Examining the impact of the Great Barrier Reef

44 Coral reefs may provide a beneficial first line of defence against tsunami hazards, though this 45 is currently debated. Using a fully nonlinear, Boussinesq propagation model, we examine the 46 buffering capacity of the Great Barrier Reef against tsunamis triggered by several hypothetical 47 sources: a series of far-field, Solomon Islands earthquake sources of various magnitudes (Mw 48 8.0, Mw 8.5, and Mw 9.0), a submarine landslide source that has previously been documented 49 in the offshore geological record (i.e. the Gloria Knolls Slide), and a potential future landslide 50 source (i.e. the Noggin Block). We show that overall, the Great Barrier Reef acts as a large51 scale regional buffer due to the roughness of coral cover and the complex bathymetric features 52 (i.e. platforms, shoals, terraces, etc.) that corals construct over thousands of years. However, 53 the buffering effect of coral cover is much stronger for tsunamis that are higher in amplitude. 54 When coral cover is removed, the largest earthquake scenario (Mw 9.0) exhibits up to a 31% 55 increase in offshore wave amplitude and estimated run-up. These metrics increase even more 56 for landslide scenarios, where they tend to double. These discrepancies can be explained by the 57 higher bed particle velocities incited by higher-amplitude waves, which leads to greater 58 frictional dissipation at a seabed covered by coral. At a site-specific level, shoreline orientation 59 relative to the reef platforms also determines the degree of protectiveness against both types of 60 tsunamis, where areas situated behind broad, shallow, coral-covered platforms benefit the 61 most. Additionally, we find that the platforms, rather than gaps in the offshore reef structure, 62 tend to amplify wave trains through wave focussing when coral cover is removed from 63 simulations. Our findings have implications for future tsunami hazards along the northeastern 64 Australian coastline, particularly as the physiological stressors imposed by anthropogenic 65 climate change further exacerbate coral die-off and reductions in ecosystem complexity. 66 Therefore, areas that experience a protective benefit by the Great Barrier Reef’s platforms 67 could be disproportionately more vulnerable in the future. 68

Tsunamis threaten low-lying coastal communities around the world. Coral reef ecosystems, 76 many of which are positioned between tsunami source regions and densely-populated 77 shorelines (Figure 1), could provide a broad, cost-effective first line of defence for coastal 78 zones (Ferrario et al. 2014). While field-based studies suggest that coral reefs induce efficient 79 energy attenuation in wind waves due to their structural complexity (Sheppard et al. 2005; along neighbouring coastlines. While there is near-universal consensus that inter-reef passages 98 (or "gaps/openings" between reefs) can amplify tsunami waves, some argue that these 99 The Great Barrier Reef (GBR), the world's largest coral reef system, is an iconic feature of 125 Australia's coastal landscape. Despite Australia's proximity to the seismically active source-126 regions (Dominey-Howes 2007; Davies and Griffin 2018), the manner in which tsunami 127 behaviour is regulated by the GBR, which partitions Australia's coastline from these 128 convergent margins, is not well understood ). Additionally, the discovery 129 of large (volume > 30 km 3 ) landslide scars and slumps on the nearby continental slope (Puga-130 Bernabéu et al. 2016, 2019) warrants an investigation into the GBR's ability to protect against 131 landslide-generated tsunamis. Though believed to occur less frequently than their coseismic 132 counterparts, landslide-generated tsunamis such as the 1998 Sissano, Papua New Guinea event 133 (Synolakis et al. 2002) can occur suddenly within close proximity to the shoreline, causing 134 significant localized damage and limiting opportunities for warning and swift response. This, 135 along with the existence of possible paleo-tsunami deposits along the adjacent coastline (Nott 136 1997), underscores an urgency to quantify the GBR's widely-speculated role as a regional 137 buffer from these hazards (Baba et  run-up, such as the extent of coral cover, the nature and proximity of the tsunami triggering 149 source, and site-specific variability in coastal bathymetry and topography. Therefore, following 150 a tsunami event, it is difficult to retrospectively ascertain the impact of coral reefs in isolation 151 from these other site-specific factors. Numerical simulations can provide additional insights 152 into tsunami behaviour (e.g., Kunkel et al. 2006), where experiments can be designed to 153 systematically test the impact of coral cover and reef platform bathymetry on tsunami 154 attenuation while keeping all other parameters, initial conditions, and boundary conditions 155 constant (e.g., Kunkel et al. 2006). Previous studies have aimed to assess the overall impact of 156 the GBR on tsunami propagation using numerical simulations (Baba et

