Content uploaded by Royce A Francis
Author content
All content in this area was uploaded by Royce A Francis on Jan 16, 2014
Content may be subject to copyright.
Climatic Change
DOI 10.1007/s10584-010-0001-9
Probabilistic life cycle analysis model for evaluating
electric power infrastructure risk mitigation investments
Royce A. Francis ·Stefanie M. Falconi ·
Roshanak Nateghi ·Seth D. Guikema
Received: 9 March 2010 / Accepted: 8 November 2010
© Springer Science+Business Media B.V. 2010
Abstract One effect of climate change may be increased hurricane frequency or
intensity due to changes in atmospheric and geoclimatic factors. It has been hypoth-
esized that wetland restoration and infrastructure hardening measures may improve
infrastructure resilience to increased hurricane frequency and intensity. This paper
describes a parametric decision model used to assess the tradeoffs between wetland
restoration and infrastructure hardening for electric power networks. We employ
a hybrid economic input–output life-cycle analysis (EIO-LCA) model to capture:
construction costs and life-cycle emissions for transitioning from the current electric
power network configuration to a hardened network configuration; construction
costs and life-cycle emissions associated with wetland restoration; and the intrinsic
value of wetland restoration. Uncertainty is accounted for probabilistically through a
Monte Carlo hurricane simulation model and parametric sensitivity analysis for the
number of hurricanes expected to impact the project area during the project cycle
R. A. Francis (
B
)
Department of Engineering Management and Systems Engineering, The George Washington
University, 1776 G St., NW #159, Washington, DC 20059, USA
e-mail: seed@gwu.edu
S. M. Falconi · R. Nateghi · S. D. Guikema
Department of Geography and Environmental Engineering, Johns Hopkins University,
313 Ames Hall, 3400 N. Charles St., Baltimore, MD 21218, USA
S. M. Falconi
e-mail: stefanie.falconi@jhu.edu
R. Nateghi
e-mail: rnategh1@jhu.edu
S. D. Guikema
Department of Civil Engineering, Johns Hopkins University,
313 Ames Hall, 3400 N. Charles St., Baltimore, MD 21218, USA
e-mail: sguikema@jhu.edu
Climatic Change
and the rate of wetland storm surge attenuation. Our analysis robustly indicates that
wetland restoration and undergrounding of electric power network infrastructure is
not preferred to the “do-nothing” option of keeping all power lines overhead without
wetland protection. However, we suggest a few items for future investigation. For
example, our results suggest that, for the small case study developed, synergistic
benefits of simultaneously hardening infrastructure and restoring wetlands may be
limited, although research using a larger test bed while integrating additional costs
may find an enhanced value of wetland restoration for disaster loss mitigation.
1 Introduction
Recently, climate change has been associated with potential increases in hurricane
frequency or intensity (Emanuel 2005). Increases in population density in coastal
areas have also been forecasted (Nicholls and Small 2002; Small and Nicholls 2003),
thus increasing the urgency of mitigating potential adverse outcomes associated with
hurricane events. These observations have led risk managers to consider the effects
of increased hurricane frequency and intensity on networks of lifeline infrastruc-
ture that support our societies (U.S. Congress Office of Technology Assessment
1990). Examples of lifeline infrastructure networks include electric power networks,
drinking water and wastewater networks, transportation networks, and oil and gas
pipeline networks. Hurricane Katrina has further focused attention in the USA on
the potential adverse impacts of increased hurricane frequency and intensity (Day
et al. 2007), and attendant risk mitigation decisions are currently being re-evaluated
(Bigger et al. 2009; U.S. Army Corps of Engineers 2004). It has been hypothesized
that coastal wetland restoration may be a cost-effective approach to mitigating the
impacts of future hurricanes (Wamsley et al. 2009, 2010).
The objective of the analysis presented in this paper is to demonstrate a method-
ology for evaluating potential risk mitigation synergies between coastal wetland
restoration and infrastructure protection, focusing on electric power network hard-
ening. We evaluate the life cycle costs associated with electric power network
hardening and wetland restoration for a model city case study for three scenarios:
undergrounding all electric power equipment; undergrounding only electric power
equipment in the commercial zone; and making no changes to the existing network
configuration. For each of these scenarios, we evaluate life cycle costs with and
without wetlands, and including and not including indirect economic and environ-
mental costs. Our results suggest that for our case study city, wetland restoration
and infrastructure hardening are not the preferred options. This result has important
implications for coastal adaption planning. Natural ecosystem restoration may be
difficult to justify as an approach for coastal adaptation without separately consider-
ing the valuation of biodiversity, ecological services, and carbon sequestration. While
these issues pose difficult valuation problems, the more directly quantifiable benefits
of wetland restoration may not offset the significant cost of restoring or building
wetlands in many locations.
2 EIO-LCA disaster mitigation framework
Risk mitigation decisions are currently evaluated using several tools, including but
not limited to: benefit–cost analysis (BCA) (Arrow et al. 1996), life-cycle cost
Climatic Change
analysis (LCA) (Chang and Shinozuka 1996), and probabilistic decision analysis
(Keeney 1982). BCA involves maximizing the ratio of benefits expected from a
decision to the expected costs incurred by taking a decision. The assumption is that
maximizing this ratio maximizes public welfare. LCA extends BCA by accounting
for the costs incurred over the life cycle of a project undertaken as a result of a
specific decision. Probabilistic decision analysis involves parameterizing the decision
choices and probability of potential adverse outcomes to be mitigated such that
the uncertainty associated with the occurrence of an adverse event is quantitatively
evaluated in the decision framework and the decisions are based on maximizing
expected utility. Risk mitigation decisions can be evaluated using a combination
of each of these tools. One recent disaster risk mitigation methodology proposed
combining each of these tools is the extended life cycle cost analysis framework
(ELCA) (Chang 2003).
ELCA extends traditional approaches to BCA by incorporating societal costs
and benefits over a risk mitigation project’s planned life with the benefits and costs
expected to accrue to the relevant lifeline agency over a project’s planned life. The
computation of societal and lifeline agency costs and benefits is facilitated by a
transparent framework accounting for four types of costs and benefits (Chang 2003):
planned costs undertaken by the lifeline agency; costs imposed on society by the
lifeline agency’s actions; expected unplanned costs undertaken by the lifeline agency;
and, expected unplanned costs imposed on society through lifeline service disruption
and restoration.
In the present paper, we extend the ELCA by incorporating supply-chain envi-
ronmental impacts and other societal impacts associated with disaster risk mitigation
projects. To incorporate these environmental and societal impacts, the economic
input–output life-cycle assessment (EIO-LCA) framework (Hendrickson et al. 2006)
is employed. We describe this new framework below in the section titled “EIO-
LCA Disaster Mitigation Framework”. We then apply the extended ELCA to a
hypothetical case study with three decision scenarios evaluating wetland restoration
and electric power network hardening to mitigate electric power infrastructure risk
in the event of increased hurricane frequency and intensity.
2.1 Extended ELCA framework
Chang and Shinozuka (1996)andChang(2003) have extended the practice of
project life-cycle cost analysis to disaster loss estimation methodology. We adapt this
framework for our purposes and present the details of its characterization here. As
described above, this framework consists of four parts:
C = C
1
+C
2
+C
3
+C
4
(1)
where: C
1
= planned costs undertaken by the lifeline agency; C
2
= costs imposed on
society by the lifeline agency’s actions; C
3
=expected unplanned costs undertaken by
the lifeline agency; and, C
4
= expected unplanned costs imposed on society through
lifeline service disruption and restoration.
