An assessment of current and future vulnerability to coastal inundation due to sea-level extremes in Victoria, southeast Australia

Authors


Abstract

Current climate 1-in-100-year storm tide heights along the coast of Victoria, southeast Australia were estimated by combining probabilities of storm surge and tide heights determined from hydrodynamic modelling. For this return period, levels lie between 1 and 2 m above mean sea level along much of the coastline. Future climate 1-in-100-year storm tide heights were estimated by adding high-end estimates of future sea-level rise from recent literature. The effect of climate change through consistent wind-speed increases was also examined and it was found that, for the late 21st Century, the contribution of wind-speed increase to the increases in extreme storm surge heights is considerably smaller, by a factor of more than 2, than the contribution of sea-level rise.

A computationally inexpensive approach to assessing current and future vulnerability to coastal inundation due to sea-level extremes is then demonstrated for the Victorian coast. A simple inundation algorithm was used with high-resolution terrestrial elevation data from a Light Detection and Ranging (LiDAR) survey of the Victorian coast to evaluate the potential vulnerability of nine coastal regions to inundation by current and future climate 1-in-100-year storm tides. The response of different regions varied from exhibiting proportional increases in inundation to sea-level rise to nonlinear responses, where the exceedance of critical sea-level thresholds led to large stepwise increases in land area or number of land parcels affected by inundation. These responses were a function of both coastal topography and the spatial density of land parcels. The low computational cost of the methodology permits different time horizons and uncertainties in future climate change to be considered using a scenario-based approach and is therefore useful in assessing options for adaptation to climate change. Copyright © 2011 Royal Meteorological Society

1. Introduction

Sea levels have been rising at an increasing rate since the early 1800s. Over the period 1993–2006 mean sea-level trends based on both reconstructed sea level (Church and White, 2006) and satellite altimetry have been 3.3 ± 0.4 mm/year, which is consistent with the upper bound of the Intergovernmental Panel on Climate Change's (IPCC) sea-level projections (Rahmstorf et al., 2007) and associated with a globally averaged sea-level increase of 0.82 m by 2100 relative to 1990 (Hunter, 2010). The IPCC has concluded that there has been a likely increasing trend in extreme high-water levels associated with storm surges across the globe during the 20th Century (Bindoff et al., 2007). These observed and projected changes in sea levels pose threats for coastal settlements that have developed during the period of relatively stable sea levels of the previous two millennia (Jansen et al., 2007).

In recent years, a number of programs and research efforts have been initiated to assess coastal vulnerability across the globe, such as the Synthesis and Upscaling of sea-level Rise Vulnerability Assessment Studies (SURVAS) discussed in Nichols and de la Vega-Leinert (2008) and the US National Assessment of Coastal Vulnerability to Sea Level Rise described in Thieler (2000). The study described in the present paper has been undertaken as part of the Future Coasts Program of the State Government of Victoria, Australia (see www.climatechange.vic.gov.au/futurecoasts) to quantify the risks associated with sea-level rise along the Victorian coast. Results from this work were also used in a recently completed first pass National Coastal Vulnerability Assessment of Australia's coast (Department of Climate Change, 2009) to identify the areas potentially vulnerable to flooding during extreme sea-level events under future sea-level rise.

Victoria has approximately 2000 km of coastline, which is highly valued for its aesthetic, ecological, recreational and economic assets. Since the late 1990s a rapid increase in the rate of internal migration from large cities to the coast has led to an increase in developmental pressure along the coast, which poses challenges for sustainable management (Harvey and Woodroffe, 2008). Approximately 85% of Victoria's population lives within 50 km of the coast and 25% within 3 km of the coast, so climate change has the potential to create significant additional challenges to coastal management through rising sea levels and possible changes to severe storms. Such issues have prompted the present study, which is concerned with the evaluation of storm tides under present and future climate conditions and their possible contribution to coastal inundation.

Studies that address the topic of storm surges and storm tides in the context of future climate change can be broadly categorized as those that aim to investigate how changes to weather conditions in the future will impact on extreme sea-level events (e.g. Debenard and Roed, 2008; Wang et al., 2008; Sterl et al., 2009) and those that are undertaken for impact assessments (e.g. Bernier et al., 2007; McInnes et al., 2009a), in which a scenario-based approach is applied to current climate risk to explore future climate conditions for the purposes of impact assessment and adaptation planning (e.g. Preston et al., 2007). This present study falls into the latter category. It develops storm tide return period maps for the coastline of Victoria in southeastern Australia to provide the basis upon which to explore the impact on inundation of future climate change and sea-level rise. The study builds on that of McInnes et al. (2009a), who used hydrodynamic modelling and extreme value statistical analysis to evaluate storm surge probabilities, which were then combined with tide data at discrete locations along the coast to estimate ‘storm tide’ return periods. Here, a hydrodynamic model is used to also evaluate the tide height probabilities so that storm tide surfaces can be evaluated. These are then used in combination with a Digital Elevation Model (DEM) of Victoria's coast, acquired as part of the Future Coasts Program, to explore the impact of a range of plausible climate change scenarios on coastal inundation under future climate conditions. The focus of this study is the effects of climate change that are likely to result in the greatest changes in sea-level extremes in the future and, hence, only the impact of future sea-level rise and wind-speed changes on storm surge heights are considered. Changes in more minor contributors to sea-level extremes, such as changes in wave behaviour and atmospheric pressure, are neglected.

The remainder of this paper is structured as follows. The following section briefly describes the background to the study and the methodology. Extreme sea-level and inundation results are presented in Section 3. Section 4 discusses the implications of these results and avenues for future research and Section 5 presents conclusions.

2. Background and methodology

There are three stages in the methodology used by this study. Firstly, current climate extreme sea levels are estimated. Then, relevant climate changes are incorporated to estimate future extreme sea levels for a range of plausible climate change scenarios. Finally, the current and future extreme sea levels are passed to a simple inundation algorithm to identify land areas and land parcels that are vulnerable to inundation under current climate conditions and the climate change scenarios.

