Aerts 2014 Science SM

www.sciencemag.
org/content/344/6183/473/suppl/DC1
Supplementary Materials for
Evaluating Flood Resilience Strategies for Coastal Megacities
Jeroen C. J. H. Aerts*, W. J. Wouter Botzen, Kerry Emanuel,

Ning Lin, Hans de Moel, Erwann Michel-Kerjan*
*Corresponding authors.
E-mail: jeroen.aerts@vu.nl and erwannmk@wharton.upenn.edu
Published 2 May 2014, Science 344, 473 (2014)

DOI: 10.1126/science.1248222
This PDF file includes
Materials and Methods

Supplementary Text
Figs. S1 to S6
Tables S1 to S38
References
1
Outline of Supplementary Material

1. Modelling Strategy ………………………………………………………………………..…………….2
2. Benefit-Cost Analysis Methodology ……………………………………………………………….…..18
3. Benefit-Cost Analysis (BCA) Results ………………………………………………………………..…20
4. Overview of Data Sources .…………………………………………………………………………..….49
Supplementary Text: Acknowledgments
References
1. Modelling Strategy
Figure S1 depicts the general modelling approach, details of which are provided below . A large set (549)
of synthetic hurricanes and their surge heights (1) has been produced by a coupled hurricane- (2) and
hydrodynamic model (ADCIRC) (3). These surge heights are used to derive flood inundation maps with
different return periods. Inundation depths are calculated using nearest neighbour interpolation of the surge
water levels at the modelling-nodes near the coastline (see Fig. S2). All inundation maps are subsequently
used to calculate flood damage, by employing a flood damage model based on the HAZUS-MH4
methodology (4). This generic model has been fine-tuned for New York City (NYC) using NYC specific
building type information (5). Hence, the HAZUS model produces damage calculations for all 549 storms.
In addition, damages from storms occurring in the future have also been calculated using future urban
growth projections resulting in ‘new land use’ with additional buildings.
Next, simulated damages for all storms with their probabilities are used to calculate expected annual
damage (EAD, or ‘flood risk’). This is calculated by plotting them on an exceedence-probability loss curve
and integrating the area below this curve (6). Inundation depths, flood damage and flood risk are calculated
for individual census blocks, the scale used by the HAZUS-MH model, but also aggregated to various other
spatial units (boroughs, FEMA zones) for analysis and presentation. The effect of climate change on future
flood risk has been assessed by adjusting the probabilities associated with the storms, based on scenarios
produced by different climate models and projections of sea level rise (see Section 1.3 of this SM).
Finally, this modelling framework is used to assess the effect on flood risk of different types of flood risk
management strategies: S1, building codes (elevating, dryproofing, wetproofing), S2 (storm surge barriers)
and S3, a hybrid solution consisting of building codes and local levees. Through these calculations, the
reduced risk can be calculated for each strategy S1,2,3, which is finally used as the ‘benefit’ in the Benefit
Cost Analysis (BCA). We now describe each one of these components along the loss curve in more detail.
Note that our analysis throughout the paper and online material does not account for possible loss of lives
nor injuries. As such our benefit calculation (loss avoided) is a lower bound. Including avoided value of
statistical life (VoL) and health care spending would increase the benefit of the proposed measures. That
said, this would most likely only affect our results marginally given the effectiveness of the early warning
system and media coverage in New York City. Integrating these two components (VoL; health care
spending) would certainly be important to consider in other cities where warning systems are not in place
and/or in places where people might not evacuate in advance of the flood, resulting in a large number of
casualties.
2
Fig. S1. Flood risk modelling and CBA approach
1.1 Hurricane Flood Hazard Simulations

We used 549 synthetic low-probability hurricanes surge events simulated by (1). This study generated a
large set of synthetic hurricane surge events for NYC with a coupled system composed of a statistical-
deterministic hurricane model and surge hydrodynamic models. The hurricane model (2) generates
thousands of synthetic tropical cyclones under given large-scale atmospheric and ocean environments for
various climate conditions; the synthetic storms for the current climate condition are in statistical
agreement with the (albeit limited) observations (2). The hydrodynamic model used is the Advanced
Circulation Model (ADCIRC) (3). From these thousands of potential hurricanes, we select 549 events that
have return periods, in terms of the surge height at the Battery in lower Manhattan, NYC, greater than 10
years.
Surge heights for the 549 storms generated by the coupled Hurricane-ADCIRC models are simulated at
numerous nodes in the NYC region (Fig. S2). From these surge heights, flood inundation depths are
estimated similar to (7), by nearest neighbour interpolation of the surge water levels at the modelling-nodes
near the coastline, and subtracting a high-resolution elevation model (2010 NYC LiDAR data). Using this
methodology, we implicitly assume that the first floor of buildings is situated at the ground level as denoted
by the elevation model. Inundation depths per census block, the level at which the analysis is performed,
are determined by taking the median water level in the census block. There are two significant novelties
though. First, in this study a large number of synthetic surge events (549) is used, with return periods
ranging from 1/10 per year to 1/15,000 per year (with respect to the water level at the Battery in lower
Manhattan). Second, we have also adjusted inundation levels around Jamaica Bay. The simulations by (1),
do not provide water levels inside the Bay. This adjustment is based on stillwater levels reported by (8).
Extra nodes were created in Jamaica Bay and adjusted water levels were assigned to these nodes based on
regression analysis with the closest nodes at the entrance of the bay, depending on the return period of the
water level.
3
Fig. S2. The number of storms that flood each of the census blocks in NYC (as a % of all storms). The dots
indicate the nodes for which the hurricane model provides water levels. The background colours denote the
elevation.
1.2 HAZUS-MH: Vulnerability and Risk Calculations

