the MMI VII range (7.0-7.5 and 7.5-8.0), we can make a conclusive comparison only for this range: combining the above noted bins from Table 4.5 and recalculating results in a MDF equal to 3.18%.  This compares reasonably well to the corresponding value for Loma Prieta of 2.17%, considering the many uncertainties involved (eg. data collection, soils, building construction, interpolation of ground motion IM).    With regards to the MMI VI data, one might speculate that the Whittier buildings performed markedly better at MMI 6.0-6.5 and markedly worse at 6.5-7.0.  However, if we again combine these two bins the resulting MDF would be 3.0%, which is much closer to the Loma Prieta value of 1.82%.  This suggests that perhaps Wiggins (1994) binned his data too finely – especially considering the following:  MMI ranges are much finer at the lower end of the scale (eg. VI≡PGA .09-.18g, IX≡PGA .65-1.24g)  Wiggins’ intensity measure was based on calculated rather than measured values  Soil conditions were assigned based only on regional mapping  4.7.2.4 Damageability Effects of Specific Building Characteristics One of the key limitations of a fragility-based approach such as ATC-13 is that all structures in the sample are assumed to have the same characteristics.  In an effort to investigate the effects of various building characteristics/irregularities, Lizundia et al. (1993) conducted statistical testing (t-testing of the mean damage factors) for various subsets of the sample to determine whether or not certain characteristics had a significant effect on the performance of the buildings.  This portion of the study considered only San Francisco buildings that had not received parapet strengthening.  The following was investigated:  Site soils  Storey height  Occupancy  Building configuration (eg. square, rectangular, irregular)  Diaphragm ratio (the plan aspect ratio of length to depth)  The presence of a “soft storey”  Year of construction  Number of storeys  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 100 Tables 4.9 to 4.11 show the resulting MDF for the various subsets and the number of buildings in the sample.  Soil type was by far the most significant attribute; in fact, its effect was so pronounced that it was necessary to sort the buildings by soils, before analyzing other attributes (Lizundia, et al. 1991).  Storey height was also significant.  This seems reasonable, since both aspects are key issues in retrofit design.  Occupancy was found to be a significant attribute; this is presumably due to the different structural forms that are associated with these occupancies.  This again appears consistent with engineering judgement, as Rutherford & Chekene (1990) separated buildings into occupancy-related prototypes for loss estimates.  Lastly, the number of stories was a significant attribute: one to three storey buildings had a lower MDF than four storey buildings.  However, the trend was inconsistent in that 5, 6, and 7 storey buildings appeared to suffer less damage (see Table 4.12).   While statistically significant differences were found for some of the remaining attributes, the damage trends were not consistent enough to warrant further investigation as part of this study and, as such, they are not shown.  Table 4.9 – Soil Type vs. MDF (From Lizundia, 1993) Soil Type MDF #Bldgs 1 (Rock/Stiff Soil <200 ft) 0.49% 45 2 (Stiff Soil >200 ft) 4.78% 660 3,4 (medium/soft clay) 6.68% 171  Table 4.10 – Wall Height vs. MDF (From Lizundia, 1993) Wall Height MDF #Bldgs < 16 feet 3.95% 637 > 16 feet 7.90% 239  Table 4.11 – Occupancy vs. MDF (From Lizundia, 1993) Occupancy MDF #Bldgs Residential, Assembly 4.42% 301 Commercial, Office 2.69% 271 Industrial 6.63% 304    Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 101 Table 4.12 – Storeys vs. MDF (From Lizundia, 1993) Number of Storeys MDF #Bldgs 1, 2, 3 4.60% 695 4 7.86% 95 5, 6, 7 2.04% 83 8 11.2% 3  While these statistics have provided some insight into the attributes that may affect the vulnerability of a given building, one key shortcoming is that no results are provided for buildings with more than one of the aforementioned attributes – what about a four-storey building of commercial occupancy, with storey heights less than 16 feet versus a 2-storey building of industrial occupancy, with storey heights greater than 16 feet?  It is unlikely that the answer is a direct superposition of the results, and the attributes themselves may be correlated.    4.7.3 Northridge 1994 Earthquake The Northridge earthquake occurred on January 17, 1994.  It was a MW=6.7 event and the epicenter was about 32km northwest of Los Angeles.  Figure 4.13 shows a map of instrumental intensity, produced by the United States Geological Survey.                   Figure 4.13 – Instrumental Instensity Map – Northridge Earthquake Source: http://earthquake.usgs.gov/earthquakes/shakemap/sc/shake/Northridge/ Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 102 4.7.3.1 Summary of Damage Survey Rutherford and Chekene (1997) compiled damage statistics for Los Angeles, which were collected by the Los Angeles Building and Safety Department.  The data was based on post-earthquake safety evaluations similar to ATC-20, including ATC-13 damage states and various damage descriptions.  Unlike the Whittier earthquake, the majority of the URM buildings in Los Angeles had been seismically upgraded by this time, in accordance with the City’s mandatory retrofit ordinance, Division 88.  Unfortunately, only a modest sample of the total population was assessed and the sample was biased in that it focused on the most heavily damaged areas of the city (Rutherford & Chekene 1997).  Data was collected as noted below:  751 of 5682 retrofitted buildings  93 of 703 unretrofitted buildings  8 of 61 partially retrofitted (tension tie only) buildings  Similar to many of the other damage surveys, the level of ground shaking experienced by the buildings was predominantly at the MMI VII level – Figure 4.14 shows the number of retrofitted buildings in each MMI category.               As aforementioned, only a portion of these buildings were actually assessed.  To address this issue, Rutherford and Chekene produced damage statistics for three scenarios:  1) For the inspected buildings only 2) For the entire population, assuming those not inspected had no damage 3) For the entire population, assuming some level of damage (dependent upon the ground motion intensity) 141 2705092176 30100020003000400050006000V VI VII VIII IX# of BuildingsMMI CategoryFigure 4.14 – MMI Breakdown for Northridge Earthquake (Reproduced from Rutherford & Chekene, 1997) Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 103 The third case fell closer to the ‘no damage’ case and Rutherford and Chekene (1997) noted that this was their best estimate of the damage.  4.7.3.2 ATC-13 Damage Statistics ATC-13 damage statistics were generated, similar to those from the Loma Prieta and Whittier earthquakes. Table 4.13 provides the DPM for the third case (i.e. some damage assumed).  Note that the geometric mean damage factor was again substituted for the central damage factor, as used in the original study.  Also, the results are not plotted at MMI IX, since only three buildings were in the sample.  Table 4.13 – Northridge DPM - Strengthened  (Modified From Rutherford & Chekene, 1997) Damage State GMDF Number of Buildings in MMI V VI VII VIII 1 None (0%) 0% 138 131 2339 58 2 Slight (0-1%) 0.5% 2 133 1503 58 3 Light (1-10%) 3.2% 1 5 1144 44 4 Moderate (10-30%) 17.3% 0 1 70 12 5 Heavy (30-60%) 42.4% 0 0 25 3 6 Major (60-100%) 77.5% 0 0 8 1 7 Destroyed (100%) 100% 0 0 3 0 MDF= 0.04% 0.37% 1.49% 3.30%  The original study also provided DPMs for unretrofitted and tension tie-only buildings. However, since the number of buildings actually inspected was so small, a ‘best estimate’ level of damage was not provided; only the case assuming no damage to inspected buildings was calculated.  Because data from several other, more reliable, sources was reviewed, there was little motivation to make use of the Northridge data for unretrofitted or tension tie-only buildings and no further discussion in presented.     4.7.3.3 Damageability Effects of Specific Building Characteristics Similar to the work of Lizundia (1993), the effects of specific building characteristics were investigated by analyzing the appropriate subsets of the sample.  The following characteristics were investigated: 1) Number of storeys 2) Presence of a basement 3) Horizontal Aspect Ratio (ratio of plan dimensions of the building footprint) 4) Vertical Aspect Ratio (ratio of height to least dimension in elevation)  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 104 Unfortunately information on soils was not available.  However, the subject buildings are concentrated in an area of relatively uniform soil conditions.  Kalkan et al. (2010) notes that extensive areas of the Los Angeles basin are underlain by Pleistocene alluvium (mean VS30=377m/s), while the center is underlain by younger alluvium (mean VS30=287).  This indicates that most of the buildings were likely founded on Site Class D, or perhaps Site Class C, soils.  In addition, the Los Angeles area is well-instrumented (ATC 2001), and so the effects of soils were deemed to be suitably well-captured for characterizing average damage.  Tables 4.14 to 4.17 below show the mean damage factors for the various subsets. Unlike the Loma Prieta set, a breakdown by ground motion intensity was included.  Note that for items 1) and 4), uninspected buildings were assumed to have no damage. For items 2) and 3), only inspected buildings were included.  The horizontal and vertical aspect ratios were analyzed in terms of both Sa(0.3s) and Sa(1.0s); the results for Sa(1.0s) are presented here, because this is consistent with our selected ground motion intensity measure.  Table 4.14 – Number of Storeys vs. MDF (Reproduced From Rutherford & Chekene, 1997) Number of Storeys Sa(0.3s) [g] Total 0.35-0.50 0.50-0.65 0.65-1.225 MDF #Bldgs MDF #Bldgs MDF #Bldgs MDF #Bldgs One to Three 0.77% 3968 0.98% 1055 2.34% 201 0.86% 5224 Four to Six 2.78% 342 1.01% 112 -- 0 2.35% 454  Table 4.15 – Presence of Basement vs. MDF (From Rutherford & Chekene, 1997) Presence of Basement Sa(0.3s) [g] Total 0.35-0.50 0.50-0.65 0.65-1.225 MDF #Bldgs MDF #Bldgs MDF #Bldgs MDF #Bldgs With  6.68% 191 2.39% 60 12.2% 5 5.78% 256 Without 8.98% 170 9.64% 74 7.71% 34 9.00% 278  Table 4.16 – Horizontal Aspect Ratio vs. MDF (From Rutherford & Chekene, 1997) Horizontal Aspect Ratio Sa(1.0s) [g] Total 0.075-0.20 0.20-0.35 0.35-0.75 MDF #Bldgs MDF #Bldgs MDF #Bldgs MDF #Bldgs < 2.0 5.72% 165 8.99% 134 8.57% 22 7.28% 321 > 2.0 5.82% 186 8.44% 185 12.7% 34 7.51% 405   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 105 Table 4.17 – Vertical Aspect Ratio vs. MDF (From Rutherford & Chekene, 1997) Vertical Aspect Ratio Sa(1.0s) [g] Total 0.075-0.20 0.20-0.35 0.35-0.75 MDF #Bldgs MDF #Bldgs MDF #Bldgs MDF #Bldgs < 0.5 0.45% 3036 1.27% 1330 3.25% 168 0.80% 4534 > 0.5 0.94% 732 3.13% 395 4.06% 17 1.75% 1144  Of the four items, only two appear particularly conclusive: 2) the presence of a basement and 4) the vertical aspect ratio.  The former of these two is interesting, but not particularly useful. It is postulated that this result is due to some correlation of other characteristics with the presence of a basement, such as quality of construction.    With regards to the number of storeys, the MDF for the two subsets is markedly different at just slightly different ground motion intensities, which seems unlikely.  While it seems intuitive that number of storeys would be a significant characteristic, one must acknowledge that two buildings with the same number of storeys could actually be quite different in form – one building could be just 20-30 feet wide, while another could run an entire city block.  By contrast, the vertical aspect ratio appears to be a much better indicator of vulnerability.  With regards to the horizontal aspect ratio, one may have expected to see a significant impact here as well.  However, there is virtually no difference except at the highest level of shaking, which is comprised of a very small sample.  This could easily have been skewed by one or two buildings with high ATC-13 damage state classifications.  Based on the results, it appears that horizontal aspect ratio is not a significant predictor of vulnerability.  One may rationalize this by noting that current retrofit provisions do attempt to address excessive diaphragm deflections.  4.7.3.4 Correlation between Ground Motion and Specific Damage Types One final item investigated by Rutherford & Chekene (1997) was concerned with determining the ground motion intensities at which specific types of damage occur.  To this end, Rutherford & Chekene provide the PGA, Sa(0.3s), and Sa(1.0s) at which one percent of the total inventory is affected and at which a sharp jump in the frequency of damage occurs.  Items investigated included:  Building/Storey Leaning  Foundation Damaged  Roof/Floors Damaged Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 106  Columns/Pilasters/Corbels Damaged  Diaphragms/Horizontal Bracing Damaged  Walls/Vertical Bracing Damaged  Cladding/Glazing Damaged  Shear Cracks  Corner Damage  Little difference was observed in the points at which the various types of damage occurred.  In general, 1% of the inventory was damaged at a PGA=.15-.20g/Sa(1.0)=.15-.20g and a “sharp jump” occurred at a PGA=.35-.40g/Sa(1.0)=.40-.45g.  Rutherford and Chekene (1997) noted that the jump occurred at a lower ground motion intensity for wall cracking, but that this may have been due to the fact that wall cracking is so easily observed.  4.7.4 Canterbury 2010/2011 Earthquake Swarm The Canterbury earthquake swarm consisted of several earthquakes, over a period of more than a year.  41 earthquakes of magnitude 5 or greater were detected from September 2010 through January 2011 (Nicholls 2012).  The four largest events were as follows (New Zealand local date):  Mw7.1 on September 4, 2010, 40km west of Christchurch  Mw6.3 on February 22, 2011, 10km southeast of Christchurch  Mw6.4 on June 13, 2011, 10km southeast of Christchurch  Mw5.8 on December 23, 2011, 26km east of Christchurch  The February earthquake was by far the most damaging, although significant damage was also sustained by URM buildings in the September and June events (Moon, et al. 2014).  