5.1 Overview of Methodology
5.1.1 Objective
5.1.2 Roles and responsibilities
5.1.2.1 Exposure
5.1.2.2 Seismic Hazards
5.1.2.3 Seismic Vulnerability of Buildings
5.1.2.4 Seismic Vulnerability of Contents
5.1.2.5 Business Interruption Loss
5.1.2.6 Use of Findings from Previous PML Investigations
5.1.2.7 Risk Assessment for Financial Stakeholders
5.1.2.8 Management of Uncertainty
5.2 Catastrophe Models for Portfolio Seismic Risk Assessment
5.2.1 Catastrophe Models Used for Earthquake Insurance
5.2.2 Damage Models Controlled by Engineers
5.2.2.1 HAZUS®MH AEBM
5.2.2.2 Code-Oriented Damage Assessment (CODA)
5.2.2.3 FEMA P-58, Seismic Performance Assessment of Buildings
5.3 Financial Stakeholder Considerations
5.3.1 Financial models and consideration of uncertainties
5.3.2 Owner risks
5.3.3 Lender risks
5.3.4 Insurer risks
5. Assessment Methodology, Approaches and Tools
5.1 Overview of Methodology
5.1.1 Objective – The objective of portfolio seismic risk assessment is to quantify the aggregate risks from future earthquakes to a geographically dispersed group of properties.
5.1.2 Roles and responsibilities – Users (e.g., real estate lenders or owners) request portfolio seismic risk studies to inform risk management decisions regarding acquisition, disposition, earthquake insurance and risk mitigation. Risk Providers work with catastrophe modeling software or catastrophe modeling consulting firms to assess the risks, identify the factors driving the risks and opportunities for risk reduction. Others such as insurers and insurance brokers may use the results from portfolio seismic risk assessments in insurance placements, and rating agencies may use the reports from portfolio seismic risk studies in rating financial instruments collateralized by a real estate portfolio.
5.1.2.1 Exposure – the locations of the real estate properties to be evaluated and their exposure values (e.g., replacement costs) are generally determined by the User (owner or lender) and transmitted to the risk Provider. It is important to note that inaccurate values directly impact the accuracy of the financial risk estimates resulting from portfolio studies. The provider should review the exposure data and request clarification or correction from the User where locations or values are missing or appear to be in error. The User may wish to engage the services of valuation professionals or automated systems to ensure accurate and consistent values for the buildings, contents and business activities at risk, to avoid systematic underestimation or over-estimation of the exposure values.
5.1.2.2 Seismic Hazards – Catalogues of earthquake simulations (“event sets”) with event probabilities and estimates of the resulting ground motion are included in the catastrophe models. These may be derived from the National Seismic Hazard Mapping Project by the United States Geological Survey [e.g., Peterson et al., 2008, 2014] or from other accepted standard.
Typical software for portfolio seismic risk assessment includes digital maps for geologic conditions such as Site Class and liquefaction susceptibility. With geocoding of the site locations, the software can identify sites with adverse geologic conditions such as soft soils or susceptibility to liquefaction, to allow further investigation by the seismic risk Provider. The software may allow over-ride of the mapped condition, to accommodate site-specific information from geotechnical investigation reports or other sources (e.g., past PML reports). In many cases, foundation systems are specifically engineered to mitigate adverse geologic conditions, and the Provider may specify modeling parameters to account for these design features.
Note that the risks resulting from certain seismic hazards are normally excluded from existing catastrophe software and require site-specific engineering investigation, such as surface fault rupture. Risks associated with special hazards such as slope instability, liquefaction effects such as lateral spreading, or tsunami effects may be addressed some catastrophe software in approximate ways. Within the context of a large, geographically distributed portfolio, the impact of losses from such hazards may be small compared to the aggregate losses to the portfolio from ground shaking. However, surface fault rupture or other special hazards may affect a critical facility functioning as a critical hub within a non-redundant system, resulting in large business interruption losses. Note that where active risk management routinely requires the screening of all significant acquisitions, the resulting real estate portfolios may systematically exclude properties with significant risks from surface fault rupture, slope instability or liquefaction effects, as these would be deemed “unstable sites."
