1
The Value of Coastal Wetlands for
Hurricane Protection
Robert Costanza1, Octavio Pérez-Maqueo2, M. Luisa Martinez2, Paul Sutton3,5,
Sharolyn J. Anderson3, Kenneth Mulder4
1.Gund Institute of Ecological Economics, Rubenstein School of Environment and Natural
Resources, University of Vermont, Burlington, VT 05405-1708, USA
2. Instituto de Ecología A.C., km 2.5 antigua carretera a Coatepec no. 351, Congregación El
Haya, Xalapa, Ver. 91070, México.
3. Department of Geography, University of Denver, Denver, CO, 80208, USA
4. W. K. Kellogg Biological Station, Michigan State University, Hickory Corners, MI 49060,
USA
5. Currently Visiting Research Fellow at Department of Geography, Population, and
Environmental Management, Flinders University, Adelaide, South Australia
in review for: BioScience
2
Abstract
Coastal wetlands reduce the damaging effects of hurricanes on coastal communities. A
regression model using 34 major US hurricanes since 1980 with the natural log of damage per
unit GDP in the hurricane swath as the dependent variable and the natural logs of wind speed and
wetland area in the swath as the independent variables was highly significant and explained 60%
of the variation in relative damages. A loss of 1 ha of wetland in the model corresponded to an
average $33,000 (median = $5,000) increase in storm damage from specific storms. Taking into
account the annual probability of hits by hurricanes of varying intensities, we mapped the annual
value of coastal wetlands by 1km x 1km pixel and by state. The annual value ranged from $100
to $15,000/ha/yr, with a mean of $3,300/ha/yr (median = $1,600/ha/yr) consistent with previous
estimates. Coastal wetlands in the US were estimated to currently provide $12.8 Billion/yr in
storm protection services. Coastal wetlands function as valuable, self-maintaining “horizontal
levees” for storm protection, and also provide a host of other ecosystem services that vertical
levees do not. Their restoration and preservation is an extremely cost-effective strategy for
society.
3
Globally, since 1900, 2652 windstorms (including tropical storms, cyclones, hurricanes,
tornadoes, typhoons and winter-storms) have been considered disasters. Altogether, they have
caused 1.2 million human deaths and have cost $381 billion in property damage1. Of these,
hurricanes, cyclones and tropical storms have resulted in $179 billion (47%) in property damage
and the loss of 874 thousand (73%) human lives. The impact of cyclones and hurricanes over the
last decades has increased, owing to an increase in built infrastructure along the coasts, an
increased frequency of category 4 and 5 hurricanes (Webster et al. 2005), and an upward trend in
tropical cyclone destructive potential (Emanuel 2005). Coastal wetlands reduce the damaging
effects of hurricanes on coastal communities by absorbing storm energy in ways that neither
solid land nor open water can (Simpson and Riehl 1981). Coastal wetlands also protect coastal
communities from other types of damages. For example, there is evidence that property damage
and loss of human lives from the 2004 tsunami that hit Southeast Asia was highly ameliorated by
coastal ecosystems (Danielsen et al. 2005).
Methods
We estimated the value of coastal wetlands for hurricane protection using data on the major
hurricanes (those considered “disasters”) that have hit the Atlantic and Gulf coasts of the US
since 1980 and for which data on total damages were available (34 of the total of 267 storms).
We combined data on the tracks of these hurricanes and their wind speeds with data on storm
damages and spatially explicit data on Gross Domestic Product (GDP) and coastal wetland area
in each storm’s swath (Figure 1). The following datasets were assembled for the analysis: 1)
tracks of all hurricanes striking the US from 1980 to 2004, which included wind-speed2. Of
these, only 34 hurricanes had sufficient information on damages (see below) to include in the
regression analysis. All of the storms were used to estimate the strike frequencies, since this did
1
www.em-dat.net/index.htm downloaded, November 2005. This data set defined a hurricane to
be a disaster if it satisfied at least one of the following criteria: 1) 10 or more people killed; 2)
100 or more people affected, injured or homeless; 3) Declaration of a state of emergency
and/or an appeal for international assistance
2
www.grid.unep.ch/data/gnv199.php
4
not require damage information; 2) Nighttime light imagery of the USA (Elvidge et al. 1999). A
1 km resolution GDP map was prepared by using a linear allocation of the national GDP to the
light intensity values of the nighttime image composite (Sutton and Costanza 2002). GDP was
adjusted to the year of each hurricane using data on national GDP by year 3; 4) The 30 meter
resolution National Landcover dataset, which included mapping of both herbaceous and woody
wetlands (Vogelmann and Howard 2001). In this study we ended up using only the area of
coastal herbaceous wetlands (marshes), since the area of woody wetlands was found not to
correlate with relative damages (see below). Wetland area for Louisiana was adjusted for years
other than the year 2000 (the year of the land use data) to take into account the recent rate of
coastal wetland loss of 65 km2/yr; 5) Total damage information for 34 hurricanes considered to
be “disasters” from the Emergency Events Database1. The damage included both direct (e.g.
