You are on page 1of 7

Journal of Coastal Research 29 1 1–7 Coconut Creek, Florida January 2013

Scenario-Based Projection of Extreme Sea Levels


Jayantha Obeysekera† and Joseph Park‡

† ‡
Hydrologic and Environmental Systems Modeling NOAA/NOS/CO-OPS
South Florida Water Management District 1305 East West Highway
3301 Gun Club Road Silver Spring, MD 20910, U.S.A.
West Palm Beach, FL 33406, U.S.A.
jobey@sfwmd.gov

ABSTRACT
Obeysekera, J. and Park, J., 2013. Scenario-based projection of extreme sea levels. Journal of Coastal Research, 29(1), 1–
7. Coconut Creek (Florida), ISSN 0749-0208.

Heavily populated urban centers and natural areas located in low-lying coastal regions are highly vulnerable to sea-level
extremes. Historical data at many tide gages suggest that changes over time in extremes generally follow the rise in
mean sea level. Assuming this observation to hold in the future, a relationship between mean sea-level rise and its
associated extremes with a generalized extreme value distribution can provide future return levels of extreme sea levels.
Current projections of future sea level, which include varying degrees of acceleration, may result in large increases in
extremes that need to be accounted for in the evaluation of existing coastal projects or in the planning of new ones.
Because precise quantitative estimates of the uncertainties in sea-level rise projections are not available, scenario-based
approaches have been suggested for project evaluation and design. Here, we propose a general method based on the
synthesis of extreme value statistics with sea-level rise scenarios that allows any combination of linear or nonlinear local
and global sea-level rise components and can accommodate the nonstationary evolution of sea-level extremes. The
temporal variation of the design level of protection for coastal projects, expressed as the return period of extreme events,
and the future behavior of the risk are explored. The concepts are demonstrated through application to tide gage data at
several locations in the United States.

ADDITIONAL INDEX WORDS: Sea level extremes, generalized extreme value distribution, climate change, risk,
nonstationarity.

INTRODUCTION Impacts of coastal extremes include, but are not limited to,
Low-lying regions and heavily developed coastal zones with shoreline erosion, submergence of coastal wetlands and
dense population are particularly vulnerable to extreme sea mangrove areas, and the reduction of discharge capacity of
levels. In a study of port cities across the globe, Nicholls et al. coastal water control structures (Obeysekera et al., 2010). In
(2008) estimated that about 40 million people, 0.6% of the view of projected increases in global mean sea level, by some
global population, are currently exposed each year to coastal estimates upwards of 1 m or more by 2100, the expected
flooding of events that may exceed the 100-year flood. Titus et extreme sea water level at a particular location will increase at
al. (2009) estimated that almost 60% of the land below 1 m least in a similar magnitude, and such increases in the future
along the Atlantic Coast of the United States is expected to be will exacerbate the impacts of storm surges and waves,
developed further, increasing future vulnerability to coastal particularly in low-lying regions of the world.
extremes. Extreme sea levels are influenced by a combination Global sea-level rise projections for the 21st century, as
of factors, which include astronomical tides, waves and reported in the Fourth Assessment Report (AR4) of the
currents, extreme weather phenomena, and teleconnections Intergovernmental Panel on Climate Change (IPCC) (IPCC,
to such global phenomena as El Niño-Southern Oscillation. 2007a), were in the range of 0.18 to 0.59 m. However, it is
These extremes are also influenced by sea-level rise associated important to note that the IPCC in its 2007 synthesis report
with long-term volume changes in the oceans attributable to qualified the projections as follows:
thermal loading and melting of ice sheets and freshwater The sea level projections do not include uncertainties in
inputs from land. Extremes relative to a land-based datum are climate-carbon cycle feedbacks nor do they include the
additionally influenced by vertical land movement (VLM) full effects of changes in ice sheet flow, because a basis in
associated with subsidence or uplift that might occur due to published literature is lacking. Therefore the upper
many factors. values of the ranges given are not to be considered
upper bounds for sea level rise. The projections include a
DOI: 10.2112/JCOASTRES-D-12-00127.1 received 25 June 2012; contribution due to increased ice flow from Greenland
accepted in revision 11 August 2012; corrected proofs received 15
and Antarctica at the rates observed for 1993–2003, but
October 2012.
Published Pre-print online 6 November 2012. these flow rates could increase or decrease in the future.
Ó Coastal Education & Research Foundation 2013 If this contribution were to grow linearly with global
2 Obeysekera and Park