Because damage was so heavy in the February event, only the September and February events are of particular interest to this study.  Figures 4.15 and 4.16  show USGS-generated maps of instrumental intensity for the September event (also known as the Darfield earthquake) and the February event (also known as the Christchurch earthquake), respectively.      Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 107                                      Christchurch Figure 4.15 – Instrumental Instensity Map – September (Darfield) Earthquake Modified From: http://earthquake.usgs.gov/earthquakes/shakemap/global/shake/2010atbj/ Figure 4.16 – Instrumental Instensity Map – February (Christchurch) Earthquake Modified From: http://earthquake.usgs.gov/earthquakes/shakemap/global/shake/b0001igm/ Christchurch Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 108 In Christchurch, where the majority of the buildings were located, shaking was predominantly at the MMI VII level for the September earthquake. For the February earthquake, shaking was predominantly at the MMI VIII/IX level.  As will be seen, the fact that many of the buildings were subjected to both earthquakes presents unique opportunities and challenges in compiling damage statistics.  Another important point is that this was the only damage survey to contain a significant number of buildings in the MMI IX range.    4.7.4.1 Summary of Damage Survey A damage survey of URM buildings was completed by researchers at the Universities of Auckland, in New Zealand, and Adelaide, in Australia (Ingham and Griffith 2011a, 2011b).  The first report characterizes and generally assesses the observed performance of URM buildings, particularly in the September earthquake.  The second report focuses solely on the performance of buildings in the February earthquake, with particular attention to the effects of strengthening.  Most of the useful damage statistics are provided in the second report.  Note that the information is not necessarily summarized here in terms of the separate reports.  Rather, the body of work is reviewed as a whole and is presented in a way that lends most usefully to the study at hand.  The database consists of 626 URM buildings. The majority were clay brick buildings, with flexible timber diaphragms, but some stone masonry buildings were also present.  According to Ingham and Griffith (2011a), there are an estimated 852 URM buildings in the Canterbury province (which more than encompasses the area affected by the earthquake), so the 626+ buildings is certainly a reasonable sample of the population.  In terms of the Christchurch central business district, Ingham and Griffith (2011b) note that 370 out of 380 total URM buildings were reviewed following the February earthquake.  For the February event, the survey was quite detailed and involved descriptions of any retrofitting work, component specific damage descriptions, building usability placards based on NZSEE (2009) (similar to ATC-20), ATC-13 damage state classifications, and Wailes and Horner damage classifications.  Data was acquired by (primarily) exterior reviews, observing buildings during demolition, aerial photograph, Google street view, and review of Christchurch City Council property records.  For the September event only about 50% of the buildings were assessed for ATC-13 damage states, although placard data was available for about 80% of the buildings.  Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 109 A variety of strengthening statuses were present, ranging from completely unstrengthened to strengthening equal or greater than current code for New Zealand.  Ingham & Griffith (2011b) classified strengthening into three categories: 1) Parapet strengthening: including the following a. the addition of a concrete ring beam, or b. the addition of structural steel bracing back to the roof structure 2) Type A Strengthening: including the following a. Installing connections between walls and the roof and floor systems so that walls no longer respond as vertical cantilevers b. Stiffening of the roof and/or floor diaphragms 3) Type B Strengthening: including the following a. Strongbacks to increase the out-of-plane resistance of the walls b. In-plane wall strengthening, such as concrete/steel frames or shotcrete overlays  Note that in-practice Type B strengthening is never performed without Type A, so the levels of strengthening would actually be considered 1) Parapet Bracing, 2) Type A, 3) Type A+B.   Ingham & Griffith (2011b) provides a breakdown of the buildings by retrofit status for the 370 buildings in the CBD as follows:  139 of 370 (38%) CBD buildings have no confirmed strengthening  149 of 370 (40%) CBD buildings have Type A strengthening  82 of 370 (22%) CBD buildings have Type A+B strengthening  Statistics on parapets were provided for each parapet, rather than for each building.  They were as follows:  149 of 435 (34%) parapets were braced  89 of 435 (21%) parapets were confirmed as not being braced  197 of 435 (45%) parapets had unknown bracing statuses  Neither of the Ingham and Griffith reports (2011a, 2011b) attempted to correlate the damage with ground motion intensity.  Instead, overall statistics for the usabiliy placards, ATC-13 damage states, and component-specific damage assessments are provided.  The following sections summarize the data.  Dr. Jason Ingham and his fellow researchers graciously provided the author with the database collected.  Further analysis was undertaken by the author and is presented in Section 4.8.4. Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 110 4.7.4.2 Placard Data Placard statistics for the two earthquakes were markedly different.  For the September earthquake, a sample of 596 buildings was analyzed, showing that 47% were tagged green, 32% were tagged yellow, and 21% were tagged red (Ingham and Griffith 2011a).  For the February earthquake, the sample of 370 CBD buildings contained 1% green, 17% yellow, and 82% red tags.  Figure 4.17 illustrates the tagging data.  No breakdown by retrofitting was provided.              4.7.4.3 ATC-13 Damage Statistics As aforementioned, Ingham & Griffith (2011a, 2011b) provided overall damage statistics for each of the two earthquakes.  The results for September and February were as shown in Figure 4.18.  The February earthquake clearly resulted in much greater damage.                 Figure 4.17 – Placard Statistics For September (Left) and February (Right) From: Ingham & Griffith 2011a, 2011b Figure 4.18 – ATC-13 Damage Statistics For September and February From: Ingham & Griffith 2011a, 2011b Green47%Yellow32%Red21%Green (1%)Yellow17%Red(82%)26%17%28%19%6%3%13%27%29%23%8%0%5%10%15%20%25%30%35%None(0-1%)Insignificant(1-10%)Moderate(10-30%)Heavy(30-60%)Major(60-100%)Destroyed(100%)% of BuildingsMMI CategorySeptember FebruaryChapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 111 Note that the ATC-13 damage classification was slightly modified: there are only six damage states, instead of seven.  In the original document, the 'Slight (1%)' damage state is effectively omitted.  However, in the figure above and in the author’s subsequent workings with the database, it was assumed that the 'None' and 'Slight' damage states were combined to produce a damage range of 0-1%.  Ultimately, this has little effect on the damage statistics.  Although not computed in the original documents, the mean damage factors (computed using the geometric damage factors) associated with the September and February earthquakes would be 22% and 43%, respectively.  At first inspection this seemed rather high compared to the California data.  However, a review of the recorded PGA's (0.3-0.4g for September and 0.7-1.2g for February) indicates that shaking could perhaps have been closer to MMI VIII, and IX for September and February, respectively (note that this issue was avoided in the author’s subsequent work, as MMI was not used).  A comparison between Loma Prieta and the September events on this basis is at least reasonable, with an MDF of about 11% for the Loma Prieta buildings (see Table 4.8) versus the value of 23% for the September (Darfield) event.  This remaining discrepancy is taken primarily as an example of the relative vulnerabilities of the building stock.  The fact that the Canterbury buildings appear more vulnerable is expected given the lack of tension ties in original construction (compared to San Francisco), and the prevalence of cavity walls, two-wythe walls, and gables (Ingham and Griffith 2011a).  4.7.4.4 Damageability Effects of Specific Building Characteristics The original report (Ingham and Griffith 2011b) also investigates the effect on damageability of various building characteristics.  The overall ATC-13 statistics from the 370 Christchurch CBD buildings in February earthquake were the basis for the investigation, including characteristics as noted below.  Note that the plots provided were reproduced directly from the original report.   1) Number of storeys 2) Row vs. Isolated buildings 3) Mid-row vs. End buildings 4) Building Typology (as defined in Ingham and Griffith 2011a)  The resulting damage appeared to be independent of the number of storeys, as shown by Figure 4.19.  Although not calculated in the original document, the resulting MDFs are shown on the figure.  Recall that previous studies from the Loma Prieta and Northridge earthquakes showed less than conclusive relationships between the number of storeys and damage, although here there appears to be no trend whatsoever.   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 112 A more promising result was obtained with regards to row versus isolated buildings, as shown by Figure 4.20, with isolated buildings suffering more damage.  Similarly, buildings at the ends of rows suffered more damage than buildings in the middle.  Lizundia et al. (1993) recorded information on adjacencies, although no work was performed as to the effects on damageability.    Finally, a reasonable trend was found with regards to the typology, as shown in Figure 4.21.  The typologies are summarized in Table 4.18.  Ingham & Griffith state that Type B and D buildings – one and two-storey row buildings, respectively – were shown to suffer less damage.  However, calculation of the MDFs reveals that Type A and B buildings were the least damaged.   Table 4.18 – New Zealand URM Typologies (From: Ingham & Griffith, 2011a) Type Description Details A 1 Storey, Isolated One storey URM buildings. Examples include convenience stores in suburban areas, and small offices in a rural town. B 1 Storey, Row One storey URM buildings with multiple occupancies, joined with common walls in a row.  Typical in main commercial districts. C 2 Storey, Isolated Two storey URM buildings, often with an open front.  Examples include small cinemas, a professional office in a rural town and post offices. D 2 Storey, Row Two storey URM buildings with multiple occupancies, joined with common walls in a row.  Typical in commercial districts. E 3+ Storey, Isolated Three+ storey URM buildings, for example office buildings in older parts of Auckland and Wellington. F 3+ Storey, Row Three+ storey URM buildings with multiple occupancies, joined with common walls in a row.  Typical in industrial districts, especially close to a port (or historic port). G Institutional, Religious, Industrial Churches (with steeples, bell towers, etc), water towers, chimneys, warehouses.  Prevalent throughout New Zealand.      Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 113                                      46% 43% 39% 39% 71%0%10%20%30%40%50%60%70%80%90%100%1 2 3 4 5% of BuildingsNumber of StoreysDestroyed(100%)Major(60-100%)Heavy(30-60%)Moderate(10-30%)Insignificant(1-10%)None(0-1%)XX% (MDF)Figure 4.19 – Damage vs. Number of Storeys From: Ingham & Griffith 2011b 50% 40%0%10%20%30%40%50%60%70%80%90%100%Isolated Row% of BuildingsRow vs. Isolated BuildingsFigure 4.20 – Damage for Row vs. Isolated (Left) and Middle vs. End (Right) From: Ingham & Griffith 2011b 34% 45%Middle EndMiddle vs. End BuildingsDestroyed(100%)Major(60-100%)Heavy(30-60%)Moderate(10-30%)Insignificant(1-10%)None(0-1%)XX% (MDF)Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 114                Note that ‘Other’ is assumed to essentially represent the Type G (institutional, religious, industrial) buildings.  It is interesting that the statistics indicate that 1-storey buildings as a whole suffer no less damage that others, yet the typology statistics indicate that the typical commercial 1-storey buildings (row or isolated) suffer less damage.  A brief review of the database indicated that a number of Type G buildings were also one-storey buildings, which is thought to be the explanation for this observation. The review also indicated that the sample of Type A buildings was quite small (just 9 buildings), further supporting this theory.  In comparing  this result to the Loma Prieta study by Lizundia et al. (1993), it should be noted that 1-storey buildings had a lower MDF than 2- and 3- storey buildings here as well (3.77%, as compared to 4.85% and 4.68%).  For a better comparison, a breakdown of the Loma Prieta buildings by the New Zealand typology was undertaken by the author and is presented in Section 4.8.2.3.  One further observation based on the MDFs is that Type C and E buildings (two and three storey isolated buildings) appear to suffer similar damage.  The same can be said of Type D and F buildings (two and three storey row buildings).     36% 33% 53% 41% 50% 39% 46%0%10%20%30%40%50%60%70%80%90%100%A B C D E F Other% of BuildingsTypologyDestroyed(100%)Major(60-100%)Heavy(30-60%)Moderate(10-30%)Insignificant(1-10%)None(0-1%)XX% (MDF)Figure 4.21 – Damage by Building Typology From: Ingham & Griffith 2011b Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 115 4.8 New Damage Statistics Results As previously noted, the fact that there is no data on the observed seismic performance of western Canadian URM buildings provided motivation to review a number of other building populations, so as to obtain a reasonable average.  