8. Level of Investigation and Impact on Uncertainty in Risks
This section describes two levels of portfolio seismic risk assessment (SRA): Desktop Portfolio Seismic Risk Assessment and Engineered Seismic Risk Assessment, where more extensive levels of investigation are intended to reduce the degree of uncertainty in the portfolio aggregate risks. We note that a portfolio may first be analyzed at the “Desktop” level, without engineering input, to indicate key risk drivers and help decide on the level of effort to be devoted to the engineering investigation. With these two levels defined, Section 8.3 outlines how to describe the quality of a portfolio seismic risk investigation that allows for non-uniform level of investigation at the sites that make up the portfolio.
8.1 Desktop Portfolio SRAs
Typical portfolio seismic risk studies for insurance placements are performed as “desktop” analyses, based on approximate values for the buildings and contents, and business interruption cost rates (dollars per unit time, e.g., for lost rents). Values may be estimated by the insurance broker, often in discussion with the owner. These studies typically use owner-provided information for the location (i.e., address), building height or number of stories, year built and occupancy (i.e., commercial, residential, warehouse, etc.). Site geological hazards are assigned by the catastrophe modeling software from digital mapping using geo-coded locations, and risks are evaluated for an exhaustive set of possible future earthquakes. The vulnerability models are approximate, based on year built, height and occupancy, but with minimal information about materials and structural system, generally with no input from structural engineering professionals (Civil or Structural Engineers), although single-site seismic risk assessment reports (“PML reports”) may be consulted.
The estimates of damage and loss that are produced for any individual site are highly uncertain. Owing to the Central Limit Theorem and the “law of large numbers," the accuracy of portfolio-wide losses tend to improve, depending upon the geographic distribution of the exposure, so that portfolios with hundreds or thousands of structures widely distributed across seismic regions should produce more accurate estimates of aggregate loss, if no net bias exists in the values and vulnerability. The catastrophe models used in earthquake insurance applications typically account for the high uncertainties involved in ways that produce aggregate loss estimates with high uncertainty, reflecting the approximate input data. Such evaluations, typical of risk studies produced by insurance brokers for earthquake insurance placements, are referred to herein as “Desktop Portfolio Seismic Risk Assessments."
8.2 Engineered Portfolio SRAs
Under the direction of a Professional Engineer, improved portfolio seismic risk results may be achieved based on more accurate values, geologic information and vulnerability modeling. Improved, current values for the buildings, contents and time-element exposures, may be found by appraisal specialists, or by using valuation software. A Professional Engineer (Civil or Structural) can assign damage models based on the structural systems found from engineering site visits and/or review of Structural design drawings, with adjustment of vulnerability parameters or assignment of “secondary modifiers." The engineer can also consult geotechnical investigation reports or published maps to improve the assignment of Site Class and local site hazards (i.e., liquefaction, landslide), and to assign foundation type (e.g., piles) for sites with potential soil failures. This engineering information may be recovered from acquisition due-diligence studies (PML reports), produced through new engineering studies, or a combination of sources.
The accuracy of the estimates of damage that are produced for any individual site will vary, depending on the quality of the information and level of investigation for that site. The accuracy of portfolio-wide losses also depends upon the geographic distribution of exposure and vulnerability. A single high-value, vulnerable property may dominate loss and uncertainty for a small portfolio, whereas portfolios with many structures (i.e., hundreds or more) in multiple seismic regions will be less sensitive to losses at any single property.
For large portfolios, uncertainties in estimates of aggregate loss are proportionally lower, especially if higher levels of investigation are completed for the concentrations of exposure that can produce high levels of portfolio-wide risk. Ranking sites using single-site risk parameters (e.g., AAL, or SEL) in dollars may provide a way to prioritize efforts to improve the input data and scope engineering investigations.
Catastrophe models used in insurance applications generally account for the increased precision of structural vulnerability models (rather than occupancy-based models) by assigning reduced uncertainties. Such evaluations, typical of risk estimates produced by engineering firms for earthquake insurance placements, are referred to herein as “Engineered Portfolio Seismic Risk Assessments."
8.3 Describing the Quality of Investigation for Portfolio Seismic Risk Assessment
The quality of the investigation underlying a portfolio seismic risk investigation shall be reported to Users along with the other findings of the study. With reference to 8.1 and 8.2, if the quality of data and the level of investigation are uniform throughout the portfolio, the overall quality of a portfolio evaluation may be described directly as "Desktop Portfolio Seismic Risk Assessment,” or “Engineered Portfolio Seismic Risk Assessment” as appropriate.
More generally, the levels of investigation (as described in E2026) will vary for each site and for each building. Furthermore, the qualifications and experience of the investigators will vary, and the capabilities of seismic risk models may differ. The Providers and Users may not have access to a catastrophe model before establishing the scope of study and identifying the sites to be examined in greater detail, and conventional models are not set up to produce an estimate of the precision of risk estimates as a function of input data quality, so an approximate way to describe the quality of investigation is needed.
The simplest way to describe the quality of investigation is to tabulate the quality of investigation at each site, using terminology derived from ASTM E2026. The information for each site may include:
- Building name or other identifier
- Building address
- Location information quality: was the address successfully geocoded? (Y/N)
- Exposure values (replacement value, contents replacement value, business interruption loss rate, etc.)
- Source and date for exposure values (owner, broker, etc.)
- Occupancy / Usage (e.g., residential, industrial, office, etc.)
- Source of site geologic information (Site Class, liquefaction susceptibility): from GIS data or geotechnical report
- Building damageability system: Structural (e.g., ATC, NEHRP, ASCE 7) or Occupancy-based, and
- Basis for damageability class (description by others, photos, visual survey by P.E., design drawing review by P.E., etc.)
- Basis for secondary modifiers (if included)
Where investigation quality varies, the quality of investigation for the overall portfolio (or for any independent seismic region) can be described as the fraction of the total exposure value meeting that quality, such as “50% of the southern California exposure was investigated to Level BD1 or better.” Such communication regarding the quality of the input data will allow the User to assess the adequacy of Provider efforts and the expected quality of the risk results.