average temperature change, the upper ranges of sea and Allan, 2010). Currently, projections based on dynamical
level rise for SRES scenarios ... would increase by 0.1 to models are not sufficiently accurate to suggest whether any
0.2m. (IPCC, 2007b, p. 45) nonstationary behavior of storminess will continue at the
historical rate or how it may change during the 21st century
As indicated previously, the IPCC report did not exclude a
due to climate change. However, it is important for a scenario-
possible larger rise due to highly uncertain ice sheet dynamics
based method to accommodate the treatment of possible
because it did not have sufficient information to adjust the
nonstationarity in the probability distribution of sea-level
projections. An alternative approach using semi-empirical
extremes (Menéndez and Woodworth, 2010; Ruggiero, Komar,
methods linking sea-level rise (SLR) to global temperature
and Allan, 2010).
has provided larger estimates of global SLR than what was
The general lack of an acceptable practice in prescribing
provided in the IPCC report (Rahmstorf, 2007; Vermeer and
fixed or probabilistic projections of SLR have led to
Rahmstorf, 2009). Various other methods of projection have
‘‘scenario-based’’ guidance, including a range of possible
provided a broad range of global SLR in the literature, but they
acceleration for planning and design of future infrastructure
do not provide sufficient information on the likelihood of a
projects (NRC, 1987; USACE, 2011). This paper uses the
particular magnitude. Few attempts have been made to
scenario-based approach to project changes in mean sea
provide probabilistic projections of global SLR (Horton et al.,
2008; Hunter, 2010; Obeysekera et al., 2012; Titus and level and then couples these projections with extreme value
Narayanan, 1995). statistics derived from tide gage data to project changes in
Through the analysis of long-term tide gage data, some design life and extreme event return periods for coastal
studies have found that historical sea-level data indicate little infrastructure. The method allows any combination of linear
or no acceleration (Houston and Dean, 2011; Watson, 2011). or nonlinear sea-level rise components, including zero or
However, Sallinger, Doran, and Howd (2012) recently found nonzero acceleration, and can easily incorporate a nonsta-
that the Atlantic coast of North America is a hotspot for tionary evolution of sea-level extremes. In the next section,
accelerated sea-level rise attributable to variations in ocean methods for predicting the nonstationary behavior of the
circulation and salinity patterns. In addition, many global and extreme sea levels, the associated probabilities of exceedance
regional studies, mostly using satellite altimetry data available or, equivalently, the return period, and the risk of a coastal
since the early 1990s, have also detected a higher rate of sea project are presented. The paper concludes by demonstrat-
level rise as compared to those indicated by the tide gages ing the application of the method for several tide gage
(Holgate and Woodworth, 2004). Due to the short length of the locations in the United States.
record it is difficult to attribute the recent increase in the rate of
sea-level rise to a particular factor, and such rates of similar METHODS
magnitude have occurred in the sea-level data reconstructed
from tide gages. Numerous projections available for the 21st
Scenario-Based Global Sea-Level Rise Projections
century consider increasing rates of sea-level rise, i.e. acceler- Scenario planning has been a useful tool for dealing with
ation (Houston, 2012; IPCC, 2007a). For example, IPCC AR4 uncertainties of projections associated with climate change and
report (IPCC, 2007a, p. 821) concludes that ‘‘In all scenarios, sea-level rise. Based on a methodology developed by the U.S.
the average rate of rise during the 21st century is very likely to National Research Council (NRC, 1987), the U.S. Army Corps
exceed the 1961 to 2003 average rate of 1.8 6 0.5 mm yr1.’’ of Engineers (USACE, 2011) has developed such a scenario-
In addition to societal and infrastructure impacts from mean based approach for considering sea-level rise across the life
sea-level rise, the consideration of future extremes is important cycle of a project. In this paper, the scenarios proposed in the
for the planning and design of coastal infrastructure (Zhang, USACE methodology are coupled with the probabilistic
Douglas, and Leatherman, 2000). Menéndez and Woodworth projections of extremes without any loss of generality in the
(2010) analyzed sea-level extremes of a quasi-global sea-level overall approach.
dataset and suggested that changes in extremes over time are The proposed method consolidates various sources of sea-
attributable to change in the mean sea level. Lowe et al. (2010) level rise at a particular location into two components: a
reviewed several previous investigations and concluded that eustatic component representing the global sea-level rise, and a
there is little evidence for extreme sea-level changes that are local component representing factors such as uplift/subsidence
significantly different from mean sea level over an extended attributable to anthropogenic and natural causes, regional
period of time and that these findings have also been oceanographic forcing, and postglacial rebound in the form of
corroborated by others (Araújo and Pugh, 2008; Park et al., glacial isostatic adjustment (GIA) (Jiang, Dixon, and Wdowin-
2010; Woodworth and Blackman, 2004). The observation that ski, 2010; Peltier, 1999):
increases in historical extremes can be related simply to
SðtÞ ¼ GðtÞ þ LðtÞ: ð1Þ
increases in mean sea level simplifies the incorporation of time-
dependent changes of the extremes into the probabilistic In Equation (1), G(t) represents the global contribution and
evaluation of future scenarios and risks (Obeysekera et al., L(t) represents the local component attributable to VLM and
2012). However, several studies on the both the east and west other aforementioned location specific factors, both assumed to
coasts of the United States have demonstrated additional vary as functions of time, t. For the global component of mean
nonstationarity in extreme sea levels due to changes in sea level, we use the quadratic form (NRC, 1987; USACE,
storminess (e.g. Komar and Allan, 2008; Ruggiero, Komar, 2011):

