Extreme water levels, waves and coastal impacts during a severe tropical cyclone in Northeast Australia: a case study for cross-sector data sharing

General Comments: This manuscript brings together observations from diverse sources and analyzes these in the context of spatial variations in coastal impacts from tropical cyclone Debbie that made landfall on the eastern Australian coast in March 2017. Compiled data of waves, water levels, and beach erosion coupled with some relatively simple and straightforward analytical calculations of wave extrapolations, wave runup, and relative contributions of low pressures and wind-setup to storm surge, and observations are used to evaluate flooding, erosion, and infrastructure damage and makes causal connections of these patterns with duration of the storm, timing with as-

barriers, either amplifying or dampening water levels (Lipari et al., 2008). In general, surge potential is maximised on open, straight coastlines with shallow shoreface slopes and slow-moving, landfalling cyclones travelling perpendicular to the coast. In the Southern Hemisphere and on east-facing coasts, the clockwise flow of low pressure systems means coastal areas on the southern limb of TCs experience the most wind, surge and wave impacts. Conversely, areas to the north often experience a suppression (setdown) of water levels and low wave impacts due to offshore-directed winds. 5 TC-induced coastal flooding is maximised when storm surge coincides with a high astronomical tide and energetic wind-waves.
Non-linear harmonics between tide and surge can further elevate total water levels especially in areas with large tidal ranges, with some research suggesting surge maxima are more likely to occur on the rising or falling tide rather than at slack water because of this dynamical coupling (Horsburgh and Wilson, 2007). Wave breaking further increases water levels at the coast (wave setup) and adds more forward momentum to the water mass (wave runup) meaning coastal foredunes can be breached even when the 10 storm tide elevation is lower than the dune crest.
The TC season in Australia typically runs from November through to April (Austral late summer to early autumn), when sea surface temperatures are warm enough for cyclogenesis. In Australia, as elsewhere, TC-affected coasts (< 28 ° S) comprise tidemodified or tide-dominated beach systems with meso-to macro-tidal range (2 to 4 m, or exceeding 4 m), low-relief dunes and a steep high-tide beach fronted by shallow gradient sand, rock or reef flats (Short, 2006). These beaches are equilibrated with a 15 predominantly low-energy hydrodynamic regime, punctuated by infrequent high-energy TC events. The large energy difference between the modal and extreme regimes leaves exposed tropical-coast locations vulnerable to significant erosion during TCs. This can have lasting impacts for beach amenity, access and tourism, and lowers the geomorphic threshold for subsequent inundation and erosion.
On 28 March 2017, a Severe Tropical Cyclone (Debbie) made landfall on the central north Queensland coast of Australia. Debbie 20 affected the Whitsunday Islands group, approximately 30 km offshore, as a low-end Category 4 tropical cyclone before making landfall on the mainland coast, near Airlie Beach, as a high-end Category 3 tropical cyclone (BoM, 2018). This area has experienced several TCs in recent years -although not as intense as Debbie -including Ului in 2010 (category 3), Anthony in 2011 (category 2), and Dylan (category 2) in 2014. Debbie intensified from a tropical low southeast of Papua New Guinea in the Coral Sea and proceeded to drift south as a category 2 system. Three days before crossing the coast, Debbie turned southwest 25 (perpendicular to the coast) and rapidly intensified to category 4 with a RMW of approximately 30 km, reducing to 15 km at landfall (BoM, 2017).
Because of the intense winds (peak gusts over 260 km h -1 ), shore-normal approach, slow forward-moving speed and coincidence with rising spring tides, significant storm tide inundation was expected. Debbie's track on approach to the coast and landfall location was in fact very similar to Severe TC Ada in 1970, an infamous category 4 system which was responsible for the loss of 30 cyclones on coastal systems is much better understood (e.g. Turner et al., 2016, Strauss et al., 2017. To address this data gap, this paper collates and analyses observations of coastal impacts and concurrent hydrodynamic conditions before, during and directly after severe tropical cyclone Debbie. Field data were collected independently by a group of organisations (Risk Frontiers/Macquarie University, Fugro Roames, Department of Environment and Science (DES) Queensland Government, James Cook University, Griffith University and Alluvium), using a range of data collection and analysis methods . 5 This work represents the only observational analysis of hydrodynamic drivers and concurrent coastal impacts for a severe tropical cyclone in Australia. The data is of value for the calibration and validation of coastal hazard modelling, both in north Queensland and for other tropical cyclone-affected east coasts in the Southern Hemisphere, such as Mozambique, Tanzania and Brazil, where the genesis of these events are similar, but coastal observations are lacking. This paper also provides a case study in data sharing and open collaboration across industry, government and academia, something which is currently lacking in a natural hazard risk 10 context in Australia.

Study area
Debbie made landfall near Airlie Beach (20.3 °S, 148.7 °E), on the Whitsunday (central north) Queensland coast, around 12:40 pm on 28 March 2017. Prior to landfall, Debbie travelled south-west from the Coral Sea over the southern end of the Great Barrier Reef (GBR), the Whitsunday Island group, and then inland as a tropical low before turning southeast towards Brisbane, the capital 15 of the state of Queensland (Figure 1). The Whitsunday and Mackay regions, from Airlie Beach in the north to Mackay in the south (~110 km shoreline length), were the focus for the groups' surveys as these were the areas within the influence of Debbie's southern limb and onshore-directed flow.