2.1. Current climate extreme sea levels

For the purposes of planning and engineering design, extreme sea-level events are commonly expressed in terms of return periods. A return period is defined as the average duration of time between events that exceed a particular level. In this study, maps of storm tide return periods are developed from separate distributions of surge and tide height using the revised Joint Probability Method of Tawn and Vassie (1989), hereafter referred to as the JPM, following the procedure illustrated in Figure 1. The development of storm surge probabilities (upper half of Figure 1) was described in detail by McInnes et al. (2009a) and so will be described only briefly here.

Figure 1.

Schematic diagram illustrating modelling approach used in this study

2.1.1. Storm surge height probabilities

The main synoptic weather systems responsible for storm surges along the coastline of Victoria are west-to-east travelling cold fronts, which occur year round but tend to be more frequent and intense in the winter months (McInnes and Hubbert, 2003). These systems impact an extensive stretch of coastline and are therefore well captured by the available tide gauge network. McInnes et al. (2009a) exploited this feature of surges along this coastline by using a selection of tide gauge records to identify significant storm surge events over the period 1966–2003, during which tide gauge records are largely complete. The hydrodynamic model GCOM2D (Hubbert and McInnes, 1999) was then used to simulate each of the storm surge events in the population using atmospheric 10-m wind and sea-level pressure forcing from the NCEP reanalyses (Kalnay et al., 1996). The hierarchy of model grids used is illustrated in Figure 2. The outermost grid has a 5-km resolution and covers Tasmania and about 2000 km of coastline of mainland Australia to the west to ensure that the frontal systems that propagate along the coastline are well captured. The currents and sea levels modelled on this grid were stored at hourly intervals and applied as boundary conditions to simulations on the 1-km grid covering Bass Strait. Similarly, simulations on four smaller grids, covering Port Phillip Bay, Westernport Bay and Corner Inlet at 100-m resolution and the northern part of the Gippsland Lakes at 50 m resolution, were nested within the 1-km grid simulations. Extreme value statistical analysis of the simulated peak storm surge heights was then undertaken to determine surge height probabilities. The r-largest Generalised Extreme Value (GEV) approach (Coles, 2001) where r = 2 (i.e. the GEV distribution was fitted to the top two modelled events per year) was used to estimate event probabilities.

Figure 2.

The domain of the 1-km-resolution Bass Strait storm surge model grid. Land area is shaded in grey. Bathymetric contours are shown as thin lines and annotated with depths in metres. The boundaries of the nested high-resolution storm surge grids are shown by dotted rectangles. The solid rectangles indicate the inundation analysis regions; 1 = Portland; 2 = Port Fairy; 3 = Ocean Grove; 4 = Queenscliff; 5 = Werribee; 6 = Williamstown; 7 = Aspendale; 8 = Tooradin; 9 = Sea Spray. The dots mark locations of tide gauges; P Portland; PL = Point Lonsdale; SP = Stony Point; PW = Port Welshpool; RI = Rabbit Island; LE = Lakes Entrance; PH = Point Hicks. Inset: The region covered by the 5-km-resolution storm surge model grid. Land area is shaded in grey. The 1-km-resolution storm surge grid is marked with a solid rectangle, and the 1-km-resolution tide model with a dashed rectangle

2.1.2. Tide height probabilities

McInnes et al. (2009a) developed tide height frequencies based on the known tide constituents analysed at tide gauge locations along the Victorian coast. In this study, for the purposes of developing spatially complete maps of tide height, a hydrodynamic model was used to generate tidal information on a grid comparable to that used for the storm surge modelling. GCOM3D, the three-dimensional, z-coordinate counterpart to GCOM2D (Hubbert, 1993) was used because it more accurately represents tidal currents through the depth of the water column, and hence the phase and amplitude of the tides, than GCOM2D. A 32-layer barotropic configuration of the model was used that neglects temperature and salinity variations and solves only the horizontal momentum and continuity equations using the solution procedure for each model layer as described for GCOM2D in Hubbert and McInnes (1999). The region covered by GCOM3D is shown by the dashed line in the inset of Figure 2. The inclusion of Tasmania on the tidal grid is important because tidal currents enter Bass Strait not only from the east but also propagate around Tasmania and enter from the west (McInnes and Hubbert, 2003).

GCOM3D simulations of tides were carried out over six-month intervals using an iterative approach to minimize root mean square (rms) errors in the amplitudes of the tide constituents at tide gauge locations within Bass Strait. For the initial six-month simulation, the tidal amplitudes and phases of the eight major tidal constituents in Bass Strait (M2, N2, S2, K2, O1, K1, P1, Q1) were obtained from a global tidal model (Le Provost et al., 1995). A tidal analysis was performed on the modelled time series of water levels to obtain amplitudes and phases of tide constituents at each grid point within the model's computational domain. At the locations of tide gauges that are situated on the open Bass Strait coast (i.e. unaffected by more complex tidal flows that are likely within embayments and estuaries, etc.), rms errors were calculated between the model simulated tides and those obtained from the tidal analysis of the measured sea levels. On the basis of these errors, several additional simulations were carried out in which minor manual adjustments were made to the phases and amplitudes along both the eastern and western boundary according to the proximity of the gauges to these boundaries. The six-month simulation and tidal analysis procedure was repeated six times yielding a reduction of approximately 10% in the average magnitude of the rms difference between tide constituents obtained from modelled and measured tides. A sample of the final rms errors, which are a function of an amplitude error and a phase error, at four tide gauges for the leading two diurnal and two semi-diurnal constituents, M2, S2, O1 and K1, is given in Table I

