Appendix 3 Validating existing surface water flood hazard and risk data
We use modelling tools to predict flood hazard (e.g. location, extent, depth, likelihood) of surface water flooding. Because surface water flooding is not a regular occurrence it cannot be fully understood simply by observing it. Instead, predictive models help us to understand where flooding could occur if there was heavy rain, examine how it might change with climate change and test the effectiveness of actions to manage the risk. Using the modelled flood hazard an assessment can then be made of the adverse impacts of that flooding, such as the types of building or infrastructure that would be affected. Flood risk is calculated in the same way for both simple and complicated models: multiple simulations of events for different likelihoods are used to estimate the adverse consequences of flooding on human health, the environment, cultural heritage and economic activity.
This section is concerned with validating SEPA’s regional pluvial modelling carried out in 2013 and shared with responsible authorities. Nevertheless, the principles set out here can be applied to other modelling that may be available, e.g. Scottish Water Section 16 modelling or modelling by local authorities.
In order to validate the regional pluvial hazard models, they should be compared with observed flood history. SEPA’s pluvial maps model an event occurring over a wide urban area. In reality, pluvial flooding can be highly localised and a flood event is unlikely to occur everywhere at the same time. The regional pluvial maps should be validated against areas known to flood that may be more localised, e.g. street and neighbourhood scale. This should help to pinpoint locations where:
- Flooding is predicted by the model and has been observed – good alignment between observed and modelled flooding locations is ideal, even if predicted flooding is not matched by observations elsewhere. Where there is good alignment and the flooding mechanisms are understood, higher confidence can be put in the modelled data.
- Flooding is predicted but has never occurred – in this case the model may be accurate but there has been no flood event in the given location to validate it. Just because a location has not experienced flooding in the past does not mean that it is not at risk of flooding.
- Flooding has occurred but is not predicted – in this case the model is failing to predict observed flooding and further information is likely to be required.
If the modelled flooding is not predicting observed flooding, professional judgment should be applied to ascertain the reasons why. Doing so will help to inform what further data collection / modelling is required.
It is likely that confidence in the model will vary throughout an area, depending on what scenarios are being modelled and the flooding mechanisms involved.
If the pluvial modelling fails to replicate observed flood history over wide urban areas, SEPA should be informed to determine whether updates to the strategic modelling can be made.
A3.1 Reasons why modelled flooding might not be predicting observed flooding
SEPA carried out regional pluvial hazard mapping using standard inputs for a range of modelling parameters. All will influence the outputs of, and hence confidence in, the pluvial flood hazard in an area. The parameters can be adjusted or new data gathered to improve validation and confidence in the model. The parameters and key influences are:
- Model resolution (and model type)
- DTM vertical accuracy, resolution, and representation of features
- Sub-surface drainage
- Percent run-off (infiltration)
- Manning’s coefficient (roughness)
- Rainfall inputs
- Other sources of flooding and interactions with other sources.
Digital Terrain Model ( DTM)
Having a DTM that accurately represents the topography and flow pathways is one of the most important factors influencing confidence in pluvial models.
The DTM used by SEPA for the regional pluvial hazard maps is based on LiDAR and NextMap, and has been processed to remove false blockages and introduce building footprints (as 0.3 m heights) as indicated on Ordnance Survey data. No ‘ground truthing’ of the DTM was undertaken. Hence in some cases the DTM may not accurately represent flow paths because of:
- Inaccuracies in the DTM – LiDAR was used for the majority of the modelling but where it was not available NextMAP was used which has lower vertical accuracy and lower resolution. In urban areas the DTM is typically processed to remove buildings and other structures. The process involves interpolation, which can introduce errors. Changes in catchment since the DTM was collected may also mean that it no longer accurately represents the current ground surface.
- Missing features – features such as kerbs and walls may not be picked up in LiDAR. Some false blockages may not have been identified. Existing flood management structures have not been explicitly added to the DTM, but some may have been picked up by LiDAR.
Known false blockages identified from mapped data can be removed from the DTM. Where it is thought that local topography or other structures are not represented in the DTM but are having an impact on surface water flow routes and flooding locations, existing surveys or new topographical surveys can be carried out and the results added to the DTM. For example, where roads are known to convey significant flows they can be modelled in the DTM as depressions. Other features too, can be added, such as flood management structures, kerb-lines / heights, low walls, additional buildings and known flow routes (Option A in Table A3.1).
Further information can be found in CIWEM Urban Drainage Group’s modelling guide. 
Flow pathways that are narrower than the model resolution may not be resolved. Generally, the model grid size has to be half the width of a flow pathway in order for it to be resolved. This can cause problems with features like vennels and closes in urban areas. Although model resolution can be increased to resolve flow pathways, doing so can significantly increase model run times. For instance, halving the grid cell size generally leads to an eightfold increase in run times.
Sub-surface drainage is represented in the regional pluvial model rainfall hyetographs by assuming a 1:5 year drainage capacity in urban areas and deducting this from the input rainfall (rural areas assume no drainage). This assumption may not accurately represent the influence of drainage in an area, and actual drainage capacity is likely to vary across an SWMP area.
The regional pluvial modelling may also be inaccurate because it does not correctly represent the dynamic interaction of above- and below-ground flows; this can occur when large sewer pipes transfer flooding from one location to another or where the catchment of the sewer system does not follow above-ground topography. Section 16 sewer flooding data and other sewer asset information can be used to infer the importance of sewer and surface interactions.
Section 16 sewer flooding mapping (where available) is useful for determining local drainage capacity. This can help to inform whether the default 1:5 year return period drainage capacity is appropriate. Scottish Water may also have other information on drainage capacity.
