M Rooke Our ref: AE/2020/125210/01-L01 Broadland District Council Your ref: 20191426 Development Control Thorpe Lodge (1) Yarmouth Road Date: 09 June 2020 Norwich Norfolk NR7 0DU

Dear Mr Rooke

CONSTRUCTION OF HOLIDAY AND LEISURE PARK COMPRISING AN ADDITIONAL 280 UNITS OF HOLIDAY ACCOMMODATION; LANDSCAPING, DRAINAGE AND ASSOCIATED INFRASTRUCTURE WORKS

LAND AT HAVERINGLAND HALL PARK, HAVERINGLAND HALL PARK, HAVERINGLAND, NR10 4PN

Thank you for your consultation dated 8 June 2020. We have reviewed the application as submitted and are raising a holding objection on Risk grounds. Further information can be found within the flood risk section below. Further clarification of foul drainage will also be required. Our comments on this can be found within the foul drainage section below.

Flood Risk

Our maps show the site is located in fluvial Flood Zone 3, the high probability zone. However, we have not undertaken any detailed modelling for the nearby Swannington Beck ordinary watercourse, so this source of flood risk has not been assessed for the purpose of the flood map.

The submitted flood risk assessment (FRA), referenced RCEF6670 and dated January 2019, does not comply with the requirements set out in the Planning Practice Guidance, Flood Risk and Coastal Change, Reference ID: 7-030-20140306. This FRA does not, therefore, provide a suitable basis for assessment to be made of the flood risks arising from the proposed development and we are raising a holding objection. In particular, the submitted FRA fails to:  Identify the impacts of fluvial flood risk from the Swannington Beck ordinary watercourse.  Assess the impact of climate change using appropriate climate change allowances. In this instance, according to ‘Flood risk assessments: climate change allowances', the allowances that should be assessed are the Higher Central of 35% and the Upper End of 65%.

Environment Agency Iceni House Cobham Road, Ipswich, IP3 9JD. Customer services line: 03708 506 506 www.gov.uk/environment-agency Cont/d..

We can confirm that the applicant requested data from us and this was provided in January 2019. In this response, we advised that the flood map in this location is JFlow and we haven’t undertaken any detailed modelling to the ordinary watercourse and this is something that would need to be undertaken.

Overcoming our Objection

The applicant can overcome our holding objection by submitting an FRA that covers the deficiencies highlighted above and demonstrates that the development will not increase risk elsewhere and where possible reduces flood risk overall. If this cannot be achieved we are likely to maintain our objection to the application. Production of an FRA will not in itself result in the removal of an objection.

Modelling Guidance

The Flood Zone maps in this area are formed of national generalised modelling, which was used in 2004 to create fluvial floodplain maps on a national scale. This modelling was improved more recently, using a more detailed terrain model for the area. This modelling is not a detailed local assessment, it is used to give an indication of areas at risk from flooding.

JFLOW outputs are not suitable for detailed decision making. Normally, in these circumstances, an FRA will need to undertake a modelling exercise in order to derive flood levels and extents, both with and without allowances for climate change, for the watercourse, in order to inform the design for the site. Without this information, the risk to the development from fluvial flooding associated with the ordinary watercourse is unknown.

In order to have fully considered all forms of flooding and their influence on the site, it will be necessary to identify the fluvial flood risk. Fluvial flood levels will be required for Swannington Beck. It may be appropriate to undertake some flow analysis such at FEH and 1D modelling to establish the level. Any revised FRA will need to consider this source of flooding and demonstrate appropriate mitigation against fluvial flood risk.

We advise that modelling should be undertaken to accurately establish the risk to the proposed development in terms of potential depths and locations of flooding. The watercourse should be modelled in a range of return period events, including the 1 in 20, 1 in 100 and 1 in 1000 year events, both with and without the addition of climate change. The flood levels on the development site should be determined and compared to a topographic site survey to determine the flood depths and extents across the site.

Please refer to the attached documents:  OI 379_05 Computational modelling to assess flood and coastal risk  Flood Estimation Guidelines  ‘Using Computer River Modelling as Part of a Flood Risk Assessment - Best Practice Guidance’ for further advice regarding modelling submissions.

We acknowledge that some of the documents above refer to outdated planning policy. However, the technical guidance and our requirements regarding computer modelling remain relevant.

We would recommend that FRAs at all levels should be undertaken under the supervision of an experienced flood risk management specialist (who would normally be

Cont/d.. 2

expected to have achieved chartered status with a relevant professional body such as the Institution of Civil Engineers (ICE) or the Chartered Institution of Water and Environmental Management (CIWEM)).

Paragraph 103 of the NPPF states:-

“When determining planning applications, local planning authorities should ensure flood risk is not increased elsewhere and only consider development appropriate in areas at risk of flooding where, informed by a site-specific flood risk assessment following the Sequential Test, and if required the Exception Test, it can be demonstrated that:  Within the site, the most vulnerable development is located in areas of lowest flood risk unless there are overriding reasons to prefer a different location; and  Development is appropriately flood resilient and resistant, including safe access and escape routes where required, and that any residual risk can be safely managed, including by emergency planning; and it gives priority to the use of sustainable drainage systems.”

Some areas of land within the site are likely to be subject to a higher risk of flooding than other areas within the site and an understanding of the susceptibility/vulnerability of land to flooding should be delivered through flood modelling and risk assessment in order to influence the layout of housing areas to avoid siting housing on areas of land that are susceptible to higher chances of flooding. This will allow a sequential “risk- based” approach to be applied to development within the site as directed by the National Planning Policy Framework.

We ask to be re-consulted with the results of the FRA. We will provide you with bespoke comments within 21 days of receiving formal re-consultation. Our objection will be maintained until an adequate FRA has been submitted.

If you are minded to approve the application contrary to this advice, we request that you contact us to allow further discussion and/or representations from us in line with the Town and Country Planning (Consultation) (England) Direction 2009.

Flood Risk Climate Change Guidance: Detailed Allowance

Climate change allowances have changed recently. The Planning Practice Guidance provides advice on what is considered to be the lifetime of the development in the context of flood risk and coastal change. Our guidance 'Flood risk assessments: climate change allowances' provides allowances for future sea level rise, wave height and wind speed to help planners, developers and their advisors to understand likely impact of climate change on coastal flood risk. It also provides peak river flow and peak rainfall intensity allowances to help planners understand likely impact of climate change on river and surface water flood risk.

For some development types and locations, it is important to assess a range of risk using more than one allowance. The extent, speed and depth of flooding shown in the assessment should be used to determine the flood level for flood risk mitigation measures. Where assessment shows flood risk increases steadily and to shallow depths, it is likely to be more appropriate to choose a flood lower in the range. Where assessment shows flood risk increases sharply due to a 'cliff edge' effect caused by, for example, sudden changes in topography or defences failing or overtopping, it is likely to be more appropriate to choose a flood level higher in the range.

Cont/d.. 3

The proposed development is classified as a More Vulnerable development, and lies within Flood Zone 3. This means the applicant must adopt a “detailed” assessment. A detailed assessment requires the applicant to perform detailed hydraulic modelling, through either re-running hydraulic models (if available) or construction of a new model by the developer. Assuming the lifetime of the development is until 2069/2115, the allowances the applicant must apply are Higher Central (35%) and Upper End 65%)

We do not currently have mode coverage for this area. You will therefore need to create your own model.

We recommend that you assess both the 35% and 65% allowances, and if possible design the development to be safe through raised floor levels in the 65% climate change allowance. If this is not possible then robust justification should be provided, and the development should be designed to be safe through raised floor levels in the 35% allowance and the safety and sustainability of the development should be assessed for the 65% and managed through flood resilient/resistant construction measures to the satisfaction of the LPA.

Other Sources of Flooding

In addition to the above flood risk, the site may be within an area at risk of flooding from surface water, reservoirs, sewer and/or groundwater. We have not considered these risks in any detail, but you should ensure these risks are all considered fully before determining the application.

Foul Drainage

No information appears to have been provided with this application in regard to foul drainage. Therefore, we request this is provided to you. The applicant should make use of our foul drainage assessment form which is available here https://www.gov.uk/government/publications/foul-drainage-assessment-form-fda1. Please re-consult us and we will update our response within 21 days. Our standard advice in relation to foul drainage can be found below.

Government guidance contained within the National Planning Practice Guidance (Water supply, wastewater and water quality – considerations for planning applications, paragraph 020) sets out a hierarchy of drainage options that must be considered and discounted in the following order:

1. Connection to the public sewer 2. Package sewage treatment plant (adopted in due course by the sewerage company or owned and operated under a new appointment or variation) 3. Septic Tank

Foul drainage should be connected to the main sewer. Where this is not possible, under the Environmental Permitting Regulations 2010 any discharge of sewage or trade effluent made to either surface water or groundwater will need to be registered as an exempt discharge activity or hold a permit issued by the Environment Agency, addition to planning permission. This applies to any discharge to inland freshwaters, coastal waters or relevant territorial waters. Please note that the granting of planning permission does not guarantee the granting of an Environmental Permit. Upon receipt of a correctly filled in application form we will carry out an assessment. It can take up to 4 months before we are in a position to Cont/d.. 4

decide whether to grant a permit or not.

Domestic effluent discharged from a treatment plant/septic tank at 2 cubic metres or less to ground or 5 cubic metres or less to surface water in any 24 hour period must comply with General Binding Rules provided that no public foul sewer is available to serve the development and that the site is not within a Groundwater Source Protection Zone.

A soakaway used to serve a non-mains drainage system must be sited no less than 10 metres from the nearest watercourse, not less than 10 metres from any other foul soakaway and not less than 50 metres from the nearest potable water supply, spring or borehole.

Where the proposed development involves the connection of foul drainage to an existing non-mains drainage system, the applicant should ensure that it is in a good state of repair, regularly de-sludged and of sufficient capacity to deal with any potential increase in flow and loading which may occur as a result of the development. Where the existing non-mains drainage system is covered by a permit to discharge then an application to vary the permit will need to be made to reflect the increase in volume being discharged. It can take up to 13 weeks before we decide whether to vary a permit.

Abstraction Licencing

Advice to applicant

If you intend to abstract more than 20 cubic metres of water per day from a surface water source e.g. a stream or from underground strata (via borehole or well) for any particular purpose then you will need an abstraction licence from the Environment Agency. There is no guarantee that a licence will be granted as this is dependent on available water resources and existing protected rights. The granting of planning permission does not guarantee the granting or variation of an abstraction licence so we would recommend contacing our permitting service at your earliest convenience. Further information can be found here: https://www.gov.uk/guidance/water- management-apply-for-a-water-abstraction-or-impoundment-licence

We trust this advice is useful.

Yours sincerely

Mr Liam Robson Sustainable Places - Planning Advisor

Direct dial 020 8474 8923 Direct e-mail [email protected]

End 5

Using computer river modelling as part of a flood risk assessment

Best Practice Guidance - Version 1 April 2006 We are The Environment Agency. It's our job to look after your environment and make it a better place - for you, and for future generations.

Your environment is the air you breathe, the water you drink and the ground you walk on. Working with business, Government and society as a whole, we are making your environment cleaner and healthier.

The Environment Agency. Out there, making your environment a better place.

Published by:

Environment Agency Rio House Waterside Drive, Aztec West Almondsbury, Bristol BS32 4UD Tel: 0870 8506506 Email: [email protected] www.environment-agency.gov.uk

© Environment Agency

All rights reserved. This document may be reproduced with prior permission of the Environment Agency.

Environment Agency 1 Notes:

This document concentrates on computer river modelling. However, many of the principles apply equally to coastal modelling.

The principles also apply to Flood Consequence Assessments carried out in Wales.

Whilst allowances should be made for Climate Change, these have not been quantified in this Guidance. These should be assessed at the time of modelling using the latest Environment Agency standards.

For all contact with the Environment Agency you should ensure that you are speaking to the office that covers the area of land in question. For further details of Environment Agency office locations please refer to our website www.environment-agency.gov.uk

You should read our Standard Notice which details our terms and conditions. If this has not been supplied to you, you can get by calling us on 08708 506 506 or from our website (search for ‘types of licence’).

If you have any queries about the content of this document or suggestions for improvement please e mail [email protected]

2 Environment Agency CONTENTS

1. Introduction 4 1.1 Purpose of this Document 4 1.2 Modelling and Flood Risk Assessment 4 1.3 Appropriate Modelling Staff Involved 4 1.4 Requirements at Specific Locations 5 2. Objectives of the Model Study 5 3. Model Building 5 3.1 Choice of Model 5 3.2 Survey Data 6 3.3 Hydrometric Data 6 3.4 Historic Information 6 3.5 Previous Modelling 7 3.6 Hydrological Assessment 7 3.7 Model Building 7 4. Model Calibration, Verification and Sensitivity Testing 8 4.1 Calibration 8 4.2 Verification 8 4.3 Sensitivity Testing 9 5. Reporting 9 5.1 General 9 5.2 Items to be Included 9 5.3 Format of Reporting 10 5.4 Other Deliverables 10 5.5 Future Use 10 6. Quality Assurance and Audit Trail 10

Glossary of terms 11 List of abbreviations 12 References 12

Environment Agency 3 1. INTRODUCTION

1.1 Purpose of this Document

This document is guidance for carrying out a flood risk assessment where computer river modelling is necessary. Flood risk assessments are carried out by individuals, developers, consultants or Local Planning Authorities for a variety of reasons (e.g. for development purposes).

The Environment Agency’s Policy is to take a risk-based approach to managing flood risk using an approach consistent with that commonly applied to other hazards. This means that flood risk management decisions are informed by flood risk assessment. It is recommended that others take the same approach.

The purpose of this document is to give general best practice guidance on the standards that should be used when carrying out computer modelling of watercourses in order to complete a flood risk assessment. Further details about undertaking Flood Risk / Consequence Assessments for the construction industry are given elsewhere, in particular in CIRIA Report C6241.

Further information may be required for land use development purposes as detailed in PPG25 (also having regard to draft PPS25) or TAN15.

It is only intended to give an overview of best practice to be considered when carrying out modelling in order to increase awareness and understanding. Further more detailed guidance for modelling for specific purposes is contained elsewhere. When starting / procuring modelling works you should always ensure you have used the appropriate detailed specification.

1.2 Modelling and Flood Risk Assessment

It should be recognised that it is not always necessary to produce a hydraulic model for all flood risk assessments. A decision on whether to construct a model should be made based on the scale and nature of the potential flood risk, as well as the scale of the project and the existing information available on flood risk. In many less complex assessments simple hydrological and hydraulic analysis may be all that is required. CIRIA Report C624 recommends a staged approach to Flood Risk Assessment. Following such a staged approach allows the need for a model, and the extent of such a model, to be determined. If there is any doubt whether a model is required, this should be discussed with local Environment Agency Staff (Development Control Teams for Land Use Planning, Flood Risk Mapping & Data Management Teams for other) at the earliest opportunity. Suitable information to assist with the modelling may also be available so early dialogue is recommended.

However, even if a model is not constructed, an assessment of the impact of any proposed development on runoff should be carried out using Flood Estimation Handbook2 (FEH) techniques in almost all cases. DEFRA/Environment Agency R&D Technical Report W5-074/A “Preliminary Rainfall Run-off Management for Developments”3 provides further information on runoff assessment for developments.

