WP 3: Benchmarking - Flood Propagation

F. ALCRUDO, P. GARCIA, P. BRUFAU, J. MURILLO, D. GARCIA, J. MULET Area de Mecánica de Fluidos, CPS-Universidad de Zaragoza, 50.018-Zaragoza, Spain.
alcrudo@posta.unizar.es

G. TESTA, D. ZUCCALÀ CESI SpA, Via Rubattino, 54, 20134 Milano, Italy.
gtesta@cesi.it

1. Introduction
2. Experimental Set Up
3. Benchmarking Program
4. Submission of results
5. Summary of files provided
6. References

Download files

The Model city flooding experiment benchmark is devoted to studying the flow characteristics of flow in urban like environments and the capabilities of different numerical models to accurately represent them. It is embodied in a broader study comprising detailed flow measurement and simulation around model building structures (The isolated building test case by S. Soares and Y. Zech 2002, see below) and mathematical modelling of an actual city flooding event, within Theme Area 3 (Flood propagation) of the EU IMPACT (Investigation of Extreme Flood Processes and Uncertainty) project.

The isolated building test case, based on experiments carried out at the Université Catholique de Louvain, by S. Soares and Y. Zech 2002, addresses the study of local flow characteristics around a building withstanding the arrival a severe dam break wave.

The benchmark described here regards the study of flow arising when such a wave sweeps across the physical model of a city, represented by a certain number of buildings arranged in an ordered pattern as happens in an actual urban area. It is based upon data obtained by ENEL-CESI at its PIS (Polo Idraulico et Strutturale) facilities in Milano, Italy.

Back to top

The ENEL-PIS facilities in Milano comprise a reduced physical model (scale 1:100) of the Toce river valley that has been extensively used for flood propagation experiments (see CADAM project). It is a 50m long concrete model of the river with quite geographical detail fitted with water depth gauges at certain locations. A general view of the model can be seen in Figure 1.

Flooding of the model is achieved by rapidly rising the water level in a feed tank connected to its upstream end by means of an electrically driven pump. The pump discharge and thus the flood intensity can be electronically controlled and recorded.

Figure 2, shows a more detailed view of the area where the model city has been placed. The model buildings can be easily identified as concrete blocks. Also the pump discharge pipe into the upstream feed tank is clearly visible.

Figure 1: General view of Toce river physical model looking upstream (top) and downstream (bottom) from a point located approximately at 3/4 of its total length.
Figure 2: Upstream part of the physical model where the model city has been located (Aligned city lay out).

Experiments have been performed with two different lay outs of the model city. In one, hereafter called aligned, buildings are placed in rows approximately parallel to the main axis of the valley. In the second, hereafter called staggered, buildings are placed in a checker board configuration. Buildings are just concrete cubes of 15cm side.

Furthermore, in order to separate the effects due to the valley bathymetry on the flood wave from those caused soleley by the city, some tests have been carried out in a modified valley model: Two masonry walls have been placed parallel to the model main axis thus providing a chanelling effect. This configuration is shown in Figure 3.

Water level can be recorded at some 10 locations by means of electrical conductivity gauges, that can be seen in Figures 3 and 4. The position of the gauges was chosen in order to observe the flow characteristics at the front arrival line, in the middle of streets and at the buildings wake. For each run, water level is recorded versus time during 60s at 0.2s intervals, hence providing enough time resolution.

Figure 3: Modified valley bathymetry by placement of two masonry walls (Staggered city lay out).

Figure 4: A typical run on the original valley with staggered city lay out. The electrical conductivity probes and their supporting spars are clearly visible.

The location and dimensions of the building are indicated on the sketch. The exact co-ordinates of the four corners of the building regarding the origin of the axis (see sketch) are summarised in the following table:

A Digital Terrain Model (DTM) of the Toce river valley physical model was performed by ENEL-CESI with a 5cm spacing. For this benchmark, only a region about 5m long close to the upstream end out of the 50m full length is used. Regardless of the buildings lay out, experiments have been conducted with two bathymetric configurations as described below:

A) The original physical model of the Toce river valley (see Figures 2, 4 or 5). In this case the bathymetry can be taken from the original DTM file. Since only a small portion of the file is needed, a shortened version of the DTM bathymetry file has been generated, covering only the useful region and its surroundings. This file containing X,Y,Z coordinates is named SHORT_TOCE_ORI.DAT. Each modeller must create his own computational mesh or meshes and then interpolate the bathymetry from that file onto the mesh.