Journal of Coastal Research, Vol. 29, No. 1, 2013


Scenario-Based Projection of Extreme Sea Levels 3

GðtÞ ¼ at þ bt2 ; ð2Þ Scenario-Based Extremes


In the theory of extremes the asymptotic distribution known
where a and b are coefficients representing a base rate and
as the generalized extreme value (GEV) distribution is widely
acceleration of sea level, respectively. The use of the
used (Coles, 2001; Tomasin and Pirazzoli, 2008). In particular,
quadratic form allows the incorporation of potential acceler-
Coles (2001) presents an example modeling sea-level extremes
ation, which is largely absent in historical data but expected
using the GEV distribution. The use of GEV has the advantage
in the future due to increased melting of land-based ice
that future nonstationarity can be included explicitly by using
sheets. However, it is noted that the acceleration itself may
time-dependent parameters. Specifically for the proposed
not be constant, as implied in the quadratic form, and such
approach, the cumulative distribution function for the GEV of
nonlinearities may further complicate future sea-level pro-
the extreme (annual maximum) relative sea-level rise (SE) is
jections. In view of the lack of information on a possible
expressed as
nonconstant acceleration term, the approach presented in the
(    )
ensuing sections assumes that the quadratic form is valid, as SE  lðtÞ 1=e
used in NRC (1987) and USACE (2011). FðSE ; tÞ ¼ exp  1 þ e ; ð6aÞ
rðtÞ
The local component in Equation (1) may exhibit significant
spatial and interannual variability, as it represents localized where l(t) is the location parameter representing the time-
VLM and oceanographic, meteorological, and geophysical dependent sea-level rise, and r(t) and e are the scale and shape
forcings (Prandi, Cazenave, and Becker, 2009). In cases where parameters of the GEV, respectively. When e  0, Equation (6)
local tide gage data are unavailable, models for the GIA and the reduces to the well-known Gumbel distribution (Coles, 2001):
estimates of VLM, possibly from measurements of Global   
SE  lðtÞ
Position Systems (GPS), can provide estimates of linear rates FðSE ; tÞ ¼ exp exp : ð6bÞ
that can be used over the typical planning horizons of coastal rðtÞ
projects. Consistent with earlier studies (NRC, 1987; Obeyse- The nonstationarity in extreme sea-level behavior, either
kera et al., 2012), it is assumed that the local component varies attributable to continuing increase in the mean sea level and/or
linearly with time, although the proposed methodology can be changing storminess, may be modeled by assuming that both
generalized easily if quantitative information regarding non- the location parameter, l, and the shape parameter, r, are time
linear variation of factors affecting relative sea-level rise dependent, as shown in Equation (6) (Menéndez and Wood-
becomes available. The only quantification of nonlinear worth, 2010). For instance, the ‘‘upper bound’’ of the annual
lithospheric uplift that the authors are aware of is the work maxima may also increase with time, and in such cases the
of Jiang, Dixon, and Wdowinski (2010), which suggested an scale parameter may have to be modeled as a function of time.
accelerating uplift in the North Atlantic region in response to Some examples are l(t) ¼ l0 þ l1t and log{r(t)} ¼ r0 þ r1t
rapid glacial ice reduction. (Ruggiero, Komar, and Allan, 2010). The parameters can also
Because L(t) is assumed to represent a linear contribution (in be modeled as functions of exogenous variables as covariates.
time), one can combine it with the linear term in G(t) so that the The shape parameter, e, is difficult to estimate with precision;
total mean sea-level rise, S, can be represented as therefore, it is not normally modeled as a function of time
(Coles, 2001). Typically, the length of the tide gauge data is not
SðtÞ ¼ ½ LðtÞ þ at þ bt2 ¼ ct þ bt2 ; ð3Þ
long enough to determine the time-dependent functions of
in which the coefficient c incorporates both linear contributions parameters, and, therefore, as a first approximation, we
of the global rise as well as the local component. With this assume only l varies with time and that both r and e are
formulation the coefficients c and b represent the current rate constant. This assumption still allows the extreme value
of sea-level rise and acceleration, respectively (Obeysekera et distribution to be nonstationary but only through the variation
al., 2012). In general, the quadratic form in Equation (3) models in the location parameter.
a nonlinear increase in SLR. In a scenario-based approach As previously discussed, evidence supports a simple linear
based on the premise that the rate of acceleration over the relationship between the evolution of mean sea level and
planning horizon is not known precisely, several scenarios for statistics of the coastal extremes. Hence, it is suggested that an
the sea-level rise can be developed by using varying values for offset relating mean sea-level rise to parameters of the extreme
the coefficient b. For example, using c ¼ 0.0017 m/y, which is value distribution at a particular location. More specifically,
equal to global average SLR rate (Church and White, 2011), the dynamic change in extremes is modeled by assuming that
USACE (2011) included three scenarios dubbed ‘‘modified the location of parameter is related to the mean sea level
NRC-I, II, and III’’ with b values equal to 2.71 3 105, 7 3 105, according to the following expression:
and 1.13 3 104, respectively. It should be noted that these
lðtÞ ¼ SðtÞ þ eðtÞ; ð7Þ
values for the coefficient b were not estimated from historical
data but rather computed from assumed scenarios for total sea- where e(t) is a positive offset to the mean sea level, which in
level rise by 2100 (USACE, 2011). The approach provides general may be a function of time, t.
multiple scenarios that may be used to plan and design coastal The reasonableness of this assumption was investigated
projects and infrastructure, which are expected to last many using hourly sea-level data for several gages along the coast of
decades into the future. Which scenarios should be considered the United States, available from the University of Hawaii Sea
in planning and design depends on many factors, for example, Level Center (UHSLC, 2011). Figure 1 shows the results of the
the planning horizon and consequences of failure. investigation for about five out of 29 gages for which the annual