damage to infrastructure, crops, housing) and indirect (e.g. loss of revenues, unemployment,
market destabilization) consequences on the local economy. We adjusted these data for inflation
to convert them to 2004 dollars based on the U.S. Department of Commerce implicit Price
Deflator for Construction.
100km x 100km hurricane swaths were then overlaid on the spatially explicit GDP and
wetland cover (herbaceous and woody wetlands were measured separately) to obtain GDP and
wetland area in each swath (Figure 1). The 100km x 100km swath was used as an average spatial
extent in order to standardize the calculations. The GDP calculated within each hurricane swath
and the reported total economic damage (TD) were used to generate a ratio, TD/GDP, which was
used to represent the relative economic damage caused by each hurricane. Pearson’s coefficients
showed insignificant collinearity between the analyzed variables. In addition no other land use
types’ area’s co-varied with either herbaceous or woody wetlands, which were also not
correlated with each other.
3
US GDP data from: http://www.bea.gov/bea/dn/home/gdp.htm
5
Results
Using ordinary least squares (OLS) we fit nine alternative multiple regression models
using the natural logs of wind speed and area of coastal herbaceous and woody vegetation as the
independent variables and TD/GDP as the dependent variable. We tested various log-log
formulations using maximum likelihood and Akaike’s Information Criterion (AIC) corrected for
small sample size. Our theoretical assumption was that TD should be proportional to GDP and
thus we used ln(TD/GDP) as the dependent variable. We did compare this to models with
ln(TD) as the dependent variable and ln(GDP) as one of the independent variable. However,
these models did not have a sufficient increase in likelihood as to merit the additional degree of
freedom as measured by AIC. The addition of woody wetlands to the model also did not
improve the model, but the addition of herbaceous wetlands yielded the best model with an AIC c
= 2.91 versus a model with just wind speed. The AIC evidence ratio implies that the model with
wetlands included has an 81% chance of being the correct model4.
The final model we used was:
ln (TDi /GDP i)=  + 1 ln(gi) +  2ln(wi) + ui
(1)
Where:
TDi = total damages from storm i (in constant 2004 $US);
GDPi = Gross Domestic Product in the swath of storm i (in constant 2004 $US). The swath was
considered to be 100 km wide by 100 km inland.
gi = maximum wind speed of storm i (in m/sec)
wi = area of herbaceous wetlands in the storm swath (in ha).
ui = error
This model had an adjusted R2 of 0.604 and was highly significant. The best fit
coefficients for the model are shown in Table 1. A 95% confidence interval was generated for
each of the parameter estimates as well as the adjusted R2 measure using a 10,000 iteration
6
bootstrap analysis. The adjusted R2 interval was (0.342, 0.832). The intervals for the other
parameters are listed in Table 1. The coefficient for wind speed (1) was significant 99.94% of
the time, while the coefficient for wetlands (2) was significant 93.04% of the time. The data
used in the model are included in Table 2.
As expected, increasing wind speed increased relative damages (TD/GDP), while
increasing herbaceous wetland area decreased them. Figure 2 shows the observed vs. predicted
relative damages from the model, with each of the hurricanes identified.
