ORI GIN AL PA PER

Analysis of tsunami impact scenarios at an oil refinery

Ana Maria Cruz • Elisabeth Krausmann • Giovanni Franchello

Received: 10 December 2009 / Accepted: 23 October 2010 / Published online: 10 November 2010� Springer Science+Business Media B.V. 2010

Abstract This work analyses the potential consequences of two tsunami scenarios and

their impacts on an oil refinery located in Sicily. Two credible tsunamis originating in the

Tyrrhenian Sea were selected based on historical data. The potential for damage and

hazardous materials releases resulting from the tsunami impacts to a refinery was assessed.

The results obtained by the JRC tsunami propagation and inundation code HyFlux2indicate that in both scenarios there would be eighteen storage tanks (of 43 located within

400 m from the shoreline) at the refinery subject to flooding. Water flow velocities were

found to be generally low, \1 m/s, except for a central section of the refinery near the

shoreline where the water flow velocities reach 3–4 m/s. These results indicate that any

damage would most likely occur due to buoyancy loads particularly in the western part of

the facility where inundation levels are higher and storage tanks are less protected.

Potential damage caused by impact of floating debris may be a problem in the central area

of the refinery near the shoreline due to high flow velocities (3–4 m/s) in both tsunami

scenarios. Small hazardous materials releases could occur due to breakage of connected

pipes and flanges caused by floating off of almost empty storage tanks or other equipment.

Salt water intrusion could affect electrical equipment, such as control panels, pumps, and

motors that are not raised above the inundation level. We conclude that in the two tsunami

scenarios analysed, the risk to nearby residents and neighbouring facilities from potential

hazardous materials releases, fires or explosions triggered by the tsunamis is likely to be

small. Nonetheless, recommendations are made on prevention measures to reduce the risk

of tsunami-triggered accidents and to mitigate their consequences if they do occur. The

results of this study are limited by the uncertainty in the input data and most importantly by

the accuracy of the elevation data and the model resolution.

A. M. Cruz (&) � E. Krausmann � G. FranchelloEuropean Commission, Joint Research Centre, Institute for the Protection and Security of the Citizen,TP 361, Via E. Fermi, 2749, 21027 Ispra, VA, Italye-mail: cruzanamaria2000@yahoo.com

A. M. CruzNatural and Technological (Natech) Disaster Risk Management, 11 rue Jean Mermoz, 33800Bordeaux, France

123

Nat Hazards (2011) 58:141–162DOI 10.1007/s11069-010-9655-x

Keywords Tsunami � Tsunami propagation and inundation � Risk assessment �Industrial facility � Flooding � Hazardous materials

1 Introduction

Engineered structures in coastal regions exposed to tsunami (or storm-surge) inundation

can be subject to a variety of concomitant damaging phenomena (Rossetto et al. 2007;

Robertson et al. 2007). These include uplifting due to submersion of structures, overturning

and displacement due to wave loading, debris impact, or foundation failure caused by

liquefaction-induced scour in sandy subsurface deposits. The impact of a tsunami on a low-

lying coastal industrial area could therefore result in damage to or collapse of industrial

buildings, tanks and other equipments. If damaged equipment or buildings are used for

storing or processing hazardous materials, these could be accidentally released with

potentially devastating consequences on the population and the environment. Furthermore,

since tsunami inundation could affect a wide swath of land, the risk in densely industri-

alized areas could be elevated.

The oil refinery fires triggered by the 1964 Niigata earthquake and tsunami in Japan

serve as an example of the potentially catastrophic effects of a tsunami when it affects a

large industrial complex located near residential areas. During this event, a 4 m tsunami

was triggered by an Mw = 7.6 magnitude earthquake, which initially caused fires in five

storage tanks and oil spills in hundreds more at two oil refineries in Niigata (Iwabuchi et al.

2006; IRIS SeismoArchives 2010). The tsunami hit the already earthquake-stricken

facilities resulting in:

• additional damage to storage tanks and plant processing equipment by collision with

tsunami-driven objects and by the hydrodynamic forces of the tsunami (Iwabuchi et al.

2006).

• the spreading of leaked oil by the tsunami current into the harbour and on inundated

land (Iwabuchi et al. 2006).

• the spreading of burning crude oil carried by the flood waters causing the fires to extend

to other parts of the plant including the heating furnace, the heat recovery boiler, the

reactor of the catalytic conversion process, the hydrolysis treatment equipment, and the

bottom of the hydrolysis reactor for the desulphurization process (Akatsuka and

Kobayashi 2008).

• the spreading of ignited crude oil carried by the flood waters into residential areas and

the destruction of 286 houses by the fire (Iwabuchi et al. 2006; Akatsuka and

Kobayashi 2008).

Oil and gas spills were also reported following a tsunami triggered by the Kocaeli

earthquake of 17 August 1999 in Turkey (Steinberg and Cruz 2004) and during the recent

Indian Ocean tsunami of 26 December 2004 (Borrero 2005; Van Dijk 2008).

Tsunami and consequent floodwaters can impose different loads on buildings, including

industrial facility buildings and equipment. These include hydrostatic loads, buoyant loads,

hydrodynamic loads, surge loads, breaking wave loads and impact loading, which results

from floating debris (Yeh et al. 2005). Current structural design codes focus on loading due

to riverine floods and storm waves, providing little guidance for loads specifically induced

by tsunami effects on coastal structures (Yeh et al. 2005). In addition, there is little

guidance on how industrial facilities can prevent or mitigate tsunami effects. This is in part

because there is scant empirical data that relate the various damage states of industrial

142 Nat Hazards (2011) 58:141–162

123

equipment versus flood water depths and water flow velocities. Furthermore, although

efforts to assess the vulnerability of buildings to tsunamis exist (e.g., Dall’Osso et al. 2009;

Dominey-Howes and Papathoma 2007), there are no specific methodologies available to

carry out vulnerability and risk assessments of tsunami impact to industrial facilities.

In this work, we analyse the potential consequences of tsunami impact on an industrial

facility that houses and processes hazardous materials. The work was performed under the

European Union’s 6th Framework Programme Project TRANSFER (2010) as part of its

Work Package 8 on scenarios of large tsunami impact and risk assessment and reduction.

This study focused on a case study of tsunami impact on the northern coast of Sicily, which

has been identified as a tsunami-prone region under the TRANSFER project. Two credible

tsunami-source scenarios were selected based on historical data and modelled, and the

potential for damage and hazardous materials releases resulting from the tsunami impacts

to an industrial facility located in this area was assessed. This work presents the results and

analysis of our study and makes recommendations for tsunami risk reduction.

2 Background

A tsunami is a sea wave of local or distant origin that results from large-scale water-body

displacements associated with large earthquakes, major submarine slides, or large volcanic

eruptions. It is characterized as a shallow water wave and differs from wind-generated

waves in its period and wavelength. Local bathymetry, undersea features and the slope at

the beach modify the tsunami as it approaches the shore, and the effects at the shoreline can

change considerably within short distances. See Fig. 1 for a definition of tsunami inun-

dation terms. Generally, run-up heights tend to be greatest near where the offshore

bathymetry is steeper. Furthermore, if a tsunami approaches the coast during astronomical

high tides, the run-up heights may be even greater.