Journal of Coastal Research, Vol. 29, No. 1, 2013


4 Obeysekera and Park

Figure 1. Locations of the tide gages across the United States used to investigate the relationship between mean sea level and the location parameter of the GEV
(upper middle). Plots of mean annual sea level, annual extremes, and the fitted lines for the mean sea level and the location parameter of the fitted GEV for the
following five locations are also shown: SF-San Francisco; SD-San Diego; GP21-Galveston Pier 21; KW-Key West; and AC-Atlantic City. (Color for this figure is
available in the online version of this paper.)

mean sea level and the annual extremes show a clear r


RE ðt; pÞ ¼ ct þ bt2 þ e  ½1  flnð1  pÞge  ð9aÞ
interdependence. For the five stations, the linear fit for the e
mean sea level and the fitted time-dependent location param- for the case of e „ 0; for e  0,
eter (l) of the GEV (both shown as solid straight lines) are
shown. It is clear that the fixed offset assumption in Equation 7 RE ðt; pÞ ¼ ct þ bt2 þ e  rlnflnð1  pÞg: ð9bÞ
relating the location parameter of the extreme sea levels with
mean sea level is reasonable. A simple t test to compare slopes It should be noted that in deference to the typical approach of
of linear fits to mean sea level and the annual extrema finds assuming stationarity in the application of extreme value
that over 60% of the stations sampled (Figure 1) did not show a distribution, the return level corresponding to a particular
significant difference at the 5% level. However, in the event exceedance probability, p, is a function of time, and hence the
that slopes are different, or behavior of extremes relative to nonstationarity can be modeled explicitly.
mean sea level is nonlinear, the methodology can be modified
easily to incorporate such differences.
Nonstationarity in Return Period and Risk
For planning and design, a return level corresponding to a Assume that a structure, such as a sea wall, is designed for a
particular probability of exceedance, p, or equivalently the particular return period s0 ¼ 1/p0 and completion at time t0.
return period, s ¼ 1/p is desired. For the GEV in Equation 6 the Equations 9(a) or 9(b) can be used to compute the design return
return level is given by (Coles, 2001) level, R0 ¼ RE(t0; p0). Because the mean sea level is increasing
with time, the future return period s(t) or the level of protection
r
RE ðt; pÞ ¼ lðtÞ  ½1  flnð1  pÞge : ð8aÞ relative to the original design will decrease with time. Return
e
period as a function of time, t, relative to t0, is given by
When e  0, the corresponding expression is
1
RE ðt; pÞ ¼ lðtÞ  rlnflnð1  pÞg: ð8bÞ sðtÞ ¼ h 1=e i ð10aÞ
e
1  exp  1þ r ðR0  ct  bt2  eÞ
Equations 8(a) and 8(b) may be combined with Equations (3)
and (7) to yield for the case of e „ 0; while for e  0,

Journal of Coastal Research, Vol. 29, No. 1, 2013


Scenario-Based Projection of Extreme Sea Levels 5

Table 1. Parameters of the fitted GEV distributions, slopes of mean sea level, and the location parameter and the computed offset, e, for the five tide gage
locations shown in Figure 1.