This led to the extensive literature review, as summarized in Section 4.7.  However, there were limitations in the results as presented by others (eg. the use of MMI) as well as opportunies for further investigation (eg. the limited statistical analyses of the Canterbury data).  In this section, the following further studies completed by the author are presented:  Converted damage statistics to a common format.  Results are presented in terms of DPMs and MDF plots with Sa(1) as the ground motion intensity measure  Analyzed the Loma Prieta database to investigate the performance of buildings with braced parapets  Analyzed the Loma Prieta database to further investigate the damageability effects of various building characteristics as defined for the Canterbury buildings  Analyzed the Christchurch database to develop DPMs and MDF plots consistent with the other three databases  Having been converted to a consistent format, the results of the databases are compared to each other and to published sources for various levels of strengthening  Developed HAZUS-compatible damage state structural fragilities for URM buildings for each strengthening level based on the observed MDF vs. Sa(1) relationships  4.8.1 Whittier 1987 Earthquake The Whittier earthquake was the only earthquake of the four for which a database was not available.  As such, all that could be done was to convert the ground motion intensity values from MMI to the chosen ground motion parameter, Sa(1).  This was performed in a two-step process.  4.8.1.1 Converting MMI to Sa(1) First, the MMI measurements were converted to Peak Ground Velocity values, based on the relationship established by Wald et al. (1999).  Note that the relationship was established by regression analysis for eight California earthquakes, including Whittier, Loma Prieta, and Northridge.  The corresponding values for MMI (or Instrumental Intensity as it is referred to for these purposes) and PGV are shown in the USGS-generated shakemaps, such as Figure 4.16.  It was assumed that each bin could be represented by its midpoint (eg. for MMI VII, the midpoint PGV of 23.4cm/s was used). Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 116 In the second step, the resulting PGV is converted to Sa(1).  It was possible to achieve this with a simple linear regression analysis because PGV and Sa(1) are generally correlated quite well.  The regression analysis was performed for each individual earthquake (i.e. Whittier in this case) using the seismic station values as listed in the United States Geological Survey shakemap archives; see the captions of the previously shown shakemaps for web links.  In this case, the correlation coefficient between PGV and Sa(1) was 0.926, with an R2 of 0.86.  Figure 4.22 provides a plot of the data as well as maps showing the spatial similarities.  A more sophisticated model could perhaps have accounted for more of the scatter in the data, but this simple model was deemed accurate enough for the task at hand, which was simply to estimate the expected Sa(1) for a given PGV, which represents a reasonably large range of values and will ultimately be binned quite coarsely.  The same process was also completed for other databases discussed shortly, with similar results.                            R2 = 0.8573010203040500 10 20 30 40Sa(1) [%g]PGV [cm/s]Figure 4.22 – Whittier PGV vs. Sa(1) Regression (left) and USGS Shakemaps (right) PGV Sa(1) Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 117 4.8.1.2 Resulting DPMs and MDF Plots With the MMI values converted to Sa(1), the DPMs can be expressed in terms of Sa(1).  This represents just a small change to Tables 4.5to 4.7 Appendix B contains the DPM.   Since we are ultimately interested in the damage at any point, rather than just the measured intensities, the resulting MDFs were plotted as a function of Sa(1) and a probability distribution was fit to the data.  The beta distribution was fitted to the data using nonlinear regression with a weighted least squares criterion, as discussed in Section 4.6.  This exercise was performed for each of the three subsets of buildings: unretrofitted, partially retrofitted, and fully retrofitted.  Figure 4.23 provides a plot of the results.                   The number next to each data point in the graph above represents the number of buildings associated with that point.  Note that the red markers represent data points that were excluded or had their weights manually reduced (beyond the typical weighting based on their standard error) because they appeared to be outliers.  The cause was usually a bin with just a few buildings, all at the same (or nearly the same) damage state, which would give an erroneous MDF value, but a standard error near zero.  In other cases, it may be due to erroneous estimates of ground motion at the site; recall that in this particular damage survey, soil effects were only accounted for on a regional basis.   10 332334312503 26592412 6 10 1173400%5%10%15%20%25%30%0.00 0.10 0.20 0.30 0.40 0.50 0.60Mean Damage Factor [%]Sa(1) [g]UNRETROFITTED UNRETROFITTED (data)PARTIAL RETROFIT PARTIAL RETROFIT (data)FULL RETROFIT FULL RETROFIT (data)Figure 4.23 – MDF vs. Sa(1) For Whittier Earthquake Data points shown in red have been excluded from the fitting process or had weights manually adjusted Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 118 The results are somewhat unusual in that they show little difference between retrofitted, partially retrofitted, and unretrofitted buildings.  Overall, the results from the Whittier database are thought to be the least reliable or useful.  This is largely due to the lack of variety in the ground motion intensity and the aforementioned issues with soils.    In the Sa(1)=0.25-0.35g range the results are consistent with expectations, showing some marginal improvements for partial retrofitting (essentially just tension ties) and more significant improvements for full retrofitting.  One may have expected a larger difference between the ‘unretrofitted’ and ‘partially retrofitted’ buildings; however, the City of Los Angeles implemented a parapet bracing program in 1949 (Rutherford & Chekene 1997), which means that many of the buildings are likely not truly ‘unretrofitted’.  Unfortunately, no discussion of this is provided in the original report (Wiggins, Breall and Reitherman 1994).  Nonetheless, the results are valuable because it contains the greatest variety of strengthening among the California databases.  4.8.2 Loma Prieta 1989 Earthquake The database for the Loma Prieta earthquake was graciously provided to the author by its original owner, Mr. Bret Lizundia of Rutherford + Chekene engineers.  Therefore, not only was the original DPM for unstrengthened buildings (see Table 4.8) converted to Sa(1), but the database was also further analyzed to determine the damage to buildings with braced parapets.  4.8.2.1 Converting MMI to Sa(1) For the buildings with unbraced parapets, a DPM using MMI was available in the original document (see Table 4.8) and was thus converted.  This task was completed in a manner identical to that described in Section 4.8.1.1.  The resulting relationship between MMI and Sa(1) was very similar.    4.8.2.2 Resulting DPMs and MDF Plots The DPM for buildings with unbraced parapets was readily constructed after converting MMI to Sa(1).  See Appendix B for the converted DPM.  The original report by Lizundia (1993) excluded buildings with braced parapets from its analysis, because the scope of the report was essentially unstrengthened buildings.  For the purposes of this study, however, the performance of buildings with braced parapets was of great interest and so the database was analyzed and DPMs were constructed.  Note that this represents only about 1000 buildings in San Francisco, not the entire Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 119 592 1533179 2751218163 7760%5%10%15%20%25%30%0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80Mean Damage Factor [%]Sa(1) [g]UNRETROFITTED UNRETROFITTED (data)PARAPETS BRACED BRACED (data)database, which was the basis of Lizundia’s work.  Since spectral seismic demands were not included in the database, it was first necessary to determine demands at the site.  This was achieved in a manner similar to that outlined by Lizundia (1993), and as described below: 1) The four closest stations having the same soil category as the subject building were identified 2) The values at the subject building were interpolated from each site using the attenuation equation from Boore, Joyner, and Fumal (1997).  Note that Lizundia used Boore and Joyner (1988).  In any case, the difference has virtually no effect, since only the relative attenuation is used 3) The values inferred from the four stations were weighted by distance.  The square of the distance was used, as described in Lizundia (1993).  While more scientifically justified methods exist, the differences again tend to be masked by the binning.  See Section 4.5.2.4 for a discussion on interpolating demands from stations versus other methods 4) The data was binned by Sa(1) in increments of 0.05g, ranging from 0.10g to 0.35g  The DPM was then constructed in the typical manner and the MDF was calculated for each bin.  As before, the beta distribution was fitted to the data.  Figure 4.24 provides the resulting curves of MDF versus Sa(1) for unretrofitted and braced parapet buildings.                 Figure 4.24 – MDF vs. Sa(1) For Loma Prieta Earthquake Data points shown in red have been excluded from the fitting process or had weights manually adjusted Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Again, the red marker represents an outlier that was excludedoriginally, plotted at 0.25g.  MDF value was so much greater the demands on these buildings.  San Francisco, the buildings were mapped along with soils.  The map in shows that these buildings (marked in blue)while most areas in San Francisco experienced MMI VI or VII level shaking, USGS maps showed that this area experienced up to MMI IX buildings were plotted at MMI 8.5, which translated into S                            Figure 4.25 – Selected San Francisco Buildings with Soils ConditionsModified From: http://nsmp.wr.usgs.gov/Presentations/EGSitSFBR/640/Pres3_4.htmlFigure 4.26 – MMI Map of San Francisco (From USGS, 1989)Statistics.  The blue marker was As there was a reasonable number of buildings and thethan expected at 0.25g, it was decided to investigate Because soils were largely responsible for variation in  were located on the poorest soil in the city; (see Figure 4.26).  As such, these a(1)=0.52g.  120  Figure 4.25    Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 121 4.8.2.3 Damageability Effects of Specific Building Characteristics The original study by Lizundia (1993) looked at the damageability effects of various characteristics.  However, as discussed in Section 4.7.2.4 many of these results were not entirely convincing.  A few items investigated by Ingham and Griffith (2011b) appeared promising and, as such, those same characteristics were investigated for the San Francisco buildings.  The items investigated were as follows: 1) Typology 2) Middle versus end buildings  4.8.2.3.1 Typology The first step was to assign the typologies.  This was accomplished by making use of several fields in the databases as noted below:  Prototype: 15 different prototypes, “A” through “O” as specified in Lizundia 1993 (eg. prototype K is two and three storey, small area, residential)  Building use: Residential, Office, Commercial, Industrial, Assembly  Number of Storeys  Adjacencies: provided for all four sides of each building, noting the presence of buildings/alleys/streets  For the purposes of classifying buildings as either row or isolated, only buildings with no buildings on any of the four sides were classified as isolated; all others were classified as row buildings.  The types were then assigned based on the following rules: 1) Buildings of assembly occupancy (Lizundia Prototype ‘O’) were assigned Type G 2) Large area, one-storey buildings (Lizundia Prototype ‘B’) were assigned Type G 3) Isolated buildings of industrial occupancy were assigned Type G 4) 1-storey isolated buildings not classified as G were assigned Type A 5) 1-storey row buildings not classified as G were assigned Type B 6) 2-storey isolated buildings not classified as G were assigned Type C 7) 2-storey row buildings not classified as G were assigned Type D 8) 3-plus storey isolated buildings not classified as G were assigned Type E 9) 3-plus storey row buildings not classified as G were assigned Type F  Note that Lizundia et al. (1993) prototypes E and F were also industrial type buildings, but these were not automatically assigned to NZ Type G.  Based on the storey height as defined by Lizundia et al. (1993), these buildings were found to be more similar to residential or commercial occupancy buildings with a similar number of storeys; this was somewhat intuitive, since many URM buildings with current-day occupancies of Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 122 residential or commercial nature were originally used for industrial purposes.  The form of a building (and thus the damageability) is obviously more closely correlated with its original occupancy than its current one.  Similarly, all Lizundia et al. (1993) prototype B buildings were assigned NZ Type G, since such buildings are typically of industrial origins. Figure 4.27 shows the resulting distribution of the San Francisco buildings.  Statistics from the Christchurch CBD and the Canterbury area (i.e. Ingham & Griffith’s entire database) are also provided6.                   Based on the graph above, it appears there exists a greater proportion of 3+ storey buildings in Francisco than in either of the two New Zealand samples.  This is at least partly due to the fact that even 5+ storey buildings are somewhat common in San Francisco (326 of 2005 buildings). A review of the Christchurch database showed that only 3 of 627 buildings were of 5+ storeys – instead, two-storey buildings are the most common.  Another key difference that is not immediately evident is that isolated buildings are somewhat more common in Canterbury than in San Francisco: Figure 4.28 shows a plot of row versus isolated buildings for the two datasets.  