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
Coastal wetlands reduce wave height and storm surge and absorb wind energy thus
moderating the damaging effects of hurricanes on coastal communities. We assembled
data on storm damage, GDP in the storm track, wind speed, and coastal wetland area for
34 major hurricanes that have struck the US since 1980 in order to estimate the effects of
coastal wetlands on moderating storm damage. A regression model 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 coastal
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 being
struck by hurricanes of varying intensities, we estimated the annual value of coastal
wetlands by state for storm protection. This value ranged from $600 to $237,000/ha/yr,
with a mean of $26,000/ha/yr (median = $8,000/ha/yr) much larger than previous estimates.
Overall, coastal wetlands in the US were estimated to currently provide $48 Billion/yr in
storm protection services, a value that may change significantly in the future if wetlands
continue to be lost or if storm frequency and intensity increase due to climate change.
2
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 hurricanes2, and an upward trend in tropical cyclone
destructive potential3. Coastal wetlands reduce the damaging effects of hurricanes on coastal
communities by absorbing storm energy in ways that neither solid land nor open water can4.
There is also evidence that property damage and loss of human lives from the 2004 tsunami that
hit Southeast Asia was highly ameliorated by coastal ecosystems5
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 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 GIS data on Gross Domestic Product (GDP) and coastal wetland area in each
storm’s swath (see Figure 1 and Methods for more details).
Using ordinary least squares (OLS) we fit the following multiple regression model6, which
had an adjusted R2 of 0.604 and was highly significant:
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.
3
gi = maximum wind speed of storm i (in m/sec)
wi = area of herbaceous wetlands in the storm swath (in ha).
ui = error
The best fit coefficients for the model are shown in Table 1. The data used in the model
are included in Table 2.
As expected, increasing wind speed increased relative damages (TD/GDP), while
increasing wetland area decreased them. Figure 2 shows the observed vs. predicted relative
damages from the model, with each of the hurricanes identified.
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 wind speed and
wetland area on total damages. 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 difference in TDi with a loss of an area a of wetlands (i.e. the “avoided cost” 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” 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
4
over $33,000/ha. The median value was just under $5,000/ha, indicating a quite skewed
distribution, mirroring the skewed distribution of damages. 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 information about the annual probability of being hit by hurricanes of various
intensities. We used data on historical frequencies by state as proxies for these probabilities7.
For each of the 19 states in the US that have been hit by a hurricane since 1980 (267 total hits),
we calculated the average GDP and wetland area in their swaths using our GIS data base. 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
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
(4)
5
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) from the first or “marginal”
hectare in the swath down to a value k, as:
wk
5
TVs   MV j  ((k)  (wsw ) )   pc,s  e gc 1  GDPsw
j1
2
2

(5)
c 0
Residual analysis of our regression results suggested that our model over-predicted the damage

mitigation for smaller wetland areas, so we only integrated MV down to k = 1000 ha. This yields
a conservative estimate of total damage avoided or total value.
We can then use this to estimate the average annual value per wetland hectare per state as:
AVs  TVs /ws
(6)
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.
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 mean of the total annual values by state (TVs) for all the state’s
wetlands, over the 19 states was $2.5 Billion/yr (± $5.5 Billion), with a median of $370 Million,
again reflecting a skewed distribution across states. The total annual value summed over all
states was almost $48 Billion/yr. TV 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 has fewer wetlands but more vulnerable infrastructure. The mean AVs by state was a
6
little over $26,000/ha/yr, with a range from $596/ha/yr (for Delaware) to $237,431/ha/yr (for
New York) and a median value of $8,000/ha/yr.
Previous estimates of the value of coastal wetlands for storm protection8 used similar, but less
spatially explicit methods and dealt only with hurricanes striking Louisiana. 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
about $3,500/ha/yr, significantly larger than the previous estimate. Our current study also 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 to built infrastructure, and storm probability,
thus providing a richer and more useful analysis of this important ecosystem service. 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 ($3,500 ha/yr) times the area,
yielding approximately $1.68 Billion/yr for services lost from wetlands lost prior to Katrina and
an additional $70 Million/yr for wetlands lost during the storm. Converting this total of $1.75
Billion/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 $58 Billion, and the lost value of the additional
wetlands loss due to Katrina of $2.3 Billion.
If the frequency and intensity of hurricanes increase 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)9, 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,
7
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.
Methods
The following datasets were assembled for the analysis: 1) tracks of all hurricanes striking the
US from 1980 to 2004, which included wind-speed10. Of these, only 33 hurricanes had sufficient
information on damages (see below) to include in the regression analysis. All of the storms were
used to estimate the strike probabilities, since this did not require damage information; 2)
Nighttime light imagery of the USA11. A one kilometer resolution GDP map was prepared by
using a linear allocation of the national GDP to the light intensity values of the nighttime image
composite12. GDP was adjusted to the year of the hurricane using national data on national GDP
by year13; 4) The 30 meter resolution National Landcover dataset, which included mapping of
both herbaceous and woody wetlands14. In this study we used the 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 from the year 2000
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 U.S.