Record Slope of Slope of Location Scale Shape


Location State Length the MSL Parameter Parameter (r) Parameter (e) Offset (e)
Atlantic City New Jersey 97 3.987 5.556 151.72 0.089 1376
Galveston Pier21 Texas 107 6.331 6.522 161.57 0.318 676
Key West Florida 95 2.322 2.604 52.56 0.199 552
San Diego California 102 2.024 2.234 58.11 0.184 1396
San Francisco California 111 1.932 2.543 90.80 0.062 1286

1 West gage is the period 1983–2001, and the reference year for
sðtÞ ¼
 : ð10bÞ
1  exp exp r1 ðct þ bt2 þ e  R0 Þ all calculations is the midpoint of NTDE, which is 1992
(USACE, 2011). In the absence of any apparent nonlinearity
A change in the return period corresponding to the original in the historical trend, the prevailing rate of sea-level rise for
design implies that the corresponding risk of failure will also be the reference year (i.e. parameter c in Equation [3]), is assumed
a function of time. Risk of failure here is defined as the to be equal to the linear trend computed for the entire period of
occurrence of one or more exceedances of the design return record. For the Key West gage, c is equal to 2.32 mm/y. The
level over the design life of a project. In the stationary case GEV fitting of the annual extremes with a linear trend for the
where s does not change with time, the expression for risk r for location parameter was accomplished by using the R-package
planning horizon of n years is given by extRemes (Gilleland and Katz, 2011), and the corresponding
r ¼ 1  ð1  pÞn ¼ 1  ð1  1=sÞn : ð11Þ results for the remaining parameters are r ¼ 52.56 (4.8) mm
and e ¼ 0.199 (0.08), where quantities in parentheses
Under nonstationary conditions, we define a new concept correspond to standard errors of the parameter estimate. The
called ‘‘remaining risk.’’ Assuming that a project has survived fixed offset e in Equation (7) is computed as the average
up to a particular time, ts, since its completion at time t0, the difference between the two trend lines for the Key West gage
risk of failure over the remaining life of the project, i.e. shown in Figure 1. The computed value for e is 552 mm.
remaining risk, will actually decrease with time even under the Following USACE (2011), three scenarios are considered: low
conditions of stationarity. However, under conditions of non- (historical rate), intermediate (corresponding to NRC-I), and
stationarity, i.e. change in s, the rate of risk reduction will be high (corresponding to NRC-III). Using the corresponding
smaller. The risk over the remaining life of the project, denoted value of the parameter b as specified in USACE (2011) and the
as rr is given by computed value for c previously, the three scenarios of mean
Yn   sea-level rise at Key West, as computed using Equation (7), are
1
rrðts Þ ¼ 1  1 : ð12Þ shown in Figure 2a. For each of the scenarios, Equations (3),
k¼t
sðkÞ
s (7), and 8(a) can now be used to compute the temporal variation
If the change in risk is unacceptable, it may be used as a of return level for a given return period, s. Figure 2b shows
measure for determining when to implement an adaptation example curves for s ¼ 100 years.
measure to retrofit or replace a project. To examine the nonstationarity in return period (s) and risk
(r), assume that the design period for a project in the vicinity of
Applications Key West is s0 ¼ 100 years, its project life is n ¼ 50 years, and its
Parameter Estimation completion date is 2010. Equations 10(a) and (12) can now be
The concepts presented in the previous section are illustrated used to explore the future reduction in return period, s(t), and
with application to several locations along the coast of the the temporal pattern of risk reduction. For this case, the
United States (Figure 1). The fitted parameters for the five variation of s(t) for each of the three scenarios is shown in
locations identified in Figure 1 are shown in Table 1. The tide Figure 3(a). The ‘‘stationary’’ case shown in Figure 3(b)
gages have an average record length of over 100 years. These assumes that the return period of 100 years will not change
stations are shown to illustrate some special cases that may over the design life.
arise in the application of the previous methodology. First, Some interesting patterns of nonstationary behavior in both
the slope of the mean sea level may be different from that of return period and remaining risk can be seen from Figure 3.
the location parameter, as is the case for Atlantic City and Return period reduction can be very drastic in the case of
San Francisco tide gages. In this case, the offset may not be ‘‘high’’ sea-level rise scenario. By 2040, i.e. 30 years after the
fixed but a function of time. Second, it is not uncommon to completion of the project, the corresponding return period may
have the fitted shape parameter that is negative or almost be as low as 57 years, 36 years, or 6 years for low, intermediate,
close to zero. In the case of the latter, the Gumbel extreme and high SLR scenarios. Clearly, depending on the scenario,
value distribution may be more appropriate. It is noted that this is a drastic reduction in the level of protection for the
the previous methodology can deal with such situations. project. The original risk of exceedance (i.e. one or more over
The results are further illustrated for the Key West gage, the design life) is 39%. By 2040, the risk over the remaining life
which is located in one of the most vulnerable locations for sea- (over 20 years left) would be about 18%. However, this risk
level rise. The National Tide Datum Epoch (NTDE) for the Key would increase to 35%, 63%, and 100% for the low, interme-