The planned costs undertaken by the lifeline agency, C
1
, are the sum of direct costs
to the utility of performing routine maintenance on the infrastructure network and
the costs to the utility of performing mitigation investments (with their associated
maintenance costs). The maintenance and mitigation investments required depend
on the nature of the network and the anticipated natural disaster. More specifically,
Climatic Change
the maintenance costs, C
m
, are calculated as the sum of the maintenance cost m
for system element i, multiplied by a discount factor for year t, z(t), for all system
elements over the planned project life.
C
m
=
t
i
m
i
(
x
i
, t
)
· z
(
t
)
(2)
The mitigation investment costs, C
mit
, are calculated as the sum of the mitigation
investment cost e for system element i, multiplied by the discount factor for year t for
all system elements over the planned project life.
C
mit
=
t
i
e
i
(
x
i
, t
)
· z
(
t
)
(3)
Both of these costs depend on properties of the system elements including their age,
materials, and tasks required.
The costs imposed on society as a result of the lifeline utility’s mitigation decisions,
C
2
, are the sum of the benefits associated with disaster mitigation activities (e.g.,
increased employment and improved operational efficiency) and the environmental
impacts associated with the activities required by the mitigation decisions. Chang
and Shinozuka (1996)andChang(2003) included only the economic portion of these
costs and benefits. Here we extend this to incorporate the life-cycle environmental
costs of agency decisions.
To compute costs and benefits, we employ the EIO-LCA approach. In short, EIO-
LCA adapts the Leontief input–output economic model to estimate the environmen-
tal emissions associated with a product or service over its lifetime (Hendrickson et al.
2006). Because the overall economic activity associated with a product or service over
its lifetime is more inclusive than the revenue associated with its purchase, we can
take advantage of this feature of the EIO-LCA model to compute the environmental
costs and economic benefits attributable to the lifeline agency’s mitigation decisions.
We propose that the benefits, beyond risk mitigation, of the lifeline agency’s decision
are the additional economic output associated with maintenance and mitigation
activities above the direct costs to the lifeline agency, EO
m
and EO
mit
, respectively.
In the event that wetland restoration is chosen, additional economic benefits accrue
due to the economic value of the ecosystem services of the restored wetland area,
EV
wet
. The costs imposed on society include the environmental costs and service
disruptions associated with the maintenance and mitigation investments. In this
paper, we ignore the costs of the service disruptions due to maintenance and
mitigation investments, including only the environmental costs, R
m
and R
mit
,ofthe
maintenance and mitigation investments, respectively.
C
2
=
t
(
EO
m
+ EO
mit
+ EV
wet
)
−
(
R
m
+ R
mit
)
t
· z
(
t
)
(4)
For the mathematical details of the EIO-LCA model, the reader is referred to
Hendrickson et al. (2006).
The expected lifeline utility costs in the event of a natural disaster, C
3
,arethe
sum of the expected repair costs associated with system element failures caused by
the natural disaster, C
r
, and the revenue loss attributable to the attendant service
disruptions, C
v
.
C
3
= C
r
+C
v
(5)
Climatic Change
As with C
m
and C
e
, the repair costs of a system element failure depend on the
element properties, age of the system element, and tasks required to perform the
repair. Chang (2003) characterizes the expected repair costs of system element i
over the project life as the sum of the product of the expected failure probability in
event of a natural disaster, F
i
, and its unit repair cost r
i
for each system element over
the life of the project. The expected failure probability due to natural disasters over
the planned life of the project is computed by integrating the product of the system
element’s fragility curve, P
F
, and the hazard curve for natural disasters, p(h), over
the range of possible disaster intensities, h. The hazard curve gives the probability
of a hazard of intensity h occurring, and the fragility curve gives the probability of
element failure as a function of h, system element age, and system element type.
C
r
=
t
i
[
F
i
(
x
i
, t
)
·r
i
]
· z
(
t
)
F
i
(
x
i
, t
)
=
h
P
F
(
x
i
, t, h
)
· p
(
h
)
dh (6)
The revenue loss attributable to attendant service disruptions, C
v
, is the sum of the
product of the expected annual unmet demand, V
t
, and unit price, p, of the service
provided over life of the project. The unmet demand is a function of the time to
service restoration, τ , the percent initial unmet demand, ω, the normal demand
volume, D, the hazard curve for natural disaster intensity, and decision-making
factors related to system repairs, w.
C
v
=
t
V
t
· p · z
(
t
)
V
t
=
h
τ
(
w, h
)
·ω
(
h
)
· D · p
(
h
)
dh (7)
Finally, the expected costs imposed on society in the event of a natural disaster, C
4
,
are the sum of the losses in economic activity that occur from utility outage and
cascading business losses directly caused by utility outages. As with the costs imposed
on society due to routine maintenance and initial mitigation investments, C
4
may be
divided into the economic output and environmental costs. We include the economic
output, EO
r
, caused by the repairs C
r
, the economic output lost, EO
l
, due to direct
business losses, C
B
, and the lost economic output, EO
ul
, associated with decreased
utility revenues, C
v
. We also account for the environmental costs associated with
repairs, R
r
, and the reduction in environmental costs due to direct business losses,
R
l
, and service interruptions, R
ul
.
C
4
=
t
{
EO
r
(
C
r
)
− EO
l
(
C
B
)
− EO
ul
(
C
v
)
}
−
{
R
r
− R
l
− R
ul
}
t
· z
(
t
)
(8)
While we have defined the repair costs and the lost lifeline utility revenues above, the
direct economic losses of business interruption are a function of the time to service
restoration, τ , the initial economic loss attributable to the natural disaster, ε,the
post-event percent unmet demand ω in area a due to an event of intensity h,the
Climatic Change
normal economic activity Q in area a, the probability of a disaster with intensity h,
and business resiliency to lifeline outage ρ.
C
B
=
t
E
t
· z
(
t
)
E
t
=
h
τ
(
w, h
)
·ε
ω
a
(
h
)
, Q
a
,ρ
· p
(
h
)
dh (9)
As a simplifying assumption, we use empirical estimates of direct business losses due
to power outages reported by LaCommare and Eto (2006), as discussed below, in
lieu of formal studies of economic resilience to power outages.
3 Case study illustration
3.1 Case study city
In order to illustrate the method, we utilize as a case study example a synthetic
small city assumed to be located in a hurricane-prone coastal area. This synthetic
city is then subjected to simulated hurricanes. Infrastructure damage is, in turn,
simulated probabilistically based on fragility curves and assumptions presented in
the literature. We examine two causes of failure: wind-induced damage of utility
poles and surge-induced damage of buried electrical lines. The costs associated
Fig. 1 Micropolis electric power network configuration
Climatic Change
with damage, repair, and outage of various electric power network components are
presented in subsequent sections. In addition, we discuss the environmental impact
and indirect economic cost assumptions for each infrastructure component. We focus
on mitigating risk to the electric power system of this city with wetland restoration
and infrastructure hardening being the two options available. We define the project
area as a small North Carolina coastal city straddling Category 3 and Category 5
hurricane storm surge zones, with the city extending approximately 1 mile inland and
1
/
2
mile along the coast. The prototype for this city is a model city called “Micropolis,”
developed at Texas A&M University as one of two model cities to be used as testbeds
for infrastructure risk research and planning (Brumbelow et al. 2007). Micropolis has
approximately 5,000 residents in a historically rural region, and details are provided
for each building in the city as well as the power and water systems for the city.
The number of electric power customers (residential, industrial, and commercial)
is specified, as is the customer electricity demand and configuration of the electric
power network. Micropolis has 434 residential customers, 15 industrial customers,
and nine commercial or other customers in the project area served by approximately
9.7 circuit-miles of overhead electric power distribution line. Micropolis’ electric
power network configuration is shown in Fig. 1.