One potential problem with this formulation is endogeneity. If wetland area and GDP are
negatively correlated (as one might expect if urban area and wetland area were “competing” for
the same fixed landscape area), then this could cause bias in the estimate of the coefficient for
wetlands or spurious correlation. For example, it might be the case that high wetland area
correlates with lower GDP which correlates with less TD. We tested for the possibility that the
reduction in damage attributable to herbaceous wetlands was spurious and caused by an
endogenous relationship with GDP. GDP and herbaceous wetland area (all variables assumed
log-transformed) were weakly and positively correlated (r = 0.19).
However, the partial
correlation coefficient of total damage upon herbaceous wetlands controlling for GDP and wind
speed was negative (r = 0.34, p=0.05). Further, if an endogenous relationship with GDP were
causing a spurious effect of herbaceous wetlands upon total damage, a similar relationship
should have been seen with woody wetlands. However, the addition of woody wetlands did not
improve any of the models we tested.
A second potential problem with the formulation used is selection bias. It might be the
case that we are using data that is selectively gathered from locations where storm protection
services are disproportionately high (and thus are being studied more intensively) leading to
overestimates of the values. Unlike the case of selection bias in typical valuation studies, in
damage mitigation studies, selection bias tends to cause underestimates of value. We tested this
One could multiply all values by 0.81 to account for this as per Burnham and Anderson’s
(2003) multi-model inference. We did not do this, but leave it for the reader to decide if this
further adjustment is warranted.
4
7
by creating a Monte Carlo simulated data set using a log-log formulation and comparing
regression results when the data was restricted to samples with damages above a given
percentile. The higher the percentile, the lower the estimated damage mitigation of wetlands
implying our estimates are conservative.
We can rearrange equation (1) to estimate the total damages from hurricane i as:


TDi  e  gi 1  wi 2 GDPi

(2)
One can clearly see from this form of the relationship the relative influence of GDP, wind speed
and wetland area on total damages. Total damages vary linearly with GDP, as one might expect
since the more infrastructure there is to be damaged the more damage one can expect. TD also
varies as the 1 power of wind speed. The value of 1 from Table 1 indicates that total damages
increase as the 3.878 power of wind speed, fairly consistent with the well-known relationship
that the power in wind varies as the cube of speed. The value of 2 in Table 1 of -0.77 indicates
that total damages decrease quite rapidly with increasing wetland area.
The difference in TDi with a loss of an area a of wetlands (i.e. the “avoided damage” per unit
area of wetland) is then:


TDi  MVi  e  gi 1  (w i  a)  2  w i

2
 GDP
i
(3)
If a is small relative to w, this represents the estimated “marginal value” (MV) per unit area of
coastal wetlands in preventing storm damage from a specific hurricane. Table 2 lists MVi for
each of the 34 hurricanes in the database for unit areas of 1 ha. The values range from a
minimum of $23/ha for hurricane Bill to a maximum of $465,730/ha for hurricane Opal, with an
average value of just over $33,000/ha. The median value was just under $5,000/ha, indicating a
quite skewed distribution, mirroring the skewed distribution of damages. For each hurricane, we
also calculated an upper and lower bound on the marginal value estimate by applying the
formula for marginal value to each combination of regression parameters and taking the 95th
percentiles (as reported in Table 1).
8
Equation 3 allows one to estimate the avoided damages from any area of wetlands (a) up
to the total area of wetlands in the swath. For example, one might be interested in the “average”
value of a larger area of wetlands, say half the wetlands in the swath. This could be estimated by
using a = 1/2 of the total area of wetlands in the swath in equation 3 and then dividing the result
by 1/2 the total area of wetlands in the swath. The average values per hectare calculated in this
way (using 50% of the wetland area) are consistently 1.8 times higher than the marginal values.
Next we estimated the annual value of coastal wetlands for storm protection. This
required an estimate of the annual probability of being hit by hurricanes of various intensities.
We used data on historical frequencies by state as proxies for these probabilities. Data for each
of the 19 states in the US that have been hit by a hurricane since 1980 (267 total hits) were used
by Blake et al. (2005) to calculate the historical frequency of hurricane strikes by storm category.
