The Scottish Government

Contents |

Listen

CHAPTER 5: HYDRAULIC MODELLING

Choice of Model

5.59 A range of hydraulic models were used in assessing the current standard of protection provided by the Flood Prevention Schemes. After an initial assessment the most appropriate model was chosen and those used are described in the following sections.

HEC- RAS

5.60 In most cases, where flood prevention scheme assets comprised defences along a river, the most common choice of hydraulic model was HEC- RAS (Hydrologic Engineering Center - River Analysis System), developed by the United States Army Corps of Engineers. The software includes two 1-D mathematical models to model gradually varied flow in networks or branched open channel systems. It includes methods to account for the head losses at structures such as bridges, culverts and weirs.

5.61 The Version used, 3.1.3 offers both steady state and unsteady (hydrodynamic) simulation modes, but the steady state methodology was used in most cases where there was limited flood plain storage.

5.62 The steady state methodology models peak flows considering a non time-varying environment, i.e. calculating maximum water levels at each cross-section from the peak flow value. The limitation with this approach is that if flood storage is significant, no account is taken of the potential attenuation or reduction in the flood peak. Hence, steady-state modelling has a tendency to provide a conservative estimate of peak water levels in volume sensitive systems - but few of the FPS schemes fall in to this category.

ISIS

5.63 ISIS was used to investigate the performance of the Earlston and Brock Burn & Levern Water FPS's. This software package was developed by Halcrow and is marketed and supported by Wallingford Software. In terms of functionality it is very similar to HEC- RAS and is based on the same hydraulic theory. Like HEC- RAS it is a 1-D model (i.e. ignores lateral variation in water surface slope across a river).

InfoWorks RS

5.64 In one case (Fettercairn), an existing model for the river reach upstream of the scheme was made available. This was available in the InfoWorks RS v6.04 software package developed by Wallingford Software. The computational engine in InfoWorks RS is the same as in ISIS.

InfoWorks CS

5.65 This package is designed for the simulation of flows within pipe or culverted networks although it can be used for open channel systems where out-of-bank storage and structures such as bridges are not important in determining water levels. It also includes several rainfall-runoff models to generate inflows to each node of the network - including the Rational Method, FEH and the Wallingford Procedure. InfoWorks CS has been used on sites with a network of culverted pipes and was chosen at Millfield of Cupar, Earnhill Road - Gourock and Skiprunning Burn - Jedburgh.

WinDes

5.66 MicroDrainage WinDes is a similar model in computational terms to InfoWorks CS. This package models storm drainage and combined sewer networks in accordance with the Wallingford Procedure. Like InfoWorks CS, WinDes allows for modelling of networks under unsteady and surcharged flow conditions by creating a run-off hydrograph for each sub-catchment, using a catchment specific rainfall hydrograph, a determined time to peak and a range of rainfall durations. The unsteady state methodology used allows a hydrograph to be constructed for system capacity exceedance, if it does occur. Representation of overland flow can be included in the model.

5.67 Once the input hydrographs had been routed through the models for the varying return periods and storm durations, the extent of flooding was examined and the storm duration that resulted in the greatest flood volume was selected.

JFLOW

5.68 JFLOW is a 2-D raster model developed by JBA Consulting to provide an overland flow model that can be used for both small-scale breach/overflow analysis and large scale catchment modelling. It provides estimates of flood depth, velocity and extent for out-of-bank flow. For a number of schemes overland flow occurred during flood events that could not be accurately represented in a 1-D hydraulic model, e.g. culvert capacity is exceeded or levels in the floodplain are lower than the bank levels. In these cases, JFLOW was used to model the depth and extent of inundation.

5.69 The hydraulic methodology is based on a diffusion wave approximation and is simpler to use and more robust than fully hydrodynamic models. It ensures accurate conservation of the volume of water, which can be a problem in some 2D hydrodynamic models (Wheater, 2001) ¹². Work by Horritt and Bates (2001) ¹³ has shown that such models are able to give as good, or even better predictions of inundation extent where flow is strongly topographically driven.

5.70 The methodology is a raster-based approach where each cell has a ground level and water depth. Water can move to any of surrounding eight cells where the water level is lower. Water will pond in low spots until the water level is high enough to spill. The velocity of movement depends on water surface slope and surface roughness.

5.71 The above points describe the basic principles of the model. The two underlying principles are:

• Mass conservation within each cell - each cell is treated like a small storage area.
• Calculation of the fluxes between the cells, using a form of the generalised weir equation.

Model Construction

Geometry

5.72 Surveyed river cross sections or culvert dimensions and relevant asset data were input to the models. Where necessary, the surveyed sections were extended using the NEXTMap DTM a 5m grid digital terrain model ( DTM). Initial checks made between the last surveyed sections and the first extension point in most cases showed varying differences in levels but often did not provide a consistent disparity for which to alter the DTM height values globally. Such variations are most likely to be due to the filtering of buildings/tree data rather than larger spatial scale errors such as datum shift.

Manning's "n"

5.73 Manning's 'n' values have been used to represent channel and culvert roughness. The initial roughness coefficient values were assessed from observations and photographs taken during walkover inspections of the modelled reaches. The values were selected on the basis of experience and by comparing the reach characteristics with standard tabulated values of Manning's 'n' in the HEC- RAS Hydraulic Reference Manual.