For each storm, damage is calculated using stage (depth)-damage curves, which represent the fraction of
damage to a building and its content based on the water level present in the census block. The damage
calculation uses the HAZUS-MH methodology (4, 9), which is the standard for flood risk assessments in
the US. However, the standard software was not used because: (i) automation of the procedure was
required given the large amount of calculations necessary for this study; (ii) the need to adjust the building
database to represent urban growth; and (iii) the need to adjust depth-damage curves to evaluate the effect
on flood risk of mitigation and protection measures.
The damage calculation uses the depth-damage curves and values at risk form HAZUS, but have been
coded into a set of Matlab® functions. HAZUS calculates damage to 33 different buildings types, including
11 types of residential buildings. Total values at risk for the buildings have been estimated per building
type, using the average square feet per building type, and the mean cost per square foot from HAZUS
[Table 14.1 in HAZUS technical manual (4)]. The value at risk per building type is given in Table S1.
Values for content and inventory in HAZUS are percentages of the value of the structure [Table 14.6 and
14.9 in HAZUS technical manual (4)]. HAZUS distinguishes damage to buildings with, and without,
basements. As such information was not readily available; we used damage curves that were weighted
averaged using the fraction of houses with, and without, basements (25% and 75% respectively). For new
buildings, it was assumed that they will not have basements, in line with the current planning policies of the
city. Furthermore, the building table of HAZUS, defining how many of each building type is present in
each census block in NYC, was not used. Rather, a more accurate NYC update of the building stock made
for the Office of Emergency Management of NYC (5) was used. This is an extensive analysis given the
high construction density in this mega-city (see Table S1 below).
4
For each storm, three damage calculations are performed: for damage in the V flood-zone (with high flow
velocities), damage in the A flood-zone (outside the V zone), and damage to vehicles. FEMA maps
(available in 2012, i.e., pre-Sandy) were used to identify census blocks in the A and V zones, as well as to
label census blocks located in the 1/100 and 1/500 flood zone (used to present the results). After damages
are calculated for all 549 storms, results are combined into a total aggregate risk level. First, probabilities
are linked to each storm based on the water level it generates at the Battery. Then the storms are re-ordered
in order of descending damage estimates, after which the cumulative probabilities are recalculated resulting
in exceedance probabilities of damage. The resulting EPL curves are subsequently used to calculate the
expected annual damage (EAD) as the area (integral) under the EPL curve (6).
5
Table S1. Building types considered in this study, including their value at risk and types of measures considered for each building type (where Y stands for yes
and N stands for no)
Measure implemented
Value at risk ($/building) Existing Buildings New Buildings**
Wetproof*
Wetproof*
Dryproof*
Dryproof*
Inventory
Structure
Elevation
Elevation
Content
Code Description
RES1 Single Family Dwelling 182492 91246 0 Y Y Y Y Y Y
RES2 Manuf. Housing 356421 178211 0 Y Y Y Y Y Y
RES3A Duplex 186630 93315 0 Y Y Y Y Y Y
RES3B Triplex / Quads 282573 141287 0 Y Y Y Y Y Y
RES3C Multi-dwellings (5 to 9 units) 873207 436604 0 N Y Y Y Y Y
RES3D Multi-dwellings (10 to 19 units) 1546338 773169 0 N Y Y Y Y Y
RES3E Multi-dwellings (20 to 49 units) 3714867 1857433 0 N Y Y Y Y Y
RES3F Multi-dwellings (50+ units) 18707261 9353631 0 N Y Y Y Y Y
RES4 Temporary Lodging 11991508 5995754 0 Y Y Y Y Y Y
RES5 Institutional Dormitory 7019251 3509626 0 Y Y Y Y Y Y
RES6 Nursing Home 8067754 4033877 0 Y Y Y Y Y Y
COM1 Retail Trade 571211 571211 41339 N Y Y N N N
COM2 Wholesale Trade 1466010 1466010 127395 N Y Y N N N
COM3 Personal and Repair Services 342290 342290 0 N Y Y N N N
COM4 Professional/Technical Services 11381918 11381918 0 N Y Y N N N
COM5 Banks 2370537 2370537 0 N Y Y N N N
COM6 Hospital 81553944 122330915 0 N Y Y N N N
COM7 Medical Office/Clinic 4904708 7357062 0 N Y Y N N N
COM8 Entertainment & Recreation 2260999 2260999 0 N Y Y N N N
COM9 Theaters 6015389 6015389 0 N Y Y N N N
COM10 Parking 180913 90457 0 N Y Y N N N
IND1 Heavy 1677769 2516654 585357 N N N N N N
IND2 Light 4095922 6143882 422805 N N N N N N
IND3 Food/Drugs/Chemicals 0 0 0 N N N N N N
IND4 Metals/Minerals Processing 0 0 0 N N N N N N
IND5 High Technology 0 0 0 N N N N N N
IND6 Construction 0 0 0 N N N N N N
AGR1 Agriculture 0 0 0 N N N N N N
REL1 Churches and Other Non-profit 1982996 1982996 0 N N N N N N
GOV1 General Services 18290701 18290701 0 N N N N N N
GOV2 Emergency Response 2874692 4312038 0 N N N N N N
EDU1 Grade Schools 7797433 7797433 0 N N N N N N
EDU2 Colleges/Universities 27175914 40763870 0 N N N N N N
*Not implemented in V-zone; **Assumed being built without basement.
6
1.3 Future Scenarios: Climate Change and Urban Growth
To explore how flood risk in NYC may change in the future, we use projections for climate change and
urban growth (‘land use change’, Fig. S3). For this Policy Forum article we look at target years towards the
2050s and 2080s, in line with the time horizons of NPCC (10). For population growth, we make use of
projections made for 2040 by the New York City Department of City Planning (NYC-DCP) (11). It is
assumed that population will be relatively stable after 2040, so it is assumed that population will remain
constant from that time onwards. The population projections are at the borough level (5 areas in NYC), and
were made spatially explicit to the level of Traffic Analysis Zones (TAZ − 1611 areas) using information
on current household distributions, the distance to subway stations and existing building and zoning plans
(11). Growth in households is assumed to follow these population growth rates per TAZ zone, and were in
turn translated into new buildings using borough differentiated percentages (as some new households will
settle in existing buildings) provided by NYC-DCP. Last, every census block was attributed a growth in
buildings similar to the TAZ zone it is located in, so we could determine the increase in the number of
buildings. Only residential buildings are assumed to increase in our projections, which make up 90% of the
existing building stock in NYC. Hence this is a reasonable assumption.
Future flood risk under climate change has been calculated using projections in storm frequency from four
global climate models (GCMs): CNRM CM3, ECHAM5, GFDL CM2.0, MIROC3.2 (1), under the IPCC
AR4 A1B emissions scenario, and three projections of sea-level rise (SLR): 30cm, 60cm and 100cm. These
SLR projections are in line with the high-end sea-level rise scenarios of the NPCC (10) for 2040 and 2080,
and the low-end rapid ice melt scenario (10). Based on this, three climate change scenarios have been
derived (low, middle and high) with two time horizons (2050s and 2080s) each. In terms of storm
frequency, prior calculations (1) included a future scenario for 2081-2100. In this study, we used their
adjusted storm climatologies for 2081-2100 for our 2080s scenario, and interpolated this effect to make
estimates for 2050s. With respect to SLR, it is shown that there is little non-linearity between SLR and
storm surges along the coast, meaning SLR can be added linearly on top of a storm surge. In this study, we
assume the linearity for the entire region and add the different SLR estimates on top of the surge tide water
levels for the different scenarios and time horizons.
It has been observed that ECHAM and MIROC show little change in return periods of surge levels (1). This
implies that changes in risk using the ECHAM and MIROC calculations mainly relate to sea-level rise. The
GFDL climatology indicates the largest increase in storm frequency and resulted in the highest increase in
surge risk. The low and middle climate change scenarios in this study have therefore been based on the
ECHAM climatology (low scenario, hardly any change in storm frequency) and GFDL climatology
(middle scenario, including a considerable increase in storm frequency), respectively; both use 30cm SLR
for the 2050s, and 60cm SLR for the 2080s. For the high climate change scenario, the GFDL climatology
has also been used, but in combination with a SLR of 100cm.
To summarize, the low scenarios reflect mainly the effect of sea-level rise, the middle scenarios reflect the
combined effect of sea-level rise and increased storm activity, and the high scenario reflects the combined
effect of increased storm activity and sea-level rise considering rapid ice melt. The effect of climate change
is included in the model by changing the probabilities associated with the 549 storms based on the flood
return levels for the Battery. Based on the calculations of (1), we estimate how probabilities change for
each of the scenarios and time horizons.
7
1.4 Enforcing Building Codes
Various types of building-level measures to reduce flood risk were used in the model by changing the
stage-damage curves (Table S1). The measures we considered are: elevation of the buildings, dryproofing
(sealing the building so no water can come in), and wetproofing (adjusting the interior and material of the
building so damage is limited when flooded). Each measure is evaluated for its effect on flood risk when
applied to 2 feet, 4 feet, and 6 feet above the surface ground level. Moreover, for dryproofing and
wetproofing, high and low estimates of the effect of these measures are used to illustrate the uncertainty
inherent in the effect of these measures. The effectiveness of dryproofing and wetproofing is estimated
using flood damage reduction factors found in international literature (12,13,14). For dryproofing reduction
factors of 87.5% and 75% were used, and for wetproofing reduction factors of 50% and 30%. Not every
measure can be implemented at every location, though. In our model, we assumed that existing multi-
dwelling apartment and commercial buildings cannot be elevated.
Dryproofing and wetproofing has been evaluated for all residential and commercial buildings, but not for
the other types of buildings (industry, agriculture, religion, government, education). For new buildings,
measures are only implemented in the model for residential buildings, as the population growth scenarios
mainly target residential buildings. For both existing and new buildings, the measures dryproofing,
wetproofing and the combination of elevation with wetproofing are not implemented in the flood-zone with
high flow velocities (V-zone).
1.5 Flood Risk Estimates
Our results estimate current flood risk to buildings and vehicles in NYC at approximately $71 million per
year. Of this risk, about 6% is due to damage to vehicles. The rest of this risk is to buildings, of which 15%
is to buildings in so-called V-zones. Figure S3 illustrates how this risk is spatially distributed over the city.
Especially the area around Jamaica Bay has high risks, with 37% of the total risk ($26 million/year) in
Queens (which includes the Rockaways), and 26% ($18 million/year) in Brooklyn. In terms of flood zones,
most of the risk comes from the FEMA 1/100 flood zone (89%), but also 4% results from areas outside the
FEMA 1-in-100 and 1-in-500 year return flood zones. This risk is the integral of many different storms
with various magnitudes. When considering the damage to buildings and vehicles from single storms, our
calculations indicate that a damage of $0.9 billion has a return period of 100 years, and damage
corresponding to a return period of 1,000 years is about $10 billion (graph in Fig. S3).
Our damage estimates can be compared to the damage Sandy caused to NYC for validation. Estimates of
recovery costs made by NY State (15) estimate damage to NYC at roughly $15 billion. Of this, $4.7 billion
has been attributed to housing ($4.5 billion to business, $2.8 billion to hospitals) (16). The exact return
period of an unprecedented event like Sandy is difficult to pinpoint as it depends on the specific location
and indicator one looks at. The return period of the storm tracks has been estimated at ~700 years (17), but
when looking at observations of the water level at the Battery, a return period of at least 1,000 years has
been calculated (18).
The hurricane simulations on which our analysis is based (1) also yield a return period of around 500 to
1000 years for the water level reached at the Battery by Sandy. When looking at our damage and risk
estimates, about 41% of the damage and risk can be attributed to residential housing (34% to commercial,
18% to other buildings and 6% to vehicles). The simulated 1/1000 per year damage of $10.3 billion thus
corresponds to about $4.2 billion dollar damage to housing in NYC, which is very much in line with the
actual damage Sandy caused to housing in NYC. These results indicate that our modelling framework
simulates the effect of storms and their resulting damage and risk to buildings in NYC very well.
8
Fig. S3. Map of NYC with its five boroughs and the risk to buildings and vehicles as estimated in this study.
Additionally, the amount of residential buildings and the total amount of buildings is given per borough
and the city as a whole. The numbers between brackets denote the change in risk and residential buildings
resulting from the population growth projections of NYC.
9
1.6 Increasing the Scale and Nature of Risk Estimates
Our risk estimates from HAZUS only pertain to buildings and vehicles. In order to also address other
damage categories, such as indirect economic damages (e.g., business interruption), and direct damage to
other assets (such as infrastructure), we use scaling factors to scale up the risk simulations by HAZUS. This
has been done using the detailed damage assessments by the U.S. government after Sandy (19), in which
indirect damages have been estimated at one-third of total direct damage. Using (19), the ratio of housing
damage to total damage in NYC (including indirect damage) has been estimated (a factor 6). This factor
has been applied to our risk estimates, assuming a constant ratio between damage and risk. In this way, our
calculated risk to buildings and vehicles of $71 mln/yr (of which $29mln/yr for buildings) can be scaled up
to $174 million/year.
Furthermore, the geographical scale of the measures considered in this study is larger than only NYC, as
the barriers would also protect parts of New Jersey (NJ). Correspondingly, the risk (reduction) in these
areas should also be considered as benefit. To include the parts of NJ that would also be protected by the
strategies considered here, we used the housing assistance data collected by FEMA in the aftermath of
hurricane Sandy (20). The fraction of the damage to housing in NYC, as compared to parts of NJ has been
derived from this database. The results reveal that the area in NJ that would be protected by the NJ-NY
connect barrier strategy (S2c) suffered in total roughly one third of the damage in NYC. For the other two
barrier strategies this is roughly one sixth.
Applying these numbers to the total risk estimated above, new risk estimates are $203 million/year and
$232 million/year for the areas protected by strategies S2a,b and S2c, respectively (19).
1.7 Future Scenario Risk Estimates
Using the risk modelling framework, we also estimated future flood risk for NYC, including the effect of
population growth and climate change for time horizons in the 2050s and 2080s. Projected population
growth alone is expected to increase the flood risk by 15% as a result of increased buildings in flood prone
zones. In addition, climate change will also have an adverse effect on flood risk in the NYC region.
Potential future changes in wealth could have important effects on future flood risk. These are, however,
not included in our analysis because of a lack of reliable projections of how potential wealth changes will
affect future flood exposure and vulnerability in NYC.
The effect of climate change is substantially larger as compared to the effect of population growth, but
there is considerable uncertainty associated with the effect of climate. Therefore, in this study we use three
scenarios (see section 1.3). The low climate change estimate (ECHAM) shows little change in storm
frequency, and as a result this estimate can be interpreted as the adverse effect of only sea-level rise. The
middle scenario (GFDL climatology), on the other hand, does show a substantial increase in storm activity,
and can thus be interpreted as the combined effect of sea-level rise and increased storminess. The high
scenario considers the same increase in storm frequency, but a higher sea-level rise in 2080 of 100cm, due
to rapid ice melting.
The results (Fig. S4) show that the projection for the 2050s yields an increase in flood risk of 80% to 480%,
implying almost a doubling of risk due to 30cm sea-level rise, which could be aggravated by increased
storm frequencies, causing an increase of almost five times. By the 2080s, flood risk may have risen by
166% (low scenario, mainly sea-level rise) to 1189% (middle scenario, sea-level rise plus increase in storm
activity) of the current risk. These projected increases in risk can mainly be attributed to climatic change,
making it a genuine threat to the NYC region. Under the high scenario (with rapid ice melt, 100cm sea-
level rise by 2080), the adverse effect of climate change on flood risk is even more substantial, yielding
increases in flood risk by the 2080s up to 2303%. Note that the increases in risk as calculated here relate to
a situation where no action is being taken to reduce risk, even if sea level would rise by 100cm.
10
Fig. S4. Projections of future flood risk to buildings and vehicles in NYC for the 2050s and 2080s and the
components which make up this future risk. The two bars for 2050s and 2080s correspond to the low
(mainly sea-level rise) and middle (sea-level rise plus increased storm activity) climate change scenarios.
Note that the high scenario is not given in this graph because its risk in 2080s is too large to display
(~2400% of current risk).
1.8 Effect of Damage-Reducing Building Codes
One way of reducing flood risk is to make buildings more flood proof by following building code
recommendations (elevating, dryproofing, or wetproofing the building up to a certain height). The effect of
such measures was evaluated here using the flood risk model for the existing building stock, and for the
projected new buildings. The reduction in risk to buildings is illustrated in Table S2 as a percentage of the
total risk to buildings and vehicles (in case of the existing building stock), and as a percentage of the
increase in risk due to new buildings. The results illustrate that building codes can result in a substantial
reduction in the flood risk of buildings in the NYC area. Depending on the height a measure is
implemented, elevating residential buildings (except existing multi-dwelling buildings for which this is not
considered possible) would reduce the flood risk up to 29%.
Wetproofing all residential and commercial buildings could reduce risks up to 30% and dryproofing could
reduce it by 42% to 49%, depending on how effective this measure is assumed to be (Table S2).
Considering the vast existing building stock and difficulties in enforcing regulations on existing buildings,
we recognize that it may prove to be difficult and costly to flood proof all existing buildings. For new
buildings, however, new legislation could very well require all new buildings to be flood proofed. By doing
so, the increase in risk that would autonomously result from the increase in exposure would actually be
reduced considerably. If no measure would be taken, the increase of new buildings would result in an
increase in risk of $10.7 million/year (Table S2). By applying strict building codes on new developments,
almost the entire increase in risk can be averted. Elevating every new building 4 feet (above ground floor)
would already go a long way and reduce the extra flood risk by almost 90%. Also wet- and dryproofing of
new buildings would greatly reduce the increase in risk.
11
Table S2. The effect of various building-level damage-reducing measures when realized on all existing
buildings (2nd column), and all new buildings as projected in this study (3rd column). The ranges
represent low and high effects of that specific measure. Note that existing multi-story buildings have not
been elevated in the simulations, resulting in a lower total decrease as compared to dryproofing.
Decrease in risk(%) Decrease in risk (%)