Table I. The phase and amplitude errors of four tide constituents evaluated from a hydrodynamic model simulation of tides at four locations along the Victorian coast
LocationErrorM2S2O1K1
PortlandAmplitude (m)0.1290.1390.1310.181
 (38.4°S, 141.6°E)Amplitude error (m)0.011− 0.009− 0.011− 0.031
 Phase error (deg)7629
LorneAmplitude (m)0.6150.1930.1450.209
 (38.5°S, 144.0°E)Amplitude error (m)− 0.0050.027− 0.015− 0.029
 Phase error (deg)− 8− 5− 31
Rabbit IslandAmplitude (m)0.6970.1690.1580.238
 (38.9°S, 146.5°E)Amplitude error (m)− 0.017− 0.039− 0.008− 0.028
 Phase error (deg)− 14− 14− 13
Point HicksAmplitude (m)0.3830.1010.0970.119
 (37.8°S, 149.3°E)Amplitude error (m)0.027− 0.0410.0330.041
 Phase error (deg)1148− 2

The tide height frequency distributions required for the JPM were then developed for each grid point of the storm surge model by running a tide prediction model (Foreman, 1977) using the derived tidal constituents and binning the resulting tide heights to develop a frequency histogram. A comparison of a frequency histogram developed using constituents derived from tide gauge data with those developed from the hydrodynamic modelling approach is presented in Figure 3 for Portland, Stony Point, Rabbit Island (at the entrance to Corner Inlet) and Point Hicks.

Figure 3.

Comparison of frequency histograms derived using tide constituents from selected tide gauges along the Victorian coast with those estimated from hydrodynamic modelling of tides

2.1.3. Joint probability analysis

The JPM (Tawn and Vassie, 1989), a commonly used approach for estimating probability distributions for storm tide heights, assumes independence between the probability distributions for tide and surge heights. In this study, the combination of these two probability distributions is achieved by randomly sampling a population of tide and surge values and summing them to develop a population of storm tide heights. Return periods for storm tide heights were estimated by ranking the sampled storm tides from largest to smallest and assigning the return period to the event height according to R = N/r, where R is the return level, N is the number of random samples and r is the rank of the event. To estimate the uncertainty in the storm tide estimates, 200 sets of 1000 storm tides were sampled, and each set of 1000 storm tides was ranked from largest to smallest. From the 200 values associated with each rank (i.e. highest, second highest and so on), the mean was taken to represent the best estimate storm tide height and the standard deviation was taken to represent the uncertainty associated with the random sampling.

The assumption that tide and surge heights are independent of each other is not strictly valid (e.g. Bernier and Thompson, 2007; Horsburgh and Wilson, 2007) as surge heights are related to water depth, with surges being larger in shallow water than in deeper water. Hence surge heights are reduced during high tides. To assess the importance of interactions between tides and surge heights along the Victorian coast, both tidal and meteorological forcing was applied to a month-long GCOM2D simulation in which several significant storm surges occurred. Two additional simulations were performed with tidal and meteorological forcing applied separately to yield independent tide and surge heights. A comparison of sea levels from the simulation with both tidal and meteorological forcing with sea levels derived by summing the independent tide and surge heights revealed that the independence assumption leads to an overestimation of storm tide heights. The month-long simulations reveal a worst-case overestimate of approximately 15% for a large surge coinciding with a spring tide at Stony Point, where the tidal range is greatest along the Victorian coast. For a more moderate tide, and away from the centre of Bass Strait, where the tidal range is smaller, the overestimation was lower. The most affected storm tides make up only a small fraction of the population of storm tides generated to estimate return periods and the estimates of 1-in-100-year levels that are the focus of this study are unlikely to be greatly affected. Indeed, McInnes et al. (2009a) compared 1-in-100-year storm tides derived from the JPM with GEV analysis of total sea levels from tide gauge records and found that they were not significantly different. Therefore we assume that the independence assumption is reasonable in this study.

2.1.4. Uncertainty analysis

In addition to the uncertainty in storm tide return levels due to the random sampling of tide and surge heights, three other significant types of uncertainty were accounted for. These were the uncertainties due to errors in the representations of: (1) storm surges by the hydrodynamic model, which also relate to the representation of pressure and surface winds in the reanalysis data used to force the model, (2) tides by the hydrodynamic model, and (3) the uncertainty associated with the statistical fits to the extreme surge heights. The estimation of uncertainties for a wide range of storm tide return levels incorporating contributions from all four sources of uncertainty is potentially complex. However, our purpose was to estimate approximate uncertainties in large, 1-in-100-year storm tide heights only, and the simplifying assumption that these would most commonly arise through the coincidence of relatively large tide heights with relatively large surge heights was made. Standard errors for each source of uncertainty were estimated on the basis of this assumption. For the simulated storm surge heights, McInnes et al. (2009a) estimate that model biases for several locations across Bass Strait range from − 0.036 to + 0.035. We conservatively assume that 95% of the biases for the modelled storm surges are encapsulated by the ± 0.036 m range, which equates to a standard error of ± 0.018 m. A selection of rms errors in the amplitude of the tidal constituents simulated by the hydrodynamic model are reported in Table I. The standard error in a large tide height, assumed to occur when the four tidal signals are approximately in phase, was estimated by summing the amplitude errors across the tidal constituents. Inspection of storm surge return level curves for several locations on the Victorian coast revealed that standard errors in return levels did not vary greatly with return period for periods greater than ten years. Noting that the width of a 95% confidence interval obtained from the two-largest GEV analysis corresponds to 2 × 1.96 standard errors, the standard error associated with the statistical fit to the extreme surge heights is obtained by dividing the 95% confidence interval for the 100-year storm surge return level by 3.92. The standard error for the joint probability sampling of tide and surge heights was obtained from the standard deviation of the 200 estimates of 100-year storm tide return levels generated by the sampling. Finally, the standard errors for each uncertainty, ΔXk for k = 1, 4, were assumed to be independent and were combined to produce the combined error, ΔX, using ΔX = √(∑(ΔXk)2, k = 1, 4). Confidence limits for the 100-year storm tide return levels were derived from these combined errors by multiplying them by 1.96 and adding and subtracting the result from the estimated return levels.

