Response of a fringing reef coastline to the direct impact of a tropical cyclone

Tropical cyclones generate extreme hazards along coastlines, often leading to losses of life and property. Although coral reefs exist in cyclone‐prone regions globally, few studies have measured the hydrodynamic conditions and morphological responses of reef‐fringed coastlines to tropical cyclones. Here, we examine the impact of Tropical Cyclone Olwyn on a section of Australia's largest fringing reef (Ningaloo Reef) using in situ wave and water level observations, topographic surveys, and numerical modeling. Despite forereef significant wave heights reaching 6 m and local winds of 140 km h−1, average beach volume change was only −3 m3 m−1. The results indicate that this erosion was due to locally generated wind waves within the lagoon rather than the offshore waves that were dissipated on the reef crest. A comparison of these volume changes to observations of tropical cyclone impacts along exposed sandy beaches quantitatively demonstrates the substantial coastal protection reefs can provide against extreme storms.

conditions over the reefs (P equignet et al. 2011), or the fate of storm deposits (Scoffin 1993). However, few studies have focused on the hydrodynamic and morphodynamic processes driving changes in reef-fringed beach morphology during a TC. In this study, we quantify the morphological change due to a direct cyclone impact along a reef-fringed coast, identify the physical mechanisms responsible for the observed beach erosion patterns, and assess the coastal protection provided by the reef relative to exposed sandy coasts experiencing similar extreme met-ocean conditions. Fringing reefs, which can range from shore-attached to several kilometers offshore, are the most prevalent class of coral reefs along tropical coasts (Hopley 2004). Reefs provide natural coastal protection by dissipating offshore wave energy through both wave breaking and bottom friction (Lowe et al. 2005;Rogers et al. 2016). Despite the abundance of reef hydrodynamic studies, few studies have quantified the processes governing coastal morphological changes in reefs, particularly during TCs. Previous studies conducted under more typical (non-TC) conditions have shown that shorelines fringed by reefs can experience morphological changes on short (weekly) to intermediate (seasonal) timescales due to changes in wind and wave direction (Eversole and Fletcher 2003;Beetham and Kench 2014); and that reeffronted beaches are generally more stable than exposed beaches (Ruiz de Alegria-Arzaburu et al. 2013). The limited studies that have directly measured TC-induced shoreline changes have suggested that shoreline response is inversely related to reef flat width, with the magnitude of beach erosion decreasing with increasing reef flat width (Mahabot et al. 2016). These same studies have also suggested that the relative coastline orientation (i.e., in relation to local wind and wave direction) can influence the magnitude of coastal erosion. However, due to the episodic nature of TCs and the rarity of capturing a direct impact with in situ instrumentation, there still remains a gap in understanding of how much coastal protection is offered by fringing reefs during TCs, and moreover, which factors specifically govern coastal responses.
Rigorously assessing the coastal protection afforded by reefs is particularly important given that a number of recent studies have suggested that the coastal protection offered by reefs is under threat due to both coral degradation (i.e., declining rates of net reef accretion and decreasing bottom roughness) and sea level rise, which will increase submergence depths and therefore expose reef-protected coastlines to larger waves (Grady et al. 2013;Quataert et al. 2015). Collectively, these studies emphasize that it is critical to understand the physical drivers of morphologic change in reef environments to better predict future changes to these coastlines. In this study, we combine in situ measurements (wave heights, water levels), repetitive topographic beach surveys, and numerical simulations from a coupled wave-current model (Delft3D/SWAN) to quantify the coastal response from the direct impact of TC Olwyn (2015) at Ningaloo Reef, Western Australia.