Measure existing buildings new buildings
Risk without $70.8 million/year $10.7 million/year
measures
Elevation 2 feet 18% 65%

Dryproof 2 feet 19 - 23% 33 - 39%
Dryproof 4 feet 34 - 39% 58 - 68%
Dryproof 6 feet 42 - 49% 68 - 79%
Wetproof 2 feet 9 - 15% 16 - 27%
Wetproof 4 feet 14 - 24% 24 - 41%
Wetproof 6 feet 17 - 29% 28 - 46%
1.9 Addressing Uncertainties
In a modelling framework as the one used here, there is inherently uncertainty present in the various
components. Uncertainty in terms of effectiveness of building codes and climate change are addressed in
different calculations and scenarios. Note, however, that some BCA categories exist which have not been
addressed in this paper. For example, costs may arise from ecological damage from different measures. But
also additional benefits may result from installing storm surge barriers, such as enhanced economic benefits
through connecting NJ and NY (Sandy Hook – Rockaways).
There are, however, more sources of uncertainty that can potentially affect risk estimates considerably and
thereby the expected benefits of measures. Two key uncertainties in flood risk assessments are related to
the water levels associated with a certain probability and the damage calculation (21, 22). Uncertainty in
the water level corresponding to a certain probability has been considered by (1), who report confidence
intervals for water levels at the battery to address the statistical uncertainty. Using the statistical confidence
intervals, risk has also been calculated using an upper (+2 standard deviations) and lower (-2 standard
deviations) confidence boundary, which reflects the 95% confidence interval. The results show that
uncertainty in water levels at certain probabilities results in a range around the resulting risk estimate from -
22% to +17%.
Furthermore, uncertainty in the damage and risk calculation can have a substantial impact on the results
(489). Therefore, in (7), three sets of damage curves and corresponding maximum damages were used to
calculate flood risk to buildings in NYC. Their results indicate a range in estimated risk ranging from -36%
up to +40% (7). It is assumed that these low and high estimates make up the entire uncertainty in the
damage curve. Uncertainty in damage curves and hydraulic boundary conditions are two of the most
important uncertainties in damage modeling (21,22) and as these two sources of uncertainty are completely
uncorrelated, their effects can be added up to derive uncertainty in the damage and risk estimate. This
results in a range for the risk estimate from -49% up to +63% for their combined effect. This range related
to uncertainty in the damage and risk estimation has been used in the BCA analysis to illustrate whether,
and how, this uncertainty would affect the evaluation of the different flood risk management strategies.
12
1.10 Flood Protection Strategies
Strategy 2a ‘Environmental dynamics’ (left, Fig. S5), is designed such that the lower NY Bay will remain
open, and the tidal currents of Jamaica Bay Inlet will not be disturbed by a storm surge barrier. Three storm
surge barriers will be installed at: Arthur Kill, Verrazano Narrows, and East River. In Strategy 2b ‘Bay
closed’ (right, Fig. S5), it is perceived that it will no longer be viable to maintain the Jamaica Bay wetlands
and its salt marshes. Hence, the large-scale protection works and marshland-stabilizing activities in Jamaica
Bay as described in S2a will be cancelled and, instead, an additional storm surge barrier will be installed
across the Jamaica Bay Inlet.
Strategy S2c (left, Fig. S6) reduces the length of the coastline of the NYC-NJ area as much as possible, to
minimize flood protection costs. Two storm surge barriers are developed: one large barrier that connects
Sandy Hook in NJ and the tip of the Rockaways in Queens, NY and a barrier in the East River. Some lower
spots (bulkheads, levees, or landfill) on the inside of the protection system will be elevated to accommodate
rising water levels caused by Hudson River peak discharges during a storm event. Strategy S3 (right, Fig.
S6) combines cost effective flood-proofing measures with local protection measures of critical
infrastructure. Such a ‘hybrid solution’ aims at keeping options open: either (a) building codes can be
further enhanced in the future with additional local protection measures or (b) storm surge barriers can be
developed.
13
Fig. S5. Strategy 2a‘Environmental dynamics’ (left) and Strategy 2b‘Bay closed’ (right)
14
Fig. S6. Strategy 2c ‘NJ-NY connect’ (left) and Strategy 3 ‘hybrid solution’ (right)
15
1.11 Event Damage versus Expected Annual Damage (‘risk’)
Comparing annual expected damages (‘risk’) with damages from a single catastrophic event is not
straightforward and can often be confusing because these differ by several orders of magnitude, simply
reflecting that large-scale disasters have a very low probability of occurring. It is, therefore, important to
note the difference between overall losses from (i) a single event (e.g., $62billion from Sandy) and (ii) the
annual expected damage (‘risk’), which is the integration of ‘probability and damage’, for all possible
storms. More precisely, the $174 million/year flood risk for NYC is the integral of many different damages
(549 storms) each having different probabilities. This risk number has been used in this study as is common
in BCA studies of this type. Obviously, that risk number is much lower than the total damage for a single
extreme event, since probability is included in calculating risk and many of these storms have actually an
extremely low probability of occurring on a yearly basis.
Damage figures for individual low-probability events such as Sandy can also be estimated. This is done in
the Table S3 for 8 different events (with return periods varying from 1/100 to 1/10,000). The first column
provides different return periods of damages to public and private buildings as well as vehicles in NYC as
calculated by our detailed HAZUS model (second column). These damage estimates are then scaled up to
reflect total damage in NYC, and the NY/NJ region (third and fourth column). The bottom row shows the
expected annual damage (i.e., ‘risk’, the integration of individual events). These figures show substantial
damages, in the order of several tens of billions of dollars, for low-probability events, in the same order of
magnitude as Sandy. Note that the last column here represents damages for the whole area of the states of
NJ and NY affected by Sandy, whereas we only used the damages and risk numbers for the areas protected
by the strategies S1,2,3, which is a smaller area, with a risk estimate of $203-232 million/year depending on
the areas considered under each strategy.
When comparing the single event damage from Sandy provided by the U.S. government with the (scaled)
damage estimates by our model, we see that the $21 billion for NYC and the $62 billion for NY/NJ
provided by the U.S. government are a bit below our damage estimates for a once-in-750 year event
(approximately the probability of Sandy).
Our analysis reveals a return period of about 660 years for the level of damage caused by Sandy. This is
very well in line with estimates of the probability of Sandy. Estimation of the water level of Sandy in our
own dataset of water levels yielded a return period of roughly 500 years, which is close to the estimated
return period of about 700 years for the storm track of Sandy by Hall and Sobel (17). So overall, our
damage and risk estimates align very well with the observations of Sandy, providing confidence in the
performance of our risk model.
16
Table S3. Individual event damages for different probabilities as estimated by our detailed model, and
scaled up to total damage in NYC and total damage in the NY/NJ region considered in this study.
Damage Buildings + Full Estimated Damage Full Estimated Damage
Return Period Vehicles NYC NYC NY/NJ
($ billion) ($ billion) ($ billion)
100 0.9 2.2 6.5
250 4.2 10.3 30.2
500 7.6 18.6 54.4
750 9.2 22.7 66.1
1000 10.3 25.4 74.0
2000 15.0 37.0 108.0
5000 18.8 46.3 135.0
10000 25.2 62.1 181.2
Risk (Expected Annual