172
Using numerical modelling, we evaluate the GBR's ability to shield the northeastern Australian 173 coastline from a range of hypothetical, though plausible tsunami sources. Firstly, we consider 174 a Solomon Islands earthquake source over various magnitudes (Mw 8.0, Mw 8.5, and Mw 9.0). 175 Additionally, we consider two near-field landslide tsunami sources: 1) the largest documented In the first of a series of tsunami propagation model runs, for each tsunami source, we 181 numerically simulate the tsunamis assuming healthy coral cover conditions (i.e. "coral-covered 182 platforms" scenarios), where reef platforms are prescribed high roughness to reflect their 183 structural complexity (Nelson 1996). Then, we simulate the tsunamis with smoothed reef 184 platforms (i.e. "smooth platforms" scenarios), where we isolate the impact of live coral cover 185 on wave attenuation (Sheppard et al. 2005 The central northeastern Australian margin is a passive margin characterised by a relatively 197 broad (~60 km) continental shelf ( Figure 2). The spring tidal range varies from north to south, 198 but the region is generally meso-to macro-tidal (Andrews and Bode 1988). Several 199 environmental factors favour coral reef growth on the mid-to outer-continental shelf, including 200 the region's tropical climate, shallow seas, far proximity from terrestrial run-off, and nutrient-201 poor oceanographic conditions. Over hundreds of thousands of years of eustatic sea level 202 fluctuations, these coral reef ecosystems have constructed large (up to ~300 km 2 ) submerged 203 and semi-submerged carbonate platforms, pinnacles, and terraces, which comprise the offshore 204 reef structure (Hopley et al. 2007;Hinestrosa et al. 2016). This reef structure, which underlies 205 the modern generation of living coral cover, extends roughly 2,300 km along the mid-to outer 206 shelf (Hopley et al. 2007). On the central margin, broad, arcuate patch reef platforms are 207 separated by relatively wide (up to ~10 km) inter-reef passages, or "gaps" (Figure 3). While 208 these passages are wide enough to allow some wind waves to propagate through to the inner 209 shelf, much of the energy transferred by wind waves is attenuated atop the reef platforms 210   incorporates the widely-implemented Okada elastic half-space formulation, which relates 280 earthquake geometric source parameters (e.g. fault width, length, strike, dip, etc.) to the initial 281 free surface deformation (Okada 1985      and were thus adopted here (see Table 1). For the Gloria Knolls Slide, slide dimensions were 309 It is important to note here that although Geowave also has the ability to simulate tsunami 360 generation and propagation by both coseismic slip and landslide sources, we opted to use 361 updated models that more explicitly resolve processes involved in landslide tsunami generation 362 coral cover is structurally complex on the meter to sub-meter scale (Nelson 1996;Graham and 383 Nash 2013). We hereafter refer to the structural complexity of coral cover as "ecosystem-scale" 384 complexity. In a modelling context, this "ecosystem-scale" complexity cannot be resolved in 385 the computational domain and must be parameterized (see Section 3.4.1). Secondly, the GBR 386 exhibits structural complexity at the >1 km scale. The reef structure itself is composed 387 primarily of completely submerged or semi-submerged carbonate platforms. These features 388 create complex positive relief on the submerged continental shelf, and much of this relief (aside 389 from smaller, deeper pinnacles and terraces), is resolved by the 100 m-resolution 3DGBR 390 bathymetric dataset (Beaman 2010). Thus, the reef structure can be adequately resolved in the 391 computational domain. We hereafter refer to complexity introduced by the reef structure as 392 "bathymetric-scale" complexity. where CD is the non-dimensional bottom friction coefficient, is the density of water, and U 403 is the particle velocity at the seabed. A variable bottom friction coefficient was established 404 throughout the domain, where it was altered according to the presence or absence of coral cover 405 on reef platforms. A value of CD=0.1522 was prescribed to reef platforms to simulate coral 406 cover (average depth of platforms  14.9 m). This value was obtained from a prior field 407 investigation of the hydraulic roughness of coral reefs, which was conducted at John Brewer 408 Reef, a reef platform within the GBR that lies close to the study region (Nelson, 1996 To test the impact of coral cover on tsunami attenuation, the "coral cover" scenarios were then 417 compared to "smooth platform" scenarios, where coral cover was effectively removed. In the 418 "smooth platform" scenarios, all areas of the bottom boundary, reef platforms included, were 419 prescribed a standard bottom friction coefficient value of CD = 0.0025. 420  for the "coral-covered platforms" scenario, 6.4 cm for the "smooth platforms" scenario, and 6.7 cm for the "no 506 reef platforms" scenario. Offshore wave amplitudes were interpolated along the 25 m isobath.

