Study cases

From Rasor Wiki
Jump to: navigation, search

Contents

Haiti

The January 2010 earthquake killed over 300,000 people, injured another 300,000 and left over a million people homeless. This disaster struck an already fragile country, and has left populations more vulnerable than ever to flooding, landslides and other natural disasters common in the hurricane-prone Caribbean. In 2004, Haiti was struck by two devastating hurricanes, and then 3 more in 2008. Again, in 2012, Sandy hit the Haitian Capital.

(thumbnail)
Earthquake in Haiti (Image courtesy of Wikimedia Commons).


The city of Gonaïves is the capital of the Artibonite department of Haiti. It has a population of about 300,000 people. It is situated on a flat plain on the coast, with a small river La Quinte running past it. During most of the year, all the water in this river is used for irrigation. However, after heavy rainfall, the small La Quinte stream can swell and overflow its banks to run through the surrounding lands including the city of Gonaïves. Moreover, if at the same time, the sea water level is high, the water accumulates near the coast and the city is severely flooded. In September 2004, Hurricane Jeanne caused major flooding and mudslides in the city. Four years later, the city was again devastated by another storm, Hurricane Hanna, which again flooded parts of the city and killed 529 people.

(thumbnail)
Gonaïves and surroundings. The La Quinte River runs from the mountains on the top left past the city and past Nan Piguel before it flows into the Baie des Gonaïves on the far right of this picture. Source: Google Earth.


Exposure

Buildings

The most of building footprints are derived by Open Street Map Database (see Figure below).

(thumbnail)
Portion of buildings Gonaives, Haiti.

Population

The data presented below represent estimates the numbers of people for hectare ('pph'), with national totals adjusted to match UN population division estimates (http://esa.un.org/wpp/) and remaining unadjusted.

(thumbnail)
number of people for hectare.

Land use

The below polygon vector layer shows Haiti's landcover, with a global legend in 14 classes. It was created from the raster data of the Global Land Cover Facility (http://glcf.umiacs.umd.edu/index.shtml), from the AVHRR satellites, with data acquired between 1981 and 1994.

(thumbnail)
Haiti's landcover.

Hazard

Gonaïves, Haiti - Hurricane storm surge and rainfall runoff flood model

A modelling chain (see Figure below) was developed to simulate storm surge, rainfall runoff and flooding of the city of Gonaïves, as a result of a hurricane storm passing by. The end user defines a hurricane track by a series of coordinates of the eye of the storm and a maximum wind velocity. This information is transformed into a time dependent wind field by WES and into a rainfall field by the R-CLIPER module. The wind field is input to a Delft-3D storm surge model of about 1500 by 1500 km (resolution of 5 to 6 km, bathymetry from GEBCO2008). The rainfall field is input to a Wflow hydrologic model, which covers the watershed of the La Quinte River and neighbouring streams at a resolution of 100 m. The DEM for this model was derived from TanDEM-X.

(thumbnail)
Gonaïves modelling chain.
(thumbnail)
Gonaïves Wflow model maps at (from top left, clockwise): Wflow subcatchments, surface flow, streamorder and DEM.

A SubGrid model was developed for the city of Gonaïves and its surroundings (see Figure) to simulate overland flow and flooding of the city. The discharge in the La Quinte River and some smaller inflows as calculated by the Wflow hydrologic model are the upstream boundary of this model. The sea water level time series from the Delft-3D storm surge model is the downstream boundary condition. The computational cells of the SubGrid model are between 25 and 800 m. A higher resolution is used for the flood prone urban areas and a lower resolution is used elsewhere for computational speed.

(thumbnail)
Gonaïves SubGrid model for simulation of 2D overland flow. The cell size varies between 25 and 800 m.

The Haiti model chain calculates the flood pattern after a hurricane whose track and intensity are defined by the user. Below is an example of a flood pattern in Les Gonaives after a hurricane that followed the same track as Jeanne in 2004. However, it is important to note that the wind and rainfall fields derived from this track are not the same as the observed wind and rainfall during the 2004 event. The WES and R-CLIPER models generate a realistic but strongly simplified hurricane model. A real storm will always deviate from this idealized storm. Nevertheless, the flood pattern as computed by the modeling chain agrees with the SPOT image (see Figures below). The La Quinte River overflows on the west bank and flows in South-West direction through the city.