1.3 Appropriate Modelling Staff Involved

Suitably qualified and experienced personnel should be used to carry out the work described in this document.

1 Lancaster, J., Preene,M. and Marshall,C. 2004, CIRIA Report C624, Development and Flood Risk – Guidance for the Construction Industry, CIRIA, London. 2 Centre for Ecology and Hydrology, 1999. Flood Estimation Handbook. Wallingford, CEH. Further details are available at http://www.nwl.ac.uk/feh/ or from CEH on 01491 838800 3 HR Wallingford (2004) Preliminary rainfall runoff management for developments: Users Guide. Defra / EA R&D Technical Report W5-074/A, HR Wallingford, Wallingford.

4 Environment Agency 1.4 Requirements at Specific Locations

Requirements at specific locations should always be discussed with local Environment Agency staff to ensure that any site-specific factors are identified, which may require special treatment when carrying out the modelling.

2.0 OBJECTIVES OF THE MODEL STUDY

The objectives and the required outputs of the modelling exercise should be defined at the outset. These should be reviewed at regular intervals and at completion.

At an early stage, the design condition should be clarified. This may, for example, include a freeboard and an allowance for climate change. Further information on freeboard is in R&D W1874.

3.0 MODEL BUILDING

A one-off request for information held by the Environment Agency at the very beginning of the project is recommended since this affects selection of method etc, and could prevent further information coming to light at a later stage and complicating matters.

3.1 Choice of Model

The modelling software chosen should be capable of producing the required output. It will generally be appropriate to choose commercial hydraulic/river modelling software that is in widespread use. In certain circumstances, for example where the applicability of a model to a specific situation has not been previously demonstrated, it may be necessary for those conducting the flood risk assessment (FRA) to have independent benchmarking tests carried out to demonstrate model performance using standard data. Examples of how this may be achieved under a range of scenarios are provided in the Defra/Environment Agency R&D Report 'Benchmarking of hydraulic river modelling software packages' (W5-105) which is available via the Joint Defra/Environment Agency Flood and Coastal Erosion Risk Management R&D Programme website.5

In reporting on any hydraulic modelling carried out as part of the FRA, a technical description of the model should be provided, including the name and version of the software used, referring to published papers/reports where appropriate to provide technical detail and to demonstrate the applicability of the model(s) to the situation in question. These references may need to be provided to the Environment Agency if required. If no publications are available then a more detailed technical description should be provided within the FRA, along with examples of relevant previous applications and/or the results produced by applying the model to standard tests (as outlined above, or similar).

Also, at this stage, the choice should be made between a fully hydrodynamic 1D or 2D model or a steady-state backwater model, flood routing model or combination of methods.

A full hydrodynamic model must be used if the study area contains either structures whose operation varies with time (e.g. pumps, sluices, and tidal outfalls) or a tidal estuary where tidal water levels increase going up the estuary 6. This should also be employed in complex tidal/fluvial situations and where the watercourse is subject to rapid increases and decreases in flow. If there is significant floodplain storage and complex flow routes on the floodplain then 2D modelling of the floodplain may be more representative. In other cases, either a steady-state or hydrodynamic model may be chosen. It should be noted that a steady-state model is unlikely to give a reasonable estimation of water levels where storage is present.

4 Environment Agency: Fluvial Freeboard Guidance Note. Technical Report W.187.2000. 5 Flood & Coastal Defence R&D Programme, Benchmarking Hydraulic River Modelling Software Packages, R&D Study: W5-105/TR1, Defra / EA, March 2004. 6 This is typically the case in estuaries of significant rivers and can be seen by inspection of the tide tables.

Environment Agency 5 3.2 Survey Data

The model should be based on a topographic survey of the watercourse. The upstream and downstream limits should be defined by the objectives of the flood risk assessment, rather than to the limits of the project / study area (see Section 3.7). The lateral extent of the survey should be sufficient to include the full extent of flooding. Guidance on this extent may come from flooding records and from the Flood Map. The extent of the survey work should be defined jointly by those undertaking the river modelling and those undertaking the survey in conjunction with advice from Environment Agency Flood Risk Mapping & Data Management staff.

The survey (and the model on which the survey is based) should continue far enough downstream so that uncertainty in the boundary condition does not significantly influence the estimated flood levels.

The cross sections surveyed should be representative of the channel and floodplain and the spacing between cross sections and orientation should be determined from the appropriate software documentation and textbooks7. Consideration shall be given to the additional survey information that may be required between cross-sections in areas where detailed flood depths or extents are needed. This can be achieved by either adding further cross sections or surveying additional spot levels.

During the survey, information on structures, flood routes, potential blockages / obstructions to the channel and channel roughness should also be gathered.

Survey data should be obtained using dual frequency GPS equipment, however, some minor and low risk developments do not justify the cost and time required to produce this type of survey. In these cases it may be acceptable to base the survey on OSBMs and this is at the discretion of the Agency’s Development Control Officer based on the appropriateness ‘test’ in PPG25.

All levels must based on Ordnance Survey Datum (further guidance on survey standards should be obtained by reference to the Environment Agency National Survey Specification). All cross sections and other survey information shall be located in plan relative to the National Grid. It is considered best practice that an insured and Chartered Land Surveyor complete the Survey.

3.3 Hydrometric Data

The Environment Agency may hold existing hydrographic and floodplain survey data which may be of use in a flood risk assessment. Environment Agency staff may be able to provide further information on the appropriateness of survey.

River flow, river level and rainfall data relevant to the model should be collected where these exist. The prime source of this data will be the Environment Agency. An understanding of the uncertainty and confidence within this data should be developed.

Another source of hydrological data is data contained within the Flood Estimation Handbook. The UK HiFLOWS Project also provides up to date information.

3.4 Historic Information

Information on historic flooding (e.g. newspaper articles, photos, flood marks) should be collected and utilised to guide the survey extent and to aid the modelling process. Such data is particularly valuable as it can provide information on historic flooding prior to the periods covered by hydrometric data. A search of the Internet can often provide useful information8. However, the effect of any alterations and additions to the watercourse and associated structures since the date

7 For example, the online manuals supplied with modelling software 8 The Chronology of British Hydrological Events, http://www.dundee.ac.uk/geography/cbhe , may contain some useful information

6 Environment Agency of the recorded event needs to be considered. Historic information is likely to be held by the Environment Agency Area office.

3.5 Previous Modelling

The Environment Agency may hold existing river models that may be of use in a flood risk assessment. Such models may, for example, have been produced during previous flood risk mapping studies, the design of flood alleviation schemes and/or previous flood risk assessments in the area.

Where existing models are available, consideration should be given as to whether these could be used as part of the flood risk assessment. You should be aware that there may be cost, licensing and intellectual property rights (IPR) issues associated with the use of models which will need to be resolved before any previous modelling is used.

If models or survey data are provided by the Environment Agency or third parties it is recommended that check surveys are undertaken at key locations to ensure that the data provided is compatible with current conditions.

The Environment Agency may not own the Intellectual Property Rights to hydraulic models that it holds. We therefore may not be able to release information with a licence for its use.

Ownership of the IPR or an approved IPR licence will be required by the Environment Agency if it is planned to use the modified model to update the Environment Agency’s flood risk mapping products and risk assessment products to represent the as built situation.

3.6 Hydrological Assessment

A hydrological assessment of the flood flows should be made using the methodology described in the Flood Estimation Handbook and the Environment Agency’s Guidelines on use of the Flood Estimation Handbook 9.

The hydrological assessment should use, wherever available, local data to improve the estimation of flood flows.

If a hydrodynamic model is used for the modelling, the hydrological assessment should include consideration of peak flows, flood volumes and shape of the flood hydrograph. If the problem includes storage (e.g. reservoir storage or a tide-locked watercourse) it is essential that the critical duration storm for storage (which often differs from the critical duration for peak flow) is identified. If a steady-state model is used, this may be limited to just consideration of peak flows.

Hydrological inputs should be estimated for a range of return periods up to and including the design flow (typically the flow with an annual probability of exceedence of 1%), and should include an appropriate allowance for climate change.

3.7 Model Building

It may be appropriate to speak to Area Environment Agency staff prior to commencing any model building.

(a) General

The model should be built to represent the key flood flow routes, flood storage and structures in the study area. The defined study area should be sufficient to demonstrate the effects of any development on locations away from the site of the proposed development.

9 Environment Agency, 2000. Flood Estimation Handbook Guidelines (Parts 1 and 2) Bristol, Environment Agency

Environment Agency 7 (b) Upstream Boundaries (Inflows)

The upstream boundary or boundaries should be developed under the hydrological assessment described in Section 3.6. For some models, one single upstream inflow per flood event may be sufficient, whilst for others, many upstream boundaries may be needed if a number of tributaries or other inflows are present. The choice of location of the upstream boundaries should be based on hydraulic considerations, not on the upstream limit of the development. The upstream boundary should be far enough upstream to allow the full impact of the development on upstream water levels to be identified.

(c) Downstream Boundary (Levels)

The downstream boundary should be at a location where the relationship between level and flow is well defined, e.g. a weir. Where this is not possible, it should be sufficiently downstream of the area of interest so that any errors in the boundary will not significantly affect predicted water levels at the proposed development site. For a typical fluvial river, a rule of thumb is that a backwater effect extends a length L=0.7D/s, where D = bankfull depth and s = river slope. Hence if the downstream boundary is greater than L from the site it is likely that any errors in the rating curve at the boundary will not affect flood levels at the site. If the downstream boundary is tidal, it should be a location where a tidal curve can be accurately defined. Any tidal boundary should take into account both the astronomical tide (i.e. the tide caused by the gravitational effects of the Moon and the Sun and reported in published tide tables) and storm surges (i.e. the elevation of tidal levels caused by weather conditions). Careful consideration of combined probabilities10 may be required in such cases. The Environment Agency holds extensive extreme tide information from Flood Risk Mapping Studies.

(d) Hydraulic Coefficients

The coefficients used in the model (e.g. channel roughness, weir coefficients) should be determined with guidance from standard textbooks. These texts should be referenced in the modelling report. Work is ongoing to produce guidance relevant to the UK, but in the meantime standard works such as Chow11 and Hicks & Mason12 can provide some guidance. Further information on roughness can also be obtained from the Defra / Environment Agency Conveyance Estimation System (CES) – http://www.river-conveyance.net/.

4. MODEL CALIBRATION, VERIFICATION AND SENSITIVITY TESTING

4.1 Calibration

Wherever practicable, the hydrological assessment and the hydraulic model should be calibrated against recorded flows and/or water levels from observed flood events. If calibration data is available, the model should be calibrated using at least three separate events. If no calibration data is available, a ‘reality check’ on the predicted levels and flows can often be carried out from photographs, historic information and anecdotal accounts of flooding.

The coefficients used in the calibration process should only be varied within the possible ranges suggested in the standard textbooks. The calibration of steady-state models should consider flow and flood levels. Calibration of hydrodynamic models should also consider the timing of the flood peak, flood volume and shape of the flood hydrograph.

4.2 Verification

If calibration is carried out, at least one separate observed event should be run through the model after the calibration to verify the adjustment of parameters.

10 Defra / EA R&D Programme. Joint probabilities - dependence mapping & best practice, FD 2308/TR1. HR Wallingford. 2003. 11 Ven Te Chow, Open Channel Hydraulics, McGraw-Hill 1959. 12 D.M.Hicks & P.D.Mason. Roughness Characteristics of New Zealand Rivers. 1999.

8 Environment Agency 4.3 Sensitivity Testing

The model should be tested by adjusting the key parameters within it. These parameters should include at least model inflows, downstream boundary condition, channel roughness and key structure coefficients. The range of parameters used in sensitivity tests should reflect uncertainties, possible changes due to climate change and variations in hydraulic coefficients (e.g. from seasonal changes or periodic maintenance).

Sensitivity to blockage of critical structures should also be tested. R&D W5A-06113 includes current understanding & some interim guidance.

5. REPORTING

5.1 General

A report should be written describing the modelling. The objective of this report is to enable an evaluation of the model and results to be carried out if necessary. It also should be a self-contained report that will provide sufficient information to allow future use of the model by the Environment Agency including if necessary replicating the work undertaken. The detail of the report should be appropriate to the complexity of the modelling.

5.2 Items to be Included

The key items to be included in the report are:

Statement of Objectives The report should provide an explanation of the reasons the modelling exercise has been undertaken and the planned objectives of the exercise. It should indicate any deviations from the original objectives or planned project outputs, and outline the reasons why these occurred.

Method statement and Justification The report should include a clear method statement, which makes it clear how the modelling has been carried out to fulfil the objectives.

A justification of the methodology should also explain why the model has been used for this application, giving detailed reasons why the modelling tool is applicable/appropriate to the situation (e.g. fully dynamic or steady-state backwater model). It should indicate any perceived advantages or disadvantages of applying the chosen tool.

Technical description Only a brief technical description is required if the tool is well known to the Environment Agency / widely applied (seek advice from Environment Agency staff). If the model is less widely known or applied, then a more detailed development history is required, giving examples of previous applications. The version number of the model used should be reported, and how the model outputs compare with those of other packages when applied to standard tests (see 3.1 above).

The schematic showing how individual parts of the model are connected should be provided.

Data sources All data used in the model must be listed in reports and made available for inspection.

Methods of data capture and/or sources of data must be made clear in the report, as should the processes by which the raw data were converted.

Any reference to earlier work should be clearly referenced, and applications or development of existing models should be subject to the same rigorous inspection methods.

13 Scoping study into the hydraulic performance of bridges and other structures, including effects of blockage, at high flow. EA/Defra R&D Programme. July 2004.

Environment Agency 9 The ownership of the data collected and the format of the data should be stated.

Uncertainty in data sources should be referenced especially where data have been discounted due to low confidence.

Parameters The derivation of the parameters (e.g. channel roughness) used within both the hydrological assessment and the hydraulic model should be stated.

Calibration/Verification Where calibration has been undertaken, the method used must be clearly illustrated and the number of independent data sets used for verification must be displayed. The model results must be presented against observed values for key locations for each verification data set, and descriptive statistics applied to describe the error band in the model.

Sensitivity Analysis The results of the sensitivity testing should be described and the potential effect these could have on the model output should be discussed.

Audit Trail The audit trail developed should be described in unambiguous detail.

Limitations Any limitations of the model or modelling technique should be highlighted. The impact of such limitations on the present or future use should be clearly stated.

Conclusions The report shall include concluding remarks, which highlight key issues from other sections and draw attention to the critical locations and/or structures within the model.

Where in the above section (5.2), the model is referred to this should be taken to include the hydrological assessment. The hydrological assessment must be reported to the same level of detail as the hydraulic modelling. The same key items will apply to both modelling and hydrology.

5.3 Format of Reporting

The report should be in a format that is easy to copy and transmit electronically, and must include all plans and schematics. Adobe pdf files are therefore preferred.

5.4 Other Deliverables

Copies of the model data files should be supplied together with sufficient instructions to allow these models to be run and viewed, for example, a text file containing timestep, runtime etc. A data file containing initial conditions should also be provided.

5.5 Future Use

A statement should accompany the report and model data on the allowable future uses of the model and its associated documentation.

Ownership of the Intellectual Property Rights (IPR) or an approved IPR licence will be required by the Environment Agency if it is planned to use the modified model to update the flood risk mapping products and risk assessment products to represent the as built situation.

