Distant tsunami inundation modelling incorporating dune failures and river flow in Christchurch

Transcription

1 Distant tsunami inundation modelling incorporating dune failures and river flow in Christchurch Prepared for Environment Canterbury November 2017 Distant tsunami inundation modelling incorporating dune failures in Christchurch

2 Prepared by: Emily Lane Alison Kohout Julian Sykes Jade Arnold Jochen Bind Shaun Williams For any information regarding this report please contact: Emily Lane Hydrodynamics Scientist Hydrodynamics National Institute of Water & Atmospheric Research Ltd PO Box 8602 Riccarton Christchurch 8011 Phone NIWA CLIENT REPORT No: CH Report date: November 2017 NIWA Project: ENC17502 Quality Assurance Statement Reviewed by: David Plew Formatting checked by: Fenella Falconer Approved for release by: Helen Rouse Distant tsunami inundation modelling incorporating dune failures in Christchurch

3 Contents Executive summary Introduction... 9 Use of this report Caveat Tsunami inundation modelling Source model and initial conditions Inundation modelling and grid Model outputs Uncertainties Model results dune failure scenarios Maximum inundation and speed based on existing dune profiles Scenario-specific maximum inundation and speed Model results river scenarios Maximum inundation and speed Avon and Heathcote rivers Maximum inundation and speed Waimakariri River Conclusion Acknowledgements References Appendix A GIS Layers Appendix B Reconstruction of Waimakariri River bathymetry from cross-sections 50 Appendix C Glossary of terms Tables Table 4-1: Bridges on Waimakariri, Avon and Heathcote rivers with heights above LVD37 and modelled maximum tsunami height at those locations. 37 Distant tsunami inundation modelling incorporating dune failures in Christchurch

4 Figures Figure 1-1: Overview of study site including modelled dune breach locations. 11 Figure 1-2: Existing (black) and modified (green) dune profiles along the 2 km stretch from Waimairi to New Brighton. 12 Figure 2-1: Difference in topography based on updated datasets between this study and Lane et al. (2014). 15 Figure 2-2: Dune breach scenario Figure 2-3: Dune breach scenario Figure 2-4: Dune breach scenario Figure 2-5: Dune breach scenario Figure 2-6: Dune breach scenario Figure 2-7: Stop-banks modelled in this scenario. 20 Figure 3-1: Maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS (Model tsunami taken from Lane et al. 2014). 25 Figure 3-2: Maximum flow speed for Christchurch assuming the largest wave arrived at MHWS (Model tsunami taken from Lane et al. 2014). 26 Figure 3-3: Modified dune scenario 1 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. 27 Figure 3-4: Modified dune scenario 1 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 28 Figure 3-5: Dune breach scenario 2 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. 29 Figure 3-6: Dune breach scenario 2 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 30 Figure 3-7: Dune breach scenario 3 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. 31 Figure 3-8: Dune breach scenario 3 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 32 Figure 3-9: Dune breach scenario 4 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. 33 Figure 3-10: Dune breach scenario 4 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 34 Figure 3-11: Dune breach scenario 5 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. 35 Figure 3-12: Dune breach scenario 5 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 36 Figure 4-1: Maximum inundation depth in the Avon and Heathcote rivers. 40 Figure 4-2: Maximum tsunami flow speeds in the Avon and Heathcote rivers. 41 Figure 4-3: Maximum inundation depth along the Waimakariri River. 42 Figure 4-4: Maximum speed along the Waimakariri River. 43 Distant tsunami inundation modelling incorporating dune failures in Christchurch

5 Executive summary NIWA was commissioned by Environment Canterbury (ECan) to carry out a numerical modelling study of inundation in Christchurch caused by a distant source tsunami originating from a magnitude earthquake at the Peru subduction zone. This work builds on tsunami inundation results associated with the same source carried out by Lane et al. (2014), by specifically incorporating river effects on inundation as well as dune failures at targeted locations along the eastern Christchurch dune system; components which were not included in the original models of Lane et al. (2014). The first part of this report models the inundation differences resulting from these potential dune failures. The second part explicitly models tsunami surge up the Avon, Heathcote, Kaiapoi and Waimakariri rivers with pre-tsunami base level flows of water in the rivers. This allows for an assessment of the influence that pre-tsunami water in the rivers has on the inland extent of tsunami surge and pattern of inundation. For this modelling, the tsunami source scenario is the same as that used by Lane et al. (2014), and represents an extreme distant earthquake scenario involving 2,500 year return period wave heights at the Christchurch coast (Power 2013). The largest wave arrivals at coast were assumed to coincide with mean high water springs (MHWS). MHWS was taken to be 1.2 m above the Lyttelton Vertical Datum 1937 (LVD37) based on the height of MHWS-10 recorded at the Sumner sea level gauge (Bell 2011; Lane et al. 2014). Simulations were conducted using present day topography as well as for five dune modification or breach scenarios developed and supplied by ECan for this study: Scenario 1 was not a dune breach but rather a hypothetical dune modification where the dune geometry was altered to reflect a 6m high dune extending over the 2 km stretch from Waimairi to New Brighton (Figure 1-1). The modified dune is lower and more continuous than existing dune profiles along this 2 km stretch (Figure 1-2). Scenario 2 incorporated eight dune breaches at dune areas adjacent to (in order of likelihood from most to least) Spit Tip, South Brighton Surf Club, Torea Lane, Neptune Place, Effingham Street, Waimairi Surf Club, Caspian-Heron Streets, Jellicoe-Halsey Streets (Figure 1-1). Scenario 3 modelled the combined effects of dune modification of scenarios 1 and dune breaches of scenario 2 above. Scenario 4 involved the five most likely dune breaches at locations adjacent to the Spit Tip, South Brighton Surf Club, Torea Lane, Neptune Place, and Effingham Street (Figure 1-1). Scenario 5 considered potential Southshore dunes only, involving breaches adjacent to the Spit Tip, Caspian-Heron Streets, Jellicoe-Halsey Streets, and the South Brighton Surf Club (Figure 1-1). Distant tsunami inundation modelling incorporating dune failures in Christchurch 7