We calculated the average GDP and wetland area in an average swath through each state using
our GIS database. We then calculated the annual expected marginal value (MV) for an average
hurricane swath in each state using the following variation of equation (3):
5

MVsw   pc,s  e gc 1  (w sw 1)  2  w sw
c1


2
 GDP
sw
(4)
where:
s = state
sw = average swath in state s
gc = average wind speed of hurricane of category c
pc,s = the probability of a hurricane of category c striking state s in a given year
GDPsw = the GDP in state s in the average hurricane swath
wsw = the wetland area in state s in the average hurricane swath
We then estimated total annual value of wetlands for storm protection as the integral of the
marginal values over all wetland areas (the “consumer surplus”). While residual analysis
suggests little correlation between wetland area and model error and we have several data points
9
with low wetland area, the log-log formulation makes marginal value calculations at low wetland
areas problematic because values go to infinity as area goes to 0. For this reason, we capped
increases in marginal value in our calculations for wetland areas below a certain number, k. In
other words, in the integration we used the marginal value at k for all wetland areas less than k
rather than allowing it to clime to infinity. Given model fit across the range of areas, we believe
this yields a conservative value estimate. We used k=10,000, but also report values in Table 3
using k = 50,000 and k = 1,000 to demonstrate sensitivity to this assumption. TV increases
dramatically as k decreases.
We can then estimate the average annual value per wetland hectare per state as:
AVs  TVs /ws

(5)
The estimated annual marginal value in an average swath (MVsw), the total value (TVs) for all
the state’s wetlands, and the average annual value per hectare (AVs) of coastal wetlands for each
state estimated in this way are shown in Table 3.
Using this technique, the mean annual marginal value per hectare in a typical swath across states
(MVsw) was almost $40,000/ha/yr, with a range from $126/ha/yr (for Louisiana) to
$586,845/ha/yr (for New York) and a median value of $1,700, indicating a quite skewed
distribution. MV varied inversely with wetland area (Figure 3A), indicating that the per hectare
value of wetlands increases as they become more scarce. The expected mean of the total annual
values by state (TVs) for all the state’s wetlands, over the 19 states assuming k=10,00 was $672
million/yr ($145 million at k=50,000 and $4.6 Billion at k=1,000), with a median of $72 million,
again reflecting a skewed distribution across states. The total annual value summed over all
states was $12.8 Billion/yr at k=10,000 ($2.8 Billion/yr at k=50,000 and $87.3 Billion/yr at
k=1,000). TVs generally increased with total area of wetlands in the state (Figure 3B).
Differences by state in both Figures 3A and 3B reflect the relative amounts of coastal
infrastructure vulnerable to damage and relative storm probabilities. For example, Louisiana has
the most wetlands, but less vulnerable infrastructure than Florida and Texas, while New York,
Massachusetts, and Connecticut had fewer wetlands but more vulnerable infrastructure. The
10
mean AVs by state was a little over $3,292/ha/yr, with a range from $108/ha/yr (for Delaware) to
$14,984/ha/yr (for New York) and a median value of $1,600/ha/yr.
We also estimated the annual values of wetlands for storm protection using spatially explicit data
on historical storm tracks to estimate probabilities of being hit by storms of a particular category
for each pixel along the coast. For this analysis we had data on 52 storms, resulting in
frequencies that did not match the state level frequencies exactly and were distributed across the
states. The advantage is that it allowed us to map wetland storm protection values at much higher
spatial resolution. For this application, a circle with radius 50 km was drawn around each 1km x
1km pixel within 100km of the coast, the wetland area and GDP in the circle was measured, and
equation 5 was applied, with the result for each pixel multiplied by the area of wetland in the
pixel divided by the area of wetland in the 50 km radius circle. Figure 4 maps the total value per
1km x 1km pixel estimated in this way. Finally, we summed over all pixels in each state to yield
estimates comparable with the “state level” analysis discussed earlier.
The state totals
aggregated from the “pixel level” analysis had an adjusted R2 of 0.88 compared with the state
totals from the “state level” analysis. Figure 4 shows wetlands of particularly high storm
protection value density at the intersection of high storm probability, high coastal GDP, and high
wetland area. For example, Southeast Florida, coastal Louisiana, and parts of Texas all show
high values. Connecticut, Massachusetts, and Rhode Island also show fairly high values, due
largely to the very high levels of coastal GDP in those states.