6. QUALITY ASSURANCE AND AUDIT TRAIL

Throughout the study, a well-defined audit trail should be defined and reported. This should include all relevant documentation and should link with the appropriate quality assurance procedures of the organisation carrying out the study. Provision should be made to make the relevant documentation available to others who may use the model in future.

10 Environment Agency Glossary of terms

Backwater Curve - The longitudinal profile of the water surface (in a non-uniform flow in an open channel) when the water surface is not parallel to the river bed. This is caused by a restriction such as a dam or weir, increasing the depth of the water above the normal water level that would result if the restriction were removed.

Backwater Effect - The effect where a dam or other restriction raises the surface of the water upstream from it above the normal water level.

Backwater Flooding - Flooding caused by downstream conditions such as a channel restriction and/or high flow in a stream at a confluence downstream of the flooding.

Backwater Model – A model built to represent the backwater effect.

Calibration – The process of adjusting parameter values in a model to try and match recorded data, so that the model can be taken as a good representation of reality.

Combined Probability – The chance of two or more independent events occurring concurrently.

Critical Duration Storm – The duration of storm necessary to produce the maximum instantaneous peak flow or volume at a specific location in a drainage system, for any given flood event probability.

Floodplain – Land adjacent to a watercourse over which water may flow in time of flood. This generally includes the defended floodplain, an area over which water would flow if flood defences were not present, or if flood defences fail.

Flood Routing Model – Process of determining progressively the timing, shape, and amplitude of the flow in a flood wave as it moves downstream at successive points along the river.

Hydrological Model – A mathematical model used to estimate the flow in a river that will result from rainfall. It will usually be based on such things as catchment size, geology and soil type, steepness, land use and storage within the catchment. The model will be calibrated and verified using recorded rainfall and flows, before using design rainfall to estimate the flows which might be expected in of different probabilities.

Hydraulic Model – A mathematical model used to predict possible future levels (and flows in a hydrodynamic model) taking into account the topography, shape and roughness of the river bed and floodplain, obstructions (e.g. weirs and bridges), and the inflows provided by the hydrological model etc. Models are calibrated using recorded historic flood data, where it is available.

Hydrograph – A graph showing the water level (stage), discharge, or other property of the flows in a river, with respect to time.

Hydrological Assessment – Carried out to understand the cycle of precipitation, consequent runoff, infiltration, and storage; eventual evaporation etc.

Intellectual Property Rights – The legal ownership of the content of the work in question.

Storage – Location where water is retained due to the lie of the land, man made influence or effect of tides / other river flows.

Steady-State Model – A hydraulic model in which the flow at any point in the model is constant with time (there can be many different flows but all are constant over time). This type of model cannot estimate the effects of storage on flood levels or downstream flows. Hydrodynamic models estimate flows and levels throughout a flood event, and can therefore take into account the effects of storage on flows and flood levels.

Environment Agency 11 Topographic Survey – Survey to measure and record the physical features of an area in horizontal and vertical dimensions.

Tributary – A river or stream that flows into a larger river.

Upstream / Downstream Boundary – The limits of the model or assessment upstream and downstream of the site of interest.

Verification – The process of checking the accuracy of the outputs of the calibrated model in comparison with recorded data. If sufficient data is available it is good practice to calibrate the model using some recorded data, and verify the model using data from other flood events. List of abbreviations

PPG25 – Policy Planning Guidance Note 25 TAN15 – Technical Advice Note 15 CIRIA – The Construction Industry Research and Information Association DEFRA – Department for Environment, Farming and Rural Affairs R&D – Research and Development 1D – One Dimensional 2D – Two Dimensional FRA – Flood Risk Assessment References

Lancaster, J., Preene,M. and Marshall,C. 2004, CIRIA Report C624, Development and Flood Risk – Guidance for the Construction Industry, CIRIA, London.

Centre for Ecology and Hydrology, 1999. Flood Estimation Handbook. Wallingford, CEH. Further details are available at http://www.nwl.ac.uk/feh/ or from CEH on 01491 838800.

HR Wallingford (2004) Preliminary rainfall runoff management for developments: Users Guide. Defra / EA R&D Technical Report W5-074/A, HR Wallingford, Wallingford.

Environment Agency: Fluvial Freeboard Guidance Note. Technical Report W.187.2000.

Defra / EA, March 2004. Flood & Coastal Defence R&D Programme, Benchmarking Hydraulic River Modelling Software Packages, R&D Study: W5-105/TR1.

Environment Agency, 2000. Flood Estimation Handbook Guidelines (Parts 1 and 2) Bristol, Environment Agency.

Defra / EA R&D Programme. Joint probabilities - dependence mapping & best practice, FD 2308/TR1. HR Wallingford. 2003.

Ven Te Chow, Open Channel Hydraulics, McGraw-Hill 1959.

D.M.Hicks & P.D.Mason. Roughness Characteristics of New Zealand Rivers. 1999.

EA/Defra R&D Programme. July 2004. Scoping study into the hydraulic performance of bridges and other structures, including effects of blockage, at high flow.

12 Environment Agency Would you like to find out more about us, or about your environment?

Then call us on 08708 506 506 (Mon-Fri 8-6) email [email protected] or visit our website www.environment-agency.gov.uk incident hotline 0800 80 70 60 (24hrs) floodline 0845 988 1188

Environment first: This publication is printed on paper made from 100 per cent previously used waste. By-products from making the pulp and paper are used for composting and fertiliser, for making cement and for generating energy.

Environment Agency 13 Environment Agency OFFICIAL

Flood risk assessments: Climate change allowances Application of the allowances and local considerations East Anglia; Essex, Norfolk, Suffolk, Cambridgeshire and Bedfordshire

1) The climate change allowances

The National Planning Practice Guidance refers planners, developers and advisors to the Environment Agency guidance on considering climate change in Flood Risk Assessments (FRAs). This guidance was updated in February 2016 and is available on Gov.uk. The guidance can be used for planning applications, local plans, neighbourhood plans and other projects. It provides climate change allowances for peak river flow, peak rainfall, sea level rise, wind speed and wave height. The guidance provides a range of allowances to assess fluvial flooding, rather than a single national allowance. It advises on what allowances to use for assessment based on vulnerability classification, flood zone and development lifetime.

2) Assessment of climate change impacts on fluvial flooding

Table A below indicates the level of technical assessment of climate change impacts on fluvial flooding appropriate for new developments depending on their scale and location. This should be used as a guide only. Ultimately, the agreed approach should be based on expert local knowledge of flood risk conditions, local sensitivities and other influences. For these reasons we recommend that applicants and / or their consultants should contact the Environment Agency at the pre- planning application stage to confirm the assessment approach, on a case by case basis. Table A defines three possible approaches to account for flood risk impacts due to climate change, in new development proposals: . Basic: Developer can add an allowance to the 'design flood' (i.e. 1% annual probability) peak levels to account for potential climate change impacts. The allowance should be derived and agreed locally by Environment Agency teams. . Intermediate: Developer can use existing modelled flood and flow data to construct a stage- discharge rating curve, which can be used to interpolate a flood level based on the required peak flow allowance to apply to the ‘design flood’ flow. . Detailed: Perform detailed hydraulic modelling, through either re-running Environment Agency hydraulic models (if available) or construction of a new model by the developer.

Table A – Indicative guide to assessment approach

VULNERABILITY FLOOD DEVELOPMENT TYPE

CLASSIFICATION ZONE MINOR SMALL-MAJOR LARGE-MAJOR Zone 2 Detailed ESSENTIAL Zone 3a Detailed INFRASTRUCTURE Zone 3b Detailed Zone 2 Intermediate/ Basic Intermediate/ Basic Detailed HIGHLY Zone 3a Not appropriate development VULNERABLE Zone 3b Not appropriate development Zone 2 Basic Basic Intermediate/ Basic MORE Zone 3a Intermediate/ Basic Detailed Detailed VULNERABLE Zone 3b Not appropriate development Zone 2 Basic Basic Intermediate/ Basic LESS Zone 3a Basic Basic Detailed VULNERABLE Zone 3b Not appropriate development Zone 2 None WATER Zone 3a Intermediate/ Basic COMPATIBLE Zone 3b Detailed Note: Where the table states 'not appropriate development', this is in line with national planning policy. If in exceptional circumstances such development types are proposed in these locations, we would expect a detailed modelling approach to be used.

Environment Agency March 2018 Environment Agency OFFICIAL

NOTES:

. Minor: 1-9 dwellings/ less than 0.5 ha | Office / light industrial under 1ha | General industrial under 1 ha | Retail under 1 ha | Gypsy/traveller site between 0 and 9 pitches . Small-Major: 10 to 30 dwellings | Office / light industrial 1ha to 5ha | General industrial 1ha to 5ha | Retail over 1ha to 5ha | Gypsy/traveller site over 10 to 30 pitches . Large-Major: 30+ dwellings | Office / light industrial 5ha+ | General industrial 5ha+ | Retail 5ha+ | Gypsy/traveller site over 30+ pitches | any other development that creates a non residential building or development over 1000 sq m.

The assessment approach should be agreed with the Environment Agency as part of pre- planning application discussions to avoid abortive work.

3) Specific local considerations

Where the Environment Agency and the applicant and / or their consultant has agreed that a ‘basic´ level of assessment is appropriate the figures in Table B below can be used as a precautionary allowance for potential climate change impacts on peak ‘design’ (i.e. 1% annual probability) fluvial flood level rather than undertaking detailed modelling.

Table B – Local precautionary allowances for potential climate change impacts

Essex, Norfolk and Suffolk

Hydraulic Model (Watercourse) Central Higher Central Upper Blackwater & Brain - 500mm 600mm 900mm Blackwater between TL7520925623 and TL7820324314 Brain between TL7373323312 and TL7683821321 Gipping – Downstream of Needham Market 400mm 500mm 850mm Gipping – Needham Market and upstream including 200mm 250mm 400mm Somersham W/C Broads (2008 Model Extent) Please use the current 1 in 1000 (0.1%) annual Bure and Ant (2012 Model Extent) probability including climate change allowance For other main rivers, tributaries and ordinary watercourses that are not stated above, basic allowances have not been calculated. In this instance you can either:  If flow data is available you can request this Other main rivers, tributaries and ordinary data from us and can conduct an watercourses intermediate assessment yourself  Or alternatively, you can choose to undertake a Detailed Assessment and “perform detailed hydraulic modelling, through either re-running our hydraulic models (if available) or constructing a new model

Environment Agency March 2018 Environment Agency OFFICIAL

Cambridgeshire and Bedfordshire

Watercourse / Model Central Higher Central Upper End Alconbury Brook 600mm 700mm 900mm River Kym Lower Ouse (Model 700mm 800mm 1100mm Extent) Mid Ouse (Cold 700mm 800mm 1100mm Brayfield to Bromham – between SP9156852223 and TL0132950919) Mid Ouse (East of 700mm 850mm 1200mm Bedford to Roxton – between TL0791848903 and TL1618854543) River Hiz and River 400mm 450mm 550mm Purwell River Ivel 500mm 600mm 750mm Pix Brook 450mm 500mm 600mm Potton Brook 500mm 600mm 700mm River Cam and 600mm 700mm 950mm tributaries (excluding the Cam Lodes and the Slade System) Great Barford (ordinary 500mm 550mm 650mm watercourses) Bromham (ordinary 550mm 650mm 850mm watercourse)

NOTES:

Urban areas excluded from the ‘basic’ approach: St Ives, Holywell, Godmanchester, Swavesey, Over, Bedford, Newport Pagnell, Buckingham and Leighton Buzzard. More detailed assessment of climate change allowances will need to be undertaken in these locations.

Use of these allowances will only be accepted after discussion with the Environment Agency.

Environment Agency March 2018 Environment Agency OFFICIAL

4) Fluvial food risk mitigation

For planning consultations where we are a statutory consultee and our Flood risk standing advice does not apply we use the following benchmarks to inform flood risk mitigation for different vulnerability classifications. These are a guide only. We strongly recommend you contact us at the pre-planning application stage to confirm this on a case by case basis. For planning consultations where we are not a statutory consultee or our Flood risk Standing advice applies we recommend local planning authorities and developers use these benchmarks but we do not expect to be consulted.

 For development classed as ‘Essential Infrastructure’ our benchmark for flood risk mitigation is for it to be designed to the ‘upper end’ climate change allowance for the epoch that most closely represents the lifetime of the development, including decommissioning.

 For highly vulnerable or more vulnerable developments in flood zone 2, the ‘central’ climate change allowance is our minimum benchmark for flood risk mitigation, and in flood zone 3 the ‘higher central’ climate change allowance is our minimum benchmark for flood risk mitigation. In sensitive locations it may be necessary to use the higher central (in flood zone 2) and the upper end allowance (in flood zone 3).

 For water compatible or less vulnerable development (e.g. commercial), the ‘central’ climate change allowance for the epoch that most closely represents the lifetime of the development is our minimum benchmark for flood risk mitigation. In sensitive locations it may be necessary to use the higher central (particularly in flood zone 3) to inform built in resilience.

For a visual representation of the above, please see Tables 1 and 2 overleaf.

5) Development in Tidal Areas There is no change to the way we respond to sites affected solely by tidal flood risk as the sea level allowances are unchanged.

6) Our Service

Non-chargeable service

We will give a free opinion on:

• What climate change allowance to apply to a particular development type

• Which technical approach is suitable in the FRA

Chargeable service:

• Review of climate change impacts using intermediate and detailed technical approaches (i.e. modelling review)

• Assessment and review of proposals for managed adaptation.

Environment Agency March 2018 Environment Agency OFFICIAL

Table 1 peak river flow allowances by river basin district (use 1961 to 1990 baseline) River Allowance category Total potential Total potential Total basin change change potential district anticipated for anticipated for change ‘2020s’ ‘2050s’ anticipated (2015 to 39) (2040 to 2069) for ‘2080s’ (2070 to 2115) Anglian Upper end 25% 35% 65% Higher central 15% 20% 35% Central 10% 15% 25% Thames Upper end 25% 35% 70% Higher central 15% 25% 35% Central 10% 15% 25%

Table 2: Using peak river flow allowances for flood risk assessments Flood Essential Highly More Less Water Zone Infrastructure Vulnerable Vulnerable Vulnerable Compatible 2 higher central higher central central and central none of the and upper end and upper end higher central allowance allowances allowances allowances allowances 3a upper end X higher central central and central allowance and upper end higher central allowance 3b upper end X X X central allowance allowance X – Development should not be permitted If (exceptionally) development is considered appropriate when not in accordance with flood zone vulnerability categories, then it would be appropriate to use the upper end allowance.

There may be circumstances where local evidence supports the use of other data or allowances. Where you think this is the case we may want to check this data and how you propose to use it.

Environment Agency March 2018 Environment Agency OFFICIAL

Appendix 1 – Further information on the Intermediate approach. 1) The methodology the chart is based on does not produce an accurate stage-discharge rating and is a simplified methodology for producing flood levels that can be applied in low risk small-scale development situations; 2) The method should not be applied where there is existing detailed modelled climate change outputs that use the new allowances. In such circumstances, the ‘with climate change’ modelled scenarios should be applied.

An example stage-discharge relationship is shown below.

Environment Agency March 2018

Flood Estimation Guidelines

Technical guidance 197_08 Issued: 21/01/2015

What’s this Offers advice to help analysts make the most of the material in the Flood document Estimation Handbook (FEH), other recent publications and older methods of about? flood estimation, when they're still applicable. It aims to ensure a consistent, robust approach, repeatable results and systematic recording of decisions made. It complements the FEH and other publications. It is not an alternative short guide.