(thumbnail)
Les Gonaives inundation map. The La Quinte River runs from North to South just right of the centre of the map. The coastline is on the left.
(thumbnail)
Les Gonaives inundation map from SPOT imagery. The La Quinte River runs on the right side.

Italy

The recent seismic phenomena occurred in May 2012 in Emilia-Romagna Region, in Northern Italy, lead to increased hydraulic risk in the affected areas. Drainage structures concerned are plants and hydraulic structures as well as levees and bridges. These territories are topographically under sea level (e.g. Ferrara Plain) or under levees/canals level, so they can be flooded due to a non-regulated drainage water-management caused by structure damages after an earthquake event. In 2012 the structure experienced significant damage: in one case, Mondine system (near Moglia, MN), the damage is so important that it’s difficult to restore its function in the short term so that there is a long term increase in flood risk. Agricultural purposes are the predominant land use. The average population density is about 100 inhabitants/km². The average height in the area varies between 17 m.a.s.l. on the western boundary to 5 m.a.s.l. at the eastern end. Dikes with heights of up to 10 m along the Po River protect the adjacent area against flooding from the river. Additionally, several storage areas between the main river channel and main dike line are available for a reduction of the water level in case of a flood event.

Exposure

Buildings

Emilia

Are partition of the buildings obtained on the basis of different types of buildings. The primary source of the partitions is CTR5. The alternative sources are either vector or digital mapping layers, that meet the requirements necessary for the DBT, available at the municipal SIT (see figure below).

(thumbnail)
Model Partition of buildings situated in Emilia.
Po River

Building exposure layer along the Po River derived from OSM and enhanced by SERTIT (see figure below).

(thumbnail)
Building along the Po River.

Land use

Emilia

This layer are derived by Corine Land Cover. The project Corine Land Cover (CLC) was born in Europe specifically for the detection and monitoring of hedging characteristics and use of the territory, with particular attention to the needs of environmental protection.

(thumbnail)
Emilia's land use.

Hazard

Po-Secchia - riverine levee breach model

Within the Po river case study, the following model software was applied:

  • the reliability transformation tool to determine the location and point in time of breaching at potential breach locations and
  • the 3Di-subgrid software for the two-dimensional hydrodynamic calculations.

Reliability transformation tool This module transforms a water level time series at a certain location of the flood defense line into a time series of failure probabilities and – in case of a failure – into a breach development over time. The transformation of water levels to time series of failure probability is done by using a fragility curve. The fragility curve expresses the reliability of a structure as a function of a defined dominant stress variable, e.g. water level at a dike section. This transformation is performed for each water level over time hw(t).

(thumbnail)
Reliability transformation with a fragility curve. The probability of failure increases for increasing water level. The probability of non-failure is 1 minus the probability of failure.

According to the time series of failure probability, the simulated water level time series and a user-defined probability threshold for each dike section a breach development is calculated. By exceeding the defined probability threshold a breach development is started in the dike section. The user can interactively change these thresholds to define different breach locations and starting times (what-if scenarios).

The breach depth development is not modeled; a defined sill height is applied (e.g. adjacent elevation to the dike). For the calculation of the breach width three different approaches are available within the tool (only the first option is used in RASOR at the moment):

  • Instantaneous breaching with a defined maximum breach width.
  • Linear breaching with a defined maximum breach width and a breach growth rate; if the water level is lower than the defined sill height the breach growth will stop.
  • Adapted semi-empirical breach growth model after VERHEIJ with a defined maximum breach width and a critical velocity representing different dike materials.
(thumbnail)
Model set-up of the Po river test case. The orange bullets indicate potential breach locations. Dike sections and dike lines are indicated by arrows.

