Processing – Definition and Simulation of Flood Scenarios

Processing refers to the second phase in hydrodynamic modeling. Flood scenarios are defined based on a hydrodynamic model discretization using the finite element mesh created in the pre-processing phase. They require additional information: flow resistance, boundary conditions, initial conditions, and control parameters. Chapter7 will give a demonstration how the simulation of flood scenarios is implemented as a geoprocessing grid service.

4.3.1. Flow Resistance

Flow resistance due to turbulence and bottom friction depends largely on the surface material properties as well as the type and height of the prevailing vegetation. These parameters are typically based on field measurements, aerial photography, as well as land use, cadastral, or geological maps. Areas with the same characteristics are assigned the same set of flow resistance parameters, or class. Recent research has focused on automatically deriving a flow resistance classification from topographic data acquired using remote sensing methods [Rat07; MC+03]. It may be argued that this task is part of the pre-processing phase.

In the numerical model, flow resistance is included in the form of a bottom shear stress. It may be expressed, for example, using equivalent sand roughness coefficients or by the approach of Gauckler, Manning and Strickler [Gau67; Man91; Man95]. The flow resistance parameters are typically specified either uniformly distributed over the region, a very simple approach, or on elements of the mesh. The latter can be automated if a map of the flow resistance classification (a coverage) is available. The step then consists of a spatial intersection or a weighted overlay of the classification onto the mesh. Flow resistance data could be given in the form of a Web Coverage Service. CORINE¹land cover data is readily available and may serve as the basis for a classification.

4.3.2. Boundary Conditions

Boundary conditions represent the course of a flood event. They are additional con-straints on the shallow water equations, which lead to a mathematical boundary value problem. Boundary conditions should be chosen so that the boundary value problem is well-posed, i. e. there exists a unique solution to the problem. Typically, some sections of the boundary are assigned water level boundary conditions, and other sections are constrained by discharge boundary conditions, while at the remaining boundary the so-called no-slip condition is applied. The no-slip condition implies that the flow velocity at the boundary is zero. This is physically justifiable because adhesive forces between water and the boundary surface are stronger than cohesive forces of the fluid.

The unsteady SWE are of parabolic or hyperbolic character, depending on the flow state (subcritical or supercritical flow). This influences the number and kind of boundary conditions that lead to well-posed problems. As a fluvial flood typically goes from upstream a river to downstream, the model needs to be sufficiently large so that the flooding process is not biased towards the downstream model boundary. Storm surges

1CORINE stands for Coordinated Information on the European Environment.

in estuaries, on the contrary, get their main influence from the downstream boundary, whereas the upstream discharge only plays a minor role.

The boundary conditions are specified at the mesh boundary in the form of water stage or discharge hydrographs, i. e. time series of discrete flow properties often measured by sensors. It is possible to apply the OGC SOS for accessing hydrological measurements and even artificial hydrographs, as they are used in future flooding scenarios.

The water level and discharge time series for a flood scenario may come from stream gauging station data or hydrological forecasts. A lot of historical data for rivers in Germany is provided by the hydrological information system PEGELONLINE¹ and the Electronic Waterway Information System (ELWIS)²operated by the German Federal Waterways and Shipping Administration (WSV). The time series are usually available in a higher resolution than needed for the simulation, so they have to be clipped to the necessary interval and filtered in order to reduce the data volume.

Once a hydrograph is available, it has to be applied to the boundary nodes or distributed to a section of the mesh. Or, thinking in reverse, the availability of hydrographs for certain locations could be queried in a catalog of data services.

4.3.3. Initial Conditions

Initial model conditions form a plausible starting flow pattern for the flood. Some models support dry starts, where there is, initially, no flow in the system. As a general rule, though, models require a more or less realistic initial flow situation for a stable execution. An often used possibility, termedlake solution, is to start with a completely wet domain and slowly lower the water level by adapting the boundary conditions to reach a quasi-steady flow state [Kin93]. In more complicated flow regimes, e. g. tidal rivers, this procedure is not going to produce a realistic result. Instead, an artificial tidal wave will have to be set as model boundary condition and many tidal cycles have to be calculated until the effect of the initial condition has been “washed out” of the system.

Obtaining an initial condition may require a considerable numerical effort. This step may be automated by setting appropriate boundary conditions that produce a desired starting flow state and calculating the lake or tidal solution beforehand. Another option is to keep a repository of pre-computed flow states that can be queried. The OGC Web Coverage Service has an extension for temporal data sets suited for this purpose. This solution decouples the providers and consumers of initial conditions and could be used

1http://www.pegelonline.wsv.de

2http://www.elwis.de

as part of a virtual organization sharing flood simulation results in a grid. The initial conditions would have to be overlaid onto the mesh prior to a simulation.

4.3.4. Control Parameters and Model Calibration

The hydrodynamic simulation is partly controlled by empirically determined parame-ters, which influence numerical aspects of the model, such as the temporal discretization and timespan, the flow resistance characterization, a turbulence model, or wetting-drying procedures. The simulation of a flood scenario requires a calibrated hydrody-namic model. The calibration process requires a series of simulations, in which the control parameters are varied. These simulations merely have the goal of finding the best approximation of a known (observed) scenario.

More often than not, the effects related to turbulent and dispersive shear stresses due to time- and depth-averaging of the Navier-Stokes equations are summarized into a single term. The magnitude of the mentioned effects depends on the spatial and temporal (in-)homogeneity of the flow field and the quality of its discretization in the mesh. For this reason, the shear stress parameter can only be guessed or taken from experience. Similarly, the quantification of bottom shear stress is error-prone because it depends on an accurate representation of vegetation and soil parameters. Furthermore, vegetation is changing with the seasons and over longer time periods, which adds to the uncertainty in roughness classification.

Analysis of the Processing Phase

The tasks in the processing phase deal with setting up a flood scenario based on a previously generated mesh. In detail, a set of flow resistance parameters, initial and boundary conditions have to be applied, and the hydrodynamic simulation has to be executed. A possible sequence of geoprocessing operations for this phase is shown in Table4.2.

OGC data services could be used to store and query a flow resistance classification, boundary conditions, and simulation results. Especially task V , the simulation, is a complex geoprocessing operation possibly taking hours or days to finish, and producing large amounts of output data. It will be shown in Chapter7how the actual simulation can profit from execution in the grid.

Table4.2.:Sequence of geoprocessing operations for the processing phase.

I Overlay flow resistance onto the mesh.

II Filter boundary condition time series to fit the simulated time span.

III Apply boundary conditions to mesh boundary sections.

IV Overlay initial conditions onto the mesh.

V Run the hydrodynamic simulation.