Wave and hydrodynamic regime
Most beaches on the central-north coast of Queensland receive little or no ocean swell waves and are only exposed to low and short-period wind waves generated primarily by south-easterly trade winds. The median long-term significant wave height, Hs, at the Mackay wave buoy (1975 -2017) is 0.7 m, peak spectral wave period, Tp, is 5.8 s, and mean wave direction, MWD, (2002 -2017) is 115 degrees True North (° TN). During cyclone events, conditions greatly exceed this, with wave heights greater than 5 25 m not uncommon. The predominately low wave energy regime is a result of effective attenuation of ocean swells by the GBR matrix, which means nearshore wave conditions are largely dependent on local wind speeds rather than far-field wave generation (Gallop et al., 2014).
The Whitsunday and Mackay coastline is situated to the lee of the widest section of the GBR shelf, with over 180 km separating the outer reefs and the mainland coast between Bowen and Mackay. In this area, water depths do not exceed 80 m but the 30 bathymetry is highly variable. The mean spring range at Hay Point (~20 km south of Mackay) is 4.9 m, whereas on narrower sections of the GBR, typical spring ranges are around 2 to 3 m.

Geomorphic setting
The tidal range at Mackay is seven times greater than the mean annual wave height (at the Mackay buoy, 35 m water depth). For this reason, beaches in this area are tide-dominated to tide-modified (Short, 2000), with a relatively steep high-tide beach and wide, 35 Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License. very low gradient sand and/or tidal mud flats. At low tide, the water line may be several hundred metres seaward of the high-tide beach. Tidal state therefore plays an important role in modulating storm surge impact in this region.
Coastal embayments in north Queensland are typically east-orientated, grading from a protected southern end dominated by tidal flats and wind-blown beach ridges, towards a more exposed northern end oriented perpendicular to the south-easterly trade winds and backed by transgressive sand dunes (Short, 2000). Longshore transport is generally northward, but can change direction in the 5 lee of headlands.

Study sites
Results from seven coastal sites are presented. These include (from north to south) Airlie Beach (20.3 °S,148.7 °E),Hamilton Island (20.4 °S,149.0 °E), Wilson Beach (20.5 °S, 148.7 °E), Conway Beach (20.5 °S, 148.8 °E), Laguna Quays (20.6 °S, 148.7 °E), Midgeton Beach/Midge Point (20.6 °S, 148.7 °E) and Seaforth Beach (20.9 °S, 148.9 °E). 10 Observations of beach morphological change and storm demand are presented at the main study sites of Wilson Beach, Conway Beach and Midgeton Beach, which were some of the most severely impacted coastal locations during Debbie.
Wilson Beach is a small coastal settlement with a 300 m-long beach that faces south across the sand flats and channel of the Proserpine River. The shore consists of a steep high tide beach, fronted by 200-m wide sand and mud flats, adjacent to the river channel (Short, 2000). The houses located closest to the shoreline are built on the sand foredune ridge, with as little as 30 m 15 separating the front doors from the beach crest, which has an elevation of around 5 m Australian Height Datum (AHD).
Conway Beach is a small residential settlement located 2 km south-east of Wilson Beach, with a 1.5-km long beach facing southeast across the Proserpine River mouth. The seafront has a low gradient high-tide beach flattening to 400 to 500 m wide, low sand flats (Short, 2000). The western portion of the beach (in front of the settlement) is backed by a rock revetment with a crest of between 5 to 6 m AHD. The houses located closest to the shoreline sit approximately 50 m behind the revetment. The eastern 20 section of the beach is backed by a vegetated coastal dune and creek system. Midgeton Beach is a 1.8-km long, south-east facing, low gradient, sandy beach. The high-tide beach is fronted by a wide, low gradient intertidal beach, with sand flats extending up to 1 km off the southern end in front of the Yard Creek mouth, which forms the southern boundary. A smaller, mangrove-fringed creek (Sandfly Creek) forms the northern boundary (Short, 2000). The community is situated along the northern half of the beach. Foreshore houses initially sit approximately 70 m behind the sparsely 25 vegetated low-lying foredune. Moving southward, the land between the houses and foredune becomes increasingly vegetated and to the south of the community the beach is backed by a foreshore reserve fringed with dense vegetation and coconut palms.

Storm tide gauges
Storm tide monitoring in Queensland was initiated in the mid-1970s and comprises a network of 36 gauges, which all now measure water levels every minute, relative to local Lowest Astronomical Tide (LAT). In this study, elevations were converted to AHD (Australian Height Datum), which approximates to mean sea level. All storm tide gauges are fitted with a barometer which records 35 the atmospheric pressure at the gauge in parallel with water level.
Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License. We define 'storm tide' in this paper as the astronomical tide plus meteorological surge components (wind setup plus barometric surge) but not wave breaking effects which are treated separately. Observations from four storm tide gauges located within a 150 km radius of Debbie's landfall site were analysed. These include (north to south); Bowen (located on the main cargo wharf at

Barometric surge
The meteorological storm surge during TC events includes wind setup and an accompanying barometric component. Under strong, onshore-directed wind forcing, the water surface is tilted upward with distance downwind, causing an increase in the water level 10 towards the coast (wind setup) (Kamphuis, 2010). A barometric surge occurs when there is a surface air pressure difference between the sea and shore. The inverse barometer (IB) effect can be approximated as; Where Δh is the change in the sea surface height at the shore (m), ρ is the density of seawater (1025 kg m 3 ), g is the acceleration due to gravity (9.81 m s -1 ), P is the central atmospheric pressure of the cyclone and Pref is the global mean atmospheric pressure (1013.3 hPa). P was taken as the minimum barometric recordings at each storm tide gauge. The effect of 1/ρg is small (approximately 0.01 m), such that Eq. (1) is often reduced to Pref -P.
To estimate wind setup, the IB effect (Eq. (1)) was subtracted from the residual between the predicted astronomical tide and the 20 observed water level. In reality, the residual may also include errors in the harmonic derivation of the astronomical tide and nonlinear tide-surge interactions (Horsburgh and Wilson, 2007). However, the contribution of these components is outside the scope of this study, so the 'residual' is taken as synonymous with the meteorological storm surge.