This difference is presumably explained by the relative size and density of the two cities, with San Francisco’s population being well over 300,000 by 1900 compared to about 50,000 in Christchurch in 1900.                                         6 Note that "Christchurch CBD" refers to the buildings in the Christchurch CBD while "Canterbury" refers to the entire sample, including buildings in various outlying areas throughout the Canterbury region. Figure 4.27 – San Francisco vs. Christchurch NZ Typology 0%10%20%30%40%50%60%A B C D E F G% of BuildingsNZ TypologySan Francisco (2005 Bldgs)Christchurch CBD (370 Bldgs)Canterbury (627 Bldgs)Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 123                   To further characterize the San Francisco buildings, the entire database of 2005 buildings was analyzed for specific characteristics, by typology.  Table 4.19 shows the results.  Unfortunately, only a few of these characteristics were recorded in the Christchurch database; however, the available data is shown in Table 4.20; note that this data primarily represents buildings from the CBD.     Table 4.19 – Average Characteristics for San Francisco (2005 buildings) Type Total Area [sq.ft.] Footprint Area [sq.ft.] Diaph. Aspect Ratio Storey Height [ft] # Reentr. Corners % Soft Storey % Parapets Braced A 1636 1636 1.32 17.2 0.30 40% 20% B 3065 2671 2.73 16.5 0.13 46% 41% C 16330 8045 2.19 14.4 1.52 27% 25% D 12097 5896 2.50 13.8 0.29 41% 43% E 31774 8013 2.19 12.7 1.93 21% 52% F 20420 4917 2.44 12.6 0.97 44% 70% G 12486 9653 2.01 18.1 0.74 30% 27% 13%87%39%61%0%10%20%30%40%50%60%70%80%90%100%Isolated Row% of BuildingsIsolated versus Row BuildingsSan Francisco (2005 Bldgs)Canterbury (626 Bldgs)Figure 4.28 – Isolated versus Row Buildings for San Francisco and Canterbury Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 124              With regards to the San Francisco buildings, the following observations can be made:  Type A, B, and G buildings have the greatest average storey heights (the values reported in the original database are the height of the building divided by the number of storeys)  Row buildings tend to have more elongated diaphragms (i.e. greater “horizontal aspect ratios” as discussed in Section 4.7.3.3) than isolated buildings  Isolated buildings tend to have more re-entrant corners (the values shown are the average number per building)  Row buildings have a higher frequency of ‘soft storeys’; this presumably represents open fronts  Row buildings had a higher frequency of parapet bracing; this is probably because they more frequently had parapets that required bracing under the city’s ordinance  In comparing the San Francisco buildings to the Christchurch buildings, the following observations can be made:  The Christchurch buildings do not exhibit the same trend for diaphragm ratios  In terms of footprint area, Type G is again the largest; however, the remaining types are somewhat similar  The trend among parapet bracing is consistent, with row buildings exhibiting the highest frequency of bracing, and Types A and G buildings exhibiting the lowest  The fact that these results are quite intuitive and reasonably consistent reinforces our confidence that buildings have been assigned correctly to the various typologies.   Table 4.20 – Avg. Characteristics for Christchurch Type Footprint Area [sq.ft.]* Diaph. Aspect Ratio** %  Parapets Braced*** A 3010 2.05 10% B 4317 2.15 31% C 3664 2.41 14% D 2669 2.29 33% E 3656 2.20 26% F 4071 2.25 46% G 7291 2.21 9% * Based on partial sample of 354 buildings ** Based on partial sample of 331 buildings *** Based on partial sample of 410 buildings Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 125 With regards to damageability, a somewhat different trend was observed for the San Francisco buildings than that found by Ingham and Griffith for the Christchurch buildings: Figure 4.29 shows the ATC-13 damage distribution to San Francisco buildings.  Note that for each typology, the MDF is shown at the top and the number of buildings in the sample is shown at the bottom.  Only buildings in the Sa(1)=0.35g bin (+0.05g) were used, to provide a fair comparison amongst the buildings.                     The number of storeys appears to be the more significant factor.  For buildings of three or more storeys buildings, row buildings (Type F) suffered less damage than isolated buildings (Type E), similar to the findings of Ingham and Griffith (2011b).  However, the trend does not hold for two-storey buildings, although this is perhaps due to the relatively small sample size for Type C.  Obviously we cannot draw conclusions about Type A buildings, as only 1 such building was available in the bin.  Another observation is that the Type G buildings appear to suffer the most damage.  On a relative scale, these differences are more pronounced than those found for the Christchurch buildings.  One possible reason is the fact that the intensity of shaking was much higher for the Christchurch earthquake.  It is intuitive that the damageability of the buildings would converge at the higher levels of damage (as is often illustrated in fragility curves).    0.10% 0.73% 0.55% 1.59% 2.30% 1.85% 3.32%0%10%20%30%40%50%60%70%80%90%100%A (1)B(92)C(30)D(281)E(110)F(676)G(144)% of BuildingsTypology (# Bldgs)Major(60-100%)Heavy(30-60%)Moderate(10-30%)Insignificant(1-10%)Slight(0-1%)None(0%)XX% (MDF)Figure 4.29 – Damage to San Francisco Buildings by Typology Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 126 4.8.2.3.2 Middle versus End Buildings The adjacency information for the San Francisco buildings also enabled the determination of middle buildings, end buildings, and corner buildings.  Figure 4.30 shows the relative proportions of middle vs. end buildings for San Francisco as well as Canterbury for comparison.  Note that the sum of the middle and end buildings equals 100% since only row buildings are included.  As can be seen, the numbers compare well.  The database also showed that 22% of the San Francisco row buildings were corner buildings – in other words, about 40% of end buildings were also corner buildings.  Unfortunately, this information was not available in the Canterbury database.                   With regards to damageability, a similar trend to that for the Canterbury buildings is observed in that end buildings suffered more damage than middle buildings, as shown in Figure 4.31.  Interestingly, corner buildings (i.e. buildings on two perpendicular sides) appear to suffer less damage than end buildings (i.e. only one adjacent building).  Engineering intuition suggests that corner buildings could be even worse than end buildings due to the eccentricity of the seismic force resisting walls in corner buildings, as the two streetfront walls tend to be highly perforated and therefore much softer and weaker.  One possible explanation is perhaps the fact that these buildings have solid walls in two directions more than makes up for the eccentricity effects in terms of global damageability. 47%53%22%43%57%0%10%20%30%40%50%60%70%Middle End Corner% of BuildingsMiddle vs. End vs. CornerSan Francisco (1735 Row Bldgs)Canterbury (378 Row Bldgs)Figure 4.30 – Middle/End/Corner Buildings for San Francisco and Canterbury Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 127    4.8.3 Northridge 1994 Earthquake This database was again provided to the author by its original owner, Mr. Bret Lizundia of Rutherford + Chekene engineers.  As the database was relatively small compared to the Loma Prieta database, no additional investigations were performed.  4.8.3.1 Converting MMI to Sa(1) As the majority of the data was for unretrofitted buildings, only this DPM was converted and used in this study.  The task was completed in a manner identical to that described in Section 4.8.1.1.  The resulting relationship between MMI and Sa(1) was slightly different, as expected.    4.8.3.2 Resulting DPMs and MDF Plots The DPM for buildings with unbraced parapets was readily constructed after converting MMI to Sa(1).  See Appendix B for the converted DPM.  Again, a beta distribution was fit to the data.  Figure 4.32 provides a plot of the results.   The results appear reasonable, in that damage is lower than both the unretrofitted and braced parapet buildings from the Loma Prieta database.   1.27% 2.26% 1.72%0%10%20%30%40%50%60%70%80%90%100%Middle(543)End(624)Corner(263)% of BuildingsTypology (# Bldgs)Major(60-100%)Heavy(30-60%)Moderate(10-30%)Insignificant(1-10%)Slight(0-1%)None(0%)XX% (MDF)Figure 4.31 – Damage to San Francisco Buildings by End/Middle/Corner Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 128                    4.8.4 Canterbury 2010/2011 Earthquake Swarm The database for Canterbury buildings was provided by Dr. Jason Ingham and his fellow researchers in New Zealand and Australia.  Because Ingham and Griffith’s work (2011a, 2011b) primarily examined raw statistics for the 370 Christchurch buildings, plenty of work was completed by the author, which included estimating IM’s, constructing the DPM’s and fitting distributions to the data.  4.8.4.1 Resulting DPMs and MDF Plots The first step in constructing the DPM was estimating the ground motions at each site.  The ground motion at each site was estimated in the same manner as for the San Francisco buildings (see Section 4.8.2.2).  The DPM was then constructed in the typical manner, the MDF was calculated for each bin, and a beta distribution was fitted to the data.  Figure 4.33 provides the resulting curves of MDF versus Sa(1).     0%5%10%15%20%25%30%35%40%45%50%0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80Mean Damage Factor [%]Sa(1) [g]RETROFFITED RETROFITTED (data)Figure 4.32 – MDF vs. Sa(1) For Northridge Earthquake Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 129                     The results again appear reasonable, with less damage for increased levels of strengthening.  Note that “Braced” refers to braced parapet buildings.  Clearly, one of the downfalls of the Canterbury database is the relatively small number of buildings.  However, it is by far the database with the strongest level of shaking.   4.8.5 Comparison of Results from North America and New Zealand Having generated results for each of the various databases on common terms, the performance of the buildings can be compared.  The following sections compare the results and provide discussion and are organized by strengthening status.    2016485 301245818858025 21 90%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40Mean Damage Factor [%]Sa(1) [g]UNRETROFITTED UNRETROFITTED (data)BRACED BRACED (data)FULL RETROFIT FULL RETROFIT (data)Figure 4.33 – MDF vs. Sa(1) For Canterbury Earthquakes Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 130 4.8.5.1 Unretrofitted Buildings As would be expected, the unretrofitted buildings suffered the most damage in each set.  Figure 4.34 shows a plot comparing the MDF vs. Sa(1) relationship for each data set.                      A wide range of performance was observed, but this is to be expected given the differing materials, construction practices, and ground motion characteristics.  Recall that an unknown, but likely significant, number of the Whittier buildings actually had received some strengthening in the form of parapet bracing.  Based on this, combined with the absence of high level shaking, it is postulated that the Whittier data is not an accurate representation of unstrengthened buildings.  The Loma Prieta and Canterbury data sets both have their merits and flaws, as previously discussed; however, their results are thought to be much more accurate representations than the Whittier results.  Based on a certain amount of engineering intuition, it is postulated that the Canterbury and Loma Prieta results represent reasonable upper and lower bounds, respectively, on the vulnerability of unretrofitted buildings.  The rationale is as follows:  Several accounts of poor mortar were reported by Ingham and Griffith (2011a, 2011b).  Poor mortar quality has also been reported in URM failures in California (Deppe 1988, Lizundia, Dong and Holmes 1993, LATF 1994); however, Lizundia Figure 4.34 – Comparison of Unretroffited Buildings 0%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40Mean Damage Factor [%]Sa(1) [g]UNRET-CHCH UNRET-CHCH (data)UNRET-WHITTIER UNRET-WHITTIER (data)UNRET-LOMA PRIETA UNRET-LOMA PRIETA (data)Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 131 (1993) reports that the mortar quality in California buildings is fair to good  Two wythe walls are reportedly somewhat common in the Canterbury region (Derakhshan 2011).  While there have been reports of two-wythe walls in California, three-wythe walls are reportedly typical in San Francisco (Rutherford & Chekene 1997).  In the author’s experience, two-wythe walls are also uncommon in Victoria, BC, although two-wythe parapets are reasonably common  Veneer/Cavity wall construction was found to be somewhat common in the Canterbury region: Ingham and Griffith (2011b) report that cavity wall construction (i.e. a gap between outer wythe and inner wythes) was encountered in “almost half of the URM buildings surveyed in Christchurch;” conversely, cavity wall construction is reportedly quite rare in California (Lizundia, 1993)  Tension ties as part of original construction were reportedly rare in Canterbury: Ingham and Griffith (2011b) report that only 1.6% (6/370) of Christchurch CBD buildings had ties from original construction.  It should be noted that 43% of buildings were identified as having retrofit ‘through ties.’  Historically, it has not been uncommon to see original ties mistakenly reported as retrofit items.  Nonetheless, original tension ties (known as ‘government anchors’) were common in California, particularly in San Francisco (Rutherford & Chekene 1990)  The cumulative effect of damage to the Canterbury buildings from the Darfield earthquake likely acted to increase damage in the February earthquake  A greater proportion of the Canterbury buildings were shown to be isolated buildings (see Figure 4.28)  4.8.5.2 Buildings with Braced Parapets As expected, buildings with braced parapets appeared to perform better than completely unretrofitted buildings.  