3.2 Hurricane simulation model
We developed a statistical model that best describes the relationship between climate
variability and North Atlantic tropical cyclone (TC) counts (Sabbatelli and Mann
2007) in the U.S. through the use of count regression analysis and data mining
techniques. In our modeling implementation we used 17 climate variables, with their
quarterly and annual averages for 73 total covariates and tropical cyclone counts in
the U.S. from 1948–2004.
We implemented three distinct approaches and compared their results to choose
the model with the most superior fit and predictive accuracy. The first two ap-
proaches involved constructing regression trees to identify the most important vari-
ables prior to fitting a Poisson Generalized Linear Model (P-GLM) on the reduced
data set consisting of the most important variables. The third approach consisted of
fitting a P-GLM without any previous data mining implementations.
Our fit and prediction results indicate that fitting regression trees prior to fitting a
P-GLM leads to better fit results and predictive accuracy. The statistics of our final
best model that was developed through this procedure is summarized in Table 1.
Our prediction errors, shown in the first column of Table 1, are calculated based
on 50 random hold-out cross validation tests. In each of 50 independent iterations,
10% of the data is randomly held out to create a validation set. The model is then
built on the remaining subset of the data, the training set, and the predictions are
tested against the validation set. The MSE (Mean Squared Error) and MAE (Mean
Table 1 Mean squared error (MSE) and mean absolute error (MAE) for the hurricane count model
for both the fitting data set and 50 repeated, random hold-out validation tests
Model prediction error Model fitting error Null model prediction error
MSE MAE MSE MAE MSE MAE
4.523 1.0523 5.747 1.873 10.377 2.660
The null model is an intercept-only model based on the historic mean
Climatic Change
Absolute Deviation) in the first columns represent the difference between the actual
TC counts and the cou nts predicted by the P-GLM model, averaged over the 50
repeated validation tests. The fit errors are the difference between the fitted values
from our model and the observed number of TC counts. The last columns shows the
errors calculated when model is replaced by the mean observed TC counts (10.4).
As can be seen in Table 1, the errors are much larger in this case. Overall the results
suggest that our model can be used to reconstruct the past history of TC counts as
well of making future predictions with reasonable predictive accuracy.
Equation 10 reports the model. Our model suggests that TC counts in the U.S.
are positively associated with annual global land and ocean temperature anomaly
(CRU) and negatively correlated with annual Sea Level Pressure (SLP) anomaly
and annual El-Nino Southern Oscillation (ENSO) anomaly, with the temperature-
related covariate (CRU) having a bigger impact on tropical cyclone counts than the
other two covariates.
ln
(
TC
count
)
= 409 +0.928
(
CRU
)
−0.401
(
SLP
)
−0.138
(
ENSO
)
(10)
In this analysis, we use a joint distribution based on historical records for CRU, SLP,
and ENSO to simulate the number of tropical cyclones impacting the project area
over the 50-year life cycle. We then downscale from these tropical cyclone counts to
the number of hurricanes that would impact our project area in three steps. First, we
multiply the number of predicted TC counts by the ratio of (1) hurricanes counts to
(2) the count of all tropical cyclones in the historical record. Second, we multiply
this number of hurricanes in the North Atlantic by the proportion of hurricanes
in North Carolina to North Atlantic hurricanes. Finally, we assume that hurricanes
making landfall within 100 nautical miles of the project area impose costs on the
infrastructure in Micropolis. This procedure is summarized in Eq. 11.
H
proj
= TC
counts
·
H
N. Atl.
TC
counts
·
H
N.C.
H
N. Atl.
·
100nm
261.6nm
(
N.C. coastline
)
(11)
3.3 Coastal wetland restoration
As discussed above, it has been hypothesized that wetland restoration and in-
frastructure hardening measures may improve infrastructure resilience to increased
hurricane frequency and intensity. Infrastructure hardening increases resilience by
decreasing the likelihood of failure for a given wind load or surge depth. Wetland
restoration acts to decrease surge depths. Here, we briefly present our assumptions
about wetland restoration techniques and the effects of wetlands on storm surge
attenuation for the simulation.
Wetland restoration projects generally fall into three classes: wetland creation
from dredged materials; manipulation of sediment flow; and, conversion of open wa-
ters to wetlands (Turner and Streever 2006). In these three classes, eight approaches
are discussed in detail by Turner and Streever (2006): crevasse splays, former
agricultural impoundment conversion, backfilling, managing spoil banks, bay bottom
terracing, dredged material wetlands, excavated wetlands, and thin-layer placement.
For more details on each technique, the reader is referred to Turner and Streever
(2006). The costs of each approach are highly variable, ranging from $0 to $44,000
per hectare; moreover, the implementation of each approach is highly dependent on
prior experience, landscape attributes, and wetland ecosystem resources, services,
Climatic Change
and inhabitants. For the purpose of our case study, we assume that wetlands may be
constructed in open ocean using dredged material wetlands, and assume the highest
value presented by Turner and Streever (2006) for the cost of wetland restoration as
the cost of wetland restoration in our model, $44,000.
Wetland ecosystems provide several economic benefits due to their intrinsic
natural processes and services. The valuation of these services is difficult, however,
and methodology for their inclusion in wetland restoration decision analyses is not
straightforward. Two difficult aspects of assessing decisions considering potential
synergies between wetland restoration and coastal infrastructure hardening are
estimating the economic value of an acre of wetlands and, separately, estimating
the amount of carbon sequestered by the wetlands. The carbon sequestration po-
tential of wetland ecosystems is very difficult to establish. While Bridgham et al.
(2006) estimate that North American wetlands are a small to moderate carbon
sink (49 Tg C/year), the uncertainty in this estimate is greater than 100%. We
assume that our restored wetlands will most resemble tidal marsh, swamp, or coastal
floodplains. Bridgham et al. (2006) and Chmura et al. (2003) estimate that the
rate of carbon sequestration in these wetland categories is 7.3 × 10
−4
MtCO
2
E/ha
(2.56 MtCO
2
E/year for our project area). In addition, methane emissions from North
American wetlands may offset the benefits of wetland concentration. Consequently,
Bridgham et al. (2006) suggest that, with the exception of estuarine wetlands, carbon
sequestration potential should not be considered in wetland restoration decisions.
Due to this uncertainty, we do not consider carbon sequestration potential of wetland
ecosystems in this analysis. We do use empirical estimates of the economic value of
tidal marsh, coastal floodplains, and other wetlands from Costanza et al. (1989)and
Costanza et al. (1997) as the economic value of the restored wetlands’ ecosystem
services in our analysis. These empirical estimates include the value of various
wetland ecosystem services and processes, as well as supported economic activities,
as intrinsic to the wetlands. From these estimates, we assume that the value of our
restored wetlands lies in the range between $6,000–$30,000 USD/ha.
Several investigators have recently studied the effect of wetlands on storm surge
attenuation rates (U.S. Army Corps of Engineers 1963; Loder et al. 2009;Resio
and Westerink 2008; Wamsley et al. 2009, 2010). However, the amount of storm
surge attenuation attributable to surge flow over wetlands remains uncertain. The
principal challenge to quantification of storm surge attenuation is quantification of
the increased drag on storm surge due to bottom friction of wetlands. This challenge
is further complicated by the spatial heterogeneity of storm surge profiles over
wetlands and the spatial variability of drag caused by wetland composition. For these
reasons, empirical rules of thumb have been employed to incorporate storm surge
attenuation into approaches to wetland restoration and valuation. These empirical
rules of thumb range from 1 m (height) surge attenuation per 4 km wetlands restored
to 1 m surge attenuation per 60 km wetlands restored. The U.S. Army Corps of
Engineers (1963) estimate, quoted by Resio and Westerink (2008), is 1 m surge
attenuation per 14.5 km wetland restored.
