A comparison between historical and recent shoreline-change rates on the U.S. east coast (based on observed shoreline positions from the last century and a half) shows that emergent, large-scale, cuspate coastline features are changing shape, becoming more asymmetrical. This change in coastline shape arises from spatial shifts in the location of erosion and accretion zones. Using a numerical model of coastline change forced by wave-driven alongshore sediment flux, we show that a previously identified shift in hurricane-generated wave climate explains the patterns of coastline change we observe. Our results reveal a previously unrecognized type of large-scale, chronic landscape response to changing forcing. Though demonstrated here for a cuspate coastline, similar large-scale morphological adjustments are likely to occur along coastlines of varying morphology in the future—as global warming continues, along with the associated intensification of storms. Our approach allows for constraining and predicting future shifts in coastline shape.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Although there is mounting evidence that our climate is changing in measurable ways [e.g., IPCC, 2007], detecting the effects of recent climate change in landscape morphology is difficult because initial responses are potentially subtle and easily masked by high frequency variance. To date there are only a few examples of climate change signals that have been detected in landscape systems—these tend to come from arctic environments where the melting of ice drives rapid and dramatic change [e.g., Overeem et al., 2011]. A potential means for improving our ability to detect initial or incipient responses to climate change is to identify and examine aspects of landscape systems that are most sensitive to shifts in climate.
 At the boundary between land and sea, dynamic coastal landscapes are particularly sensitive to changes in environmental conditions. As our climate warms into the future, rising sea level [e.g., IPCC, 2007; Church and White, 2006] and increases in hurricane intensity [e.g., Knutson et al., 2010; Bender et al., 2010] will greatly alter coastal areas. A persistent increasing trend in the height of the largest summer (i.e., hurricane-generated) waves in the Atlantic Ocean spanning three decades was identified by Komar and Allan , based on wave records from deep water wave buoys east of Cape Hatteras, North Carolina, and Charleston, South Carolina, on the eastern seaboard of the U.S. Komar and Allan  attribute this trend to a recognized increase in hurricane activity since 1970. Such changes in wave climate, especially the associated changes in the statistical distribution of wave influences from different directions, averaged over at least decadal time scales—which are likely to occur globally as climate change alters storm intensity and/or frequency—will tend to reshape sandy coasts [Slott et al., 2006]. However, detection of coastline response to such chronic changes in storm activity is difficult since the amplitude of high frequency variations in storm size and occurrence tends to dominate local coastal change signals.
 Dynamically, detecting coastal change in response to long-term trends in storm activity requires investigating characteristics of coastlines that evolve over similarly long intrinsic time scales [Werner, 1999]. The shape of cuspate coastlines, which emerges from feedbacks between coastline shape and gradients in alongshore sediment flux [Ashton et al., 2001], provides an example of such a characteristic, featuring dramatic shifts in shoreline orientation over scales of 1 – 100 km. Although their large scale prevents such coastline shapes from changing significantly during individual storms, these morphological features are particularly sensitive to subtle changes in long-term wave climate [Slott et al., 2006]. We turn, therefore, to the cuspate North Carolina coastline, along the U.S. mid-Atlantic coast (Figure 1) between the Hatteras and Charleston Buoys, as the best place globally to detect the influence of a documented change in storm wave climate. Because we expect the changes in shape of such large-scale features over decades to be subtle, we focus on the associated shifts in coastline-change rates. Here, we present an analysis of historical and recent shoreline change for Cape Hatteras and Cape Lookout on the North Carolina coastline and compare these observations to model-generated predictions of shoreline change in response to the alterations in wave climate observed by Komar and Allan . Through this analysis, we identify shifts in observed patterns of coastline change, which have resulted in corresponding shifts in shoreline orientation that are consistent with the changes in Atlantic Ocean wave climate identified by Komar and Allan .