Six dike sections were selected as potential breach locations. They are all located at the south bank of the river in the middle part of the model area. The two-dimensional model domain covers about 1.300 km². It includes about 85 km of the Po River as well as the adjacent area south of the river. The resolution of the SubGrid computational elements varies between 144 and 576 m, according to the variability of the local variations in the DEM. The underlying subgrid raster has a resolution of 12 x 12 m and is based on TanDEM-X data.

The end-user of the RASOR platform can set the boundary conditions as a maximum discharge or a time series at Borgoforte and the breach locations. The downstream boundary condition is calculated automatically by a stage-discharge relationship. Below is an example flood pattern after a dike breach at Pieve di Coriano. The breach location is indicated by an arrow.

(thumbnail)
Po river inundation map. Dikes are indicated as red lines, breach location by an arrow.

Indonesia

Bandung

Bandung is the capital of West Java province in Indonesia and Indonesia's third largest city, with a population of 2.4 million. It is located 768 m above sea level and is surrounded by volcanic mountains. Regular flooding in Bandung presents a real and dangerous ongoing problem. The areas south of the city center are most prone to flooding from the Citarum River and its tributaries.

Cilacap

The town of Cilacap is a sea port on the southern coast of the island of Java. The port is accessible to relatively large ships, making it an important cargo hub. The South coast of Java is prone to tsunami hazard because of the Alpide belt, a seismic active zone that extends along the southern margin of Eurasia, stretching from Java to Sumatra through the Himalayas, the Mediterranean, and out into the Atlantic. The Alpide belt is the second most seismically active zone in the world, with frequent earthquakes, volcanic eruptions and tsunamis in Indonesia and the surrounding areas. Although Cilacap is relatively protected from tsunami impact by the Island of Nusakambangan, the 2004 tsunami took 147 lives, devastated beaches, damaged 435 fishing boats and inflicted material losses amounting to about Rp 86 billion (around $9 million).

Jakarta

Jakarta is the capital and largest city of Indonesia with a population of more than 10 million. It is located on the northwest coast of Java, in a low, flat basin, at elevations between −2 to 50 meters above sea level. About 40% of the city is below sea level and is prone to coastal and fluvial flooding. Moreover, Jakarta is sinking at a rate of 5 to 10 cm annually and even more in the coastal areas. There are plans to build a dike around Jakarta Bay, which will be equipped with a pumping system and retention areas to defend against seawater.

Exposure

Buildings

The most of building footprints are derived by Open Street Map Database. Before, database have been downloaded in SERTIT computers, then the attributes were analyzed and regrouped in order to fit the Rasor nomenclature. The comparison between the Dajarta test site, the Bandung and the Calicap one well illustrated the heterogeneity of OSM database. For example Cilacap is much more detailed and documented than Bandung.

(thumbnail)
OSM buildings over Bandung, Java, Indonesia, integrated within the platform, 21 189 buildings are documented
(thumbnail)
OSM building over Cilacap, Java, Indonesia integrated within the Rasor platform, 12390 building are individualized, part of them are annoted
(thumbnail)
OSM building extracted from OSM database and load within the Rasor Platform, 288 665 elements for all Jakarta AOI area

Land use

Land cover information derived from remote sensing data has been identified, and are has been partially downloaded into the razor palteform.

(thumbnail)
Giava's land use.

Hazard

Bandung-fluvial flood model

A Wflow hydrologic model at 250 m grid resolution was developed to simulate the rainfall runoff from the mountains that surround Bandung. The DEM for this model was derived from SRTM90. This runoff is than later used as input to a smaller area SubGrid 2D flood model (see below).

(thumbnail)
Bandung wflow model DEM.
(thumbnail)
Bandung river network.

The end-user can define uniform rainfall scenarios through the RASOR platform web interface or select a historical period of TRMM rainfall. The user can also select an initial soil moisture condition (dry, medium or wet) that represent respectively typical conditions during the dry season, annual average and conditions during the wet season (December-January).

The river flows in several rivers and larger streams as computed by the Wflow model are ingested into a smaller area SubGrid flood model (see Figure) at a varying grid resolution of 100 to 800 m. The water depths from this model are later downscaled to 50 m based on the TanDEM-X DEM. Several versions of the model were made to represent subsidence scenarios for 2010, 2020, 2050 and 2100. The user can select either of these models through the RASOR web interface to run a simulation for a particular subsidence scenario.