Although we use a simple linear relationship for storm surge attenuation over
wetlands, this rule of thumb is known to have several weaknesses as indicated by
recent studies. For example, Loder et al. (2009) find that marsh elevation and bottom
friction contribute to storm surge attenuation, while marsh continuity may amplify
storm surge. These findings are consistent with the general findings of Wamsley
et al. 2009: storm surge attenuation is nonlinearly dependent on landscape
Climatic Change
characteristics (e.g., bathymetry, wetland attributes, presence of structures, etc.) and
storm characteristics (e.g., storm speed, size, track, and intensity). The nature of
storm surge attenuation over wetlands is also discussed in a concurrent paper in the
present special issue of Climatic Change (Gedan et al. 2010). Gedan et al. (2010)
indicate that nonlinearities may emerge in wetlands’ abilities to attenuate hurricane
storm surge due to biological and physical characteristics of the wetlands, including
differences in the identity, phenology, and morphology of the species comprising
the wetland system. Moreover, Gedan et al. (2010) also indicate that variation in
storm characteristics and coastal geography may also overcome the attenuation
effects attributable to the wetland system. These studies indicate that wetland storm
surge attenuation is a complex function of vegetation, bathymetry, and topology.
Approximation of this function as a linear trend does not accurately reflect these
complex relationships and may, in turn, not accurately estimate risk reductions in
coastal areas. Nonetheless, we summarize results from these studies using general
ranges of attenuation rates, as this complexity is not the present focus of our paper.
Consequently, our model incorporates this uncertainty by employing a triangular
distribution on the storm surge attenuation rate attributable to wetlands, with
minimum and maximum values of 1 m:60 km [surge attenuated: wetlands restored]
and 1 m:4 km, respectively. The mode of this triangular distribution is the U.S. Army
Corps of Engineers empirical rule, 1 m:14.5 km. To incorporate the attenuation
rate into our simulation analysis, we employ a simplified rule-based approach to
determine the amount of storm surge height attenuation. First, we draw a rate from
this triangular distribution. Next, the Saffir-Simpson storm category is determined by
the simulated wind speed. The simulated non-attenuated storm surge amount is then
the midpoint of the range of storm surge heights expected for the Saffir-Simpson
category. Finally, we subtract the amount of storm surge reduction implied by the
simulated attenuation rate from the midpoint of the Saffir-Simpson category surge
to obtain the attenuated storm surge. It must be noted, however, that this simplified
approach used in our simulation may not be valid for some events, as the relative
level of surge reduction may diminish as the overall surge potential increases (Loder
et al. 2009).
For our case study, our wetland restoration option involves the restoration of
wetlands sufficient to attenuate storm surge height by 2 ft using dredged material
wetlands (Turner and Streever 2006) under the assumption of the U.S. Army Corps
empirical rule of thumb, ignoring uncertainty in the attenuation rate. This leads to
restoring to wetlands to a distance of 2.5 miles out from the coast. Under these
assumptions, the wetland area we have restored in our case study is 3500 ha, requiring
an initial investment of $140,000,000.
4 Electric power network hardening
The electric power system for our case study city is shown in Fig. 1, while a descriptive
overview is provided in Table 2. The as-is overhead network equipment is indicated
by numbered poles, while existing underground equipment are designated by nodes
with no numbers. A transmission line runs along the railroad in the middle of the
city. The railroad also separates the city into the distinct hurricane storm surge zones
listed above: Category 3 east of the railroad, Category 5 west of the railroad. In this
Climatic Change
Table 2 Project area
descriptive figures
Miles of circuit line 9.6 mi
Depth of micropolis area inland 1 mi
Project area 0.5 mi
2
Number of residential customers 434
Number of commercial, industrial, 24
other customers
Hurricane category (surge zone) 3 (East of railroad)
5 (West of railroad)
analysis, we consider three scenarios: undergrounding all electric power equipment
east of the railroad (Scenario 1); undergrounding only the electric power equipment
in the commercial zone east of the railroad (Scenario 2); and making no changes to
the existing network configuration (Scenario 3).
First, we discuss the fragility curve assumptions we employ in our model. The
fragility curves for underground and overhead power network components are
of critical importance for evaluating the impacts of hurricanes on electric power
infrastructure. While overhead electric power network infrastructure is primarily
impacted by the wind associated with hurricanes, underground infrastructure is
primarily impacted by storm surge. As storm surge inundates inland areas, pad-
mounted transformers, buried lines in unsealed conduits may be damaged, and
underground equipment may be uncovered as the surge recedes. In our model we
then used a connectivity-based approach for estimating the impacts of power system
component failures on power supply to individual buildings. We assume that if a
building is connected to the substation, it can receive power. A failure of a line,
transformer, or pole is assumed to break the path on which that component resides.
While this approach does not capture power load flow balance and short-term system
dynamics, it is a reasonable first approximate for a low-voltage, radial-topology
power distribution system such as the one used in our test case. For a high-voltage
transmission system, a full power load flow model would likely be needed.
To estimate the fragility curves for underground and overhead power network
components, we use data primarily from Han et al. (2009), and Brown (2009). Han
et al. (2009) uses data from a large investor-owned utility (IOU) in the Gulf Coast
region to estimate power outages from hurricanes and to estimate a fragility curve for
wooden poles. The fragility curve for a distribution pole, as a function of wind speed,
is found to be approximated by the cumulative distribution function (cdf) of a normal
distribution with mean parameter 154 and shape parameter 27. These assumptions
are presented in Table 3. Three cases for the project area are now evaluated, using
probability and cost estimates for underground equipment failure from Xu and
Brown (2008) and Brown (2009): (1) all existing equipment is overhead (no changes);
(2) all equipment in the storm surge category 3 zone is placed underground; and,
(3) only the electrical equipment in the commercial area in storm surge category 3
zone is placed underground. According to Xu and Brown (2008) and Brown (2009),
reasonable estimates of the annual maintenance costs per circuit mile for an overhead
system are $4,500, including tree-trimming, and the annual maintenance costs per
circuit-mile for underground equipment is $4,000. The initial investment costs for
undergrounding existing equipment is estimated as $1.333 million per circuit-mile
(Xu and Brown 2008;Brown2009). This includes the cost of undergrounding non-
electric equipment such as cable and telephone lines. No initial investment is in-
cluded in this analysis for overhead circuitry because the system currently exists as an
Climatic Change
Table 3 Cost and storm-condition reliability assumptions for undergrounding analysis
Initial investment for undergrounding $1,333,333/circuit-mile
existing overhead equipment
Repair cost for overhead system component failure $4,000/component
Repair cost for underground system component failure $60,000/component
Cost of residential customer interruption hour $2.70/h
Cost of commercial customer interruption hour $886/h
Cost of industrial customer interruption hour $3,853/h
Economic output stimulated by $1 million investment n $1,160,000
in electric power network constructio
Economic output stimulated by $1 million investment $1,160,000
in electric power network maintenance
Greenhouse gas emissions produced by $1 million 676 MtCO
2
E
investment in electric power network maintenance
and construction
Greenhouse gas emissions produced by $1 million 503 MtCO
2
E/$1 million GDP-PPP
general economic activity
Monetary value of greenhouse gas emissions 16,000,000 $/MtCO
2
E
Probability of pole/span failure as function p =
x
μ = 154,σ
2
= 27
of windspeed, x (mph)
Probability of underground equipment failure 0.13
Time to service restoration for overhead line failure 4 h
Time to service restoration for underground line failure 10 h
Time to service restoration for transformer failure 6.5 h
Average residential rate $0.11/kWh
Average commercial/industrial rate $0.16/kWh
overhead system. We can then estimate the amount of economic output stimulated
by investing in underground power line construction using a multiplier obtained
from the EIO-LCA model (The Green Design Institute 2009). We also estimate the
equivalent greenhouse gas emissions from this amount of construction from the EIO-
LCA model as 676 Mt CO
2
equivalent (MtCO
2
E) emissions per $1 million invested
in electric power network construction (NAICS Sector 237130). The environmental
impact and economic output stimulated by electricity production is slightly different,
$0.6 million and 9,160 MtCO
2
E per $1 million of economic activity, also estimated
using the EIO-LCA model. These later numbers are needed for estimating the
reduction in environmental impacts due to lost electricity production when power
systems fail after hurricanes. We consider only the contributions to global warming
potential as environmental impacts, though other emissions would be associated with
construction and power generation activities. To convert the environmental impacts
to a monetary value, we assign the spot price for a 1 MtCO
2
E certified emission
reduction (CER) on the European Climate Exchange (2009) as the economic value
of 1 MtCO
2
E. On 30 December 2009, the price of a CER in USD is $16.00 per metric
ton ($16 million per Mt). Because of the possibility of cap and trade being adopted
in the U.S., we assume the cost of 1 Mt CER is the economic value of 1 MtCO
2
E. To
estimate the private losses associated with interruptions due to system component
failures, we assume the cost of an interruption-hour for residential, commercial, and
industrial customers is $2.70, $886, and $3,853, respectively (LaCommare and Eto
2006).
