Coastal flooding in the Northeastern United States due to climate change

Abstract

With dense population and development along its coastline, the northeastern United States is, at present, highly vulnerable to coastal flooding. At five sea level stations in the United States, from Massachusetts to New Jersey, sea level rise (SLR) trends and tidal effects were removed from the hourly sea level time series and then frequency analysis was performed on the positive remaining anomalies that represent storm surge heights. Then using eustatic SLR estimates for lower and higher greenhouse gas emissions scenarios and assumed trends in local sea level rise, new recurrence intervals were determined for future storm surges. Under the higher emissions scenario, by 2050, the elevation of the 2005 100-year event may be equaled or exceeded at least every 30 years at all sites. In more exposed US cities such as Boston, Massachusetts and Atlantic City, New Jersey, this could occur at the considerably higher frequency of every 8 years or less. Under the lower emissions scenario, by 2050, the elevation of the 2005 100-year event may be equaled or exceeded at least every 70 years at all sites. In Boston and Atlantic City, this could occur every 30 years or less.

Appendix A: GEV frequency analysis

For large samples, the cumulative distribution function for the maximum values of many probability distributions converges to one of three extreme value distributions (EV type I, II, or III) described by Gumbel (1958). The Generalized Extreme Value (GEV) distribution is a general mathematical form which incorporates Gumbel’s type I, II and III distributions for maxima (Stedinger et al. 1993). The parameters of the GEV distribution are ξ (location parameter), α (scale parameter) and κ (shape parameter). The Gumbel (EV type I) is obtained when κ = 0. For κ > 0, the distribution has a finite upper bound at ${\xi + \alpha } \mathord{\left/ {\vphantom {{\xi + \alpha } \kappa }} \right. \kern-\nulldelimiterspace} \kappa$ and corresponds to the EV type III distribution for maxima that are bounded from above. For this study, GEV parameters were estimated using the method of L-moments (Stedinger et al. 1993; Hosking and Wallis, 1997) and are shown in Table 6. GEV goodness-of-fit was confirmed using a probability plot correlation coefficient (PPCC) hypothesis test (Chowdhury et al. 1991) and an L-moment diagram. The PPCC hypothesis test indicated that the GEV adequately fit the annual maximum time series for all sites except Boston. However, the L-moment diagram indicated that the GEV could reproduce the moments of the detrended Boston data. In addition, a visual inspection of PPCC plots (Fig. 6) showed that the GEV fit the upper tail of the Boston time series reasonably well after the largest (1991) data point was removed; hence the GEV was deemed appropriate for all sites. For the Atlantic City and Boston sites, parameters were estimated both with and without trend removal.

Table 6 GEV parameters estimated using the method of L-moments, following Hosking and Wallis (1997)

Using the parameters estimates in Table 6, GEV quantile estimates, x _p , for specified cumulative probabilities, p, were computed from

$$x_{p} = \xi + \frac{\alpha }{\kappa }{\left\{ {1 - {\left[ { - \ln {\left( p \right)}} \right]}^{\kappa } } \right\}}$$

(A-1)

Cumulative probabilities (p) were converted to exceedance probabilities, p_e = 1 − p. The return periods shown in Table 3 are 1/p_e.

Fig. 6

Probability plot correlation coefficient (PPCC) plots for annual maximum time series. The Generalized Extreme Value (GEV) distribution was selected based on the results of the PPCC test and L-moment diagrams