Optimizing Tsunami Evacuation Plans Through the Use of Damage Scenarios

EarthzineArticles, Disaster Management Theme, In-Depth, Original, Sections, Themed Articles

A photo showing the author turning over a cover crop of winter rye on Jan. 2, 2012. In three of the past six years, ground freeze-up, which historically occurred by November, has been delayed into January.

Stefan Scheer, European Commission, Joint Research Centre, Ispra, Italy

Richard Guillande, Geosciences Consultants, Paris, France

Annalisa Gardi, Geosciences Consultants, Paris, France


Civil protection authorities need guidance in order to establish comprehensive emergency plans for tsunami-prone communities. In particular, the safe evacuation of all potentially affected persons prior to the arrival of the first devastating tsunami wave should be the primary goal in case of a tsunami alert. Evacuation is usually done on basis of well developed evacuation plans that operate on a given topography. This may be counterchecked through mathematical simulation, and further optimization of the plan may be achieved through the inclusion of additional safe areas (shelters) and/or appropriate escape routes within the plan. Recent research carried out in the framework of the European SCHEMA project suggests the inclusion of hazard and damage scenarios as these may identify suitable vertical shelters, suitable escape routes and even expected accumulation of debris.


Tsunami hazard scenarios are built up by specifying the various characteristics of one or more tsunamigenic sources (possibility of earthquakes and sub-marine landslides, historical earthquakes and sub-marine landslides). Through computer simulations, the oceanic tsunami wave propagation (arrival time of first tsunami wave on land, time intervals between the various waves) and its local effects on land (extent of inundation, wave heights) can be calculated. Hence, as basic input to the generation of tsunami evacuation plans, the expected flooded areas and the expected maximum wave height in these areas will be exploited in order to define the maximum number of affected persons and the time constraints to evacuate these persons onto safe areas.

Map of the maximum sea elevation computed for the Catania area. It shows a small difference between the maximum inundation line (for the worst-case scenario) and the black-lined inundation boundary according to an aggregated scenario (from Tinti et al., 2011)

Figure 1: Map of the maximum sea elevation computed for the Catania area. It shows a small difference between the maximum inundation line (for the worst-case scenario) and the black-lined inundation boundary according to an aggregated scenario (from Tinti et al., 2011).

The creation of suitable evacuation plans is done through simulation, too. Each simulation comes closer to a scenario that should demonstrate whether a number of affected persons can evacuate within a certain time toward the next dedicated shelter. Obviously, a couple of refinement steps can be applied during each simulation considering variants in escape routes and/or shelters. In particular and prior to their inclusion within an evacuation plan, vertical shelters (buildings, platforms) have to successfully pass damage scenarios that check their stability during a tsunami.

Evacuation Plan Generation

A tsunami evacuation plan is a plan that will be invoked if a tsunami alarm has been triggered. Hence, such a plan will affect preparedness measures among which the evacuation of the population is the most important. Concretely, all potentially affected persons should have sufficient time available to evacuate along a prescribed escape route toward a location declared as safe. Obviously, different escape routes may be attributed to differently located persons.

Consequently, the population distribution, their exposure to inundation (see Figure 1), the available time to evacuate, the availability of suitable escape routes, and the declaration of safe locations have to be known and to be used as starting point within a simulation exercise. A mathematically-based, time-cost algorithm does the necessary simulation; as positive outcome a complete evacuation of the affected population is guaranteed, while in the negative case, the simulation will show that some parts of the affected population may still be affected by tsunami waves.

An instance of a tsunami evacuation plan is a valid scenario that demonstrates that all affected persons will start moving toward a shelter; the scenario will stop with the arrival of the first tsunami wave [Scheer et al., 2010]. Such a scenario takes as granted that:

‰Û¢ Each person has her/his ‰ÛÏown‰Û shelter, usually the nearest;

‰Û¢ Each person has her/his dedicated escape route in order to reach the shelter;

‰Û¢ Each person has sufficient time to move (considering human speed, considering local topography),


‰Û¢ A maximum number of persons (considering temporarily residing persons);

‰Û¢ That evacuation is only done by foot, not by using vehicles;

and taking into account:

‰Û¢ Specific bottlenecks (which may reduce the throughput of persons in that place) and other local particularities.