Journal of Coastal Research, Vol. 29, No. 1, 2013


6 Obeysekera and Park

Figure 2. (a) Mean sea level and (b) return level corresponding to return Figure 3. (a) Future reduction in the 100-year design period of a project
period ¼ 100 years for low, intermediate, and high mean sea-level rise completed in 2010 based on historic tide gauge data from Key West, Florida.
scenarios based on historic tide gauge data from Key West, Florida. (Color for (b) Variation of risk remaining until the end of design life for project with a
this figure is available in the online version of this paper.) 50-year design life. The green curve shown in (b) plots the risk reduction in
the stationary case if the return period remains 100 years throughout the
design life.

diate, and high scenarios, respectively. It should be noted that


for the high scenario the risk remains at 100% throughout the intermediate scenarios may be adequate. The approach
design life due to the fact that toward the end of the design life, presented here also provides a means for planning adaptation
the return period decreases to 1 year, implying that the strategies. The nonstationary, extreme value analysis present-
exceedance of the design return level is certain. Because the ed here quantifies the risk of exceeding the design return level
remaining risk is an aggregation of all remaining years of the of the project as a function of time, which may be used to
project, the risk remains high throughout the planning horizon. determine the timing of changes necessary to accommodate
future sea-level rise. For example, if the future sea-level rise is
DISCUSSION observed to follow a particular scenario, planners and engi-
The approach presented here is general in that the neers can obtain a general sense of when a particular design
methodology allows one to relax some of the assumptions risk level would be exceeded and could plan up front for any
regarding linearity and nonstationarity without a major effort. retrofit or a replacement. In considering such a strategy, risk
Using tide gage data most parameters necessary for application tolerance can be incorporated.
of the method can be obtained. However, the parameters of the
GEV distribution may have significant uncertainties in case of CONCLUSIONS
short records. Gage records less than 60–70 years may not be Future sea level is an important constraint on the planning
sufficient to obtain reliable of estimates of long-term trends and retrofitting of coastal infrastructure. In view of the recent
primarily due to the presence of multidecadal variations. In focus on nonstationarity in climate (Milly et al., 2008) and
such cases, it may be possible to model the extremes using the associated sea-level regimes, past information from tide gauges
generalized Pareto distribution using the peaks-over-threshold may have limited utility. Planners and engineers need to
method (Coles, 2001; Mendez et al., 2006). An alternative accommodate the potential for accelerating sea-level rise using
approach is to use long-term gages in the region and transfer a scenario-based approach and consider future changes to the
information to a particular location of interest with necessary level of protection in the design of their project. As sea level and
adjustments for localized conditions. Furthermore, storm- climate dynamics continue to evolve in the coming decades, the
surge modeling may help identify the offset between the mean impact of extreme events on coastal infrastructure will require
sea level and the extremes. careful scrutiny. Synthesis of sea-level rise projections with
Because of the uncertainties in predicting future sea-level extreme value statistics of sea-level exceedences, as demon-
rise and more importantly the potential magnitude of expected strated in this paper, will allow the designer to explore changes
acceleration, a scenario-based approach is deemed appropriate. in future risk based on parameters determined by historical
Scenario selection for a particular project depends on many tide records. One of the primary conclusions conveyed by
factors and, in some cases such as projects near the coast application of the method is the significant (some might say
designed for low probability–high consequence events (e.g. a alarming) nonlinearity and reduction in extreme sea-level
nuclear power plant), consideration of a high sea-level rise return periods over time. Effectively, the time to failure is being
scenario may be warranted. In others cases, use of low and compressed in time in an exponential fashion. This reduction in