(thumbnail)
Bandung SubGrid model reference DEM (2010) and grid.
(thumbnail)
Regular grid DEM at 50 m resolution for downscaling of the Bandung SubGrid model water depths.

The Bandung SubGrid model calculates the flood pattern caused by overflow of the Citarum River. Below is an example flood extent map after intense rainfall in December 2014.

(thumbnail)
Bandung SubGrid model reference DEM (2010) and grid.

Cilacap tsunami model

The RASOR case study allows the end-user to define a tsunami wave height off the coast and calculate the inundation depths and flow velocities in and around the city of Cilacap. An empirical model called FAST calculates the wave runup over transects that run from offshore to onshore locations (see Figure). From each of the input point a number of transects and their bathymetry/elevation profiles are set. For each of this transect the empirical relation is applied to determine the flood depth based on the bottom gradients. Subsequently, the maximum calculated flood depths at each grid point of the gridded DEM are transferred to a finer resolution (in this case 100 m).

(thumbnail)
FAST model concept of offshore points (left) and wave runup calculated over transects (right).

The Flooding ASsessment of Tsunami (FAST) tool uses relatively simple, empirical expressions that relate tsunami wave height on the coast to the run-up onto linearly sloping land (Blaas et al, 2008). The FAST model requires only the bathymetry and DEM for a given area to run. For Cilacap, the bathymetry was taken from 1km gridded GEBCO2008. Two versions of the model were made, one based on SRTM DEM and one on the TanDEM-X intermediate product (see Figure).

(thumbnail)
Bathymetry and TanDEM-X DEM of the FAST model in Cilacap. Depths are in m.

For the determination of the risk, the (maximum) velocity estimates on land are also required, which is not calculated by the original FAST model. Therefore, a number of methods to estimate the maximum runup velocity from literature were reviewed. The method used by Carrier and Greenspan (1958) and Cousins et al (2007) was found most applicable to narrow bays/channels as the relation is derived using linear approximation of 1D equation (among others) for sine tsunami wave.

Tsunami wave heights for a number of return periods can be found in the table below. An example inundation pattern after a 2.1m tsunami (return period 100 years) is shown in the Figure below.

Tabella flo wd 2.1.png
(thumbnail)
Cilacap maximum flood depths for a wave of 2.1m (flood depths in m).

The results from FAST were compared to observed tsunami levels. Tsuji et al. (2005) lists observed run-up and tsunami levels for the 1994 tsunami for several locations. The simulation results show a good overall agreement to these observations (see Figure), expect for one outlier (20 m water depth) that may have been caused by a local funneling effect. The SRTM and TanDEM-X versions of the model perform more or less equal, with SRTM reproducing the observed inundation depths slightly better for the western locations.

(thumbnail)
Cilacap locations (top), runup distance (middle) and flood depths (bottom). Observed vs simulated.

Jakarta - coastal levee breach

Jakarta is the capital and largest city of Indonesia with a population of more than 10 million. It is located on the northwest coast of Java, in a low, flat basin, at elevations between −2 to 50 meters above sea level. About 40% of the city is below sea level and is prone to coastal and fluvial flooding. Moreover, Jakarta is sinking at a rate of 5 to 10 cm annually and even more in the coastal areas. There are plans to build a dike around Jakarta Bay, which will be equipped with a pumping system and retention areas to defend against seawater.

A SubGrid model was developed for a coastal area of Jakarta called Pluit. The model has a variable grid resolution of 50 to 800 m, but the water depths that are calculated by this model are downscaled to 50 m based on the TanDEM-X DEM (see Figure). Several versions of the model were made to represent subsidence scenarios for 2010, 2015 and 2030. The RASOR end-user can define a sea water level time series, select a subsidence scenario, run a simulation and view the results or use them in a risk assessment.

(thumbnail)
Jakarta DEM at 50 m resolution based on TanDEM-X. Green colours indicate land below mean sea level.
(thumbnail)
Jakarta DEM at 50 m resolution based on TanDEM-X. Green colours indicate land below mean sea level.
(thumbnail)
Jakarta flood map at 50 m resolution.