Damage, EAD; 70.8 174 508
in $ million per year)
17
2. Benefit-Cost Analysis Methodology
2.1 General Benefit-Cost Analysis Method
The different flood risk management strategies S1,2,3 are assessed using BCA (Fig. S1). In particular, for
each strategy the Benefit/Cost Ratio (B/C ratio) and the Net Present Value (NPV) are estimated. Formally,
where Bt is the benefit of a flood risk management strategy in year t and Ct its cost, r is the discount rate,
and the investment horizon is T years. The benefit in year t in this context is the avoided flood damage in
year t, and the cost includes the initial investments or construction costs and yearly maintenance costs. A
positive NPV, thus, indicates that the sum of the discounted benefits exceeds the sum of the discounted
costs over time, which implies that a strategy is beneficial in economic terms. A related indicator of
economic efficiency of a flood risk management strategy is the B/C ratio:
∑ ∑ .
If NPV is positive, then the B/C ratio is higher than 1. Both indicators are provided here since while the
B/C ratio shows the economic efficiency in terms of relative benefits per dollar invested in a strategy, the
NPV provides the amount of net economic benefits that a strategy generates. All BCAs are conducted over
an investment horizon which starts in year 2015 (t=0) and ends after 100 years (T=100) for the building
code measures and after 100 or 150 years for the storm surge barriers strategies. All BCAs are conducted
for the expected reduced flood risk by the strategy, and for the uncertainty interval reported in Section 1.9
which reflects both uncertainty of the water level as well as of the damage calculation.
An important variable is the discount rate r which reflects the opportunity costs of public investments, and
is an important parameter in BCAs of projects with a long time horizon. It should be realized that the
opportunity costs of public investments is often considered higher than a social discount rate which is only
based on the weighting of welfare values of incomes at different points in time. The discount rate is an
uncertain exogenously determined parameter. A broad range of discount rate values has been used in BCAs
of investments by the public sector. All BCAs are here conducted using the recommended discount rate by
(23) of 7% (r=0.07). However, this high discount rate may be debated for the use of flood risk management
strategies with a very long time horizon since it has the effect that monetary values far in the future receive
a very small weight in estimating their present value (and the NPV). Moreover, an implicit assumption
behind using a high discount rate of 7% is that productivity growth in the (far) future − and the opportunity
costs of capital − is high, while, in fact, this is uncertain. Therefore, a recent high level expert panel of
economists has advised the U.S. Environmental Protection Agency to apply a discount rate that is lower,
and perhaps even declining over time, for BCAs of public investments with long-term benefits and costs
(24). However, this panel did not agree yet on an exact value of the discount rate. Therefore, we apply as a
lower bound the discount rate of 4% (r=0.04) used by the Netherlands government for long-term projects
which reduce societal risks, such as public investments in flood risk management and climate change
policy. It should be noted that, although we treat the discount rate of r=0.04 as a lower bound, in practice
U.S. agencies have used even lower values. For instance, the U.S. Interagency Working Group on the
Social Cost of Carbon adopted a lower value of 2.5%, while the Office of Management and Budget at the
White House (OMB) recommends using a range of 7% and 3% for BCA used for regulation purposes.1 In
comparison, our lower bound of 4% may be regarded as being conservative.
1
The final report, “Social Cost of Carbon for Regulatory Impact Analysis” can be found at
http://go.usa.gov/3fH
18
2.2 BCAs of Building Code Strategies
The BCAs of building code strategies for S1 and part of S3 pertain to three main categories of measures –
elevation, wet floodproofing, and dry floodproofing – which are an enhancement of actual building codes
policies in NYC (25). We explore the costs and benefits of the application of each measure for 2 feet, 4
feet, and 6 feet above the current height of existing buildings. A distinction is made for each strategy
whether it applies to existing buildings or only new residential buildings. For the BCAs it is assumed that
the building stock grows linearly until year 2040 (see Section 1.7). This implies that every year the same
number of new homes is built according to the building codes, which results in a constant yearly cost until
2040 and an annual flood risk reduction, which becomes larger over time until 2040 all new homes are built
and flood-proofed. It should be noted that the measures are not applied to all building types since certain
building types are unsuitable for a particular measure (see Section 1.4). Applying building codes to existing
buildings results in an investment cost in the first period and benefits in terms of annual flood risk
reduction in the subsequent periods in all BCAs that are conducted for existing buildings. Nevertheless, wet
floodproofing of existing buildings may in practice be best undertaken during planned reconstructions of
the house (26).
A further distinction is made between application of the measures in the 1/100 and 1/500 FEMA flood
zones, based on the maps that were available in 2012. Current building code policies and insurance
regulations pertain only to the 1 in-100-year flood zone. But, because of sea level rise and possible changes
in storm surge climatology, the area of the current 1-in-100-year flood zone in NYC is expected to increase.
As a result, the current 1 in-500-year flood zone may become the future 1-100-year flood zone (25). For
this reason, it may be considered prudent to apply building code regulations already in the current 1-in-500
year flood zone. Elevation is applicable to both 1 in-100 A and V zones, while wet and dry floodproofing is
analyzed only for application in A zones since these stand-alone measures are less effective to cope with
high velocity waves in V zones, especially if flood depths are high (25).
The estimation of the costs of the building code strategies is discussed in detail in (19). These cost
estimations are based on the number and types of buildings in the NYC flood zone and costs of elevating
them, and wet- and dry floodproofing of buildings from FEMA studies. The annual risk reductions that
these measures deliver have been estimated by the flood risk model described in Section 1. Potential
damage reductions from wet- and dry flood-proofing are particularly uncertain, which is why lower and
upper bounds are used for estimating their effectiveness and resulting risk reduction.
2.3 BCAs of Hybrid (S3) and Storm Surge Barriers Strategies (S2a,b,c)
The hybrid strategy S3 applies building code measures in the 1-in-100 year flood zone– namely the
elevation of new houses (4 feet in A-zone, 6 feet in V-zone) and wet floodproofing (2 feet) of existing
buildings – as well as local protection measures (19). The costs of the storm surge barrier strategies S2a,b,c
include the costs of installing the large storm surge barriers and additional flood protection measures as
well as their annual operational and maintenance costs. The costs of these protective measures (both
barriers and local protection) are estimated in (19), based on engineering studies, expert interviews and
assessments of construction costs of international storm surge barriers. Moreover, (19) discusses cross
border issues which are associated with flood protection strategies that are implemented in different state,
county and local jurisdictions, which are not repeated here. The reduced risk of all protective measures is
assumed to be the total risk behind these local protection structures, assuming that these structures offer
complete protection and do not fail. In the BCAs, it is assumed that the initial investment costs occur in the
first period while operational and maintenance costs occur in all subsequent periods. We use the midpoint
of the range of total NYC-NJ investment costs reported in Table 1 of the main paper. The annual risk
reduction is achieved after the barrier and local protection systems are installed. These benefits are based
on the total reduced risk in both NYC and NJ. This is done by firstly simulating the current annual
prevented flood risk to buildings in NYC of $71 million/year (Section 1.5). This number is further up-
scaled in Section 1.6 to include additional damage savings to sectors other than buildings, indirect
economic damage, and damage savings in parts of NJ that are also protected by the strategy. The baseline
total prevented current flood risk is $197million/yr for the levees in S3, $203million/year for storm surge
barriers S2a,b and $232million/year for storm surge barrier S2c.
19
2.4. The Inclusion of Future Trends in The Estimation of Risk Reduction: Climate Change and Urban
Growth
The risk reduction of all flood risk management strategies has been estimated under current conditions as
well as future scenarios. Future scenarios include one projection of urban growth (new buildings) and
different climate change projections. All BCAs are conducted under the assumptions of low, middle and
high (rapid ice melt) scenarios due to climate change as well as a socio-economic scenario depicting the
increase in new buildings in the coming decades, as discussed in Section 1.7. These projections are of
interest, since the benefits of flood risk management in terms of avoided annual flood risk are higher if, for
example, climate change increases the probability of damaging floods. Both the low and middle climate
change scenarios use the same assumption for sea level rise, namely 30cm in 2050, and 60cm in 2080,
following the upper boundaries of the NPCC sea level rise projections for New York (Section 1.7). The low
and middle climate change scenarios differ in how the frequency of storms may change under climate
change. Moreover, as an upper bound of climate change effects, flood risk has been estimated under a high
scenario which entails sea level rise of 100cm in 2080 (10), in addition to the increase in storminess. This
scenario has only been used as an upper bound of climate change effects in the BCAs of the storm surge
barriers strategies. While this flood risk model provides estimates of annual flood risk reduction under
climate change for the years 2050 and 2080 (Section 1.7), the BCAs assume that the risk reduction benefits,
between periods 2015-2050, 2050-2080 and afterwards can be calculated by assuming linear growth
between these time intervals.
3. Benefit-Cost Analysis (BCA) Results
Table T4 summarizes the main CBA results, which will be discussed in detail in the following sub-sections.
It shows the initial investment costs of the flood management strategies S2a,b,c and S3, which include costs
for barriers and local levees and beach nourishment for NYC and NJ. The overall combined investment
costs for NYC and NJ are maximum $14.7 to $23.8 billion for S2a,b,c and $11.6 billion for S3 (19).
Furthermore, yearly maintenance costs are higher for the barrier strategies than for S3. The costs of
implementing building codes for S1 depend on the type of measure and the area considered (19). From this
family of S1 strategies, the most efficient ones are incorporated in S3 (Section 3.2).
Under current climate conditions, none of the barrier strategies, nor the hybrid strategy have a B/C ratio of
>1 (Table T4). The only strategy for which benefits exceed costs revolves around the elevation of new
buildings in flood zones. The results of the BCA for the various measures of S1 (Section 3.1) show that for
newly constructed homes, elevation always results in a higher NPV than wet or dry flood-proofing. In
particular, elevating new buildings +6 feet in the 1/100 V zones has a B/C ratio of >1 under both discount
rates if the most likely risk estimates are used (Section 3.1). Moreover, elevating new buildings +4 feet in
1/100 A zones has a B/C ratio based on the most likely risk estimates larger than 1 if a low (4%) discount
rate is adopted, while +2 feet elevation is cost-effective in both discount scenarios.
Under a low climate change scenario none of S2a,b,c nor S3 has a B/C ratio > 1 (Section 3.2). S3 shows the
highest NPV and B/C ratio (0.26 and 0.45 for 7% and 4% discount rates, respectively). Assuming low
climate change conditions, wet floodproofing existing buildings in the 1/100 A zones up to +2 feet has a
B/C ratio under low discounting conditions (Section 3.1), in addition to the measures of S1 that are already
cost effective under current climate conditions.
Under the middle climate change scenario and a high discount rate (7%), S3 and S2c are the only strategies
that are cost effective, according to the most likely risk estimates. When a low 4% discount rate is
considered, all strategies make good economic sense, but only S3 and S2c have a B/C ratio >1 for the whole
uncertainty interval. S3 has the highest B/C ratio of 2.45, though S2c comes close with 2.24 (Table T4).
20
Table S4. Summary of total costs and main BCA results of flood management strategies S2a,b,c and S3
S2a Environmental S2b S2c S3 Proposed
Dynamics Bay closed NJ-NY connect Hybrid Solution a
Total initial investment cost NYC $16.9 – 21.1bn $15.9 – 21.8bn $11.0 - 14.7bn $6.4 - 7.6bn
Total initial investment cost NJ $2bn $2bn n.a. $4bn
Total initial investment cost NJ-NYC $18.9 – 23.1bn $17.9 – 23.8bn $11.0 – 14.7bn $10.4 – 11.6bn
Maintenance cost protection NJ-NYC 98.5mln/yr 126mln/yr 117.5mln/yr $13.5mln/yr
Current climate: Current climate: Current climate: Current climate:
Annual risk reduction
$203mln/yr $203mln/yr $232mln/yr $197mln/yr
(uncertainty interval in parentheses) b
(103; 331mln/yr) (103; 331mln/yr) (117; 379mln/yr) (100; 322mln/yr)
$-19.5bn $-19.8bn $-11.2bn $-8.5bn

NPV using a using high (7%) and low
(-20.9; -17.7bn) (-21.2; -18.1bn) (-12.9; -9.1bn) (-10.0; -6.6bn )
(4%) discount rate
$-18.4bn $-19.0bn $-10.0bn $-6.3bn
(-20.9; -15.3bn) (-21.4; -18.8bn) (-12.9 -6.4bn) (-9.1; -3.2bn)
B/C ratio using high (7%) and low (4%)

0.13 (0.07; 0.21) 0.12 (0.07; 0.20) 0.23 (0.12; 0.37) 0.26 (0.13; 0.43)
discount rate
0.21 (0.11; 0.35) 0.21 (0.11; 0.34) 0.36 (0.18; 0.59) 0.45 (0.23; 0.73)
Middle climate change Middle climate change Middle climate Middle climate
scenario: c scenario: c change scenario: c change scenario: c
$1.1bn
NPV using a high (7%) and low (4%) $-8.9bn $-9.1bn $0.9bn
(-5.2; 9.0bn)
discount rate. (-15.6; -0.4bn) (-15.9; -0.6bn) (-6.7; 10.7bn)
$16.9bn
(uncertainty interval in parentheses) b $7.5bn $7.0bn $19.3bn
(2.8; 35.0bn)
(-7.8; 27.0bn) (-8.3; 26.5bn) (2.1; 41.9bn)
B/C ratio using a high (7%) and low (4%)
0.60 (0.30; 0.98) 0.60 (0.30; 0.97) 1.06 (0.54; 1.74) 1.09 (0.55; 1.78)
discount rate
1.32 (0.67; 2.16) 1.29 (0.65; 2.11) 2.24 (1.14; 3.67) 2.45 (1.24; 4.00)
Notes n.a. stands for not applicable; B/C ratio stands for benefit cost ratio; if B/C ratio> 1, then the measure is cost effective.
a
BCA results are shown for the scenario of high effectiveness of wet flood-proofing.
b
Range is due to modeling uncertainty (Section 1.9).
3.1 Results of Building Code Policies
The BCA results of the building code policies are shown in Tables S5-S22. Results are shown in separate
tables per measure, and furthermore a distinction is made whether the measure is applied to existing or new
buildings. Different tables show results that are derived using the most likely estimates of reduced risk or
the lower, or higher, estimates which reflect the uncertainty interval presented in Section 1.9.
The BCA results show that the elevation of existing buildings is not cost-effective under any of the most
likely risk estimates and climate conditions studied here (Table S5). The reason is that the costs of lifting
existing houses are very high (26) and apparently cannot be recouped by damage savings over time. Only
the middle climate change scenario and higher bound risk estimates could justify elevation of existing
buildings in the 1/100 zone (Table S7).
Elevating new houses in V zones is cost-effective for all heights, discount rates, and climate change
scenarios considered using the most likely risk estimates (Table S8). Moreover, under current climate
conditions, elevating new houses in 1/100 A zones is cost-effective for + 2 feet elevation, while it is also
cost-effective for + 4 feet elevation if the low discount rate is applied (Table S8). Higher elevation is
economically feasible if climate change occurs. Elevating new buildings in the 1/500 year flood zone is
only cost-effective under the middle climate change scenario and use of a low discount rate, and is never
cost effective under current climate conditions or the low climate change scenario, not even when the
higher bound of risk is used (Table S10).
Wet floodproofing existing buildings is not economically feasible under current climate conditions, but 2
feet and 4 feet of wet floodproofing in the 1/100 year flood zone have a B/C ratio >1 under the scenario of
high effectiveness and low climate change and low discounting if the most likely risk estimates are used
(Table S11). Wet floodproofing up to 6 feet has a B/C ratios >1 under the middle climate change scenario
21
(Table S11), even if the lower bound of risk estimates is used (Table S12), while using the upper bound
estimates could justify this already in the low climate change scenario (Table S13). A similar picture
emerges for wet floodproofing of new buildings (Tables S14-S16).
Dry floodproofing existing buildings is not cost-effective under current climate conditions (Table S17),
except if a low discount rate and the higher bound risk estimates are used (Table S19). Dry floodproofing 4
feet or 6 feet in the 1/100 flood zone may be cost-effective under a low scenario of climate change (Table
S17). Moreover, dry floodproofing existing buildings in the 1/100 year flood zone has B/C ratios >1 under
all combinations with the middle climate change scenario using the most likely risk estimates (Table S17)
and according to the lower bound risk estimates and low discount rate (Table S18). A similar picture
emerges for dry floodproofing of new homes (Tables S20-S22).
Overall results show that for new buildings, the highest NPVs can be obtained with the elevation strategy,
irrespective of the uncertainty in risk estimates (Section 1.9). For existing buildings, the highest NPVs can
be obtained using dry floodproofing, although overall BCA results between dry- and wet floodproofing do
not differ much. Nevertheless, we advise the wet floodproofing of existing buildings for Strategies S1 and
S3, because current building code regulations allow dry floodproofing only for commercial and mixed
commercial residential building uses, and not for residential buildings (26). Given the potential cost-
effectiveness of dry floodproofing, future research could examine whether it is feasible to take up this
measure in building code regulations for residential houses.
22
Table S5. BCA results of elevation of existing buildings using the most likely estimates of (reduced) risk (in
US$). Numbers in bold represent calculations with B/C ratio > 1.
High discounting Low discounting