6 The findings suggest that the breaching of dunes, specifically in scenarios 2 to 5, result in more widespread inundation with greater depths in the New Brighton, Southshore, Taylors Mistake, Sumner, Moncks Bay, McCormacks Bay, Ferrymead and Heathcote Valley compared with previously modelled results by Lane et al. (2014). Of the modelled dune breach scenarios, scenarios 2, 3 and 5 produce the most widespread inundation and hazard exposures, implying the worst cases in the context of this study. Including a representative flow in the Avon, Heathcote, Kaiapoi and Waimakariri rivers had significant effects on the modelled inundation. The river effects modelling for the Avon and Heathcote rivers shows similar inundation depths in southern coastal areas compared with the original scenario (e.g., Sumner, Moncks Bay and McCormacks Bay) with more inundation in Southshore and New Brighton. The extent of inundation is more widespread in upstream areas such as Avondale/North Aranui and Burwood along the Avon River, and in places in the Heathcote Valley. These upstream areas do not appear to get inundated if river flow is not considered in the modelling. The effects described above for the Avon and Heathcote rivers would most likely be exacerbated if breaching of the dunes is taken into account. Results for the Waimakariri River in this study show major differences with previously modelled results for these areas by Lane et al. (2014). Notable is the overtopping of dunes at the Pines and the stop-banks along the Waimakariri and Kaiapoi rivers causing significant inundation of the seawards side of Kaiapoi and the land between Kaiapoi and The Pines. Brooklands, Spencerville and inland along Lower Styx Road also suffer significant inundation. This is considerably more inundation than is seen when flow is not included. Major differences between the combined results in this study compared with previously modelled results by Lane et al. (2014) are mainly due to the bathymetry and flow of the Waimakariri River and the more refined grid used in this study. Some difference may be due to differences between the models, especially in their representation of friction and the initialisation of the wave at the edge of the grid. The results for areas specifically modelled in this study are considered more robust and should supersede the interpretations for these areas made by Lane et al. (2014), regardless of uncertainties outlined in section 2.4 of this report. 8 Distant tsunami inundation modelling incorporating dune failures in Christchurch

7 1 Introduction Christchurch lies along the eastern New Zealand South Island coast, and is exposed to distant-source tsunami originating at the subduction zone off the South Peru / North Chile coast (Power 2013; Lane et al. 2014). For this work, NIWA was commissioned by Environment Canterbury (ECan) to build on the work of Lane et al. (2014) in two ways: by modelling first the effect that dune breach failure along the eastern coast from Waimairi Beach to Southshore has on inundation; and second, the effect that including a representative river flow in the Avon, Heathcote, Waimakariri and Kaiapoi rivers has on inundation (i.e., beyond just a static water level). These scenarios were beyond the original scope of the modelling carried out by Lane et al. (2014). The purpose of this current study is to enable a better understanding of the effects that potential dune breaches and inland tsunami surge through natural waterways (i.e., Avon and Heathcote Rivers) could have on exacerbating the inundation and consequent exposure in Christchurch to an extreme distant-source scenario. In turn, this helps to better inform emergency and evacuation planning decisions. The tsunami source scenario used was the same as that used by Lane et al. (2014), and involves tsunami originating from a moment magnitude (Mw) earthquake at the South Peru / North Chile Subduction Zone. This scenario was based on the findings of Power (2013), and represents a 2,500 year return period event; this is considered an extreme scenario. This report presents results for a modelled tsunami generated from the source scenario, with five dune breach scenarios encompassing dune breaches and modifications at selected sites between Waimairi Beach and Southshore (Figure 1-1). Soil erosion is not explicitly modelled in these scenarios but rather the topography is altered to represent dune breaches in likely locations. The modelling assumes the tsunami coincides with mean high water spring tide (MHWS), which was taken to be 1.2 m above Lyttelton Vertical Datum 1937 (LVD37); consistent with the MHWS value used in previous modelling (Bell 2011; Lane et al. 2014). MHWS was set as the baseline water level for the modelling and represents the case where the largest wave arrives in conjunction with high tide for an average spring tide. Simulations were conducted using existing/intact dune topography as well as for five dune modification or breach developed and supplied by ECan. Figure 1-1 shows the locations of the breaches used in the various scenarios with the inset topographies showing those breaches. The five scenarios were as follows: Scenario 1: Not a dune breach scenario but a hypothetical dune modification where the dune geometry is altered to reflect a continuous 6 m high dune with the profile shown in Figure 1-2 extending over the 2 km stretch from Waimairi to New Brighton (Figure 1-1). The modified 6m dune is lower and more continuous than existing dune profiles along this 2 km stretch. Scenario 2: This scenario incorporates eight dune breaches at areas adjacent to (in order of likelihood from most to least); Spit Tip, South Brighton Surf Club, Torea Lane, Neptune Place, Effingham Street, Waimairi Surf Club, Caspian-Heron Streets, Jellicoe- Halsey Streets (Figure 1-1). Scenario 3: This scenario combines scenarios 1 and 2 above, incorporating both the modified 6 m high dune and the eight dune breaches. Distant tsunami inundation modelling incorporating dune failures in Christchurch 9