Greece

The island of Santorini in the Aegean Sea is subject to seismic hazards from several submarine faults in the caldera and surrounding sea bed and from volcanic activity. An earthquake can cause direct damage to buildings and indirect damage from tsunamis that are caused by landslides from the cliffs of the caldera rim. The tsunami waves travel across the caldera and can propagate even to the outer shores of the island. This forms a flood hazard to the population of the low-lying areas and to cruise ships that are often moored in the Caldera. The travelling time of the tsunami wave to the most vulnerable locations is a few minutes, which probably too short to issue a warning. This case study can help to improve risk awareness and support contingency planning.

Exposure

Buildings

The building footprints are derived by Worldview2 (see Figure below).

(thumbnail)
Santorini Buildings.

Land use

Land cover information derived from Globcover 2009 landuse/landcover converted in CLC code.

(thumbnail)
Santorini land use.

Hazard

Santorini - landslide-induced tsunami model

To accurately model the process of a landslide causing a tsunami wave requires sophisticated numerical models connected in a complex way. An example is given by Tinti et al (2006) who simulate a submarine landslide and tsunami near Stromboli (Italy). To develop such a model for Santorini is beyond the scope of this RASOR case study and would require too much computing time for rapid risk assessments. Instead, we have adopted a simplified approach for the tsunami source by making a few assumptions on the wave shape, length and height of the tsunami approximately 60 seconds after the slide impact (see figures below from Tinti, 2006). We assume the initial wave to be radially symmetrical from the source point specified. In this approach it is assumed that the tsunami wave can be treated as long wave.

(thumbnail)
Tsunami wave pattern and N-wave shape 60 seconds after slide (from Tinti et al, 2006).

To simulate the tsunami wave propagation in the waters in and around the Santorini caldera we apply Delft3D-FLOW, which is a hydrostatic non-linear shallow water solver that calculates non-steady flow and transport phenomena that result from tidal and meteorological forcing on a rectilinear or a curvilinear, boundary fitted grid. For RASOR, only the 2D functionality is used. The model area is about 50 by 60 km. A grid resolution of approximately 75 m is used.

Bathymetry information near the main islands has been obtained from the Greek Institute of Geodynamics. Missing data in this map near the Santorini coastline (see Figure) were filled with a minimum depth value of 20m and interpolated using the available data. Outside the area of data coverage, the existing data is interpolated towards the freely available global SRTM data.

(thumbnail)
Bathymetry of the area (source Institute of Geodynamics) and the Delft3D model bathymetry.

In general, a tsunami wave generated by a slide will contain smaller wave lengths, i.e. wave lengths that are in the same order of magnitude to the local depth value. The dynamics of these waves will not be represented properly on the coarse grid that we are using and not solved properly by a hydrostatic, non-linear shallow water solver. This would require a very high resolution model that has the capability to resolve the wave dispersion (i.e. either a code that is capable of handling non-hydrostatic terms or a Boussinesq model). The tsunami wave (heights) as computed in our case study will slightly be overestimated. Given the objectives of this study, i.e. rapid risk assessment, this is deemed acceptable.

The model was implemented into the RASOR platform. Through the RASOR web interface, the user can select an arbitrary landslide location and initial water perturbation.

To set a realistic initial water perturbation, a literature search was done to find a relationship between landslide characteristics and initial wave parameters. Several (related) studies were found that use a general parameter P to characterize sub aerial landslide tsunamis (Fritz et al, 2003; Fritz and Hager, 2010; Heller and Hager, 2014). This approach is simple and straightforward and is therefore very convenient for this study. However, it should be mentioned that many other equations exist that relate landslide characteristics to tsunamis (e.g. Law and Brebner, 1968; Papadopoulos and Kortekaas, 2003; McAdoo and Watts, 2004). This highlights the fact that the process is complicated and a simple and universal expression for landslide tsunamis is not readily available. Moreover, the parameter P was derived from landslide observations which were much larger than the one listed in the example here (e.g. the Lituya Bay landslide of 1958, which had an estimated mass of kg). This could mean that this equation is not particularly suitable for smaller landslides, which in turn could explain the relatively small value that was found for the wave period in the example above. It is important to stress that the user is not bound by the results of the equation and is free to enter other values describing the initial wave for the model.