Discussion
Previous estimates of the value of coastal wetlands for storm protection used similar, but less
spatially explicit, methods and dealt only with hurricanes striking Louisiana (Costanza et al.
1989).
They estimated an annual average value of storm protection services of about
$1,000/ha/yr (converted into $2004). Our current analysis included a much larger database of
more recent hurricanes covering the entire US Atlantic and Gulf coasts and was able to utilize
much more spatially explicit data on hurricane tracks, wetland area, and GDP.
Our
corresponding estimated value for Louisiana was also about $1,000/ha/yr, very consistent with
the previous estimate. Our current study allows one to assess not only the value of wetlands for
storm protection, but how that value varies with location, area of remaining wetlands, proximity
11
to built infrastructure, and storm probability, thus providing a richer and more useful analysis of
this important ecosystem service. It also allows us to state the ranges of values that would result
from varying parameter assumptions and the confidence intervals on the estimates.
The results also allow straightforward assessments of impacts on wetlands. For example,
Louisiana lost an estimated 480,000 hectares of coastal wetlands prior to Katrina and 20,000
hectares during hurricane Katrina itself. The value of the lost storm protection services from
these wetlands can be estimated as the value per hectare ($1,000 ha/yr from Table 3) times the
area, yielding approximately $480 million/yr for services lost from wetlands lost prior to Katrina
and an additional $20 Million/yr for wetlands lost during the storm. Converting this total of
$500 million/yr to present value terms using a 3% discount rate implies a lost value for just the
storm protection service of this natural capital asset of $16 Billion, and the lost value of the
wetlands loss due to Katrina of $666 million.
If the frequency and intensity of hurricanes increases in the future, as some are predicting as a
result of climate change, then the value of coastal wetlands for protection from these storms will
also increase. Coastal wetlands provide “horizontal levees” that are maintained by nature and
are far more cost-effective than constructed levees. The experience of hurricane Katrina provided
a tragic example of the costs of allowing these natural capital assets to degrade. Coastal
wetlands also provide a host of other valuable ecosystem services that constructed levees do not.
They have been estimated to provide about $11,700/ha/yr (in $2004) in other ecosystem services
(excluding storm protection – Costanza et al 1997), and experience (including the current study)
has shown that as we learn more about the functioning of ecological systems and their
connections to human welfare, estimates of their value tend to increase.
Investing in the
maintenance and restoration of coastal wetlands is proving to be an extremely cost-effective
strategy for society.
Acknowledgements.
MLM and OPM are grateful to Instituto de Ecología, A.C. for financial support during their
research stay at the Gund Institute of Ecological Economics (University of Vermont). We are
grateful to Joshua Farley, Brendan Fisher, and the students in the Advanced Course on
12
Ecological Economics (University of Vermont) for their helpful comments on earlier versions of
this text.
References
Blake, E. S., Rappaport, E. N., Landsea, C. W., and Jerry G. J. 2005. Deadliest, Costliest, and
Most Intense United States Tropical Cyclones from 1851 to 2004 (NOAA Technical
Memorandum NWS TPC-4). Miami, Florida: National Weather Service, National
Hurricane
Center,
Tropical
Prediction
Center.
Updated
August.
http://www.aoml.noaa.gov/hrd/Landsea/dcmifinal2.pdf
Burnham, K. P. and D. Anderson. 2003. Model selection and multi-model inference. Springer,
New York.
Costanza, R., d'Arge, R., de Groot, R., Farber, Grasso, S. M., Hannon, B., Naeem, S., Limburg,
K. Paruelo, J. O'Neill, R.V., Raskin, R., Sutton, P., and van den Belt, M. 1997. The value
of the world's ecosystem services and natural capital. Nature 387: 253-260.
Costanza, R., Farber, S. C. and Maxwell. J. 1989. The valuation and management of wetland
ecosystems. Ecological Economics. 1: 335-361.
Danielsen, F., et al. 2005. The Asian tsunami: a protective role for coastal vegetation. Science:
310: 643.
Elvidge, C.D., Baugh, K.E., Dietz, J.B., Bland, T., Sutton, P.C. and Kroehl, H.W. 1999.