Wave buoys
Ocean wave monitoring in Queensland began in the mid-1970s and now comprises a network of 16 waverider buoys. Observations 25 from one buoy located to the north of the landfall site, Abbot Point (14 m water depth), and two to the south, Mackay (34 m) and Hay Point (10 m), were used in this paper (Table 1, locations Figure 1).
The buoys measure directional wave spectra, from which parametric data are derived at 30-minute intervals. In this study, we refer to the 0.5-hrly significant wave height, Hs (m), the maximum wave height, Hmax (m), the peak spectral period, Tp (s), and the mean wave direction, MWD (° TN, degrees True North). Hs approximates to the average of the highest third of all waves measured in 30 30 minutes and is a common descriptor of the bulk wave climate; Hmax is the maximum wave height recorded during a half-hour; Tp is the wave period pertaining to the primary energy peak of the wave spectrum (the most energetic waves in 30 minutes); and MWD is the mean wave direction at the primary energy peak.

Wave data extrapolation
Due to extreme conditions leading up to Debbie, the Mackay buoy failed at around 05:30 on 28 March (~ 7 hrs before landfall). 35 Prior to failure, a strong correlation was observed between wave conditions at Mackay and Hay Point (R 0.82, p ≤ 0.05 for Hs), To do this, a cumulative distribution function (CDF) mapping approach was used (Brocca et al., 2011). This method compares the Mackay and Hay Point CDFs and adjusts the Hay Point CDF to best match the Mackay CDF. The Mackay-adjusted Hay Point data was then used as a surrogate for the missing data at Mackay. Figure 2 illustrates this process for extrapolating Hmax; this was 5 repeated for Hs, Tp and MWD.

Wave runup estimation
Water levels at open-coast sites include the effects of swash, wave-induced setup and runup, in addition to the storm tide. Swash 10 is generally defined as the time-varying location of the intersection between the ocean and the beach. Wave-induced setup is the super-elevation of the still water level due to the presence of waves. Wave runup is the maximum vertical extent of wave uprush on a beach or structure. Most calculations of runup include the effects of swash and wave-induced setup, and are therefore a measure of the maximum elevation of wave influence above the still water level (or in this case, storm tide). It follows, therefore, that the maximum inundation extent -as evidenced by the most landward detritus lines observed in the field -represents the 15 elevation reached by wave runup above the storm tide.
Wave runup was estimated empirically from wave buoy and beach profile observations, using the equation of Stockdon et al.
Where R2% is the elevation above the storm tide level that is exceeded by 2 % of the wave runups, tanβ is the beach slope, and Hs0 and L0 are the offshore (i.e. deepwater, where water depth, d ≥ 0.5 L) significant wave height and wave length, respectively. This equation has been applied here because it is valid for a broad range of sandy beach types and has been applied for Hurricane-type wave conditions (Stockdon et al., 2007). 25 We used Hs at the time of maximum storm tide (not the peak-storm Hs) to calculate R2%, since the storm tide component accounts for most of the total water level and thus represents the assumed time of maximum inundation. Hs was taken from the Mackay buoy, to the south of Debbie (wave conditions to the north were offshore-directed, thus not contributing to wave runup). The wavelength at the time of the maximum storm tide was first estimated: From this, wave conditions at the time of the peak storm tide were shown to be in intermediate water depths ( Where cg and cg0 are the wave group speeds at the buoy and in deepwater, respectively, given as: This resulted in an Hs0 of 5.0 m and L0 of 131 m at the time of peak storm tide, which was used to estimate R2% at study sites on the south side of Debbie (L0 is equal to L in Eq. (3) as Tp is assumed not to change during shoaling). 10 The beach slope used in Eq. (2) is defined over the area of significant swash activity (Stockdon et al., 2006). However, for applications to cyclone-induced runup where large waves likely move the swash zone higher up the beach profile, the upper (hightide) beach slope (tanβ) is a more relevant measure (Stockdon et al., 2007). tanβ was estimated for each study site, from crossshore beach profiles, from the break of slope at the toe of the high-tide beach up to the berm crest or toe of the coastal defence (where present). 15

Wave power estimation
Wave power is a measure of the energy flux potential in a wave of a given height travelling at a given speed. The maximum power of waves generated during tropical cyclones is an important statistic for the design of coastal and offshore structures. Likewise, the cumulative power of waves during a TC event is an important indicator of beach and dune erosion potential (Splinter et al., 2014).
The deepwater wave power, P0, can be estimated from: 20 From this, the cumulative storm wave power, Pc, integrated over the duration of the storm, D, is: Where P0 and Pc are expressed in megawatt hours per metre crest-length (mWh m). Storm wave conditions were defined as those exceeding a threshold Hs of 2 m, which approximates to the 95% percentile long-term wave height at Mackay, a threshold that conforms with other studies of storm waves in East Australia (e.g. Splinter et al., 2014, Goodwin et al., 2016 Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License.

Airborne terrestrial LiDAR surveys
Fugro Roames undertook aerial LiDAR (Light Detection And Ranging) surveys eight months prior to Debbie (23 July -3 August 2016) and four to five days after Debbie (1 -2 April 2017), at locations across Mackay and the Whitsundays. The primary purpose was to provide a rapid assessment of energy distribution networks, but flight paths also covered coastal sections of interest to this study. Data were collected from an altitude of approximately 1850 ft (564 m), with a track spacing of 350 m. Reported vertical 5 accuracy was ± 0.15 m (to one RMSE, or 68% confidence). Because the LiDAR surveys were flown for terrestrial applications, data were only available landward of the waterline. The Digital Elevation Models (DEMs) interpolated from the LiDAR points were processed to remove vegetation and buildings.