Figure 4.35 provides a plot comparing the results from the Loma Prieta and Canterbury buildings.  Data points in red were excluded from the fitting process (as discussed in Section 4.8.1.2).  It can be seen that there is less variation between the two data sets, which is intuitive.  For the same reasons as previously discussed, these damage relationships are thought to be reasonable bounds.  4.8.5.3 Partially Retrofitted Buildings Samples for partially retrofitted buildings were available for the Whittier and Canterbury databases.  Note that partial retrofitting is somewhat loosely defined, but the intent is to include tension ties at all floors as a minimum.  Figure 4.36 shows the results for these two data sets.   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 132                          Figure 4.35 – Comparison of Braced Parapet Buildings 0%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40Mean Damage Factor [%]Sa(1) [g]PARA. BRACED-CHCH BRACED-CHCH (data)PARA. BRACED-LOMA PRIETA BRACED-LOMA PRIETA (data)3 2659 241101445321120%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40Mean Damage Factor [%]Sa(1) [g]WHITTIER WHITTIER (data) ChCh ChCh (data)Figure 4.36 – Comparison of Partially Retrofitted Buildings Data points shown in red have been excluded from the fitting process or had weights manually adjusted Data points shown in red have been excluded from the fitting process or had weights manually adjusted Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 133 The results for the partially retrofitted buildings are somewhat disappointing.  For the Whittier data, we saw that the partially retrofitted category suffered only slightly less damage. For the Canterbury data, the damage actually exceeded that for buildings that had only parapet bracing (see Figure 4.35).  There are many potential explanations for this result: the “partial retrofits” could simply have been poorly conceived. Another issue is the completeness of the scope. For instance, a building could be classified as “partially retrofitted” because one wall had been connected to the floors/roofs, while other walls had not been. If any of the remaining unrestrained walls collapsed, the building performance as a whole would be judged to have been poor.  4.8.5.4 Fully Retrofitted Buildings Samples for fully retrofitted buildings were available for the Whittier, Northridge, and Canterbury databases (see Figure 4.37).  Note that fully retrofitted buildings can be taken to include strengthening measures for out-of-plane (eg. tension/shear ties, strongbacks) and in-plane demands (eg. steel frames, concrete shear walls).                  The results are encouraging in that both damage and variability appear to have been reduced with strengthening.  Note that only the Canterbury results include higher levels of shaking.  In looking at the Northridge curve, the slope appears somewhat flat at the levels of higher shaking; this may represent an underestimate of damage for the Northridge buildings.  Nonetheless, the range of results is relatively small. 0%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40Mean Damage Factor [%]Sa(1) [g]FULL RETROFIT.-CHCH FULL RETROFIT-CHCH (data)FULL RETROFIT-WHITTIER FULL RETROFIT-WHITTIER (data)FULL RETROFIT-NR FULL RETROFIT-NR (data)Figure 4.37 – Comparison of Fully Retrofitted Buildings Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 134 4.8.6 Comparison of Results to Published Sources In an effort to further validate the results, it was decided to make comparisons to published motion-damage relationships.  These include: 1) “ATC-13” (ATC 1985) 2) “Expected Seismic Performance of Buildings” (EERI 1994) 3) “HAZUS-MH 2.1” (FEMA 2012)  ATC-13 contains an appropriate relationship for unretrofitted buildings and EERI (1994) contains a relationship for buildings retrofitted to the Uniform Code for Building Conservation, which is essentially the same as current American and Canadian URM retrofit standards.  HAZUS contains an appropriate relationship for unretrofitted buildings and, although it does not provide one for strengthened URM buildings, it suggests (in passing) the use of its relationship for low code reinforced masonry in lieu.    4.8.6.1 Unretrofitted Buildings ATC-13 and HAZUS were both found to contain appropriate motion-damage relationships for unretrofitted buildings.  Both sources contained multiple URM types; “Facility Class 75” from ATC-13 and “URMLR (pre-code)” were selected for comparison.  Figure 4.38 shows the two relationships along with the relationships from the observed statistics (reproduced from Figure 4.34).                   0%10%20%30%40%50%60%70%80%90%100%0.00 0.50 1.00 1.50 2.00 2.50 3.00Mean Damage Factor [%]Sa(1) [g]UNRET-CHCHUNRET-CHCH (data)UNRET-LOMA PRIETAUNRET-LOMA PRIETA (data)HAZUS-URMLR (w/ NSC's)ATC-13 (FC=75)Figure 4.38 – Published vs. Observed Damage for Unretrofitted Buildings Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 135 The published relationships are not highly consistent with the observed results, or with one another.  At the low-intensity end, ATC-13 is similar to the Canterbury results, but the Loma Prieta results indicate significantly less damage, which was also noted by Lizundia (1993), in his comparison to ATC-13.  The HAZUS curve has a somewhat different shape and indicates much higher damage at the low-intensity end.  Again, the Whittier results have been excluded, since they are thought to be an inaccurate representation.  Also note that the results have been shown to a much higher intensity level than is truly plausible so as to provide a better sense of the relative shapes of the curves.  However, the values above Sa(1)=2g are of virtually no consequence for this study as the probability of exceedance is so low as to not affect the subsequent cost-benefit analyses.  4.8.6.2 Retrofitted Buildings EERI (1994) and HAZUS were both found to contain appropriate motion-damage relationships for retrofitted buildings. As aforementioned, the EERI relationship was specifically developed for rehabilitated URM buildings, while HAZUS simply recommends the relationship for Low Code Reinforced Masonry.  Figure 4.39 shows the two published relationships along with the relationships from observed statistics.                       Figure 4.39 – Published vs. Observed Damage for Fully Retrofitted Buildings 0%10%20%30%40%50%60%70%80%90%100%0.00 0.50 1.00 1.50 2.00 2.50 3.00Mean Damage Factor [%]Sa(1) [g]FULL RETROFIT.-CHCHFULL RETROFIT-CHCH (data)FULL RETROFIT-WHITTIERFULL RETROFIT-WHITTIER (data)FULL RETROFIT-NRFULL RETROFIT-NR (data)HAZUS-RM1L (w/ NSC's)EERI 1994Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 136 For the retrofitted buildings, there is an even greater discrepancy between the published and observed relationships.  Rutherford and Chekene (1997) also noted significant differences in their comparison of the Northridge buildings and the EERI relationship.      4.8.6.3 Structural versus Non-Structural Damage The aforementioned discrepancies between the observed data and published relationships raised questions as to what exactly the published and observed relationships represented.  One of the most important distinctions is which building elements are included in the motion-damage relationships.  Possible elements include: 1) Structural components 2) Non-structural components (NSC’s) 3) Building contents (eg. furniture, computers)  In the case of ATC-13, building elements are simply grouped into two categories: the “facility” (structural and non-structural components) and “equipment” (contents).  The previously shown relationship for ATC-13 Facility class 75 (URM Low Rise) is for the facility and therefore includes the non-structural components.  In the case of HAZUS, separate relationships are provided for all three elements, and NSC’s are further broken down into acceleration-sensitive and drift-sensitive components.  The previously shown relationship includes the structural and all non-structural components and was obtained by combining the three MDF vs. Sa(1) relationships.  To combine the relationships, the relative values of the structural and non-structural components must be established.  HAZUS provides the following values for a retail trade (COM1) occupancy, which is the most common occupancy for URM buildings in Victoria:  Structural Value = 29%  Acceleration Sensitive NSC’s = 43%  Drift Sensitive NSC’s = 28%  These values add to 100%, representing the ‘building value’, which is also specified in the document.  Contents values are specified as equal to 100% of the structural value for this occupancy class and there is also a small value prescribed for business inventory.  For the time being however, only the three above noted percentages are important, as these were used to produce a weighted average of the three separate MDF vs. Sa(1) curves, yielding the results shown in Figures 4.38 and 4.39.  Of course, the relationship for each element is significantly different.  Also note that some work was required to convert the Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 137 HAZUS results to Sa(1), since it was developed in terms of spectral displacement.    The significance of the foregoing discussion is appreciated when considering the meaning of the observed damage statistics.  In Figure 4.40, the structural-only damage relationships from HAZUS are plotted, along with the relationships developed from the Canterbury data.  Both unretrofitted and fully retrofitted curves are presented.                    Here, the HAZUS curves are in better agreement with the observed results, although they still generally exceed even the Canterbury data by some margin in the regions of plausible shaking.  Although the damage surveys were intended to capture non-structural components, it is easy to rationalize the better fit with structural-only relationships, since post-earthquake evaluations are often performed primarily (if not exclusively) from the exterior.  Furthermore, for rapid safety evaluations, the inspectors would instinctively have their attention drawn to structural damage since this is more likely to produce serious life safety threats (eg. due to collapse).  4.9 Motion-Damage Relationships for Victoria Having completed the extensive review of damage to URM buildings in other regions, the focus is now shifted to Victoria and an attempt to define appropriate motion damage 0%10%20%30%40%50%60%70%80%90%100%0.00 0.50 1.00 1.50 2.00 2.50 3.00Mean Damage Factor [%]Sa(1) [g]UNRET-CHCHUNRET-CHCH (data)FULL RETROFIT.-CHCHFULL RETROFIT-CHCH (data)HAZUS-URMLR (Struct. Only)HAZUS-RM1L (Struct. Only)Figure 4.40 – Published vs. Observed Damage for Unretrofitted Buildings Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 138 relationships is made.  First, the level of vulnerability for Victoria relative to the other regions is rationalized on a subjective basis and then the curves for various levels of retrofitting are presented.    Based on the discussion of Section 4.8.6.3, it was concluded that the observed data better represented structural-only damage and so the data is used to define curves for the structural-only damage.  Motion damage relationships for the non-structural components and building contents will be those as defined in HAZUS.  4.9.1 Relative Vulnerability of Victoria URM Buildings For reasons discussed in Section 4.8.5, it is postulated that the vulnerability of Victoria buildings lies somewhere between those of the California and Canterbury buildings.  Beyond this, there is little that can be said with a high degree of confidence about where the vulnerability of Victoria’s URM buildings should fall.  In Table 4.21, we make a subjective comparison of Victoria to the California and Canterbury characteristics as noted in Section 4.8.5. An ‘X’ denotes which of the two regions to which Victoria is more similar.  Note that items #1 through #7 are presented in what the author believes to be the order of importance to seismic performance.  Additionally, a particular emphasis was placed on the Loma Prieta buildings for the California results, as this was the database that was considered more representative of unretrofitted URM buildings.  For retrofitted buildings, the Northridge data is the best available Californian sample.   Table 4.21 – Victoria Buildings vs. Other Regions Characteristics California7 Canterbury 1 Prevalence of original tension ties  X 2 Prevalence of Cavity Construction X  3 Wall Thickness X  4 Mortar Strength X  5 Number of Stories  X 6 Cumulative Earthquake Damage  X 7 Prevalence of Row Buildings  X 8 Retrofit Standards X   Clearly there are mixed results.  See Chapter 7 for a characterization of Victoria buildings.  Note that Victoria was classified as more similar to the Canterbury buildings in terms of cumulative earthquake damage because of the potential for a subduction                                       7 Note that "California" refers to the Loma Prieta results (Section 4.8.2) for unretrofitted buildings and to the Northridge results (Section 4.8.3) for retrofitted buildings. Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 139 earthquake, which comprises a significant portion of Victoria’s seismic hazard.  Such a scenario would be expected to generate a long duration of shaking and many aftershocks.  On a deterministic basis, one could rank this higher in order of importance, but again, this represents only a portion of the seismic hazard.  Of course, the retrofit standards category is an important consideration for retrofitted buildings. As previously noted, Canadian and American URM retrofit standards are quite similar.  Another issue to be considered in deciding where Victoria’s buildings should fall relative to the various results is that the Canterbury data is the only one with extremely high shaking.  Even if intuition suggests that Victoria’s buildings are more similar to the California buildings, it seems imprudent to put more stock in extrapolated results than actual observed damage.   4.9.2 Building Specific Damageability Effects As seen in earlier sections, attempts have been made to link many specific building characteristics to building damageability.  For the most part, the efforts have been either unfruitful or inconsistent among different databases.  