Who does this All staff carrying out flood estimation in the Environment Agency. apply to? Staff supervising studies or reviewing those carried out externally. Managers of flood estimation studies, who should read at least the executive summary. Consultants carrying out work for us or carrying out work requiring our approval.

Contact for • Craig Jones Team Leader Inland Hydrology queries and • Inland Hydrology team feedback • Anonymous feedback for this document can be given here

Doc No 197_08 Version 5 Last printed 12/02/15 Page 1 of 110

Table of contents

Contents Executive summary ...... 4 Introduction ...... 7 Development of the Flood Estimation Guidelines ...... 7 Using the FEH and these guidelines ...... 9 Competencies and Training ...... 10 Hydrometric data and catchment descriptors ...... 13 Selecting and examining flood peak data ...... 13 Rating Reviews ...... 17 Flood Event Data ...... 20 Flood History ...... 21 Catchment Descriptors ...... 23 Choice of methods ...... 26 Overview ...... 26 A framework for choosing a method ...... 27 The need to think ...... 29 Preparing method statements ...... 29 Choosing between the FEH methods ...... 30 Hybrid methods ...... 32 Checking results ...... 34 Conclusion ...... 35 Advice and cautions on FEH and ReFH methods...... 36 Overview ...... 36 Design rainfall ...... 37 Statistical Method - general ...... 38 Statistical method – index flood, QMED ...... 38 General information ...... 39 From catchment descriptors ...... 39 Urban adjustment ...... 40 Data Transfer...... 42 Statistical methods - growth curves ...... 49 Pooling groups ...... 49 Pooled growth curves ...... 51 Growth curves for sites with flood peak data ...... 53 Rainfall-runoff approaches ...... 56 General information ...... 57 Lumped or distributed approach? ...... 57 The ReFH method ...... 60 FEH rainfall-runoff method ...... 64 Continuous simulation - an alternative rainfall-runoff approach ...... 65 Assumptions, limitations and uncertainty ...... 65 Overview – a common criticism...... 65 Why bother with uncertainty? ...... 66

Doc No 197_08 Version 5 Last printed 12/02/15 Page 2 of 110

Typical assumptions ...... 67 Typical limitations ...... 68 Assessing uncertainty ...... 69 Application-specific guidance ...... 72 Overview ...... 72 Flood mapping and hydrodynamic models ...... 72 Catchment-wide studies ...... 74 Post-event analysis ...... 75 Modelling effects of land-use change on flooding ...... 76 Pumped and other low-lying catchments ...... 76 Water level management plans and short return period estimates...... 79 Small catchments and greenfield runoff ...... 80 Rational method ...... 81 ADAS Report 345 ...... 82 Institute of Hydrology Report 124 ...... 83 Development control and urban catchments ...... 84 Up to moderately urbanised catchments ...... 85 Heavily or very heavily urbanised catchments ...... 86 Extremely heavily urbanised catchments ...... 87 Flood FReQuency SIMulation (FRQSIM) ...... 87 Example application of revised ReFH method ...... 89 Permeable catchments ...... 92 Catchments containing reservoirs ...... 95 Flood estimation for reservoir safety ...... 97 Estimating long return period floods (150-1000 years) ...... 99 Audit trail ...... 102 Overview ...... 102 Flood estimation calculation record ...... 102 Filling in the calculation record ...... 102 Presenting results ...... 104 Recording the data used ...... 104 List of acronyms ...... 105 Related documents ...... 107

Doc No 197_08 Version 5 Last printed 12/02/15 Page 3 of 110

Executive summary

Why we've This executive summary gives a brief overview, intended mainly for included this managers of flood estimation studies. summary

If you think it's Although you can apply many of the FEH methods using straightforward easy – you're software, flood estimation is a complex process with many aspects. not looking Practitioners need many skills, including statistics, mathematical modelling, deeply enough fluvial hydraulics and meteorology, and hydrology. An enquiring mind is essential and a determination to challenge assumptions and seek out facts. Analysts need to think, at all stages, about the problem they are solving. So it's essential to ensure that those carrying out studies have the right knowledge, skills and experience and that they are allowing enough time for the task. Half a day may be just adequate for a preliminary assessment. But thorough flood estimation studies can take many days or weeks - the FEH suggests allowing between five and 50 days. Table 2 shows indicative levels of staff competence and timescales for different types of flood estimation studies. You must take a risk-based approach when considering the required competence and the time needed to carry out a study.

What to expect We've designed these guidelines to complement the FEH and other and not expect publications. They're not an alternative short guide. Analysts still need to consult the FEH. We encourage all users to read at least Volume 1, which has only 61 pages, including a thought-provoking and frank interlude. In line with the philosophy of the FEH, the guidelines offer few prescriptive instructions. Example: In many situations, there's a choice of FEH methods and alternatives, sometimes giving a wide variety of results. These guidelines don't tell users which method to choose. But they do offer a framework for choosing a method and they give advice on: • the ranges of applicability of each method; • how to write a method statement; • factors to consider when choosing a method; • how to reconcile results from different methods; • which methods to prefer for various unusual types of catchment; • How to record and justify the choice of method. The guidelines are intended mainly for river management and reservoir safety applications. They cover estimation of design floods over a range of annual exceedence probabilities up to the probable maximum flood.

How do I make Much of our involvement with flood estimation comes from reviewing studies sense of this carried out by consultants. Before we revised these guidelines in 2006- hydrology 2008, we consulted a sample of Environment Agency staff. They mentioned report? 18 typical shortcomings in flood hydrology reports. The most common were lack of information on assumptions, limitations of the methods and poor justification for the choice of method.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 4 of 110

The guidelines now address these and other comments by including sections on assumptions and limitations (see Chapter 5,), a new flood estimation calculation record (SD01) and a Checklist for reviewing flood estimates (SD03). The flood estimation record is for use on all Environment Agency studies, whether carried out internally or by our consultants. As well as assisting reviewers and project managers, it is also designed to help analysts ensure that they have thought through the choice of approach and applied the methods correctly. Analysts have a responsibility to establish this audit trail. Project managers are responsible for defining the purpose of the flood estimates they need and ensuring that they are used appropriately.

One minute There are two principal techniques available: overview of • the FEH statistical method; flood estimation • the Revitalised Flood Hydrograph (ReFH) method. methods This has replaced the FEH rainfall-runoff method for most applications. You can apply these on any UK catchment with an area larger than 0.5 km2. Difference between the two The statistical method gives just a peak flow. The rainfall-runoff techniques (ReFH or FEH) produce hydrographs. Because it is more direct and based on a larger dataset, users often prefer the statistical method. Using a hybrid method If a hydrograph is needed, you can use a hybrid method to fit a hydrograph shape to the peak flow from the statistical method. Other older approaches On smaller catchments (see Small catchments and greenfield runoff) or extremely heavily urbanised areas (see Development control and urban catchments), older approaches are sometimes applied, such as the Institute of Hydrology Report 124 method for small catchments or the ADAS Report 345 method for greenfield runoff estimation. These guidelines recommend that FEH methods should now be used in preference. The FEH also provides rainfall frequency estimates, which are most often used to provide input to rainfall-runoff models for flood estimation.

Catchment The FEH software enables rapid estimation of design floods from catchment descriptors are descriptors. However, these are rarely likely to be the best estimates. a last resort The first of the FEH’s six maxims states that flood frequency is best estimated from gauged data. For this reason, the guidelines offer advice on how to obtain flow data (the principal source is the HiFlows-UK dataset, now hosted by CEH) and how to review data quality, in particular the accuracy of rating equations. The availability and quality of flow data can be the greatest influences on the accuracy of the resulting flood estimate. On ungauged catchments, users can often apply data transfers by seeking nearby hydrologically similar catchments for which flow data are available. Selecting donor catchments is a subjective process. So the guidelines offer advice drawn from the FEH, more recent research and the accumulated experience of many users. Quite, quite Even the 50 days of work the FEH suggests won't produce a definitive sure? statement on the magnitude of a 1% flood or the rarity of an observed event. By its very nature, flood estimation is an uncertain business and the Doc No 197_08 Version 5 Last printed 12/02/15 Page 5 of 110

uncertainty is probably greater than many hydrologists realise. These guidelines offer advice on identifying sources of uncertainty. Confidence limits for flood estimates are difficult to calculate and remain a subject for research. But the FEH offers advice on the uncertainty of some parts of the process and analysts should quote this information. It's important to realise that a wide confidence interval doesn't necessarily mean that the best estimate is wrong. Analysts should aim for the best estimate at each stage in the flood estimation process. This is better then making successive decisions that are biased on the conservative side that could result in a final answer that lies a long way above the best estimate. If required, they can add a factor of safety to the outcome of the design process, such as a freeboard allowance that raises the design height of a flood defence. A degree of pragmatism can be important in flood estimation. Since the answer is always uncertain, the analyst must be able to judge when they've found sufficient information and explored enough options to give a result suitable for the purpose of the study.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 6 of 110

Introduction

Development of the Flood Estimation Guidelines

Revisions The table below describes the publishing history of the Flood Estimation Guidelines.

Version Authors, content and changes 1 Written by Bullen Consultants, with input from the Environment Agency and the Rivers Agency. Issued in August 2000.

2 Produced by the Environment Agency with the help of JBA Consulting. Based on version 1 with extensive revision and updates. Included new material, such as advice on non-FEH methods, reservoir safety methods, guidance on uncertainty and a checklist for reviewing calculations. Merged the two parts of version 1 into a single volume, with Part 1 (Overview) condensed into an executive summary. Issued in 2008.

3 Produced by the Environment Agency with the help of JBA Consulting. Includes research, software and datasets released between 2007 and 2009. Issued in 2009.

4 Produced by the Environment Agency with the help of JBA Consulting. Includes research and datasets released in 2010- 12 and feedback from users. Issued in 2012.

5 Produced by the Environment Agency. Includes research and datasets released since 2012. Issued 2015

Purpose of These guidelines offer advice to help analysts make the most of the material these in the FEH and later publications, as well as older methods of flood guidelines estimation where they are still applicable. Their aim is to ensure a consistent and robust approach, repeatable results and systematic recording of the decisions made. They provide a framework in the form of: • a Flood estimation calculation record (SD01, SD02) to enable robust recording and quality assurance of the results; • and a Checklist for reviewing flood estimates (SD03). Other aspects the guidelines address include levels of competence and supervision.

Scope As Figure 1 below shows, these guidelines concentrate mainly on methods used for flood estimation for river management and reservoir safety, that is, the FEH procedures and the more recent ReFH method. These guidelines also review alternative methods for unusual catchments, such as small ones or lowland areas with pumped drainage. They only briefly mention sewer design methods and alternative approaches to flood estimation, such as continuous simulation.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 7 of 110

Figure 1 This diagram shows applications and methods covered by the guidelines.

Relationship to These guidelines complement the FEH and other publications. They are not FEH an alternative short guide. Analysts: you must read and consult the FEH and other relevant publications. References to the FEH follow conventions used in the FEH. Example: The reference 1 2.2 in these guidelines refers to Volume 1, Section 2.2 in the FEH. In line with the approach adopted by the FEH, these guidelines do not offer prescriptive methods.

Precedence Analysts or project managers: you may sometimes need to depart from these guidelines. When you do, the Project Brief and the Proposal must make this clear. The Project Brief and the Proposal then takes precedence over these guidelines. But, in all cases of apparent difference, consultants and Environment Agency analysts must first seek clarification from the Environment Agency’s Project Manager.

Revisions The original version of these guidelines was published in August 2000. Since then, use of the FEH has become widespread and users have accumulated a great deal more experience. There have been developments in research (for example, the ReFH method and improvements to the statistical method) and data management (such as HiFlows-UK, now hosted by CEH as peak flow data) that have changed the way we use the FEH. There is also an increasing emphasis on catchment-scale flood estimation. For these reasons, we comprehensively revised the guidelines in 2006-7. We broadened the scope to include non-FEH methods when these are still applicable, particularly for very small catchments. We updated the guidelines again in 2009, 2012 and 2014. There will be future revisions following any major changes in methodology or at least every three years.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 8 of 110

Presenting These guidelines quote the frequency of a flood in terms of a return period. return periods Definition The return period of a flood is the average interval between floods of that magnitude or greater. We use return periods to remain compatible with the previous version of the guidelines and with the FEH. See also Note on the definition, immediately below. Alternative expression Alternatively, we can express flood frequency in terms of an annual Exceedence probability (AEP). This is the inverse of the return period. Example: A 1% AEP flood has a 1% chance of being exceeded in any year. Presenting results to non-specialists Use the alternative expression. Non-specialists may associate the concept of return period with a regular occurrence rather than an average recurrence interval. Table 1, below, provides a quick conversion between return periods and AEPs. Note on the definition Strictly speaking, this is the return period on the peaks-over-threshold scale. There is an alternative definition, based on annual maximum floods, which the FEH uses more widely (1 Appendix A). The difference is only important at short return periods, under 20 years. The AEP is the inverse of the annual maximum return period.

Table 1 Return period 2 5 10 25 50 75 100 200 1000 (in years)

AEP (in 50 20 10 4 2 1.33 1 0.5 0.1 percentages)

Using the FEH and these guidelines

Finding The Environment Agency’s focal point for discussion and review of technical information aspects of flood estimation is the Hydrology team within the Modelling and and sharing Forecasting service. Send any suggestions to improve these guidelines by experience e-mail to [email protected]. Consult the FEH page on the Easinet for information relating to FEH technical and software support. It also includes information on our policies, these guidelines and details of training courses. Inconsistencies in flood data are sometimes identified when carrying out flood studies. FEH analysts should provide feedback to the hydrometric section of the relevant gauging authority for these to be investigated. They should submit any errors or suggestions relating to the HiFlows-UK dataset, using the feedback form on CEH's website.

FEH webpages Information about the FEH is provided on the FEH website and the CEH Wallingford website. They include news on updates, frequently asked questions and information on training courses. Select this link for a list of FEH errata/corrigenda on the CEH Wallingford website. Analysts: it is recommended that you make hard-copy corrections to your copy of the FEH.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 9 of 110

Software Currently, the latest releases of the FEH software packages are: • FEH CD-ROM version 3.0 (released in September 2009); • WINFAP-FEH version 3.0.003 (released in November 2009); • ReFH spreadsheet version 1.4; • ReFH Design Flood Modelling Software (released in July 2007). • A number of hydraulic models have the facility to implement the FEH rainfall-runoff method. Notes: Report any installation errors to the Corporate Information Services (CIS) help desk (tel. 8080).

Competencies and Training

Range of skills Flood estimation is complex. There are many aspects to the process. Practitioners need many skills including statistics, mathematical modelling, fluvial hydraulics and meteorology, and hydrology. An enquiring mind is essential and a dogged determination to challenge underlying assumptions in datasets and seek out facts. It is essential, therefore, to ensure that: • the people carrying out studies have the correct knowledge, skills and experience; • and that sufficient time is allowed for the task. See Table 2 for more details.