8 Scenario 4: This involves the five most likely dune breach locations adjacent to the Spit Tip, South Brighton Surf Club, Torea Lane, Neptune Place, and Effingham Street (Figure 1-1). Scenario 5: This scenario considers Southshore dunes only, involving breaches adjacent to the Spit Tip, Caspian-Heron Streets, Jellicoe-Halsey Streets, and the South Brighton Surf Club (Figure 1-1). In the second round of modelling, the dunes remain intact; however, the Avon, Heathcote, Kaiapoi and Waimakariri rivers are explicitly modelled with a representative flow in each river (i.e., an inflow is set at the upstream end of the river rather than the river just filling up with water as far as the baseline level of the modelling dictates). Use of this report The main purpose of this report is to help inform evacuation and emergency management planning. The distant source scenario modelled has a long estimated return period (c. 2,500 years) and represents an extreme scenario. Information provided in this report may also be useful for strategic development and infrastructure planning as it may, when used with other hazard and risk information, highlight areas of higher vulnerability that are potentially unsuitable for future development. The report is not appropriate for setting land use planning rules. Maps of the inundation extents should not be used at scales finer than 1:25,000. The overview maps are intended as a guide only and should not be used for interpreting inundation. Caveat This report is based on state-of-the-art knowledge and modelling capabilities of tsunamis and tsunami inundation at the time of writing. While every effort was made to provide accurate information, there are many uncertainties involved including knowledge of potential tsunami sources, source characteristics, bathymetry and topography (see Section 2.4 of this report for details). In addition, while the hydrodynamic models capture much of the physics involved in tsunami propagation and inundation, they also include some simplifying assumptions, as with all models. The information provided in this report is of a technical nature and should be considered with the above limitations in mind. 10 Distant tsunami inundation modelling incorporating dune failures in Christchurch

9 Figure 1-1: Overview of study site including modelled dune breach locations. Black lines represent the extents of the breaches in each of the inset topographies. Distant tsunami inundation modelling incorporating dune failures in Christchurch 11

10 Figure 1-2: Existing (black) and modified (green) dune profiles along the 2 km stretch from Waimairi to New Brighton. Image supplied by ECan. 12 Distant tsunami inundation modelling incorporating dune failures in Christchurch

11 2 Tsunami inundation modelling 2.1 Source model and initial conditions Power (2013) provides a wave height at the Christchurch coast of m for the 2,500 year return period tsunami at the 84 th percentile uncertainty level 1. The de-aggregation of this wave height identified South Peru / North Chile as a major source of the hazard and that an earthquake of Mw was required to produce that wave height at the Christchurch coast (Lane et al. 2014). For consistency with previous work by Lane et al. (2014), this source and corresponding modelling approach is used to model the tsunami in this study. The overarching scenario was guided by the GNS Science tsunami database and uses source segments from Tang et al. (2010). In Lane et al. (2014) four earthquake rupture scenarios were modelled. All had the same moment magnitude (Mw), but differed in configurations of rupturing segments and corresponding displacements. For this report we model the largest of those four scenarios. 2.2 Inundation modelling and grid The far-field propagation of the tsunami across the Pacific Ocean was modelled using Gerris (Popinet 2003; 2011) in Lane et al. (2014). We did not repeat this modelling but used the same input conditions taken from it for ocean boundary conditions (the height and timing of the approaching tsunami waves) driving the inundation modelling in this report Dune Breach Scenarios The results from the transpacific modelling were used to form the boundary conditions at the edge of the higher resolution inundation grid (197 E or 163 W), to model the inundation using RiCOM (River and Coastal Ocean Model) at the locations of interest (see Figure 1-1). RiCOM is based on the Reynolds-averaged Navier-Stokes (RANS) equations and incompressibility criterion, which have been shown to adequately model tsunami inundation in the case of non-breaking waves (Walters and Casulli 1998; Walters 2005a and 2005b). An initial quiescent sea state, set at a level chosen to represent the tidal height at the arrival of the maximum wave (in this case MHWS), was used in the inundation modelling. Tidal amplitudes in the open ocean are small relative to water depth, with low velocities that are unlikely to significantly influence tsunami wave speed. Thus, the trans-pacific modelling using Gerris was carried out at zero sea level which was adequate to form the boundary conditions to model the tsunami arriving and inundating during MHWS using RiCOM. Refer to Lane et al. (2014) for further details on the modelling approach used. In terms of the model grid used (bathymetry and topography), this study involved the use of updated and more refined digital elevation data than the original study by Lane et al. (2014). A RiCOM grid encompassing updated and more accurate bathymetry in the Avon and Heathcote rivers obtained from NIWA was used as the base grid for the modelled dune breach scenarios. Because the bathymetry of the Avon and Heathcote rivers is included in this modelling, there is water in these 1 In probabilistic tsunami hazard assessment, the 84th percentile uncertainty level refers to the confidence level (or uncertainty) regarding the water level due to the contributions of different fault sources and uncertainty in the parameters of those sources (see Power 2013 for further details). Distant tsunami inundation modelling incorporating dune failures in Christchurch 13