Beach profile and surge limit surveys
Cross-shore beach profiles were taken through the LiDAR DEMs pre-and post-Debbie. In addition, Differential GPS (DGPS) 10 surveys were undertaken by Risk Frontiers/Macquarie University (RF/MQU) along the same profile lines approximately five months after Debbie. DGPS measurements of estimated maximum inundation extent were also made directly after Debbie by RF/MQU and DSITI. These measurements were largely based on observed debris lines (usually pumice, vegetation or coral).
An on-the-fly GNSS system with post-processing was used for the RF/MQU DGPS surveys. The mean vertical accuracy was ± 0.33 m (to one RMSE, or 68% confidence), because of not being able to calibrate against a permanent survey marker (PSM). For 15 this reason, more data points were recorded over the same locations to increase the precision of the mean value. DSITI used Real-Time Kinematic (RTK) GNSS with post-processing. At each location, calibration points were taken on PSMs adjacent to the work area, giving a better (mean) vertical accuracy of ± 0.04 m.

Structural damage surveys
A team from the Cyclone Testing Station at James Cook University collected land-based and geo-tagged photographic evidence 20 of damage caused to buildings in the areas affected by storm tide inundation and waves. These surveys were focussed on Wilson Beach and Hamilton Island as the two locations receiving the most water damage to buildings. Figure 3 shows the water level variations and surge residuals at the four gauges before and after landfall (12:40 on 28 March). 25

Storm tide and surge conditions
[ Figure 3 here]

Wind setup and barometric surge
Using Eq. (1), the IB and wind setup contributions to the total surge were estimated at each gauge (Table 2). Figure 4 shows these estimates plotted as a function of distance around Debbie at landfall. 30 [ Table 2

Wave conditions
Wave observations during Debbie are shown in Figure 5. Maximum wave heights and wave conditions at the time of peak storm tide conditions are given in Table 3. The locations of the three wave buoys are shown in Figure 1.

Wave runup and maximum water levels
Wave runup, R2%, and maximum water levels were estimated empirically using Eq. (2) for the three main impact sites to the south of Debbie; Midgeton Beach, Conway Beach and Wilson Beach ( Figure 6, locations Figure 1). Given that storm tide was the major component of total water levels, R2% was calculated for wave conditions at the time of the highest storm tide (rather than maximum 10 wave conditions). The maximum water level was defined as R2% plus the maximum storm tide level, which was taken from the nearest gauge (4.4 m AHD, Laguna Quays).
At each site, three cross-shore profiles were taken through the pre-Debbie (July/August 2016) LiDAR DEM, extending from the landward edge of the foredune to as far seaward as possible. Because the LiDAR was not water-penetrating, the seaward extent of profiles depended on the location of the waterline at the time of data capture. A single R2% value was calculated for each site based 15 on the mean slope (tanβ) of the three profiles.
[ Figure 6 here] Results are summarised in Table 4. At sites where both groups captured data over the same area, the means compared well (a difference of 0.02 m). 25

Field observations of maximum water levels
[ Table 4 here]

Volumetric beach erosion
Digital Elevation Model (DEM) surfaces were created from pre-and post-Debbie LiDAR, at the three main study sites (Midgeton,30 Wilson and Conway Beach). The pre-Debbie DEM was subtracted from the post-Debbie DEM to derive a difference model at each location showing areas of erosion and accretion ( Figure 7). The seaward extents of the difference models extended to 2 m AHD after analysis of the LiDAR data indicated this was the approximate position of the waterline during the pre-Debbie data capture. Although we cannot be sure all morphological change in Figure 7 is attributable to Debbie (because the pre-storm data was collected eight months prior) it most probably did, especially on the upper beach area of interest, because a low wave energy regime with no storm conditions occurred during the eight-month period prior to Debbie.
Erosion volumes (m 3 ) were calculated from these difference models for zones of interest at each site (Table 5, locations of erosion zones in Figure 7). Indicative transport pathways were derived from a qualitative assessment of results. 5 [ Table 5  The LiDAR-derived transects were used to calculate erosion and accretion volumes above 2 m AHD for each profile ( Table 6).
The DGPS data were not used to derive cross-sectional area change, because of the high error (± 0.33 m) and because the surveyed 20 lines were not exactly aligned to the LiDAR-derived transects. However, their inclusion clearly shows there has been very little (to no) beach recovery five months on from Debbie.