Competency A disciplined framework for carrying out studies ensures good quality flood framework estimates. It is essential that those who work on, supervise and approve flood studies have suitable training, professional qualifications and experience. Table 2 below, provides an indicative hierarchy of flood estimation studies and the time required for different types of studies. It aims to help: • managers and analysts to discuss the levels of effort and competence required; • and team leaders to allocate staff to studies

Table 2 The table below provides indicative levels of competence and supervision for flood estimation staff. Notes • The values in all columns are indicative. • FM: flood mapping; CFMP: Catchment Flood Management Plan. • Interpret the competence criteria as minimum levels. • An analyst who has not carried out or supervised the study must give approval. • Level 1: hydrologist with minimum approved experience in flood estimation. • Level 2: senior hydrologist. • Level 3: senior hydrologist with extensive experience of flood estimation. Doc No 197_08 Version 5 Last printed 12/02/15 Page 10 of 110

Complexity Example of Value of Indicative Competence of the flood a study flood timescale criteria estimation defence for flood Analyst Super- study works or estimatio vision damage n and s approval Simple Preliminary - <1 day Level 1 Level 2 assessment; culvert capacity check

Routine Low-risk <£50k 1 - 2 days Level 1 Level 2 development application

Moderate Small FM <£250k 2 - 10 Level 2 Level 3 study or days medium-risk development application

Difficult Medium FM <£1 2 - 4 Level 2 Level 3 study or million weeks CFMP or pre- feasibility

Very difficult Major >£1 >1 month Level 3 Level 3 scheme million design or large FM study/ CFMP

Training All Environment Agency users of the FEH must attend an approved training courses course in flood estimation methods. We offer two FEH courses: • FEH Awareness, a 1-day course for project managers and others needing an overview; • FEH Users, a 2-day course for those who will be using FEH methods. The users’ course introduces all the basic techniques and software. It should enable most analysts to reach Level 1 in Table 2. This is a minimum requirement. For complex studies, analysts may require more detailed training in one or more of the FEH techniques or have gained experience under the supervision of senior colleagues. You can find further information on these courses on the FEH page on the Easinet. Neither course covers flood estimation for reservoir safety. More advanced training courses are available from various consultants and academic institutions. Information on other internal hydrological courses can be found in the Learning and Development course directory

Doc No 197_08 Version 5 Last printed 12/02/15 Page 11 of 110

Getting There is no substitute for experience to develop familiarity with the experience challenges of flood estimation. Analysts: you will find that time spent on the worked examples in the FEH is repaid by additional insight into many facets of the FEH methods.

Supervision Supervision, by a more experienced colleague, can provide support and create the opportunity to learn. It enables problems to be shared. This, in turn, may provide reassurance when handling the more knotty aspects of a difficult study. Supervision also provides a quality control mechanism on a day-to-day basis. Project Managers and team leaders: you are responsible for ensuring that staff experienced in flood estimation are adequately supervising all flood studies.

Managing Project Managers: When commissioning a study, you must discuss your studies requirements with the hydrologists (within the Environment Agency or consultants) who will be carrying out and supervising the study. These discussions enable both parties to identify the options available for the study and agree a specification. You can record this specification, usually as the Project Brief and in a Proposal. You can use the Environment Agency’s SFRM Model Report Performance Scope as a starting point. Completing the calculation record establishes an audit trail for every flood estimation study. But there is still a need to monitor the execution of studies to ensure that they are technically correct and meet your needs.

Signing off Supervisors: you must sign off completed studies to certify their technical responsibility basis and validity. Analysts: you must sign off the results of the flood estimation to confirm that they are fit for the purposes of the study.

Consultants Consultants must be able to demonstrate that staff who carry out flood estimation have appropriate qualifications, training, experience and supervision to meet the aims described above in this chapter.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 12 of 110

Hydrometric data and catchment descriptors

Selecting and examining flood peak data

Rationale The availability and quality of flow data can be the greatest influences on the quality of the resulting flood estimate. A review of hydrometric data is, therefore, vital at the outset of most studies. The most useful type of data in flood estimation is normally a peak flow series. However, other sorts of data can also be valuable, including records from stations that measure only water levels.

Available data The FEH provided flood peak data for 1000 gauging stations when it was first published. This dataset ended at about 1995. In 2005, the HiFlows-UK dataset version 1 was released. It was an updated version of the FEH dataset. It contained approximately 1000 sites. Some of the original FEH stations were removed and others were added. HiFlows-UK has since been updated in 2008, 2009 and 2011. The current version (3.3.2 released in April 2014) which contains annual maximum and POT data up to the end of water year 2011-12. All FEH users: you must use the HiFlows-UK dataset as your primary source for flood peak data. You can download the latest version from the HiFlows-UK page on the CEH website. You should overwrite the dataset provided with WINFAP-FEH.

Description of HiFlows-UK, now hosted by CEH under the title of "Peak flow data" builds HiFlows-UK on the dataset assembled for the FEH research. It includes suitable flow sites from all the UK gauging authorities. The website provides peak flows, levels, rating histories, photographs and information on each gauging station. It provides: • guidance on the quality of data; • and a statement indicating whether each station is considered suitable for: • estimating QMED; That is, moderate floods. • and/or pooling. That is, extreme floods. This suitability considers only data quality, not record length or the nature of the catchment.

Guidelines The guidelines and advice in the table below are included to help users. Item Guideline or advice 1 There are two main uses for the HiFlows-UK dataset: • You can use stations suitable for pooling to create pooling groups, by downloading the dataset and saving it to a directory used by WINFAP-FEH. Most users find it convenient to ensure that WINFAP-FEH only uses the subset of stations that are classed as suitable for pooling when it constructs pooling

Doc No 197_08 Version 5 Last printed 12/02/15 Page 13 of 110

groups. You can do this by browsing to the appropriate directory under the Load Options tab in the General options menu in WINFAP-FEH.

• You can consider stations suitable for QMED as potential donor sites. You can locate these using the search facility on the website. 2 For most lower risk studies, you can use the HiFlows-UK dataset without any need for further review or searching for data.

3 If you are using the data in more detailed studies, there are limitations in the dataset to address: • there are other sources of flow data not in the HiFlows-UK dataset; Examples: Recently installed stations, temporary flow loggers and stations that were not judged to be of suitable quality at the time of compiling the dataset. You should investigate all gauging stations at or near the reach of interest because even if their high flow data is inaccurate or uncertain, it may still result in better estimates of QMED than those made solely from catchment descriptors. Even level gauges can be useful sources of evidence for flow magnitudes, for example if you are able to derive an approximate rating equation using spot gaugings or a hydraulic model. • the dataset will typically lag a year or two behind the present, so there will often be scope to update flood peak series; • some stations have flows in HiFlows-UK that currently differ from the data held on the Environment Agency’s Wiski database; • the data quality classification is 'indicative'. More detailed rating reviews are often worthwhile and can result in changes to the classification of stations. 4 In some studies, it is worth updating the flood peak records for stations on the study reach and at donor sites. This is more worthwhile at times when Hiflows-UK is less up to date or when there has been a recent major widespread flood.

5 Temporary flow loggers such as portable ultrasonic meters are worth installing for some studies, particularly if they can be installed at least two years in advance. This provides a long enough flood peak record to give an estimate of QMED that is more reliable than that obtainable from catchment descriptors (3 2.2). On 95% of typical catchments, you can expect catchment descriptors to give an estimate of QMED within about a factor of 2.0 of the real value. With just 2 years of flow data available, this uncertainty reduces to within about a factor of 1.7 of the real value (3 13.8.2. With 5 years of data, the factor drops to 1.4. So installing a temporary flow monitor could make a large difference to the outcome of a study, such as the number of people thought to be at risk of flooding or the level to which a flood defence should be constructed. On unusual catchments such as highly permeable or urban ones, an even shorter period of flow data may provide a more reliable estimate of flood frequency than catchment descriptors, due to the influence of local hydrological features that are not well represented, for example in the UK-average regression formula for QMED. In some unusual catchments you will have to accept a Doc No 197_08 Version 5 Last printed 12/02/15 Page 14 of 110

huge uncertainty in design flood estimates unless you obtain some flow data. Example: Within a month of its installation in 2010, a temporary flow logger installed on a small Magnesian limestone catchment near Doncaster recorded a flood peak that was more than twice the catchment-descriptor estimate of QMED. Although the flow record was too short to draw any statistically significant conclusions, the data cast serious doubt on the FEH estimates and supported the use of an alternative method (continuous simulation).

6 Visual examination of flood peak data is always worthwhile, see Figure 2. Plotting a time series of flood peaks can reveal features such as: • outliers; These are a typical feature of flood peak data but you should investigate them if additional information is available (1 Interlude, p. 33-35). • apparent upper bounds on the magnitude of flood peaks; These may be genuine features due to storage in the catchment or an artefact due, for example, to bypassing the gauging station. • trends or fluctuations; These may be due to climate or land use. Investigation (3 21.4) may reveal no obvious cause for non-stationarity. But when you find a cause, data adjustment or curtailment may be relevant (3 21.1.3). One of the main findings to emerge from an analysis of trends in the original FEH dataset is that national trends in flood peaks, associated with land use change or climate change, cannot be easily identified or readily dismissed (3 21.5.4). • step changes; These may indicate a sudden change in the catchment (such as the construction of a reservoir or flood storage area) or a change in the station or rating which has altered the apparent flows. • unusually small annual maximum flows. This can occur, for example, on a highly permeable catchment that has not experienced a flood in a particular water year. These catchments require special treatment (3 11.2). Small flows may otherwise be due to missing data. You should investigate years with missing data to see if the annual maximum may have occurred in the missing data period and the year excluded or included accordingly. Investigation methods include comparison of flows with another station(s) on the same or neighbouring river, or comparison with rainfall data. 7 Correlation plots between flood peaks at upstream and downstream gauging stations, or those on adjacent tributaries, are another useful tool for examining data. They can help identify patterns or inconsistencies in hydrological behaviour (see Figure 3).

8 The recommended methods for growth curve estimation, in 3 Table 8.3, assume that the flood record at the subject site is of average quality. You should informally reduce the record length in the table if the gauged record is considered unusually poor or increase it if the record is particularly good (3 8.2).

Doc No 197_08 Version 5 Last printed 12/02/15 Page 15 of 110

Figure 2: The graph below illustrates a flood peak time series on the River Stour at Example flood Langham, Essex/Suffolk. peak time series 100 90 Outlier: Sept 1968 flood No obvious trend or step changes 80 Low annual maxima - 70 years?

/s) 60 Check for missing data? 3 50 Few flood peaks above 40 m3/s 40 Flow (m 30 20 10 0 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002

Figure 3: The graph below shows a flood peak correlation plot, using flood peaks Example flood (from POT data) on adjacent tributaries of the River Stour in Essex/Suffolk. peak correlation plot

The catchments are similar in size, soils and geology. But the Stour Brook at Sturmer is affected by urbanisation and a major flood storage scheme. The correlation coefficient is 0.84, indicating close correlation. Flood peaks at Broad Green are generally higher than those at Sturmer, although the 1968 event (pre-scheme) is an exception. One possible explanation is that the scheme is reducing flood peaks to less than those expected from a rural catchment.

Feedback and We strongly encourage users to feed back any further information and any errors errors they find to HiFlows-UK. You can do this on the feedback page on CEH's website. It is often worth informing the gauging authority directly as well, for example through the Project Manager if the Environment Agency runs the gauge.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 16 of 110

Rating Reviews

Rationale At most, flow gauging stations, rating curves are used to transform water level into flow. Accurately calculating flood flows is problematic but of great importance.

Description Flood rating curves, particularly those that represent out-of-bank conditions, are often based on a small number of measurements or on extrapolation from the highest flow gauging. There are comments on ratings at most stations in the HiFlows-UK dataset. These are an important source of information. They should act as a prompt for users to enquire further, if appropriate. Analysts: you must take into account any more recent rating reviews or high flow gaugings, which may not yet have been incorporated into HiFlows-UK. If there has not been a review and there are questions over the rating, it is often worth carrying out a review.

Requirements Most flood estimation studies will require a review of rating equations at each gauging station used in the study (whether within the study reach or as a donor site), unless a recent review is available from another study. Our Hydrometry and Telemetry teams, with input from hydrologists in other teams, carry out full reviews and revisions of ratings, which are complex procedures. This section gives guidance on what you might expect in a typical rating review, carried out as part of a flood estimation study. References: For further guidance on rating reviews, see the Operational Instruction on flow derivation methods (OI 188_07). For guidance on extending ratings, see Ramsbottom, D.M. and Whitlow, C.D. (2003) listed in Related documents.

Guidelines The guidelines and advice in the table below are included to help users. Select references that are linked to see details in Related documents. Item Guideline or advice 1 The person carrying out the rating review needs: • a knowledge of: • hydrometry; Example: See Herschy, R.D. (1998). • and hydraulics. • an understanding of the value of flood data. 2 Rather than being purely a statistical exercise, the review should take into account the nature of the gauging station. Current information about existing stations is available from the measurement authority (within the Environment Agency, from the Hydrometry and Telemetry and/or Hydrology teams) and any review should always involve staff from these teams.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 17 of 110

3 A site visit often provides valuable insight into the way the station might perform during flood flows.

4 For detailed studies, it can be useful to obtain details of closed stations or information about the history of existing stations. You can find this in various sources, such as: • the teams mentioned above; • the station files held at CEH Wallingford; • reports on earlier flood studies; • and reports on previous hydrometric improvements. 5 The information to seek from all the sources listed in Item 4 above includes: • investigating the history of the station, such as its original purpose and any changes in the channel, structure or rating equations; • checking whether the rating is solely theoretical, checked by current meter gaugings or based solely on gaugings (empirical); • establishing whether the rating is theoretical, by finding out how it was derived; Example: By hydraulic theory or physical model tests. • establishing whether the rating is empirical, by finding out how it has been extrapolated for measuring flows above the calibrated range; Note: Straight line extrapolation on a log scale is the normal method used but there are better techniques. For example, extrapolating the velocity rather than flow and using measured channel cross-sections is a better method but this is only the simplest of the possibilities. See Ramsbottom, D.M. and Whitlow, C.D. (2003). • finding: • how flow measurements are taken; Example: By current meter (wading or cableway) or by an ADVP device that can be towed across the river. • and whether the measurements include flow through parallel channels or the floodplain; • comparing the valid range of the rating curve relative to the physical characteristics of the site, such as the bank levels and the levels recorded in flood conditions; • finding whether there have been any additional gaugings (or measurements, such as float runs or using portable ultrasonic flow meters) which current databases may not list; • assessing the potential for bypassing during flood flows; • checking for non-modular flow due to backwater effects; • checking for susceptibility to hysteresis (looped ratings due to storing flood water); • finding how the station is classified, according to the Gauging Station Data Quality system. Note: This assesses whether measurements for flows around half of QMED are reliable, based on site and station factors, and checks gaugings. See JBA Consulting (2003). 6 You can summarise some of the information, listed in Item 5

Doc No 197_08 Version 5 Last printed 12/02/15 Page 18 of 110

above, on a plot showing the rating curve against flow gaugings. A plot like Figure 4 shows: • the scatter in the gaugings (a measure of uncertainty); • and how much the rating has been extrapolated for measuring the highest flow on record and for QMED. Adding the bank level can help to explain any changes to the slope of the rating curve, which often occur at bankfull flow. It can also be worthwhile plotting the channel cross section on a second x-axis.

7 You can statistically assess the accuracy of the rating if needed, but do this with caution. Example: Goodness-of-fit statistics such as R2 tend to be dominated by the large number of low flow gaugings and may not reflect the quality of the rating for high flows.