B) In order to simplify the bathymetry and concentrate on the effects of buildings on the flood, the model valley was modified according to the points below:

Firstly two masonry walls were erected along the sides of the central part of the original model valley (see Figure 3). The coordinates (in m) of the edge points of the masonry walls with respect to an arbitrary origin consistent with all the other coordinates given in this document are:

Secondly, the meandering Toce river bed was filled up with concrete. However the DTM original file was not modified accordingly because no new valley survey was performed. Therefore the river bed has been artificially filled on a new, modified DTM file named SHORT_TOCE_MOD.DAT. The actual filling process consisted of running a steady flow model on the original DTM bathymetry with increasing discharge until spilling over the river banks was achieved. Then the new bed function was set to the water free surface.

Finally, while placing the two masonry walls the model valley was extended upstream with a concrete slab a few cm long that protrudes into the feeding reservoir. Hence the data points on the new DTM file had to be extended (extrapolated) upstream a few cm to cover the concrete slab protruding into the feed tank. This has been done on a row by row basis and hence the resulting extrapolated bathymetry looks somewhat grooved on its upstream end. Despite the bad looking appearance this has no influence on the hydraulics because it affects only about one cm at the upstream end.

Hence, in order to run this bathymetric configuration the modeller must generate first his computational mesh bounded by the edges of the masonry walls listed above (see Figures 6 and 7) and then interpolate the bed function (Z coordinate) from the SHORT_TOCE_MOD.DAT file.

The model city is represented by concrete cube blocks of 15cm long edges that have been positioned in either an aligned or a staggered lay out as previously stated. The number and position of the buildings depends on the city and bathymetry lay outs. The different possibilities are explained below.

i) Original Toce river valley model (Bathymetry A)

In case of the original Toce river valley bathymetry, the aligned city lay out uses 20 buildings and the staggered one uses only 18. In fact only the second and fourth rows have been modified. A general sketch of the upstream area of the valley where the model city is located can be seen in Figure 5, below, for the aligned city lay out. Buildings are represented as numbered green squares. Also visible are the probe locations, marked as numbered crossed black circles.

Figures 5: The model city area corresponding to the aligned lay out.

The staggered arrangement is obtained by displacement of the second and fourth rows (buildings 6 to 10 and 16 to 20) parallel to themselves in the vertical direction approximately one block. In order to keep a symmetrical pattern, buildings 10 and 20 are therefore eliminated. The coordinates (Xi,Yi) in m with respect to an arbitrary origin consistent with the rest of this document of the four corners of each building are listed in Tables 1 and 2, below, for the aligned and staggered building lay outs respectively.

The separate file CASE_ALLINEATE_lista_coordinate.doc contains the same information as Table 1 in MS Word text format.

The separate file CASE_SFALSATE_lista_coordinate.doc contains the same information as Table 2 in MS Word text format.


ii) Modified Toce river model valley (Bathymetry B, i.e. channelled with masonry side walls)

In the valley bathymetry channelled with two masonry walls, the space available is narrower and the row of buildings closest to the right wall had to be removed.

Hence the aligned city lay out comprises only 16 buildings resulting after removing blocks 1, 6, 11 and 16 . This can be seen in Figure 6 where the two side walls are depicted as two red lines.

Figure 6: Aligned lay out in the channelled valley geometry.

As regards the staggered arrangement, comprises only 14 buildings resulting after removing also blocks 1, 6, 11 and 16 . This is shown in Figure 7 where the two side walls are also visible as red lines.

Figure 7: Staggered arrangement in the channelled valley geometry

As stated before, the model is equipped with electrical conductivity gauges that are capable of measuring water depth versus time during the flooding experiments. Water depth is recorded at 0.2s intervals for a duration of 60s every run.

The gauges positions can be spotted in any of the previous figures as black crossed circles. As can be easily realized, their location is kept fixed irrespective of the bathymetry or model city arrangement. This is so because of the burden of drilling new holes and passing wires and hardware across the concrete model bed.

The coordinates of the 10 probes with respect to an arbitrary origin consistent throughout this document are tabulated in below.

Probe number 1 is located inside the feeding tank and therefore has no use for benchmarking. However it can be used, if so desired, as an alternative to impose the upstream boundary condition instead of the discharge, or as a cross check.

Gauge number 2 is at the entrance of the model and gives information about the flood intensity entering the valley.

Gauges 3 and 4 are placed just upstream of the first row of buildings and will first detect the arrival of the flood, while probe number 5 is amidst the main or central street. Gauges 6, 7, 8 and 9 are located around building number 13 irrespective of the bathymetry and city lay out.

Finally probe 10 lies at the downstream end of one of the streets parallel to the valley axis in the aligned city lay out or in the wake of building 17 in the staggered one.

Initial condition

The initial state of the model valley will be considered as dry, although no information has been given about the possibility of a very thin film of water or moisture wetting the concrete bed due to previous runs during a session.