Finally, we must estimate the influence of reliability on costs. We assume that
the repair cost for overhead system element failures is $4,000 per failure, while the
Climatic Change
repair cost for underground system element failures is $60,000 per failure (Xu and
Brown 2008). We ignore differences in non-storm reliability between overhead and
underground systems as a first approximation. We estimate the hurricane failure
probability of poles from Han et al. (2009). The failure probabilities of spans and pad-
mounted transformers for underground and overhead equipment and the time to
service restoration for overhead and underground system elements are approximated
based on empirical rules of thumb (Xu and Brown 2008). The cost per customer
interruption hour for residential and commercial customers is estimated from Florida
Public Utilities Commission data (Xu and Brown 2008).
5Results
5.1 Micropolis case study results
Here we present the results of our analysis for the three scenarios. For the purpose of
this base case, the portion of Micropolis east of its railroad is classified as a hurricane
storm surge category 3 zone, while the portion of Micropolis west of the railroad
is a category 5 zone. Because we make no changes to west Micropolis, and the
storm surge zone category for this area is much higher than the intensity of storms
experienced in this area, surge-induced failures in this zone will not occur in our
simulation model. Consequently, we neglect the cost of underground maintenance
and failure in this zone. The scenarios proposed are illustrated as a decision tree
in Fig. 2. Finally, our base case scenario assumes a time horizon of 50-years and a
discount rate of 8% applied to all utility investments and imposed costs. We reserve
discussion of the impact of choice of discount rate for a later section.
Fig. 2 Decision tree illustrating proposed scenarios for analysis with and without wetland restoration
Climatic Change
5.1.1 Life cycle cost results, no environmental costs considered
The life cycle cost results for the base case with no wetlands restoration and no
environmental costs included is shown in Table 4.Table4 reports the costs of each
option without considering environmental impacts over the project life cycle. This
table suggests that Scenario 1 (S1), with no wetland restoration, has the highest
project life cycle costs of any option. Although these costs are primarily the cost
of the initial undergrounding investment, we see that Scenario 1 (the complete
underground conversion) costs more than Scenario 3 (S3, the completely overhead
case) in terms of expected private losses (e.g., business losses). This increased
relative cost may be due to the large costs of downtime for an underground system
component failure. Although the expected utility costs are less for Scenario 1, the
initial mitigation investment excluded, these costs savings do not appear to justify
the undergrounding investment. Similar observations may be made for Scenario 2
(S2, underground conversion of commercial zone only) relative to Scenario 3. While
the utility expected costs are lower for Scenario 2 than Scenario 3, any advantage
these savings might suggest is tempered by the large initial investment and increased
private losses for Scenario 2 relative to Scenario 3.
Table 5 reports the costs among the options including wetland restoration. This
table suggests that when wetland restoration is included in the life cycle cost analysis,
the choice among options is similar to the decision excluding wetlands. The cost
difference and allocation among the cost components is close to the results reported
in Table 4, with the exception of the substantially higher cost of wetland restora-
tion construction. Although including wetlands reduces the costs to private losses
approximately 29%, thus reflecting a reduction in the failure rate of underground
components, this synergistic cost reduction does not justify the investment in disaster
mitigation by undergrounding and wetland restoration.
While Tables 4 and 5 report average costs, these tables do not indicate the
amount of variability in the simulation cost estimates. Figure 3 reports the empirical
cumulative density function for the absolute value of the life cycle project losses. In
this figure, expected life cycle losses are the sum of the expected utility costs and the
Table 4 Average costs for 50-year life cycle of three hardening scenarios, excluding wetland
restoration and induced economic output and environmental costs (Million USD, 8% discount rate)
Scenario 1, Scenario 2, Scenario 3,
no wetlands no wetlands no wetlands
Maintenance costs $(0.490) $(0.527) $(0.534)
Mitigation investment $(8.744) $(1.355) N/A
Planned utility costs, C1 $(9.234) $(1.882) $(0.534)
Planned costs imposed on society, C2 $– $– $–
Cost of repairs $(0.0145) $(0.0215) $(0.0157)
Lost revenue due to Disaster $(0.00587) $(0.0326) $(0.0307)
Expected Utility Costs, C3 $(0.0203) $(0.0542) $(0.0464)
Expected Private Losses $(0.0396) $(0.0332) $(0.00606)
Disaster costs imposed on society, C4 $(0.0396) $(0.0332) $(0.00606)
Total life cycle costs $(9.294) $(1.969) $(0.586)
Average number of hurricanes in 3.44 3.51 3.41
50-year project life cycle
Costs are averaged over the N = 1,000 simulations of the 50-year life cycle
Climatic Change
Table 5 Average costs for 50-year life cycle of three hardening scenarios, including wetland
restoration while excluding induced economic output and environmental costs (Million USD, 8%
discount rate)
Scenario 1, Scenario 2, Scenario 3,
with wetlands with wetlands with wetlands
Maintenance costs $(0.490) $(0.527) $(0.534)
Mitigation investment $(138) $(131) $(130)
Planned utility costs, C1 $(139) $(132) $(130)
Economic value of wetland $467 $467 $467
ecosystem services
Planned costs imposed on society, C2 $467 $467 $467
Cost of repairs $(0.0145) $(0.0206) $(0.0157)
Lost revenue due to disaster $(0.00587) $(0.0282) $(0.0329)
Expected utility costs, C3 $(0.0203) $(0.0488) $(0.0486)
Expected private losses $(0.0396) $(0.0280) $(0.00648)
Disaster costs imposed on society, C4 $(0.0396) $(0.0280) $(0.00648)
Total life cycle costs $328 $335 $336
Average number of hurricanes 3.44 3.51 3.41
disaster costs imposed on society (e.g., C3 + C4). Although Fig. 3 does not include
environmental costs or induced economic activity, Fig. 3 gives a more illustrative
picture of the variability in lifecycle costs associated with each scenario.