Figure showing Generalized scheme of evacuation planning (Nagao 2005)

Figure 2: Generalized scheme of evacuation planning (Nagao 2005)

At first sight, one will choose those higher located areas as safe locations (shelters) that will not get flooded according to predictions of the tsunami hazard scenario. While this may well be the case in remote and probably less inhabited areas, the perspective may totally change for built-up areas. There will be a higher number of evacuees which, however, mostly goes in line with the possibility to reduce the length of the escape routes for these evacuees by selecting suitable vertical shelters nearby. Figure 2 shows a generalized scheme of evacuation planning.

The decision to declare a scenario instance as a valid instance (suitable to be taken up within a tsunami evacuation plan) depends on whether all affected persons will have arrived at ‰ÛÏtheir‰Û safe location within the given time span. In the contrary case, those parameters should be changed in such a way that an improvement (with respect to full evacuation) can be measured within a simulation repetition. Practically, this means to reduce the time-to-shelter. This could be partly done through the selection of better escape routes and/or through an increase of the throughput (of persons) along roads and/or bottlenecks; however, in most cases this would mean to increase the number of available shelters, in particular the number of vertical shelters.

Damage Scenarios

There are many objects along a coastline that may be hit by a tsunami. Among the most important ones in terms of evacuation measures are lifelines, such as roads usable as escape routes, and buildings, usable as additional vertical shelters. The flow of debris brought forward and left behind by tsunami waves is also of importance as evacuation may be tampered and rescue operations may be hindered dramatically.

The intensity with which buildings become affected with depends on several factors (in decreasing order of importance):

‰Û¢ The intrinsic resistance of a single building (due to its structural characteristics);

‰Û¢ The wave height hitting the building;

‰Û¢ The hydrostatic forces in the moment when hitting the walls of a building

‰Û¢ The proximity of the building to the shoreline;

‰Û¢ The particular location of the building giving its exposition to the waves (e.g. if in front of the sea or not);

‰Û¢ The particular location of the building with respect to be hit by debris;

‰Û¢ The total number of waves and backwash events hitting a building;

‰Û¢ The duration of submersion.

Table showing classes of buildingsAs a calculation that takes into account all these parameters could become extremely complex, a more pragmatic approach could be elaborated. Hence in most cases the direct damage to buildings is a function of wave height (water column hitting a building) and construction type of a building [Peiris 2006; Leone et al., 2010; Valencia et al. 2011].

The best way to make certain features (like building types, building height, potential damage to buildings, inundated area, and many more) visible is through the creation of thematic maps; on that basis a decision maker gets a variety of tools to further assess a local situation in a case-by-case manner. As such, one could check the explicit exposure of each single building, and/or one could select those buildings that have sufficient height above the expected tsunami wave height. On the other hand, damage level calculations could be done by considering the building type and the inundation depth thus having only two, but nevertheless strong parameters to reflect on.

Several vulnerability classes of buildings have been worked out. As typical building typology, the one proposed by [Leone et al. 2006], has been taken as basis and extended in the SCHEMA project. Four main classes (plus sub-classes) of buildings have been identified (see Table 1) mentioning two parameters: construction type and number of stories.

Table showing damage levels on buildings, their usability as shelters and the possibility to detect such damages through earth observationBeyond a building classification, it is useful to define levels of damage to buildings on a scale which has been adapted from [Peiris, 2006; Leone et al., 2006 and 2010]; it is summarized in Table 2, mentioning the damage (on a qualitative scale), the potential use as shelter, and the possibility to identify such a damage through earth observation. The latter, though not being of primary importance for the tsunami evacuation planning, could be of use during a first range of response actions by identifying those buildings that could have hosted evacuees.