2 Historical and Recent Shoreline-Change Observations
 To conduct a shoreline-change analysis for the regions extending 20–30 km to the north and south of Cape Hatteras and Cape Lookout, we compile shoreline positions from three sources (Table S1) for the time period between 1849 and 2004, and following methods described in Moore , we calculate shoreline-change rates for a historical time period (1850 – 1974; seven shorelines per cape) and recent time period (1974 – 2004; five shorelines per cape). Since wave observations do not extend far enough back in time to identify when changes in wave climate began to occur, the 1974 cutoff between time periods intentionally corresponds to the beginning of the time period of observations analyzed by Komar and Allan . Because we seek to quantify broad-scale (km) longer-term (decadal) shoreline changes on the north and south flanks of the capes, we focus our analysis on stretches of coast that are not directly impacted by inlet processes and cape tip dynamics, which alter shoreline position on short time scales (daily to seasonal), especially in response to storms.
 In ArcGIS, we project all shorelines to the North Carolina State Plane, NAD83 reference system. Using the Digital Shoreline Analysis System [Thieler et al., 2009], we construct an offshore baseline and cast transects landward from this baseline at 100 m intervals using a 300 m smoothing of transect casting direction to insure transect intersection at a near perpendicular angle. We cast transects for all shoreline stretches that are greater than 1 km away from a cape tip across all shoreline positions and that are not within the zone of immediate inlet influence, which we classify as all stretches of coast exhibiting (at some point in the record) a landward curving shoreline position on either side of a gap in shoreline position. We then calculate a linear regression rate (LRR) of shoreline change for the historical and recent time periods for each cape and apply a 5 km Gaussian filter to the LRR values. Uncertainty estimates (Uavg, see supporting information) for filtered LRR rates range from ± 2.3 to 0.040 m/yr with an average uncertainty of ± 0.35 m/yr. We subtract historical shoreline-change rates (Rh) (Figures 2a and 2b) from recent shoreline-change rates (Rr) (Figures 2c and 2d) (Difference = Rr − Rh) revealing an overall pattern of widespread increased erosion on the northern flanks and widespread decreased erosion on the southern flanks of Cape Hatteras and Cape Lookout (Figures 2e and 2f). Historical and recent average shoreline-change rates and shoreline-change rate differences for each flank of each cape qualitatively support this general trend (Figures 2a–2f), despite the moderately high rates of accretion in a localized zone at the north end of the north flank of Cape Hatteras and the localized areas of increased erosion at the east and west end of the south flank of Cape Lookout, which reduce the reflection of the overall trend in average values. The observed shift in patterns of erosion and accretion equates to a subtle change in coastline shape, increasing the asymmetry of cape features.
3 Simulations of Coastline Response to Wave Climate Change
 To assess whether or not the observed shifts in shoreline response are consistent with the observed changes in Atlantic wave climate, we utilize a numerical model [Ashton et al., 2001] to determine the response of a modeled cuspate coastline forced by a changing wave climate. Because the model has been described in detail previously [Ashton and Murray, 2006a], we include only critical components here.
3.1 Model Description
 The flux of wave-driven sediment Qs moving parallel to a local shoreline is given by
where Hb is the breaking-wave height, φb is the breaking wave angle, and θ is the local shoreline angle. The parameter K1 is an empirical constant set to 0.4 m1/2/s [Ashton et al., 2001]. Conservation of sediment implies that gradients in this flux of alongshore sediment will alter the position, η, of the coastline as
where D is the cross-shore depth over which sediment is deposited or eroded, and x is the alongshore coordinate. The calculations of shoreline angle are made considering local shoreline orientation.
 Waves are forced in the model from a given offshore angle and then refracted across assumed shore parallel contours until a breakpoint has been reached. With complex coastline shapes, protruding portions of shoreline can shadow other shoreline segments from some wave-approach directions. Sediment transport does not occur on shadowed shoreline segments.
 Model experiments [Ashton and Murray, 2006a] show coastline shape depends primarily on two wave-climate parameters: the proportion of wave influences from “high-angle” waves (those with angles between offshore wave crests and the shoreline greater than approximately 45°) which affects the cross-shore/alongshore aspect ratio of coastline shapes; and the asymmetry (the proportion of wave influences from the left, looking offshore) which affects the asymmetry of coastline shapes. These two parameters combine to define a four-binned depiction of wave climate.