The parameter P for subaerial landslide tsunamis is defined as follows:

P=V_s \cdot g^{-1/2} \cdot h^{-3/2} \cdot s^{1/2} \cdot (\frac{m_s}{p_w \cdot b_s})^{1/4} \cdot [cos(\frac{6}{7}) \cdot \alpha]^{1/2}

Where:

Vs Slide impact velocity [m/s]
ms Slide mass [kg]
bs Slide width [m]
s Slide thickness [m]
α Hill slope angle [°]
h Still wather depth [m]
pw Water density [kg/m³]
g Gravitational acceleration [m/s²]

From this parameter, P, all relevant wave parameters can be calculated:

am = (4 / 9)P4 / 5h

Hm = (5 / 9)P4 / 5h

Tm = 9P1 / 2(h / g)1 / 2

With, aM, being the maximum amplitude, [m], Hm the corresponding height, [m], and, Tm, the corresponding period, [s];

(thumbnail)
Schematization of a subaerial landslide tsunami with all relevant parameters.

An example scenario was created for a landslide near Imerovigli, which is the highest point on the caldera rim. The landslide properties are:

Vs [m/s]
ms 1 \cdot 10^6 [kg]
bs [m]
s [m]
α [°]
h Still wather depth [m]


These values lead to an impulse product parameter of P=0.12. The initial wave amplitude and period are respectively 8 meters and 10 seconds. This initial wave was entered as a starting condition for a Delft3D-FLOW simulation (see Figure).

(thumbnail)
Initial water level perturbation, invoked by a landslide near Imerovigli.

The wave propagation was simulated for 20 minutes. The computing time is about 30 minutes. The Figure shows an example of the wave pattern two minutes after the landslide. The wave height increases to 1 m or more in small inlets of the coastline that act as funnels. Within the caldera, the arrival time at any locations is below 3 minutes. This leaves hardly any room for early warning or evacuation.

(thumbnail)
Santorini wave pattern about 2 minutes after the landslide. The location of the landslide is indicated by a red arrow.
(thumbnail)
Wave arrival times in minutes.

The Santorini model provides an insight into the flood risk from tsunami waves caused by landslides that can occur on the inner slopes of the caldera. The model is able to estimate the location of the high waves and the arrival time with respect to the location of the landslide. It has a potential for early warning and emergency applications.

Rotterdam

The area of Rotterdam is located in the western part of the Netherlands. Approximately 1.3 million people live in the greater Rotterdam area with an average population density of about 3000 Inhabitants/km². Housing, industrial and agricultural purposes are the predominant land use. Rotterdam is situated at the delta of the Rhine and Maas rivers, about 35 km from the North Sea. Due to this location in a deltaic area and an average elevation of about MSL, Rotterdam is a highly flood prone area with a very high flood impact potential. Coastal and riverine floods or combined events are probable. However, large flood protection structures, like the Maeslant Barrier which protects the Rhine Meuse Delta from the North Sea in case of extreme surges and the high protection standards for the Dutch dike rings (up to a 10.000 yearly flood event), reduce the flood risk to a low level in this area.

(thumbnail)
Model set-up of the Rotterdam test case.

Exposure

Buildings

The building footprints are derived from Open Streep Map Database.

(thumbnail)
Building exposure layer over the Feyenoord derived from OSM and enhanced by SERTIT.
(thumbnail)
Building exposure layer in Zwijndrecht.

Land use

Land cover information derived from Globcover 2009 landuse/landcover converted in CLC code.

(thumbnail)
Rotterdam's Land use.

Hazard

Rotterdam - Storm surge and levee breach model

Within the Rotterdam case study, the following model software is applied:

  • a Rotterdam water levels script to determine the required boundary condition,
  • the reliability transformation tool to determine the location and point in time of breaching at potential breach locations and
  • the 3Di-subgrid software for the two-dimensional hydrodynamic calculations.