Southern Italy has been (and could be in the future) affected by tsunamis generated both

in the Eastern Mediterranean (e.g., in the Hellenic Arc) and also locally (in the Tyrrhenian

Fig. 1 Definition of tsunami inundation terms (adapted from Anderson 2007)

Nat Hazards (2011) 58:141–162 143

123

and Ionian Seas) (Lorito et al. 2008). Tsunamis in these regions are associated with

earthquakes, volcanic eruptions and landslides (Papadopoulos and Fokaefs 2005).

Although tsunamis originating in the Mediterranean Sea are less frequent than those in the

Pacific or Indian oceans, they have caused extensive damage and loss of life (Lorito et al.

2008). Devastating tsunamis occurred following earthquakes of equivalent Mw = 6.9 and

7.2 in 1783 and 1908, respectively, in the Messina Straits in Southern Italy (Papadopoulos

and Fokaefs 2005; Tinti et al. 2004). The most updated tsunami catalogue for the Italian

region, put together by Tinti et al. (2004), contains data for 67 tsunamis observed in the

Italian seas. More than 50% of these affected Southern Italy.

The northern coast of Sicily, of interest for this study, could be at risk by tsunamis

generated by earthquakes along the southern Tyrrhenian source zone. Lorito et al. (2008)

estimated the average maximum wave height to be around 0.2 m. Waves of 0.5 m and

higher would affect only few localities around the northern coast of Sicily such as Palermo,

Trapani and Milazzo (Lorito et al. 2008). Tsunami waves resulting from volcanic erup-

tions, volcano-associated submarine slides and landslides, and mass failure (submarine and

land slides) have historically affected the coasts of Italy as well. Approximately 18% of the

tsunamis in the updated catalogue of Italian tsunamis (Tinti et al. 2004) are volcano

associated; less than 3% are due to mass failure. One recent example occurred in December

2002 following a phase of explosive activity of the Stromboli volcano. The tsunami was

caused by a massive submarine landslide followed by a sub-aerial landslide from an

elevation of 650 m above sea level on the north-west slope of Stromboli Island (Maramai

et al. 2005). Tsunami waves as high as 8–10 m were recorded in Stromboli, and smaller

waves were observed as far as 170 km at Mondello, northern Sicily and on the island of

Ustica.

3 Methodology and assumptions

3.1 Data and model description

Based on tsunami source data for two credible tsunami scenarios provided by the Uni-

versity of Bologna (Armigliato 2008, personal communication), tsunami flow velocities

and maximum water surface levels along the northern coast of Sicily were simulated using

the JRC code HyFlux2 (Franchello 2008, 2010; Franchello and Krausmann 2008),

developed to simulate severe inundation scenarios due to dam breaks, flash floods and

tsunami-wave run-up.

The code solves the shallow water equations using a finite volume method. The

interface flux is computed by a Flux Vector Splitting method based on a Godunov-type

approach. A second-order scheme is applied to the water surface level and velocity, pro-

viding results with high accuracy (Franchello 2008). Physical models are included to deal

with bottom steps and shorelines. The second-order scheme together with the shoreline-

tracking method makes the model well-balanced with respect to mass and momentum

conservation laws, providing results that are reliable and robust (Franchello 2010).

HyFlux2 was validated on 1D analytical test cases (Franchello 2008), a 2D run-up

numerical case study and finally with the Okushiri tsunami run-up problem (Franchello

2010). A validation with the Malpasset dam-break case study was also carried out

(Franchello and Krausmann 2008).

The numerical model was run in a cascade of nested simulations from coarse to fine grid

size: a grid size of 400 9 400 m in the abyssal plain, 100 9 100 m in the continental

144 Nat Hazards (2011) 58:141–162

123

shelf, a smaller grid size (20 9 20 m) near the shore, and a fine grid size (5 9 5 m) in the

run-up zone. The simulation results with the coarser grid were used as boundary condition

for the simulation with the finer grid.

The two scenarios were (a) an earthquake along the Capo Vaticano fault in Calabria and

(b) a landslide at Stromboli. Source data for both scenarios were provided by the

TRANSFER project partners at University of Bologna and included for scenario (a) the

water surface elevation at 0 s and for scenario (b) the velocity and water surface elevation

fields at 100 s, calculated by a finite element model.

The simulations for both scenarios were run assuming that the waves arrive when the

sea is at the maximum tide level (worst-case scenario). Data collected for the simulations

included digital elevation model (DEM) data at a 100 9 100 m grid size (SRTM DTED,

3 arcseconds), bathymetry data in the far field at a 1,000 9 1,000 m grid size

(SRTM30_plus, 30 arcseconds), and in the near field vector data corresponding to iso-

lines—digitized from a nautical map—for the northern coast of Sicily. Isolines were also

digitized for the specific area where the refinery is located, and then interpolated to a raster

grid, assuming a constant upward slope from the shoreline (0 m) to the railroad tracks that

cross the refinery from west to east. Figure 2 presents a map of the DEM used for the study

region in the near field. The tanks were modelled as bottom steps (see Franchello 2008)

without additional DEM refinement. The hydrostatic and hydrodynamic load exerted by

the wave on the tank wall was evaluated using the expression F/b = � qh (gh ? cDv2),

where F is the force [N], b the obstacle width [m], q the water-debris density [kg/m3], h the

water depth [m], g the acceleration of gravity [m/s2], v the velocity normal to the obstacle

[m/s] and cD a drag coefficient. This load was calculated at each time step, and a maximum

value was determined for each cell of the entire space domain.

Data about the studied refinery, whose name and location are not disclosed to protect its

identity, were obtained from previous work carried out by Giardina (2000) and through

review of safety reports and other pertinent documents. These data were compared with

recent satellite images of the industrial facility obtained from Google Earth. Storage tank

dimensions and maximum storage quantities were estimated based on measurements taken

from the Google Earth satellite image of the refinery using typical height to diameter ratios

for large atmospheric storage tanks, and assuming tanks were filled to 75% capacity

Fig. 2 Map of DEM data (in metres) for the study region (total map width = 2,315 m). Dots show thelocation of storage tanks. The tank walls were modelled as bottom steps

Nat Hazards (2011) 58:141–162 145

123

(Austin 1988; Sinnott 1989). The refinery’s official website was also reviewed. Data on

total number of storage tanks and quantity of chemicals stored and processed reported on

the refinery’s website were used to check the accuracy of the estimated tank dimensions.

The weight of each tank was estimated to determine buoyancy of empty tanks subjected

to flooding. In general, weight loads considered for design purposes include vessel shell,

vessel fittings (e.g., manways and nozzles), internal fittings (e.g., plates, heating and

cooling coils), external fittings (e.g., ladders, platforms), auxiliary equipment, which is not

self-supported (e.g., condensers, agitators), insulation and the weight of the liquid to fill the

vessel (Sinnott 1989). In this particular case, we were interested in the total weight of the

shell and fittings excluding the contents. A more detailed discussion on the method used to

estimate the tank shell thickness and tank weight is presented in Cruz et al. (2009). In

general, when refinery specific data were not available, typical generic data for this type of

facility were used for the calculations.

A field visit was carried out in order to collect data on altitudes (elevation above sea

level) and latitude/longitude coordinates at various reference points to compare and vali-

date topographic (DEM) data used for the tsunami inundation modelling. Furthermore,

features that might influence tsunami wave propagation and inundation were documented.

The field visit included data collection offsite as the refinery did not grant an on-site visit.