Radiance Calibration of DMSP-OLS low-light imaging data of human settlements,
Remote Sensing of Environment 68: 77-88
Emanuel K. 2005. Increasing destructiveness of tropical cyclones over the past 30 years. Nature,
436: 686-688.
Simpson, R.H. and Riehl, H. 1981. The hurricane and its impact. Louisiana State University
Press, Baton Rouge, LA.
13
Sutton, P. C. and Costanza, R. 2002. Global estimates of market and non-market values derived
from nighttime satellite imagery, land use, and ecosystem service valuation. Ecological
Economics 41: 509-527
Vogelmann, J. E. and Howard S. M. 2001. Completion of the 1990's National Land Cover
Dataset for the conterminous United States from Landsat Thematic Mapper Data and
Ancillary Data Sources. Photogrammetric Engineering and Remote Sensing 64: 45-57.
Webster, P., Holland, J., Curry, G.J., and Chang, H. R. 2005. Changes in tropical cyclone
number, duration, and intensity in a warming environment. Science, 309:1844-1846
14
Table 1. Regression model coefficients including lower and upper 95% confidence intervals.
Parameter
Coefficient
Lower 95%
CI
Upper 95%
CI
Standard
Error
t
P

-10.551
-20.06
-3.36
3.29
-3.195
0.003

3.878
2.45
5.34
0.706
5.491
<0.001

-0.77
-1.06
-0.247
0.16
-4.809
<0.001
15
Table 2. Maximum wind speed, herbaceous wetland area, GDP, Total Damage and calculated
marginal value (per hectare of wetland) for storm protection for each hurricane used in
the regression analysis. Lower and upper 95% confidence intervals on the marginal
values are also shown.
Hurricane
Alberto
Alicia
Allen
Allison
Allison
Andrew
Bill
Bob
Bonnie
Bret
Chantal
Charley
Charley
Danny
Dennis
Year
1994
1983
1980
1989
2001
1992
2003
1991
1998
1999
1989
1998
2004
1997
1999
Elena
Emily
Erin
1985
1993
1995
Floyd
1999
Fran
Frances
Gaston
Gloria
Hugo
Irene
1996
2004
2004
1985
1989
1999
Isabel
Isidore
2003
2002
Ivan
Jeanne
Jerry
Katrina
Keith
Lili
Opal
2004
2004
1989
2005
1988
2002
1995
States hit
GA, MI, FL, AL
TX
TX
TX, LA, FL, NC, PA, VA
TX, LA, FL, NC, PA, VA
FL, LA
LA, MS, AL, FL
NC, ME, NY, RI, CT, MA
NC, SC, VA
TX
TX
TX
FL
OH, PA, IL, NY, NJ
NC
FL, AR, KY, SD, IO, MI, IN,
MO
Max
wind
Herbaceous
speed Wetland area
(m/sec) in swath (ha)
28.3
4,466
51.4
93,590
84.9
26,062
23.1
167,494
25.7
100,298
69.4
901,819
25.7
642,544
51.4
68,465
51.4
49,774
61.7
29,695
36.0
104,968
25.7
55,126
64.3
358,778
36.0
271,317
46.3
22,752
GDP in
swath hit
year (2004
$US
millions)
5,040
100,199
13,151
149,433
185,610
83,450
70,669
122,358
15,840
2,043
81,319
18,775
483,281
66,711
17,669
Observed
total
damage
(2004 $US
millions)
305
2,823
1,674
63
6,995
34,955
17
829
373
35
111
33
6,800
111
45
Estimated
marginal
value (MV)
per ha (2004
$US/ha)
15,607
14,449
127,090
348
1,611
699
23
30,683
6,984
4,557
2,400
470
15,347
367
20,704
MV lower
95% CI
(2004
$US/ha)
1,254
5,316
29,204
85
377
318
8
9,982
2,008
1,119
763
90
7,918
158
4,199
MV upper
95% CI
(2004
$US/ha)
59,663
22,146