5.1.2.3 Seismic Vulnerability of Buildings – the basic structure of a building is composed of a frame with gravity load carrying elements (slabs or framed floors, columns and/or bearing walls, and foundations), and a lateral force-resisting system (e.g., moment frames, braced frames, shear walls, etc.). In addition, there are architectural elements (e.g., cladding, partitions, and ceilings) and building service equipment. Some damage models do not segregate these elements, but estimate damage to the building as a whole (e.g., ATC-13). Other models (e.g., HAZUS™®, FEMA P-58) distinguish the structural frame from the nonstructural elements (architectural elements and building service equipment).
Building damage models utilized in catastrophe modeling software developed for property insurance applications typically rely on building height (or number of stories), year built and building type or class. Structural classification systems vary, as does the ability of insurance models to accommodate the findings from engineering reviews. The models can be used with building classification systems based on occupancy (i.e., residential, office, commercial, industrial, etc.). Weighted average damage relationships may be derived from logic trees from candidate structural classes (i.e., steel moment frame, steel braced frame, masonry shear wall, etc.) with an associated high uncertainty. In seismic risk assessments involving engineers (civil or structural), determinations of the materials of construction and the gravity and lateral systems allow for direct assignment of the structural class, with improved accuracy and reduced uncertainty. Where catastrophe modeling software offers damage relationships based on engineering building classification systems, “secondary modifiers” may be input to modify the models to make them more building-specific. Users and Providers are referred to the catastrophe modeling vendors for the specific details of the modeling and the opportunity each tool affords for engineering input.
Damage models that follow engineering methods (e.g., HAZUS™®, FEMA P-58) typically distinguish the structural frame from the nonstructural elements. A structural response model is used to estimate building peak responses (i.e., displacements and accelerations) from ground motions, and these responses are converted to engineering demand parameters (EDPs). Damage to the structural system is usually caused by the inter-story drifts that occur under lateral earthquake loads, so drift is the engineering demand parameter used to predict the damage states for structural drift-sensitive elements. Architectural elements may be drift-sensitive (e.g. full-height partitions) so drift is the EDP used to predict damage states. Nonstructural elements such as roof-mounted equipment, or suspended ceilings are acceleration sensitive, so the engineering demand parameters used to predict damage to nonstructural elements are floor accelerations for above-grade items, and ground accelerations for items at or below-grade. The amount of damage is found using a fragility model, relating damage states (with associated repair costs) to median values of the relevant engineering demand parameter. For example, in HAZUS, the damage states are described as ‘none,’ ‘slight,’ ‘moderate,’ ‘extensive’ and ‘complete.’ Each damage state is associated with a repair cost and an expected time to complete repairs, and may be tied to life-safety consequences (injury rates and fatality rates). The fragility models include a dispersion parameter (β) to account for the variability of damage state as a function of the relevant engineering demand parameter.
5.1.2.4 Seismic Vulnerability of Contents – Some damage models (e.g., ATC-13) relate ground motions to contents damage without consideration of the building in which they are contained. Damage models that follow engineering methods (e.g., FEMA P-58) use ground accelerations or floor accelerations as the EDP, so that the building response is considered when estimating damage to contents located above grade. Engineering methods use the EDP with fragility models to predict the damage state, and the damage state is associated with a repair cost.
5.1.2.5 Business Interruption Loss Assessment – Damage models derived from ATC-13 use Social Function Classes and building damage factor to predict restoration time for facility function. Damage models that follow engineering methods use structural (or in some cases nonstructural) damage states to predict restoration time for facility function.
5.1.2.6 Use of Findings from Previous PML Investigations – The major catastrophe models currently used in insurance risk assessment (see Section 5.2) are capable of producing a damage estimate for individual buildings, as well as listing the corresponding hazard levels. This allows for the provider to check results, or for some degree of calibration, e.g. adjusting “secondary modifiers” to achieve better agreement with previous seismic risk assessment estimates. The user should be careful to utilize similar outputs (e.g., Probable Loss) with similar uncertainty levels, and to disable secondary loss features (e.g., fire-following earthquake, or demand surge) when comparing results to the results of PML studies. Other models may be able to re-use the vulnerability relationships developed in detailed single-site studies directly in portfolio seismic risk assessments, maintaining calibration.