Storm tide conditions
The storm surge to the south of Debbie's landfall location (Airlie Beach) was considerably larger than to the north, because of the clockwise rotation of low pressure systems in the Southern Hemisphere, east-facing coast and consequential onshore air flow to the south of the eye (Figure 3). The largest surge residual of 2.7 m and a maximum storm tide level of 4.4 m AHD was measured at Laguna Quays, in Repulse Bay, approximately 40 km south of Airlie Beach (locations Figure 1). This was the highest water 30 level recorded at this site, since the gauge was installed in November 1994. By comparison, the maximum surge reached at Bowen, 80 km to the north, was about two metres less because of the offshore-directed air flow (Table 2).
According to Hardy et al. (2004) and Haigh et al. (2014), the storm tide level at Laguna Quays equate to a recurrence of approximately 1,000 years. The storm tide during Debbie was similar to the level reached during TC Yasi, one of the most powerful cyclones to have affected the Queensland coast, which made landfall further north near Cairns in 2011 with a storm tide level of 35 4.5 m AHD at Cardwell. The Cairns area has a smaller tidal range than Mackay and Yasi made landfall on a falling tide, so the Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License. storm tide level comprised mostly surge (3.5 m), and thus was a much rarer event (Hardy et al. (2004) and Haigh et al. (2014) both estimate the Yasi storm tide to have a > 10,000-year recurrence).
It is important to note that these estimates are based on synthetic extensions of ~ 45 years of satellite derived TC track and intensity data (BoM, 2017) -which may not capture the longer-term variance of cyclone behaviour in the Coral Sea (South West Pacific).
For example, alternative methods that are not constrained by a short observational record (such as statistical-or paleo-based 5 approaches) estimate the Yasi water level to have a lower 1,000 year recurrence (Nott andJagger, 2013, andLin andEmannuel, 2015). Extreme water levels as produced by Debbie, therefore, may be unprecedented with regards to recent observations, but not necessarily over the longer term.
Either side of Repulse Bay, water levels during Debbie were more than a metre lower, suggesting a more regionally representative surge residual was 1.1 to 1.2 m. The higher water level recorded at Laguna Quays, and inundation observed at sites within Repulse 10 Bay, suggest the surge was amplified in this area. Several geomorphological features may have been responsible for this, including; a shallow shoreface slope, concave shoreline planform and exposure to east and south-easterly winds with little sheltering by offshore islands (see Figure 1). Indeed, it may have been these features that contributed to the strong ebb current that first 'repulsed' Captain Cook in 1770, giving the bay its name.
South of landfall, wind setup was the biggest contributor to surge (accounting for 54 to 85 % of the total surge residual) because 15 wind flow was onshore-directed. Conversely at Bowen, north of landfall, the inverse barometric (IB) effect was the largest contributor to surge (85 %). This was probably because the wind was offshore-directed at this location, causing a set-down in residual water levels ( Figure 4). By comparison, McInnes and Hubbert (2003) found wind setup contributed ~ 90 % of the total surge during extra-tropical cyclone events (more than here), because the atmospheric pressure drop is not as significant.
Prior to Debbie crossing the coast, landfall was forecast to occur close to high spring tides -triggering evacuation advice for over 20 4,000 properties within the Whitsunday storm tide zone (IGEM, 2017). It eventuated that the slowing of Debbie on approach to the coast meant that landfall occurred approximately two hours after the spring high water, averting more widespread inundation.
Maximum water levels occurred (in most locations) prior to landfall, coincident with the high tide, whereas the maximum surge residual occurred 1 -2 hours after landfall, on the falling tide ( Figure 2). This may be because of Debbie's south-west trajectory ( Figure 1) and slow forward speed, meaning the southern limb (and strongest winds on the south side) reached the coast sometime 25 after the eye made landfall at Airlie Beach.
Another reason for the delay of the surge peak may be the role of tide/surge interactions. Analyses of tide gauges in the North Sea suggest that, in shallow water environments, tide/surge interactions can cause a phase shift in the total water level that means the maximum surge residual (which is the sum of the meteorological surge plus phase-shift effects) is more likely to occur on the rising, and sometimes, falling tide, but almost never coincident with high tide (Horsburgh and Wilson, 2007). Tide/surge 30 interactions are important determinants of surge maxima elsewhere in Australia (e.g. Bass Strait, Victoria -McInnes andHubbert, 2003, andBroome, Western Australia -Haigh et al., 2014), but their role in modulating extreme water levels along the Mackay-Whitsunday coast requires further investigation.

Wave conditions
Wave conditions to the south of Debbie peaked 2 to 4 hours before landfall, broadly coinciding with high tide and peak water 35 levels ( Figure 5). The largest wave height recorded at Mackay (before buoy failure) was Hmax 8.7 m, but extrapolation suggests On Debbie's north side, waves were much smaller and still coming from the ENE (onshore) as Debbie approached the coast (Abbot Point buoy, Figure 5), whereas the wind direction was blowing in a broadly offshore direction. This suggests the most energetic 5 waves to the north of Debbie, until approximately 3 hours prior to landfall, were still being produced by ambient weather, rather than the approaching cyclone. At 10:00 on 28 March, MWD abruptly swung from 68 ° (ENE, onshore) to 257 ° (WSW, offshore) as the offshore winds strengthened as Debbie moved closer to the coast, becoming the main generation source for local waves. At  Shand et al., 2011 andSplinter et al., 2014, using the same storm identification method as here). Debbie, therefore, can be regarded a significant wave power event not only locally, but for the whole east coast of Australia.

Coastal inundation and wave runup
Maximum limits of coastal inundation were estimated by summing the local measured storm tide with an empirically-derived estimate of wave runup, R2%, using the approach of Stockdon et al. (2006). These estimates were then compared with observations 20 made in the field directly after Debbie.
At Midgeton and Conway Beach, R2% and the maximum water level were estimated at 0.93 m and 5.33 m AHD ( Figure 6). This water level would have breached the dune at profiles MP1 and CB1 (southern end of Midgeton and Conway Beach) where the dune crest is lower. The difference plots in Figure 7 also indicate dune breaching and overwash occurred in these areas. The estimate of ~ 5.3 m AHD at Conway Beach broadly agrees with field evidence at this location (~ 5.1 m AHD, Table 4). 25 At Wilson Beach, R2% and the maximum water level were estimated higher, at 1.60 m and 6.00 m AHD. This is because the upper  (Table 4). This water level was still able to breach the dune crest across the whole of the beach frontage. The empirical over-estimate at Wilson Beach may be due to the large difference in the slope of the upper (tanβ 0.13) and lower (tanβ 0.01) beach, and that Eq. (2) assumes incident waves (no refraction) which is most likely not the case here. Stockdon et al. (2006) note that use of Eq. (2) for reflective beach types with variable cross-shore slopes can indeed lead to large errors in the estimation of runup. In addition, the storm tide peak at Laguna Quays may have reduced by the time it reached 35 Wilson Beach due to the shallow bathymetry at the Proserpine River entrance and attenuation of the tidal wave.
At Midgeton Beach, field measurements suggest inundation was lower than at other sites, the empirical estimate, and the local storm tide measurement, reaching only 4.3 m AHD (Table 4). This is likely to be because Midgeton town sits at a lower elevation than the fronting dunes. For the water to have reached the locations at which debris lines were measured, it must have first breached the dune crest, which is between 5 and 6 m AHD.
At Laguna Quays, the storm tide displaced pontoons from their moorings and deposited them on adjacent grass land, providing a 5 useful inundation marker, at around 4.6 m AHD (Figure 9). Laguna Quays marina is sheltered from most wave effects; therefore, the total water level only includes the storm tide with perhaps some small additional wave motion. For this reason, the field estimates are close to the maximum storm tide level recorded at the gauge (4.4 m AHD).
At Seaforth, maximum water levels were around 5.  (Table 4). By comparison, previous work suggests the storm tide contributes 15 between 65 to 75 % of the total inundation of extreme paleo-cyclones along the North Queensland coast (Nott et al., 2009;Forsyth et al., 2010). The higher contribution of storm tide during Debbie reflects the macro-tidal range of the Mackay area, which significantly reduces further north along the coast. The large tide makes it difficult to make accurate TC storm tide forecasts, because a few hours error can translate to hundreds of metres of difference in the cross-shore position of the waterline.
Wave effects were responsible for 2 to 30 % of total water levels, depending on coastal exposure, with a mean of 16 %. The 20 empirical estimate of Stockdon et al. (2006) did reasonably well in replicating the relative contribution of waves to total water levels. If a regional upper beach slope of 0.02 -0.03 is used (as per Short (2006) and measured at Midgeton and Conway Beach), and the mid-shelf wave record is de-shoaled to deep-water values, then wave effects are estimated at approximately 17 % of the total inundation. This appears to capture the mean, regional contribution of waves to total water levels, assuming sites receive incident wave energy and do not deviate far from the regional mean slope. 25