Study area
Ningaloo Reef is Australia's largest fringing reef, stretching 270 km south from Australia's North West Cape (Fig. 1a). The morphology of Ningaloo is characterized by a steep (1 : 20) forereef slope, a relatively narrow reef flat (100s of m), and a wide (1-5 km), but shallow (< 5 m deep) lagoon (Collins et al. 2003). The study area is located at the northern end of Ningaloo Reef at Tantabiddi (Fig. 1c) and features a shoreline salient inland from the reef flat; a common feature along the reef-fringed coastline of Western Australia, including at Ningaloo (Sanderson 2000). The beach has a variable slope (calculated between the 0.5 m and 1.5 m contours), ranging from 0.03 north of the salient to 0.15 south of the salient, is relatively narrow ( 100 m), and is backed by a dune system (heights from 3 m to 10 m). Ningaloo Reef experiences small tides (average range 0.8 m) with typical incident waves heights of 1-2 m from a southwesterly direction; however, waves often reach 4 m during austral winter swell events (Taebi et al. 2012). Depthinduced breaking on the shallow reef crest drives onshore flow across the reef flat toward the lagoon; within the lagoon, currents typically diverge at the salient toward the adjacent channels (Taebi et al. 2011). The sediment at Tantabiddi is dominated by biogenic material (70-95%), with an average density of 2.58 g cm 23 (Cuttler et al. 2017). The spatial distribution of sediment at Tantabiddi generally follows the mean current patterns, with the coarsest material (median grain size, D 50 , 0.5 mm) found on the outer reef flat, and finer material found within the lagoon (D 50 0.35 mm) and on the beach (D 50 0.23 mm) (Cuttler et al. 2017).
Ningaloo Reef is located along Australia's Northwest Shelf, which is the most cyclone prone region of Australia, experiencing 1-2 TCs per year (Drost et al. 2017). In order to opportunistically capture the impact of a TC on this reeffronted coastline, five pressure sensors (Fig. 1c

Meteorological and hydrodynamic data
Meteorological data (atmospheric pressure, wind speed, and wind direction) were measured 18 km south of the study site at the Milyering weather station, maintained by the Australian Institute of Marine Science. A cross-shore array of pressure sensors (RBRvirtuoso), extending from the forereef ( 18 m depth) to the shoreline recorded observations of waves and water levels ( Fig. 1c,d). All sensors logged continuously at 1 Hz until 19 March 2015 when they were recovered following TC Olwyn. Atmospheric pressure was removed from the pressure observations and the tidal component of the water level was removed using the time series predicted by T_TIDE (Pawlowicz et al. 2002) with constituents derived from the forereef pressure sensor. Hourly estimates of the sea surface energy-frequency spectrum were obtained from Fourier transforms of 1-h de-meaned and detided data segments and block averaged using a Hamming window (1024 samples) with 50% overlap, yielding 17 degrees of freedom (Thomson and Emery 2014). Hourly significant wave heights were estimated from the variance of sea surface elevation in the sea-swell (SS; 0.04 Hz < f < 0.4 Hz, where f is frequency) and infragravity (IG; 0.002 Hz < f < 0.04 Hz) energy bands (H sig,SS and H sig,IG , respectively) using linear wave theory.
The pressure observations were also used to estimate setup following: where g denotes setup (due to both waves and wind), h tot is total measured depth (assuming hydrostatic pressure), g t denotes tidal elevation, h o is still water level, and overbars denote hourly averaging. Still water level (h o ) is a constant at each site and was calculated as the depth at each site when forereef waves were low (H sig,SS < 0.8 m), wind speed was low (< 10 km h 21 ), and tidal elevation was highest (g t > 0.6 m; i.e., assuming no wave setup) (Raubenheimer et al. 2001). All measurements were then referenced to the forereef sensor to remove any water level fluctuations due to shelf-scale water level variations (Lowe et al. 2009).