283,261
938
4,042
1,247
54
48,743
11,585
8,368
4,351
1,254
24,062
622
40,396
FL, AL, MS
56.6
51.4
41.2
50,568
615
264,226
14,240
6
132,138
1,774
38
821
8,835
5,795
1,278
2,629
272
610
14,698
24,748
1,967
NC, FL, SC, VA, MD, PA,
NJ, NY, DE, RI, CT, MA, VT
69.4
188,637
420,940
7,259
56,214
26,075
93,566
54.0
64.3
30.9
64.3
72.0
46.3
9,033
340,051
100,502
87,863
32,906
692,219
10,471
150,986
82,063
188,531
13,684
114,903
3,900
4,400
62
1,451
1,391
104
114,389
5,272
1,439
72,229
46,288
319
16,760
2,710
402
27,350
11,868
179
259,346
8,270
2,999
119,390
89,485
488
72.0
56.6
37,942
574,157
35,068
64,990
5,406
79
92,176
547
24,987
296
175,244
830
74.6
56.6
38.6
78.2
33.4
64.3
66.9
504,033
404,769
98,540
708,519
222,324
224,504
7,261
226,150
133,657
86,173
214,277
55,856
24,439
12,652
6,000
7,000
49
22,321
44
295
3,521
6,996
2,088
3,717
4,363
328
1,779
465,730
3,204
1,148
1,209
1,847
126
881
64,749
12,550
3,096
6,450
8,429
594
2,798
1,111,043
52
53
17
218,995
100,400
243,111
99,905
75,994
110,816
3,561
825
7,001
33,268
4,914
83,466
7,356
1,231
13,403
71,962
8,398
196,765
NC
NC, SC, VA, MD, VA, PA,
OH, Washington DC
FL, NC, SC, OH
VA, SC, NC
NC, NY, CT, NH, ME
SC
FL
NC, MD, VA, Washington
DC
LA, MS, AL, TN
AL, LA, MS, FL, PA, MD,
NJ, OH, NC, VA, GA, TN
FL
TX
AL, LA, MS, GA, FL
FL
LA
FL, GA, AL
Mean
Median
S.D.
16
Table 3. Coastal wetland area in each state, wetland area and GDP in the average 100km swath, estimated annual storm probabilities
by category, and calculated MVsw, TVs and AVs .
Wetlands
Wetland
Probability of state being hit by a storm of the
GDP in
within 100
area in
given category in a year by Storm Category
average
km of coast average
swath ($
by state Ws swath W sw
millions/year)
(ha)
(ha)
State
Alabama
Connecticut
Delaware
Florida
Georgia
Louisiana
Maine
Maryland
Massachusetts
Mississippi
New Hampshire
New Jersey
New York
North Carolina
Pennsylvania
Rhode Island
South Carolina
Texas
Virginia
Mean
Median
S.D.
16,759
21,591
33,964
1,433,286
140,556
1,648,611
60,388
60,511
49,352
25,456
19,375
69,001
5,306
64,862
7,446
3,638
107,894
448,621
71,509
6,388
12,601
12,089
186,346
29,120
370,299
15,500
16,011
16,801
6,048
9,905
21,864
2,117
21,295
2,994
1,759
39,177
79,110
23,588
9,499
65,673
10,488
70,491
7,356
36,250
14,670
21,924
67,266
3,890
23,051
78,703
90,770
13,023
93,117
12,810
15,367
63,661
27,786
1
7.14%
2.60%
1.30%
27.92%
7.79%
11.04%
3.25%
0.65%
3.25%
1.30%
0.65%
1.30%
3.90%
13.64%
0.65%
1.95%
12.34%
14.94%
5.84%
225,691
60,388
475,157
45,948
16,011
89,175
38,200
23,051
31,083
6.39%
3.25%
7.01%
2
3
3.25% 3.90%
1.95% 1.95%
0.00% 0.00%
20.78% 17.53%
3.25% 1.30%
9.09% 8.44%
0.65% 0.00%
0.65% 0.00%
1.30% 1.95%
3.25% 4.55%
0.65% 0.00%
0.00% 0.00%
0.65% 3.25%
8.44% 7.14%
0.00% 0.00%
1.30% 2.60%
3.90% 2.60%
11.04% 7.79%
1.30% 0.65%
3.76%
1.30%
5.25%
3.35%
1.95%
4.39%
4
0.00%
0.00%
0.00%
3.90%
0.65%
2.60%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.65%
0.00%
0.00%