Patterns of coastal erosion
At Wilson Beach, the LiDAR surveys suggest Debbie caused erosion to the whole of the upper high tide beach (displacing a volume of approximately 4,500 m 3 , Table 5 Table 4). The LiDAR data suggests most of the eroded material accumulated at the toe of the high-tide beach in a depositional lobe about 300 m in length. This feature probably extended further seaward than was captured in the DEMs, 5 and may have resulted from a rip-cell circulation that was set up here, transporting eroded material alongshore and then offshore in front of the central beach.
There is little evidence of much landward transport (overwash) of sand at Midgeton. Likewise, the dunes did not experience any notable roll-over as seen at Wilson Beach (Figure 8 A -C). Instead, the beach responded similarly to the revetted section at Conway Beach, where eroded sand was transported seaward. This may be related to the reinforced stability of the coastal foredunes at 10 Midgeton; a community-led initiative had ensured they were well-vegetated and in most part underlain by a geotextile mat prior to Debbie (Zavadil et al., 2017).

Another commonality between Conway and Midgeton Beach was the major erosion that occurred near tidal creeks. At Conway
Beach, storm water transported seaward down the creek caused large, localised erosion at the mouth (~ 3,700 m 3 -more than the rest of the entire frontage), and was deposited at the toe of the upper beach. The difference plot is cut off at + 2 m AHD, but 15 suggests a 200 m-wide ebb tidal delta formed in this area.
Similarly, at Midgeton Beach, significant erosion occurred at the two creek entrances, with the southern creek eroding an almost identical volume to the Conway Beach creek (~3,700 m 3 ). Sand from both the northern and southern creeks looks to have washed out in a south-westerly direction and been deposited offshore.

Storm erosion demand and offshore transport 20
The storm erosion demand during Debbie was estimated from the beach profiles extracted from the pre/post storm LiDAR DEMs.
Storm demand refers to the erosion that occurs to the upper beach during a storm event (typically above 0 m AHD in Australia). It is an important parameter used for the design of coastal setback. Here, the storm demand can only be calculated above 2 m AHD because this was the limit of the LiDAR data. However, beach profiles show that all erosion occurred well above 2 m AHD ( Figure   8); therefore, volumes should be comparable to those calculated to 0 m AHD in other studies. 25 The average storm demand at Midgeton Beach was 15.6 m 3 m, and at Conway and Wilson Beach it was roughly half this (7.3 and 9.2 m 3 m, respectively) ( Table 6). The lower volumes at Conway and Wilson Beach may be because these locations are afforded some protection from easterly waves by Cape Conway, whereas Midgeton Beach is open to both easterly and south-easterly wave attack (Figure 1).
These volumes are an order of magnitude smaller than those of wave-dominated coasts (e.g. 150 -250 m 3 m is a typical range for 30 New South Wales and Southeast Queensland) but comparable with modelling studies of the Mackay coast, which suggest the 100year ARI erosion volume is between 1 and 16 m 3 m, depending on aspect (Mariani et al., 2012). To our knowledge, this study represents the only set of observations of TC storm erosion demand for north Queensland.
Most profiles saw a net loss of sand above 2 m AHD (average 78 % loss, Table 6), suggesting that the eroded volume was accumulated below (seaward of) this level. The exceptions were CB2 and CB3 at Conway Beach which experienced a net gain 35 above 2 m AHD, because the depositional lobe that accumulated at the toe of the upper beach extended into the surveyed portion of the profiles. This suggests that at all study sites, sand transport was predominately offshore. At Midgeton and Conway Beach, this may have been in part due to the stabilisation of the coastal frontage. Even at Wilson Beach, however, where the coastal Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License. foredune was not stabilised, results indicate that most sediment was transported seaward of the high-tide beach (beyond the limits of our data capture), with only a small portion of the eroded volume overwashed landward.
Our results suggest that while overwash and landward transport clearly occurred during Debbie, this only accounted for a small portion of the total eroded volume of the upper beach, and the majority was transported offshore and placed around the toe of the high-tide beach. This may have been because at most sites, coastal water levels did not fully breach the foredune, inducing offshore 5 transport. Similarly, Sallenger et al. (2006) found that the net direction of sediment transport during Hurricane conditions along the US East Coast is a function of the water level relative to the foredune height (i.e. overwash occurs when this ratio is positive, else there is offshore transport). While this may be true at the peak water level, as the tide recedes and storm wave conditions continue, erosion is focussed lower down the profile and sediment may be transported offshore. The result can be a mix of on-and off-shore transport even at sites inundated at the peak water level (as was the case at Wilson Beach). 10