Figure 3 shows that, while wetland restoration does not induce large life cycle loss
differences under any scenario, Scenario 1’s cost curve is shifted to the left of the
curves for Scenarios 2 and 3. Thus, the larger average cost of Scenario 3 reported in
Tables 4 and 5 may be due to the larger tail of Scenario 1’s distribution relative to
Scenarios 2 and 3. Indeed, Fig. 4 seems to contradict the results reported in Tables 4
Fig. 3 Empirical CDFs for 1,000 simulations of the 50-year life cycle losses (C3+C4) under each
scenario
Climatic Change
Fig. 4 Empirical CDFs for 1,000 simulations of the 50-year life cycle private losses (C4) under each
scenario
and 5, especially for private losses (C4). Figure 4 reports the cost curves for private
losses under each scenario, excluding induced economic output and environmental
costs. Figure 4 suggests two observations. First, in the event of a disaster, most costs
will be borne by Micropolis’ power utility. Second, although the average private loss
is largest under Scenario 1, Scenario 1 has the largest probability that no private
losses will be incurred when compared to Scenarios 2 and 3 (70–80% vs. 55% and
40%, respectively).
5.1.2 Life cycle costs, induced economic and environmental costs included
In, Table 6 we report the results of our base-case analysis including environmental
costs, but excluding the economic value of wetlands’ ecosystem services. These
results include economic activity and greenhouse gas emissions induced (or averted)
by maintenance and mitigation activities, repair costs, revenue losses, and private
losses.
Generally, the case study results including environmental costs reinforce the ob-
servations made from the life cycle analyses excluding environmental costs. This may
be attributable primarily to the price of carbon emission reductions relative to the
amount of carbon emitted by general economic activity. First, consider private losses
under each scenario. While private losses remain greater in Scenarios 1 and 2 relative
to Scenario 3, the carbon emissions averted by private losses is only 503 MtCO2E per
$1 million private losses in terms of MtCO
2
E/$GDP-PPP. On the other hand, the
amount of carbon emitted by electricity production is nearly 20 times greater than
the amount of carbon emitted by general economic production (9,160 MtCO
2
Efrom
electricity generation:503 MtCO
2
E from economic activity). Because the private
losses are greater in the full undergrounding case, and electric power production
forces much more greenhouse gas production than general economic production, the
favorability of the status quo is enhanced by the inclusion of environmental costs.
Climatic Change
Table 6 Average costs for 50-year life cycle of three hardening scenarios, including wetland
restoration, and induced economic output and environmental costs (Million USD, 8% discount rate)
Scenario 1, Scenario 2, Scenario 3,
with wetlands with wetlands with wetlands
Maintenance costs $(0.490) $(0.527) $(0.534)
Mitigation investment $(138) $(131) $(130)
Planned utility costs, C1 $(139) $(132) $(130)
Economic output induced $0.569 $0.611 $0.619
by maintenance
Economic output induced by mitigation $160 $151 $150
Environmental costs of maintenance $(5,305) $(5,703) $(5,776)
Environmental costs of mitigation $(1,496,000) $(1,416,000) $(1,402,000)
Economic value of wetland $467 $467 $467
ecosystem services
Wetland carbon sequestration $504 $504 $504
Planned costs imposed on society, C2 $(1,500,000) $(1,421,000) $(1,406,000)
Cost of repairs $(0.0145) $(0.0206) $(0.0157)
Revenue loss due to disaster $(0.00587) $(0.0281) $(0.0329)
Expected utility costs, C3 $(0.0203) $(0.0488) $(0.0486)
Economic output induced by repairs $0.0168 $0.0239 $0.0182
Lost economic output induced $(0.00939) $(0.0451) $(0.0526)
by revenue loss
Direct private losses $(0.0396) $(0.0280) $(0.00647)
Environmental costs of repairs $(157) $(223) $(170)
Environmental costs averted $860 $4,129 $4,821
by revenue loss
Environmental costs averted $0.000319 $0.000225 $0.000052
by private losses
Disaster costs imposed on society, C4 $703 $3,906 $4,652
Total life cycle costs $(1,500,000) $(1,418,000) $(1,402,000)
Average number of hurricanes 3.44 3.51 3.41
in project cycle
In future applications of this methodology, more case-specific cost assumptions
must be made in lieu of our demonstrative assumptions. Important considerations
include construction of a hybrid EIO-LCA model for the specific construction
processes involved and the local economy impacted, wetland valuation methods
specific to the wetland habitat to be restored, and more realistic surge attenua-
tion modeling employing SLOSH or ADCIRC (e.g., Loder et al. 2009; Resio and
Westerink 2008; Wamsley et al. 2009, 2010).
5.2 Sensitivity analysis
We evaluate the sensitivity of our results to: 1.) The number of hurricanes expected to
impact the project area over the 50-year life of the project; 2.) The rate of storm surge
attenuation observed over marsh wetlands; and, 3.) The separation between social
and private decision makers represented by the choice of different social and private
time discounting rates. We choose these parameters for sensitivity analysis because
we expect that overall costs are most influenced by the total number of hurricanes
making landfall in the project area, while the potential synergies between wetland
restoration and electric power network hardening would be most influenced by the
Climatic Change
rate of storm surge attenuation observed. Furthermore, the choice of a time discount
rate is quite controversial, and we would like to understand more about how the
dichotomy among decision maker classification might influence our results.
5.2.1 Sensitivity to number of hurricanes impacting project area
First, Fig. 5 reports the average life cycle costs, excluding environmental and eco-
nomic impacts, for the three scenarios including wetland restoration under different
assumptions about the number of hurricanes expected to influence the project area.
Figure 5 assumes a wetland storm surge attenuation rate of 1 m:14.5 km, and an 8%
discount rate. Although North Carolina may be reasonably expected to receive an
average of 3–5 hurricanes in a 50-year planning horizon, we show results for a range
of 3–15 hurricanes over a 50-year planning horizon for extension to other local cases.
These results reflect the intuitive idea that the life cycle costs increase as the number
of hurricanes increases. The top panel of Fig. 5 illustrates that under Scenarios 1 and
2, utility costs are less than the costs imposed on society through private losses, while
for Scenario 3, private losses are always less than utility costs. In addition, Fig. 5
suggests that private losses under Scenario 3 are most sensitive to increases in the
number of hurricanes. The bottom panel of Fig. 5 reports the probability that no
loss is incurred over the 50-year project life cycle. Although the probability that no
loss is incurred decreases substantially for each scenario as the number of hurricanes
increases, the probability that no private loss is incurred remains much higher under
Scenario 1 than the other Scenarios proposed. For Scenario 1, the probability that
no private losses are sustained remains higher than 0.5 until the number of expected
hurricanes over the 50-year project life cycle increases beyond 10.
5.2.2 Sensitivity to wetland storm surge attenuation rate
On the other hand, our results are much less sensitive to wetland storm surge
attenuation. Figure 6 reports the average life cycle costs, excluding environmental
and economic impacts, for the three scenarios. Figure 6 assumes that five hurricanes
are expected to impact the project area over the 50-year life cycle, and that a discount
rate of 8% applies. In short, Fig. 6 suggests that life cycle costs for Micropolis are not
sensitive to changes in the wetland storm surge attenuation rate over the range of
reported surge attenuation rates. Although utility and private losses are decreased
by approximately 30% over the range of attenuation rates under Scenario 1, the
probability that no costs will be incurred is much less sensitive. This may be due
to the small range between the minimum and maximum storm surge attenuation
rates observed. Nonetheless, these results, when considered alongside the literature
investigating storm surge attenuation over wetlands, reinforce the importance of
detailed wetland modeling for local risk analysis and investment planning purposes.
While our results show that damages in Micropolis are not sensitive to wetland
storm surge attenuation rate, Gedan et al. (2010) report that 60% of the variation
in damages inflicted on US coastal communities in 34 major hurricanes since 1980 is
explained by differences in coastal wetlands.