2.2. Future climate extreme sea levels

Climate change has the potential to change the magnitude and frequency of sea-level extremes through mean sea-level rise and changes in the behaviour of the weather systems that generate storm surges. Along the Victorian coast, these systems are generally eastward-moving low-pressure systems tracking to the south of the Australian mainland with a frontal trough extending northward towards the coast from the low-pressure centre (McInnes and Hubbert, 2003). The winds arising from the low-pressure system and the falling pressure over the coast due to the trough (the so-called inverse barometer effect) both contribute to storm surges. The inverse barometer effect could potentially be enhanced in the future due to an intensification of the troughs or a change in the geographical position of the low-pressure systems. However there is no basis for changes in the intensity of the troughs that act to increase the relative importance of the inverse barometer effect to be considered in this study. Climate models suggest a poleward movement of the storm track in the future (e.g. Lim and Simmonds, 2009; McInnes et al., 2011), which would be associated with a weakening of the troughs passing over Bass Strait. Therefore we assume that the relative contribution of the inverse barometer effect to storm surges will continue to be less than that of winds under future climate change and, consistent with our exclusion of the contribution of waves to extreme sea levels, we neglect changes to the inverse barometer effect contribution. Hence this study considers only the impact of future changes in sea level and wind speed. The selected future sea-level rise and wind speed change scenarios and their incorporation into the methodology for evaluating storm tide return periods are described below.

2.2.1. Sea-level rise

The values of future sea-level rise used by this study are increases in the global-mean sea level. Though uncertain, estimates of regional departures from global sea-level rise for Bass Strait are expected to be small relative to the large uncertainty in global-mean sea-level rise (e.g. Church et al., 2011; their Figure 4). Model-based projections provided by the IPCC's Fourth Assessment Report (IPCC, 2007) give the likely range of values for global-mean sea-level rise between the 1980–1999 and 2090–2099 periods as 0.18–0.59 m. This range includes uncertainties associated with future greenhouse gas emissions, thermal expansion of the oceans and the melting of glaciers and ice sheets. An additional contribution to sea-level rise over this period of 0.2 m, related to the potential for a rapid dynamic ice-sheet response to climate change, is also suggested by the Fourth Assessment Report. Though this contribution is highly uncertain, and larger values cannot be excluded, a figure of 0.79 m can be regarded as a high-end IPCC (2007) estimate of sea-level rise by 2090–2099. Since the intention of the present study is to inform coastal planning decisions, the high-end IPCC (2007) estimate is selected as the basis for the analysis of future sea-level extremes and coastal inundation. Consistent estimates for the years 2030, 2070 and 2100, sourced from Hunter (2010), who estimated values for these and other years throughout the 21st Century by scaling the IPCC (2007) sea-level rise estimates for 2090–2099, were considered.

Figure 4.

The spatial pattern of 1-in-100-year storm tide heights for the Victorian coast under current climate conditions

The Victorian government in its coastal management strategy recommends that a value close to this high-end IPCC (2007) estimate of sea-level rise, 0.8 m by 2100, should be considered for planning purposes (Victorian Coastal Council, 2008). Since the publication of the Fourth Assessment Report, several statistical models of sea-level rise have been developed using observations of 20th Century sea levels. These have resulted in somewhat larger projections of sea-level rise for the 21st Century (e.g. Rahmstorf, 2007; Horton et al., 2008; Grinsted et al., 2009; Vermeer and Rahmstorf, 2009). Rahmstorf (2007) suggests a likely range of values for sea-level rise over the 1990–2100 period of 0.5–1.4 m. A study undertaken for The Netherlands Delta Committee, which assessed post-Fourth Assessment Report publications on the impacts of recent warming trends on ice-sheet dynamics, derives an upper bound of sea-level rise of 1.1 m by 2100 (Katsman et al., 2011). In recognition of the possibility that sea-level rise may exceed IPCC (2007) estimates, the 1.1 m estimate of Katsman et al. (2011) and the 1.4 m estimate of Rahmstorf (2007) are explored (see Table II).

Table II. For the nine analysis regions shown in Figure 2, the area of land (in km2) and (in brackets) the number of land parcels potentially exposed to inundation more frequently than once every 100 years under current climate conditions and plausible conditions for 2030, 2070 and 2100
 Current203020702100
Scenario 12121234
Sea-level rise (m) 0.150.150.470.470.820.821.101.40
Wind speed change (%) 41319
Portland0.81.01.01.31.51.72.12.02.5
 (168)(178)(189)(180)(192)(210)(220)(228)(272)
Port Fairy3.33.74.25.011.312.414.214.015.5
 (162)(168)(189)(314)(565)(680)(781)(812)(968)
Ocean Grove54.256.056.559.660.861.969.969.572.8
 (345)(460)(498)(693)(751)(821)(2559)(2591)(2825)
Queenscliff6.58.99.410.911.867.873.373.375.3
 (222)(623)(819)(1379)(1667)(1937)(2094)(2094)(2254)
Point Wilson9.110.410.914.317.121.024.023.926.9
 (62)(64)(65)(71)(75)(79)(96)(97)(100)
Williamstown6.97.88.310.013.115.523.523.232.0
 (303)(415)(491)(1187)(2240)(3373)(10409)(9531)(15568)
Aspendale2.12.32.44.07.813.322.620.928.8
 (257)(425)(484)(1905)(6995)(9891)(16793)(15520)(19648)
Tooradin42.649.152.565.476.281.796.497.5109.8
 (554)(711)(822)(1161)(1343)(1425)(1587)(1574)(1704)
Sea Spray27.430.531.336.839.640.742.642.644.1
 (289)(548)(579)(1867)(2720)(2984)(3341)(3346)(3763)

To incorporate the contribution of sea-level rise to future storm tide return levels for the scenarios described in Table II, the relevant estimate of sea-level rise was simply added to the storm tide return levels estimated for current climate conditions. This approach makes the assumption that sea-level rise will have little effect on tide heights. To verify that this assumption is reasonable, a Bass Strait storm surge and tide simulation incorporating a sea-level increase of 1.5 m was performed. This indicated that the impact of mean sea-level rise on the simulated temporarily-enhanced sea levels was negligible, other than the addition of the mean sea-level increase. This result is consistent with those of studies of other regions, such as the North Sea (Lowe et al., 2001; Sterl et al., 2009). However, we note that an increase in mean sea level may increase the tidal range within Port Phillip Bay due to reduced frictional attenuation of tidal flows across the shallow bathymetry inside the entrance to the bay, enabling the exchange of a greater volume of water on flood and ebb tides. A hydrodynamic modelling study of Port Phillip Bay by Black et al., (1990) indicated that a 0.5 m sea-level rise, comparable to the high-end IPCC (2007) sea-level rise estimates for 2070 (see Table II), could lead to a 7% increase in the tidal range. For tidal amplitudes typically experienced in Port Phillip Bay, this would result in high tides being about 0.03 m higher and lower tides being about 0.03 m lower than for current mean sea level.