Beach recovery and implications for coastal management
Beach profile measurements taken five months after Debbie show very little, if any, beach recovery occurred over this period.
While the quality of the DGPS measurements was not sufficient to derive volumes, their inclusion in Figure 8 shows the profiles are very similar to those captured directly after Debbie.
This can be attributed to the wave climate of the region. Tropical cyclones often impact coastal areas that are equilibrated with a 15 low-energy, tide-dominated regime because of their latitude and small regional wind-wave climate. This is particularly the case along the North Queensland coast, where a significant amount of wave energy is dissipated over the Great Barrier Reef (Gallop et al., 2014). This is exemplified in the wave buoy record pre-and post-Debbie; prior to Debbie, storm wave conditions (Hs ≥ 2 m) had not been exceeded since the start of January (3 months prior). Likewise, after Debbie, it took over three months -to mid-June -to accumulate the same amount of wave power as was exerted over nine days during Debbie. It was at this point (June 2017) that 20 the beach recovery surveys were undertaken. The large energy difference that exists between modal and cyclonic wave conditions in this region means TC erosion impacts have a permanent impact on the landscape and require a management response to restore beach amenity and access. In some geomorphic settings natural rebuilding of the frontal dune will occur over time, however this may take a much longer period than between successive cyclones. For example, photographic evidence 12 months on from Debbie suggests some minimal sand accretion had 25 begun to occur at Midgeton Beach (Zavadil pers. comms.) Where natural recovery processes occur, coastal management actions should be designed to encourage and accelerate natural rebuilding of the beach. For example, the shoreline at Midgeton Beach has eroded and recovered many time over recent decades (Zavadil et al., 2017) and the underlying tendency for many beach ridge systems in North Queensland is progradation. Natural recovery processes become important for coastal management particularly when engaging local communities about erosion 30 management options.

Damage to buildings
Although Debbie's surge did not coincide with the highest tide, some buildings at Wilson Beach and Hamilton Island suffered varying degrees of damage from storm tide and waves. The most severe structural damage was observed at Wilson Beach. Field observations indicated three processes were responsible; 1) overland flow causing scour around footings, foundations, sub-floor 35 structures and piles; 2) lateral forces exerted by waves on external cladding elements such as doors and windows, and; 3) direct inundation causing damage to internal wall linings, floors and building contents.
Fast-flowing water associated with in-flowing (landward) and out-flowing (seaward) storm tide inundation can cause localised scour around a building and its foundations. In-flow tends to be more uniform across an area, but is exacerbated by wave action.
Post-Debbie surveys suggested that the out-flow was not powerful enough to develop overland ebb channels, but that both in-flow and out-flow combined caused some scouring of building footings with subsequent settlement (Figure 10 D).

5
[ Figure 10 here] In areas affected by tidal creeks, the in/out flow pattern was further complicated. At Wilson Beach, the storm tide entered from two directions; the south (beach) side, where buildings were directly exposed to wave action, and the north (creek) side, which was responsible for most inundation of the community (Figure 10 A). Wave action on beach-side properties caused damage to external 10 cladding elements and broke windows and doors on some houses. Wave action also caused substantial drag forces on the substructure of buildings with suspended floors. The window and cladding damage shown in Figure  At present, building damage curves that relate cyclone intensity to structural damage treat wind and water damage separately. Field inspections post-Debbie suggest that the upward wind action exerted on roofs combined with the upward wave and storm tide action on the underside of floors, in some instances, combined to wash buildings off their stumps. This highlights the need to develop TC building damage relationships that account for the combined impacts of water and wind, rather than treating them separately. 20

A shared information platform for coastal extremes
Aside from the technical findings of this research, this paper provides a case study of data sharing across industry, government and academia in a natural hazard risk context. The research was facilitated through an information-sharing event (the TC Debbie forum, organised by the Bureau of Meteorology and hosted by the Queensland Fire and Emergency Service) and is built on the principles of openness and collaboration. Each group collected data independently, according to their own time and cost constraints. Not all 25 data was collected with coastal hazard monitoring as the primary purpose (e.g. Fugro Roames LiDAR). The value added by unifying this information came at minimal additional cost, but was essential to understanding the relationships between the extreme hydrodynamic forcing and coastal impacts.
At a high level, the economic value of data sharing is well recognised (e.g. Deloitte, 2014, World Bank, 2014), but in Australia cross-sector data exchange during natural disasters is lacking (Gissing, 2017). Data sharing can lead to direct and (mostly) indirect 30 economic benefits such as efficiency gains, a reduction in competitive bias, better planning and prediction by regulatory agencies, improved price signalling by insurance, and facilitates research and innovation. However, reluctance to share information, a lack of co-ordination and standardisation and high costs of data collection often restrict data sharing and encourage the continuance of a piece-meal approach.
In Australia, it is estimated that the centralisation of key natural peril data through the development of open data platforms could 35 save the economy over $2 billion over the period to 2050 (Deloitte, 2014). Because of rising coastal population density (both globally and in Australia), coastal extremes are likely to become particularly costly events. We believe a shared information platform is needed for coastal extremes in Australia, to provide a repository of observational coastal data for historical extremes.
In this way, all relevant data and metadata is indexed on an event-by-event basis in a single location. This may include information Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License. on the meteorology (winds, air pressure), hydrodynamics (waves, water levels), coastal response (erosion, inundation) and structural (water) damage aspects of individual events that have impacted the Australian coast. At present, a large number of organisations collect this type of information (and most is, in principle, publicly available) but it is often siloed within a discipline or sector. The centralisation of this data would simplify access and increase its collective value, as has been demonstrated in this paper. 5