3.2 Study limitations

The study has a number of limitations, which are due to the assumptions made and the

uncertainties in the input data. The most important limitation concerns the accuracy of the

bathymetry and DEM data used in the simulations. As was discussed in Sect. 3.1, the DEM

was obtained by digitizing a nautical map. Thus, uncertainty results in the modelled

tsunami maximum water surface levels. The present study could be improved by using a

more accurate bathymetry and topography at the refinery as well as data concerning

containment walls, building structures and other plant features, and re-running the simu-

lation model with the 20 9 20 m and 5 9 5 m grid size DEM. Even very small changes in

local bathymetry and the slope at the beach can result in very different effects at the

shoreline within short distances. These inaccuracies will undoubtedly introduce an error in

the estimation of tsunami loads that depend on the maximum water surface levels and

velocity estimated at each tank.

The source data for the tsunami simulations provided by the University of Bologna are

based on simulations using historical data. While these results are certainly subject to

parameter and modelling uncertainties, we assume them to be negligible with respect to the

above-mentioned DEM uncertainty. The uncertainty in the model used for simulating the

tsunami propagation and inundation derives from the fact that HyFlux2 is a 2D model, and

it is therefore unable to capture 3D wave effects, such as turbulence and wave breakdown

on the shore. However, the model was tested validated against experimental tsunami wave

run-up output showing satisfactory results (Franchello 2010). Recently, the code was

assessed with respect to the 29 September 2009 Samoa tsunami (Annunziato et al. 2009),

showing good agreement with DART buoys and tide gauge measurements, and preliminary

results suggest that the model indicates inundation in areas where inundation was observed.

Finally, there is uncertainty in the present study concerning the exposed elements

(location, dimensions, distances, etc.) at the refinery such as storage tanks, pipes, pro-

cessing equipment, building structures and other plant features. An updated floor plan of

the refinery including accurate DEM data at each tank and major process equipment would

provide more reliable results on the vulnerability assessment.

146 Nat Hazards (2011) 58:141–162

123

4 Scenario development and analysis

4.1 Description of the oil refinery

The refinery occupies an area of about 2 km2 along a stretch of coast line about 1.7 km

from west to east in a bay. On its eastern side, the refinery is bordered by a power plant.

The south-eastern corner is bounded by a small road and a highway. The southern part of

the refinery is bounded by a small road, becoming residential towards the west. The

distance between the shoreline and the refinery fence line at the eastern end is about 20 m,

at an elevation of 2–2.5 m. The distance from the shoreline to the refinery fence line in the

western most tip of the refinery is 80 m at an elevation of 1–1.5 m. For the most part, the

western end of the refinery on the shoreline is flat, and in fact the storage tanks and other

equipment that are visible across the fence line are slightly below this level. The refinery is

bounded by residential areas and the port area on its western end and by residential areas

along the western and south-western edges. Thus, any oil spills on the shoreline during a

tsunami (from a source to the east or north such as in the present tsunami simulations)

would likely be transported by wave action into the nearby port area and inundate resi-

dential areas to the west of the refinery. Figure 3 shows a satellite photo of the refinery.

The field observations from outside the fence line of the refinery indicated that the DEM

data used for the tsunami modelling sufficiently approximates the elevations near the

railroad tracks but was less accurate in representing elevations near the shoreline (±2 m).

This undoubtedly adds uncertainty to the tsunami modelling results. Nonetheless, the DEM

data used were the best available data for the simulations.

Fig. 3 Satellite photo of the refinery. (Source: Google Earth). 1. North-west side of the refinery, elevation:1–1.5 m. 2. North-east side of the refinery, elevation: 2–2.5 m. 3. Pump station, elevation: 1 m. 4. East sideof the refinery neighbouring with the power plant, elevation: 6 m. 5. South-east side of the refinery,elevation: 16 m. 6. Small buildings neighbouring the refinery, elevation 1.5 m. 7. Power plant, elevation:1.5–2 m. 8. Open canal on south side of the refinery, elevation: 15 m. 9. Warehouse, pumping station andwater storage tanks. 10. The two refinery port terminals. 11. Distillation unit. 12. Watch point

Nat Hazards (2011) 58:141–162 147

123

The refinery storage farm is composed of 127 storage tanks with capacity of over

3.75 million m3 for prime materials, intermediate products and finished products. Figure 4

shows the main storage tanks at the refinery indicating corresponding tank identification (ID)

numbers. There are over 30 large storage tanks near the shoreline (less than 200 m from the

shoreline) at the refinery. Other features of the refinery located on or near the shoreline include

a warehouse, a pumping station, three large water storage tanks and the distillation unit.

The storage tanks on the eastern side (both near the shoreline and towards the southern

end) of the refinery (tank ID 52–58), used for crude oil storage, all have earthen con-

tainment dikes of 3–4 m in height. Storage tanks on the western side of the plant (tank ID

0, 1, 6, 7, 10, 11, 15, 18, 120, 121 and 31), used for the storage of intermediate and finished

products, appear to have 0.8–1 m high brick and concrete walls, but no earthen dikes.

During the field visit, containment walls or dikes were not always visible from our vantage

point outside the refinery. We assumed that the walls were present around all storage tanks

as this is general practice (see Fig. 5).

At the shoreline, there are no visible tsunami protection measures such as break walls,

or other types of barriers. On the eastern end, there is a natural earthen barrier about 4–5 m

in height, 2–3 m in width and about 30–40 m in length. See Fig. 6. This earthen barrier

could provide some protection from debris impact at this point of the plant.

The entire refinery is surrounded by a fence. On the shoreline, the fence is supported by

a small brick and concrete wall (0.5 9 0.15 m), designed to keep intruders out, but not

water. The support wall for the fence is no longer visible in some areas (due to sand-

banking, see Fig. 6). The fence support wall on the southern and eastern sides of the

refinery (inland) stands high, about 2 m in height.

There are two port terminals or jetties used for loading and unloading tankers. The two

jetties penetrate into the sea to a distance of about 400 m. There are no break walls or

Fig. 4 Storage tanks and tank identification (ID) numbers

148 Nat Hazards (2011) 58:141–162

123

natural features to protect the port terminals and docked ships from wave action. The port

terminals are in operation all year round, having a crude oil reception rate capacity of

15,000 tons per hour and moving about 570 ships per year. Figure 7 shows the two port

terminals. Loading arms are visible from the shoreline.

4.2 Earthquake-triggered tsunami scenario

The first scenario is a tsunami triggered by an Mw 7.0 earthquake along the Capo Vaticano

fault near Calabria studied by Piatanesi and Tinti (2002). This earthquake has a return

period of 1 in every 475 years (10% in 50 years probability of exceedance). This fault line

Fig. 5 Storage tanks for finished product on western end of the refinery. Concrete containment dike(0.8–1 m) is visible around tank no. 53 in this photo

Fig. 6 Natural earthen barrier. Soil erosion is evident from this photo as well as sandbanking over the fenceline (bottom right side of the photo)

Nat Hazards (2011) 58:141–162 149

123

is believed to be the source of an earthquake and tsunami in Calabria on 8 September 1905,

which devastated many towns and villages and resulted in more than 500 victims. The

tsunami source conditions were provided in the form of an ANSI file containing data on

longitude, latitude and wave amplitude in metres (Armigliato 2008, personal communi-

cation). The wave amplitudes were input into the HyFlux2 model as initial conditions for

the tsunami propagation and run-up simulation. Figure 8 shows the initial conditions

for this tsunami scenario.