508
For the hypothetical Mw 8.5 Solomon Islands earthquake scenario, the GBR, both in terms of 509 its ecosystem-scale and bathymetric scale complexity, appears to have slightly more impact on 510 offshore tsunami amplitudes and estimated run-up. When coral cover is present (Figure 8a), 511 wave amplitudes landward of the GBR range from ~5-10 cm, with an Rmax estimate of ~26 cm.

527
The GBR has a much more substantial impact on the propagating tsunami when considering 528 the hypothetical Mw 9.0 Solomon Islands earthquake source. Overall, the Mw 9.0-generated 529 tsunami is significantly larger in amplitude than its smaller-magnitude counterparts. When  tsunami attenuation is sizeable for the landslide-generated cases considered on this margin. 590 Turning firstly to the previously-termed "worst-case scenario" for the Gloria Knolls Slide 591   Solomon Islands earthquake scenario (Figure 9), the Gloria Knolls submarine landslide 703 scenario ( Figure 11) and the Noggin Block potential submarine landslide scenario (Figure 12). 704 These declines in wave amplitude are driven by elevated frictional dissipation over coral-705 covered reef platforms. We eliminate the possibility that wave breaking contributed to energy 706 dissipation, as wave-breaking was not detected in any of the simulations due to the tsunamis' 707 large wavelengths in comparison to their amplitudes. These results reaffirm the prevailing 708 notion that the GBR acts as a regional buffer to tsunamis (Baba et  which allows the cumulative impact of frictional dissipation to dominate. Therefore, we 713 propose that the effect of live coral cover should be directly incorporated into future hazard 714 assessments of the northeastern Australian margin, as we anticipate it will have a detrimental 715 impact on propagating tsunamis. 716

717
The energy-diminishing impact of coral cover becomes most apparent when comparing the 718 "coral-covered platforms" simulations with the "smooth platforms" simulations. When coral 719 cover is removed, amplitudes increase across each source scenario tested here. Notably, run-720 up projections increase as much as 24% for the Mw 9.0 earthquake source (Figure 10  This implies that the degree of coral-induced frictional dissipation at bed is different across 740 source scenarios. Our findings demonstrate that these differences in frictional dissipation are 741 directly related to wave amplitude (and thus, wave energy). Particle velocity (note: this is 742 different to wave celerity) is a function of wave amplitude (Nielsen 1992), and therefore, waves 743 of differing amplitudes experience different degrees of dissipation due to shear stress at bed. 744 This amplitude-mediated discrepancy in particle velocity is best exemplified by comparing 745 earthquake scenarios, where tsunami amplitude was altered by changing the magnitude and 746 slip displacement of the initial coseismic source (Figure 4, see Table 1 for source parameters). 747 For the Mw 8.0 Solomon Islands earthquake scenario, bed particle velocities are relatively low 748 (< 1 cm/s) throughout the computational domain given the relatively low tsunami amplitudes 749 produced by the source. However, for the Mw 8.5 and Mw 9.0 earthquake scenarios, particle 750 velocities are much higher on the shelf (> 5 cm/s). Moreover, in their corresponding "smooth 751 platforms" simulations, particle velocities are more elevated atop the reef platforms than in the 752 "coral-covered platforms" simulations, which further reflects the dissipative effect of coral 753 cover. As wave energy dissipation through shear stress is proportional to the square of the 754 particle velocity (see Eq. 1), the higher velocities computed for higher-magnitude earthquake 755 tsunamis result in greater overall wave energy dissipation via bottom friction when coral cover 756 is present. This also explains why a relatively large degree of attenuation is observed for the 757 landslide-generated tsunamis, both of which produce similarly high waves (9 m and 3.5 m for 758 the landward-propagating waves, respectively). Our results show that tsunami amplitude, 759 which ultimately depends on the magnitude and proximity of the triggering source, should also 760 be considered when examining the buffering capacity of natural defences such as coral reefs. 761