8 It is also worth plotting a time series of the deviations between predicted and measured flows and showing the cumulative deviation. This can reveal any drift in the gaugings, which might suggest that the rating needs to be recalculated. Further investigations, if required (for example, if the gaugings are very scattered) could include separating the gaugings by: • season, to investigate vegetation growth; • or rising/falling stage, to investigate any hysteresis.

Figure 4 The graph below shows a rating curve plotted against flow gaugings on the River Kent at Bowston, Cumbria

2.0

1.5

1.0 Stage (m)

Existing rating 0.5 Gaugings Highest recorded level QMED Bank level 0.0 0.0 20.0 40.0 60.0 80.0 100.0 120.0 Flow(m 3/s)

Result of the The review should result in a conclusion about the suitability of the rating for review high flow measurement and possibly recommendations for further work. In some cases, it is appropriate to develop a new rating, if there have been additional recent high flow gaugings. Always carry this out in consultation with the Hydrometry and Telemetry team and ensure any revisions to the

Doc No 197_08 Version 5 Last printed 12/02/15 Page 19 of 110

rating are fed back into the Environment Agency’s WISKI archive. In reaching the conclusion, it is important to realise that high flow measurement is uncertain at nearly all gauging stations. Before rejecting a station, consider what the alternatives are, bearing in mind their uncertainty. This is particularly the case if the alternative is to base a flood estimate solely on catchment descriptors, which the FEH describes as a last resort.

When to revisit You will sometimes need to revisit the rating review later if the study goes the review on to develop a hydraulic model of the reach that includes the gauging station. This may reveal the influence of downstream water levels on the high flow rating. Examples: Constrictions at structures or inflows from downstream tributaries. It may also show the effects of hysteresis, which is often due to storage of water on the floodplain.

Flood Event Data

Rationale Similar to flood peak data, visually examining flood event data can reveal much about the hydrological behaviour of a watercourse. It is also vital for checking the quality of data. Example: Spotting spurious peaks or periods of missing data. It can be useful to plot rainfall and flow together, as this may identify problems which may cause an event to be rejected (4 A.4).

Description Model parameters for the ReFH method (and the FEH rainfall-runoff method) are best estimated from flood event data, which is normally recorded at a time step of 15 minutes. The ReFH method requires flow and rainfall data. It does not include the provision to use river level data for deriving time to peak, as in the FEH rainfall-runoff method. However, given the wider availability of river level recorders, there are likely to be some situations where analysts judge that level data are helpful in guiding the selection of parameters for the ReFH method.

Guidelines The guidelines and advice in the table below are included to help users. Item Guideline or advice 1 Flood event analysis needs to be based on catchment-average rainfall data. On smaller catchments with a nearby recording raingauge, it is often acceptable to treat the data from that gauge as the catchment average. On larger catchments, you should average data over several

Doc No 197_08 Version 5 Last printed 12/02/15 Page 20 of 110

recording gauges. If these are not available, it is possible to use daily raingauges to improve the averaging.

2 Radar-derived rainfall data can provide a valuable additional source of information, when used with measurements from at least one raingauge (4 A.4.1).

3 The ReFH method can also use potential evaporation data. These are required for setting the initial soil moisture when estimating model parameters from observed data or simulating observed events. One option is to use an annual sinusoidal series, which only needs the annual mean daily potential evaporation. Another option is to enter a potential evaporation time series, which can be obtained from the ’s MORECS or MOSES systems. For more guidance on how to obtain this data, see 414_07 Accessing Hydrological Data and Information, on Easinet.

Flood History

Rationale You can often make flood estimates at longer return periods much more reliable by carrying out a historical review and incorporating floods before the period of gauged records. Reference: Bayliss and Reed (2001). Most studies need an estimate of the 100-year flood, which is not that likely to have occurred during most gauged records (see The risk equation below). Historical reviews, similar to pooled analysis, can supply a wider perspective (1 C). Uncovering forgotten information can also add credibility to the analysis and contribute to public understanding of flood risk (1 C.2).

The risk The probability p that a T-year return period flood (or larger) will occur at equation least once in an N year period is given by the risk equation: N p = 1 - (1 - 1/T)

A typical record length for flood peak data is 40 years. The risk equation gives the probability of a 100-year flood occurring during this period as 33%. In other words, there is a one in three chance that the 40-year record will include the 100-year flood.

Description Historical reviews are often required in flood estimation studies. In many studies, they are too often left out or only given lip service. Perhaps they are seen to need more effort and thought than a pooled analysis that can be carried out using the FEH software. However, historical reviews can be rewarding as well as valuable and they can have a large influence on the design flows. For example, a study (Black and Fadipe, 2009) found that 100-year flood flows at three out of four sites increased by more than 50% as a result of incorporating reliable historical information. Doc No 197_08 Version 5 Last printed 12/02/15 Page 21 of 110

There is a great deal of historical flood information available. Archer (1999) suggests that you may obtain useful information for a period of at least 150 years in virtually every flood-prone catchment in England. MacDonald (2009) describes how relatively good records of flooding are available for large catchments since 1500, and very good records since 1750. MacDonald and Black (2010) present a reassessment of flood risk at York using documentary records dating back to 1263AD. The study showed that the FEH estimates of 100-year flow (whether from single-site or pooled analysis) were implausibly high as the estimated flow rates had not been reached in the entire 737-year historic series. The preferred estimate of 100-year flow was nearly 20% smaller than the FEH pooled estimate. Going even further back, historical reviews can extend into palaeoflood investigations which use evidence such as sediment deposits, tree rings and pollen to develop very long-term records of major floods. You can find an example in Brown (2009) who developed a 1500 year record of flood flows on the River Trent using geomorphological and geoarchaeological data. Palaeoflood techniques have particular potential in making assessments of the very largest floods that a landscape has experienced. You should consider using palaeoflood methods for high-risk studies such as those involving the safety of dams or facilities handling catastrophically dangerous materials (Bayliss and Reed, 2001).

Guidelines Item Guideline or advice 1 Project Managers and analysts: you must agree at the start of a study whether to include a historical review. For all except simple or routine studies (see Table 2) you should normally include a historical review or an update of a previous review, if it will supplement an existing gauged flow record. While the scale of the study should dictate the effort employed, experience suggests that a thorough review of historical sources may take no more than two to eight days.

2 For information, you can refer to: • the FEH (1 C) and Bayliss and Reed, 2001 for advice on carrying out a historical review; • the BHS Chronology of British Hydrological Events, although rather qualitative, is a very useful resource for information up to 1935; • other websites can add information, usually on more recent events; • numerous other sources, including local newspapers, local history books, the British Rainfall publication series and flood marks on buildings. It can be important to go back to the original sources of historical data and to critically assess their quality (see Bayliss and Reed, 2001) They can provide source information for studies on neighbouring catchments (for example, dates of flooding) and should be added to the BHS Chronology website, to enable wider access.

3 In some cases, historic information can be used to guide the choice between a single-site and a pooled growth curve, without any need for quantitative data. One way to approach this is to rank historic events, or classify them as major, moderate or minor Doc No 197_08 Version 5 Last printed 12/02/15 Page 22 of 110

floods. You can then compare the results with the size of the highest floods within the gauged record, to see whether the single-site growth curve is consistent with the longer-term history. The FEH recommends an informal method (1 C.3.3) for incorporating historical flood data into estimation of the flood frequency curve. Archer (1999) outlines an example of using practical informal methods. More detail appears in Bayliss and Reed (2001), which reviews various methods for incorporating historical data in a flood frequency analysis and advocates using simple methods. Take care if there have been substantial changes to the catchment that would affect its flood behaviour. You may also need to consider the effects of climatic fluctuations, although MacDonald and Black (2010) point out that, once long periods are considered (over 250 years), climatic variability becomes inescapable, and that inclusion of flood rich and flood poor periods leads to more robust flood frequency estimates.

4 Bayliss and Reed (2001) recommend particular care in cases when the historical flood data suggest that the preferred frequency curve is too high, because of the scope to overlook floods. The FEH suggests giving greater respect to historical flood data when they suggest that the preferred frequency curve may be too low. You should adjust the fitted distribution to acknowledge the historical data (1 C.3.3).

Catchment Descriptors

Information The FEH CD-ROM version 1 offered 20 catchment descriptors for sites available within the resolution of the underlying digital terrain model, IHDTM. Version 2 of the CD-ROM provided three additional catchment descriptors based on an improved and more recent land cover map, in particular URBEXT2000. URBEXT2000 is defined differently from URBEXT1990 and typically has a higher value for the same degree of urbanisation Reference: See Bayliss, A.C., Black, K.B., Fava-Verde, A., Kjeldsen, T.R. (2007) listed in Related documents. It is based on three land cover types: urban, suburban and inland bare ground (which in urban areas corresponds to gravel car parks, railway sidings, derelict areas and so on). Therefore, do not use URBEXT2000 in the original FEH equations for urban adjustments. Only use it in equations developed specifically for URBEXT2000. Version 3 of the CD-ROM adds another three catchment descriptors. Reference: Kjeldsen, T.R., Jones, D. A. and Bayliss, A.C. (2008) listed in Related documents): • FPEXT: floodplain extent, the fraction of the catchment inundated by a 100-year flood, used when selecting pooling groups; • FPLOC: floodplain location relative to the catchment outlet; • FPDBAR: mean depth of water on floodplains in a 100-year event.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 23 of 110

Guidelines Item Guideline or advice 1 Ten descriptors are used in flood estimation procedures. The others provide extra information for the analyst to use when comparing catchments for data transfer and selecting pooling groups.

2 Do not use catchment descriptors, obtained from the FEH CD- ROM, without, at least, a rudimentary check. In particular, confirm catchment boundaries, based on the IHDTM, and therefore area (AREA), urban extent (URBEXT) and the effect of reservoirs and lakes (FARL). Use information such as OS maps, digital elevation models (DEMs) and local knowledge. Analysts: you may find that a site of interest will not be found within the resolution of the FEH CD-ROM data. Version 2 of the CD-ROM corrected some of the more major errors, but you will find places where the catchment boundaries are still in error. Checking is particularly important for small catchments, see Figure 5.

3 It's particularly worthwhile to verify catchment boundaries: • in fenland areas; • or when there are artificial influences; Examples: Reservoir catchwaters, diversion channels or embankments. • or groundwater interactions. You should also investigate any other local anomalies that might affect hydrological response.

4 The best way to check a catchment boundary is usually with GIS. But, due to data licensing restrictions, it is not possible to import boundaries from the FEH CD-ROM to a GIS package. CEH Wallingford can provide digital boundaries on request for a fee. Or you can visually compare boundaries from the CD-ROM with those derived in a GIS from information such as the Nextmap DEM or contours on an OS map. You can print a map at a user- defined scale and you can import shape files and add them to the map display.

5 As well as catchment boundaries, you should normally check soil characteristics from the HOST classification. This is particularly important on small catchments, where the use of SPRHOST may be inappropriate due to the 1 km resolution of the summary HOST data (5 5.4). You can check soil characteristics against soil and geology maps. Note: The Soil Survey of England and Wales (now the National Soil Resources Institute) published a 1:250,000 Soil Map of England and Wales in 1983 and have larger-scale maps of some areas, see the Landis website. For an online summary of the 1:250,000 map see this Soilscapes page. For important studies on smaller catchments, a site survey of soil properties may be worthwhile. Appendix C of FEH Volume 4 lists the HOST classes allocated to each soil association shown on the soil maps. You can derive Doc No 197_08 Version 5 Last printed 12/02/15 Page 24 of 110

SPRHOST and BFIHOST from the HOST classes, using 5 Table 5.1. You should always view low values of SPRHOST in what appear to be relatively impermeable areas with suspicion. Example: Some Pennine catchments where soil associations on the Soil Survey map indicate slow-draining soils having SPRHOST below 20%, whereas 30-50% would be expected.

6 It is worth carrying out a quick check of the FARL value. For most catchments, this will be close to 1.0, indicating no significant attenuation from lakes or reservoirs. Many flood storage reservoirs are not included in the dataset on which FARL is based and there are some errors in the CD-ROM where outflows from water bodies are in the wrong location. You can correct these omissions or errors by manually calculating FARL (5 4.3); see Item 7.

7 When you find any FEH CD-ROM catchment boundaries are incorrect, you will need to manually adjust the descriptor values (5 7.2.1). You can adjust many of the catchment descriptors using a simple area weighting method (5 7.2.2). However, this is not applicable to all descriptors. You cannot adjust FARL by area weighting. You can estimate DPLBAR approximately by regression on the catchment area. You must apply adjustment procedures with care. Analysts: you should take account of the derivation and purpose of the descriptor and record the adjustment fully.

Figure 5 Example: The maps below show a catchment boundary error around Wacton Stream, Norfolk. FEH CD-ROM: catchment area is 0.55 km2. Catchment boundary from Nextmap DEM: area is 2.01 km2

© Crown Copyright. All rights reserved. Environment Agency, 100026380, (2009).

Doc No 197_08 Version 5 Last printed 12/02/15 Page 25 of 110

Notifying Notify any errors in catchment boundaries or descriptors to CEH errors Wallingford, by e-mail, to [email protected].

Choice of methods Overview

Basic methods The basic methods available are: available • the FEH statistical method; • the ReFH method; • the FEH rainfall-runoff method, sometimes known as the FSR rainfall- runoff method because the FEH made few changes. Superseded by ReFH in most cases but versions of the method are still applicable for reservoir safety work and on pumped catchments. • various older methods used on very small catchments or for greenfield runoff estimation.

Six maxims The FEH offers six maxims (1 2.2), summarised below. These should guide the choice of method. • Flood frequency is best estimated from gauged data. • While flood data at the subject site are of greatest value, data transfers from a nearby site, or a similar catchment, are also useful. • Estimation of key variables from catchment descriptors alone should be a method of last resort. Some kind of data transfer is usually feasible and preferable. • The most appropriate choice of method is a matter of experience and may be influenced by the requirements of the study and the nature of the catchment. Most importantly, it will be influenced by the available data. • In some cases, a hybrid method, combining estimates by statistical and rainfall-runoff approaches, is appropriate. • There is always more information. An estimate based on readily available data may be shown to be suspect by a more enquiring analyst.

Analysts: The six maxims stress the need for you to think, at all stages, about the approach to problem you are solving and not to simply feed data into software packages. choosing a These guidelines further promote this philosophy. You have to make method decisions and you may have to improvise. You have to rely on judgement based on experience, the nature of the problem and, not least, the available data and time. When necessary, you should seek assistance from a senior colleague. Prescriptive rules on choice of method are neither feasible nor desirable. The FEH says that choice of method is 'both complex and subjective'. It acknowledges that 'different users will obtain different results, by bringing different data and experience to bear' (1 5.1).

In this chapter This chapter gives guidance on the basic choice between approaches. For many studies, this means deciding between a statistical and a rainfall-runoff

Doc No 197_08 Version 5 Last printed 12/02/15 Page 26 of 110

approach. It includes a suggested framework for decision-making and emphasises the importance of starting with a method statement. For information on the limitations of various methods, see Chapter 5. For guidelines on choosing a method for unusual catchments, see Chapter 6.

