Valuing the storm surge protection service of US Gulf Coast wetlands

Abstract

Current plans for US Gulf Coast wetland restoration assume that wetlands can save lives and property by reducing storm surges. However, there have been few economic valuations of this benefit for Gulf Coast wetlands. We develop a methodology for estimating the value of wetlands in reducing expected property damages from hurricane flooding that relates damages to the distribution of storm events and incorporates both the wetland characteristics of wave attenuation and offshore storm surge properties. We apply this methodology to value the storm surge protection service of coastal marshes, in terms of reducing expected property damages, along the path of a storm south-east Louisiana, which includes New Orleans. We conclude by discussing the implications of this analysis for further research on the economic value of wetlands in protecting coastal property and for restoration policy.

Keywords:

• coastal wetlands
• environmental valuation
• storm protection
• storm surge
• US Gulf Coast
• wetland restoration

JEL code:

• Q57

Acknowledgements

Research for this paper was funded by the Coastal Restoration and Enhancement through Science and Technology (CREST) Program, an alliance of academic institutions within southern Louisiana and Mississippi, under Grant Contract # CREST10-6. We thank Denise, Reed of the University of New Orleans, for sub-contracting our economic analysis through the CREST grant and for providing valuable advice. We are grateful to Ioannis Georgiou for extracting the hydrodynamic data on storm surge levels and for providing estimates of wetland continuity and roughness used in our analysis. We also thank Keven Lovetro of the US Army Corps of Engineers (USACE), New Orleans, LA, for providing us with the USACE Residential Depth-Damage Function data base for southern Louisiana, and the Hurricane Protection Office (HPO) for providing access to storm surge simulations.

Notes

1. Using historical storm frequencies for the Louisiana Gulf Coast, Farber (1987) estimates the expected wind damage to property from the loss of intervening marsh. The present value of the loss of a 1 mile strip of wetlands amounts to between $1.1 and $3.7 million (1980 dollars). The increased cost to property damage amounted to between $7 and $23 per acre. However, the study by Farber (1987) did not estimate the value of the marsh in protecting against hurricane storm surge.

2. Laso-Bayas et al. (2011) is not strictly an economic valuation study, as it examines the impacts of storm surge on casualties and structural damage, with the latter is represented solely by a binomial variable.

3. If the time period is sufficiently large (e.g., a hurricane season, year, or longer), the household might experience more than one storm surge flooding event.

4. Although we ignore any possible psychological effects due to property losses from storm damages (Merz et al. 2010), this is a standard assumption in a variety of approaches to modelling such damages (e.g., see Barbier 2007; Bengtsson and Nilsson 2007; Farber 1987; Smith, Carbone, Pope, Hallstrom, and Darden 2006). As we approach the valuation problem from the standpoint of a household's ex-ante decision making, it is impossible for the household to know in advance the likely psychological consequences of any storm damage to property.

5. Although the assumption of a Poisson distribution is employed in Barbier (2007), the Poisson process is not explicitly modelled. In contrast, we do model explicitly the Poisson process, showing how our model may be obtained from the original assumptions in Barbier (2007). See also Katz (2002) for a more in-depth discussion of the expected value of periodic damage.

6. Formally, using the definition of the two-dimensional Poisson process, the sequence of surge events can be described as a point process on some set, A where τ_z is the intra-period arrival time of the zth surge, and . Z_t is therefore equal to the number of points in set A and represents the number of storm events in period t during which g > a (i.e. the number of surge-damage events), and for the purpose of this analysis, may be assumed independent of per surge damage F.

7. We are grateful to Ioannis Georgiou of the Department of Earth and Environmental Sciences and Pontchartrain Institute for Environmental Sciences, University of New Orleans for extracting the hydrodynamic data on storm surge levels for the hypothetical hurricane simulations and storm surge transect in the Caernarvon Basin and for providing estimates of wetland continuity and roughness for the 12 segments of the transect used in our analysis.

8. The relative characteristics of the four simulated hurricanes are

Storm A = Central pressure of 96 kilopascals (kPa), radius to maximum winds (R_max) of 67 km, forward speed of 20.5 km/hr, return period is less than 50 years.

Storm B = Central pressure of 93 kPa, radius to maximum winds (R_max) of 47 km, forward speed of 20.5 km/hr, return period is less than 50 years.

Storm C = Central pressure of 96 kPa, radius to maximum winds (R_max) of 46 km, forward speed of 20.5 km/hr, return period is more than 50 years.

Storm D = Central pressure of 93 kPa, radius to maximum winds (R_max) of 33 km, forward speed of 11.1 km/hr, return period is less than 50 years.

9. The transects do not include any dry land; i.e. they are either open water or marsh wetland.

10. Note that our empirical estimation further controls for the possible influence of coastal locational characteristics on storm surge levels as they pass through wetlands by using the same storm track for each of the four storms we analyse. In addition, we examined for the effects of distance on storm surge attenuation by including transect segment distance x as an independent variable, as well as interacted with W_C and W_V, in different versions of regression model (13), but these specifications of the model proved to be considerably less robust and the coefficients for x were not significant.

11. The 15 parishes are Assumption, Iberville, Jefferson, Lafourche, Livingston, Orleans, Plaquemines, St. Bernard, St. Charles, St. James, St. John the Baptist, St. Martin, St. Tammany, Tangipahoa, and Terrabonne.

12. As a result, we use 1 as the initial value for the associated parameter in the non-linear regression algorithm.

13. Substituting parameter values from the non-linear regression in Table 4, mean elevation, and property value, and taking the derivative of Equation (17) with respect to S we obtain

14. The SURGEDAT data-set is available from the Climate Center at Louisiana State University, and can be downloaded from the site http://surge.srcc.lsu.edu/.

15. Specifically, we may now evaluate the following integral:

16. As the β_i coefficients estimated in regression (13) are for the change in surge measured in metres (see Tables 1 and 2), the coefficients were converted to feet, to be consistent with the regression (17) based on USACE (2006) flood damage estimates, as the latter data indicate the effect of flood level in feet on property damage. However, we report the final estimates in per metre units in Table 3(B) to be consistent with Tables 1, 2, and 3(A).