Elevation existing buildings
B/C ratio NPV B/C ratio NPV
No climate change
1/100 flood zone +2 feet 0.07 -2,184,227,401 0.11 -2,072,811,631
1/100 flood zone +4 feet 0.09 -2,267,440,113 0.15 -2,107,489,106
1/100 flood zone +6 feet 0.10 -2,365,254,919 0.16 -2,185,280,870
1/500 flood zone +2 feet 0.01 -1,312,674,438 0.03 -1,298,366,914
1/500 flood zone +4 feet 0.02 -1,391,405,558 0.03 -1,373,148,241
1/500 flood zone +6 feet 0.02 -1,460,764,561 0.03 -1,441,905,546
Low scenario
1/100 flood zone +2 feet 0.10 -2,104,162,145 0.19 -1,890,145,088
1/100 flood zone +4 feet 0.13 -2,175,273,558 0.24 -1,882,275,293
1/100 flood zone +6 feet 0.14 -2,258,317,179 0.26 -1,928,471,076
1/500 flood zone +2 feet 0.02 -1,304,899,215 0.04 -1,280,461,513
1/500 flood zone +4 feet 0.02 -1,385,593,744 0.04 -1,356,376,299
1/500 flood zone +6 feet 0.02 -1,455,215,308 0.04 -1,425,188,676
Middle scenario
1/100 flood zone +2 feet 0.27 -1,700,948,174 0.59 -958,324,418
1/100 flood zone +4 feet 0.35 -1,621,609,946 0.78 -556,203,612
1/100 flood zone +6 feet 0.38 -1,627,774,318 0.83 -438,028,124
1/500 flood zone +2 feet 0.07 -1,242,231,196 0.15 -1,131,438,384
1/500 flood zone +4 feet 0.07 -1,317,994,849 0.16 -1,188,321,655
1/500 flood zone +6 feet 0.07 -1,386,898,743 0.16 -1,254,292,942
23
Table S6. BCA results of elevation of existing buildings using the lower bound estimates of (reduced) risk
(in US$). Numbers in bold represent calculations with B/C ratio > 1.

No climate change
1/100 flood zone + 2 feet 0.03 -2,260,955,632 0.06 -2,204,579,252
1/100 flood zone + 4 feet 0.05 -2,377,592,902 0.08 -2,296,657,693
1/100 flood zone + 6 feet 0.05 -2,489,196,893 0.08 -2,398,130,024
1/500 flood zone + 2 feet 0.01 -1,322,527,540 0.01 -1,315,287,933
1/500 flood zone + 4 feet 0.01 -1,403,978,747 0.02 -1,394,740,545
1/500 flood zone + 6 feet 0.01 -1,473,752,120 0.02 -1,464,209,458
Low scenario
1/100 flood zone + 2 feet 0.05 -2,220,442,613 0.10 -2,112,149,982
1/100 flood zone + 4 feet 0.06 -2,330,956,625 0.12 -2,182,699,503
1/100 flood zone + 6 feet 0.07 -2,435,086,396 0.13 -2,268,184,268
1/500 flood zone + 2 feet 0.01 -1,318,593,277 0.02 -1,306,227,800
1/500 flood zone + 4 feet 0.01 -1,401,037,970 0.02 -1,386,253,942
1/500 flood zone + 6 feet 0.01 -1,470,944,198 0.02 -1,455,750,722
Middle scenario
1/100 flood zone + 2 feet 0.14 -2,016,416,343 0.30 -1,640,648,722
1/100 flood zone + 4 feet 0.18 -2,050,802,838 0.39 -1,511,707,233
1/100 flood zone + 6 feet 0.19 -2,116,031,708 0.42 -1,514,020,134
1/500 flood zone + 2 feet 0.03 -1,286,883,260 0.08 -1,230,822,097
1/500 flood zone + 4 feet 0.04 -1,366,832,929 0.08 -1,301,218,293
1/500 flood zone + 6 feet 0.03 -1,436,376,016 0.08 -1,369,277,481
24
Table S7. BCA results of elevation of existing buildings using the higher bound estimates of (reduced) risk

No climate change
1/100 flood zone + 2 feet 0.11 -2,086,064,967 0.19 -1,904,234,431
1/100 flood zone + 4 feet 0.15 -2,126,515,896 0.25 -1,865,475,852
1/100 flood zone + 6 feet 0.16 -2,206,689,479 0.27 -1,912,971,831
1/500 flood zone + 2 feet 0.02 -1,300,068,849 0.04 -1,276,718,970
1/500 flood zone + 4 feet 0.03 -1,375,320,020 0.05 -1,345,524,079
1/500 flood zone + 6 feet 0.03 -1,444,148,899 0.05 -1,413,370,987
Low scenario
1/100 flood zone + 2 feet 0.16 -1,955,398,470 0.31 -1,606,122,633
1/100 flood zone + 4 feet 0.21 -1,976,100,078 0.40 -1,497,926,910
1/100 flood zone + 6 feet 0.22 -2,032,167,088 0.43 -1,493,858,247
1/500 flood zone + 2 feet 0.03 -1,287,379,685 0.06 -1,247,497,356
1/500 flood zone + 4 feet 0.04 -1,365,835,139 0.07 -1,318,152,269
1/500 flood zone + 6 feet 0.03 -1,435,092,518 0.07 -1,386,089,055
Middle scenario
1/100 flood zone + 2 feet 0.45 -1,297,353,269 0.96 -85,391,299
1/100 flood zone + 4 feet 0.57 -1,072,521,064 1.27 666,222,073
1/100 flood zone + 6 feet 0.62 -1,003,121,138 1.36 938,544,651
1/500 flood zone + 2 feet 0.11 -1,185,105,478 0.25 -1,004,291,610
1/500 flood zone + 4 feet 0.11 -1,255,513,743 0.26 -1,043,887,091
1/500 flood zone + 6 feet 0.11 -1,323,599,884 0.26 -1,107,187,217
25
Table S8. BCA results of elevation of new residential buildings using the most likely estimates of (reduced)
risk (in US$). Numbers in bold represent calculations with B/C ratio > 1.

Elevation new buildings in
flood zones to different levels B/C ratio NPV B/C ratio NPV
No climate change
1/100 V flood zone +2 feet 2.71 1,332,067 4.60 3,663,757
1/100 V flood zone +4 feet 1.89 1,390,747 3.22 4,505,919
1/100 V flood zone +6 feet 1.40 946,283 2.39 4,233,206
1/100 A flood zone +2 feet 1.16 6,304,226 1.98 49,086,693
1/100 A flood zone +4 feet 0.81 -14,746,711 1.37 37,559,323
1/100 A flood zone +6 feet 0.56 -51,063,821 0.95 -7,795,108
1/500 flood zone +2 feet 0.14 -13,899,798 0.24 -15,999,911
1/500 flood zone +4 feet 0.09 -29,385,369 0.16 -35,513,607
1/500 flood zone +6 feet 0.06 -45,579,498 0.11 -56,597,338
Low scenario
1/100 V flood zone +2 feet 4.56 2,777,653 12.42 11,604,413
1/100 V flood zone +4 feet 3.14 3,337,845 8.55 15,343,593
1/100 V flood zone +6 feet 2.32 3,084,324 4.42 10,416,703
1/100 A flood zone +2 feet 1.80 30,852,506 3.42 121,452,113
1/100 A flood zone +4 feet 1.25 18,974,466 2.37 136,982,605
1/100 A flood zone +6 feet 0.86 -16,230,822 1.63 94,868,607
1/500 flood zone +2 feet 0.21 -12,812,905 0.57 -9,147,017
1/500 flood zone +4 feet 0.14 -27,942,665 0.37 -26,458,963
1/500 flood zone +6 feet 0.09 -44,125,763 0.17 -52,319,754
Middle scenario
1/100 V flood zone +2 feet 13.10 9,437,293 37.48 37,076,556
1/100 V flood zone +4 feet 9.24 12,858,376 26.61 52,055,567
1/100 V flood zone +6 feet 6.83 13,650,677 14.71 41,800,086
1/100 A flood zone +2 feet 5.86 186,991,892 12.99 601,159,539
1/100 A flood zone +4 feet 4.13 240,929,533 9.17 818,828,738
1/100 A flood zone +6 feet 2.86 214,676,145 6.35 804,436,234
1/500 flood zone +2 feet 0.77 -3,708,238 2.25 26,449,538
1/500 flood zone +4 feet 0.51 -15,804,929 1.50 21,017,282
1/500 flood zone +6 feet 0.34 -31,906,153 0.77 -14,888,339
26
Table S9. BCA results of elevation of new residential buildings using the lower bound estimates of
(reduced) risk (in US$). Numbers in bold represent calculations with B/C ratio > 1.