3.2 Wave Forcing
 Since historical records of offshore wave conditions from buoy data are incomplete and do not extend as far back in time as our observations of shoreline change (~30 years vs. ~160 years), we have used wave climates based on data from recent decades to represent historical wave conditions.
 We force the model with a wave climate that is derived from wave records for U.S. Army Corps of Engineers Wave Information Study hindcasts (WIS, based on 1980 – 1999, wis.usace.army.mil/wis.html, stations 507, 508, and 509). Forcing with any of the WIS wave climates results in a large-scale cuspate coastline roughly similar to that found along the Carolina coast (supporting information). Using WIS station data as a starting point, we run a series of simulations, varying the distribution of incoming wave angles, to search for the wave climate that produces a cuspate shoreline in the model that most closely matches the present coastline (Figure 3), (i.e., the “best fit” wave climate).
 In the coastline model, a new offshore incident wave angle is chosen every model day from a probability distribution function representing the wave climate. Because the model involves the simplified assumption of wave transformations over shore-parallel contours, the most appropriate location to collect wave data to force the model is at the seaward limit of the shoreface. However, because wave-shadowing effects are important along a complex-shaped coastline, wave data from any single shallow-water location is not representative of wave conditions affecting the whole coastline. On the other hand, wave transformations between the edge of the continental shelf and the seaward limit of the shoreface are not included in model simulations, so that deeper-water WIS stations (e.g., stations 507, 508, and 509), while providing a more representative wave climate, are also not fully appropriate as model input (supporting information). Therefore, although we generate a reference probability distribution for daily incoming wave directions based on WIS data (from station 509, the best of the three deeper-water stations, see supporting information), we then alter that climate to produce the best fit wave climate (supporting information).
 In generating the reference distribution, we utilize the WIS hourly records of significant wave height and mean wave direction to generate a probability distribution consisting of four equally spaced wave-angle bins, where angles represent the relative angle between incident waves and the general trend of the coastline along the Carolina coast [Ashton and Murray, 2006a]. We weigh the contribution from each wave observation in such a way that the probability of choosing waves from each wave-angle bin corresponds to the relative contribution of waves from those wave angles to alongshore sediment flux [Ashton and Murray, 2006b, supporting information].
 Previous authors [Komar and Allan, 2008] have documented an increase in wave height from wave events that occur in the summer months, especially when filtered for waves above 3 m in height. Specifically, for NOAA buoy 41002 located offshore of Charleston, South Carolina, these authors found that large, summer wave height events, which correspond to hurricanes, are increasing at a rate of 0.054 m/yr. We incorporate this in model forcing by adjusting the probability distributions such that all summer month wave heights above 3 m are increased by 2.7 m (representing application of the trend over 50 years). Based on the directions from which summer waves higher than 3 m approach in the WIS station 509 data, the adjustment in wave height produces an increase of 0.03 in the wave asymmetry and a decrease of 0.02 in the wave highness [Ashton and Murray, 2006b] (supporting information). (Using the smaller trend in hurricane wave height observed for the Hatteras Buoy produces qualitatively the same results; supporting information.) We apply the corresponding change to the four probability bins of the best fit wave climate to simulate an increase in hurricane waves (Figure 3a). The wave-angle bins 0 to 90 represent waves approaching from the left, when looking offshore (relative to the overall coastline orientation), and the increase in the wave influences from 0 to 45 degrees shifts the imbalance between influences from left-approaching and right-approaching waves. (The increased waves from the left, or E - NE, are generated by onshore-directed winds on the northern flanks of nearby hurricanes. Note that Slott et al.  incorrectly assumed a relative increase in waves from the right in a scenario of increasing hurricane strength; supporting information.)