5.2.3 Sensitivity to public and private discount rates
The choice of discount rate for evaluating public projects is controversial, especially
with respect to evaluating climate change mitigation investments. In fact, one of
Climatic Change
Fig. 5 Sensitivity of average 50-year life cycle costs (excluding environmental and economic impacts)
and probability that no losses are incurred to the number of hurricanes expected over the 50-year
project cycle
the principal streams of criticism of the Stern Review on the Economics of Climate
Change (Stern et al. 2006) concerns the assumptions made surrounding the choice of
an appropriate discount rate (Nordhaus 2007;Weitzman2007; Tol and Yohe 2009).
Climatic Change
Fig. 6 Sensitivity of average 50-year life cycle costs (excluding environmental and economic impacts)
and probability that no losses are incurred to the rate of wetland storm surge attenuation
While we do not wish to enter this debate, this discussion indicates the gravity with
which time preferences must be considered when evaluating private investments
that impose considerable public costs and benefits. Consequently, we do consider
Climatic Change
it necessary to discuss the role of the social and private decision makers when
evaluating the potential synergy between wetland restoration and infrastructure
disaster loss mitigation. As indicated, we have assumed a 50-year time horizon for
evaluating the electric power hardening investments. This time horizon is symbolic
of the difficult negotiation among public and private benefits, as the utility is
expected to develop and manage the infrastructure, while the public must reckon
with the externalities. In addition, this time horizon amplifies the importance of this
negotiation because this relatively short time horizon implicitly requires that the
private (utility) decision makers and the public (social/government) decision makers
be represented separately since their benefits accrue on different time scales. For the
private utility investments and expected costs of repairs and lost revenues, then, we
assume an 8% discount rate. This reflects the private utility’s preference to accrue
benefits earlier in the project life cycle, as only approximately 2% of the expected
costs and benefits accrued at the end of the 50-year life cycle are included in the
present value. On the other hand, the choice of a social discount rate for the costs
imposed on society by service interruptions and utility investments is assumed to
Table 7 Sensitivity of average costs for 50-year life cycle of three hardening scenarios to two-decision
maker scenario, including wetland restoration, and induced economic output and environmental
costs. (Millions USD) assumes 8% private, 2% social discounting rate
Scenario 1, Scenario 2, Scenario 3,
with wetlands with wetlands with wetlands
Maintenance costs $(0.490) $(0.527) $(0.534)
Mitigation investment $(138) $(131) $(130)
Planned utility costs, C1 $(139) $(132) $(130)
Economic output induced $1.461 $1.571 $1.591
by maintenance
Economic output induced by mitigation $170 $161 $159
Environmental costs of maintenance $(13,629) $(14,649) $(14,836)
Environmental costs of mitigation $(1,584,700) $(1,500,100) $(1,484,500)
Economic value of wetland $1,199 $1,199 $1,199
ecosystem services
Wetland carbon sequestration $1,295 $1,295 $1,295
Planned costs imposed on society, C2 $(1,596,000) $(1,512,000) $(1,497,000)
Cost of repairs $(0.0148) $(0.0211) $(0.0158)
Revenue loss due to disaster $(0.00374) $(0.0298) $(0.0360)
Expected utility costs, C3 $(0.0186) $(0.0509) $(0.0518)
Economic output induced by repairs $0.0475 $0.0595 $0.0484
Lost economic output induced $(0.0166) $(0.107) $(0.152)
by revenue loss
Direct private losses $(0.0565) $(0.0714) $(0.0189)
Environmental costs of repairs $(442) $(555) $(451)
Environmental costs averted s $1,521 $9,828 $13,919
by revenue loss
Environmental costs averted $0.000454 $0.000574 $0.000151
by private losses
Disaster costs imposed on society, C4 $1,078 $9,273 $13,467
Total life cycle costs $(1,594,707) $(1,502,913) $(1,483,394)
Average number of hurricanes 3.41 3.51 3.44
in project cycle
Climatic Change
be 2%. This reflects the public’s general unwillingness to bear the costs imposed
on them by electric power infrastructure construction, repairs, or failure. An 8%
discount rate incorporates approximately 37% of the expected costs and benefits
accrued to society at the end of the 50-year life cycle are included in the present
value. This public–private tradeoff implies that costs over the duration of the project
life-cycle and beyond should more greatly influence public decision-making when
considering electric power infrastructure hardening investments. We report results
for this case, including wetland restoration investment, and induced social, economic,
and environmental costs, in Table 7.Table7 shows that, while the environmental and
social costs and benefits associated with risk mitigation investments take on increased
importance if a social rate of discounting is assumed, the decision is still dominated
by the price of carbon. We also evaluated 16 pairwise combinations of social discount
(0%, 1%, 2%, 3%) and private discount (6%, 8%, 10%, and 12%) rates. These
results are not reported here, as the results are not sensitive to changes in the
discount rates over these ranges. Nonetheless, our sensitivity analysis indicates that
the decision is still dominated by the price of carbon emitted during construction and
maintenance of restored wetlands and electric power network hardening. Thus, when
evaluating risk mitigation and natural restoration decisions, the choice of discount
rate may be most important for larger projects, especially if the price of carbon is not
included.
6 Lessons learned
In summary, our multi-criteria life cycle framework suggests several interesting
implications for future research and planning for wetland restoration projects when
applied to disaster loss mitigation. Table 8 summarizes the results for each scenario
for each of the three cases examined. For all cases, no infrastructure hardening and
no wetland restoration is justified in our case study location. However, we do suggest
a few specific areas for additional research:
1. The life cycle analysis must consider non-monetized ecological benefits of wet-
land restoration and account for the CO
2
production they avert. For example,
when considering CO
2
emissions in the analysis, the potential for the restored
wetland to become a carbon sink must be evaluated. Even modest amounts of
wetland CO
2
sequestration may improve the attractiveness of the project when
Table 8 Summary of the results for each infrastructure hardening scenario for each case
Case: 1 2 3
Wetland restoration: Excluded Included Included
Induced economic output: Excluded Excluded Included
Environmental impacts: Excluded Excluded Included
Scenario E[NPV] E[NPV] E[NPV]
1: Underground all eligible components ($9,294) ($0.0396) ($1,500,000)
2: Underground only the commercial district $(1,969) ($0.0280) ($1,418,000)
3: No undergrounding ($0.543) ($0.006) ($1,402,000)
Preferred alternative Do nothing Do nothing Do nothing
Costs are in million of US dollars
Climatic Change
considered in conjunction with the economic value of the wetlands ecosystem
services and the CO
2
emissions averted through ecosystem process production.
2. In the future, simulation of storm surge propagation over constructed or natural
wetlands using more sophisticated models such as SLOSH or ADCIRC may be
the most accurate method of disaster risk assessment. This approach has been
performed in several of the studies cited here (Loder et al. 2009; Resio and
Westerink 2008; Wamsley et al. 2009, 2010), and may more accurately capture the
effects of spatial and temporal heterogeneity in wetland topology, composition,
and continuity. Furthermore, storm surge simulation may allow for more flexible
examination of the influence of factors not yet considered, including wetland
morphology and wave setup. In addition, because the response of wetlands to
storms, and storm surge to wetlands, may be interactive in nature, stochastic
simulations may be necessary to capture any potential feedback cycles between
wetland morphology and storm surge attenuation (e.g., consider findings of
Wamsley et al. 2009)
3. This work must be replicated using either a larger model city testbed, or a
testbed with a different distribution of residential, commercial, and industrial
customers. This work suggests that the attractiveness of a wetland restoration
project for disaster loss mitigation may be sensitive to the customer mix, although
we do not evaluate this sensitivity in this paper. Furthermore, detailed regional
economic resilience models should be developed and employed when applying
our framework to a real case. The economic model may become more important
when extending the framework to interdependent infrastructures.