762
While the GBR generally acts as a buffer to tsunami wave energy, despite its namesake, the 763 GBR itself does not form a continuous barrier on the mid-to outer shelf, especially in the 764 central region (Figure 3). As a result, the buffering effect offered by coral cover varies 765 considerably alongshore. Turning again to the Solomon Islands earthquake scenarios ( Figure  766 10), when coral cover is removed, the largest increases in wave amplitude and run-up tend to 767 occur landward of broad reef platforms (see also Figure 9a, b). On the other hand, areas that 768 lie between inter-reef passages, or gaps, exhibit smaller increases in amplitude and run-up. This 769 phenomenon is consistent across source scenarios, and it is particularly pronounced in cases 770 where tsunami amplitudes are relatively high. This implies that the protectiveness offered by 771 coral cover varies alongshore because of platform placement; if coral-covered platforms 772 (particularly broad platforms) are positioned between the incoming tsunami and the shoreline, 773 they are more inclined to dampen the tsunami. 774

775
To summarise, reef cover contributes substantially to the overall buffering capacity of the 776 GBR, which is consistent with previous findings (e.g., Kunkel et al. 2006). However, the 777 GBR's buffering capacity for any given location alongshore depends on various site-specific 778 factors, including the presence of coral cover, the relative positioning of the platforms, and 779 tsunami amplitude.   coral cover appears to fully or partially counteract these focussing effects, where waves 799 subsequently dampen after growing in amplitude over the platforms (e.g., Figure 9). 800 Consequently, smoothing the domain tends to enhance the platforms' ability to focus wave 801 energy. This is demonstrated by the higher-amplitude, landward wave trains shown in wave 802 amplitude distributions (e.g., Figure 9). Some platforms appear to more effectively focus wave 803 energy than others, and we suspect this is due to factors such as reef morphology, size, and 804 submergence depth. A more systematic investigation of platform characteristics warranted to 805 test this hypothesis, particularly as coral reef cover is expected to decline in the future. simulations, porous gaps in the reef structure certainly permit wave energy to pass through to 822 the coastline. However, there is little evidence to support the notion that the gaps amplify 823 waves. In fact, due to focussing, amplification of wave amplitudes occurs over the platforms 824 rather than between them (e.g., Figure 9, Figure 11). In the case of the GBR, many of the 825 platforms appear to be wide enough, deep enough, far enough apart, and far enough from the 826 coastline such that the inter-reef gaps do not pose a significant hazard. This is in contrast to 827 many fringing reef systems, where gaps can be quite narrow, shallow, and close to shore. We 828 therefore suggest that for the GBR, the wave focussing ability of platforms may be of greater 829 concern for the northeastern Australian coastline than the presence of gaps in the reef structure. 830