Department of Commerce implicit Price Deflator for Construction6.
100km x 100km hurricane swaths were then overlaid on the spatially explicit GDP and wetland
cover 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 of TD/GDP, which was used to represent the relative economic damage caused by each
8
hurricane. Pearson’s coefficients showed insignificant collinearity between the analyzed
variables. Six alternative models were tried using the natural log of wind speed and various
combinations of linear and log forms of the area of coastal herbaceous and woody vegetation.
Akaike’s Information Criterion estimated the probability of the model we used being the
“correct” model at 97.9%. All models with linear forms of wetland area had less than 0.01%
chance of being the correct model.
Equation (5) estimates the avoided damages for each hurricane category in an average swath and
multiplies by the probability of a given storm striking a state in a year. Value is thus only
ascribed to wetlands that are expected to be in the swath of a hurricane with average value being
the expected damage mitigation divided amongst all wetlands in the state.
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) Webster, P., Holland, J., Curry, G.J., and Chang, H. R., Changes in tropical cyclone number,
duration, and intensity in a warming environment. Science, 309, 1844-1846 (2005)
3) Emanuel K. Increasing destructiveness of tropical cyclones over the past 30 years. Nature,
436, 686-688. (2005)
4) Simpson, R.H. and Riehl, H. The hurricane and its impact. Louisiana State University Press,
Baton Rouge (1981)
5) Danielsen, F., et al. The Asian tsunami: a protective role for coastal vegetation. Science, 310,
643. (2005)
9
6) We fit several alternative models and this was the best according to the AIC criteria (see
Methods)
7) Blake, E. S., Rappaport, E. N., Landsea, C. W., and Jerry G. J. 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 2005.
http://www.aoml.noaa.gov/hrd/Landsea/dcmifinal2.pdf
8) Costanza, R., Farber, S. C. and Maxwell. J. The valuation and management of wetland
ecosystems. Ecological Economics 1, 335-361. (1989)
9) 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. The value
of the world's ecosystem services and natural capital. Nature 387, 253-260 (1997).
10) www.grid.unep.ch/data/gnv199.php
11) Elvidge, C.D., Baugh, K.E., Dietz, J.B., Bland, T., Sutton, P.C. and Kroehl, H.W. Radiance
Calibration of DMSP-OLS low-light imaging data of human settlements, Remote Sensing of
Environment 68, 77-88 (1999)
12) Sutton, P. C. and Costanza, R. Global estimates of market and non-market values derived
from nighttime satellite imagery, land use, and ecosystem service valuation. Ecological
Economics 41, 509-527 (2002)
13) US GDP data from: http://www.bea.gov/bea/dn/home/gdp.htm
14) Vogelmann, J. E. and Howard S. M. 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 (2001).
10
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 Ecological Economics (University of Vermont)
for their helpful comments on earlier versions of this text.
Correspondence and requests for materials should be addressed to RC (e-mail: Robert.Costanza@uvm.edu)
11
Table 1. Regression model coefficients.
Coefficient
Std. Error
t
P

-10.551
3.29
-3.195
0.003
1
3.878
0.706
5.491
<0.001
2
-0.77
0.16
-4.809
<0.001
12
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.
Hurricane Year
Alberto
1994
Alicia
1983
Allen
1980
Allison
1989
Allison
2001
Andrew
1992
Bill
2003
Bob
1991
Bonnie
1998
Bret
1999
Chantal
1989
Charley
1998
Charley
2004
Danny
1997
Dennis
1999
Elena
1985
Emily
1993
Erin
1995
Floyd
1999
Fran
Frances
Gaston
Gloria
Hugo
Irene
Isabel
Isidore
1996
2004
2004
1985
1989
1999
2003
2002
Ivan
Jeanne
Jerry
2004
2004
1989
2005
1988
2002
1995
Katrina
Keith
Lili
Opal
Max wind
speed
(m/sec)
States hit
Herbaceous
Wetland
area in
swath (ha)
GDP in swath
in year hit
(2004 $US)
(millions)
Observed total
Estimated
damage (2004 marginal value
$ US) (millions) (2004 $US/ha)
FL, AL, MS
28.3
51.4
84.9
23.1
25.7
69.4
25.7
51.4
51.4
61.7
36.0
25.7
64.3
36.0
46.3
56.6
51.4
41.2
4,466
93,590
26,062
167,494
100,298
901,819
642,544
68,465
49,774
29,695
104,968
55,126
358,778
271,317
22,752
50,568
615
264,226
$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
$14,240
$6
$132,138
$305
$2,823
$1,674
$63
$6,995
$34,955
$17
$829
$373
$35
$111
$33
$6,800
$111
$45
$1,774
$38
$821
$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
$8,835
$5,795
$1,278
NC, FL, SC, VA, MD, PA, NJ,
NY, DE, RI, CT, MA, VT
69.4
188,637
$420,940
$7,259
$56,214
54.0
64.3
30.9
64.3
72.0
46.3
72.0
56.6
9,033
340,051
100,502
87,863
32,906
692,219
37,942
574,157
$10,471
$150,986
$82,063
$188,531
$13,684
$114,903
$35,068
$64,990
$3,900
$4,400
$62
$1,451
$1,391
$104
$5,406
$79
$114,389
$5,272
$1,439
$72,229
$46,288
$319
$92,176
$547
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
52
53
17
218,995
100,400
243,111
$99,905
$75,994
$110,816
$3,561
$825
$7,001
$33,268
$4,914
$83,466
GA, MI, FL, AL
TX
TX
TX, LA, FL, NC, PA, VA
TX, LA, FL, NC, PA, VA
FL, LA
LA, MS, AL, FL
NC, ME, NY, RI, CT, MA
NC, SC, VA
TX
TX
TX
FL
OH, PA, IL, NY, NJ
NC
FL, AR, KY, SD, IO, MI, IN, MO
NC
NC, SC, VA, MD, VA, PA, OH,
Washington DC
FL, NC, SC, OH
VA, SC, NC
NC, NY, CT, NH, ME
SC
FL
NC, MD, VA, Washington DC
LA, MS, AL, TN
AL, LA, MS, FL, PA, MD, NJ,
OH, NC, VA, GA, TN
FL
TX
AL, LA, MS, GA, FL
FL
LA
FL, GA, AL
Mean
Median
S.D.