2.2.2. Wind-speed changes

The estimates of future wind-speed changes considered in this study were sourced from a recent report giving climate change projections for Australia (CSIRO and Australian Bureau of Meteorology, 2007). In this report, the method described by Watterson (2008) is used to develop projections of various climate variables for the SRES B1, A1B and A1FI emission scenarios (Nakićenović and Swart, 2000) from the output of a set of climate model simulations that were undertaken for the IPCC's Fourth Assessment Report (IPCC, 2007). To be consistent with the high-end IPCC (2007) estimate of sea-level rise considered in this study, the high (90th percentile) estimates for changes in annual average wind speed for the A1FI emissions scenario were used. Percentage wind speed changes relative to the period 1980–1999 for Bass Strait, taken as the region bounded by 140–150°E and 38–41°S, were obtained by spatially averaging the gridded changes. Wind-speed increases of 4, 13 and 19% were calculated for 2030, 2070 and 2100, respectively. However, it should be noted that, because of the large uncertainties in wind-speed change in this region as represented by the set of Fourth Assessment Report climate models, the Australian projections did not preclude the possibility of future decreases in annual average wind speed in Bass Strait (see also McInnes et al., 2011).

McInnes and Hubbert (2003) noted that storm surge heights in Bass Strait respond approximately linearly to changes in wind speed, with a 1% increase in wind speed corresponding to a 2% increase in storm surge height. This property was used in this study to estimate future storm tide return levels for climate change scenario 2 described in Table II, which incorporates both wind-speed increase and sea-level rise. The effect of wind-speed increase was accounted for by simply increasing the storm surge heights estimated for current climate conditions by a percentage equal to twice the relevant percentage wind-speed increase prior to input to the JPM.

2.2.3. Uncertainty analysis

As regards future storm tide return levels, we take a simple scenario-based approach to representing uncertainties in the effects of mean sea-level rise and wind-speed change on return levels and do not attempt to represent these uncertainties as confidence intervals. We expect the modelling and statistical fitting uncertainties associated with the current climate storm tides to be broadly applicable to storm tides for each individual future climate scenario considered. We note that an extra contribution to the storm surge modelling uncertainty arises due to the incorporation of wind-speed increases. However, this will be small compared to the uncertainty associated with the changes in mean sea level and wind speed spanned by the future climate scenarios and we have chosen to neglect it.

2.3. Inundation analysis

After 1-in-100-year return levels for total sea level (storm surge + tide + mean sea-level rise) were evaluated across the Bass Strait region, a DEM was used to develop GIS layers identifying low-lying coastal land vulnerable to inundation (in this study, we regard land potentially subject to storm tide inundation at least every 100 years, on average, as being vulnerable).

2.3.1. Data and selection of analysis regions

The digital elevation data that was used in the inundation calculations was obtained from an airborne terrestrial Light Detection and Ranging (LiDAR) survey of the Victorian coast from the coastline to approximately the 10-m contour on the landward side of the coast.

Nine regions along the coast (Figure 2) were selected for inundation analysis on the basis that either they contained extensive areas of terrain below a 2 m elevation that will be potentially vulnerable to increasing sea levels in the future, or because the results of this study were likely to complement management activities being undertaken in these regions. Model grids were set up over these regions at 10-km horizontal resolution and the 1-in-100-year storm tide surface was interpolated to each of these grids. Land subdivisions or ‘parcels’ that are vulnerable to inundation were identified by comparing the GIS layers of vulnerable land areas with the Government of Victoria's Vicmap property database (see http://www.land.vic.gov.au/vicmap for more information). In this study, we use land parcel data, which is readily available from the Vicmap database, as an indicator of population and asset density, for which data were not readily available. In general, the number of parcels inundated is related to affected population and assets, with urban areas being characterized by a density of small parcels, corresponding, for example, to individual households and business premises, and rural areas being characterized by large parcels, corresponding, for example, to farms.

2.3.2. Inundation algorithm

A simple non-dynamical but computationally inexpensive approach, the so-called ‘bathtub’ technique, was used to estimate the area of land likely to become inundated. The inundation algorithm used identified land area adjacent to the coast that lies below the level of a storm tide. Dry land points were reclassified as inundated points by carrying out sweeps across the model grid and comparing the height of the coastal sea level in the 1-in-100-year storm tide surface to the adjacent land point. If the sea level exceeds the land height, the land point is inundated to the level of the adjacent sea point.

A further check was carried out to ascertain if there were other adjacent points that could flood the point in question. For example, on a grid extending from i = 1, ni in the east-west direction and j = 1, nj in the north-south direction, if a south to north sweep of the grid (j increasing) yielded a land point at (i, j + 1) at a lower elevation than the sea level at the point (i, j) it was provisionally assigned a water depth of the (i, j) point minus the terrain height. A check was then made on points to the left and right of the targeted point, i.e. the (i1, j + 1) and (i + 1, j + 1) points and if one or both of them were also wet, the water level of the land point at (i, j + 1) was then assigned the average depth of the adjacent wet points minus the terrain height.

3. Results

This section presents the results of the analysis of extreme sea levels along the Victorian coast and describes the potential inundation that may occur from a 1-in-100-year storm tide at the nine low-lying locations shown in Figure 2. Extreme sea levels and potential inundation under current climate conditions and for the climate change scenarios described in Table II are discussed.