Moreover, no effort has been put forth to investigate the effect of multiple simultaneous characteristics.  However, the breakdowns by New Zealand typology for both the Canterbury and Loma Prieta buildings gave results that were reasonably in line with engineering intuition and can, indirectly, account for the presence of multiple characteristics.  Based on those results, the following observations were accounted for:  Isolated buildings suffered more damage than row buildings  Type G buildings (industrial or religious types) suffered more damage than others  Type A and B buildings (1-storey structures) suffered less damage than others  The observations were accounted for by adjusting the base curves shown in Figure 4.41.  In order to define the adjustment factors, it was necessary to first designate one of the typologies as the anchor point: upon calculating the weighted averages of the mean damage factors for both datasets, the resulting average MDF compared most closely with the Type F buildings and so this was selected as the anchor.    Another observation was that the differences in damage were not as pronounced for the Canterbury buildings.  It was postulated that this was primarily due to the increased level of shaking (i.e. at some point the shaking is intense enough that the differences in vulnerability are insignificant and all the structures sustain heavy damage).  As such, it was decided to calculate separate factors at the two different levels of ground motion Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 140 that were investigated (Sa(1)=0.35g and Sa(1)=0.9g, respectively) and to eliminate any difference at a shaking intensity of 50% greater than that which was observed in Christchurch (i.e. Sa(1)=1.35g).  By taking the ratio of the MDF of each typology with Type F for the two databases (shown in Figures 4.21 and 4.29), the following adjustment factors were developed.  Table 4.22 shows the resulting adjustment factors to be applied to the base MDF curve.  It should be noted that the adjustment factors were limited to a +50% adjustment so as to limit the effect on the subsequent cost-benefit analysis.  The adjustment factors were also reduced where the confidence level (from t-testing) was less than 90%.  Table 4.22 – Typology MDF Adjustment Factors Typology Sa(1)=0.30 Sa(1)=0.90g Sa(1)=2.0g A 1 storey, Isolated 0.75 0.90 1.00 B 1 storey, Row 0.75 0.90 1.00 C 2 storey, Isolated 1.25 1.10 1.00 D 2 storey, Row 1.00 1.00 1.00 E 3+ storey, Isolated 1.25 1.10 1.00 F 3+ storey, Row 1.00 1.00 1.00 G Institutional, religious, industrial 1.25 1.10 1.00  It should be noted that the differences in the MDF were indeed significant, particularly for the Loma Prieta data: the MDF’s were typically at least one to two standard errors (i.e. standard deviations of the mean) away from one another.  Although the testing showed less than 90% confidence in differentiating Types C and E from F, the consistent results between the Loma Prieta and Canterbury data suggest there is in fact a significant difference.  These adjustment factors were applied to unretrofitted buildings; reduced factors (eg. 25% increase reduced to 12.5%) were applied to buildings with braced parapets and no factors were applied to further retrofits because presumably the underlying deficiencies would be addressed.  The following section presents the “base curves” to which these factors were applied.  4.9.3 Defining Fragility Curves for Victoria Recognizing the uncertainty in the process, three weighted average cases were defined to represent the vulnerability.  Case #2 is considered the best estimate and will be the main focus; Cases #1 and #3 will be used in a sensitivity analysis in Chapter 5.   1) Upper Bound:100% weight on Canterbury buildings 2) Best Estimate: 67%/33% weights on Canterbury and California, respectively 3) Lower Bound: 50%/50% weights on Canterbury and California, respectively Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 141 For unretrofitted and braced-parapet buildings, “California” refers to the Loma Prieta results.  For retrofitted buildings, “California” refers to the Northridge results.  The partial retrofit classification is the most difficult to accurately define as the scope of work is highly variable.  Because no reliable results for partially retrofitted buildings were available from California, this curve was taken as the weighted average of the braced parapet and fully retrofitted buildings, with weights of 67% and 33%, respectively; the rationale was that in most of the data encountered, “partial retrofits” were found to provide limited benefits over braced parapet buildings, but logic dictates some nominal increase in performance must be achieved.  Obviously, this decision was highly subjective, but this approach was thought to be better than the alternatives (such as not including a partial retrofits category).  Figures 4.41 to 4.43  provide the resulting structural MDF vs. Sa(1) curves for Victoria for the best, upper, and lower estimates.  Note that these curves do not yet account for the damageability effects of specific characteristics as discussed in Sections 4.7 and 4.8.  Also, only the mean damage factor is addressed for the time being. The distribution of damage into the various damage states (in the form of fragility curves) is also of interest, but was of secondary importance to the study at hand.                     Figure 4.41 – Basic Structural MDF vs. Sa(1) Curves (Best Estimate) 0%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00Mean Damage Factor [%]Sa(1) [g]Unretrofitted (Best)Braced Parapets (Best)Partial Retrofit (Best)Full Retrofit (Best)Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 142                                  0%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00Mean Damage Factor [%]Sa(1) [g]Unretrofitted (Upper)Braced Parapets (Upper)Partial Retrofit (Upper)Full Retrofit (Upper)0%10%20%30%40%50%60%70%80%90%100%0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00Mean Damage Factor [%]Sa(1) [g]Unretrofitted (Lower)Braced Parapets (Lower)Partial Retrofit (Lower)Full Retrofit (Lower)Figure 4.42 – Basic Structural MDF vs. Sa(1) Curves (Upper Bound) Figure 4.43 – Basic Structural MDF vs. Sa(1) Curves (Lower Bound) Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 143 Ultimately, it was decided to not generate the individual damage state fragilities based on the observed data.  The reasons are as follows:  1) In some cases, there were not a sufficient number of buildings to obtain a statistically reliable estimate for each damage state at every intensity level 2) Fitting directly based on data often provides less than satisfactory results, as was the case in even a more focused and sophisticated study by King (2005) 3) It was desired that the final motion-damage relationships be compatible with HAZUS, as it is now the standard in earthquake loss estimation.  Because all the data was in terms of ATC-13 damage states, it would have been necessary to map the ATC-13 damage states to the HAZUS damage states (through some subjective conversions)  Instead of generating individual damage state fragilities from the observed data, it was decided to generate the fragilities by matching the MDF from our observed results to the MDF associated with the fragilities, while still maintaining a distribution among the damage states that was reasonably similar to the observed data.  Figure 4.44 shows an example result for Unretrofitted buildings.                      Figure 4.44 – HAZUS Struct. Fragility for Unretrofitted Type D/F Victoria 0%10%20%30%40%50%60%70%80%90%100%0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2Probability of Exceeding Damage State/MDFSa(1) [g]SlightModerateExtensiveCompleteMDFChapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 144 For example, at Sa(1)=0.9g, about 50% of the Canterbury buildings were in damage states of Heavy (Damage Ratio=45%) or greater (see Appendix B).  Our HAZUS fragilities were proportioned to have approximately 50% of buildings in damage states of Extensive (Damage Ratio=50%) or greater – all while achieving the closest possible match between the MDF vs. Sa(1) relationships based on the observed damage statistics (i.e. Figures 4.41 to 4.43) and those resulting from the combination of the damage state fragilities and their associated damage factors.  Notice that the MDF curve in the above figure (i.e. the black dashed line) is nearly identical to the Unretrofitted, best estimate curve (i.e. the blue line) of Figure 4.41.  The damage factors are specified in HAZUS and represent the loss as a fraction of the structural value of the subject building.  They are as follows:  None: 0%  Slight: 2%  Moderate: 10%  Extensive: 50%  Complete: 100%  To understand the relative performance improvements, it is perhaps more instructive to compare the individual damage states for various degrees of retrofitting.  Figure 4.45 shows the various fragility curves for each damage state for a given retrofit level.  As expected, parapet bracing and partial retrofits achieve nearly the same structural damage reduction as full retrofits at low levels of shaking, but with increased shaking intensity they converge to the unretrofitted case.  Note that the fragility curves were developed in terms of lognormal distributions so as to be compatible with the overall HAZUS framework in future loss estimates (although it would be necessary to convert back to terms of spectral displacement).  While the preference would obviously be to produce fragilities based entirely on observed data, the results provided are thought to be a significant improvement over the existing data in HAZUS.  Of course, care and judgment should be exercised before making use of these results in other studies because they could be at odds with the remaining structure type fragilities of HAZUS in a relative sense (i.e. a URM should be worse than a modern building).  Results for the remaining building categories are provided in Appendix C.   Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 145                      4.10 Summary The purpose of this chapter was to establish motion-damage relationships for URM buildings based on observed damage statistics.  To this end, the following was undertaken and was presented in this chapter:  Damage assessment methods for URM buildings and sources of damage information were reviewed:  The ATC-13 damage scale was the primary scale used in collecting the damage data (which was performed by others)  Other scales included were those from Wailes and Horner (1933) and ATC-20 (1989) (or highly similar variants)  Various ground motion intensity measurements and their appropriateness for use in developing motion-damage relationships was discussed: 0%10%20%30%40%50%60%70%80%90%100%0 0.5 1 1.5 2Probability of Exceeding DSSa(1) [g]Unretrofitted (Slight)Parapets Braced (Slight)Partial Retrofit (Slight)Full Retrofit (Slight)0%10%20%30%40%50%60%70%80%90%100%0 0.5 1 1.5 2Sa(1) [g]Unretrofitted (Moderate)Parapets Braced (Moderate)Partial Retrofit (Moderate)Full Retrofit (Moderate)0%10%20%30%40%50%60%70%80%90%100%0 0.5 1 1.5 2Probability of Exceeding DSSa(1) [g]Unretrofitted (Extensive)Parapets Braced (Extensive)Partial Retrofit (Extensive)Full Retrofit (Extensive)0%10%20%30%40%50%60%70%80%90%100%0 0.5 1 1.5 2Sa(1) [g]Unretrofitted (Complete)Parapets Braced (Complete)Partial Retrofit (Complete)Full Retrofit (Complete)Figure 4.45 – Comparison of Damage State Fragilities by Retrofit Type Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 146  It was noted that because the purpose of this study was to produce loss estimates, it was necessary to make a compromise between more advanced IMs (eg. cumulative absolute velocity) and simpler parameters (eg. PGV, Sa)  More advanced IMs likely would be better predictors of damage.  Simpler IMs (eg. PGV, Sa) although likely not as accurate predictors of damage could be more readily related to existing information on seismic hazard and local (soils) effects  Ultimately, Sa(1) was selected as the IM for use in this study  Methods of estimating ground motion intensity at a site were reviewed and discussed:  It was found that the weighted interpolation method as developed by Rutherford and Chekene (1990) was sufficiently accurate for the purposes of this study  The process of developing damage probability matrices (DPMs) and fragility curves from observed damage statistics was discussed.  URM damage statistics collected by others were obtained by the author and reviewed:  Damage statistics from the 1989 Loma Prieta and 1994 Northridge earthquakes were kindly provided by Mr. Bret Lizundia of Rutherford and Chekene Engineers  Damage statistics from the 2010/2011 Canterbury earthquake sequence were kindly provided by Dr. Jason Ingham of the University of Auckland  Results from the Whittier earthquake were obtained from Wiggins et al. (1994)  Using the existing raw damage data, new motion-damage relationships were developed:  Relationships for structural damage vs. Sa(1)were developed for various URM strengthening levels for each of the data sets  The relationships were defined in terms of HAZUS fragilities, since this is the current standard in loss estimates for North America  As expected, the results indicated that limited retrofits (eg. parapet bracing, tension ties) achieve nearly the same results as full retrofit for low levels of shaking; however, the improvements disappear at high levels of shaking  The results from the various data sets were compared and it was found that the Canterbury buildings suffered greater damage; however, the variation decreased with increased strengthening, as would be expected  Relationships for structural damage vs. Sa(1) were defined for Victoria:  A qualitative comparison was made between URM buildings for Victoria and Canterbury/California Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 147  Low, high, and best estimates were defined using weighted average of the Canterbury and California databases  The best estimate was ultimately based on 67%/33% weights for Canterbury and California, respectively; this was partially because the California databases included just low to moderate levels of shaking and it seemed imprudent to place more confidence in extrapolations than actual damage data  4.11 Conclusions 4.11.1 General Conclusions The results showed that there can be significant differences in motion-damage relationships for similar types of buildings (i.e. URM) in different regions, which is thought to be due to both differences in construction and differences in the seismic demands that are not accounted for through our simple intensity measure (which is the 5% damped spectral acceleration at a period of 1 second).  