Elevation new buildings
No climate change
1/100 V flood zone + 2 feet 1.37 288,650 2.33 1,351,734
1/100 V flood zone + 4 feet 0.96 -67,035 1.63 1,275,740
1/100 V flood zone + 6 feet 0.71 -677,310 1.21 635,620
1/100 A flood zone + 2 feet 0.59 -15,814,123 1.00 76,457
1/100 A flood zone + 4 feet 0.41 -45,469,959 0.70 -30,517,802
1/100 A flood zone + 6 feet 0.28 -82,850,478 0.48 -78,228,554
1/500 flood zone + 2 feet 0.07 -15,042,108 0.12 -18,531,064
1/500 flood zone + 4 feet 0.05 -30,886,618 0.08 -38,840,102
1/500 flood zone + 6 feet 0.03 -47,089,658 0.05 -59,943,578
Low scenario
1/100 V flood zone + 2 feet 2.31 1,020,116 6.28 5,369,706
1/100 V flood zone + 4 feet 1.59 918,197 4.33 6,759,603
1/100 V flood zone + 6 feet 1.17 404,539 2.23 3,764,469
1/100 A flood zone + 2 feet 0.91 -3,392,694 1.73 36,693,360
1/100 A flood zone + 4 feet 0.63 -28,407,043 1.20 19,790,379
1/100 A flood zone + 6 feet 0.43 -65,224,981 0.83 -26,280,714
1/500 flood zone + 2 feet 0.11 -14,492,140 0.29 -15,063,499
1/500 flood zone + 4 feet 0.07 -30,156,610 0.19 -34,258,452
1/500 flood zone + 6 feet 0.05 -46,354,068 0.09 -57,779,121
Middle scenario
1/100 V flood zone + 2 feet 6.63 4,389,894 18.96 18,258,610
1/100 V flood zone + 4 feet 4.68 5,735,585 13.46 25,335,862
1/100 V flood zone + 6 feet 3.46 5,751,113 7.44 19,644,461
1/100 A flood zone + 2 feet 2.97 75,613,836 6.57 279,425,317
1/100 A flood zone + 4 feet 2.09 83,902,221 4.64 364,804,522
1/100 A flood zone + 6 feet 1.45 51,613,945 3.21 332,760,506
1/500 flood zone + 2 feet 0.39 -9,885,179 1.14 2,948,358
1/500 flood zone + 4 feet 0.26 -24,014,915 0.76 -10,235,472
1/500 flood zone + 6 feet 0.17 -40,170,945 0.39 -38,838,825
27
Table S10. BCA results of elevation of new residential buildings using the higher bound estimates of

Elevation new buildings
No climate change
1/100 V flood zone + 2 feet 4.42 2,666,966 7.51 6,621,650
1/100 V flood zone + 4 feet 3.09 3,255,763 5.25 8,638,455
1/100 V flood zone + 6 feet 2.29 3,023,430 3.90 8,835,786
1/100 A flood zone + 2 feet 1.90 34,601,386 3.23 111,788,047
1/100 A flood zone + 4 feet 1.32 24,559,145 2.24 124,653,944
1/100 A flood zone + 6 feet 0.91 -10,397,491 1.55 82,314,077
1/500 flood zone + 2 feet 0.23 -12,438,380 0.40 -12,761,676
1/500 flood zone + 4 feet 0.15 -27,464,742 0.26 -31,257,849
1/500 flood zone + 6 feet 0.10 -43,647,471 0.17 -52,316,319
Low scenario
1/100 V flood zone + 2 feet 7.44 5,026,161 20.26 19,580,800
1/100 V flood zone + 4 feet 5.12 6,433,428 13.95 26,325,539
1/100 V flood zone + 6 feet 3.78 6,512,714 7.21 18,927,254
1/100 A flood zone + 2 feet 2.94 74,664,177 5.59 229,888,413
1/100 A flood zone + 4 feet 2.03 79,592,105 3.86 286,912,740
1/100 A flood zone + 6 feet 1.40 46,449,964 2.66 249,861,261
1/500 flood zone + 2 feet 0.34 -10,664,570 0.93 -1,577,752
1/500 flood zone + 4 feet 0.23 -25,110,250 0.61 -16,480,669
1/500 flood zone + 6 feet 0.15 -41,274,975 0.28 -45,335,302
Middle scenario
1/100 V flood zone + 2 feet 21.37 15,894,694 61.16 61,151,337
1/100 V flood zone + 4 feet 15.08 21,970,933 43.42 86,239,481
1/100 V flood zone + 6 feet 11.15 23,757,001 24.00 70,144,934
1/100 A flood zone + 2 feet 9.56 329,483,655 21.21 1,012,770,933
1/100 A flood zone + 4 feet 6.74 441,822,775 14.96 1,399,685,630
1/100 A flood zone + 6 feet 4.67 423,290,135 10.36 1,407,875,628
1/500 flood zone + 2 feet 1.26 4,194,245 3.68 56,515,824
1/500 flood zone + 4 feet 0.84 -5,301,464 2.44 61,000,562
1/500 flood zone + 6 feet 0.56 -21,332,572 1.25 15,752,769
28
Table S11. BCA results wet of flood-proofing existing buildings using the most likely estimates of (reduced)
Wet floodproofing existing buildings in flood zones High discounting Low discounting
to different levels under low and high effectiveness
scenarios in saving flood damage B/C ratio NPV B/C ratio NPV
No climate change
1/100 flood zone +2 feet low 0.31 -171,247,557 0.53 -116,896,578
1/100 flood zone +2 feet high 0.51 -120,371,666 0.88 -29,525,931
Low scenario
1/100 flood zone +2 feet high 0.70 -73,621,771 1.45 111,154,427
Middle scenario
1/100 flood zone +2 feet low 1.30 73,635,534 2.86 460,097,155
1/100 flood zone +2 feet high 1.99 245,223,320 4.78 934,472,646
1/100 flood zone +4 feet high 1.69 350,464,661 3.74 1,398,110,236
1/100 flood zone +6 feet low 0.64 -352,192,029 1.41 404,088,044
29
Table S12. BCA results wet of flood-proofing existing buildings using the lower bound estimates of
Wet floodproofing existing buildings in flood High discounting Low discounting
zones to different levels under low and high
effectiveness scenarios in saving flood damage B/C ratio NPV B/C ratio NPV
No climate change
1/100 flood zone + 2 feet low 0.16 -208,677,219 0.27 -181,175,623
1/100 flood zone + 2 feet high 0.26 -182,934,018 0.45 -136,966,076
Low scenario
Middle scenario
1/100 flood zone + 2 feet low 0.66 -84,766,374 1.45 110,783,206
1/100 flood zone + 2 feet high 1.01 2,057,045 2.42 350,817,204
1/100 flood zone + 4 feet high 0.85 -75,132,374 1.89 454,976,287
30
Table S13. BCA results wet of flood-proofing existing buildings using the higher bound estimates of
(reduced) risk (in US$). Numbers in bold correspond to calculations with B/C ratio > 1.
Wet floodproofing existing buildings in flood High discounting Low discounting
No climate change
1/100 flood zone + 2 feet low 0.50 -123,361837 0.86 -34,661,038
Low scenario
Middle scenario
1/100 flood zone + 2 feet low 2.12 276,287,369 4.67 906,992,733
1/100 flood zone + 2 feet high 3.25 556,318,635 7.81 1,681,173,535
1/100 flood zone + 4 feet low 1.65 332,607,357 3.65 1,355,791,567
1/100 flood zone + 6 feet low 1.05 44,296,069
2.31 1,278,545,148
31
Table S14. BCA results of wet flood-proofing new residential buildings using the most likely estimates of
Wet floodproofing new buildings in flood zones High discounting Low discounting
to different levels under low and high
No climate change
1/100 flood zone +2 feet high 0.60 -12,951,240 1.02 893,089
Low scenario
Middle scenario
1/500 flood zone +2 feet high 0.44 -7,711,312 0.97 -510,147
32
Table S15. BCA results of wet flood-proofing new residential buildings using the lower bound estimates of
Wet floodproofing new buildings in flood zones High discounting Low discounting
No climate change
Low scenario
Middle scenario
33
Table S16. BCA results of wet flood-proofing new residential buildings using the higher bound estimates of
Wet floodproofing new buildings in flood zones to High discounting Low discounting
different levels under low and high effectiveness
No climate change
1/100 flood zone + 2 feet low 0.59 -13,147,399 1.01 458,438
1/100 flood zone + 2 feet high 0.98 -650,019 1.67 28,150,356
1/100 flood zone + 6 feet high 0.43 -72,857837 0.74 -44,097,353
Low scenario
Middle scenario
1/500 flood zone + 2 feet low 0.43 -7,735,608 0.96 -666,136
34
Table S17. BCA results of dry flood-proofing existing buildings using the most likely estimates of (reduced)
Dry floodproofing existing buildings in flood zones High discounting Low discounting
No climate change
Low scenario
Middle scenario
1/100 flood zone +4 feet low 1.48 386,132,035 3.24 1,802,785,286
1/100 flood zone +6 feet low 1.55 535,179,344 3.37 2,314,211,030
35
Table S18. BCA results of dry flood-proofing existing buildings using the lower bound estimates of
Dry floodproofing existing buildings in flood zones to High discounting Low discounting
No climate change
Low scenario
Middle scenario
36
Table S19. BCA results of dry flood-proofing existing buildings using the higher bound estimates of
Dry floodproofing existing buildings in flood High discounting Low discounting
No climate change
Low scenario
Middle scenario
1/100 flood zone + 2 feet low 1.70 444,656,290 3.74 1,743,421,072
1/100 flood zone + 4 feet low 2.41 1,139,568,503 5.28 3,451,546,609
1/100 flood zone + 4 feet high 2.51 1,218,429,709 5.64 3,743,705,683
1/100 flood zone + 6 feet low 2.53 1,490,634,430 5.50 4,394,014,141
1/100 flood zone + 6 feet high 2.95 1,905,639,780 6.42 5,289,152,073
37
Table S20. BCA results of dry flood-proofing new residential buildings using the most likely estimates of
Dry floodproofing new buildings in flood zones until
different levels under low and high effectiveness High discounting Low discounting
No climate change
Low scenario
Middle scenario
38
Table S21. BCA results of dry flood-proofing new residential buildings using the lower bound estimates of
Dry floodproofing new buildings in flood zones High discounting Low discounting
until different levels under low and high
No climate change
Low scenario
Middle scenario
39
Table S22. BCA results of dry flood-proofing new residential buildings using the higher bound estimates of
Dry floodproofing new buildings in flood zones until High discounting Low discounting
No climate change
Low scenario
Middle scenario
40
3.2 Results of the Storm Surge Barrier Strategies S2a,b,c and Hybrid Strategy S3
The BCA results of the storm surge barrier strategies S2a,b,c and the hybrid strategy S3 are shown in
Tables S23-S34. Results are shown in separate tables depending on the strategy under consideration.
Different tables show results that are derived using the most likely estimates of reduced risk or the lower, or
higher, estimates which reflect the uncertainty interval presented in Section 1.9 of this SM.
None of the hybrid S3 and storm surge barriers strategies S2,a,b,c have a B/C>1 under current levels of
flood risk and the low climate change scenario using the lower bound or most likely risk estimates. Only
S2c and S3 have B/C ratios that slightly exceed 1 under the low climate change scenario if a low discount
rate is used and the higher bound of the risk estimate S3 has the highest NPV under the current and low
climate change conditions, independently of the uncertainty of the risk estimate.
Under the middle climate change scenarios, all strategies have a B/C >1 if a low discount rate is used, and
only S2c and S3 have a B/C>1 if a high discount rate and the most likely or higher bound risk estimates are
used. Strategy S2c results in the highest NPV in that scenario, independently of the uncertainty of the risk
estimate.
All storm surge barriers have a B/C>1 if flood risk develops according to the high (rapid ice melt) scenario,
using the most likely and upper bound risk estimates. In particular, S2c is economically the most attractive
investment in that case, with a B/C ratio larger than 1 for the whole uncertainty range of the risk estimate.
In that scenario storm surge barrier S2c results in a NPV of about $64 billion up to $68 billion if a low
discount rate and the most likely risk estimate are used, which indicates a very high economic return.
Table S23. BCA results of storm surge barriers strategy 2a using the most likely estimates of (reduced) risk
Discounting Discounting
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -19,506 -18,434 -19,504 -18,390
B/C ratio 0.13 0.21 0.13 0.22
Low scenario
NPV (in $ million) -17,515 -13,908 -17,502 -13,537
B/C ratio 0.22 0.41 0.22 0.42
Middle scenario
NPV (in $ million) -8,904 7,506 -8,815 9,985
B/C ratio 0.60 1.32 0.61 1.43
High scenario
NPV (in $ million) 9,717 46,319 9,857 50,180
B/C ratio 1.43 2.98 1.44 3.14
41
Table S24. BCA results of storm surge barriers strategy 2a using the lower estimates of (reduced) risk (in
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -20,938 -20,894 -20,938 -20,892
B/C ratio 0.07 0.11 0.07 0.11
Low scenario
NPV (in $ million) -19,931 -18,604 -19,925 -18,437
B/C ratio 0.11 0.21 0.11 0.21
Middle scenario
NPV (in $ million) -15,574 -7,768 -15,529 -6,535
B/C ratio 0.30 0.67 0.31 0.72
High scenario
NPV (in $ million) -6,151 11, 871 -6,081 13,804
B/C ratio 0.73 1.51 0.73 1.59
Table S25. BCA results of storm surge barriers strategy 2a using the higher estimates of (reduced) risk (in
US$). Numbers in bold correspond to calculations with B/C ratio > 1.
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -17,674 -15,287 -17,670 -15,188
B/C ratio 0.21 0.35 0.21 0.35
Low scenario
NPV (in $ million) -14,425 -7,901 -14,402 -7,268
B/C ratio 0.36 0.66 0.36 0.69
Middle scenario
NPV (in $ million) -371 27,048 -224 31,120
B/C ratio 0.98 2.16 0.99 2.33
High scenario
NPV (in $ million) 30,019 90,390 30,248 96,719
B/C ratio 2.34 4.86 2.35 5.12
42
Table S26. BCA results of storm surge barriers strategy 2b using the most likely estimates of (reduced) risk
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -18,958 -19,908 -19, 747 -18,925
B/C ratio 0.12 0.21 0.13 0.21
Low scenario
NPV (in $ million) -17,758 -14,432 -17,745 -14,072
B/C ratio 0.22 0.40 0.22 0.41
Middle scenario
NPV (in $ million) -9,146 6,982 -9,058 9,499
B/C ratio 0.60 1.29 0.60 1.39
High scenario
NPV (in $ million) 9,475 45,795 9,614 49,645
B/C ratio 1.42 2.91 1.42 3.07
Table S27. BCA results of storm surge barriers strategy 2b using the lower estimates of (reduced) risk (in
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -21,159 -21,418 -21, 428 -18,925
B/C ratio 0.07 0.11 0.11 0.21
Low scenario
NPV (in $ million) -20,173 -19,128 -20,168 -18,972
B/C ratio 0.11 0.20 0.11 0.21
Middle scenario
NPV (in $ million) -15,861 -8,292 -15,772 -7,070
B/C ratio 0.30 0.6536 0.3037 0.7053
High scenario
NPV (in $ million) -6,394 11,347 -6,324 13,269
B/C ratio 0.72 1.47 0.72 1.55
43
Table S28. BCA results of storm surge barriers strategy 2b using the higher estimates of (reduced) risk (in
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -18,108 -18,811 -15, 724 -18,925
B/C ratio 0.20 0.34 0.34 0.21
Low scenario
NPV (in $ million) -14,667 -8,424 -14,644 -7,804
B/C ratio 0.35 0.65 0.35 0.67
Middle scenario
NPV (in $ million) -613 26,524 -467 30,584
B/C ratio 0.97 2.11 0.98 2.27
High scenario
NPV (in $ million) 29,777 89,866 30,005 96,183
B/C ratio 2.31 4.75 2.32 5.01
Table S29. BCA results of storm surge barriers strategy 2c using the most likely estimates of (reduced) risk
High Low High Low
100 years 150 years
Current climate
NPV (in mln) -11,213 -10,039 -11,211 -9,990
B/C ratio 0.23 0.36 0.23 0.37
Low scenario
NPV (in mln) -8,938 -4,866 -8,922 -4,443
B/C ratio 0.38 0.69 0.39 0.72
Middle scenario
NPV (in mln) 904 19,608 1006 22,439
B/C ratio 1.06 2.24 1.07 2. 42
High scenario
NPV (in mln) 22,185 63,965 22,345 68,376
B/C ratio 2.53 5.07 2.54 5.33
44
Table S30. BCA results of storm surge barriers strategy 2c using the lower estimates of (reduced) risk (in
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -12,850 -12,850 -12,850 -12,850
B/C ratio 0.12 0.18 0.12 0.19
Low scenario
NPV (in $ million) -11,699 -10,232 -11,692 -10,043
B/C ratio 0.19 0.35 0.20 0.36
Middle scenario
NPV (in $ million) -6,719 2,151 -6,668 3,559
B/C ratio 0.54 1.14 0.54 1.23
High scenario
NPV (in $ million) 4,050 24,596 4,129 26,803
B/C ratio 1.28 2.56 1.28 2.69
Table S31. BCA results of storm surge barriers strategy 2c using the higher estimates of (reduced) risk (in
High Low High Low
100 years 150 years
Current climate
NPV (in $ million) -9,119 -6,442 -9,115 -6,331
B/C ratio 0.37 0.59 0.37 0.60
Low scenario
NPV (in $ million) -5,406 2,000 -5,379 2,721
B/C ratio 0.63 1.13 0.63 1.17
Middle scenario
NPV (in $ million) 10,656 41,941 10,823 46,592
B/C ratio 1.74 3.67 1.75 3.95
High scenario
NPV (in $ million) 45,387 114,332 45,649 121,563
B/C ratio 4.12 8.27 4.14 8.70
45
Table S32. TBCA results of the hybrid strategy 3 using the most likely estimates of (reduced) risk (in US$).
Numbers in bold represent calculations with B/C ratio > 1.
High Low High Low
Effectiveness wet-proofing Low High
Current climate
NPV (in $ million) -8,561 -6,569 -8,510 -6,319
B/C ratio 0.26 0.44 0.26 0.45
Low scenario
NPV (in $ million) -6,994 -2,958 -6,933 -2,835
B/C ratio 0.39 0.75 0.40 0.76
Middle scenario
NPV (in $ million) 870 16,534 1,042 16,942
B/C ratio 1.08 2.41 1.09 2.45
High scenario
NPV (in $ million) 16,644 49,801 16,816 49,801
B/C ratio 2.44 5.26 2.46 5.26
Table S33. BCA results of the hybrid strategy 3 using the lower estimates of (reduced) risk (in US$).
High Low High Low
Current climate
NPV (in $ million) -10,022 -9,094 -9,996 -9,050
B/C ratio 0.13 0.22 0.13 0.23
Low scenario
NPV (in $ million) -9,229 -7,267 -9,198 -7,205
B/C ratio 0.20 0.38 0.20 0.38
Middle scenario
NPV (in $ million) -5,249 2,595 -5,162 2,802
B/C ratio 0.54 1.22 0.55 1.24
High scenario
NPV (in $ million) 2,731 19,222 2,818 19,429
B/C ratio 1.24 2.65 1.24 2.66
46
Table S34. BCA results of the hybrid strategy 3 using the higher estimates of (reduced) risk (in US$).
High Low High Low
Current climate
NPV (in $ million) -6,692 -3,337 -6,608 -3,195
B/C ratio 0.42 0.71 0.43 0.73
Low scenario
NPV (in $ million) -4,134 2,554 -4,034 2,755
B/C ratio 0.64 1.22 0.65 1.24
Middle scenario
NPV (in $ million) 8,701 34,365 8,980 35,032
B/C ratio 1.76 3.94 1.78 4.00
High scenario
NPV (in $ million) 34,443 87,992 34,723 88,658
B/C ratio 3.99 8.53 4.01 8.59
3.3 BCA and the Timing of Investing in Barriers