Journal of Coastal Research, Vol. 29, No. 1, 2013


Scenario-Based Projection of Extreme Sea Levels 7

design life and increase in risk is associated with a sensitive Milly, P.C.D.; Bettencourt, J.; Falkenmark, M.; Hirsch, R.M.;
dependence on the acceleration of the sea-level rise scenario. Kundezewicz, Z.W.; Lettenmaier, D.P., and Stouffer, R.J., 2008.
Stationarity is dead-Whither water management. Science, 319,
Future work may require the acceleration-based scenarios to
573–574.
be expressed in terms of probabilistic projections and then Nicholls, R.J.; Hanson, S.; Herweijer, C.; Patmore, N.; Hallegatte,
integrated into the proposed methods. S.; Corfee-Morlot, J.; Chateau, J., and Muir-Wood, R., 2008.
As a final comment, the concepts presented here may be used Ranking Port Cities with High Exposure and Vulnerability to
for others cases where nonstationarity associated with climate Climate Extremes: Exposure Estimates, OECD Environment
Working Papers, No. 1, OECD Publishing. doi: 10.1787/
change is present. A classic situation is the design of a flood
011766488208.
control project using the concepts of design floods with a NRC (National Research Council), 1987. Responding to Changes in
particular return period. Sea Level: Engineering Implications. Washington, D.C.: National
Academy.
LITERATURE CITED Obeysekera, J.; Irizarry, M.; Park, J.; Barnes, J., and Dessalegne, T.,
2010. Climate change and its implications for water resources
management in South Florida. Stochastic Environmental Research
Araújo, I. and Pugh, D.T., 2008. Sea levels at Newlyn 1915–2005:
& Risk Assessment, 25(4), 495–516.
analysis of trends for future flooding risks. Journal of Coastal
Obeysekera, J.; Park, J.; Irizarry-Ortiz, M.; Barnes, J., and Trimble,
Research, 24, 203–212.
P., 2012. Probabilistic Projection of Mean Sea Level and Coastal
Church, J.A. and White, N.J., 2011. Sea-level rise from the late 19th
Extremes. Journal of Port, Coastal, Waterway and Ocean Engi-
to the early 21st Century. Surveys in Geophysics, 32, 585–602.
neering. doi: 10.1061/(ASCE)WW.1943-5460.0000154.
Coles, S., 2001. An Introduction to Statistical Modeling of Extreme
Values. London: Springer-Verlag, 208p. ISBN 1-85233-459-2. Park, J.; Obeysekera, J.; Barnes, J.; Irizarry, M., and Park-Said, W.,
Gilleland, E. and Katz, R.W., 2011. New software to analyze how 2010. Climate links and variability of extreme sea level events at
extremes change over time. Eos, 92(2), 11 January 2011, 13–14. Key West, Pensacola, and Mayport Florida. ASCE Journal of Port,
Holgate, S.J. and Woodworth, P.L., 2004. Evidence for enhanced Coastal, Waterway and Ocean Engineering, 136(6), 350–356.
coastal sea level rise during the 1990s. Geophysical Research Peltier, W.R., 1999. Global sea level rise and glacial isostatic
Letters, 31:L07305, doi: 10.1029/2004GL019926. adjustment. Global and Planetary Change, 20, 93–123.
Horton, R.; Herweijer, C.; Rosenzweig, C.; Liu, J.; Gornitz, V., and Prandi, P.; Cazenave, A., and Becker, M., 2009. Is coastal mean sea
Ruane, A.C., 2008. Sea level rise projections for current generation level rising faster than the global mean? A comparison between
CGCMs based on the semi-empirical method. Geophysical Research tide gauges and satellite altimetry over 1993–2007. Geophysical
Letters, 35, L02715, doi: 10.1029/2007GL032486. Research Letters, 36(5).
Houston, J.R., 2012. Global sea level projections to 2100 using Rahmstorf, S., 2007. A semi-empirical approach to projecting future
methodology of the Intergovernmental Panel on Climate Change. sea-level rise. Science, 15(5810), 368–370.
Journal of Waterway, Port, Coastal, and Ocean Engineering doi: Ruggiero, P.; Komar, P.D., and Allan, J.C., 2010. Increasing wave
http://dx.doi.org/10.1061/(ASCE)WW.1943–5460.0000158, heights and extreme value projections: the wave climate of the U.S.
Houston, J.R. and Dean, R.G., 2011. Sea-level acceleration based on Pacific Northwest. Coastal Engineering, 57(5), 539–552.
