Geotechnical News • September 2018
35
GEOHAZARDS
the rarer they will be. For example, a
Class V hurricane will occur at a much
lower frequency than a Class III hur-
ricane, a Richter scale 8 earthquake at
a much lower frequency than a Richter
scale 6 earthquake, and so on. The
same principle applies to landslides.
A document titled: Subdivision
Preliminary Layout Review – Natural
Hazard Risk (MoTI, 2015) stipulates
that up to the 1:10,000-year event
should be integrated in landslide
hazard or risk assessments in B.C..
The 1:10,000-year event was also
considered in an expert panel review
for the Cheekeye River development
(Cheekye Review Panel #2, 2015)
and in a paper by Cave (1992/1993).
This article examines the associated
challenges.
Most of British Columbia was covered
by glaciers during the last ice age that
eroded or obliterated the evidence of
most pre-glacial landslides. Conse-
quently, most recognizable landslide
source areas and deposits are less than
about 10,000 to 11,000 years old. For
recurrent landslide processes such
as debris flows or rockfalls, various
techniques exist to characterize their
frequency and magnitude. The estima-
tion precision will be a function of
many variables, such as the physical
evidence available to decipher past
events, the preservation and strati-
graphic complexity of the deposits,
and the practicality of accessing the
data archive. These issues govern the
types of methods available and the
cost of applying some or all of such
methods and to what detail. With
significant investment, it is sometimes
possible to estimate a statistically-
based frequency-magnitude relation
for events that have occurred for some
period since deglaciation. In most
instances, however, regional landslide
inventories, slope stability analyses,
assessments of current and anticipated
site conditions, statistical methods
and/or other inputs are combined with
professional judgement to estimate
landslide frequency and attendant
magnitude, mobility, and intensity.
Various documents exist to guide the
hazard threshold that should be con-
sidered in geohazard safety analysis
(MoTI, 2015; EGBC, 2010, 2012). In
BC, guidance ranges from the 1:200-
year event for floods, the 1:300-year
event for snow avalanches and up to
a 1:10,000-year event for landslides.
The reason for this sliding scale may
be attributed to differences in the per-
ceived rate of change in the destruc-
tiveness and lethal potential of a given
geohazard with changes in probability.
With respect to landslides, according
to provincial guidance (MoTI, 2015),
a life-threating event ought to con-
sider up to the 1:10,000-year event.
This threshold was first referenced in
work by Dr. Peter Cave (1993) and
has been followed by at least one
regional district in BC. The 1:10,000-
year threshold is now also stipulated
in a Ministry of Transportation and
Infrastructure (MoTI) brief (2015) for
its subdivision approval officers. To
determine whether such an event has
occurred, or to estimate the charac-
teristics of a future event with this
probability of occurrence, a gamut of
absolute dating methods and vari-
ous approaches to reconstitute and/or
extrapolate event magnitude must be
employed. However, the practitioner
is invariably confronted with trying to
estimate the magnitude (volume) and
intensity (impact force) of an event for
which there may not be any historical
precedent, or it may not be practical to
recover evidence of such an event in
the field.
One of the fundamental issues with
the 1:10,000-year event lies in the
accuracy of its estimate. The accuracy
and precision of estimating the mag-
nitude of a landslide is proportional to
its return period: The longer, the more
uncertain, to the point where the error
bars (judgement or statistically-based)
are too large to be credible.
Another statistical issue emerges from
the fact that landslide-generating
mechanisms are not self-similar over
a wide range of frequency-magnitude.
The processes generating a 1:100-year
debris flow, may be very different
from those generating a 1:1000-year
or 1:10,000-year debris flow, hence
each perceived process type deserves
its own frequency-magnitude relation-
ship.
Assuming the data from past landslide
events exists or can be reconstructed,
one school of thought promotes
only relying on data to assign event
frequencies, and dismisses statistical
wizardry to extrapolate, interpolate or
impute data. This is reasonable only
(a) for cases that are characterized by
long and continuous records, (b) when
there is a thorough understanding of
the geomorphic processes and engi-
neering geology and (c) when it can be
reasonably assumed that the processes
and process rates that generated the
record have been constant and will
prevail in the future. Unfortunately,
these prerequisites are hardly ever met
in BC or elsewhere.
Statistical analysis and extrapola-
tion of known age and size pairs over
a limited period can yield variable
outcomes depending on the chosen
distribution and the knowledge of the
practitioner of the engineering geol-
ogy and geomorphology of a slope
or basin, which may limit the maxi-
mum credible event volume. This is
especially the case when extrapolating
to the 1:10,000-year event using only
a few hundred years of record recon-
structed.
Problems with geomorphic reconstruc-
tion invariably arise. For example,
most valley bottom alluvial fan
settings in settled parts of the prov-
ince have been logged, limiting the
use of dendrochronological methods
for frequency analysis. Moreover,
hundreds of developed fans are along
marine or lake shorelines where much
of the fan is below water level, which
precludes test trenching and sampling
organic materials for radiocarbon dat-
ing. Methods are available to estimate
sediment yield from the watershed and
channels, but it is hugely challenging