Additionally, Section 16 results can be used to determine the critical duration of drainage exceedance. Knowing the critical duration, which varies with gradient and other factors, will help to inform which SEPA mapping scenario – a one-hour or three-hour storm event duration – would be more appropriate to use. If, in consultation with Scottish Water and other partner agencies, it is clear that none of the default drainage capacity or event duration scenarios is suitable, further pluvial modelling using data supplied by SEPA (Option A in Table A3.1) should be considered.
Percent run-off (infiltration)
Percentage run-off is represented in SEPA’s regional pluvial model rainfall hyetographs by assuming 70% in urban areas and 55% in rural areas. (Whether an area is designated urban or rural areas is based on the 2007 land cover map.) In reality, infiltration rates will vary at a smaller spatial scale and over the course of a flood event. If the infiltration rates used in the regional pluvial modelling are considered inappropriate, the models can be re-run with different infiltration rates. See Option A in Table A3.1 for further information on re-running the regional pluvial hazard models.
Other sources of flooding and interactions
SEPA’s regional pluvial modelling does not show flooding from culverts or pipes. It may, though, represent flooding from smaller urban burns that are not culverted and are featured in the DTM. Higher confidence in the model is likely where the drainage capacity is exceeded and most flooding is from overland flow.
Neither is the impact of high river or sea levels on pluvial flooding or the drainage network taken into account in the modelling. These interactions can impede discharge from surface water drainage outfalls, resulting in a locally reduced drainage capacity. Such dynamic interactions require a more detailed type of model that can represent above- and below-ground interactions. This type of model, often called an ‘integrated urban drainage model’, should be developed for high risk areas or where the option appraisal requires more detailed understanding of these interactions (Option B or C in Table A3.1).
SEPA’s 2013 pluvial modelling used the Flood Estimation Handbook ( FEH) 1999 depth duration frequency ( DDF) model for rainfall. A new version of the DDF model ( FEH 2013) was released in 2015, after SEPA’s pluvial hazard maps were published, and replaced the existing DDF model ( FEH 1999) for an entire range of return periods and durations. FEH 2013 incorporates a significant amount of additional data and uses an enhanced statistical model. As a result, it has greater depths for short-duration rainfall (< 6 hours) for most locations in Scotland up to the 0.1% annual exceedance probability ( AEP) event (1:1000 year rainfall event). Further information on the development of the FEH 2013 model can be found in Defra’s technical report Reservoir Safety.  Rainfall depths in FEH 1999 and FEH 2013 could be compared for one- and three-hour durations at a number of locations across the SWMP area to determine whether the FEH 1999 in SEPA’s pluvial modelling is significantly underestimating rainfall depths.
A3.2 Options for further modelling
This section describes options for further modelling in more detail. As described above, further modelling will be required if SEPA’s flood hazard maps do not reflect observed flooding. The type of modelling required will depend on why it is not reflecting observed flooding. Further information on undertaking a flood modelling study can be found in SEPA’s flood modelling guidance for responsible authorities. 
A risk-based approach should be adopted to select the modelling method. The approach applied should be the simplest one that allows subsequent decisions to be made with confidence. Modelling can be improved for more localised areas (e.g. highest risk neighbourhoods) or be recommended as a future requirement in an SWMP.
SEPA should be contacted if the pluvial modelling fails to replicate observed flood history over wide urban areas, in order to determine whether updates to the strategic modelling can be made.
It is anticipated that most SWMPs can be developed effectively with the SEPA 2013 regional pluvial flood modelling and Section 16 sewer flooding, without having to undertake further modelling in the first stages.
Table A3.1 Options for further hazard modelling ( NB: once new hazard maps are available the adverse impacts of flooding / flood risk will need to be assessed)
|A) Re-run regional pluvial hazard model
|The regional pluvial hazard mapping models can be re-run with updated information on:
In order to gain a more accurate representation of pluvial flooding, the model parameters and key influences listed above and in Section A3.1 can be adjusted to better reflect real conditions using data or knowledge from local authorities or Scottish Water. ( SEPA can also be contacted for advice.) This can either be via an adjustment to the SEPA model, that can be supplied to the local authority, or the local authority can apply an alternative modelling approach using the available data.
Where both observed flood event data and rainfall data are available for a particular event, models may be run to compare observed rainfall data with observed flooding. This may help to increase confidence in the model and identify the source of any discrepancies between the modelled and observed flooding.
SEPA agrees with the principle of sharing models and model data, but recognises that the ability to do so is dependent on licensing conditions. Licensing conditions will apply to the model themselves, the inputted datasets and the outputs generated by the contractor. This may limit what information SEPA can share until licensing conditions are agreed with licensors.
Re-running the regional pluvial models is the simplest approach. However, as the regional pluvial modelling does not include an explicit representation of the drainage system, it will only be appropriate in areas where the sub-surface drainage system is believed not to be an important factor influencing flooding.
|B) Sewer and pluvial modelling
|The coupled 1D (underground sewer network) and 2D (above ground) model allows water to flow across the modelled urban surface and re-enter the sewer network where there is an inlet and underground capacity. This will be appropriate in areas where the sub-surface drainage system is important but where there are minimal interactions with other sources of flooding (e.g. rivers, sea).
|C) Integrated Catchment modelling
|This usually involves combining existing sewerage models with watercourse models and a 2D representation of the urban surface. It can also be used to model the influence of other sources of flooding, including river and coastal flooding, on surface water flooding. This approach is costly and time-consuming, and requires a high degree of collaboration between partner agencies. It is already being applied in areas of very high risk (e.g. Glasgow) and in other Integrated Catchment studies.
Gordon Robertson: Flooding_Mailbox@gov.scot
There is a problem
Thanks for your feedback