4. Although our case study reports that undergrounding and wetland restoration
may increase social costs relative to a completely overhead electric power
network configuration, the probability that no social costs are imposed is greatly
increased under the undergrounding and wetland restoration scenarios proposed
in this paper. Future investigation may find this property of electric power
network hardening valuable, and should be investigated in conjunction with
storm surge simulation and interdependent infrastructure evaluation.
Overall, our results have important implications for coastal adaptation. Natural
ecosystem restoration may be difficult to justify as an approach for coastal adaptation
without considering the valuation of biodiversity, ecological services, and carbon
sequestration. While these issues pose difficult valuation problems, the more directly
quantifiable benefits of wetland restoration may not offset the significant cost of
restoring or building wetlands in many locations.
Acknowledgements We gratefully acknowledge Drs. Kelly Brumbelow and Alex Sprintson of
Texas A&M University for providing the Micropolis case study, and Dr. Steven Quiring of Texas
A&M University for providing the climatic data used in developing the hurricane count model. We
also acknowledge the funding sources for this work, the National Science Foundation (grant ECCS-
0725823), the U.S. Department of Energy (grant BER-FG02-08ER64644) and the Whiting School
of Engineering. However, all opinions in this paper are those of the authors and do not necessarily
reflect the views of the sponsors.
References
Arrow KJ, Cropper ML, Eads GC, Hahn RW, Lave LB, Noll RG, Portney PR, Russell M,
Schmalensee R, Smith VK, Stavins RN (1996) Is there a role for benefit cost analysis in envi-
ronmental, health, and safety regulation? Science 272:221–222
Climatic Change
Bigger JE, Willingham MG, Krimgold F (2009) Consequences of critical infrastructure interdepen-
dencies: lessons from the 2004 hurricane season in Florida. Int J Critical Infrastructures 5:199–219
Bridgham SD, Megonigal JP, Keller JK, Bliss NB, Trettin C (2006) The carbon balance of North
American wetlands. Wetlands 26:889–916
Brown RE (2009) Cost-benefit analysis of the deployment of utility infrastructure upgrades and
storm hardening programs. Quanta Technology, Raleigh
Brumbelow K, Torres J, Guikema SD, Bristow E, Kanta L (2007) Virtual cities for water distrib-
ution and infrastructure system research. In Proc. World Environmental and Water Resources
Congress 2007: Restoring Our Natural Habitat, ASCE/EWRI
Chang SE (2003) Evaluating disaster mitigations: methodology for urban infrastructure systems. Nat
Hazards Rev 4:186–196
Chang SE, Shinozuka M (1996) Life-cycle cost analysis with natural hazard risk. J Infrastruct Syst
2:118–126
Chmura GL, Anisfeld SC, Cahoon DR, Lynch JC (2003) Global carbon sequestration in tidal, saline
wetland soils. Glob Biogeochem Cycles 17:1111–1123
Costanza R, Farber SC, Maxwell J (1989) Valuation and management of wetland ecosystems. Ecol
Econ 1:335–361
Costanza R, d’Arge R, de Groot R, Farber S, Grasso M, Hannon B, Limburg K, Naeem S, O’Neill
RV, Paruelo J, Raskin RG, Sutton P, van den Belt M (1997) The value of the world’s ecosystem
services and natural capital. Nature 387:253–260
Day JW Jr, Boesch DF, Clairain EJ, Kemp GP, Laska SB, Mitsch WJ, Orth K, Mashriqui H, Reed
DJ, Shabman L, Simenstad CA, Streever BJ, Twilley RR, Watson CC, Wells JT, Whigham DF
(2007) Restoration of the Mississippi Delta: lessons from Hurricanes Katrina and Rita. Science
315:1679–1684
Emanuel K (2005) Increasing destructiveness of tropical cyclones over the past 30 years. Nature
436:686–688
Gedan KB, Kirwan ML, Wolanski E, Barbier EB, Silliman BR (2010) The present and future role of
coastal wetland vegetation in protecting shorelines: answering recent challenges to the paradigm.
Clim Change. doi:10.1007/s10584-010-0003-7
Han S-R, Guikema SD, Quiring SM (2009) Improving the predictive accuracy of hurricane power
outage forecasts using generalized additive models. Risk Anal 29:1143–1153
Hendrickson CT, Lave LB, Matthews HS (2006) Environmental life cycle assessment of goods and
services: an input–output approach. Resources for the Future, Washington, DC
Keeney RL (1982) Decision analysis: an overview. Oper Res 30:803–838
LaCommare KH, Eto JH (2006) Cost of power interruptions to electricity consumers in the United
States. Energy 31:1845–1855
Loder NM, Irish JL, Cialone MA, Wamsley TV (2009) Sensitivity of hurricane surge
to morphological parameters of coastal wetlands. Estuar Coast Shelf Sci 84(4):625–636.
doi:10.1016/j.ecss.2009.07.036
Nicholls RJ, Small C (2002) Improved estimates of coastal population and exposure to hazards
released. Eos 83:301–304
Nordhaus WD (2007) A review of the Stern Review on the Economics of Climate Change. J Econ
Lit XLV:686–702
Resio DT, Westerink JJ (2008) Modeling the physics of storm surges. Phys Today 61:33–38
Sabbatelli TA, Mann ME (2007) The influence of climate state variables on Atlantic Tropical
Cyclone occurrence rates. J Geophys Res 112:D17114. doi:10.129/2007JD008385
Small C, Nicholls RJ (2003) A global analysis of human settlement in coastal zones. J Coast Res
19:584–599
Stern N, Peters S, Bakhshi V, Bowen A, Cameron C, Catovsky S, Crane D, Cruickshank S, Dietz S,
Edmonson N, Garbett S-L, Hamid L, Hoffman G, Ingram D, Jones B, Patmore N, Radcliffe H,
Sathiyarajah R, Stock M, Taylor C, Vernon T, Wanjie H, Zenghelis D (2006) The economics of
climate change. HM Treasury, London
The European Climate Exchange (2009). http://www.ecx.eu. Accessed 15 February 2010
The Green Design Institute (2009) Economic input–output life cycle analysis model (EIO-LCA).
Carnegie Mellon University, Pittsburgh, PA. http://www.eiolca.net
Tol RSJ, Yohe GW (2009) The Stern Review: a deconstruction. Energy Policy 37:1032–1040
Turner RE, Streever BJ (2006) Approaches to coastal wetland restoration: Northern Gulf of Mexico.
SPB Academic Publishing, The Hague
U.S. Army Corps of Engineers, New Orleans District (1963) Interim survey report, Morgan City, LA
and vicinity. New Orleans, LA
Climatic Change
U.S. Army Corps of Engineers, New Orleans District (2004) Louisiana coastal area ecosystem
restoration study. New Orleans, LA
U.S. Congress Office of Technology Assessment (1990) Physical vulnerability of electric systems to
natural disasters and sabotage, OTA-E-453, U.S. Government Printing Office, Washington, DC
Wamsley TV, Cialone MA, Smith JM, Ebersole BA, Grzegorzewski AS (2009) Influence of land-
scape restoration and degradation on storm surge and waves in southern Louisiana. Nat Hazards
51:207–224
Wamsley TV, Cialone MA, Smith JM, Atkinson JH, Rosati JD (2010) The potential of wetlands in
reducing storm surge. Ocean Eng 37(1):59–68. doi:10.1016/j.oceaneng.2009.07.018
Weitzman ML (2007) A review of the Stern Review on the Economics of Climate Change. J Econ
Lit XLV:703–724
Xu L, Brown RE (2008) Undergrounding assessment Phas 3 Report: Ex Ante Cost and Benefit
Modeling. Quanta Technology, Raleigh