Examples include duration, directivity and the possibility for multiple earthquakes in a sequence.    The results showed that, as expected, limited strengthening measures such as parapet bracing and tension ties provide levels of damage reduction at low intensity of shaking that are similar to that achieved by full retrofits.  However, the improvements disappear at higher levels of shaking.  Some differences were also noted between the observed results and the relationships from published sources such as ATC-13 (1985) and HAZUS (FEMA 2012), which all generally overestimated damage to some degree.  The results presented herein are thought to be an improvement over the existing results, especially since no results for braced-parapet buildings were available.  4.11.2 Conclusions for Victoria The most important conclusion for Victoria is likely that parapet-bracing can offer significant overall damage reduction for low-intensity shaking.  The performance can be similar to more comprehensive retrofits at a fraction of the cost and this “low-hanging fruit” has not yet been pursued in Victoria.    Of course, it is also important to recall that the effects of parapet bracing disappear at higher levels of shaking, which could certainly occur.    Chapter 4 – Quantifying Building Vulnerability Through Observed Damage Statistics 148 4.11.3 Future Research Opportunities Damageability effects of various building characteristics (eg. number of stories) have shown to be inconsistent to a certain degree in the literature.  This is likely due to limitations in the data (both quality and quantity) as well as differences in the level of shaking.  Much the same as the benefits of parapet upgrading disappear with increased shaking, it is postulated that differences in vulnerability due to slightly different building forms also likely disappear with increased shaking.  This was accounted for conceptually in this study, but a more rigorous investigation is needed.  The selection of an intensity measure and interpolating the demands at each given site remains a challenge.  The goal of this study was to prepare loss estimates using observed damage statistics – the goal was not to necessarily select the best possible IM or develop the best possible motion-damage relationship.  However, improvements here would obviously lead to improved loss estimates (among many other improvements).  149 Chapter 5  Cost-Benefit Analysis for URM Seismic Rehabilitation 5.1 Purpose and Scope In Chapter 3 it was seen that several communities had made the decision to implement URM seismic risk mitigation programs, which in many cases included some form of mandatory strengthening.  However, these decisions were often founded on an emotional and/or political response to past earthquake losses.  The purpose of this chapter is to provide an alternative, more rational motivation for URM seismic strengthening and thus avoid needless additional loss of life and property damage.  To that end, cost-benefit analyses were undertaken and are presented herein. The analyses made use of the motion-structural damage relationships in Chapter 4, supplemented with additional relationships from other sources and cost data derived herein.  The results are specific to Victoria, but the methodology can readily be extended to other communities.  5.2 Background Quantification of the costs and benefits of URM seismic rehabilitation is a topic that has received some attention in the past, particularly in the United States during the 1980’s and 1990’s when many jurisdictions in California were in the midst of implementing URM retrofit ordinances.  Example studies include those by Rutherford and Chekene (1990) and Recht Hausrath & Associates (1990, 1993).  More recently, the City of Seattle has embarked on the process of implementing a URM seismic retrofit ordinance and commissioned a cost-benefit study (Gibson Economics 2014).  While these studies illustrate the precedence for completing such an exercise, it was deemed worthwhile to perform a cost-benefit analysis in this study for the following reasons:  No such study has been performed for a western Canadian city (and at the outset of this study, none in the pacific northwest)  None of these studies were based on observed damage statistics – note that the Seattle study was based entirely on HAZUS results  None of the studies included a “Parapets Braced” type of retrofit alternative, which was of particular interest in the current study  Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 150 In this chapter, a summary is first provided of the costs and benefits typically considered as well as which stakeholders bear the costs and accrue the benefits.  Subsequently, several relevant studies on URM cost-benefit analysis are summarized, the loss estimate and cost-benefit analysis methodologies used for this study are defined, and the results of a cost-benefit analysis for URM buildings in Victoria, BC are presented.  Finally, a sensitivity analysis is presented and some of the limitations of cost-benefit analysis and decision-making based on expected cost are discussed.  5.3 Types of Costs and Benefits to be Considered Table 5.1 summarizes the typical costs and benefits of a seismic upgrade. The primary cost in the analysis is the cost of retrofitting, while the benefits are essentially due to reduced expected losses in the form of damage, casualties, and downtime.  There are a variety of stakeholders affected, including building owners, tenants, and the general public.  Note that impacts may be costs for some and benefits for others; the manner in which they are presented is arbitrary.   Table 5.1 – Costs & Benefits for URM Seismic Upgrades Cost Description Borne by Building Retrofit  Construction costs  Consulting costs  Permit costs  Building owners  Building tenants (increased rent)  Taxpayers  (eg. grants/loans) Disruption  Loss of income due to closure  Loss of income due to reduced rent  Noise pollution  Building tenants  Building owners Loss of Historic Fabric  Demolitions due to seismic damage  Demolitions instead of retrofits  General Public Benefit Description Accrued by Reduced Damage  Reduced “expected” damage to building  Reduced “expected" damage to contents    Building owners  Buildings tenants  Insurers Reduced Casualties  Reduced deaths  Reduced treatment of injuries  Reduced pain and suffering  Buildings tenants  General public  Insurers Increased resilience  Reduced loss of sales during recovery  Reduced loss of rent during recovery  Reduced regional/national funding  Reduced loss of industry (businesses may choose to rebuild elsewhere after earthquake)  Buildings tenants  Building owners  General public Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 151 Other costs and benefits could be defined. For example, local construction companies would likely benefit due to the high demand. However, the aforementioned impacts are those most likely to influence decision-making and so we limit our treatment to these specific impacts.  The degree to which different stakeholders are affected by these impacts will vary depending on a number of factors, including the building’s use.  Note that throughout the cost-benefit analysis only buildings of commercial (retail trade, or “COM1” as defined in HAZUS) occupancies are considered, as this is the most representative of the subject buildings in Victoria.  The choice of occupancy affects the replacement value of the building and the relative proportions of the costs (i.e. structural, non-structural, contents). The difference between commercial and residential occupancies is of interest, since rehabilitated buildings are often converted from commercial to residential. Note that this applies more so to “full” rehabilitations, while a change of occupancy is much less likely to accompany a partial rehabilitation.  An examination of the 2012 HAZUS technical manual (FEMA 2012) shows that the replacement value, including contents, of a residential (RES3) building is about 25% greater than that of a commercial (COM1) building.  The difference is mainly due to increased drift-sensitive non-structural components (i.e. partition walls).  The increased exposed assets would increase losses and, thus, curtail the benefits of a retrofit to some degree.  With regards to life-safety, a review of the literature (NRC 1993, Rutherford & Chekene 1990) suggests that the occupant exposure (i.e. the number of occupants times the duration of occupancy) for residential buildings is equal to or less than that for commercial buildings.  As such, there is likely some reduction in loss associated with the reduced occupant exposure and, thus, increased life-safety benefits.  Overall, it was felt that accounting for a conversion of occupancy associated with full retrofits in our cost-benefit analysis was not warranted, due to the lack of reliable information on the parameters driving the changes such as fragility of drift-sensitive non-structural components (NSC's) in URM buildings and detailed occupancy data.  This level of evaluation is one that should be undertaken on a building-by-building basis.  Of course, another issue not accounted for in the cost-benefit analysis is that commercial-residential occupancy changes are typically performed at the prospect of generating additional revenue. Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 152 5.4 Literature Review for Cost-Benefit Analyses The opening section mentioned a few of the most relevant cost-benefit studies that were encountered in the literature.  In this section, a brief summary of their methodologies and results is provided.  In particular, we will review the studies completed for San Francisco (Rutherford & Chekene 1990) and Seattle (Gibson Economics 2014).   5.4.1 San Francisco Study Two studies into the costs and benefits of URM seismic rehabilitation were commissioned by the City of San Francisco in 1990, when it was in the midst of implementing its URM retrofit ordinance.  The work by Rutherford and Chekene (1990) characterized various retrofitting alternatives, including costs, disruption, and seismic performance.  The work by Recht Hausrath and Associates (1990) addressed the socioeconomic impacts, including the expected number of building demolitions and effects on housing costs.  Although neither of these reports explicitly provided a cost-benefit analysis, they provide essentially all of the components, with the exception of the economic analysis parameters (eg. time value of money).   The study by Rutherford and Chekene was quite broad in that it defined 15 different prototype buildings (recall Figure 2.2) and four strengthening statuses.  The four strengthening statuses were defined as: 1) Status Quo: Unstrengthened (although many buildings had received parapet strengthening) 2) Strengthening Option #1: Out-of-plane wall strengthening (including tension anchors, shear anchors, and strongbacks for slender walls); commonly referred to as “bolts-plus”  3) Strengthening Option #2: Retrofit to the 1991 Uniform Code for Building Conservation (ICBO 1991) – this is similar to the current Canadian practices 4) Strengthening Option #3: Retrofit to San Francisco Building Code, Section 104f – this is similar to current code requirements for new buildings of the day  The purpose of the highly detailed prototyping was to better characterize the costs of seismic upgrading.  This was possible and reasonable for the San Francisco buildings because Rutherford and Chekene had a complete database of the 2005 URM buildings in San Francisco and no dominant occupancy was observed.  No such database is available for the Victoria buildings.  However, a small sampling of buildings in a targeted area was completed as part of this thesis (see Section 7.2.2). Comparing the occupancies of this survey to the San Francisco database (Figure 5.1) shows that commercial buildings are Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 153 more predominant in Victoria.  It should be noted that the Victoria statistics are based on floor area, while the San Francisco results are based on the primary occupancy for each building.  The Victoria statistics may actually overestimate the prevalence of residential occupancies, since upper stories of buildings were assumed to be residential if no signage indicating otherwise was available at the ground level.  These results reinforce the decision for selecting commercial occupancy as the only type for our analysis.                  The damage relationships were somewhat less refined: rather than defining 15 different relationships (for each strengthening status), the prototypes were grouped in terms of similar damageability based on engineering judgement and motion-damage curves were assigned to each group.  Since retrofits were thought to homogenize the risk, fewer groups were developed for increased levels of retrofitting. Twelve damage curves (MDF vs. MMI) were developed in total, ranging between the ATC-13 values for URM (presumably low-rise) to Reinforced Masonry.  Figure 5.2 provides a plot of the damage curves, including the ATC-13 curves.  The nomenclature for the curves is as follows: the first character represents the strengthening status and the second character represents the enumeration of damage curve within a given strengthening status (eg. “1-2” represents the second damage curve for strengthening option #1).  The higher enumeration indicates a higher vulnerability within that strengthening status (eg. 1-2 is more vulnerable than 1-1).  Note that there is some overlap in vulnerability between the strengthening levels (eg. 1-2 is slightly more vulnerable than U-1).  