Overbank flow options

5.74 1-D models using extended cross-sections do not allow differential water levels between the river and out-of-bank areas. Where appropriate, an improvement to this was made by applying lateral weirs to represent the bank defences. This allows the calculation of weir flow over the defences/banks. For steady state models, the loss of flow overbank was calculated and then deducted from the downstream reach, thus improving the confidence in water levels in the downstream reach.

5.75 Where unsteady hydrodynamic models were used ( ISIS, InfoWorks RS), overbank flow was modelled using spill sections along the main channel flowing into floodplain storage areas. Floodplain sections represented flow between storage areas.

Boundary conditions

5.76 Ideally, the boundary conditions of a model are taken at points where the discharge and water level can be accurately estimated, for example a known flood mark at a control structure such as a weir. Failing this, the boundaries of the model should be taken at a distance sufficiently far from the site of interest as to allow any effects of uncertainty within the boundary conditions to have a minimal effect on the water levels at the site. In many cases, no obvious structures were present in the study reach to provide an accurate boundary condition, so a boundary condition of normal depth was assumed.

Joint Probability

5.77 Several river reaches modelled had tidally influenced water levels at the downstream boundary: Arbroath, Conon Bridge, Bridge of Earn, Carnoustie, Port Logan. For other schemes including Denholm, Earlston, Innerleithen, Lauder, Millfield of Cupar, Peebles, Rannie Burn and Skiprunning Burn, the modelled watercourse discharges into a larger river, the level of which could potentially affect water levels along the scheme reach. This introduces the consideration of joint probability: the probability that a flood event in the scheme reach will be concurrent with a flood event in the downstream river or with extreme sea levels. As an initial investigation, the effect of varying the downstream boundary was assessed. If this was found not to significantly affect levels along the scheme reach, then no further assessment was deemed necessary. For a few schemes, further investigation was required, and is discussed in Appendix B.

Table 5-1: Example of Joint Probability Marginal return period

5.78 Hydraulic modelling can also be used to determine the critical combination of marginal return periods for each joint return period. This method was used to determine the current standard of protection for Prestonpans by assessing the joint probability of extreme sea level and wave height, and for Port Logan by assessing the joint probability of river flow and extreme sea level.

5.79 However, for a number of schemes, the relevant parameters were both river flows within subcatchments. The joint probability research detailed above has included no analysis of the dependence between river flows within subcatchments, and as such cannot be used for this scenario. The FEH does not provide any practical guidance; rather it recommends avoiding it where possible. In any case, carrying out the necessary analysis to provide a robust joint probability function is both beyond the scope of this project, and likely to be impossible without gauged records on each tributary. In most cases, preliminary investigation of the sensitivity to the downstream boundary indicated that levels along the study reach were either totally unaffected, or in one case totally controlled by the downstream boundary level.

Modelling assumptions

5.80 In general, for accurate current assessment of the standard of protection the channel and floodplain roughness values were based on current observations.

5.81 In a few cases, the condition assessment of a scheme indicated the presence of minor faults or gaps in the defence assets. Due to the minor nature of these faults and relatively small cost of repair, it was assumed that these are repaired.

Model runs

5.82 A number of model runs were carried out for the assessment of current flood risk at the schemes. These included:

• The current undefended case assuming the defences as part of the scheme do not exist, but no account made for dredging or channel realignments in place as part of any of the schemes.
• The 'as surveyed' model run with no freeboard taken into account representing the current best estimate for the onset of flooding.
• The current standard of protection case assuming that the total calculated freeboard is subtracted from the current defence height. A series of flows are run to provide the current standard of protection (expressed as a return period) that will be withstood by the scheme. The reduction in defence levels using a freeboard allows for a high degree of confidence to be given to this standard despite any uncertainties in modelling and hydrology.
• The standard of protection case with an allowance for climate change.

5.83 The modelling determines only the fluvial flood risk, and does not consider secondary flooding from drainage networks unless the drainage system forms a component part of the FPS.

Calibration

5.84 Calibration is necessary to improve confidence in the hydraulic model predictions of flood stage levels and to test the uncertainty in the parameters used. Models should be calibrated using historic flood information, ideally in the form of peak water levels at specific locations corresponding to peak recorded river flows.

5.85 Unfortunately there was frequently no recorded discharge or level data available for previous flood events. For a couple of schemes, there were anecdotal records of flooding, such as wrack marks or photographs, but with no associated discharge data. Calibration using records of flood data was only carried out at the two sites in Galashiels, where there were wrack marks on structures and associated flow data recorded at the gauging station upstream for the 1984 event, and at Collier Street where the model had previously been calibrated against the 1990 and 1994 events.

Sensitivity testing

5.86 Sensitivity testing was carried out to test the sensitivity of the model and Standard of Protection to changes in the modelling parameters. Sensitivity to channel and floodplain roughness (Manning's 'n'), peak flows and downstream boundary conditions were checked. Changes in water levels resulting from changes in roughness allowed comment to be made on scheme maintenance. Uncertainty associated with estimated peak flows and the models sensitivity to this parameter is accounted for in the calculation of freeboard ( see Chapter 7).

Model Results

5.87 The results from each model are recorded in the report for each scheme. The hydraulic model has been supplied to both the relevant local authority and Scottish Executive. The flood outlines generated from the models are recorded in the database and in the model report. A summary of the results for each cross section can be found in the database under the surveyed cross section drawings and an example is shown in Figure 5-1. The information includes a rating curve at each section which helps future proof the database. New flows can be assessed against the rating curve to determine the flood level at a given section.

Figure 5-1: Hydraulic Model Summary Sheet

Contents |