Beach morphologic changes
Annual beach surveys have been conducted at Tantabiddi since August 2013 using a backpack-mounted differential GPS (DGPS) receiver. The pre-cyclone morphology was recorded in July 2014, approximately 8 months prior to the cyclone, and a "post-cyclone" survey was conducted 10 d after TC Olwyn. Two subsequent "cyclone recovery" surveys were conducted in October 2015 and June 2016. Although the pre-cyclone survey was conducted 8 months prior to TC Olwyn, there were no significant storms (i.e., TCs) in this period, typical large winter swell events (H sig 4-5 m) have minimal erosion potential ("Model simulations" section), and analysis of historical aerial imagery (from 1969 to 2014) of the Tantabiddi salient using the "Digital Shoreline Analysis System" (Thieler et al. 2009) as well as previous work along the greater Ningaloo coastline (Sanderson 2000) provides no evidence to suggest significant seasonal variability in shoreline position (Supporting Information Fig. S1).
Positions from the DGPS surveys (uncertainty of 0.05 m in the vertical and horizontal directions) were organized into a triangulated irregular network and gridded to produce a 1 m cell-size topographic surface of the surveyed beach. Successive topographic surfaces were differenced to determine (c) study site and bathymetry with the cross-shore array of pressure sensors indicated by the white triangles, the "inner" Delft3D-FLOW and SWAN domain show in gray and the "shore" Delft3D-FLOW grid shown in black. TC Olwyn track is the red line in (a) and (b). (d) Cross-shore depth profile from the coastline to the forereef, with the reef crest marked with a red "x" and pressure sensors denoted by red triangles.
the location and extent of erosion or accretion. For the entire 3.5 km beach length studied, 69 cross-shore transects were extracted at 50 m alongshore intervals from the survey grids to assess beach elevation and volume change. Height data were referenced to the Australian height datum (approximately mean sea level).

Numerical model
The numerical model Delft3D-FLOW (Lesser et al. 2004) coupled with the wave model SWAN (Booij et al. 1999) was used to simulate the hydrodynamic conditions during the cyclone. Delft3D solves the unsteady shallow-water equations in two-or three-dimensions (Lesser et al. 2004) and has been used effectively in coral reef environments under both non-storm and storm conditions (Grady et al. 2013;Hoeke et al. 2015). Numerical details, governing equations, and underlying assumptions of Delft3D are described in detail in Lesser et al. (2004); brief details of the model resolution and settings used for this study are described below. Further details of model settings and validation are provided in the Supporting Information.
Delft3D-FLOW (version 6.02.02.5562M) was run in depthaveraged mode and consisted of three domains ("outer," "inner," "shore") of varying resolution that were two-way coupled using "domain decomposition" (Hummel and de Goede 2000). The "outer" domain had a 50 3 50 m resolution, the "inner" domain had a 17 3 17 m resolution, and the "shore" domain had a 5 3 5 m resolution (Fig. 1b,c). The hydrodynamic model was two-way coupled with SWAN (version 40.72) over two nested domains that had the same resolutions as the "outer" and "inner" circulation domains. Stationary SWAN simulations were run at 60 min intervals using updated currents, water levels, and wind from Delft3D-FLOW; wave simulation results were then passed back to Delft3D-FLOW. Although SWAN does not model IG energy, which has been shown to be important to lagoonal processes under non-TC conditions (Pomeroy et al. 2012), it does account for wind-wave growth, which we show below to be a more critical factor in determining the coastal response from TC Olwyn. Model performance was evaluated using the Murphy (1988) Skill Score (Supporting Information Text S2). Inclusion of wind growth yielded the most accurate model results (Supporting Information Table S1, Supporting Information Fig. S2); therefore, all model results presented below (unless otherwise stated) are for simulations with wind growth.