Tables S35-S37 show the BCA results of investing in storms surge barriers S2a,b,c in 2040 under the
climate change scenarios. In other words, these are the results for delaying the investment by 25 years (27).
NPVs are expressed in current net present values. The results show that it is economically rational to invest
in a storm surge barrier system in 2040 if climate change develops according to the high climate change
scenario, while such an investment is not economically rational if the low climate change scenario occurs.
Investing in S2c yields the highest NPV.
Table S35. BCA results of delaying investment in storm surge barriers strategy 2a until 2040 using the
most likely estimates of (reduced) risk (in US$). Numbers in bold represent calculations with B/C ratio > 1.
High Low High Low
100 years 150 years
Low scenario
NPV (in $ million) -2,913 -4,018 -2,910 -3,852

B/C ratio 0.29 0.54 0.30 0.56
Middle scenario
NPV (in $ million) 398 11,243 418 12,393
B/C ratio 1.10 2.28 1.10 2.41
High scenario
NPV (in $ million) 6,240 32,915 6,271 34,630
B/C ratio 2.51 4.75 2.52 4.94
47
Table S36. BCA results of delaying investment in storm surge barriers strategy 2b until 2040 using the
most likely estimates of (reduced) risk (in US$). Numbers in bold represent calculations with B/C ratio > 1.
High Low High Low
100 years 150 years
Low scenario
NPV (in $ million) -2,957 -4,215 -2,954 -4,053
B/C ratio 0.29 0.53 0.29 0.55
Middle scenario
NPV (in $ million) 353 11,047 373 12,192
B/C ratio 1.08 2.23 1.09 2.35
High scenario
NPV (in $ million) 6,196 32,719 6,226 34,429
B/C ratio 2.48 4.64 2.49 4.82
Table S37. BCA results of delaying investment in storm surge barriers strategy 2c until 2040 using the most
likely estimates of (reduced) risk (in US$). Numbers in bold represent calculations with B/C ratio > 1.
High Low High Low
100 years 150 years
Low scenario
NPV (in $ million) -1,287 -455 -1,284 -266
B/C ratio 0.52 0.92 0.52 0.96
Middle scenario
NPV (in $ million) 2,496 16,987 2,519 16,480
B/C ratio 1.93 3.88 1.94 3.78
High scenario
NPV (in $ million) 9,173 41,755 9,208 43,714
B/C ratio 4.43 8.08 4.44 8.39
48
4. Overview of Data Sources
In an integrated study such as this one, data from a wide variety of sources is integrated into an overall risk
analysis. Table S38 gives an overview of the data sources used, along with links for more information.
Table S38. Overview of data used in this study
Data Type Description Source Contact - Link