Conclusions
Tropical Cyclone (TC) Debbie was potentially one of the most powerful wave-and surge-producing cyclones to have made landfall on the northeast coast of Australia since monitoring began. Despite localised impacts, Debbie was a near-miss in terms of widespread coastal flooding. If landfall had occurred two hours earlier (i.e. if the surge had coincided with high tide), then the maximum storm tide may have been ~ 18 % higher. Total water levels would likely have exceeded 6 m AHD on the southern side, 10 which is higher than most dune crest elevations. It eventuated that Debbie slowed on approach to the coast meaning landfall occurred on the falling tide, averting widespread inundation.
The maximum recorded storm tide was 4.4 m AHD, and the maximum recorded surge was 2.7 m (Laguna Quays). These were the highest levels recorded at this location (since the gauge was installed in 1994), with previous work suggesting a storm tide recurrence of approximately 1,000 years. These maxima are likely the result of surge amplification within Repulse Bay with sites 15 north and south recording levels over a metre lower. The considerable variation in water levels over relatively small spatial scales (~ 20 km between gauges) highlights the need for suitably high resolution coastal models and data that can capture the effects of local-scale geomorphology on the hydrodynamics.
Waves, water levels and coastal impacts were considerably larger south of the landfall site, because of the clockwise rotation of low pressure systems in the Southern Hemisphere, east-facing coast and onshore air flow to the south of the eye. An average coastal 20 water level (storm tide plus waves) for the region Airlie Beach to Mackay was 5.2 m AHD, ranging between 3.8 m to 5.9 m AHD depending on exposure. Total inundation varied dramatically close to tidal creeks. At Wilson Beach, flooding was caused by the storm tide entering the creek and inundating the town from the behind, in addition to coastal flooding. The largest beach erosion volumes were also found near to tidal creeks at Conway Beach and Midgeton Beach.
Results suggest the storm tide contributed ~ 84 % of total water levels, which is higher than previous estimates for tropical cyclones 25 in Australia (65 -75 %), but may reflect the locally large tidal range. Wave runup was responsible for ~ 16 % of total water levels, varying between 2 and 30 % at sites depending on coastal exposure. Wind setup contributed between 54 and 85 % of the total surge residual south of landfall, but probably contributed to a set-down in water levels north of landfall.
Debbie may have produced the most extreme wave conditions in this area for at least the past 40 years, with an (extrapolated) maximum wave height exceeding 11 m. The slow forward-speed led to an unusually long duration of storm waves and large 30 cumulative wave power (7.4 mWh m), that began three days before landfall and continued for another eight days. This caused significant beach erosion in areas exposed to easterly and south-easterly waves, with storm demand volumes between 6 and 19 m 3 m. Most of the eroded material was placed near the toe of the high-tide beach. While landward transport clearly occurred, this only accounted for a small portion (~ 25 %) of the total eroded volume of the upper beach. In most places, coastal barriers were not breached and this may have been the cause for the lack of overwash deposits. 35 The large energy difference that exists between modal and cyclonic wave conditions in this region (and many other TC-affected global coastlines) means TC erosion impacts are often a permanent feature on the landscape within planning timeframes and require Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org/10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License. a management response to restore beach amenity and access. Surveys undertaken five months on from Debbie indicated very little beach recovery had occurred. Photographic evidence 12 months on also indicated very small-scale geomorphic change.
To our knowledge, this study represents the only observational analysis of hydrodynamic drivers and concurrent coastal impacts for a tropical cyclone in Australia, and the only set of observations of TC storm erosion demand in north Queensland. The data were collected on a largely ad-hoc basis on a limited budget. We advocate for a more formalised collaborative approach to 5 collecting and archiving TC coastal impact data in Australia, as simple as a publicly available, standardised data depository, to help improve our understanding and prediction of TC-related coastal hazards in the future.

Author contribution
TM conceived the idea for the paper and prepared the manuscript. TM prepared the manuscript with contributions from all coauthors, in particular DM who provided significant input on hydrodynamic forcing. Coastal impact surveys were carried out and 10 analysed by TM, JS, SBL, JM, GB and EZ. IDG contributed to the analysis of field observations and coastal process implications.       Figure 7). Maximum storm tide elevation was taken from the nearest gauge (Laguna Quays) (grey solid line), and maximum water elevation (grey dashed line) was defined as R2% + storm tide elevation. R2% corresponds to the expected level of exceedance of 2% of the wave runups during wave conditions at the time of maximum storm tide elevation (Hs0 5 m L0 131 m, at 12:00 28 Mar). R2% was calculated for each location 5 based on the mean slope (tanβ) of the profiles. The locations at which tanβ was calculated for each profile are shown (base of dune to toe of upper beach, black circles).
Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci.    Figure 9). Green circles show alongshore-averaged DGPS elevations along the same profile lines ~ 5 months after Debbie. Only the lower 'recovery' profile is shown at Midgeton Beach because heavy vegetation hindered DGPS data collection. Only the lower 'recovery' profile is shown at Conway Beach because the upper portion is a revetment. No 5 'recovery' profiles are shown for Wilson Beach because beach recharge occurred in the interim. Also shown is the estimate of maximum water elevation from Figure 6 and storm demand. (F) indicates the revetment has moved during Debbie along profile CB3, consistent with observations (DSITI, 2017). The storm demand calculation for CB3 does not include the revetment change.
Nat. Hazards Earth Syst. Sci. Discuss., https://doi.org /10.5194/nhess-2018-110 Manuscript under review for journal Nat. Hazards Earth Syst. Sci. Discussion started: 24 April 2018 c Author(s) 2018. CC BY 4.0 License.      ** at each site, the residual between the mean of the maximum water level observations and the storm tide height from the nearest gauge was used 5 to estimate the contribution of wave effects. At Midgeton, the storm tide is greater than the total water level estimate, for reasons described in the text.