1.30%
4.55%
0.00%
Annual
expected
marginal
value per
average
swath MV sw
($/ha/yr)
Average
Total annual value per state (TV s) annual value
for k=50,000, 10,000, and 1,000
of wetlands
($ millions/yr)
per ha per
state (AVs)
at k=10,000
($/ha/yr)
k=50,000
k=10,000
k=1,000
5
0.00%
0.00%
0.00%
1.30%
0.00%
0.65%
0.00%
0.00%
0.00%
0.65%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
14,155
14,428
222
1,684
630
126
715
445
8,422
7,154
1,095
583
586,845
5,072
11,651
95,193
1,281
3,901
1,555
2.4
15.9
0.2
1,579.4
7.1
439.4
1.4
0.9
20.5
1.0
0.6
3.0
4.6
23.8
0.2
0.4
32.6
609.2
9.7
40.9
263.2
3.7
6,453.9
72.3
1,665.7
21.3
14.3
301.2
17.8
10.7
37.0
79.5
304.1
4.1
7.7
265.0
3,087.1
115.6
749.5
2,705.2
38.6
40,010.3
542.2
10,107.2
196.0
129.4
2,673.1
341.5
131.7
298.9
3,473.3
2,477.0
141.3
377.0
1,880.0
20,144.4
914.1
2,440.7
12,191.7
107.5
4,502.9
514.0
1,010.4
353.0
236.0
6,103.8
697.9
550.4
536.1
14,983.7
4,687.8
554.2
2,122.4
2,456.5
6,881.2
1,617.1
0.72% 0.14%
0.00% 0.00%
1.40% 0.35%
39,745
1,684
134,195
144.9
4.6
384.6
671.8
72.3
1,594.0
4,596.4
749.5
9,848.0
3,292.0
1,617.1
4,190.4
Totals
2,752.4
12,765.0
87,330.7
17
Figure 1. Typical hurricane swath showing GDP and wetland area used in the analysis
18
1.0000
Emilly 1993
Opal 2005
Allen 1980
Fran 1996
Hugo 1989
Bret 1999
0.1000
Isabel 2003
TD/GDP predicted
Gloria 1985
Elena 1985
Floy d 1999
Bonnie 1998
Dennis 1999
Lili 2002
Bob 1991
Charley 2004
Ivan 2004
Alicia 1983
Frances 2004
Katrina 2005
Alberto 1994
0.0100
Isidore2002
Jerry 1989
Jeanne 2004
Andrew 1992
Chantal 1989
Irene 1999
Gaston 2004
Danny 1997
Keith 1988
Charley 1998
Erin 1995
Allison 2001
0.0010
Allison 1989
Bill 2003
0.0001
0.0001
0.0010
0.0100
0.1000
TD/GDP observed
1.0000
10.0000
Figure 2. Observed vs. predicted relative damages (TD/GDP) for each of the hurricanes used in the analysis.
19
$1,000,000
A
Annual expected marginal value per average swath MV sw
($/ha/yr)
New York
$100,000
Rhode Island
Alabama
Connecticut
Pennsylvania
$10,000
Massachusetts
Mississippi
North Carolina
Texas
Virginia
New Hampshire
South Carolina
$1,000
Maine
Florida
Georgia
New Jersey
Maryland
Delaware
Louisiana
$100
1,000
10,000
100,000
1,000,000
Area of coastal wetlands in average swath (ha)
Total Annual Value of Coastal Wetlands (TVs Million$/yr)
$100,000
y = 0.0093x
R2 = 0.5788
Florida
B
Texaas
$10,000
Louisiana
New York
Connecticut Massachusetts
North Carolina
$1,000
South Carolina
Virginia
Alabama
Georgia
Mississippi
Rhode Island
New Jersey
Maine
$100
New Hampshire
Pennsylvania
Maryland
Delaware
$10
1,000
10,000
100,000
1,000,000
10,000,000
Total Area of Coastal Wetlands (ha)
Figure 3. Area of coastal wetlands (A) in the average hurricane swath vs. the
estimated marginal value per ha (MVsw) and (B) in the entire state vs.
the total value (TVs) of coastal wetlands for storm protection.
20
Figure 4. Map of total value of coastal wetlands for storm protection by 1 km x 1 km pixel.