3.1. Extreme sea levels

The spatial pattern of 1-in-100-year storm tide heights for the Victorian coast under late 20th Century climate conditions estimated using the JPM is shown in Figure 4. The highest coastal values, in excess of 2 m, occur in and around Western Port Bay, and values of 1.8–2.0 m extend from just west of Port Phillip Bay to Wilson's Promontory, the southernmost point on the Victorian coast. These high values are the result of a large contribution from both storm surges, due to the exposure of this stretch of coastline to the westerly winds most commonly associated with surges, and astronomical tides, which are largest in central Bass Strait. In Port Phillip Bay the low storm tide heights of 1.0–1.2 m are due to the strong attenuation of the tides across the entrance to the bay. It is noteworthy that the 1-in-100-year storm tide heights in the west of the state and in the west of Port Phillip Bay are lower than the Katsman et al. (2011) and Rahmstorf (2007) estimates of sea-level rise for 2100 (Table II). If these estimates prove to be accurate, then by 2100, the mean sea level will lie above the current 1-in-100-year sea level.

For selected locations along the Victorian coast, Figure 5 illustrates the variation in late 20th Century 1-in-100-year storm tide heights and the associated uncertainties. Also shown are late 20th Century 1-in-100-year storm tide heights from McInnes et al. (2009a) and 1-in-100 storm tide heights estimated by the present study for the climate change scenarios presented in Table II. The McInnes et al. (2009a) estimates lie within the 95% confidence intervals of the estimates from this study, reflecting the general agreement between tide height frequency histograms generated from the model-derived tide constituents and those derived from observations.

Figure 5.

1-in-100-year storm tide heights for selected locations along the Victorian coast under current climate conditions and climate change scenarios as indicated in Table II. 95% confidence intervals in current climate storm tide heights are shown as shading around the solid black line. Relative locations of the points are indicated on the coastal outline at the bottom of the panel. The bold line on the bottom axis indicates locations that are within Port Phillip Bay. In the legend, sea-level rise values in metres are given in brackets, asterisks refer to the inclusion of the appropriate wind-speed increase given in Table II, McI09 refers to values obtained from McInnes et al. (2009a) and HAT refers to Highest Astronomical Tide

The relative contributions of wind-speed increases and sea-level rise to changes in extreme sea levels can also be examined in Figure 5 by comparing 1-in-100-year storm tide heights for scenario 2, (wind-speed change and sea-level rise), with those for scenario 1, (sea-level rise only). For the locations included in the figure, 100-year storm tide return levels for scenario 2 compared to scenario 1 are 0.05–0.10 m higher in 2030, 0.13–0.22 m in 2070 and 0.18–0.33 m in 2100, respectively. These results indicate that, for the high-end IPCC (2007) sea-level rise estimates considered, the contribution to storm tide heights of consistent high-end estimates of wind-speed increase is considerably smaller, by a factor of more than two in the late 21st Century, than the contribution of sea-level rise. In this region it therefore seems likely that climate change will have a greater impact on extreme storm tide heights through sea-level rise than through wind-speed changes. It follows that sea levels currently attained only during severe storms will be reached during much less extreme storm conditions in the future.

3.2. Inundation

Table II shows, for the nine analysis regions shown in Figure 2, the area of land and the number of land parcels potentially exposed to inundation more frequently than once every 100 years under current climate conditions, assuming our best estimates of 1-in-100-year storm tide heights. Results for current climate conditions and plausible conditions for 2030, 2070 and 2100 are given. Figures 6, 7 and 8 show the spatial distribution of land potentially exposed to inundation for the Ocean Grove, Sea Spray and Aspendale analysis regions respectively.

Figure 6.

The Ocean Grove analysis region, showing land potentially exposed to inundation more frequently than once every 100 years under current climate conditions and various scenarios of future sea-level rise (SLR)

Figure 7.

The Sea Spray analysis region, showing land potentially exposed to inundation more frequently than once every 100 years under current climate conditions and various scenarios of future sea-level rise (SLR)

Figure 8.

The Aspendale analysis region, showing land potentially exposed to inundation more frequently than once every 100 years under current climate conditions and various scenarios of future sea-level rise (SLR)

3.2.1. Inundation under current climate conditions

Under current climate conditions the areas most vulnerable to inundation from a 1-in-100-year storm tide are generally beach fronts, low-lying wetlands, coastal parks and wildlife reserves. The analysis regions where these are extensive exhibit the greatest vulnerability in terms of area of land vulnerable to inundation. Of the nine analysis regions considered, only the Ocean Grove, Sea Spray and Tooradin regions contain over 20 km2 of land vulnerable to inundation. Almost all of this is low-lying land located adjacent to the sea, rivers or behind coastal dune systems (Figures 6 and 7). However in the Tooradin region, which consists of reclaimed swamp land, vulnerable land extends as far inland as key regional road and rail links.

Six regions, Ocean Grove, Sea Spray, Tooradin, Aspendale, Williamstown and Queenscliff, contain more than 200 land parcels vulnerable to inundation. The number of land parcels in a region that are vulnerable to inundation is a function of both land area vulnerable to inundation and density of parcels. The urban Aspendale and Williamstown analysis regions are part of the Melbourne conurbation and contain many areas with a high density of small parcels. The other analysis regions are more rural in character and high parcel densities are confined to more limited areas corresponding to regional settlements. Significant parts of these settlements are generally not vulnerable to inundation by sea water under current climate conditions, the exception being in the Queenscliff region, in which some of the settlement of Queenscliff is vulnerable. Although less than 10 km2 of land is vulnerable to inundation in each of the Aspendale and Williamstown regions, the numbers of land parcels vulnerable to inundation are similar to the number for the Sea Spray region, which contains 27 km2 of land vulnerable to inundation.

Compared with the other six analysis regions, the Point Wilson, Portland and Port Fairy regions exhibit relatively little vulnerability in terms of either land area or number of land parcels vulnerable to inundation. Each of these regions contains less than 10 km2 of vulnerable land and less than 200 vulnerable land parcels.