13
Table 3. Coastal wetland area in each state, wetland area and GDP in the average 100km swath, estimated annual
storm probabilities by category, and calculated MVsw, TVs and AVs .
State
Alabama
Connecticut
Delaware
Florida
Georgia
Louisiana
Maine
Maryland
Massachusetts
Mississippi
New Hampshire
New Jersey
New York
North Carolina
Pennsylvania
Rhode Island
South Carolina
Texas
Virginia
Mean
Median
S.D.
Wetlands
within 100 Wetland area
GDP in
km of coast in average average swath
by state W s swath W sw
($
(ha)
(ha)
millions/year)
16,759
6,388
$9,499
21,591
12,601
$65,673
33,964
12,089
$10,488
1,433,286
186,346
$70,491
140,556
29,120
$7,356
1,648,611
370,229
$36,250
60,388
15,500
$14,670
60,511
16,011
$21,924
49,352
16,801
$67,266
25,456
6,048
$3,890
19,375
9,905
$23,051
69,001
21,864
$78,703
5,306
2,117
$90,770
64,862
21,295
$13,023
7,446
2,994
$93,117
3,638
1,759
$12,810
107,894
39,177
$15,367
448,621
79,110
$63,661
71,509
23,588
$27,786
225,691
60,388
475,157
45,944
16,011
89,160
$38,200
$23,051
$31,083
Probability of state being hit by a storm of
the given category in a year.
Storm Category
1
7.14%
2.60%
1.30%
27.92%
7.79%
11.04%
3.25%
0.65%
3.25%
1.30%
0.65%
1.30%
3.90%
13.64%
0.65%
1.95%
12.34%
14.94%
5.84%
6.39%
3.25%
7.01%
2
3
3.25% 3.90%
1.95% 1.95%
0.00% 0.00%
20.78% 17.53%
3.25% 1.30%
9.09% 8.44%
0.65% 0.00%
0.65% 0.00%
1.30% 1.95%
3.25% 4.55%
0.65% 0.00%
0.00% 0.00%
0.65% 3.25%
8.44% 7.14%
0.00% 0.00%
1.30% 2.60%
3.90% 2.60%
11.04% 7.79%
1.30% 0.65%
3.76%
1.30%
5.25%
3.35%
1.95%
4.39%
4
0.00%
0.00%
0.00%
3.90%
0.65%
2.60%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.65%
0.00%
0.00%
1.30%
4.55%
0.00%
5
0.00%
0.00%
0.00%
1.30%
0.00%
0.65%
0.00%
0.00%
0.00%
0.65%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.72% 0.14%
0.00% 0.00%
1.40% 0.35%
Annual
expected
Average annual
marginal value Total annual
value of
per average value per state wetlands per
swath MV sw
TVs
($
ha per state
($/ha/yr)
millions/yr)
AV s
($/ha/yr)
$14,155
$372
$22,209
$14,428
$1,425
$65,999
$222
$20
$596
$1,684
$22,419
$15,641
$630
$296
$2,105
$126
$5,682
$3,446
$715
$104
$1,729
$445
$69
$1,141
$8,422
$1,430
$28,968
$7,154
$168
$6,616
$1,095
$68
$3,522
$583
$162
$2,342
$586,845
$1,260
$237,431
$5,072
$1,338
$20,626
$11,651
$60
$8,071
$95,193
$118
$32,534
$1,281
$1,033
$9,578
$3,901
$11,202
$24,970
$1,555
$496
$6,930
$39,745
$1,684
$134,195
$2,512
$372
$5,526
$26,024
$8,071
$53,662