Although the Capo Vaticano fault is about 75 km from the area of interest, the earth-

quake would be felt in the north-eastern part of Sicily. Modified Mercalli Intensity (MMI)

values in this area for the earthquake were estimated at IV-IIV (Guidoboni et al. 2007)

based on historical records of the 8 September 1905 earthquake. Previous earthquake

experience shows that the likelihood of moderate damage (e.g., failure of some connected

Fig. 7 Loading arms are visible in this photograph at the port terminals of the refinery

Fig. 8 Initial conditions of theearthquake-triggered tsunamialong the Capo Vaticano fault(Armigliato 2008, personalcommunication)

150 Nat Hazards (2011) 58:141–162

123

pipes, repairable damage to tank support systems, moderate likelihood of release of tank

contents) at these MMI levels would be less than 5% (Seligson et al. 1996). Thus, the

facility would most likely experience no damage or just light damage such as slight

movement of tanks from tank supports due to the earthquake forces. No hazardous

materials releases would be expected due to the earthquake shaking.

The results of our tsunami propagation modelling show that the earthquake would

produce a large tsunami wave, which would propagate across the south-eastern Tyrrhenian

region affecting Calabria, the Aeolian archipelago and northern Sicily. The water starts to

recede after 13 min, while the first positive wave would arrive at the area of interest

approximately after 27 min (see Fig. 9, left side). The tsunami wave would be approaching

the coastline laterally from east to west, impacting the port terminals perpendicularly. The

maximum water surface level at the port terminals would be about 1.2 m above high tide,

70 min after the earthquake, with about a 2.5 m differential between the wave crest and

trough.

The maximum water surface level at the shoreline of the refinery would vary between 1

and 1.2 m. The water flow velocity at the shoreline would vary between 0.5 and 4 m/s. The

area immediately to the west of the warehouse and pump station would experience the

highest water flow velocities.

The distance to the shoreline of the maximum water run-up is estimated at about 140 m;

the area between the two port terminals would be inundated the greatest distance inland.

This area would also suffer the highest maximum water surface levels. In general, the

western half of the refinery would experience greater water surface levels, between 1.2 and

1.3 m, than the eastern end, which would be subject to maximum water surface levels of

1–1.2 m.

Fig. 9 Water surface level (top) and velocities (x, y) at Watch point 12 (see Fig. 3) for Capo Vaticano (left)and Stromboli (right) tsunami scenarios

Nat Hazards (2011) 58:141–162 151

123

There would be eighteen large storage tanks subject to flooding. Figure 10 (left side)

maps the maximum water surface level. Figure 11 (left side) shows the maximum water

flow velocities, and Fig. 12 (left side) shows the maximum loads. The circular, non-

coloured areas on the map show the presence of storage tanks. The western half of the

refinery is subject to higher maximum water surface levels and higher water flow velocities

than the eastern half. For this scenario, maximum water loads appear uniform along the

coast. However, loads near the shoreline are higher than further inland. Thus, the storage

tanks closer to the shoreline (ID. 18, 120 and 121) would experience higher loads.

4.3 Landslide-triggered tsunami

Recent research has demonstrated that the Stromboli Volcano has the potential to produce

large catastrophic tsunamis and that its effects would be felt along the northern coast of

Sicily (Tinti et al. 2003). In fact a landslide-triggered tsunami affecting the northern coast

of Sicily occurred fairly recently at Stromboli in 2002 (Maramai et al. 2005). Therefore,

the second scenario is a tsunami triggered by a landslide at the Stromboli Volcano. The

data provided represent the tsunami field at a time (100 s) after which it is reasonable to

Fig. 10 Map of storage tanks near the shoreline showing the maximum water surface level [m] for (left)Capo Vaticano and (right) Stromboli tsunamis (Map width = 2,315 m)

Fig. 11 Map of storage tanks near the shoreline showing the maximum water flow velocities [m/s] for (left)Capo Vaticano and (right) Stromboli tsunamis (Map width = 2,315 m)

Fig. 12 Storage tanks near the shore and maximum loads (hydrostatic ? hydrodynamic, N/m) for (left)Capo Vaticano and (right) Stromboli tsunamis (Map width = 2,315 m)

152 Nat Hazards (2011) 58:141–162

123

assume that the forcing due to the landslide is over and that what remains is the pure

tsunami propagation (Armigliato 2008, personal communication). The wave amplitude and

velocity components for the second scenario were then input into the HyFlux2 model for

the tsunami propagation and run-up simulations. Note that for the landslide scenario not

only the wave amplitude but also the velocity components were input to the model.

Figure 13 shows the location and initial conditions of the landslide-triggered tsunami.

The results show that the modelled landslide at Stromboli would generate a large

tsunami wave, which propagates towards the south-eastern Tyrrhenian region affecting

Calabria, the Aeolian archipelago and northern Sicily. The water starts to recede after

14 min, while the first positive wave would impact the refinery after about 18 min. See

Fig. 9, right side. The tsunami wave would approach northern Sicily from north to south,

thus impacting the coastline at the refinery perpendicularly. The maximum wave height at

the port terminals would be between 0.5 and 0.8 m above high tide with about a 1.2–1.8 m

differential between the wave crest and trough.

The maximum calculated water surface level at the shoreline at the refinery would be

lower than for the Capo Vaticano scenario, varying between 0.4 and 1.0 m. However,

the maximum water flow velocity is generally higher, varying between 1.5 m/s to about

2.4 m/s, with a peak of 4 m/s in the area immediately to the west of the warehouse and pump

station. The distance to the sea of the maximum water run-up varies between about 80 m to

a little over 160 m. The area between the two port terminals would be inundated with the

greatest distance inland, experiencing also the highest maximum water surface level. In

general, and similar to the Capo Vaticano scenario, the western half of the refinery would

experience higher maximum water surface levels, between 1.2 and 1.6 m, while the eastern

half is subject to maximum water surface levels of 0.8–1.0 m.

In this scenario, there would also be eighteen storage tanks exposed to flooding. The

maximum water surface level is shown in Fig. 10, right side. The storage tanks located on

Calabria

Aeolian Archipelago

Stromboli

Sicily

N

Fig. 13 Initial conditions of the landslide-triggered tsunami at Stromboli input into the HyFlux2 tsunamimodel (Armigliato 2008, personal communication). Time = 0 for our tsunami modelling with HyFlux2.However, the input data used for the modelling represents the tsunami field at a time = 100 s after which itis reasonable to assume that the forcing due to the landslide is over and that what remains is the pure tsunamipropagation

Nat Hazards (2011) 58:141–162 153

123

the western half of the refinery, particularly the area between the two port terminals is

subject to a higher maximum water surface level. The area beyond the eastern side of the

refinery, where the power plant begins, would be subject to a maximum water surface level

of 1.4 m.

Figure 11 (right side) shows the maximum water flow velocities. The areas subject to

the highest maximum water surface levels are also subject to the highest water flow

velocities, as in the Capo Vaticano tsunami scenario. However, in this case, the highest

water loads (hydrostatic ? hydrodynamic) are not evenly distributed, but would be higher

in the western half of the refinery and the eastern end beyond the refinery fence line. See

Fig. 12, right side.

5 Discussion

In these two hypothetical, but credible, tsunami scenarios, there could be up to 18 storage

tanks affected by flooding. Both tsunami scenarios are consistent in that they indicate

higher exposure to flooding of the western half of the refinery, and in particular, the area

between the two port terminals. Overall, the Stromboli tsunami produces slightly higher

water flow velocities, while the Capo Vaticano tsunami results in greater water depths at

the exposed tanks.