 We then compare the pattern of shoreline erosion and accretion that occurs in response to the altered wave climate with the case in which the model was forced with an unaltered wave climate (Figure 3, and supporting information). Results using the best fit wave climate (Figure 3c) show that the shift in wave climate results in increases in erosion updrift of capes and decreases in erosion downdrift of capes. This response is consistent, in both spatial pattern and magnitude, with the shifts in shoreline-change patterns identified in the observational record of shoreline position. This shift in shoreline-change patterns alters overall coastline shape by increasing the asymmetry of the cuspate form, which is consistent with an increasingly asymmetric wave climate. (Mechanistically, a relative increase in the sediment flux near the cape tip on the updrift side, and a corresponding decrease on the downdrift side, result in increased sediment-flux divergence on the updrift flank and increased sediment-flux convergence on the downdrift flank).
4 Discussion and Conclusions
 There are limitations to our model and observational approaches. The WIS hindcast stations do not contain data prior to 1980 and the data do not completely represent the entire wave spectrum. In the numerical simulations, the empirical alongshore sediment transport relation used is one of many existing relations [see Ashton and Murray, 2006a for sensitivity tests] and the numerical model involves simplifying assumptions about wave transformations across the continental shelf. For simplicity, we have neglected the approximately alongshore-uniform component of shoreline change related to sea level rise [Wolinsky and Murray, 2009], which will not alter the spatial patterns in our shoreline-change results. We have used coarsely binned wave climates (see supporting information for tests of sensitivity to bin resolution), and in choosing a wave climate that produces a “best fit” with the current Carolina coastline, the results depend in detail on the alongshore extent of the actual coastline considered (see supporting information for tests of sensitivity to wave-climate parameters). For these reasons, we consider our model results qualitatively, focusing only on shoreline-change patterns and how these patterns change as a function of wave climate rather than on specific values of shoreline change. We did not attempt to tune the wave climate or the wave-climate change to produce the best match with shoreline-change observations, and we present the model runs with the best fit to the initial coastline not as quantitatively reliable predictions, but to illustrate the main characteristics of the way the shoreline-change patterns along a coast like the Carolina Capes shift with this change in wave climate (more erosional updrift of a cape tip, and less erosional/more accretional downdrift).
 Likewise, shoreline-change analysis is limited by the accuracy of shoreline observations, though increasing the number of shorelines (here 5–7) from the minimum of two per time period of interest improves the reliability of results [Dolan et al., 1991]. Our observational analysis may also be limited by the fact that due to a lack of wave data for this region prior to 1974, it is not possible to know when the changes in wave climate identified by Komar and Allan  began. However, if the shift in wave climate began prior to the beginning of the observational record (i.e., prior to 1974), our selection of this year as the breakpoint between historical and recent time periods results in a conservative analysis: if the trend began before 1974 then what we are considering the historical shoreline pattern includes some of the altered pattern, which would tend to make the historical and recent patterns more similar than they would be if we knew the actual year that the trend began. Human shoreline stabilization and alterations to sediment budgets can affect coastline change rates. We neglect such influences in our analysis because our study areas consist of protected coastlines with minimal human manipulations. Despite these limitations, the qualitative agreement between observations and simulations of the shift in shoreline-change patterns and the consistency of shifts in patterns of shoreline change across both cuspate landforms suggest a connection between changes in Atlantic Ocean wave climate and changes in shoreline position.
 Our results suggest that the effects of changing Atlantic Ocean wave climate are already detectable on the cuspate Carolina coastline. Regardless of whether the recently observed regional trend in wave climate continues unchanged (or is part of a natural, decadal-scale variation), changes in storm behaviour, and therefore shifts in wave climate, are likely to occur globally as our climate continues to warm [e.g., Bender et al., 2010; Knutson et al., 2010; Emanuel, 2013]. Alterations in storm climate will likely produce future changes in the shape of many coastlines (analogous to the scenarios explored here for cuspate coastlines), which will be accompanied by shifting and intensifying large-scale zones of coastal erosion hazard. The sensitive cuspate coastal landforms therefore may be providing early warnings of landscape changes that will become more prominent as the effects of climate change on wave forcing become more pronounced.
 The authors thank Peter Haff and Marco Marani for helpful comments on an earlier draft and Matt Wolinsky and anonymous reviewers for comments that helped to improve the manuscript. Support for this project was provided by UVA and the National Science Foundation (EAR-1053151).
 The Editor thanks two anonymous reviewers for their assistance evaluating this manuscript.