In parallel, fragility curves (for each building type) have been generated within the SCHEMA project [Valencia et al., 2011] that allow attributing an expected damage level (Table 2) to buildings depending on the water elevation. These functions have been developed empirically from a database compiled in the southwest area of Banda Aceh (Indonesia) being hit by the 2004 Indian Ocean tsunami (see Figure 3). The following damage matrix (see Table 3) has then been extracted from fragility curves (for building classes A ‰ÛÒ E1).

Figure 4 shows the application of the Table 3 values to the concrete case of Rabat (based on scenarios worked out for that region [Tinti et al., 2011]).

On basis of the matrix of Table 3 and on basis of the to-be-expected tsunami wave height, a decision maker has sufficient means to create an inventory of buildings suitable to be included in the list of additionally to be considered vertical shelters.

Figure showing fragility curves for building types

Figure 3: Fragility curves for building types A – E

Inventory of Vertical Shelters & Tsunami Damage Maps

The inventory of vertical shelter buildings provides a good basis for an optimization of evacuation plans. However, especially for inhabited and built-up areas, more aspects to consider are of great value. Such areas, for example, pose the risk that many other objects are hit by tsunami waves thus creating additional problems.

In a first category, one can include all those objects that can easily be mobilized and carried away by tsunami currents, to avoid huge amounts of debris that can significantly increase the destructive power of atsunami. Main elements of debris are vehicles on land, yachts and boats in harbors and many other non-fixed objects on streets and pedestrian walks.

Moreover, the vulnerability of lifelines (gas, water, electricity, waste water, power stations, telecommunication, etc.), rescue-relevant objects (hospitals, fire brigades, emergency equipment, etc.) and response-related objects and networks (airports, main roads, airports, bridges, etc.) have to be taken into consideration. Though quantitative assessment of tsunami damage to these elements could become quite complex, a qualitative assessment could at least be of importance with respect to the feasibility of an evacuation plan. All results should be kept within damage maps indicating potential accumulation of debris and/or potential damage to critical infrastructures.

Table showing damage matricesAs defined in the SCHEMA project, damage maps could also be enhanced by a quantitative assessment of the wave velocities (at each location) as an additional outcome of the tsunami hazard scenario; alternatively, if possible, a qualitative assessment could be made for each location considering its exposure to the waves. In that sense, for example, a shoreline vertical shelter must demonstrate a higher resistance than others, and a distant-located class-B building that gets just inundated but not really hit, may not really collapse.

The inventory of vertical shelter buildings may again be checked, often on a case-by-case approach, deselecting those constructions that may encounter heavy impacts due to debris flow and other concerns. One characteristic here is that ‰ÛÒ after a tsunami ‰ÛÒ debris may block access to (and exit from) that building. Another aspect for deselecting could be the neighborhood, to potentially damaged gas or electricity lines.

Similar considerations should anticipate potential situations after a tsunami has struck:

‰Û¢ appreciate the importance of access roads as these should be made usable as soon as possible. In that sense, a road classification could be made available within a thematic layer;

‰Û¢ appreciate important facilities that should be made usable in an easy way; necessary emergency equipment must be accessible promptly, and airports are among the first-order facilities to be used for response actions. The basis of any of these reflections is the map that shows the flow direction and extent of inundation — as it can be presumed that the amount of debris left behind is higher along the inundation boundaries or in bottle-necked points (or in general where velocities of water are expected to decrease under certain thresholds).

Thematic map of damages to building and approximate to-be-expected wave heights.