In order to understand the potential impact of the modelled tsunami scenarios at the

refinery, the vulnerability of the exposed equipment needs to be assessed. However, there

are only scarce simplified equipment damage models available in the literature to assess

damage probabilities of industrial equipment due to tsunami impact or flooding.

Antonioni et al. (2009) have proposed flood vulnerability curves for industrial equip-

ment (see Fig. 14). The equipment vulnerability curves were developed based on past flood

events correlating different damage states for different types of industrial equipment with

the maximum water depth and the square of the water flow velocity. Thus, the computed

maximum water depth and the maximum water flow velocities for both simulations were

used to determine an equipment vulnerability index ranging from 0 (low probability of

damage) to 5 (high probability of damage) for each storage tank subject to flooding. The

index values were read off Fig. 14.

The vulnerability index values were zero for almost all storage tanks for both tsunami

scenarios (except for two tanks during the Stromboli tsunami), which corresponds to a

0–5% probability of damage according to Antonioni et al. (2009). Two storage tanks, ID

Fig. 14 Plot for the calculation of the equipment vulnerability in the case of floods. V2 square of water flowvelocity; h maximum water depth. The vulnerability zones indicate the equipment vulnerability index(Antonioni et al. 2009)

154 Nat Hazards (2011) 58:141–162

123

120 and 121, which are located to the west of the centre of the refinery near the shoreline,

had index values of 1 for the Stromboli tsunami, which translates to a probability of

damage higher than 5%. Because the equipment vulnerability curves in Fig. 14 were

developed based on limited empirical data (mostly from riverine and hurricane related

flood events), the values computed in the study provide only an indication of potential

consequences and should not be considered definitive.

Water flow velocities are generally small for the Capo Vaticano scenario, but may reach

values of 1 m/s or higher at some tanks for the Stromboli tsunami. Under slow submersion

(water flow velocities \1 m/s), damage would be restricted to failure of flanges and

connections mostly due to floating off of unanchored empty or almost empty storage tanks

and attached appurtenances and pipes.

Antonioni et al. (2009) identified three damage threshold limit values for structural

damage with loss of containment and with consequent release of hazardous materials: (a)

high water level condition: water level, h [ 1 m and minimum velocity required:

v = 0.25 m/s; (b) high water flow condition: v [ 2 m/s and water level, h = 0.5 m; and

(c) high risk condition: water flow velocity, v C 1 m/s, and water level, h C 1 m. Thus,

under high water level conditions and low water velocity, buoyant loads are mostly

responsible for any damage and resultant releases of hazardous materials.

The forces exerted on each tank by the flood waters and the resulting buoyant force

were estimated for each tank assuming that the flood waters had overtopped existing

concrete or earthen containment walls around tanks.1 While these barriers serve the pur-

pose of retaining accidental releases from the tanks, they have not been designed to keep

flood waters out. However, it could be assumed that they would provide some level of

flood protection to the individual tanks as long as the walls are not damaged or overtopped.

The ratio, r, of the buoyant force to the weight of the tank was calculated for both

tsunami scenarios for all storage tanks subject to flooding. The results are shown in

Table 1. Empty (or almost empty) storage tanks with an r [ 1 could float off if not

properly anchored and suffer damage. There would be 14 storage tanks in the Capo

Vaticano scenario and 13 storage tanks in the Stromboli scenario with r [ 1. These storage

tanks are located near the shoreline just to the west of the centre of the refinery protected

by low concrete walls and they could float off if the walls are overtopped by the flood

waters. In both scenarios, buoyancy loads could also affect low-lying pipes, particularly

those located at the pump station and that are 10–20 m from the shoreline where water

depths are greater than 1 m and are not protected by containment dikes or walls. Floating

off of pipes could result in pipe breaks and consequent releases of hazardous materials if

these are in operation or if they have not been properly de-inventoried before the tsunami.

The flood waters run inland as far as 140 m in both scenarios. Inundation in this area is

likely to result in salt water intrusion on low-lying equipment such as pumps and motors,

electrical panels and electronic control equipment, particularly at and near the pump station

and warehouse, as well as at the western half of the refinery where water depths are greater.

Damage to electric and/or electronic control equipment could result in process upsets and

possible accidental releases of hazardous materials. The inundation may also overcome the

internal plant drainage system, possibly causing waste oil to be lifted by the floodwaters,

which may spark fires and explosions upon contact with an ignition source (e.g., hot

refinery parts).

1 The containment wall around the tank and other equipment was not modelled. As mentioned above, thetopography is considered smooth due to the linear interpolation of isolines with the exception of the tankwalls, which were modelled as bottom steps.

Nat Hazards (2011) 58:141–162 155

123

Tab

le1

Rat

io(r

)o

fb

uo

yan

tfo

rce

tow

eig

ht

of

emp

tyst

ora

ge

tan

kin

dic

atin

gb

uo

yan

cyo

fem

pty

tan

ks

for

the

Cap

oV

atic

ano

and

Str

om

bo

lits

un

amis

Cap

ov

atic

ano

Str

om

bo

li

IDD

(m)

F(m

)t

(mm

)w

(kg)

hm

ax(m

)fm

axX

Y(N

)fm

axZ

(N)

rhm

ax(m

)fm

axX

Y(N

)fm

axZ

(N)