831
Overall, the GBR's underlying bathymetric structure contributes significantly to its buffering 832 capacity, and this becomes apparent when platforms are removed from simulations (see Figure  833 10 and Figure 13). When platforms are removed, waves are permitted to propagate smoothly 834 and uninterruptedly across the shelf, highlighting the highly obstructive nature of the platforms Indeed, our simulations showcase shoaling and focussing on platforms, which locally augment 873 wave amplitudes at the intra-platform scale. A more rigorous inundation study would be needed 874 to confirm whether this translates to increased hazard within the lagoons, shoals, and islands 875 that rest within the platforms. Therefore, coral reefs could have either beneficial or detrimental 876 effects on the overall hazard depending on the type reef system in question and the proximity 877 of coastal communities and assets to the site of the most severe shoaling/focussing. In the 878 debate surrounding reef protectiveness against tsunamis, a distinction must be made between 879 fringing reef systems and offshore barrier systems, as they have different implications for 880 proximity to wave focussing effects, and therefore, exposure. basis to support the fact that coral reefs can dissipate tsunami wave energy, reducing the 886 tsunami hazard. However, this overall reduction in hazard may not be sufficient to completely 887 reduce the physical vulnerability and exposure of coastal communities (Uslu et al. 2010). When 888 discussing the buffering role of reefs, many have highlighted that despite being within close 889 proximity to reefs, coastal assets have nonetheless been destroyed during tsunami events (e.g., 890 Baird et al. 2005), leading some to conclude that coral reefs provide no protective benefit to 891 coastal communities. In these cases, the reefs could very well have buffered the overall tsunami 892 hazard, reducing the overall inundation and run-up extent. However, this protective benefit 893 may not have been sufficient to completely shield coastal communities that were situated close 894 to shore. Care must be taken when retrospectively interpreting the role that coral reefs may 895 have played in reducing tsunami hazard along a shoreline, and a clear distinction should be 896 made between hazard reduction and risk reduction, which lies at the intersection between 897 hazard, exposure, and vulnerability. 898 899 5.5. Study limitations and future work 900 Uncertainties persist that could complicate such future tsunami hazards assessments in coral 901 reef environments. Firstly, at the ecosystem-scale, the relationship between coral rugosity and 902 community composition requires more precise quantification on an intra-reef platform scale 903 (Rogers et al. 2016). This will continue to be a pressing task in the future, as profound 904 ecological shifts may be precipitated by both the immediate aftermath of the tsunami impact 905 and longer-term environmental changes, thus affecting ecosystem-scale structural complexity 906 rugosity is still not precisely known. These ecosystems should be stringently monitored to 910 better assess how coastal hazard severity as a whole will be transformed in these areas. 911 Additionally, the approach used to parameterize bottom shear stress, though very common both 912 in the field and in modelling studies, may need to be reconfigured to account for more complex 913 tsunami interactions and subgrid turbulent dissipation within the 3D reef structure (Lowe et al. Finally, our study was not designed to provide a reappraised, comprehensive hazard assessment 931 for the northeastern Australian coastline although our findings suggest that the reef's role 932 should be considered in future assessments. That being said, we stress the need for a robust 933 parameterization of reef roughness (Nelson 1996; Rosman and Hench 2011). Furthermore, as 934 indicated by sensitivity analyses (see Online Resource 2), these propagation simulations 935 require high spatial resolution (200 m for earthquake sources and 100 m or less for landslide 936 sources) in order to properly capture the reef structure and to resolve complex tsunami-reef 937 interactions. While this increases computational demand, we nonetheless deem it worthwhile 938 to consider the role of the reef, as current assessments may be over-estimating tsunami risk in 939 northeast Australia. Additionally, a more meaningful assessment of the submarine landslide 940 tsunami hazard is needed to better understand the timing, frequency, and magnitude of these 941 events. In the future, it may be worth considering more complex failure dynamics (i.e. landslide 942 deformation and two-way coupling with the water column), which could alter the run-up results 943  This study demonstrates the nuanced interactions between tsunamis and coral reef systems. In 950 agreement with previous work we find that the Great Barrier Reef (GBR), both in terms of 951 coral cover and larger-scale bathymetric complexity, acts as a large-scale regional buffer 952 against tsunamis. However, the reef appears to provide greater protection against higher-953 amplitude tsunamis due to the larger computed particle velocities at bed, which directly dictates 954 the degree of frictional dissipation through shear stress. Additionally, we find that the 955 protectiveness offered by the GBR locally depends on coral cover and platform distribution. 956 We also find that wave focussing by reef platforms could pose a greater hazard than the gaps 957 between platforms, which have been previously thought to amplify waves. In the context of 958 the larger debate about whether coral reefs reduce tsunami hazards for coastal communities, 959 we conclude that differing interpretations can be reconciled when considering site-specific 960 factors. 961 962