Comm'l40%Res.39%Other22%Comm'l58%Res.31%Other11%Figure 5.1 – Occupancy for San Francisco (left) vs. Victoria (right) URMs Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 154                    It is not explicitly stated, but these curves are assumed to represent damage costs as a fraction of the structural plus non-structural values, since this is the case in ATC-13 (i.e. contents were not included).  All curves fall below the ATC-13 URM curve, consistent with our previous conclusions (Section 4.8.6) that the ATC-13 values overpredicted damage.  However, it is interesting to note that the reinforced masonry case was selected as a lower bound; given the observation that ATC-13 overestimated damage to URM, it would seem plausible that damage for reinforced masonry is also overestimated.  It is unclear if there should in fact be any relationship between damage to reinforced masonry and retrofitted URM buildings, as the structural systems are typically very different, with the possible exception of hollow core retrofits of URM walls.  Note that the curves used in the San Francisco study imply a significant degree of reduction not only for structural damage, but non-structural as well: a damage reduction of about 35% (50% to 15%) likely represents a value that is greater than the entire structural value.  Structural components typically represent 15-25% of the building value (Onur 2001, Thibert 2008).  At the opposite end of the spectrum, HAZUS (FEMA 2012) suggests modest reductions in non-structural damage as a result of retrofitting, particularly for acceleration-sensitive components, as will be shown in Sections 5.5.3.2 to Figure 5.2 – Motion-Damage Relationships Used in San Francisco Study (After Rutherford and Chekene 1990)  0%10%20%30%40%50%60%70%80%90%100%6 7 8 9 10 11 12Damage Ratio [%]MMIU-1U-2U-3U-4U-51-11-21-32-12-23-13-2URMRFMURM and RFM from ATC-13Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 155 5.5.3.4.  The issue of non-structural loss mitigation from structural retrofits is an important question, because NSC’s can easily represent 75% of the building value.  This issue will be addressed later in this chapter when the non-structural/contents damage curves are defined for this thesis.  Rutherford and Chekene (1990) also accounted for casualties and downtime.  The casualty estimates were developed by including outdoor casualties and calibrating ATC-13 fatality rates to match historical Californian earthquakes. Table 5.2 shows the fatality rates used.  The study also provided indoor and outdoor occupant densities to which the fatality rates were applied.  Finally, hospitalized injuries were assumed to be four times the number of fatalities.   Table 5.2 – Fatality Rates Used in San Francisco Study (From Rutherford and Chekene 1990) Strengthening Status Location Mean Damage Factor 0% 5% 20% 45% 80% 100% Unretrofitted Building 0 0.000010 0.00035 0.0035 0.035 0.20 Street 0 0.000200 0.00300 0.0700 0.120 0.15 Retrofit Alt. #1 Out-of-Plane Building 0 0.000009 0.00033 0.0034 0.035 0.20 Street 0 0.000018 0.00280 0.0650 0.120 0.15 Retrofit Alt. #2 UCBC Building 0 0.000007 0.00023 0.0028 0.035 0.20 Street 0 0.000014 0.00200 0.0550 0.120 0.15 Retrofit Alt. #3 SFBC 104(f) Building 0 0.000008 0.00030 0.0032 0.035 0.20 Street 0 0.000016 0.00270 0.0600 0.120 0.15 Note: trailing zeros have only been shown to permit easy comparison of the relative values  The fatality rates above are based on the overall MDF (i.e. structural plus non-structural damage). Conversely, HAZUS makes use of the structural-only damage states to estimate casualties.  While it could be argued that structural-only damage is better correlated with fatalities, the fact that the above rates were calibrated to actual earthquakes lends a significant amount of merit.  Estimates of downtime were accounted for using the ATC-13 data, with the exception that a “critical loss level” was included; this represents the damage level above which a building is likely to be demolished because repair is not economically viable.  The result is that buildings below the critical loss level will be repaired (with an associated downtime), while those above this point will be demolished.  Rutherford and Chekene (1990) chose to set the critical loss level at 40% and 50% for unstrengthened and strengthened buildings, respectively.  Some studies (EERI 1989) have suggested critical Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 156 loss ratios of up to 65% - lower values were used by Rutherford and Chekene (1990) because URM buildings have a variety of other deficiencies aside from seismic vulnerability that may also require remediation and may be more likely to be not economically viable prior to the earthquake.  In any case, the notion of a critical loss ratio appears intuitive and realistic and this will be implemented in the study at hand.  The study also quantified costs for the various retrofit options.  The cost analysis was quite detailed in that square-foot costs for each prototype were generated based on data from the complete database of URM buildings in San Francisco and unit cost estimates submitted by a cost consultant. For example, Rutherford and Chekene had access to the height-to-thickness ratios of walls for each building, analyzed each prototype to see what fraction would require out-of-plane retrofits (eg. strongbacks) and included the cost of such work on a probabilistic basis for each prototype.  Cost adjustments for work in occupied space, historically sensitive buildings, and work concurrent with architectural renovations were also provided.  Table 5.3 shows the basic square-foot cost estimates developed by Rutherford and Chekene; note that the costs have been adjusted to 2014 Canadian dollars using RSMeans’ historical cost index (Reed Construction Data 2012) for comparison with the author’s results later in this chapter.    Table 5.3 – Square-Foot Costs for Retrofits in San Francisco Study Adjusted to 2014 Canadian Dollars (Modified From: Rutherford and Chekene 1990) Prototype (see Figure 2.2) Retrofit Alt. #1: Out-of-Plane  Retrofit Alt. #2: UCBC Retrofit Alt. #3: SFBC 104(f) Seismic Arch. Seismic Arch. Seismic Arch. A 18.88 1.38 20.51 1.54 27.30 1.83 B 10.95 0.56 15.44 0.76 18.18 0.97 C 9.06 1.32 14.64 2.20 23.12 5.13 D 10.95 0.74 14.99 1.38 26.65 3.56 E 18.59 0.37 21.50 0.95 29.34 2.08 F 8.46 0.21 14.13 1.46 17.13 2.39 G 23.37 1.98 25.99 2.14 34.35 3.62 H 10.65 0.78 15.71 1.24 19.75 3.27 I 16.99 1.65 28.99 2.24 40.89 4.69 J 8.65 0.80 16.14 1.63 25.74 4.02 K 20.53 3.77 22.55 4.12 32.02 6.18 L 11.88 2.20 15.36 2.68 20.92 4.84 M 11.84 1.65 29.63 2.97 33.54 5.19 N 7.68 1.03 18.49 1.44 29.10 4.88 O 14.50 2.00 20.57 2.10 27.84 4.82 Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 157 As can be seen, the costs were broken into “seismic” and “architectural” costs: the seismic costs represent the work necessary to complete the retrofit (eg. including removal of finishes) while the architectural costs represent the work necessary to return the space to its original condition (eg. to re-instate or replace finishes that were removed).  The architectural costs do not include extensive renovation works, which are frequently combined with comprehensive retrofit works.    Notice that the jumps in cost for increased strengthening vary by prototype.  This is largely dependent on which elements would need work for a particular retrofit option.  Prototype A, for example, has a relatively small jump between options #1 and #2; this is because the UCBC retrofit standard (i.e. Option #2) allows 1-storey buildings to have seismic resisting elements on only three sides, despite having a flexible roof diaphragm – as a result, new steel frames or concrete shearwalls would likely not be required.  Obviously, the level of architectural finishes also affects the cost.  The most relevant prototypes for our study are likely prototypes ‘G’ and ‘H’ (two/three storey office and commercial buildings, small and large area, respectively).  Again, such a database was not available for Victoria buildings and so this level of detail was not possible.  Rather, we will focus on the “typical” commercial-occupancy type of building.  In our study, we will develop unit cost estimates for certain selected retrofit works similar to those defined by Rutherford and Chekene.  Unfortunately, Rutherford and Chekene did not provide cost-benefit results for the various retrofit alternatives. However, the aforementioned parameters represent essentially all those required, except economic ones such as the time value of money.  5.4.2 Seattle Study During the course of the writing of this thesis, the City of Seattle commissioned and released a cost-benefit analysis on URM seismic rehabilitation (Gibson Economics 2014).  Below are some key points summarizing the analyses and the results.  A discussion of the results and the merits and flaws of the study is provided below.  Results:  The study’s best estimate of the cost-benefit ratio for retrofitting to bolts-plus was 7.6%, suggesting that the costs of retrofitting greatly outweigh the benefits;  About 45% of the benefits were ascribed to economic savings (reduced damage and downtime) and 55% was ascribed to casualty savings Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 158  A sensitivity analysis, including many parameters, suggested that the benefit-cost ratio could vary from 1.7% (worse case) to 66.8% (best case)  Analyses:  The report considered only two levels of strengthening: “bolts plus” (most similar to San Francisco Option #1) and “reinforced” (most similar to San Francisco Option #2)  The report contained an explicit cost-benefit analysis for each retrofit alternative  The loss estimates were based on HAZUS data (eg. fragility curves, replacement values breakdown between structural and non-structural components)  The study used three representative earthquakes and their annual probabilities to “approximate” the hazard curve  Retrofit costs were based on discussion with local engineers and building owners  Building values were specified based on local assessed values and assumptions were made regarding the contents values  The number of stories of a building and the soils were accounted for through modified fragility curves  5.4.2.1 Merits The greatest general merits of the study are that the results are current and that the economic analysis appears reasonably rigorous in that an extensive sensitivity analysis was performed. With regards to our study, it is particularly relevant because Seattle is likely the best proxy city for Victoria, due its relatively similar economic, social, and seismic environments.  5.4.2.2 Flaws Despite the merits discussed above, there are also – in the author’s opinion – a number of flaws (i.e. opportunities for improvement in the current study).  The foremost flaw is likely the lack of technical rigor in defining the performance of the buildings.  Although HAZUS is a very powerful and detailed tool, it appears that the ‘unretrofitted’ and ‘fully retrofitted’ strengthening statuses were characterized by default data (although a sufficient amount of information was not presented in the report to confirm this).  In Chapter 4, we saw that the HAZUS curves overestimated damage in comparison to the observed data.  As HAZUS tends to overestimate structural damage in both cases (both retrofitted and unretrofitted), the consequences may or may not be significant.   Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 159 Additionally, there appeared to be inconsistencies in damage trends for the acceleration-sensitive components and building contents – under some of the scenario earthquakes, the results indicated a slight increase in damage for the ‘bolts plus’ retrofit versus unretrofitted.  In the author’s opinion, this is highly unlikely since a bolts-plus retrofit would not add sufficient stiffness so as to induce higher accelerations.  In fact, one would expect less damage due to a lower probability of structural failure (since the NSC’s will certainly be demolished if collapse occurs).  Compounding the aforementioned issue is that the report states that the value of the building contents was assumed to be equal to 50% of the building value (i.e. the structure plus nonstructural components).  However, the HAZUS technical manual commonly specifies a contents value of 50-100% of the structure value (FEMA 2012).  A contents value of 45% of the building value would be considered an upper bound, appropriate for high-technology buildings, while a more common value for contents would be 10-20% of the building value (Thibert 2008).  Figure 5.3 shows breakdowns of the building value similar to that indicated in the report and a more plausible breakdown (within the 10-20% range noted by Thibert).  Since the contents are also acceleration sensitive, they will be subject to the same discrepancy noted above.  It is also significant because HAZUS specifies that the maximum loss of contents is 50% (because contents may be salvaged).  Again, the report did not provide sufficient information so as to confirm the exact effects, but it appears that a more accurate representation could be achieved in our study.               Struct.(14.7%)Accel NSC (21.6%)Drift NSC(13.8%)Contents(50.0%)Figure 5.3 – Components for Seattle (left) and a More Common Assumption (right) Struct.(15%)Accel NSC (35%)Drift NSC(30%)Contents(20%)Chapter 5 – Cost-Benefit Analysis for URM Seismic Rehabilitation 160 Another potential flaw in the Seattle study lies in the retrofit costs.  The best estimate cost for a bolts-plus retrofit is stated to be $40/sq.ft., with upper and lower bounds of $20/sq.ft. and $60/sq.ft., respectively.  This was reportedly based on discussion with owners and engineers.  A review of available data for Victoria (see Appendix A) shows that $40/sq.ft. exceeds the mean cost for a complete retrofit to current code.  Similarly, the data in Table 5.3 shows that the cost of a ‘bolts-plus’ type retrofit ranges between $10/sq.ft. and $25/sq.ft.  While it may be pointed out that these costs do not include extensive renovations, this also appears to be the case in the Seattle study.  The treatment of the effect of soils is also worth discussing: in the study, the effect of soils is accounted for through adjusted fragility curves for “liquefiable” or “non-liquefiable” soils.  The study does not describe how these adjuste