12 rivers up to the high tide extent (1.2m above LVD37 2 ) which does transmit some of the tsunami wave further upstream, however pre-tsunami water in the rivers is not explicitly modelled as it is in the river scenario below. This was the same grid used for modelling three regional tsunami scenarios in Kohout et al. (2015). Differences in digital elevation based on updated datasets between this study and Lane et al. (2014) are shown in Figure 2-1. The main differences are that the bathymetry for the Avon and Heathcote Rivers obtained from NIWA (Measures and Bind 2013) is included in the more recent grid. There are also differences in the position of the estuary outflow and the height of the sewerage ponds (as the ponds are not inundated in any of the scenarios, this does not affect the results). There were also differences in the topography on the Port Hills, mostly due to slight shifts in the placement of the grid. These topography differences are large in steep regions, but as they are above the level of inundation, they have no impact on the modelling. Tsunami inundation was simulated on the original grid (with the updated elevations as described above) as well as on a further five grids developed according to the scenarios described below. A patch grid was created for each of the dune modifications used in the five scenarios (see insets in Figure 1-1) with specified bathymetry through the dune. Topography was also truncated to the level of the dune breach behind the dune as far as the road, although lower topography was not increased in height. These patches were then used to re-depth the inundation grid according to the scenarios described below. Scenario 1: A dune modification of a hypothetical continuous 6 m high dune with profile shown in Figure 1-2 extending over the 2 km stretch from Waimairi to New Brighton (Figure 2-2). The modified 6 m dune is lower and more continuous than existing dune profiles along this 2 km stretch (Figure 1-2). Scenario 2: Eight dune breaches at dune areas adjacent to (in order of likelihood from most to least); Spit Tip, South Brighton Surf Club, Torea Lane, Neptune Place, Effingham Street, Waimairi Surf Club, Caspian-Heron Streets, Jellicoe-Halsey Streets (Figure 2-3). Scenario 3: The combined effects of scenarios 1 and 2 above (Figure 2-4). Scenario 4: The five most likely dune breach locations adjacent to the Spit Tip, South Brighton Surf Club, Torea Lane, Neptune Place, and Effingham Street (Figure 2-5). Scenario 5: Southshore dune breaches only, involving breaches adjacent to the Spit Tip, Caspian-Heron Streets, Jellicoe-Halsey Streets, and the South Brighton Surf Club (Figure 2-6). 2 LVD37 is the Lyttelton Vertical Datum. This is a local fixed survey datum which was created in 1937 and derived from mean sea level at Lyttelton from 9 years of measurements prior to Due to sea level rise over the intervening period, the present mean sea level is approximately 0.17m higher than the Lyttelton Vertical Datum 14 Distant tsunami inundation modelling incorporating dune failures in Christchurch

13 Figure 2-1: Difference in topography based on updated datasets between this study and Lane et al. (2014). Zero and minus five metre contours with respect to LVD37 are given to orient the picture. Red represents areas where the more recent grid is lower than the previous grid and blue represents areas where it is higher. Figure 2-2: Dune breach scenario 1. Existing topography is shown in black and white, insets show scenario topography. The modified continuous 6 m dune profile which is lower than existing dune profiles along this 2 km stretch is indicated by the red arrow. Distant tsunami inundation modelling incorporating dune failures in Christchurch 15

14 Figure 2-3: Dune breach scenario 2. Existing topography is shown in black and white, insets show scenario topography, breach locations are indicated by red arrows. Figure 2-4: Dune breach scenario 3. Existing topography is shown in black and white, insets show scenario topography, breach locations (as in Figure 2-3) are indicated by red arrows and the lower, more continuous design dune as seen in Figure 2-2 is also incorporated in this scenario. 16 Distant tsunami inundation modelling incorporating dune failures in Christchurch

15 Figure 2-5: Dune breach scenario 4. Existing topography is shown in black and white, insets show scenario topography. Breach locations are indicated by red arrows. Figure 2-6: Dune breach scenario 5. Existing topography is shown in black and white, insets show scenario topography. Breach locations are indicated by red arrows. Distant tsunami inundation modelling incorporating dune failures in Christchurch 17

16 Development of these scenarios were based on the following assumptions (Todd, Cope and Gadsby 2016, pers. comm.): The eight breach locations were chosen based on the length of foredune which did not exceed 5.5 m anywhere across the dune profile or the narrowest of the dune that did exceed 5.5 m (elevation below 5.5 m is shown as red on the elevation plots for each scenario in Figure 1-1). Secondary consideration was given to the dune/land elevations around and landward of the potential breach sites. The beach walls at North Brighton and New Brighton were not considered based on the assumption that the previous inundation modelling by Lane et al. (2014) did not explicitly model these walls. The breach cross-section for the modelling was estimated as a constant slope from the base elevation at the landward end of the dune to the 2 m foreshore contour. It was assumed that this profile was representative of the whole breach length and that all sand above this breach profile gets removed by the tsunami. No account was made in the profile regarding the redistribution of this material either landward or seaward due to the tsunami. The scour of the mouth of the estuary due to high tsunami velocities, followed the same assumptions from Risk & Reality (Christchurch Engineering Lifelines Group 1997) that the width and depth of the entrance channel would increase by 20 per cent each, with the width increase primarily due to erosion of the Spit (as hard rock on the Clifton side would likely limit erosion on that bank). Based on this assumption, these entrance increases would result in a 44 per cent increase in entrance cross-sectional area. Thus, for the Spit Tip breach in particular, the tsunami is assumed to have widened the estuary channel and breached the dunes to the level indicated in the profile south of the black line drawn on the Digital Elevation Model (DEM) plot (see Figure 1-1). The results presented in this study show the influence these dune breach scenarios have on tsunami inundation Rivers scenarios Inundation modelling capturing the effect of (pre-tsunami) water in rivers was beyond the original scope of the modelling carried out by Lane et al. (2014). For this part of the study however, river flow is explicitly modelled in the Avon, Heathcote, Kaiapoi and Waimakariri rivers in order to provide a more realistic scenario of the potential tsunami surge effect through natural waterways. The Basilisk model (Popinet 2015), a successor to Gerris (Popinet 2011, 2012), was used to model the inundation together with the river flow. These models have been used for both tsunamis (Popinet 2011, 2012) and flood inundation (Bind et al 2014, Smart et al 2017, Smart 2017) which is why Basilisk, the more recent of the two, was chosen for this modelling. A feature of Basilisk and Gerris is that these models use an adaptive grid where the spatial resolution varies throughout the duration of the simulation, with the grid resolution automatically increasing in areas of interest (as defined by Popinet 2012 and 2014), while lower resolution is maintained elsewhere, improving computational efficiency. The model grid was confined to Christchurch and Kaiapoi, extending from Taylors Mistake in the south to The Pines in the north. The model grid was rotated by 13.8 clockwise so that the edge of the grid was approximately parallel with the shoreline. The model was forced with the incoming wave height at its outer edge taken from the previous modelling. The grid is square in shape with the length of a side of the grid being 27 km. The finest resolution of the grid is 13.2 m. 18 Distant tsunami inundation modelling incorporating dune failures in Christchurch