Upstream boundary condition

As for the upstream boundary condition, the pump discharge into the feeding tank is known versus time. It can be assumed to a good approximation as the global discharge entering the model valley through its upstream end.

The pump discharge versus time is provided in MS Excel file Inflow.xls with different excel sheets corresponding to every experimental run each. Every sheet contains two columns: The first one corresponding to time elapsed in seconds and the second to inflow discharge in m3/s. They have the same name as their corresponding test case (see Section 3 and Table 4): For example inflow hydrograph for test case 1a will be in sheet 1a and so on.

An alternative can be the use of Probe number 1 reading as the water level history of the feeding tank. The Probe 1 readings for the different test cases are provided as individual sheets in file Probe1.xls.

The inflow angle, a_inflow, will be taken perpendicular to the inflow section. It can easily be seen that the inflow angle with the X-axis:

a_inflow = -23.5º

Downstream boundary condition

The physical model extends some 45 m downstream of the studied model city area. Since it is not feasible to use the whole model length, the computational domain in the benchmarking process is limited to the approximately 5m long model city area. Therefore numerical boundary conditions need to be imposed at the downstream end of the computational domain.

It is suggested that each modeller chooses the boundary conditions scheme that best suits his model. As a hint, and although it is not proved that the flow leaves the model city area at the downstream section in supercritical conditions, preliminary tests with extrapolation of all variables have yielded satisfactory results.

The experimental team at ENEL-CESI suggested a bed Manning roughness coefficient for the concrete bed of:

n = 0.0162

Back to top

ENEL-CESI performed in all more than ten different tests on the Toce river valley physical model varying the city lay out, bathymetry and flooding conditions. Furthermore every test was run twice in order to check for repeatability. The set of test changes comprise:

i) Model bathymetry, either original (A) or channelled (B) with masonry walls

ii) Model city lay out, either aligned or staggered

iii) Flood intensity, characterised by the inflow hydrograph peak.

Due to the large number of experiments performed only some have been selected to serve as benchmarks for the mathematical modellers. The choice has been made with the aim of highlighting special features or effects or in order to make the tests more exacting on the mathematical models.

The following criteria have been used to select the benchmarking test cases:

i) The importance of focusing onto the effects of the flood on the model city, and conversely on the city influence on the flood. Hence most of the cases selected regard the bathymetry with lateral masonry walls (B or channelled bathymetry).

ii) The strong fundamental differences between the two model city lay outs. This makes it necessary to include both city arrangements for comparison in most cases.

iii) The modelling experience indicates that for a given flood event (in this case a strong inertial flood) it is more difficult to reproduce lower than higher submersion levels. Friction and bed variation effects are more noticeable, mass conservation errors more apparent and absolute deviations translate into larger relative errors, hence easier to spot. With these ideas in mind the low hydrograph has been selected out of the three available for the channelled bathymetry (low, 60 l/s, medium, 80 l/s, and high, 100 l/s) in all cases except one where the maximum hydrograph (100 l/s) has been chosen for comparison.

The inclusion of buildings in the mathematical models can be done in several ways. The most straightforward approach is to perform a detailed meshing around and then treat them as solid walls (i.e. as a boundary condition). In order to test simpler, less costly strategies it is proposed that buildings be represented as abrupt bottom elevations or even as local areas of very low conveyance (i.e. high apparent friction coefficients). It is suggested that those modellers willing to do so test also different building treatments.

The different benchmarking options have been grouped in order to test one effect at a time when possible as it is explained below.

1. Tests on the effects of city model lay out

It is intended to have the simplest configuration in order to make apparent the differences due to the two model city lay outs. Hence the following features: Low hydrograph (60 l/s peak discharge), channelled model valley bathymetry and building boundaries treated as solid walls. The two variants are then:

a) Aligned building arrangement
b) Staggered building arrangement

2. Tests on the effects of model bathymetry

It is not possible to make a direct comparison between results obtained with the original and the modified bathymetry holding all the other features the same. This is so because the construction of the two masonry walls has a strong influence on many other issues. For instance several houses had to be removed; the inflow hydrograph must be decreased considerably in order to have comparable submersion levels etc…

However it is interesting to see how mathematical models predict the flood propagation both in a simplified valley (channelled with the masonry walls) and in an actual valley (the original Toce valley model bathymetry).

Hence the tests proposed here are the counterpart of tests 1a and 1b but on the original valley, with the following characteristics: Low hydrograph (for the original bathymetry runs this means some 90 l/s peak discharge), original model valley bathymetry, building boundaries treated as solid walls. The two variants are:

a) Aligned city lay out
b) Staggered city lay out

3. Tests on the effects of the flood intensity

Common features include: Staggered city arrangement, channelled model valley bathymetry. Building boundaries will be treated as solid walls. The two cases are:

a) Low hydrograph (60 l/s peak discharge)
b) High hydrograph (100 l/s peak discharge)

4. Tests on the numerical treatment of buildings

This set is not dictated by the different experiments performed by ENEL-CESI, but rather by the different alternatives in the treatment of buildings within the modelling framework.