3.2.2. Inundation under future climate conditions

For all regions a sea-level rise of 0.15 m, the value considered here for 2030, results in an increase in land area vulnerable to inundation by 0–40%. It is noteworthy that these percentages can also be regarded as an approximate guide to the uncertainty in current climate inundation since Figure 5 shows that the 1-in-100-year storm tide heights estimated for 2030 approximately coincide with the upper bounds of the 95% confidence intervals for current climate storm tide heights. Of the three regions that exhibit the greatest potential for inundation under current conditions, Ocean Grove, Sea Spray and Tooradin, Tooradin undergoes the greatest increase under greater values of sea-level rise. In this region, a sea-level rise of 1.4 m, the largest value considered for 2100, results in a 160% increase in land area vulnerable to inundation, and a tripling in the number of vulnerable land parcels. The corresponding increases in land area vulnerable to inundation are much more modest for Ocean Grove and Sea Spray, 30 and 60%, respectively. However, the additional area potentially subject to inundation coincides with the coastal settlements of Ocean Grove, Barwon Heads, Sea Spray and The Honeysuckles (Figures 6 and 7). This results in disproportionate increases in the number of vulnerable land parcels, an 8-fold increase for Ocean Grove and a 13-fold increase Sea Spray. It is notable that in the Ocean Grove region, a tripling of the number of parcels vulnerable to inundation occurs between sea-level rise values of 0.82 and 1.1 m with much of the inundation potential occurring within the township of Barwon Heads. This is an example of a highly nonlinear response to sea-level rise, where a step change in the number of vulnerable land parcels has occurred as a result of the exceedance of a threshold value of mean sea level.

Of the six regions that contain the largest number of land parcels that are currently vulnerable to inundation, Ocean Grove, Sea Spray, Tooradin, Aspendale, Williamstown and Queenscliff, the largest increases in the number of land parcels vulnerable to inundation by 2100 occur for the urban regions of Aspendale and Williamstown. In the Aspendale region, a five-fold increase in the number of vulnerable parcels occurs for a sea-level rise between 0.47 m and 0.82 m, associated with the overtopping of embankments along the Patterson River and increased vulnerability of residential areas around the Edithvale-Seaford wetlands (Figure 8). Owing to the topography of the settlements in the Queenscliff region, increases in land parcel vulnerability under a sea-level rise of 1.4 m are more limited than for the Aspendale and Williamstown regions.

As for current climate conditions, by 2100, the Point Wilson, Portland and Port Fairy regions exhibit relatively little vulnerability in terms of either land area or number of land parcels vulnerable to inundation. Under a sea-level rise of 1.4 m, each of these regions contains less than 30 km2 of land and less than 1000 land parcels vulnerable to inundation.

The additional effect of wind-speed increase on the area of land vulnerable to inundation is slight for 2030. The land area vulnerable to inundation for the combination of a sea-level rise of 0.15 m and an increase in wind speed of 4%, the values considered for 2030, is generally less than 10% greater than the land area vulnerable to inundation for the same sea-level rise with no wind-speed change. With the exception of the Queenscliff region, the number of land parcels vulnerable to inundation for the combination of a sea-level rise of 0.15 m and an increase in wind speed of 4% is less than 20% greater than the number vulnerable to inundation under a 0.15 m sea-level rise alone. Although the additional effect of wind-speed increase on the area of land vulnerable to inundation is greater for 2070 than it is for 2030, this does not always translate into a greater additional effect on the number of land parcels vulnerable to inundation. The additional effect of wind speed increase for 2070 is greatest for the Aspendale and Port Fairy regions. The area of land vulnerable to inundation is doubled for these regions and the increases in the number of land parcels vulnerable to inundation are approximately four-fold and 80%, respectively. When potential inundation of either land area or land parcels for 2100 is considered, the effect of the combination of a sea-level rise of 0.82 m and an increase in wind speed of 19% is approximately equivalent to the effect of a sea-level rise of 1.1 m with no wind-speed change.

4. Discussion

The evaluation of inundation using a simple algorithm that identifies land area adjacent to the coast that lies below the level of a storm tide, colloquially known as the bathtub approach, contrasts with dynamical approaches that simulate the flow of sea water over the land surface with a hydrodynamic model (e.g. Hubbert and McInnes, 1999; McInnes et al., 2003). The simple approach was used here for several reasons. Firstly, the hydrodynamic model was used to simulate tide and surge separately. No storm tide simulations of the two contributions combined were performed. Hence the dynamic simulation of inundation could not be performed without the running of an additional set of hydrodynamic model simulations. Secondly, the simple approach is computationally inexpensive and permits a range of different storm tide heights to be explored. Thirdly, the inundation analysis can be carried out at much higher resolution than the hydrodynamic modelling used to estimate storm tide heights. For example, in this study, the inundation was estimated on 10-m-resolution grids using storm tide levels estimated on the basis of hydrodynamic modelling at grid resolutions ranging from 100 m to 1 km. Nevertheless, it should be noted that the 10-m resolution used for the inundation calculations will not resolve all flood protection measures that may protect from inundation, or resolve small-scale features such as drainage systems that may allow flood water penetration. Therefore it must be stressed that the areas indicated by the algorithm provide broad guidance only.

The use of the simple non-dynamical inundation algorithm means that there are also processes not considered in this study that may affect the extent of inundation. The algorithm does not account for the flood water currents and how these would be modified by friction at the terrain surface, or how wind stress may increase or impede the current flow. Therefore, for areas that contain tracts of low-lying land that penetrate far inland, inundated areas may be different to those indicated in this study. There is also no accounting for the duration of the higher storm tide levels and the effect this would have on the inland penetration of inundation. If the storm tide sea levels peak for only a short period of time, then the inland penetration of flood water may be limited. However this is a limitation only for the storm tide component of the extreme sea levels rather than the sea-level rise component. The former is a transitory process while the latter is a long-term change which may lead to permanent inundation of low-lying coastal terrain.

