Scotland has an area of roughly 79,000km2 and a population of five million. The northern highlands is the least populated area, but covers two thirds of the land, with the central lowlands the most populous zone and the fertile southern uplands dedicated mainly to farmland. In all, the country has some 50,000km of rivers and numerous lochs, including major water bodies such as Loch Lomond and Loch Katrine, which supply water to Glasgow, and Loch Ness.
With such plentiful water resources and the highest rainfall in Britain, a number of Scotland's major cities are vulnerable to flooding and have suffered inundation over recent decades. To provide adequate flood mapping facilities for the country's extremely varied terrain SEPA, the Scottish Environmental Protection Agency, began to search for a software solution that would be capable of becoming the second generation flood map for Scotland, a searchable map providing web-based flood outlines that the public could access.
The maps would have to indicate to members of the public the flood risk that exists at any particular location (that is, whether there might be a one in 100, 200 or 1000 year event flood risk). The final output will consist of indicative rather than accurate flood outlines that are automatically generated, as the timescale and funding for the project is naturally limited.
Criteria included requirements for the software:
- To be robust and defensible;
- To be fully automated;
- To be capable of being updated with new or improved data;
- To have the facility to be extended to include coastal floodplains, asset data and flood risk assessment;
- To be able to support regional and national scenario analysis.
The invitation to tender was issued on 30 January 2004 and the joint HR Wallingford and Wallingford Software proposal was submitted in March. The project was awarded in April with a start date of 1 May and completion data of 30 September 2005 - InfoWorks RS from Wallingford Software proved an obvious choice for the task.
The project divides the country into three model areas and the processing order will be determined from the availability of the validated NextMap Digital Terrain Model (DTM) data. This is likely to be available for central Scotland, followed by the South and finally the Northern Highlands and the Islands. In all, some 50,000km of river will be modelled. The project is divided into six work packages - the first will undertake raw data processing to provide clean data for InfoWorks RS; the second will provide InfoWorks RS data processing and importing to create models with the data required for normal depth calculations.
A third package will provide normal depth model simulation to provide corrected normal depth flood outlines. The fourth work package focuses on backwater model simulation, to provide backwater model flood outlines, and the fifth package undertakes an aggregation of the results to enable merged flood outlines. Lastly, internet delivery means enabling the flood outlines to be accessible in the public domain via the SEPA website. The methodology for the project has been agreed and it is due to begin shortly.
In more detail, the first work package delineates catchment boundaries, eliminating small catchments of less than 3km2. Flowgrid points are being aligned with centre lines and a stream order established. The stream ordering is used to divide the catchment into separate model reaches for InfoWorks RS. This is to ensure that no connected tributaries appear in the model. If these did occur, cross-sections from the two different reaches at a junction may overlap. This would result in two different calculated water levels at a given point, which is physically impossible. The solution is to order the streams such that stream order one is the main catchment flow path. Stream order two contains all the tributaries that join stream order one – forming a separate InfoWorks model. Stream order three contains all the tributaries that join stream order two and so on for higher number orders. Rivers serving a catchment of less than 3km2 total area are deemed too small to be modelled.
Centrelines will be generalised and smoothed using a bezier smoothing function based on longitudinal slope. This allows more smoothing in the heavily meandering downstream reaches and less smoothing in the mountainous regions. The smoothing is introduced to provide a generalised line (that is, less irregular) for creating the perpendicular cross-sections. If this did not occur, the cross-sections would be created perpendicular to the original Digital River Network (DRN), resulting in numerous intersecting cross-sections. The initial cross-sections are created and then checked for intersecting cross-sections, for instance around a meander bend. These are removed using a hierarchical system, such that cross-sections with the greatest number of intersections are removed first. The number of intersections for each cross-section is then recounted, and again the cross-sections with the greatest number of intersections is removed. This process is repeated until there are no-longer any intersecting sections along a given reach.
The Land Cover Map 2000 (LCM) is used to define 22 different categories of roughness types. Unit roughness estimates are assigned to each category, for example classes covering coniferous woodland, arable land, natural grassland or bracken would each have a different unit roughness value. The LCM is provided as a polygon vector data set, which enables the roughness values to be extracted at each cross-section survey point at the same time that the DTM elevation is extracted. The catchment Digital Terrain Model (DTM) provided by Intermap is used to generate ASCII grid files for use in InfoWorks RS.
Further Package Requirements
The second work package involves manipulating the pre-processed cross-section data to convert these to MUSK sections, set BED markers, determine bed slopes, and model boundary conditions in preparation for importing to InfoWorks RS. A referenced table of flow points will be used to create initial conditions and flow inputs, and generate further files ready for use in InfoWorks, and ASCII grid files will be imported into InfoWorks to provide Digital Terrain Model data. A further step is to import the cross-sections into the Conveyance Estimation System software to generate rating curves that will be used in the InfoWorks RS simulation.
Package three will take network and event data and run models to confirm their operation. The models and results will be exported, boundary conditions updated and models re-run to provide an InfoWorks database of models and results. InfoWorks DTM data will be used to create flood mapping models, confirm their operation, correct for normal depth discrepancies, export and review the output and produce flood outlines to create flood contours.
The fourth work package involves backwater calculations for a few (~20) reaches as prescribed by SEPA for instance at Lochs or in certain tidal locations. For these reaches, previous MUSK sections will be converted to river sections and flood mapping models generated.
The final output from all the InfoWorks RS modelling (Work Package 3 and 4) will be shape files of flood outlines for each stream order for the three events (Q100, Q200 and Q1000).
Package five takes the normal depth and backwater flood contours, and merges the flood outlines for all the stream orders, creates smoothed boundaries at tributaries and removes isolcated pockets of water. These outlines are then reviewed for any discrepancies. There is a particular challenge with this package in merging water stream data at the intersections.
The last package will take the data to a form that can be used by the public on the SEPA website. The processed flood outlines will be added to multimap basemap (that is, overlaid on the National Ordnance Survey (OS) maps to show the location of the flood outlines). MultiMap will provide the specified functionality and lay-out and “look and feel” according to SEPA's specified web style sheet. Initially a trial SEPA Flood Map web page will created, tested and debugged, and given supplementary functionality where necessary. The final SEPA Flood Map web page is likely to be available at the end of 2005.
It is important to realise the flood outlines produced from this approach are considered to provide indicative rather than detailed accurate modelling results, as the method is fully automated and is to be applied at a national scale. National data sets will be used (for instance the DRN – Directional River Network,). The results of mapping will be attributed with uncertainty flags based on topography (such as the DTM); hydrology (CEH flow grid); the hydraulic model/methodology and output, that is, how the results compare to local knowledge and detailed models.
The method provides the potential for more detailed localised models and information regarding flood defences to be incorporated at a later stage. These initial flood outlines will provide a sound basis and starting point for producing and later updating flood outlines at a national scale.