17 The same LiDAR data obtained from ECan as used for the previous modelling was used for the land area. For the Avon and Heathcote rivers we were able to access bathymetry developed postearthquake for ECan (Measures and Bind 2013). This bathymetry was also incorporated into the grid for the dune breach scenario and is responsible for much of the difference between the original as modelled in this project versus the original as modelled by Lane et al. (2014). ECan provided bathymetry for the lower reaches of the Waimakariri River (the lowest 3 km) but this was not enough for the purposes of this modelling. We were also able to obtain cross-sections that were used to reconstruct river bathymetry further upstream. Cross-sections were also used to develop bathymetry for the Kaiapoi River. Details of this reconstruction technique are provided in Appendix A. Stop-bank height levels were provided for the Waimakariri, Kaiapoi and Avon stop-banks and the Sumner seawall for previous projects (Lane and Arnold 2013; Lane et al. 2014; Kohout et al. 2015; Williams et al. 2015). The flood gates where the Cam River joins the Kaiapoi River are modelled as open. In order to ensure that these features were resolved within Basilisk s adaptive refinement, the cells containing these features were refined to the highest level while other regions were allowed to adaptively refine according to details of the flow and whether they were inundated or not. Figure 2-7 shows the stop-banks specifically modelled in this way and the bathymetry from 0 to 10 m above LVD37. The effect of the grid can be seen in that dry regions where inundation didn t reach have big cells (visible as large square blocks of colour) while areas where the inundation reached (or close by) have much more refined bathymetry indicative of the fine underlying grid. The bathymetry at these stop-banks was initialised according to the stop-bank specifications. The four rivers were initialised with flows of 1.0, 1.7, 2.0 and 60 m 3 /s for the Heathcote, Avon (Cameron 1992; Orchard and Measures 2016), Kaiapoi and Waimakariri rivers respectively. The black squares in Figure 2-7 indicate where the flow entered the rivers. The Waimakariri River was originally initialised with a flow of 120 m 3 /s (Heslop 2011); however, with the bathymetry we had, the river burst its banks and flooded the surrounding area at that level of flow. This is most likely due to discrepancies in the bathymetry given the limited data that we had. Because the intent of this exercise was to model the effect of presence of flow in rivers on inundation by tsunami, and because we did not have better bathymetry data for the Waimakariri, we reduced the river flow until it stayed within the banks. Kaiapoi mean flow is reported as 3 m 3 /s in Hudson (2011) but this is below the confluence with the Cam River so it was reduced to 2 m 3 /s. Likewise the Heathcote was originally modelled with a flow of 1.2 m 3 /s based on NIWA (2016) but this was reduced to 1.0 m 3 /s. Orchard and Measures (2016) also report 0.8 m 3 /s as a median flow for the Heathcote. The model was allowed to run for 1 day of model time to allow the rivers to fill up with water and reach a steady state. During this time the outer boundary was also gradually raised from LVD37 up to 1.2 m above LVD37. This gave us a baseline height of 1.2 m above LVD37 (i.e., MHWS) without resulting in ponding of low lying regions of Christchurch that are below 1.2 m above LVD37 but are not connected to the sea. After 24 hours simulation time, a steady state river flow and sea level was reached and the tsunami forcing was initiated. Distant tsunami inundation modelling incorporating dune failures in Christchurch 19

18 Figure 2-7: Stop-banks modelled in this scenario. The black lines show where stop-banks for the Waimakariri, Kaiapoi and Avon rivers are specified as well as the Sumner seawall. Black circles indicate where water was added to create flow down the rivers. The background shows the topography. Dark blue indicated topography below LVD37, dark red is topography greater than 10 m above LVD37. The adaptive grid can be seen in the different resolution of the topography (in some places where inundation does not reach the grid is very blocky but it is highly refined in other places). This grid has been rotated 13.8 clockwise from north so that the beach is approximately parallel with the edge of the grid. 2.3 Model outputs The outputs listed below were produced using the numerical models: A. Detailed spatial data depicting the inundation depth and extent at the locations of interest for the largest wave arriving at MHWS for each dune breach and river effect scenario. B. Detailed spatial data depicting the maximum flow speeds offshore and over land in the locations of interest for the largest wave arriving at MHWS for each dune breach and river effect scenario. These outputs describe the propagation and magnitude of the tsunami arriving at the Christchurch coastline and the inundation at the locations of interest. The maps are presented and discussed in this report. Detailed spatial data (in digital ArcGIS format) has been provided to ECan (see Appendix A). 20 Distant tsunami inundation modelling incorporating dune failures in Christchurch