5.1.2.7 Risk Assessment for Financial Stakeholders – When earthquake damage occurs, there may be a number of parties with a financial stake in the loss: apart from the owner, there may be a lender and/or an insurer. Other stakeholders may include tenants, owners or occupants of neighboring buildings, etc. The models that apportion the losses to the various parties at risk are commonly called stakeholder models or loss allocation models. Since damage estimates from earthquake models are uncertain, the predicted losses have a statistical distribution that must be considered in allocating the financial consequences. For example, in the case of a mortgage lender, the owner is unlikely to default on the mortgage if the loss amount is less than the owner’s equity in the building (where equity is found as market value minus mortgage balance). If the statistical distribution of losses includes loss levels above the owner’s equity, then a lender’s loss is predicted. Expected losses to the parties are found by examining all potential loss levels and their probabilities. A similar approach is used for earthquake insurance, where the statistical distribution of losses is used with the specified deductibles and limits of liability to allocate losses between the owner and insurer. The user should identify the key stakeholders and the outputs required to meet their needs, as well as the insurance coverage details or information (e.g., owner equity) to allow appropriate stakeholder risk analysis.
5.1.2.8 Management of Uncertainty – Major contributors to uncertainty in portfolio seismic risk assessments include the uncertainties in exposures (i.e., the locations and values of the properties), earthquake hazards (e.g., ground shaking and the attendant geotechnical effects such as liquefaction), the vulnerability of buildings and equipment, and business interruption with its financial impacts. The available catastrophe models (see Section 5.2 below) all attempt to consider these uncertainties in a comprehensive way. Since portfolio-wide or aggregate losses are strongly affected by effects that are correlated portfolio-wide, care is needed where the effects of uncertainty are correlated, for instance in accounting for interevent uncertainty in ground motion prediction equations.
5.2 Available Catastrophe Risk Models (Tools) for Portfolio Seismic Risk Assessment
5.2.1 Catastrophe Models Used for Earthquake Insurance
Commercial earthquake insurance catastrophe modelers offer catastrophe models for use in assessing risks to real estate portfolios for earthquakes and other perils. The models all utilize large inventories of earthquake simulations (“event sets”) to represent the ground shaking and other hazards at each site within a portfolio, and then use damage relationships and the values of the building and contents, and the costs of business interruption, to predict the resultant losses at each site, and total portfolio-wide losses. Uncertainties are tracked and managed, and financial loss-allocation models are used to distribute the losses to each stakeholder (e.g., owner, insurer, or lender). These models are specifically developed for insurance risk assessment, rather than for engineering applications, but they establish a baseline for the state-of-the-art and practice of portfolio seismic risk assessment. The models offer various ways to include and consider site-specific geologic conditions that may differ from values obtained from their digital maps of local hazards, and building-specific features that affect damageability, and Providers who use these models must adapt the scope of their investigations to suit the specific capabilities of the model they choose.
5.2.2 Damage Models Controlled by Engineers
Some earthquake damage models allow full control of the damageability of each building by an engineering service Provider. Such models may be used to calibrate damageability of some or all of the buildings to previous results from engineering seismic risk assessments. Several of the publicly-available models are described below. Other models may be proprietary, offered by particular engineering service Providers. Additionally, commercial earthquake insurance catastrophe modelers may allow users to specify a customized vulnerability relationship.
5.2.2.1 HAZUS®MH AEBM
The seismic damage model for buildings in HAZUS®MH was originally developed for use by emergency planners and government decision-makers, looking at large, regional building portfolios subjected to individual earthquake scenarios. There is also an "engineer's" version of HAZUS, called the Advanced Engineering Building Module or AEBM, suitable to model the expected performance of particular buildings. The HAZUS model building types correspond roughly to the common engineering building types defined in ASCE 41-13 (e.g., W1, W2, S1, S2, etc.), with subtypes defined by height (Low-rise, Mid-rise, and High-rise) and usage as defined by an Occupancy Class. HAZUS models are further discussed in the Technical Appendix. HAZUS outputs include probabilities of different damage states (“none”, “slight”, “moderate”, “extensive”, or “complete”), duration of downtime, and casualties (injuries and deaths).
As provided by FEMA, HAZUS can produce estimates of expected loss for one or more specified earthquake scenarios, which the user can define by specifying the location of the causative fault rupture segment and event magnitude, and by selecting an attenuation relationship, or through use of a map depicting the distribution of ground shaking. HAZUS can also provide estimates of average annual loss based on the probabilistic ground motions such as those provided by the U.S. Geological Survey’s National Seismic Hazard Mapping Project. Some Providers also offer versions of HAZUS that extend these capabilities, to allow probabilistic portfolio analysis and insurance risk analysis.