Rotterdam water levels script

For the Rotterdam case study, a SubGrid model was developed that has three water level boundaries: Maassluis on the sea side of the flood prone area and Dordrecht and Krimpen a/d Lek on the east side. The model requires water level time series at these locations as boundary conditions.

The extreme water levels in this area are mainly determined by storm surges at sea. The required water level time series at the three model boundaries are linked to a reference location at the coast: Hoek van Holland. The extreme sea water levels at this coastal station have been studied in depth and the water level exceedance frequency curve at this location is a reference for most flood studies in the area.

The water level time series at the three SubGrid model boundaries are derived in three steps:

The maximum sea water level at Hoek van Holland is entered by the RASOR user. An exceedance frequency curve is available to link this level to a return period. A water level time series at Hoek van Holland is generated by scaling the observed water levels from a November 2007 event to the target maximum level.

The water levels at the three model boundary locations (Maassluis, Dordrecht and Krimpen a/d Lek) are generated from the Hoek van Holland series by a regression model. The last step involves both a water level height scaling and a time-delay of one to a several hours, depending on the location. The regression of observed values showed that a peak water level at the inland locations typically occurs later than at the coast.

(thumbnail)
Exceedance frequency curve for Hoek van Holland water levels.

Reliability transformation tool

This module transforms a water level time series at a certain location of the flood defense line into a time series of failure probabilities and – in case of a failure – into a breach development over time. The transformation of water levels to time series of failure probability is done by using a fragility curve. The fragility curve expresses the reliability of a structure as a function of a defined dominant stress variable, e.g. water level at a dike section. This transformation is performed for each water level over time hw(t).

(thumbnail)
Reliability transformation with a fragility curve. The probability of failure increases for increasing water level. The probability of non-failure is 1 minus the probability of failure.

According to the time series of failure probability, the simulated water level time series and a user-defined probability threshold for each dike section a breach development is calculated. By exceeding the defined probability threshold a breach development is started in the dike section. The user can interactively change these thresholds to define different breach locations and starting times (what-if scenarios).

The breach depth development is not modeled; a defined sill height is applied (e.g. adjacent elevation to the dike). For the calculation of the breach width three different approaches are available within the tool (only the first option is used in RASOR at the moment):

  • Instantaneous breaching with a defined maximum breach width.
  • Linear breaching with a defined maximum breach width and a breach growth rate; if the water level is lower than the defined sill height the breach growth will stop.
  • Adapted semi-empirical breach growth model after VERHEIJ with a defined maximum breach width and a critical velocity representing different dike materials.

Within the dike rings of Pernis and IJsselmonde, 15 dike sections are selected as potential breach locations. The two-dimensional model domain for the hydrodynamic model covers about 500 km². The resolution of the SubGrid computational elements ranges between 100 and 400 m, according to the local variability of the DEM. The underlying DEM has a resolution of 25 x 25 m. The dike rings of Pernis and IJsselmonde are covered by the model domain. To include also the surrounding rivers of the dike rings, the model the area is extended beyond the dike rings. Thus, interactions (e.g. back water effects) between rivers and floodplains are represented in the hydrodynamic model.

The end-user of the RASOR platform can set the maximum sea water level at Hoek van Holland and the breach locations. He can also set the functioning of the Maeslant barrier to either ‘functional’ (will close when the water reaches a level of 3.6m at Rotterdam) or ‘non-functional’ (barrier will not close). The Hoek van Holland water levels for a range of return periods are given in the table below:

return period (years) water level (m+MSL)
10 3.0
20 3.2
50 3.4
100 3.6
200 3.8
500 4.1
1000 4.3
2000 4.5
5000 4.8
10000 5.1

The Figure below shows an example flood pattern after a 4.5m storm surge (T=2000 yrs), a failure of the storm surge barrier and a double levee breach. Breach locations are indicated by arrows.

(thumbnail)
Rotterdam inundation map. Dikes are indicated as red lines, breach locations by arrows.
Personal tools
Namespaces

Variants
Actions
Navigation
Tools