Figure 4: Thematic map of damages to building and approximate to-be-expected wave heights. Courtesy of CRTS (Morocco) and ACRI-ST (France)

The final list of potential vertical shelter buildings provides suitable options to use for further improving a tsunami evacuation plan. Finally, a filed survey is necessary in order to check and list the potential difficulties of access to those buildings (e.g. due to the presence of security codes or locked doors). Hence the location of those vertical shelters should guarantee 1) a minimum of accessibility during the post-tsunami operations, 2) that the total time of occupancy (during the inundation) is kept to a minimum, and 3) that for the time of the occupancy the evacuees have access to a minimum of lifelines.

Optimization of evacuation maps

A valid instance of a tsunami evacuation plan is a scenario that allows evacuating all affected people toward safe locations in time. Safe locations are by definition situated outside the inundated area; however, with the availability of additional vertical shelters, it is advisable for inhabited regions to augment the number of nearby shelters thus reducing the time for evacuation and making evacuation less complex.

Prior to the selection of vertical shelters, an evacuation plan may be a valid instance; in that case the selection of additional vertical shelters may improve the performance of an evacuation plan. On the other hand, if an evacuation plan does not demonstrate that all affected persons could evacuate in time, and that all other parametric options like choosing appropriate escape routes have been considered, the selection and inclusion of additional vertical shelters becomes mandatory.

In either case, the role of local decision makers is to choose suitable buildings from the list of vertical shelters. On the basis of additional shelters, the evacuation simulation procedure should be launched again and eventually produce satisfying results. Hence a valid instance of an evacuation plan is created step by step, including more and more vertical shelter buildings into the plan [Scheer et al., 2011].

Other constraints, obviously, apply when reflecting on the number of persons that a vertical shelter building may host; moreover such a building should also provide with sufficient sanitary facilities in order to provide acceptable shelter for the time of the flooding.

Cases for which the inventory of vertical shelters was not sufficient have been reported as well. It could happen that none of these buildings is close enough to beach areas from which ‰ÛÒ depending on the season ‰ÛÒ a significant number of persons has to be evacuated in (short) time. On Okushiri Island (Japan), for example, artificial elevated platforms had been constructed thus providing appropriate nearby shelter for beach tourists [UNESCO-IOC]. This can help also in case of a short distance tsunami (i.e., generated from a near source) which could strike the area of interest only few minutes after its generation — not allowing a tsunami alert to be triggered, or leaving a too short time span for authorities to activate the evacuation plan.

Mid-term maintenance of an existing evacuation plan consists in constantly checking the availability as well as the accessibility (including the escape routes) of horizontal and vertical shelters. Among the many preparedness tasks for authorities, there is in particular the proper training of residents and proper education of specific parts of the population (children, elderly, handicapped, etc.), on evacuation measures on top of well-elaborated instruction and divulgation of the existing evacuation plan.

Long-term maintenance consists of counterchecking an existing plan against its acceptance within the population in addition to the postulation of changes within the basic parameters resulting from the tsunami hazard scenario. The latter, in particular, could easily make an evacuation plan obsolete which, in turn, requires a restart of the whole evacuation plan generation procedure.


Evacuation plan generation should consider, on top of local topologies, the outcome of tsunami hazard analysis. This basically means to overlay the expected maximum inundation zone with the topology and to set a basic time constraint for a full evacuation to take place. A secondary step overlay is produced with the topology of buildings and constructions taking into account the expected inundation depth for each specific location. From this overlay a number of vertical shelter buildings could be derived that may serve as input for a simulation and optimization procedure.

Within a simulation procedure it can be checked whether the affected population can evacuate toward shelters (horizontal shelters located outside the inundation zone; vertical shelters located within the inundation zone but with sufficient height to provide with safe areas) within the maximum of time given by the basic time constraint. Within various optimization steps, each simulation procedure can be fine-tuned by adding suitable vertical shelters to the map of shelters within the evacuation map.