19 2.4 Uncertainties Inherent uncertainties associated with the tsunami modelling stem from the bathymetric resolution of the grid used, the fault rupture scenarios used, the numerical equations and the solver used for the modelling. These uncertainties are described in more detail below. The quality of the topographic data and the bathymetric data in inshore waters strongly influences the simulation of inundation. For this modelling, we used available DEM data for the land topography and near-shore bathymetry in the vicinity of Christchurch, but in order to capture some of the more complex small scale processes in harbours and embayed areas, higher resolution data is required. Known topographic uncertainties associated with the DEM used for this study include the bathymetry of the Waimakariri River. Also, in order to resolve the Waimakariri and Avon stop-banks and Sumner seawall, we refined cells in these regions to the highest resolution and then set their elevation to the specified stop-bank and seawall heights. Given that the finest resolution is 13.2 m, this means that the stop-banks and seawalls were modelled as having at least that width. In the case of the Sumner seawall, this is wider than the actual sea wall. Other uncertainties in the modelling study include the gridded representation of a continuous coastline (grid-stepping), which can deform the shape of bays and estuaries, and the effects of building and land features on form drag. The latter could substantially modify the onshore propagation of tsunamis. Improving the drag representation remains a goal of current research. Eradication of the other errors is constrained by limitations of data quality and the practicalities of grid resolution; models always represent an approximation of reality. Basilisk and RiCOM use different drag representations which affects the modelled inundation extent. The two models used in this work are based on different underlying grids. The underlying grid used in the RiCOM model is triangular. Most of the triangles are chosen to be as close as possible to equilateral, but this allows the edge of the triangles to be whatever length is required down to a maximum resolution of around 15m. Basilisk, however, is based on squares. It allows variable resolution by subdividing squares into four smaller squares until the desired resolution is achieved. In this modelling the maximum resolution of the Basilisk grid is 13.2 m. These differences in grid can affect the results. RiCOM is able to better represent coastlines and features like stop-banks where the grid can be specifically chosen to line up with the features. Basilisk can represent square features that line up with its grid directions better. In this modelling, the Basilisk grid was rotated so that the shoreline and the roads would be more in line with the grid. Model uncertainty can be quantified by running multiple simulations with small variations in key parameters, an approach known as ensemble prediction or sensitivity analysis. Such an approach provides an envelope of predicted solutions, rather than single worst-case or scenario-type predictions, on which to base emergency response procedures. However, running many simulations increases the computational and research costs, and, in any event, model forecasts can never be certain because our knowledge of all the geophysical processes involved in tsunami generation, propagation and inundation remains incomplete. Differences between models can be caused by a range of factors including the underlying structure of the grid, the resolution of the grid, the digital elevation model (DEM), the representation of friction and the treatment of boundary conditions. Especially in locations where there are dunes or other barriers to the flow the results can be quite sensitive to these factors as there will be very different answers depending on whether or not the dunes are overtopped. Differences in the DEM (see Figure 2-1) between the earlier modelling (Figures 4.15 and 4.17 in Lane et al 2014) and the Distant tsunami inundation modelling incorporating dune failures in Christchurch 21

20 current RiCOM modelling (Figure 3-1) highlight this sensitivity. The Basilisk modelling including river flow in Section 4 also shows difference in the inundation around the dunes in New Brighton and Southshore (see Figure 4-1) due to the sensitivity of this area to differences in the model set-up. Quantitative calibration of the tsunami inundation model against real measurements is difficult due to the uncertain nature of tsunami impact data from New Zealand and the consequent difficulty in identifying events from the past. Nevertheless, the RiCOM, Gerris and Basilisk models have been continuously validated against standard analytical test cases (e.g., Walters and Casulli 1998; Walters 2005a; Walters 2005b; Popinet 2003, 2011, 2012, 2015). 22 Distant tsunami inundation modelling incorporating dune failures in Christchurch

21 3 Model results dune failure scenarios 3.1 Maximum inundation and speed based on existing dune profiles The maximum predicted tsunami inundation without considering dune breaches and river effects for the Christchurch coast is shown in Figure 3-1. These results are similar to those modelled by Lane et al. (2014), but incorporate more accurate and updated elevation data. The values shown include the MHWS offset. Inundation depths significantly greater than 2.5 m are experienced in Sumner, Moncks Bay, along the New Brighton coastline including the Southshore Spit, and in eastern areas adjacent to the Avon estuary. Inundation depths up to 2.5 m are experienced in localised areas of New Brighton Township, Southshore, McCormacks Bay, and in areas adjacent to the Avon River ~1km upstream of the estuary. Figure 3-2 shows the maximum predicted tsunami flow speeds without considering dune breaches and river effects for the Christchurch coast. Flow speeds exceeding 5 m/s are experienced at the entrance of the Southshore Spit (between McCormacks Bay and the Spit), along the Sumner beach and in localised areas in McCormacks Bay and Southshore. Speeds up to 5 m/s are experienced on land in localised areas of McCormacks Bay, Sumner and Southshore. The subtle differences apparent in the distribution of both maximum predicted inundation and speed between this study and that originally modelled by Lane et al. (2014) are due to the more refined bathymetry and topography data used in this study; with the results of modelling presented in this report more accurate than previous work. 3.2 Scenario-specific maximum inundation and speed Scenario 1 Maximum tsunami inundation depth and maximum flow speeds predicted for this scenario are provided in Figure 3-3 and Figure 3-4. The pattern of maximum predicted inundation for this scenario does not differ significantly from the expected maximum inundation without dune modification as shown in Figure 3-1. The overall extent of inundation landward of the New Brighton township appears slightly less in this scenario, although more over-toping occurs immediately north of the township. This suggests that modifying the dune to the 6 m design profile along the 2 km stretch from Waimairi to New Brighton increases the extent of inundation in this localised area, but also influences the overall pattern of inundation resulting in a slightly less inundation extent landward of the township. Maximum predicted flow speeds for this scenario show a similar pattern to maximum inundation, with the area over-topped between Waimairi and New Brighton showing greater extents of flow speeds up to 1 m/s. The extent of flow speeds up to 1 m/s is also slightly less landward of the township Scenario 2 Maximum tsunami inundation depth and maximum flow speeds predicted for this scenario are provided in Figure 3-5 and Figure 3-6. Significantly greater inundation depths and extents are experienced in this scenario along New Brighton and Southshore, including Moncks Bay and McCormacks Bay as well as areas adjacent to the Heathcote River and Charlesworth Reserve. Inundation depths significantly greater than 2.5 m are also experienced in dune-breached areas in Southshore and New Brighton. Erosion of the Spit-tip and dune-breached areas in Southshore causes inundation depths exceeding 2.5 m along the western coast of the Spit. Erosion of the Spit-tip also Distant tsunami inundation modelling incorporating dune failures in Christchurch 23