Another factor that could contribute to coastal inundation that is not considered in this study is the effect of waves, which may enhance the amount of inundation that occurs during a storm tide event. Waves can contribute to extreme sea levels through wave setup and wave runup. Estimating the contribution of waves to extreme sea levels, which is generally much smaller than that of a storm surge, was beyond the scope of the present study. However a preliminary study of wave setup during storm surge events has shown that coastline orientation in relation to the prevailing winds and waves during a storm event will strongly determine the magnitude of wave setup at the coast (McInnes et al., 2009b; O'Grady and McInnes, 2010). Future work should aim to quantify the contribution of waves to sea-level extremes along the Victorian coastline. High-resolution bathymetric LiDAR datasets that are being developed as part of the Future Coasts Program will be important to this endeavour.

The storms associated with storm tides are often accompanied by significant rainfall in river catchments intersecting the coastal zone. Hence, in addition to coastal inundation due to extreme sea levels, a storm tide may also be accompanied by flooding due to rainfall. This additional contribution to inundation was not considered in this study, but would potentially increase the area affected by inundation.

This study has regarded the topography of the coastline as being constant throughout the 21st Century. However during this time period environmental processes, such as the erosion of beaches and soft cliffs, have the potential to change the morphology of the shoreline. Indeed many of these processes will be affected by increases in mean and extreme sea levels associated with climate change and hence climate change itself may result in significant changes in the shoreline in the future. Superimposed on these environmental processes will be the adaptive responses of society to changes in the shoreline. Such adaptive responses may include the continuing re-nourishment of beaches to retain the existing coastline, the building of sea walls along the coast and embankments along water courses to inhibit erosion and inundation, and the infilling of low-lying land.

The omission of relevant processes from this study and the limited resolution of the data used means that the inundation analysis and maps shown should be considered only as indicative of areas that would be vulnerable to inundation under the scenarios and storm conditions considered. However such information is often valuable to decision makers as the resourcing requirements of higher-resolution analyses incorporating a more comprehensive set of processes is often prohibitive.

5. Conclusions

In this study, 1-in-100-year storm tide heights along the coast of Victoria, southeast Australia, have been evaluated by using the JPM to combine storm surge and tide heights simulated using a hydrodynamic model. The study extends an earlier study of individual locations along the coast by McInnes et al. (2009a) by developing continuous spatial surfaces of the 100-year return level. These surfaces are used to investigate the inundation that may arise from extreme sea-level events under current climate conditions and under several plausible future climate change scenarios at nine low-lying coastal locations.

This study finds that the 1-in-100-year storm tide levels vary considerably along the Victorian coastline due to variations in both storm surge heights and astronomical tides. Storm tides are highest in and around Western Port Bay, where they exceed 2.0 m under current climate conditions. They are also high along the open coastline from just west of Port Phillip Bay to Wilsons Promontory, exceeding 1.8 m under current climate conditions. In comparison, in the west of the state and for much of Port Phillip Bay, 1-in-100-year storm tides are low, with magnitudes of around 1.1 m under current climate conditions. While values obtained in the present study were generally consistent with those obtained in McInnes et al. (2009a), the areas of largest difference coincided with those areas for which there are large uncertainties associated with errors in surge and tide modelling and in the statistical estimation of recurrence intervals.

Four climate change scenarios were used to estimate storm tide return levels under future climate conditions and, hence, to investigate the impact of future climate change on coastal inundation. Sea-level rise values from the IPCC's high-end estimates (Hunter, 2010) and higher estimates of sea-level rise from Katsman et al. (2011) and Rahmstorf (2007) formed the basis of three climate change scenarios that incorporated sea-level rise only. A fourth scenario incorporated consistent high-end estimates of changes in annual average wind speed (CSIRO and Australian Bureau of Meteorology, 2007).

For the high-end sea-level rise estimates considered, the contribution of consistent high-end estimates of wind-speed increase to the increases in extreme storm surge heights is considerably smaller, by a factor of more than two, than the contribution of sea-level rise. It therefore seems likely that climate change will have a greater impact on extreme storm surge heights through sea-level rise than through wind-speed changes. It follows that sea levels currently attained only during severe storms will be reached during much less extreme conditions in the future.

The potential inundation caused by a 1-in-100-year storm tide under various climate change scenarios was investigated. Nine regions along the Victorian coast were selected for inundation analysis on the basis that they contained extensive areas of terrain below 2 m elevation. The main findings of the inundation analysis were that, under a 1-in-100-year storm tide under current climate conditions, the most vulnerable areas were generally beach fronts, low-lying wetlands and coastal reserves. Under plausible values of sea-level rise by 2100 extensive additional areas of land were found to become vulnerable in some regions. In other regions more modest increases in land area vulnerable to inundation were found to result in disproportionate increases in the amount of vulnerable property if the additional areas potentially subject to inundation coincided with coastal settlements. In some cases, highly nonlinear response to sea-level rise were identified, where a step change in the amount of property potentially vulnerable to inundation occurs as a result of the exceedance of a threshold value of mean sea level. The additional effect of a high-end estimate of wind speed increase on the area of land vulnerable to inundation was found to be slight for 2030, but significant for 2100.

It should be noted that the methodology developed in this study, as well as the resolution of the hydrodynamic models and bathymetric and atmospheric datasets used, have been guided by the desire to provide data for the entire Victorian coast within the limitations of the available computing resources. It is possible that higher-resolution studies of sections of the coastline utilising different methodologies and datasets may yield different return levels. Despite various limitations, the analysis presented in this study illustrates how different climate change scenarios may affect the degree of inundation that could potentially occur due to extreme sea levels in the future. This study identifies areas that will be most vulnerable to coastal inundation in the future and highlights thresholds of sea-level rise that are important in the context of vulnerability and adaptation.

Acknowledgements

This work was undertaken as part of the Australian Climate Change Science Program, funded jointly by the Department of Climate Change and Energy Efficiency, the Bureau of Meteorology and CSIRO, and the Victorian Department of Sustainability and Environment.

Ancillary