r

04

01

54

.79

1,6

43

,359

0.1

73

37

,69

12

,13

4,8

85

1.3

30

.191

37

,69

12

,348

,40

71

.46

14

01

54

.79

16

4,3

59

0.1

18

18

,38

81

,45

7,0

25

0.9

00

.166

31

,42

72

,042

,66

41

.27

24

01

54

.79

16

4,3

59

0.0

00

01

20

.00

0.0

00

01

20

.00

34

01

54

.79

16

4,3

59

0.0

00

01

20

.00

0.0

00

01

20

.00

44

01

54

.79

16

4,3

59

0.0

00

01

20

.00

0.0

00

01

20

.00

54

01

54

.79

16

4,3

59

0.0

00

01

20

.00

0.0

00

01

20

.00

64

01

54

.79

16

4,3

59

0.4

39

12

5,3

31

5,4

04

,784

3.3

60

.346

86

,86

04

,263

,87

82

.65

74

01

54

.79

16

4,3

59

0.3

45

81

,53

74

,24

6,2

95

2.6

40

.305

71

,62

83

,756

,73

62

.33

84

01

54

.79

16

4,3

59

0.0

14

63

51

69

,14

60

.11

0.0

34

4,5

89

41

5,3

76

0.2

6

94

01

54

.79

16

4,3

59

0.0

23

1,5

85

27

8,3

09

0.1

70

.023

2,1

33

28

7,7

91

0.1

8

10

40

15

4.7

91

64

,35

90

.187

33

,91

92

,30

2,7

47

1.4

30

.205

37

,15

32

,526

,44

31

.57

11

40

15

4.7

91

64

,35

90

.166

29

,22

92

,03

9,9

80

1.2

70

.166

27

,16

52

,047

,40

31

.27

12

40

15

4.7

91

64

,35

90

.000

01

20

.00

0.0

00

01

20

.00

13

40

15

4.7

91

64

,35

90

.000

01

20

.00

0.0

00

01

20

.00

14

40

15

4.7

91

64

,35

90

.000

01

20

.00

0.0

00

01

20

.00

15

40

15

4.7

91

64

,35

90

.060

6,5

20

73

5,2

96

0.4

60

.075

8,5

31

92

5,3

06

0.5

7

16

40

15

4.7

91

64

,35

90

.000

01

20

.00

0.0

00

01

20

.00

17

40

15

4.7

91

64

,35

90

.000

01

20

.00

0.0

00

01

20

.00

18

55

16

7.0

34

11

,95

30

.503

25

5,3

56

11

,72

6,2

38

2.9

00

.403

18

0,1

64

9,3

96

,13

12

.33

19

55

16

7.0

34

11

,95

30

.000

02

30

.00

0.0

00

02

30

.00

20

55

16

7.0

34

11

,95

30

.000

02

30

.00

0.0

00

02

30

.00

21

45

17

6.1

12

66

,06

30

.000

01

60

.00

0.0

00

01

60

.00

22

45

17

6.1

12

66

,06

30

.000

01

60

.00

0.0

00

01

60

.00

31

55

16

7.0

34

11

,95

30

.626

36

7,3

27

14

,59

0,5

28

3.6

10

.327

10

4,1

22

7,6

17

,76

91

.89

156 Nat Hazards (2011) 58:141–162

123

Tab

le1

con

tin

ued

Cap

ov

atic

ano

Str

om

bo

li

IDD

(m)

F(m

)t

(mm

)w

(kg)

hm

ax(m

)fm

axX

Y(N

)fm

axZ

(N)

rhm

ax(m

)fm

axX

Y(N

)fm

axZ

(N)

r

32

55

16

7.0

34

11

,95

30

.000

02

30

.00

0.0

00

02

30

.00

33

55

16

7.0

34

11

,95

30

.000

02

30

.00

0.0

00

02

30

.00

34

55

16

7.0

34

11

,95

30

.000

02

30

.00

0.0

00

02

30

.00

52

60

15

7.1

94

75

,46

60

.379

16

8,6

81

10

,50

7,7

88

2.2

60

.180

44

,76

95

,000

,72

81

.07

53

60

15

7.1

94

75

,46

60

.000

02

80

.00

0.0

00

02

80

.00

54

60

15

7.1

94

75

,46

60

.000

02

80

.00

0.0

00

02

80

.00

55

60

15

7.1

94

75

,46

60

.000

02

80

.00

0.0

00

02

80

.00

56

60

15

7.1

94

75

,46

60

.000

02

80

.00

0.0

00

02

80

.00

57

60

15

7.1

94

75

,46

60

.397

17

6,9

71

11

,01

2,1

09

2.3

60

.162

43

,27

14

,502

,66

30

.97

58

80

15

9.5

81

,03

3,1

11

0.4

43

32

8,5

78

21

,81

3,9

42

2.1

50

.186

12

7,5

09

9,1

44

,60

70

.90

59

80

15

9.5

81

,03

3,1

11

0.0

00

04

90

.00

0.0

00

04

90

.00

60

80

15

9.5

81

,03

3,1

11

0.0

00

04

90

.00

0.0

00

04

90

.00

11

92

58

1.6

02

0,0

56

0.4

38

77

,37

52

,10

8,0

28

10

.73

0.2

69

33

,94

61

,296

,71

76

.60

12

02

58

1.6

02

0,0

56

0.7

46

21

9,2

53

3,5

91

,919

18

.28

0.4

10

68

,71

61

,971

,73

71

0.0

3

12

12

58

1.6

02

0,0

56

0.7

82

24

0,1

65

3,7

65

,960

19

.16

0.4

58

84

,85

72

,204

,37

61

1.2

2

12

22

58

1.6

02

0,0

56

0.3

86

61

,66

51

,85

8,2

18

9.4

50

.167

16

,47

18

03

,98

64

.09

12

33

51

23

.35

84

,85

00

.000

09

0.0

00

.000

09

0.0

0

12

43

51

23

.35

84

,85

00

.000

09

0.0

00

.000

09

0.0

0

12

53

51

23

.35

84

,85

00

.000

09

0.0

00

.000

09

0.0

0

Inad

dit

ion

tor,

the

tab

leco

lum

ns

sho

wth

eta

nk

iden

tifi

cati

on

nu

mb

er(I

D),

dia

met

er(D

),h

eigh

t(H

),ta

nk

wal

lth

icknes

s(t

),ta

nk

wei

gh

t(W

),m

axim

um

forc

eco

mp

on

ent

inth

eX

and

Ydir

ecti

on

(fm

axX

Y),

max

imum

forc

eco

mponen

tin

the

Zd

irec

tio

n(f

max

Z)

(bu

oyan

tfo

rce)

.F

max

Zis

div

ided

by

the

gra

vit

atio

nal

acce

lera

tio

n(9

.81

m/s

2)

bef

ore

com

pu

tin

gr

Nat Hazards (2011) 58:141–162 157

123

Water flow velocities at the shoreline just to the west of the pump station and warehouse

are high, reaching 3–4 m/s in both scenarios. Debris (e.g., drift wood, small boats, lumber)

being transported by the tsunami wave in this area close to the shoreline could result in

damage to the refinery fence line along the beach front. If the fence line was breached or

overtopped, debris could result in possible damage to building structures, storage con-

tainers, pipes or other exposed elements in this area.

High water flow velocities (2–4 m/s) and water depths[1 m in the Stromboli scenario,

combined with possible turbulence near the shoreline, could result in soil erosion and

localized scour particularly around the area just to the west of the warehouse and pump

station, up to the second port terminal, and at the far eastern side of the refinery where

some soil erosion and scouring is already present. The refinery fence line support wall

could also be subject to scouring in this area. Weakening of the support wall would

increase the likelihood of the fence line being overturned. Overall, erosion and scour would

most likely not represent a large threat to storage tanks at the refinery, but would mainly

undermine protection barriers such as the natural protection barrier, and the refinery fence

line that could serve to filter small debris, as well as the foundations of some equipment

located near the shoreline.

The likelihood of major hazardous materials releases from damage to storage tanks is

low in the previous scenarios. However, small leaks from damaged flange connections and

broken pipes due to floating off of empty or almost empty storage tanks are possible in both

tsunami scenarios. Damage from debris impact, erosion and scour appears not to be a

major issue in terms of potential releases, or at least not as a direct result. Oil spills and

other hydrocarbon releases (e.g., LPG) could result from broken loading/unloading arms

and pipe breaks at the port terminals, particularly during the Capo Vaticano tsunami. Since

the Capo Vaticano tsunami impacts the coastline laterally, the two port terminals would

receive the wave impact perpendicularly. Thus, any ships docked on the east side of the

port terminals could have a forcing effect on the pier structures and could cause some

damage, while a ship moored on the west side would be pushed away from the pier

possibly tearing pipe connections and leading to oil spills. In fact, an oil spill at this

refinery was observed following the Stromboli tsunami of December 2002. Two oil tankers

moored at the wharfs were at risk of having their mooring lines broken when the tsunami

wave displaced them 10 m laterally. A small oil spill was reported as a result of the single

tsunami wave observed (Maramai et al. 2005). It is not clear if the spill was caused because

the moored tankers were still in the process of (un)loading or if transfer was stopped but

they were still connected to the wharf via the transfer arms.