U.S. tide gauges and extensions of previous global-gauge analyses. Sallinger, A.H.; Doran, K.S., and Howd. P.A., 2012. Hotspot of
Journal of Coastal Research, 27(3), 409–417. accelerated sea-level rise on the Atlantic coast of North America.
Hunter, J., 2010. Estimating sea-level extremes under conditions of Nature Climate Change, 24 June 2012. doi: 10.1038/
uncertain sea-level rise. Climatic Change, 99, 331–350. NCLIMATE1597.
IPCC (Intergovernmental Panel on Climate Change), 2007a. Climate Titus, J.G.; Hudgens, D.E.; Trescott, D.L.; Craghan, M.; Nukols,
Change 2007: The Physical Science Basis. In: Solomon, S.; Qin, D.; W.H.; Hershner, C.H.; Kassakian, J.M.; Linn, C.J.; Merritt, P.G.;
Manning, M.; Chen, Z.; Marquis, M.; Averyt, K.B.; Tignor, M., and McCue, T.M.; O’Connell, J.F.; Tanski, J., and Wang, J., 2009. State
Miller, H.L. (eds.), Contribution of Working Group I to the Fourth and local governments plan for development of most land
Assessment Report of the Intergovernmental Panel on Climate vulnerable to rising sea level along the US Atlantic coast.
Change. Cambridge, United Kingdom, and New York: Cambridge Environmental Research Letters, 4(044008), doi: 10.1088/
University Press. 1748–9326/4/4/044008.
IPCC, 2007b. Climate Change 2007: Synthesis Report. Intergovern- Titus, J.G. and Narayanan, V., 1995. The Probability of Sea Level Rise.
mental Panel on Climate Change. http://www.ipcc.ch/pdf/ U.S. Environmental Protection Agency, 186p. EPA 230-R95-008.
assessment-report/ar4/syr/ar4_syr.pdf; accessed on 26 September
Tomasin, A. and Pirazzoli, P.A., 2008. Extreme sea levels in the
2012.
English Channel: calibration of the joint probability method.
Jiang, Y.; Dixon, T.H., and Wdowinski, S., 2010. Accelerating uplift in
Journal of Coastal Research, 24(4C), 1–13.
the North Atlantic region as an indicator of ice loss. Nature
UHSLC (University of Hawaii Sea Level Center), 2011. University of
Geoscience, 3, 404–407.
Hawaii Sea Level Center. http://uhslc.soest.hawaii.edu; accessed on
Komar, P.D. and Allan, J.C., 2008. Increasing hurricane-generated
21 April 2011.
wave heights along the U.S. Atlantic Coast and their climate
controls. Journal of Coastal Research, 24(2), 479–488. USACE (U.S. Army Corps of Engineers), 2011. Sea-Level Change
Lowe, J.A.; Woodworth, P.L.; Knutson, T.; McDonald, R.; McInness, Considerations in Civil Works Programs. Department of the Army
K.L.; Woth, K.; von Storch, H.; Wolf, J.; Swail, V.; Bernier, N.B.; Engineering Circular No. 1165-2-212, 1 October 2011. U.S. Army
Gulev, S.; Horsburgh, K.J.; Unnikrishnan, A.S.; Hunter, J.R., and Corps of Engineers, CECW-CE, Washington, D.C., 20314-1000.
Weisse, R., 2010. Past and future changes in extreme sea levels and Vermeer, M. and Rahmstorf, S., 2009. Global sea level linked to global
waves, Chapter 11. In: Church, J.A.; Woodworth, P.L.; Aarup, T., temperature. PNAS, 106(51), 21527–21532.
and Wilson, W.S., (eds.), Understanding Sea-Level Rise and Watson, P.J., 2011. Is there evidence yet of acceleration in mean sea
Variability. Chichester, UK: Wiley-Blackwell, 375p. level rise around mainland Australia? Journal of Coastal Research,
Mendez, F.J.; Menéndez, M.; Luceń, A., and Losada, I.J., 2006. 27(2), 368–377.
Estimation of the long-term variability of extreme significant wave Woodworth, P.L. and Blackman, D.L., 2004. Evidence for systematic
height using a time-dependent Peak Over Threshold (POT) model. changes in extreme high waters since the mid-1970s. Journal of
Journal of Geophysical Research, 111, C07024. Climate, 17, 1190–1197.
Menéndez, M. and Woodworth, P.L., 2010. Changes in extreme high Zhang, K.; Douglas, B.C., and Leatherman, S.P., 2000. Twentieth-
water levels based on a quasi-global tide-gauge data set. Journal of century storm activity along the U.S. East Coast. Journal of
Geophysical Research, 115, C10011. Climate, 13, 1748–1761.

Journal of Coastal Research, Vol. 29, No. 1, 2013

You might also like