Field observations
The eye of TC Olwyn passed 10 km offshore of the study site on 12 March 2015 (Fig. 1b) and resulted in incident waves on the forereef reaching 5.8 m (H sig,SS ). To place the TC Olwyn conditions in the context of typical wave conditions (H sig,SS 1-2 m, Pomeroy et al. 2012), the observed waves and water levels during the cyclone were compared to a "typical" swell event (H sig,SS 5 1.8 m) that occurred prior to TC Olwyn (Fig. 2). The cross-shore distribution of wave heights and setup (due to both wind and waves) were considerably different during the TC compared to typical swell conditions. Under typical swell conditions, after dissipation of incident wave energy commences near the reef crest, H sig,SS continues to decrease across the reef flat and lagoon toward shore with H sig,SS and H sig,IG becoming comparable in magnitude (Fig. 2b). However, at the peak of TC Olwyn, although H sig,SS decreased near the reef crest by 89% due to depth-induced wave breaking, H sig,SS increased across the lagoon toward the shoreline, with H sig,SS reaching 1.5 m ( 1.5 times reef flat H sig,SS ), or approximately six times larger than H sig,IG (Fig. 2c).
The shoreward growth in H sig,SS corresponded to a change in the wind direction that increased the shoreward-directed  S3d). The observed wave spectra show that energy was primarily enhanced at high frequencies (0.1-0.3 Hz) indicating local growth of wind waves within the lagoon (Fig. 2f). Maximum TC waves occurred during high tide (Supporting Information Fig. S3b), which would have allowed more wave energy to propagate into the lagoon and aided wind growth by increasing the lagoon depth. A similar pattern was observed in the cross-shore setup profile. Under typical conditions setup decreased across the lagoon toward the shoreline whereas during the cyclone setup increased toward shore, coincident with the shoreward increasing wave heights and onshore winds (Fig. 2d).
The observed changes in beach morphology were remarkably modest in response to the TC, with alongshoreaveraged, sub-aerial profile volume changes of only 23 m 3 m 21 and net volume change (integrated over the entire subaerial beach) of 214,457 m 3 (28%); however, the beach response was variable alongshore (Fig. 3a,b). The north side of the salient showed slight accretion (13 m 3 m 21 on average), whereas the south side was predominantly eroded (28 m 3 m 21 ; Fig. 3a,b). In regions where erosion occurred, it corresponded to a flattening of the cross-shore beach profile, primarily due to erosion of the dune (removing up to 1.5 m elevation; Supporting Information Fig. S4).
The second post-cyclone survey (October 2015) showed limited beach recovery, with a total sub-aerial beach volume change of 13885 m 3 (12%) from to March 2015; which is still 26% compared to the pre-cyclone beach volume. However, the limited recovery was likely impacted by TC Quang, which passed by the study area 6 weeks after TC Olwyn and had a similar magnitude and storm track.  complete beach volume recovery, with a total sub-aerial beach volume change of 126,677 m 3 (15% net accretion) compared to the pre-cyclone (July 2014) beach volume.

Model simulations
To further investigate how the nearshore wave fields likely drove the observed spatial variability in beach changes, we used the numerical model to extend the in situ results over the entire study area. Modeled wave heights were extracted along the 2 m isobaths (as an indicator of the wave energy along the beach), which is farther offshore on the south side of the salient due to the southern portion of the lagoon generally being shallower than the north (Fig. 3a).
Modeled wave heights during TC Olwyn were slightly larger on the northern side of the salient, despite the observations indicating accretion in this area ( Fig. 3d; Supporting Information Fig. S5). Examination of the alongshore beach morphology showed that TC-induced beach volume change was significantly correlated with the pre-cyclone beach slope (r 2 5 0.60, p 0.01); with northern (accreted) beaches having shallower slopes and the southern (eroded) beaches having steeper slopes (Fig. 3b,c). Under typical wave conditions, the shoreline on the southern side of the salient (y 5 21000 m to 0 m) is exposed to smaller waves than the northern side (y > 0 m; Fig. 3d), likely explaining the typical alongshore variability in beach slope (Wright and Short 1984). Note, there are no significant alongshore differences in beach sediment characteristics (Cuttler et al. 2017). The largest waves within the lagoon, and likely the majority of the beach morphology changes, occurred when the onshore component of the wind was strongest (Supporting Information Fig. S3); at this time, the wave direction (along the 2 m isobath) was approximately orthogonal to the shoreline (Supporting Information Fig. S5). These results suggest that this reef-fronted beach is in equilibrium with the prevailing conditions, and that the morphological response to TC Olwyn was determined by how much the cyclone waves locally deviated from typical conditions, i.e., the alongshore beach morphological response was related to the spatial pattern in the disequilibrium of TC wave heights (H sig,TC /H sig,Typical ) from a typical swell event (r 2 5 0.48, p 0.01; Fig. 3e).
To confirm the role of wind-wave growth within the lagoon, SWAN was run with and without wind growth. Without wind growth included in the model, the nearshore wave heights from the cyclone swell are predicted to be approximately two times larger than during a typical incident swell and disequilibrium wave heights are similar on both sides of the salient. However, with wind growth enabled, nearshore TC wave heights are up to five times larger than those from a typical swell, and the magnitude of disequilibrium from typical conditions is larger on the southern side of the salient, consistent with the observed beach erosion (Fig. 3e). Thus, despite the lagoon being relatively shallow (< 5 m), locally generated wind waves within the lagoon were a dominant mechanism driving the observed beach changes.