6 Consequence analysis

The quantity of hazardous materials releases triggered by either of the two tsunamis is

likely to be small, thus also reducing the likelihood of major fires or explosions. Therefore,

the risk of potential harmful consequences from chemical releases, fires or explosions to

residential areas and neighbouring facilities is likely to be small. Nonetheless, even small

amounts of oil can form a thin film over flood waters and can be dispersed and transported

quickly throughout large areas. Any damage in this case will be restricted to coating of

surfaces with oil, although oil spills have been reported to disperse completely leaving

almost no trace. This would be the case if high turbulence occurs as was observed during

the Indian Ocean tsunami (Van Dijk 2008). If floating oil catches fire (e.g., due to contact

158 Nat Hazards (2011) 58:141–162

123

with very hot surfaces or another ignition source), the fire could be spread over large areas

very quickly. In this case, the refinery itself and its workers could be at risk.

Releases of flammable gases such as LPG due to damaged flanges or broken pipe

connections may occur. In the case of LPG, the port terminals are particularly vulnerable.

Any release could lead to the formation of a vapour cloud and possibly vapour cloud

explosion (VCE). According to an accident scenario taken from Giardina (2000) involving

a break of the transfer hose and the resulting overpressure, nearby residents would prob-

ably not be affected unless the vapour cloud is carried as far as the nearest residences about

500 m away. An explosion, however, could cause damage to the port terminals and other

ships, and could trigger domino effects.

7 Tsunami risk reduction measures

Our results indicate that the risk of damage and hazardous materials releases caused by the

two modelled tsunami scenarios would be low. Nevertheless, these are just two of the

possible credible scenarios that could affect the refinery. In addition, other industrial

establishments are located in tsunami-prone regions all over the globe. The study results

highlighted the possible weak points in the design and protection of industrial infra-

structures when subject to tsunami waves. Therefore, measures to prevent tsunami-trig-

gered chemical accidents and/or to mitigate their consequences need to be put into place to

ensure effective risk reduction.

Limiting industrial development in tsunami-prone areas is the most efficient way to

minimize the hazard associated with the tsunami impact. However, land-use-planning

restrictions for existing installations are difficult to implement. In this case, supplementary

measures are required to protect hazardous facilities. Offshore breakwalls or other types of

barriers onshore can be efficient in reducing the tsunami force (Ergin and Balas 2006;

Jayappa 2008; Maheshwari et al. 2005; Harris and Wilson 2008). They could also keep

tsunami-driven debris from washing into the plant. The studied refinery does not have any

breakwalls, which leaves the two oil terminals fully exposed to potential tsunami wave

action. In the eastern part, there is a stretch of earthen barrier that would protect any piping

and equipment lying behind it from the full force of the incoming waves. It does not,

however, run along the full length of the refinery and would therefore provide little

protection from inundation. This situation is exacerbated by the fact that the modelled

tsunamis are generated in the near field, with resulting short tsunami warning lead times.

According to our analysis, the first wave would arrive at the refinery after 18 min in the

Stromboli scenario and 27 min from Capo Vaticano. This time may be sufficient for

de-inventorying of pipes and emergency shutdown only. The time required for de-inven-

torying of connected pipes and pipelines will depend on the characteristics of the substance

being transported, the diameter and length of the pipe, and whether or not the pipeline is

feeding a process that can be quickly and safely shutdown. Careful analysis of the pro-

cesses and emergency planning and preparedness may help to reduce potential losses due

to tsunami impacts.

Tankers moored at the refinery’s oil terminals would require a warning lead time of a

few hours to stop (un)loading and move into deep waters to reduce the risk of a major oil

spill (Eskijian 2006). In addition, in the Capo Vaticano scenario the incoming tsunami

waves would hit the terminals laterally. Any moored tanker would suffer vertical but also

lateral displacement. Transfer needs to be rapidly stopped and the loading arms discon-

nected to minimize the risk of an oil spill during the tsunami.

Nat Hazards (2011) 58:141–162 159

123

In the absence of external barriers, all structures containing hazardous materials and all

systems critical for the safety of the installation need to be protected from wave-load

damage and water intrusion. Submersion and subsequent buoyancy are of particular con-

cern for storage tanks that may float off their foundations, thereby tearing pipe connections

and resulting in the release of possibly flammable and/or toxic materials. Adequate

anchoring should keep tanks and other equipment from floating off under most conditions.

In addition, a minimum quantity of material could be left in these to ensure that they would

not suffer buoyancy in the case of flooding. Since it was not possible to visit the studied

refinery, we could not determine the type of anchoring or restraint systems utilized on-site.

We noticed earthen containment dikes and brick or concrete walls around the storage tanks

visible from the plant perimeter. While their purpose is to retain accidental releases from

the tanks rather than keeping the flood waters out, they would probably provide some flood

protection if the walls are not overtopped and if there is no erosion that could compromise

the integrity of the dikes.

Structural tsunami risk reduction measures need to be supplemented by organizational

measures to effectively minimize the risk of a tsunami impact. This includes the drawing

up of tsunami hazard management plans both at plant and community level, monitoring of

construction practices and compliance with building codes. Moreover, when planning for

preventing tsunami-triggered chemical accidents, the impact of a possibly preceding

earthquake needs to be taken into account. This is applicable to both shore protection

systems and industrial facilities, which when weakened would have less resistance to an

impacting tsunami wave.

8 Conclusions

This paper analyses the potential impacts of two tsunamis originating in the Tyrrhenian Sea

and their consequences at an industrial facility located on the coast in north-eastern Sicily.

The results of the tsunami simulations indicate that in both scenarios, there would be 18

storage tanks at the refinery subject to flooding, with tanks closer to the shoreline suffering

close to 1 m inundation.

The study results indicate that there is low likelihood of damage to storage tanks due to

hydrodynamic loads in either scenario. Water flow velocities in most areas are less than

1 m/s, indicating that any damage would likely occur due to hydrostatic uplift forces due to

buoyancy loads particularly in the western part of the facility where inundation levels are

higher and storage tanks are less protected. Nevertheless, storage tanks in this part of the

refinery do have concrete containment dikes and these may provide some flood protection.

Thus, damage due to buoyancy loads would occur only if the containment dikes were

damaged by flood waters or if they were overtopped.

Debris impact loads could be a problem particularly in the area to the west of the

pump station and warehouse in both tsunami scenarios where water flow velocities reach

3–4 m/s. Debris impact could result in some damage to building structures, storage

containers, pipes or other exposed elements.

The refinery port terminals would be at risk of tsunami wave loads and possible damage,

particularly during the Capo Vaticano tsunami because the wave would impact the piers

perpendicularly. Thus, any ships docked at the port terminals could have a forcing effect on

the pier structures possibly causing some damage. Furthermore, ships moored on the west

side would be pushed away from the pier possibly tearing pipe connections and leading to

oil spills.

160 Nat Hazards (2011) 58:141–162

123

The likelihood of hazardous materials releases from inundated storage tanks is low.

However, small releases could occur due to breakage of connected pipes and flanges due to

buoyancy loads. Flooding of electrical equipment, such as control panels, pumps and

motors, not raised above the inundation level could result in salt water intrusion leading to

possible short circuit, hampering of safety and mitigation systems, process upsets and

possible hazardous materials releases.

We conclude that the risk of harmful consequences of tsunami-triggered hazardous

materials releases, fires or explosions to nearby residents and neighbouring facilities is

likely to be small. Nonetheless, uncertainty in the data used in the simulations, particularly

bathymetry and DEM means that different inundation values could be observed than those

obtained for the present study. Whatever the case, we recommend that the minimum of

preventive and preparedness measures be taken to reduce the risk of tsunami-triggered

damage, in particular, tsunami-induced chemical accidents and to mitigate their conse-

quences if these events do occur.