22 exacerbates inundation extents and depths significantly greater than 2.5 m in Moncks Bay, McCormacks Bay and in the Heathcote Valley. The extent of maximum predicted flow speeds up to 1 m/s is also greater along the Avon and Heathcote rivers by up to 1 km inland in some cases when compared with the previous scenarios. Significantly greater flow speeds are experienced in Southshore for this scenario compared with the previous scenarios, and in some localised breached Southshore locations flow speeds up to, and exceeding, 5 m/s are experienced Scenario 3 Maximum tsunami inundation depth and maximum flow speeds predicted for this scenario are provided in Figure 3-7 and Figure 3-8. There are subtle changes in the extent of maximum predicted inundation around the New Brighton and Avon River areas for this scenario compared with the previous scenarios. Notable is the slight increase in inundation extent and depth between the 2 km stretch from Waimairi to New Brighton resulting from over-topping due to erosion of the 6 m design dune in that area. Similarly, subtle changes in the extents of maximum flow speeds are apparent for this scenario compared with previous scenarios, including flow speeds up to 3m/s potentially reaching the road adjacent to the 2 km stretch from Waimairi to New Brighton Scenario 4 Maximum tsunami inundation depth and maximum flow speeds predicted for this scenario are provided in Figure 3-9 and Figure Whilst the extent of inundation and maximum predicted inundation is less for this scenario compared with scenarios 2 and 3, they are greater in extent than inundation expected in scenario 1 as well as the original scenario which does not consider the breaching of dunes. A similar pattern is also observed with maximum predicted flow speeds for this scenario Scenario 5 Maximum tsunami inundation depth and maximum flow speeds predicted for this scenario are provided in Figure 3-11 and Figure The maximum extent of predicted inundation including maximum inundation depth for this scenario are similar to inundation patterns observed in scenarios 2 and 3. The main difference in this scenario is the lower inundation extent in areas north of Waimairi. Further, this scenario as well as scenarios 2 and 3 suggest that breaching of the dunes and Spit-tip in Southshore results in greater surges of water into the estuary, significantly increasing the extent and depth of inundation in Southshore as well as Moncks Bay, McCormacks Bay, Heathcote valley and in areas adjacent to the Avon River. This observation is corroborated by the maximum predicted flow speeds shown in Figure 3-12 for this scenario. 24 Distant tsunami inundation modelling incorporating dune failures in Christchurch

23 Figure 3-1: Maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS (Model tsunami taken from Lane et al. 2014). Inundation depths are only shown for inundated land. Distant tsunami inundation modelling incorporating dune failures in Christchurch 25

24 Figure 3-2: Maximum flow speed for Christchurch assuming the largest wave arrived at MHWS (Model tsunami taken from Lane et al. 2014). 26 Distant tsunami inundation modelling incorporating dune failures in Christchurch

25 Figure 3-3: Modified dune scenario 1 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. Distant tsunami inundation modelling incorporating dune failures in Christchurch 27

26 Figure 3-4: Modified dune scenario 1 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 28 Distant tsunami inundation modelling incorporating dune failures in Christchurch

27 Figure 3-5: Dune breach scenario 2 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. Distant tsunami inundation modelling incorporating dune failures in Christchurch 29

28 Figure 3-6: Dune breach scenario 2 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 30 Distant tsunami inundation modelling incorporating dune failures in Christchurch

29 Figure 3-7: Dune breach scenario 3 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. Distant tsunami inundation modelling incorporating dune failures in Christchurch 31

30 Figure 3-8: Dune breach scenario 3 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 32 Distant tsunami inundation modelling incorporating dune failures in Christchurch

31 Figure 3-9: Dune breach scenario 4 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. Distant tsunami inundation modelling incorporating dune failures in Christchurch 33

32 Figure 3-10: Dune breach scenario 4 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 34 Distant tsunami inundation modelling incorporating dune failures in Christchurch

33 Figure 3-11: Dune breach scenario 5 maximum inundation depth for Christchurch assuming the largest wave arrived at MHWS. Distant tsunami inundation modelling incorporating dune failures in Christchurch 35

34 Figure 3-12: Dune breach scenario 5 maximum flow speed for Christchurch assuming the largest wave arrived at MHWS. 36 Distant tsunami inundation modelling incorporating dune failures in Christchurch

35 4 Model results river scenarios The river scenarios were modelled using Basilisk rather than RiCOM, see Section 2.4 for a discussion about what affects this may have had on the results. Inundation in the Waimakariri and Christchurch areas were modelled on one grid. However, the results are shown separately to allow closer views and also to allow more ready comparison with the previous modelling. Table 4-1 gives the location and heights of the bottom of bridges on the Kaiapoi, Waimakariri, Avon and Heathcote rivers, their heights above LVD37 (where we were able to find them) and the modelled tsunami heights above LVD37 at each location. Most of the bridges are above the height of the tsunami, except for the Bamford Street foot bridge on the Heathcote River. We were not able to obtain heights for all locations, however, and at some locations we were unsure of the datum. Table 4-1: Bridges on Waimakariri, Avon and Heathcote rivers with heights above LVD37 and modelled maximum tsunami height at those locations. Bridge Longitude Latitude Height above LVD37 (m) Kaiapoi River Williams St Bridge Kaiapoi River Davie St Footbridge Kaiapoi River Bridge St Footbridge Kaiapoi River SH1 Kaiapoi River Railway Bridge Waimakariri River Main North Rd Waimakariri River SH1 Avon River Bridge Street Avon River Pages Road Avon River Anzac Drive Avon River Avondale Road Heathcote River Ferry Road Heathcote River Tunnel Road Tsunami height above LVD37 (m) Distant tsunami inundation modelling incorporating dune failures in Christchurch 37