5.2.2.2 Code-Oriented Damage Assessment (CODA)
CODA, for Code-Oriented Damage Assessment [Graf & Lee, Earthquake Spectra, EERI, 2009] is an adaptation of ATC-13, estimating expected building damage as a function of a demand-to-capacity ratio (DCR). The CODA model uses ground motion response spectra at the fundamental building structural period as the “demand,” while “capacity" is defined as Cs x R, where Cs is the base shear coefficient and R is the response modification factor as defined in current codes. The uncertainty in building damage for a given demand-to-capacity ratio is a function of the evel of investigation, similar to the “Levels” defined in ASTM E2026 e.g., BD0, BD1, etc.). A scaling factor may be used to achieve calibration of the damage relationship for a specific building with past seismic risk assessment results at a relevant pre-defined hazard level.
5.2.2.3 FEMA P-58, Seismic Performance Assessment of Buildings
FEMA P-58 [FEMA, 2012] is the state-of-the-art methodology developed to be used in performance-based seismic design. “The procedures are probabilistic, uncertainties are explicitly considered, and performance is expressed as the probable consequences, in terms of human losses (deaths and serious injuries), direct economic losses (building repair or replacement costs), and indirect losses (repair time and unsafe placarding) resulting from building damage due to earthquake shaking. The methodology is general enough to be applied to any building type, regardless of age, construction or occupancy; however, basic data on structural and nonstructural damageability and consequence are necessary for its implementation.” (from the Preface of FEMA P-58-1, 2012)
5.2.2.4 TZR, The Revised Thiel-Zsutty Earthquake Damage Model
The empirical Thiel-Zsutty damage estimation model was originally published in 1987 and was revised in 2017, [Thiel, Zsutty, 1987, 2017]. The revised Thiel–Zsutty (TZR) model provides a Beta probability distribution function (applicable for PL evaluation) for the damage ratio of a building as a function of peak ground acceleration and parameters that represent the site soil, building damageability. Parameter base values are given for ATC-13 building types and for ASCE 7 structural systems and soil classes. The model assigns uncertainty according to level and quality of investigation related to ASTM E2026 investigation levels.
5.3 Financial Stakeholder Considerations
5.3.1 Financial models and consideration of uncertainties. Portfolio seismic risk assessment generally includes consideration of stakeholder financial positions. For example, studies performed to assist in decisions for the purchase of earthquake insurance will typically allocate losses between the property owner and the insurer(s), and studies for lenders may allocate earthquake losses between the lender and the property owner. The allocation models must follow statistically sound methods, addressing the statistical distribution of losses subject to uncertainty.
5.3.2 Owner risks – these are the losses retained by an owner. Where earthquake insurance applies, the policy will specify coverage with a deductible and limit of liability. The owner’s (retained) loss is that portion of the loss below deductible and above the limit, with the balance paid by the insurer. Deductibles typically apply per event, per building and per unit of coverage. Limits may apply portfolio-wide, and site-by-site limits may also apply. In order to account for this in allocation of losses, an event-by-event approach is typically used, with an “event set” or a stochastic catalog of earthquake simulations.
5.3.3 Lender risks – these are the losses that accrue to a lender when owner(s) default on a loan. Loans are typically made for an amount less than the market value of the property, so that the owner’s equity (market value minus loan balance) provides an incentive for the owner to meet mortgage payment obligations and avoid default. After an earthquake, an owner may be faced with repair costs and loss of revenue that prevent them from meeting mortgage payments, or which result in “negative equity” for the owner, where repair and sale of the property would not provide the funds to pay off the mortgage. In such cases, the owner may elect to default on the mortgage. When a lender chooses to foreclose on a loan that is in default, the lender takes possession of the property and may then sell the property to pay off some or all of the loan balance. In effect, owner equity serves as a buffer to limit lender risks. In portfolio seismic risk analysis, where the cost of damage and loss of property revenue exceed the owner’s equity, the lender is exposed to a loss equal to the mortgage balance and the cost to repair and sell the property, net of the price of the sale of the property. The model to estimate lender loss must follow well-defined logic to predict defaults and address the statistical distribution of losses subject to uncertainty.
5.3.4 Insurer risks – these are the losses retained by an earthquake insurer, as specified under a policy in place at the time of an earthquake. These are less than the “ground-up” losses, as reduced by the application of policy deductibles and limits of liability. The insurer allocation model must implement the logic of the earthquake insurance policy, while following statistically sound methods, addressing the statistical distribution of losses subject to uncertainty.