Storms Track, intensity, and size of the Lin et al., 2012 dx.doi.org/10.1038/nclimate1389
549 storms emanuel@mit.edu
Surges Estimated surges for the 549 Lin et al., 2012 dx.doi.org/10.1038/nclimate1389
storms and associated nlin@princeton.edu
probabilities
DEM High resolution (3 foot) digital NYC Mayor’s Office NYC Mayor’s Office of Long-
elevation model (LiDAR) for Term Planning and Sustainability
NYC
Inundation Median inundation depth per This study hans.de.moel@vu.nl
storm for each census block
Exposure Exposure data on buildings and NYC OEM NYC Office of Emergency
vehicles in HAZUS classes Management
specifically for NYC
Damage curves Depth-damage curves of HAZUS-MH4 www.fema.gov/media-
different building types library/assets/documents/16579?id
=3726
Risk estimates Risk calculations for different This study hans.de.moel@vu.nl
measures and under different
future scenarios
Damage Sandy to Assessments on the damage States of NJ and NYC, dx.doi.org/ 10.1111/nyas.12200
sectors caused by Sandy for various quoted in Aerts et al., hans.de.moel@vu.nl
sectors 2013
Housing assistance Damage assessments of houses FEMA https://www.fema.gov/media-
data NJ and NY for illustrating the spatial Used by Aerts et al., library/assets/documents/30714?id
Sandy distribution of damage by Sandy 2013 =6963
dx.doi.org/ 10.1111/nyas.12200
hans.de.moel@vu.nl
Costs building Cost estimates of flood proofing Aerts et al., 2013 dx.doi.org/ 10.1111/nyas.12200
codes measures for HAZUS building wouter.botzen@vu.nl
types
Costs barriers Cost estimates of different Aerts et al., 2013 dx.doi.org/ 10.1111/nyas.12200
barriers, levee and other jeroen.aerts@vu.nl
protection systems
Climate change Change of surge probabilities Lin et al., 2012 dx.doi.org/10.1038/nclimate1389
impacts on surge based on four climate model nlin@princeton.edu
risk projections (IPCC AR4 A1B
emissions scenario)
Population growth Demographic projections for New York www.nymtc.org/project/forecastin
2040 at the scale of Traffic Metropolitan g/SED_products/2040%20Forecast
Analysis Zones (TAZ) Transportation Council s/2040%20Final%20Draft_TAZ_r
eport.pdf
BCA Guide for conducting flood- FEMA www.fema.gov/library/viewRecor
related BCAs d.do?id=4185
49
Supplementary Text: Acknowledgments
J.C.J.H.A, W.J.W.B., and H.d.M. would like to thank the financial support from the KfC Theme
1, Netherlands Organization for Scientific Research Innovational Research Incentives Scheme
Veni (NWO-VENI) and the European Union Framework Programme 7 ENHANCE and TURAS
projects, the C-40 and Connecting Delta Cities networks. N.L. acknowledges support from
Princeton University's Andlinger Center for Energy and the Environment (Innovation Fund) and
School of Engineering and Applied Science (Project X Fund). NYC Department of City Planning,
NYC Mayor’s Office of Long-Term Planning and Sustainability, NYC Department of Buildings
and Office of Emergency Management also provided support. E.M.-K. acknowledges support
from the National Science Foundation (NSF grant# SES-1062039), the Zurich Insurance
Foundation, the Department of Homeland Security Center of Excellence CREATE at the
University of Southern California, the Travelers Foundation and the Wharton Risk Center’s
Extreme Events Project.
References in Supplementary Material

1. N. Lin, K. Emanuel, M. Oppenheimer, E. Vanmarcke, Physically based assessment of
hurricane surge threat under climate change. Nat. Clim. Change 2, 462–467 (2012).
doi:10.1038/nclimate1389
2. K. Emanuel, Climate and tropical cyclone activity: A new model downscaling approach. J.
Clim. 19, 4797–4802 (2006). doi:10.1175/JCLI3908.1
3. R. A. Luettich, J. J. Westerink, N. W. Scheffner, “ADCIRC: An advanced three-dimensional
circulation model for shelves, coasts, and estuaries. Coastal Engineering Research Center,
Vicksburg, Report 1: Theory and methodology of ADCIRC-2DDI and ADCIRC-3DL”
(Dredging Research Program Tech. Rep. DRP-92-6, prepared for U.S. Army Corps of
Engineers, Washington, DC, 1992); www.dtic.mil/dtic/tr/fulltext/u2/a261608.pdf.
4. Federal Emergency Management Agency, “HAZUS-MH MR4 flood model technical manual”
(FEMA, Washington, DC, 2009); www.fema.gov/media-
library/assets/documents/16579?id=3726.
5. Applied Research Associates, “Level 2 exposure data compilation for New York City Office of
Emergency Management” (Prepared for National Institute of Building Sciences, ARA,
Albuquerque, NM, 2007).
6. P. J. Ward, H. de Moel, J. C. J. H. Aerts, How are flood risk estimates affected by the choice of
return-periods? Nat. Hazards Earth Syst. Sci. 11, 3181–3195 (2011). doi:10.5194/nhess-
11-3181-2011
7. J. C. J. H. Aerts, N. Lin, W. Botzen, K. Emanuel, H. de Moel, Low-probability flood risk
modeling for New York City. Risk Anal. 33, 772–788 (2013). doi:10.1111/risa.12008
Medline
8. C. I. Moore, S. A. Dendrou, R. S. Taylor, “Total stillwater frequency-elevations,
implementation and results” (New York City flood insurance study report no. 7, Prepared
for New York Department of Environmental Conservation, Albany, NY, 1981).
9. C. F. A. S. C. E. Scawthorn, P. Flores, N. Blais, H. Seligson, E. Tate, S. Chang, E. Mifflin, W.
Thomas, J. Murphy, C. Jones, M. Lawrence, HAZUS-MH Flood loss estimation
50
methodology. II: Damage and loss assessment. Nat. Hazards Rev. 7, 72–81 (2006).
doi:10.1061/(ASCE)1527-6988(2006)7:2(72)
10. R. Horton, V. Gornitz, M. Bowman, R. Blake, Chapter 3: climate observations and
projections. Ann. N. Y. Acad. Sci. 1196, 41–62 (2010). doi:10.1111/j.1749-
6632.2009.05314.x Medline
11. New York Mtropolitan Transpotation Council and New York City Departmetn of City
Planning, “2040 forecasts modeling methodology TAZ allocation. Demographic and
socioeconomic forecasting,” (Technical memorandum, NYMTC and NYCDCP, New
York, 2012);
www.nymtc.org/project/forecasting/SED_products/2040%20Forecasts/2040%20Final%2
0Draft_TAZ_report.pdf.
12. International Commission for the Protection of the Rhine, Non-Structural Flood Plain
Management: Measures and Their Effectiveness (ICPR, 2002)
13. H. Kreibich, A. H. Thieken, T. Petrow, M. Müller, B. Merz, Flood loss reduction of private
households due to building precautionary measures – lessons learned from the Elbe flood
in August. Nat. Hazards Earth Syst. Sci. 5, 117–126 (2005). doi:10.5194/nhess-5-117-
2005
14. Environment Agency and the Department for Environment Food and Rural Affairs,
“Developing the evidence base for flood resistance and resilience: Summary report”
(R&D Technical Report FD2607/TR1. Environment Agency and DEFRA, London,
2008).
15. State of New York, Governor Cuomo holds meeting with New York's Congressional
Delegation, Mayor Bloomberg and regional County Executives to review damage
assessment for the state in the wake of Hurricane Sandy [press release] (State of New
York, Albany, NY, 26 November 2012); www.governor.ny.gov/press/11262012-
damageassessment.
16. State of New York, New York State Hurricane Sandy Recovery Needs Summary (State of
New York, Albany, NY, 2012);
www.governor.ny.gov/assets/documents/sandyimpactsummary.pdf.
17. T. M. Hall, A. H. Sobel, On the impact angle of Hurricane Sandy’s New Jersey landfall.
Geophys. Res. Lett. 40, 2312–2315 (2013);
http://onlinelibrary.wiley.com/doi/10.1002/grl.50395/abstract.
18. A. McCulloch, Hurricane Sandy: a look back. Significance: Statistics Making Sensei, 2012;
www.significancemagazine.org/details/webexclusive/3004001/Hurricane-Sandy-a-look-
back.html.
19. J. C. J. H. Aerts, W. J. W. Botzen, H. de Moel, M. Bowman, Cost estimates for flood
resilience and protection strategies in New York City. Ann. N. Y. Acad. Sci. 1294, 1–104
(2013). doi:10.1111/nyas.12200 Medline
20. FEMA, Housing assistance data Hurricane Sandy New York and New Jersey (2013);
www.fema.gov/media-
library/assets/documents/30714?fromSearch=fromsearch&id=6963.
21. H. de Moel, L. M. Bouwer, J. C. J. H. Aerts, Uncertainty and sensitivity of flood risk
calculations for a dike ring in the south of the Netherlands. Sci. Total Environ. 473-474,
224–234 (2014). doi:10.1016/j.scitotenv.2013.12.015
51
22. H. de Moel, N. E. M. Asselman, J. C. J. H. Aerts, Uncertainty and sensitivity analysis of
coastal flood damage estimates in the west of the Netherlands. Nat. Hazards Earth Syst.
Sci. 12, 1045–1058 (2012). doi:10.5194/nhess-12-1045-2012
23. FEMA, BCA Reference Guide (Federal Emergency Management Agency (FEMA)
Department of Homeland Security, Washington, DC, 2009); www.fema.gov/media-
library/assets/documents/18870?id=4185.
24. K. J. Arrow et al., “How should benefits and costs be discounted in an intergenerational
context? The views of an expert panel” (Discussion paper RFF DP 12-53, Resources for
the Future, Washington, DC, 2012).
25. J. C. J. H. Aerts, W. J. W. Botzen, Flood-resilient waterfront development in New York City:
bridging flood insurance, building codes, and flood zoning. Ann. NY Acad. Sci. 1227, 1–
82s (2011). doi:10.1111/j.1749-6632.2011.06074.x Medline
26. FEMA, Homeowner’s Guide to Retrofitting (U.S. Department of Homeland Security: Federal
Insurance and Mitigation Administration (FEMA), Washington, DC, ed. 2, 2009);
www.fema.gov/media-library/assets/documents/480?id=1420.
27. N. Ranger, A. Millner, S. Dietz, S. Fankhauser, A. Lopez, G. Ruta, “Adaptation in the UK: A
decision-making process” (Policy Brief September 2010, Environment Agency, London,
2010); www.cccep.ac.uk/Publications/Policy/docs/PB-adaptationUK-ranger.pdf.
52

Aerts 2014 Science SM

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Aerts 2014 Science SM

Încărcat de

Drepturi de autor:

Formate disponibile

www.sciencemag.

Supplementary Materials for

Evaluating Flood Resilience Strategies for Coastal Megacities

Jeroen C. J. H. Aerts*, W. J. Wouter Botzen, Kerry Emanuel,

E-mail: jeroen.aerts@vu.nl and erwannmk@wharton.upenn.edu

Published 2 May 2014, Science 344, 473 (2014)

This PDF file includes

Materials and Methods

Materials and Methods

Materials and Methods

1.1 Hurricane Flood Hazard Simulations

1.2 HAZUS-MH: Vulnerability and Risk Calculations

Value at risk ($/building) Existing Buildings New Buildings**

1.5 Flood Risk Estimates

1.7 Future Scenario Risk Estimates

1.8 Effect of Damage-Reducing Building Codes

Decrease in risk(%) Decrease in risk (%)

Elevation 2 feet 18% 65%

1.9 Addressing Uncertainties

Risk (Expected Annual

2.1 General Benefit-Cost Analysis Method

3. Benefit-Cost Analysis (BCA) Results

$-19.5bn $-19.8bn $-11.2bn $-8.5bn

B/C ratio using high (7%) and low (4%)

3.1 Results of Building Code Policies

High discounting Low discounting

High discounting Low discounting

High discounting Low discounting

High discounting Low discounting

High discounting Low discounting

High discounting Low discounting

B/C ratio 0.73 1.51 0.73 1.59

B/C ratio 2.34 4.86 2.35 5.12

B/C ratio 0.72 1.47 0.72 1.55

B/C ratio 4.12 8.27 4.14 8.70

B/C ratio 2.44 5.26 2.46 5.26

B/C ratio 1.24 2.65 1.24 2.66

3.3 BCA and the Timing of Investing in Barriers

NPV (in $ million) -2,913 -4,018 -2,910 -3,852

B/C ratio 2.48 4.64 2.49 4.82

B/C ratio 4.43 8.08 4.44 8.39

Table S38. Overview of data used in this study

Data Type Description Source Contact - Link

References in Supplementary Material

S-ar putea să vă placă și