36 Heathcote River Bamford Street Heathcote River SH 74A Heathcote River Connal Street Heathcote River Radley Street Maximum inundation and speed Avon and Heathcote rivers The maximum predicted tsunami inundation depths above ground level which incorporate the Avon and Heathcote rivers effects for Christchurch are shown in Figure 4-1. The values assume that the largest wave arrived at MHWS (1.2 m above LVD37). Close to the coast, the pattern of inundation extent is broadly similar to that in the previous modelling shown in Figure 3-1, however there is more inundation in Southshore and New Brighton where the dunes are overtopped in most places. The differences in dune overtopping in New Brighton are most likely due to the differences in the models used rather than the effect of including the river flows in the modelling. Most of New Brighton and Southshore is inundated with details of individual roads being inundated observable in parts. Further inland there are major differences where the tsunami is transmitted upstream by the water modelled in the Avon and Heathcote rivers. In addition, the tsunami spills out from the Avon River inundating Aranui to over 1m depth in some places. The Avon River it also spills its banks further up, most of the way to Avonside, inundating up into Travis Wetland and Horseshoe Lake. Many of these inundated areas are within the residential red zone designated after the Christchurch February 22 nd 2010 earthquake, but the tsunami goes beyond the red zone in places. The surge from the tsunami can be observed within the river many kilometres upstream, as far as Manchester Street in the city centre, even though the tsunami does not overtop the riverbanks this far upstream. There is also considerable inundation around the Heathcote River. The low lying paddocks southwest of the oxidation ponds are inundated as far as Dyers Road. Ferrymead shops are inundated to Tunnel Road. Inundation also covers Ferrymead Park and in towards Heathcote Valley. The tsunami travels up the both the Woolston Cut and the Heathcote River and overtops the banks in low lying areas there and along Clarendon and Richardson Terraces. Influence of the tsunami extends up the river as far as Beckenham. As in previous modelling, all the low-lying land in Redcliffs, Moncks Bay, Sumner and Taylors Mistake is inundated to over 2.5 m in many places. Much of the extreme inundation occurs from later waves which, although not as big as the earlier waves, ride on the water already in the estuary and up the river which has not been able to drain after the earlier waves. Maximum predicted tsunami flow speeds shown in Figure 4-2 are consistent with the pattern of inundation extent and depth for this scenario. Speeds over 5 m/s are observed in places where the tsunami overtops the dunes in New Brighton and Southshore, the Sumner seawall and at the estuary mouth. Flows exceed 1 m/s in much of the estuary. Further up the Avon and Heathcote rivers the flow speeds are generally below 1 m/s though they reach up to 3 m/s in the constrictions at the Bridge St and Ferry Road bridges. Whilst dune breaches were not explicitly considered in this scenario, the findings imply that breaching of the dunes (especially scenarios 2 to 5 as modelled in Section 3.2) coupled with river effects shown here would result in greater inundation depths over potentially more widespread areas than this scenario suggests. 38 Distant tsunami inundation modelling incorporating dune failures in Christchurch

37 4.2 Maximum inundation and speed Waimakariri River Maximum tsunami inundation depth above ground level and maximum flow speeds predicted for this scenario are provided in Figure 4-3 and Figure 4-4. Depths shown in the river are above the initial water level in the river (i.e., show the increase in water depth). Given that the previous model had ground at the river water level the results should be comparable. The findings as modelled in this scenario show major differences to the pattern of inundation along the Waimakariri River from previous modelling (Lane et al. 2014) with far more inundation overall when flow is included. In this scenario, inundation heights above 2m are experienced as far up the Waimakariri River as State Highway 1 (SH1), resulting in overtopping of the stop-banks up to 3 km inland, and tsunami surge up the river and river banks up to 8km upriver (more than 1 km inland of SH1). Kairaki and The Pines are severely inundated with depths over 2.5 m in many places, and much of the land on the northern shores of the Waimakariri and Kaiapoi rivers is inundated to some extent. The Kaiapoi River stopbanks appear to protect much of Kaiapoi township. The flooding directly to the north of Kaiapoi is mostly via the Cam River, which is modelled with the flood gates open. Inundation extends to the northern edge of the inundation grid in the low-lying fields to the northwest of Kaiapoi, and so could be expect to continue further north. Inundation is also widespread in the Brooklands and Brooklands Lagoon areas, with inundation depths over 2.5 m in localised areas near the estuary and Styx River. Spencerville is severely inundated, as is the land around Lower Styx Rd inland of it. The tsunami travels up the Styx River as far as Marshlands Road (note this is despite the Styx River not being explicitly modelled). Maximum tsunami flow speeds are consistent with the observed pattern of inundation. Speeds exceed 5 m/s at the mouth of the Waimakariri River, and also at many place where the dunes are overtopped. Flow speeds on the flats generally are slower further inland and away from the river. High speeds (up to 3 m/s) are still observed up the river at least as far as SH1. The key differences between the findings in this scenario with previously modelled results by Lane et al. (2014) is due to including the river bathymetry and flow, as well as the different grid with higher resolution. As such, the findings here are considered more robust and should supersede previous interpretations. Distant tsunami inundation modelling incorporating dune failures in Christchurch 39

38 Figure 4-1: Maximum inundation depth in the Avon and Heathcote rivers. The modelling assumes that there is flow in the Avon and Heathcote rivers and that the tidal height at the arrival of the maximum wave was MHWS (1.2 m above LVD37). On land areas maximum inundation depth is shown, on the rivers the maximum wave height (height of the tsunami above the river level) is shown. 40 Distant tsunami inundation modelling incorporating dune failures in Christchurch

39 Figure 4-2: Maximum tsunami flow speeds in the Avon and Heathcote rivers. The modelling assumes that there is flow in the Avon and Heathcote rivers and that the tidal height at the arrival of the maximum wave was MHWS (1.2 m above LVD37). Distant tsunami inundation modelling incorporating dune failures in Christchurch 41

40 Figure 4-3: Maximum inundation depth along the Waimakariri River. The modelling assumes that there is flow in the Kaiapoi and Waimakariri rivers, that the Cam River flood gates are open and that the tidal height at the arrival of the maximum wave was MHWS (1.2 m above LVD37). On land areas, maximum inundation depth is shown, on the river the maximum wave height (height of the tsunami above the river level) is shown. 42 Distant tsunami inundation modelling incorporating dune failures in Christchurch