Discussion and conclusions
This study highlights the effectiveness of coral reefs at providing natural coastal protection. Despite the extreme offshore wave conditions, beach morphology changes were relatively modest; particularly when compared to the responses of open, sandy coastlines exposed to comparable wave conditions (Fig. 4). For example, average beach volume loss at Tantabiddi was 3 m 3 m 21 , whereas, areas in Florida, U.S.A. that were impacted by similar offshore wave heights from Hurricane Ivan experienced 5-10 times greater beach erosion (Wang et al. 2006), with comparable erosion rates also observed during other Atlantic Ocean TCs (blue symbols, Fig. 4). Although the magnitude of beach response in reef environments is much less than sandy environments, the underlying forcing mechanism appears to be similar, with the magnitude of divergence (or disequilibrium) of nearshore wave heights from typical nearshore wave heights a key factor determining storm-induced beach response (e.g., Yates et al. 2009).
The magnitude of beach response observed here agrees well with the few previous observations of TC-induced beach morphology changes in fringing reefs (Fig. 4, red x's) (Mahabot et al. 2016). These previous observations were conducted in shore-attached (i.e., no significant lagoon) reef systems sandy coasts (blue symbols) and reef-fronted coasts (red symbols). For studies that did not report offshore H sig , wave heights from WaveWatch-III hindcast model predictions (Tolman 2009 and highlighted the importance of wave direction and reef flat width in determining morphological response; however here, in a system with a significant lagoon, we suggest that wind-driven processes can also be very significant (or dominant) in driving beach morphodynamics during a TC. We note that the influence of these wind-driven processes will likely scale with distance from the TC and lagoon width. For example, if TC Olwyn had passed further offshore of the study site, incident wave heights may have been similar (i.e., 6 m) but local winds would have been much less. Therefore, nearshore wave heights would have been significantly smaller and beach volume changes likely would have been even less than those observed here.
These results show that the offshore TC-generated waves were largely dissipated by the reef and played a minimal role in driving the beach changes. Instead, locally generated wind waves played a critical role in determining the morphological change, with the wind waves reaching the shoreline ultimately being larger than the residual offshore waves that were transmitted across the reef. Although future climate changes are likely to increase wave transmission over coral reefs, primarily due to sea level rise but also due to degradation of reef habitats (Grady et al. 2013;Quataert et al. 2015), the results presented here show that while these structures still exist, they can provide natural coastal protection from offshore waves, even under extreme conditions. However, our results also reveal that even over the small scales (order 1 km) of a nearshore reef-lagoon system, local wind wave generation can play a dominant role in sediment transport and erosion responses during TC conditions, a process traditionally neglected in nearshore model applications to predict beach response to extreme events (e.g., models that can only consider remotely generated waves, such as Xbeach). Furthermore, because the reef geomorphology at Ningaloo (e.g., shallow reef crest that gradually deepens to a shallow lagoon) represents a "standard" reef (Falter et al. 2013), accounting for the local wind growth of waves is expected to be critical to determining beach response to TCs in reef systems worldwide.