Defining the Characteristics of a PML Study
March 2007
Probable maximum loss (PML) studies have long
been a staple in the insurance, real estate, and financial industries for quantifying
the risk due to extreme hazards and the resulting financial losses.
by Nathan
C. Gould, D.Sc., P.E., S.E.
ABS Consulting
The use of PML studies that provide probabilistic methods of estimating natural
hazards and the magnitude of the resulting estimated losses have combined to
alter the nature of the PML that is used today.
Deterministic versus Probabilistic
PML studies that have been a staple in the insurance and financial industries
to help quantify the risk due to earthquake and extreme wind loads have often
focused on using a deterministic event, as opposed to a probabilistic event,
to quantify the natural hazard risk. The fundamental difference between a deterministic
and probabilistic analysis is that a deterministic analysis does not consider
the probability associated with a hazard, whereas a probabilistic analysis incorporates
the hazard probability.
In a deterministic analysis, a controlling fault (earthquake) or hurricane
(extreme wind) for a site is specified. An event with specified parameters (such
as magnitude for earthquakes or wind speed for hurricanes) for the desired return
period is assumed to have occurred. This
approach can be expected to generate a conservative "worst-case" scenario for
loss, especially when combined with a 90 percent confidence level on the loss
estimate. Deterministic events may include either scenario or historic events.
From an extreme wind perspective, the historical analysis will typically focus
on a relatively recent period of time (i.e., the past 100 years).
In contrast, a probabilistic evaluation considers a statistical combination
of all the events that have the potential to occur, but have not necessarily
been observed at the site. As an example, a probabilistic hurricane evaluation
should include a robust statistical distribution of the main storm parameters
that can be developed from historical data. This distribution can then be used
to simulate climatologically plausible hurricane tracks from genesis to decay.
A probabilistic analysis should result in a more complete and realistic evaluation
of the potential loss.
Selection of Return Periods
The event of choice for the standard earthquake PML has historically been
an event with a 475-year return period, which can also be expressed as an event
that has a 10 percent chance of exceedance in 50 years. There are many thoughts
as to the origin of the classic 475-year return period event; however, this
return period has long been used to define design basis earthquake in several
of the primary building codes in the United States that preceeded the new International
Building Code (IBC).
The 2006 IBC (through reference to the ASCE 7-05 Standard) uses two-thirds
of the maximum considered earthquake (MCE) ground motion as the design earthquake.
The MCE ground motions are typically defined as the maximum level of earthquake
ground shaking that is considered reasonable for typical structures to resist.
For many parts of the United States, the MCE will correspond to an event with
an approximate 2,500-year return period (2 percent chance of occurrence in 50
years). The selection of a 2,500-year return period has resulted in significant
impacts in building code requirements in many regions of the United States and—if
also projected into PML studies—can be expected to have a similar effect on
the resulting projected losses.
The selection of a wind event for hurricane loss evaluations has varied greatly
over the past years. One traditional approach was to use deterministic events
that had wind speeds corresponding to the 50-year and 100-year mean recurrence
interval (MRI) event, which reflected older wind design standards for new buildings.
These standards typically used the 50-year MRI as the basis for design wind
speed for ordinary buildings and the 100-year MRI as the basis for the design
wind speed of buildings with a higher level of importance. Other approaches
have ranged from the simplistic, such as using an event with a return period
equal to that used for the classic seismic PML, to high level site-specific
analyses to determine controlling wind events over a full range of return periods.
Recently there has been a push to use return periods in conjunction with
probabilistic loss analysis that had previously been considered to be at the
far upper end of the wind loss curve. Probabilistic loss analysis for return
periods of 1,000, 5,000, and 10,000 year, which correspond to annual probabilities
of exceedance of 0.1, .02, and .01 percent, respectively, may be required where
financial instruments or other considerations require the development of the
upper spectrum of the loss curve. It is important to understand how the loss
curve "flattens out" at the higher return periods. An example of this phenomenon
can be observed in
Figure
1 that shows a normalized loss curve (normalized to the 475-year
loss) over a range of return periods for a probabilistic hurricane evaluation
of a site in the Caribbean.
Is a Site Visit Required?
An often-asked question is whether a visit to the facility under review is
required. There are PML studies that are performed without a site visit. These
are typically referred to as "desktop" studies, and they involve using only
the information that can be sent to the reviewers. The material provided typically
includes design drawings, a property condition assessment report, and financial
information related to the facilities insured value. However, this information
alone may not be adequate to provide a complete picture of the important characteristics
of the facility relative to the risk assessment. Issues such as significant
architectural and/or structural upgrades or renovations, architectural/structural
deterioration, and site infrastructure often need to be observed and reviewed
at the site.
In addition, for seismic and wind hazard evaluation, there are critical hazard-specific
issues that should be addressed during the site visit. For a seismic evaluation,
these issues include site-related conditions such as slope stability and ground
surface faulting. For an extreme wind evaluation, areas of potential water inundation,
and sources of wind-borne debris are items that are often reviewed during the
site visit.
Summary
Performing a seismic or extreme wind risk assessment for a facility requires
knowledge not only of the hazard, buildings, and infrastructure at the facility
under consideration, but also requires an understanding of how the risk assessment
and corresponding loss data will be utilized. Issues relating to probabilistic
and deterministic loss, return periods to reviewed, site visits, and other differentiating
factors should be discussed and understood by all parties involved prior to
commencing an assessment.
Opinions expressed in Expert Commentary articles are those of the author and are
not necessarily held by the author's employer or IRMI. Expert Commentary articles
and other IRMI Online content do not purport to provide legal, accounting, or other
professional advice or opinion. If such advice is needed, consult with your attorney,
accountant, or other qualified adviser.