Those vertical shelters buildings are kept on a list of suitable vertical shelters, which is one of the outcomes of a damage scenario computation. The main factors to consider herein are hydrological forces, brought forward by the energy of the arriving waves and the expected wave height onto the structure of each building. With a classification scheme of buildings and with a damage level classification scheme, a damage matrix is derived indicating for each building class the minimum and maximum wave heights necessary to create a certain damage (according to the damage classification scheme).

The proposed damage matrix serves as input for decision makers who, on basis of the expected inundation depth, select those buildings that could withstand a to-be-expected tsunami and which could serve as additional vertical shelter buildings.


Leone F., Denain J. C., Vinet F. and Bachri S. (2006). Analyse spatiale des dommages au bÌ¢ti de Banda Aceh (Sumatra, IndonÌ©sie): contribution ÌÊ la connaissance du phÌ©nom̬ne et ÌÊ l’Ì©laboration de scÌ©narios de risque tsunami. Scientific report of Tsunarisque (2005-2006) programme.

Leone F., Lavigne F., Paris, R., Denain J.-C., Vinet F. (2010): A spatial analysis of the December 26th, 2004, tsunami-induced damaged: lessons learned fro a better risk assessment integrating buildings vulnerability. Applied Geography, doi:10.1016/j.apgeog.2010.07.009.

Nagao I. (2005): Disaster Management in Japan, Fire and Disaster Management Agency (FDMA), Ministry of Internal Affairs and Communication, Presentation as of 28/02/2005, Japan.

Peiris, N. (2006): Vulnerability functions for tsunami loss estimation. First European Conference on Earthquake Engineering and Seismology (a joint event of the 13th ECEE & 30th General Assembly of the ESC), Geneva, Switzerland. Paper number 1121.

Scheer, S., Eftichidis, G., Guillande, R. (2010): A Generic Framework for Tsunami Evacuation Planning, European Geosciences Union (EGU) General Assembly, Vienna, 2 ‰ÛÒ 7 May 2010.

Scheer, S., Gardi, A., Guillande, R., Eftichidis, G., Varela, De Vanssay, B., Colbeau-Justin, L. (2011): Handbook on Tsunami Evacuation Planning. SCHEMA project no. 030963, EUR 24707.

Tinti, S., Tonini, R., Bressan, L., Armigliato, A., Gardi, A., Guillande, R., Scheer, S. (2011): Handbook on Tsunami Hazards and Damage Scenarios, SCHEMA project no. 030963. EUR 24691.

UNESCO-IOC, http://ioc3.unesco.org/itic/printer.php?id=20

Valencia, N., Gardi, A., Gauraz, A.L., Leone, F. and Guillande, R., (2011). New tsunami damage functions developed in the framework of SCHEMA project: application to Euro-Mediterranean coasts, Nat. Hazards Earth Syst. Sci., under revision.


SCHEMA (Scenarios for Hazard induced Emergencies Management) was a FP6-funded research project worked out in the period 2007- 2010 by a consortium of 11 partners led by Geosciences Consultants.

The authors:

Stefan Scheer is a senior research at the European Commission’s Joint Research Centre Ispra (Italy). He has a university degree in computer science, software engineering and economics from the Technical University of Munich (Germany). He was involved in projects related with lessons learning from natural disasters and general principles of disaster reduction.

Richard Guillande holds a Phd in structural geology. He has been post-doctorate researcher on application of remote sensing to natural disasters for the French National Space Center (CNES). Since 1990, he has been manager of Geosciences Consultants (Paris, France). He has developed an eclectic knowledge and practice of assessment and mitigation of natural disasters and has been leading various national and international projects related to Disasters Risk Reduction. He was coordinator of SCHEMA project from 2007 to 2010.

Annalisa Gardi has been a researcher engineer at Geosciences Consultants (Paris, France) since 2006. She holds a PhD in Earth Sciences (University of Milan, Italy, and Ecole Normale Sup̩rieure of Paris, France). She is an expert in numerical modelling of seismotectonic processes and works on natural hazard and vulnerability assessment, especially focusing on earthquakes and tsunamis.