Transcript Topics covered

Flooding & Landsliding—How do we
define and cope with Risk?
By
Bob Gerber, P.E. & Certified Geologist
Ransom Consulting, Inc.
[email protected]
207-838-1418
1
Real v. Perceived Risk
• Variables that may appear to create differences between perceived risk
and real risk:
–
–
–
–
–
–
Number of lives, dollars or environmental damage at stake
Amount of control one has over the risk factors
Whether the risks are static or change with time
Ability to avoid risk or insure against it
How rare the event may be
Whether the individual factors contributing to the risk are independent of
each other or are linked or coupled
– Whether one is considering the chance of a single occurrence in a year or the
chance of at least one occurrence over 100 years
– Extent of knowledge or ability to estimate the risk
• Most risks can be quantified (e.g., as annual probabilities of occurrence)
and compared with other risks, but most people are still uncomfortable
with risks defined as probabilities. Expert witnesses are often asked to
opine whether an event is more likely than not to occur under a certain
set of circumstances (i.e., is the probability of occurrence >50%?)
2
Deterministic v. Probabilistic Modeling
• Predictive groundwater modeling and slope stability modeling for many
regulatory programs is primarily done from a deterministic point of view.
No error bars or confidence intervals are defined. A slope is usually
considered “safe” if the Factor of Safety (FS) against failure is >1.5. The
physical properties of the slope are treated as if they are fully known with
certainty everywhere in the slope.
• RCRA closure, from the human health risk assessment perspective, is
almost all done using statistics and probabilistic approaches where cancer
risks are being evaluated.
• Groundwater quality regulatory compliance at a landfill is determined by
comparing upgradient vs. downgradient groundwater quality using
statistical and probabilistic approaches.
• The probability of an earthquake with a magnitude of a certain recurrence
interval is usually done using a complex probabilistic approach although
some geologic interpretation is embedded in the method.
3
Evaluating risks with large consequences
associated with complex technologies
• Interesting reference: Charles Perrow, Normal Accidents,
Living with High-risk Technologies (1984, reprinted 1999
with some new material)
• The complexity issue & coupling
– Accident probability goes up with complexity of the system
– Linear systems have lower likelihood of major accidents than
systems that have tight coupling leading to unforeseen
interactions between separate systems
– Normal Accident Theory (NAT) implies that accidents are normal
or inevitable in highly coupled, complex systems
• The size of the potential harm, the number of people
potentially affected, and the nature and length of the harm
affects the degree of risk that is tolerated by society
4
Perrow
(1999),
Afterword, p.
385
Perrow
unwittingly
predicts our
2008 financial
meltdown when
he looks at
financial
practices
through the lens
of accident
theory
5
Flood Risk
• Coastal Flooding
– Flooding caused by storm surge and/or large waves
generated and driven primarily by high wind velocities.
Because of the spatial variability of where an extreme
ocean storm might hit, it is difficult to estimate recurrence
intervals at any specific point, particularly with changing
storm frequencies and intensities.
• Riverine Flooding
– Flooding caused by large precipitation and/or snowmelt
events that may be exacerbated by debris or ice jams. The
statistical approach to estimating flood elevations of
specific recurrence intervals is well established using
gaging stations with long (>20 years) continuous records.
This assumes the risk is relatively constant through time.
6
Components of Risk in Ocean Flooding
• Wind (Frequency of extreme winds seem to be increasing in recent times)
• The component of water level related to storm surge, which is related to
wind, and the component related to tide elevation (i.e., neap or spring
tide and height of tide at the height of storm)
• Wave Heights—either propagated from large offshore waves or locally
wind-generated waves
• Wave Setup (momentum transfer, raises water levels 1’ to 5’ at the
immediate shoreline)
• Wave Runup (deep water and steep slopes at shore create the largest
wave runup)
• Sea Level Rise (backward looking only and rate of rise has been slow
compared with effects of recent increase in storm intensity and frequency)
• Loss of land due to wave erosion (assumed with sand dunes, soft soil
slopes, and man-made structures as part of runup analysis)
7
The state-of-the-art method of estimating ocean surge
from extreme events
• Use very advanced coupled ocean surge and wave models that
simulate the tracks of all historical storms (~100 years of storm
data); only done to date with hurricanes; need to add northeasters
• Develop statistics on water elevation at the shore for each point of
interest on the shore based on the model simulations of each storm
• Can derive the annual probability of occurrence of any given water
elevation at each point from the statistics
• May need to make adjustments for recent increase in severity and
frequency of ocean storms
• This method differs from that currently used by FEMA in southern
Maine where such variables as peak tide stage and peak storm
surge are assumed to occur simultaneously. The joint probability of
occurrence of two independent probabilities of 1% each is much
less than 1% (actually the product of 1% x 1% where events are not
linked in any way).
8
Variable Components of Risk in River
Flooding
• Precipitation (both precipitation and runoff
have been documented to be increasing since
1970)
• Land Use changes (e.g., adding impervious
area) increases runoff
• Joint probability of a melting snowpack in
combination with heavy rain
• Unpredictable obstructions in the channel
(e.g., chance of ice jams or debris jams)
9
Yarmouth Historical Society Risk
Evaluation
• A building was offered for free for a new historical
society repository on edge of Royal River 100-yr
floodplain
• How do you help the client decide whether the risk is
acceptable? Besides the hydraulic modeling involved,
it required trying to describe the difference between
single event and cumulative probability theory
• When I was with Sebago Technics we did the best we
could to account for apparent increases in flood flow
with time as a function of recurrence interval flooding
events
10
11
12
1961
TP 40
Comparison
of 100-year
rainfall in 24
hours based
on data up to
1961, vs
current
estimate of
that value
from the
Northeast
Regional
Climate
Center at
Cornell; 25%
increase
13
100-yr
flood
64%
500-yr
flood
14
15
April 1996 Rockland Harbor Landslide
Landslide risk is usually stated in terms
of factor of safety against sliding
• Factor of Safety = Ratio of forces resisting failure to forces causing
failure
• Variables include:
– Topography (shore slopes can be over-steepened by wave erosion)
– Geology (depth to hard substrate, type of soil and nature of
layering)
– Groundwater pore pressure distribution
– Soil strength spatial variability
• Can calculate annual probability of failure if you can gather enough soil
strength data so that geostatistics can be applied but this usually
would cost too much for a typical landowner
• One approach is to monitor slope movement and manage the risk in
real time by adjusting a variable. This was done in the Ft. Halifax dam
removal on the Sebasticook River in Winslow where reservoir level was
the variable that was adjusted because it translated to groundwater
pore pressure distribution
17
Horizontal Slope Movements below
Dallaire Street Monitored in real time
July 2008, just downstream of Dallaire St. houses
Summary
• To understand risk objectively, try to put it in probabilistic
terms that can be compared against other risks you are more
familiar with that can be characterized probabilistically
• You can make choices that can reduce risk (move away from
risk, build protective structures, buy insurance, reduce
complexities, etc.)
• Short-term risks of high consequences can often be managed
in real time through monitoring and adjustment of variables
• Learn the difference between annual occurrence probability
and cumulative probability over multiple years
• In man-made systems, highly coupled systems are more likely
to fail than unlinked linear systems
20