Acknowledgments This work was performed in the frame of and co-funded by the European Union’s 6thFramework Programme project Tsunami Risk and Strategies for the European Region (TRANSFER). Theauthors also acknowledge Dr. Alberto Armigliato from the University of Bologna for his help with preparingthe input data set for the tsunami propagation and inundation modelling.

References

Akatsuka H, Kobayashi H (2008) Fire of petroleum tank, etc. by Niigata earthquake. Failure knowledgedatabase, Japan science and technology agency. http://shippai.jst.go.jp/en/Detail?fn=0&id=CB1012035&kw=Environment Accessed Apr 2008

Anderson C (2007) State of Hawai’i multi-hazard mitigation plan update 2007. Hazards, climate andenvironment program, University of Hawaii with the Hawaii state hazard mitigation forum, the Hawaiistate earthquake advisory committee, the Hawaii state hurricane advisory committee, and the fourcounty planning efforts, including the county of Kauai, the county of Maui, the county of Hawaii, andthe city and county of Honolulu. http://www.scd.state.hi.us/2007_hmp.html Accessed 16 Dec 2008

Annunziato A, Franchello G, Ulutas E, De Groeve T (2009) 29 Sept 2009 Samoa Tsunami, EUR report24068 EN, European communities

Antonioni G, Bonvicini S, Spadoni G, Cozzani V (2009) Development of a general framework for the riskassessment of Na-Tech accidents. Reliab Eng Saf Syst 94:1442–1450

Austin GT (1988) Shreve’s chemical process industries, 5th edn. McGraw-Hill International Editions,Chemical Engineering Series, Singapore, p 859

Borrero JC (2005) Field survey northern Sumatra and Banda Aceh, Indonesia, after the tsunami andearthquake of 26 December 2004. Preliminary report. Earthquake Engineering Research Institute,Feb 9

Cruz AM, Franchello G, Krausmann E (2009) Assessment of tsunami risk to industry in southern Italy. EURReport 23801, European communities

Dall’Osso F, Gonella M, Gabbianelli G, Withycombe G, Dominey-Howes D (2009) Assessing the vul-nerability of buildings to tsunami in Sydney. Nat Haz Earth Syst Sci 9:2015–2026

Dominey-Howes D, Papathoma M (2007) Validating a tsunami vulnerability assessment model (the PTVAModel) using field data from the 2004 Indian Ocean tsunami. Nat Haz 40(1):113–136

Ergin A, Balas CE (2006) Damage risk assessment of breakwaters under tsunami attack. Nat Hazards39:231–243

Eskijian ML (2006) Mitigation of seismic and meteorological hazards to marine oil terminals and other pierand wharf structures in California. Nat Hazards 39:343–351

Franchello G (2008) Modeling shallow water flows by a high resolution Riemann solver. EUR 23307 EN,European communities

Franchello G (2010) Shoreline tracking and implicit source terms for a well balanced inundation model. Int JNum Meth Fluids 63:1123–1146

Nat Hazards (2011) 58:141–162 161

123

Franchello G, Krausmann E (2008) HyFlux2: a numerical model for the impact assessment of severeinundation scenario to chemical facilities and downstream environment. EUR 23354 EN, EuropeanCommunities

Giardina G (2000) Applicazione della metodologia HARIA2 alla pianificazione di emergenza nell’area diMilazzo. Master thesis, University of Pisa

Guidoboni E, Ferrari G, Mariotti D, Comastri A, Tarabusi G, Valensise G (2007) Catalogue of StrongEarthquakes in Italy (461 B.C.-1997) and Mediterranean Area (760 B.C.-1500). CFTI4Med, INGV-SGA. Available from http://storing.ingv.it/cfti4med/

Harris SP, Wilson DO (2008) Mitigating hurricane storm surge perils at the DeLisle plant. Process Saf Prog27(3):177–184

IRIS SeismoArchives (2010) http://www.iris.edu/seismo/quakes/1964niigata/. Accessed 24 June 2010Iwabuchi Y, Koshimura Shunichi, Imamura Fumihiko (2006) Study on oil spread caused by the 1964

Niigata earthquake tsunami. J Disaster Res 1(1):157–168Jayappa KS (2008) Coastal problems and mitigation measures including the effects of tsunami. Curr Sci

94(1):14–15Lorito S, Tiberti MM, Basili R, Piatanesi A, Valensise G (2008) Earthquake-generated tsunamis in the

Mediterranean Sea: scenarios of potential threats to southern Italy. J Geophys Res 113:B01301. doi:10.1029/2007JB004943

Maheshwari BK, Sharma ML, Narayan JP (2005) Structural damages on the coast of Tamil Nadu due totsunami caused by December 26, 2004 Sumatra earthquake. ISET J Earthq Technol 42(2–3):63–78

Maramai A, Graziani L, Alessio G, Burrato P, Colini L, Cucci L, Nappi R, Nardi A, Vilardo G (2005) Near-and far- field survey report of the 30 December 2002 Stromboli (Southern Italy) tsunami. Mar Geol215:93–106

Papadopoulos GA, Fokaefs A (2005) Strong tsunamis in the Mediterranean Sea: a re-evaluation. ISET JEarthq Technol 42(4):159–170 Paper No. 463

Piatanesi A, Tinti S (2002) Numerical modelling of the September 8, 1905 Calabrian (southern Italy)tsunami. Geophys J Int 150(1):271–284

Robertson IN, Riggs HR, Yim SCS, Young YL (2007) Lessons from hurricane Katrina storm surge onbridges and buildings. J Waterw Port Coast Ocean Eng 2007:463–483

Rossetto T et al (2007) The Indian Ocean tsunami of December 26, 2004: observations in Sri Lanka andThailand. Nat Hazards 42:105–124

Seligson HA, Eguchi RT, Tierney KJ, Richmond K (1996) Chemical hazards, mitigation and preparednessin areas of high seismic risk: a methodology for estimating the risk of post-earthquake hazardousmaterials release. Technical report NCEER-96-0013, Buffalo, New York

Sinnott RK (1989) Chemical engineering volume 6 (in SI units): an introduction to chemical engineeringdesign. In: Coulson JM, Richardson JF (eds) Volumes in the chemical engineering. Pergamon Press,Oxford, p 838

Steinberg LJ, Cruz AM (2004) When natural and technological disasters collide: lessons from the Turkeyearthquake of August 17, 1999. Nat Hazards Rev 5(3):121–130

Tinti S, Pagnoni G, Zaniboni F, Bortolucci E (2003) Tsunami generation in Stromboli Island and impact onthe south-east Tyrrhenian coasts. Nat Hazards Earth Syst Sci 3:299–309

Tinti S, Maramai A, Graziani L (2004) The new catalogue of Italian tsunamis. Nat Hazards 33(3):439–465TRANSFER (2010) Tsunami risk and strategies for the European region. www.transferproject.eu. Accessed

24 June 2010Van Dijk S (2008) Environmental impact assessment: tsunami Indonesia. United Nations, office for the

coordination of humanitarian affaires, Banda Aceh, Aceh, Indonesia. http://humanitarianinfo.org/sumatra/reference/assessments/doc/other/report_def_draft_send_2601.pdf Accessed March 2008

Yeh H, Robertson I, Preuss J (2005) Development of design guidelines for structures that serve as tsunamivertical evacuation sites. Washington State Department of Natural Resources, Washington Division ofGeology and Earth Resources, open file report 2005-4, November

162 Nat Hazards (2011) 58:141–162

123