A framework for choosing a method

Summary The diagram in Figure 6 illustrates a framework for decision-making. Choosing the method occurs at several stages: • the analyst makes an initial choice, which often involves a number of possible approaches, during preparation of the method statement; • they then derive initial flood estimates, using the selected methods, often just at example locations, such as gauging stations or important confluences or flood risk areas; • by comparing results, they select the preferred method (or methods) and apply this at all locations; • finally, they check the results and, if necessary, they revisit the calculations. If analysts follow this framework, there is no need to carry out calculations at numerous sites several times over. This takes a lot of time and tends to result in multiple tables of results, with the potential for misinterpretation.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 27 of 110

Figure 6 The diagram below illustrates a framework for decision-making that is intended to guide analysts through the thought processes that are required. It shows the main stages they should follow in flood estimation for a typical study, involving multiple flow estimation points (such as a flood mapping study or CFMP). They can apply a simpler version to smaller-scale studies. The right-hand column of the diagram, in light green, shows the outputs that they should produce. Select links in the diagram to move to sections in this document providing more details.

Assemble information: • the brief; • maps; • hydrometric data; • flood history;

Think: • type of problem; Write a method statement. • type of catchment; Agree with the client, if • type of data

Analysis at selected sites.

Select preferred method. Record the choice of method.

Agree with the client, if

Analysis at all sites. Record the calculations.

Check results for sensibility Record the results. and consistency.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 28 of 110

The need to think

Three factors Choice of method is important and rarely straightforward. The many factors to think about to consider can be grouped into three categories. Select the links to read more details in Chapter 6 on specific issues. • type of problem; Examples: Is a hydrograph needed? How will the flows be applied to any hydraulic model? Is the flood estimate for a reservoir spillway assessment? What return period is required? • type of catchment; Examples: Is it large? Permeable? Urban? Pumped? Are there disparate subcatchments? (4 9.2) Is there a reservoir? (4 8) Are there extensive floodplains? (1 3.1.2) type of data. Examples: Is there a flood peak record? How good are the high flow measurements? Are flood event data available? What about flood history?

Show how It is often helpful to include a section in a hydrological report dealing with factors have each of the above three factors. It aids the thinking process and it influenced demonstrates that you have considered all the factors that might influence choice the choice of method.

Preparing method statements

Time needed Preparing a method statement helps analysts to plan their studies carefully. While half a day may be adequate for a preliminary assessment, thorough flood estimation studies can take many days, even weeks. The FEH suggests allowing five to 50 days (1 Interlude, p 37). Major flood studies need planning in advance, with time to review and update data. There are many factors to consider when choosing the approach to adopt. Analysts: you should agree the level of detail required with the Project Manager at the start of a study. It will depend on the application and its importance, and on available data.

Description The method statement represents an opportunity to develop a conceptual understanding of the catchment. It may help to visualise what conditions are likely to lead to flooding of the areas of interest (sometimes referred to as the 'design condition'). See Examples of conditions.

Examples of • Consider these examples: conditions • is flooding likely to be dominated by the magnitude of peak flows or are flood volumes or tide levels also likely to have an effect? • will it be a joint probability problem, for example due to the presence of tributaries with different hydrological characteristics, or a combination of high flows and high groundwater levels? Doc No 197_08 Version 5 Last printed 12/02/15 Page 29 of 110

• is there a possibility that the most severe floods could arise from runoff generated on only part of the catchment? Examples: An area downstream of a reservoir or an impermeable portion of a geologically mixed catchment. • is the catchment likely to be vulnerable to snowmelt floods? is there an additional risk posed by landslides, bridge collapses or flood debris creating temporary dams that could collapse? Example: See the report on the 2004 flood, listed in Related documents.

Guidelines Item Guideline or advice 1 If river flow or level data are available, it is worth carrying out some initial analysis. Example: Plotting time series and looking at hydrograph shapes. If there are several gauging stations, then it can be worthwhile looking at travel times and correlations between peak flows, and the relative seasonality of flood peaks at different stations (as floods that occur in different seasons tend to arise from different processes). On permeable catchments, you can investigate the importance of baseflow. Example: By plotting daily mean flow data.

2 Review: • the quality of data; See Selecting and examining flood peak data, • and the availability and quality of historical data. See Flood history. 3 For lengthy or high-risk studies (for example, those in the bottom two rows of Table 2), it is advisable to agree the method statement with the Project Manager before going any further. Example: You could sketch a conceptual model of the system and present it to Area staff who are familiar with the catchment.

Choosing between the FEH methods

Background For the first six years after the FEH was released, the most difficult choice was often between the FEH statistical and rainfall-runoff methods because they can give such different results. The ReFH method has now superseded the FEH rainfall-runoff method for most applications. It tends to give results that are more consistent with the statistical method. However, the choice of method can still have a major influence on the results.

Factors The statistical method is likely to be preferred in many cases, particularly favouring the when any of these apply: statistical • there are more than two or three years of flood peak data on the method watercourse (even if not at the sites of interest), from a gauging station

Doc No 197_08 Version 5 Last printed 12/02/15 Page 30 of 110

suitable for high flow measurement; • the catchment is larger than 1000 km2; Rainfall-runoff approaches assume a catchment-wide design storm, which is less realistic for large catchments. the catchment is highly permeable (approximately BFIHOST>0.65). Neither ReFH nor the FEH rainfall-runoff method work well on permeable catchments. See Permeable catchments for more details.

Factors Examples of factors that might favour a rainfall-runoff approach (in most favouring the cases, the ReFH method but you may consider FEH rainfall-runoff in some ReFH method situations) include: • there is no continuous flow record, but rainfall and flow data are available for five or more flood events; • the problem involves flood storage and/or routing (for example, reservoirs or an unusually extensive floodplain) and there is no flood peak data that implicitly account for the effects of the storage; • the return period is long, for example 1000 years; Estimating long return periods, discusses the applicability of ReFH for long return periods. ReFH will not always be the best choice in this case and it is important to compare with the results of the statistical method. • the study involves designing works to counter the effects of a new urban development and/or storm sewer design; • the catchment includes subcatchments with widely differing flood responses; • the catchment is low-lying, with pumped drainage. See Pumped catchments.

Guidelines Item Guideline or advice 1 Because the statistical method is based on a much larger dataset of flood events, and has been more directly calibrated to reproduce flood frequency on UK catchments, you should often prefer it to any rainfall-runoff approach (1 5.6). However, the choice is not always clear cut. Sometimes you will choose both approaches for different reasons, such as those listed in Factors favouring the statistical method and Factors favouring the ReFH method. It will often be worth deriving results at example sites using several methods. In doing so, additional information may emerge which can help the final decision. The FEH suggests that sometimes an intermediate estimate can be adopted (1 5.6).

2 Like the FEH rainfall-runoff method, the ReFH method’s design procedure was calibrated with a dataset much smaller than that available for the statistical method: 100 catchments compared with around 960 in the HiFlows-UK dataset that the statistical method can draw on for flood frequency estimation. Note: Although several catchments were added to the flood event archive during the ReFH research, many were found to have insufficient event data for large floods. Figure 7, illustrates this. It shows, in particular, the lack of small, large or urban catchments available for calibrating ReFH. A similar

Doc No 197_08 Version 5 Last printed 12/02/15 Page 31 of 110

range of catchments was used in calibrating design events in the FEH rainfall-runoff method.

3 It's important to understand that the quality of flood frequency estimates, from design event methods such as ReFH or FEH rainfall-runoff, is influenced by the appropriateness of the 'design package', (that is, the combination of storm depth, duration, profile and soil moisture) to the catchment. It is not just influenced by the quality of the rainfall-runoff model parameter estimates (1 12.2). The ReFH method gives more information on when you should, and should not, use ReFH.

4 The FEH discourages users from choosing a method because: • it gives the highest or lowest flow (3 Box 7.1); • or it gives results that match those from a previous study (1 5.8). 5 Analysts: there will be times when the FEH methods are inappropriate and you may need to consider an alternative method (see Small catchments and greenfield runoff )

Figure 7 The graph below shows a range of catchment types that the FEH statistical, ReFH and IH Report 124 methods draw on.

0.5

0.45

0.4

0.35

0.3

0.25

URBEXT 0.2

0.15

0.1

0.05

0 -3.0 -1.0 1.0 3.0 5.0 7.0 9.0 ln (AREA)

Hiflows-UK (suitable for QMED) Used in ReFH Used in IH Report 124

Hybrid methods

Description When you need a design hydrograph, the preferred approach will sometimes be a hybrid method. A hybrid method combines a hydrograph shape with an estimate of peak flow by the statistical method (1 5.6, 3 10 and 4 7.3). Hybrid methods are used commonly in hydrodynamic modelling studies.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 32 of 110

Possible The FEH suggests three hybrid methods, listed below. Others, such as (d) methods below, are used occasionally. Possible options Description and guidelines (a) Generating This is the quickest method and often the best. You the hydrograph can apply it to gauged or ungauged catchments. from the ReFH The disadvantage is that it is rather a 'brutal' method, then application of the ReFH method, losing the information scaling it to match on runoff volume. the statistical estimate. It is not well suited to large catchments or those dominated by storage, where hydrograph shapes are less likely to resemble the simple ReFH hydrograph. However, it can sometimes be applied in these catchments by splitting them up into subcatchments and routing the resulting hydrographs.

(b) Adjusting the This might appear more elegant than option (a) but parameters of the you should use it with caution. It is only valid if the ReFH model until parameters have not already been estimated from the simulated local flood event data. It assumes that the reason for peak flows match the ReFH method giving a poor answer is that the the preferred model parameters have been poorly estimated, which values (3 10.2). is not always the case. It may prove difficult to match the statistical results over a range of return periods, because the ReFH method may give a different growth curve.

(c) Using a This constructs a symmetrical hydrograph, using a simplified model parameter defining the width of the hydrograph at half of the hydrograph the peak flow. You can estimate this from recorded shape (3 10.4). events or from Tp(0). This approach is rarely used.

(d) Basing the You can derive a shape by averaging the hydrographs hydrograph shape of major events, standardised by their peaks. You can on gauged flow do this by: data. • simple averaging of the hydrograph ordinates (see Figure 8 below); • or using a more sophisticated procedure, such as deriving the duration of Exceedence of selected percentiles of peak flow. Reference: Archer, D., Foster, M., Faulkner, D. and Mawsdley, J. (2000) listed in Related documents. The above paper recommends using observed events on catchments with significant storage (in aquifers, lakes or floodplains), unless the storage is to be modelled explicitly as part of the study. A simpler alternative is to use the shape of the largest flood on record, particularly attractive if the peak is thought to have a return period similar to that of the required design event. This approach is only possible at a gauging station or shortly up or down river.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 33 of 110

Figure 8 The graph below shows the average flood hydrograph shape on the River Ore at Beversham.

1.0 27-Apr-81 0.9 30-Dec-81 02-Apr-83 0.8 19-Apr-83 27-Jan-84 0.7 07-Feb-84 22-Jan-85 0.6 12-Apr-85 25-Jun-85 0.5 27-Dec-85 28-Feb-87 0.4 26-Aug-87 16-Oct-87 Proportion of peakflow of Proportion 0.3 10-Jan-92 07-Jan-93 0.2 13-Oct-93 31-Oct-94 0.1 20-Dec-97 30-Oct-00 0.0 20-Sep-01 -60 -40 -20 0 20 40 60 23-Dec-02 Time after peak (hours) Average

Checking results

Questions to It is vital to check that flood estimates are sensible. This can sometimes ask help in choosing between results from alternative methods. Some questions to ask are listed in the table below. Select the links in the table to read more detail in Chapter 6. If there are multiple flow estimation points, some of the questions are best answered graphically. Examples: Plotting long sections of specific discharge against location or Doc No 197_08 Version 5 Last printed 12/02/15 Page 34 of 110

maps of growth factors.

Item Question 1 Are the results spatially consistent between upstream and downstream points and at confluences?

2 Are the growth factors sensible? In the FSR regional growth curves, the ratio of the 100-year to the 2- year flow varies from 2.1 to 4.0. You should investigate 100-year growth factors that fall significantly outside this range. You can sometimes justify much higher growth factors on highly permeable, clay or urban catchments (or catchments containing mixtures of these characteristics), where they are consistent with the flood history.

3 What specific discharge (that is, flow in litres/second/hectare) do the results equate to? Can you explain the variations in specific discharges between different locations across the catchment?

4 What return period do the results imply for major events during the gauged record? This can help in the choice between single site and pooled curves. 5 Are the results consistent with the longer-term flood history?

6 Are flows generated by a hydrodynamic or routing model, consistent with those estimated from a lumped catchment FEH estimate, at locations within the model reach? If not, the inconsistency needs to be explained and you will need to make a decision about the preferred method for flood estimation.

Using the You can use the Checklist for reviewing flood estimates (SD03) which checklist includes the questions above and others. This checklist can be used by: • analysts checking their own work; • supervisors carrying out internal reviews; • and project managers reviewing calculations.

Conclusion

Use the six Use the six maxims to guide all aspects of the choice of method. maxims as a As the sixth one says, 'there is always more information'. Some pragmatism guide is needed in deciding when a flood estimate is good enough for the needs of the study.

No The reconciliation of estimates by different methods is a skilled task. It is not prescriptive possible to give a prescriptive set of rules. set of rules Part of the skill is in knowing when - having explored the possibility – to

Doc No 197_08 Version 5 Last printed 12/02/15 Page 35 of 110

accept or reject a particular adjustment.

Adopting Sometimes the best flood estimates are derived from approaches, which do unusual something out of the ordinary. approaches Examples: Incorporating historical data or accounting for unusual flood- generating processes. If you are adopting an approach that deviates from normal practice, it is all the more essential to justify the decisions made and check that the answers are sensible by following the advice given in this chapter. Too often, an unusual approach results in flood estimates that are difficult to defend and no better (or even worse) than could be obtained using more conventional methods.

Incorporating Sometimes the only indication that the design flows need altering comes information once they have been applied to a hydraulic model. Water levels or flood from hydraulic extents are easier to visualise than flow rates. If the model results are not models consistent with local knowledge or flood history, then this can act as a prompt to revise the design flows, as long as there is enough confidence in the hydraulic model structure and parameter values. This last point is important because sometimes it is the model or the modeller’s assumptions that need to be altered. Flow rates inferred using an uncalibrated hydraulic model should not be treated with the same level of confidence as those derived from a rating curve at a gauging station.

Advice and cautions on FEH and ReFH methods

Overview

Reminders, There are many opportunities for choice when applying the FEH methods, guidance and including somewhere the unwary might miss a subtle variation in the options latest research facing them. The sections in this chapter aim to help less experienced analysts use the FEH and act as a reminder to more frequent users. They concentrate mainly on areas that FEH users tend to find difficult, or areas that tend to have the largest effects on the results. The sections also highlight findings from more recent research, giving advice on when and how it to put into practice.

Analysts: You should establish what previous flood studies have been carried out for general advice the subject site or within its catchment. These are often worth examining. They may provide information on data sources and accuracy, catchment conditions and flood history. You should note results for comparison and investigate unexpected discrepancies, except where you consider, and record, that this is unnecessary.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 36 of 110

Design rainfall

Description of The depth-duration-frequency (DDF) model provided on the FEH CD-ROM the DDF model enables the estimation of design rainfalls for any location in the UK or the return period of an observed rainfall. The DDF model is fitted to rainfalls with durations from one hour to eight days (2 12.1).