Common features for this set of tests comprise: Staggered city lay out, low hydrograph (peak discharge 60 l/s) and channelled model bathymetry. The building treatment options are:

a) Building boundaries treated as solid walls
b) Buildings treated as an abrupt bottom elevation
c) Buildings included as a reduced conveyance area (therefore an unusual high friction coefficient to be set by the modellers).

Table 4 below summarizes the benchmarking program designed and also the fact that, due to redundancies, the total number of variant cases is reduced to seven as last column shows.

The test cases listed above are considered as the standard set for all participants in the program. It is a decision of individual modellers to run all or only some of the cases. The group at University of Zaragoza will perform a systematic analysis of the submitted results by comparing results from different modellers and with experimental data only for the standard set of tests.

If a particular modeller wishes to perform other tests than those listed above and compare his results with the experimental data he will be able to do so after the blind phase of the benchmark program is finished. The deadline for submission of results and hence the end of the blind phase of the program is set to 31 May 2003. Therefore the complete set of experimental data can be released to those organizations requesting them after the deadline of 31 May 2003.

Back to top

The modellers participating in the program are requested to send their results files to the following e-mail address:alcrudo@posta.unizar.es not later than May 31, 2003 with the following nomenclature and content.

Modellers are requested to send to the above mentioned e-mail address a short MS Word or plain text file containing a short description of their modelling technique and any features judged interesting or deserving explanation. In particular a short description of the numerical method used, the type and size of grid or grids used, time step, boundary treatment or any other consideration together with comments on the results if any. References to previous or other authors' work can also be made to help understand his simulations.

4.2 Output files

It is strongly recommended that the submitted data files be MS Excel books (extension .xls) containing the different cases as distinct sheets. In case that Excel format can not be used then plain text (ASCII, extension .txt) files can be submitted.

MS Excel files

If the MS Excel format is chosen there will be only one file per modeller with multiple sheets, each one containing the results of one single test case. Each sheet will be named according to the test case (i.e.: Test1a, Test2c, or simply 1a, 2c, 3b etc ...).

The file name will be a suitable, easy to understand acronym of the organisation (preferably in capitals), the individual name or a combination of both. Examples of suitable filenames could be:

Each excel sheet of the file, named after the corresponding test case, will contain 10 columns. The first column corresponding to the elapsed time in seconds, the second to the computed water depth at probe number 2 in meters (recall that probe number 1 lies in the feeding reservoir and is not a simulation output), the third to the computed water depth at probe number 3 and so on, according to the example below:

Experimental data have been recorded every 0.2s for a total duration of 60s. However there is no need to write simulation results on the output files every such short time intervals. Modellers are free to choose longer output time steps provided resolution of their simulation data is not lost. Suggested values lie around 0.2s to 0.5s.

Plain ASCII text files

Modellers choosing this option are requested to send one file per test case with the same format as the Excel sheets (i.e.: with ten columns, the first one corresponding to the time elapsed in seconds and the following to subsequent probe readings in meters, see Table 5). The naming of the file will follow the same rules as for Excel files save for the fact that the test case identification must be added. Hence examples of suitable filenames could be:

Back to top

Bathymetry files
	SHORT_TOCE_ORI.DAT	Original Toce river valley physical model (scale 1:100) digital terrain model (DTM) cut to some 6.5m long.
	SHORT_TOCE_MOD.DAT	Modified Toce river model physical model (filled river bed, extrapolation of points to cover upstream protruding concrete slab) bathymetry cut to some 6.8m long.
Building coordinates files
	CASE_ALLINEATE_lista_coordinate.doc	List of the coordinates of the four corners of each building for the Aligned city lay out.
	CASE_SFALSATE_lista_coordinate.doc	List of the coordinates of the four corners of each building for the Aligned city lay out.
Boundary conditions files
	Inflow.xls	MS Excel file (book) containing the inflow hydrograph for each test case as different sheets.
	Probe1.xls	MS Excel file (book) containing probe 1 readings for each test case as different sheets.

Back to top

Soares-Frazao S., Morris M., Zech Y., (2000), Concerted Action on Dam Break Modelling (CADAM): Objectives, Project Report, Test Cases, Meeting Proceedings. CD-ROM, Université Catholique de Louvain (Belgium), Civil Engineering Department, Hydraulics Division.

To download a zip file containing all the files for the Model City Flooding Experiment click here.

back to WP3

back to Benchmarks Page

back to Top