Guidelines Item Guideline or advice 1 You can rely on the results for durations as short as 30 minutes. For shorter durations, you must revert to the Flood Studies Report (FSR) rainfall statistics, listed in Related documents. 2 Rather than using the FSR statistics directly, we suggest that you use them to factor down to shorter durations from the 1-hour FEH rainfall depth. Example: You require a 15-minute depth. Calculate the ratio of the 15-minute and 1-hour depths from FSR statistics. Then multiply it by the 1-hour depth from FEH.

3 Design rainfalls produced by the DDF model are for sliding durations, which are durations that start at any time (2 2.5). There is an option to adjust rainfall depths to convert between fixed (duration starts at discrete times only) and sliding durations. You will normally only need this if you are estimating the return period of a storm that has been measured only at daily raingauges.

4 Flood estimates from rainfall-runoff approaches need a catchment-average rainfall. You can estimate catchment rainfall automatically using the DDF model provided on the rainfall model part of the FEH CD-ROM. The areal reduction factor formula is in 2 3.4.

5 Definition: The index variable, RMED, is the median of annual maximum rainfalls (for a given duration) at a site. Digital maps of RMED on a 1 km grid were developed for combination with rainfall growth curves. Catchment average values of RMED, on the catchment descriptor part of the FEH CD-ROM, are not for use in rainfall frequency estimation (2 7.1). Analysts: you should always use the rainfall DDF model rather than the RMED values given with catchment descriptors.

6 The FEH recommends that, generally you should not use local data for refining design rainfall estimates, even where rainfall records are long (2 12.2). However, given that 17 years has elapsed since the FEH rainfall data was collected, you may find sites where you can calculate a more reliable estimate of RMED than the FEH value, particularly for durations shorter than 1 day.

Revised rainfall frequency statistics have been developed as part of a research project at CEH Wallingford; see Stewart, Lisa and others (2010). They are expected to be available in a future release of the FEH CD-ROM. Doc No 197_08 Version 5 Last printed 12/02/15 Page 37 of 110

The main focus of the project was improving estimates for extreme return periods (see Flood estimation for reservoir safety) but design rainfalls for return periods of 100 years or less will also change.

Statistical Method - general

Data and Flood estimates from the statistical method depend on the quality and extent Software of available gauged data: • at subject sites or donor sites to estimate QMED (1 5.3); • and at pooled gauging stations to construct the pooled growth curve. The statistical method is usually applied using WINFAP-FEH. The current version of the software is v3, released in September 2009. WINFAP-FEH is not a straightforward package to use. Some analysts find it convenient to record their calculations in a spreadsheet, which they can also use to calculate QMED and design flows given the growth curve parameters produced by WINFAP-FEH.

Figure 9 The diagram below illustrates the main options available to analysts. There are no short cuts to choosing the most suitable method and analysts need to consult the FEH (1 5 and all of Volume 3).

Flood peak data at the subject site No flood peak data at the subject site

Record length Record length Record length >13 years 2 to 13 years <2 years

QMED as QMED from QMED by QMED from catchment descriptors, median of POT data data transfer adjusted by transfers whenever possible annual from donor maxima catchment

Best estimate of QMED

Record length 2T years 2T years

Pooled analysis Single-site analysis Investigate both

Complete the flood frequency analysis for return period of primary interest = T years

Statistical method – index flood, QMED

Doc No 197_08 Version 5 Last printed 12/02/15 Page 38 of 110

General information

Guidelines

Item Guideline or advice 1 When you estimate QMED from flood peak data, the gauged record at the subject or donor sites should be of sufficient length and quality (1 5 and 3 2.2, 12).

2 You should consider the possibility of trends in flood peaks (3 21.3.3, 21.5.4). If a trend is identified, there may be a case to sacrifice length of record for realism (3 2.2, 21).

3 Climatic variability can result in flood-rich or flood-poor periods. In QMED estimation, it is important to watch out in case such a period distorts the estimate from gauged data. The FEH recommends that QMED is adjusted for climate variation if the station’s record is shorter than 14 years (3 2.2, 20).

4 The presence of tied values (identical annual maxima) in a flood series can compromise the estimate of QMED (3 2.3). You can identify these by examining the ranked flood peak data.

5 The FEH includes two additional approaches to estimate QMED that are rarely used: • continuous simulation modelling • channel dimensions (3 5.2). The second method can form the basis for a second opinion on QMED. QMED is estimated from the bankfull channel width. The method is only applicable on natural reaches with a bankfull width not much less than 5 m. Someone with geomorphological knowledge will need to visit the site to ensure that the conditions in FEH 3 5.2 are met.

From catchment descriptors

The revised The original QMED equation provided in the FEH was superseded in 2008 QMED by the revised equation in Science Report SC050050, listed in Related equation documents. It was developed using the longer and higher quality flood peak records available from the Hiflows-UK dataset and more advanced regression techniques. This revised QMED equation is:

1000 2 QMED = 8.3062 AREA0.8510 0.1536 SAAR FARL3.4451 0.0460BFIHOST

This equation performs significantly better than the original FEH equation. It gives lower QMED estimates at most sites in East Anglia and the English Midlands (where SAAR is low), and higher estimates in most other locations, apart from where SAAR is very high. This revised equation was Doc No 197_08 Version 5 Last printed 12/02/15 Page 39 of 110

developed from data on 602 rural catchments, with catchment descriptors covering the following ranges: • AREA: 1.6 - 4590 km2; • SAAR: 560 - 2850 mm; • FARL: 0.645 - 1.000; • BFIHOST: 0.20 - 0.97.

Using the You should use the revised equation for all new studies. It is included in v3 revised QMED of WINFAP-FEH. equation

Guidelines Item Guideline or advice 1 You should only consider estimating QMED from catchment descriptors as a last resort.

2 You should not: • extrapolate the formula beyond its calibration range; • or rely on it when FARL<0.9 due to reservoirs (3 3.3, 13). 3 The model for QMED cannot account for all catchment features. Example: You should not use it on karst catchments and generally avoid it on artificially drained fenland catchments.

4 The FEH includes suggestions on using locally derived values for some of the variables in the catchment descriptor equation for QMED, for example SPR or BFI (3 13.7.3). The suggestions on using SPR are no longer relevant because the revised QMED equation excludes SPRHOST. It would be possible to substitute a gauged estimate of BFI for BFIHOST in the above equation, for example at a gauging station, which provides reasonable data for average flows but poor flood peak data. However, there is no specific recommendation available on the value of this adjustment.

Urban adjustment

The issues Urbanisation modifies the natural flood response. In the absence of flood peak data for the site of interest, both QMED and the growth curve need to be adjusted for urbanisation (3 9). The guidance below explains some important issues with the adjustment of QMED within WINFAP-FEH. Although the FEH only mentions performing the urban adjustment for urban catchments, it makes sense to apply it on all catchments to avoid a discontinuity when URBEXT2000 exceeds the threshold value of 0.030. For more general advice on urban catchments, refer to Development control and urban catchments.

Using To adjust QMED for urbanisation, multiply the rural estimate of QMED (from WINFAP-FEH catchment descriptors) by an urban adjustment factor, UAF.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 40 of 110

to adjust Due to a bug in WINFAP-FEH, if QMED is estimated using either of the QMED for “Donor station” or “User defined value” options then WINFAP does not apply urbanisation the urban adjustment to QMED. The work-around is to calculate the urban adjusted QMED value outside WINFAP (e.g. in Excel). This adjusted value can then be entered into WINFAP as a “User defined value” to allow the flood frequency curve to be derived, or the flood frequency curve can be calculated in a spreadsheet using the growth factors taken from WINFAP. The UAF used in v3.0.003 of WINFAP-FEH is calculated from a pair of formulae that have not been published together. Despite the message given within WINFAP-FEH, the adjustment method is not quite the same as that published by Kjeldsen (the reference is given as 2009 but the paper actually came out in 2010; it is listed in Related documents). UAF is calculated by WINFAP-FEH using Equation 8 from Kjeldsen (2010): UAF = (1+URBEXT2000)0.37 PRUAF2.16 where PRUAF is the percentage runoff urban adjustment factor, which quantifies the effect of urbanisation on percentage runoff. In Kjeldsen (2010) PRUAF is calculated from a formula based on BFIHOST (Equation 6 in the paper). However, when WINFAP-FEH v3 was developed it used an earlier version of the formula from Bayliss et al., 2007 in which PRUAF was calculated from SPRHOST:  70  PRUAF = 1 + 0.47URBEXT2000  −1  SPRHOST  It is noted that the BFIHOST formula can give near-infinite values for PRUAF as BFIHOST approaches 1. There are plans that in the next release of WINFAP-FEH the full Kjeldsen (2010) method will be implemented with a modification, as proposed by CEH, to avoid the discontinuity in catchments with a BFIHOST value of 1. Although there is some concern that the UAF formula applied in WINFAP- FEH was developed using values of PRUAF that were defined differently to their implementation in WINFAP-FEH, in fact the alternative formulae for PRUAF result in similar values of UAF on average for most catchments. The exceptions to this are permeable catchments.

Urban The two methods start to diverge for highly permeable catchments. Figure adjustment on 10 shows an example of the difference between urban adjustment factors highly for a chalk catchment with BFIHOST = 0.90. The actual urban extent for permeable this catchment is 0.15, for which the urban adjustment factor calculated from catchments BFIHOST is 1.5 times that calculated from SPRHOST.

Figure 10 Comparison of urban adjustment factors derived from BFIHOST and SPRHOST

Doc No 197_08 Version 5 Last printed 12/02/15 Page 41 of 110

Rhee at Ashwell: SPRHOST 11%, BFIHOST 0.90 16 14 12 10 8 UAF 6 4 2 0 0 0.1 0.2 0.3 0.4 0.5 URBEXT2000

UAF from Kjeldsen (2010), PRUAF from BFI UAF from Kjeldsen (2010), PRUAF from SPR (WINFAP-FE H v3)

Urban If you are dealing with an urbanised catchment for which BFIHOST > 0.8, adjustment of you need to be aware UAF calculated using the two PRUAF methods may QMED on be very different, and the UAF calculated in WINFAP-FEH is likely to be an highly underestimate. permeable Note that WINFAP-FEH v3 does not offer the option to calculate UAF from catchments an alternative method. The option to select versions 1 to 3 for urban adjustment applies only to the growth curve, not to QMED. You should not revert to the Bayliss et al. 2007 method for adjusting QMED because it was based on the original FEH regression formula for QMED. It is important to understand that urban, highly permeable catchments are outside the range of the vast majority of catchments from which FEH methods have been developed. There is very little data on the effects of urbanisation on highly permeable catchments and so it is difficult to know which method is more appropriate. On such catchments the only reliable flood estimates are likely to be derived from local flow data.

Additional notes on When using alternative software such as a spreadsheet to calculate QMED, adjusting you should record and justify what method you use for the urban adjustment QMED with the above issues in mind.

Adjustment of Recent research (Kjeldsen, 2010) has led to a revision of the urban growth curves adjustment procedures. The research recommends that the UAF is used only for the adjustment of QMED. Unlike in earlier methods, growth curves are to be adjusted using a new method that does not include UAF. You should use this new method for adjusting growth curves (see Pooled growth curves).

Data Transfer

Doc No 197_08 Version 5 Last printed 12/02/15 Page 42 of 110

The issues The main area where difficulty or disagreement can arise in QMED estimation is the selection of donor catchments, which are intended to improve the estimate of QMED. The FEH provided some guidance on selecting donor and analogue catchments (for example, 1 3.3 and 3 4), and further ideas emerged through the experience of FEH users. Science Report SC050050, see the FSR listed in Related documents, presents a revised method for applying donor catchments. This gives a more structured way of selecting donors, but data transfer is still a subjective process with no universally applicable rules. There is scope for disagreement even between experienced hydrologists.

Guidelines Item Guideline or advice 1 It is important to take heed of the latest research. The original FEH data transfer method, at least as interpreted by some analysts, has been shown to give estimates of QMED that are worse than those obtained from catchment descriptors in many cases. Note: The performance of various data transfer approaches was tested by applying them at the sites of gauging stations and then comparing the resulting adjusted QMED estimates with the best estimates, that is, those obtained directly from flood peak data. To perform well, a data transfer approach needs to give an adjusted QMED value that is closer (than the initial catchment descriptor estimate) to the best estimate of QMED.

2 The data transfer method presented in Science Report SC050050 stops using the term analogue catchment and uses a single local donor. This is selected purely on the basis of distance between catchment centroids. There is no requirement for the donor to be on the same watercourse as the subject site, although in practice this is said to be likely if the catchment centroids are close. The adjustment ratio is not applied in full. Instead it is moderated by a power term, a, so that the adjusted QMED at the site of interest is given by:

a  QMED  QMED = QMED  g,obs  s,adj s,cds  QMED   g,cds  where: • QMEDs,adj: adjusted QMED at the site of interest; • QMEDs,cds: initial estimate from catchment descriptors at the site of interest; • QMEDg,obs: estimate from observed data at the gauging station (donor site); • QMEDg,cds: estimate from catchment descriptors at the gauging station (donor site). 3 The FEH procedure for data transfer used the same equation, apart from the power term a. This reduces with distance between the catchment centroids. The adjustment has its full effect when the subject site is at a gauging station. The effect declines to quite small once the centroids are more than 10 km apart; see Figure 11.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 43 of 110

4 The research underlying the revised data transfer method involved comparing the performance of alternative techniques for selection of donor or analogue catchments. It found that identification of potential donor catchments should be based on geographical closeness rather than on hydrological similarity as defined by catchment descriptors. It did not examine the option of considering both distance and similarity, partly because it was considered difficult to automate the subjective process that analysts might adopt in selecting donors in order to test the process on a national scale. However, when considering an individual application it makes hydrological sense to consider the physical similarity of catchments as well as their proximity. See Figure 12 for an example from a real study.

5 As in the FEH data transfer method, particular caution is required when proposing a transfer to or from a catchment affected by urbanisation, reservoir development, opencast mining, forest drainage or other major artificial influence (3 4.6). You should also be careful if flow is known to be out-of-bank below QMED in either the subject or donor catchments, resulting in attenuation of QMED. One way to estimate the potential for significant attenuation is to check the value of FPEXT.

6 A donor site should have good quality flood data, which will generally mean it is classed as suitable for QMED by HiFlows- UK. However, a review of the rating is worthwhile for high risk studies. Donor sites with longer records are preferable to those with short records, especially if the short records are thought to cover an atypical period in terms of flood frequency.

7 In some cases, there will be several suitable donors at similar distances from the subject catchment. Analysts: You should calculate adjustment factors for two or three potential donors in this case, rather than automatically selecting the one that happens to be nearest by what could be a small margin. If the various donor sites give similar adjustment factors, then this should strengthen confidence in the resulting estimate of QMED. If there is a wide variation in adjustment factors, then it is worth carrying out a more detailed review of the suitability of the potential donor catchments, in terms of both data quality and relevance to the subject site, before making a final choice.

8 There is no current definitive guidance on how to calculate an adjustment factor based on more than one gauge. Science Report SC050050 mentioned it as a possible topic for future research. The FEH suggests using a weighted average (3 4.4), where the weights reflect the suitability of the donor and the quality of the QMED estimate. It also recommends using a geometric mean rather than an arithmetic mean, which is not appropriate for the averaging of ratios.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 44 of 110

It would be possible to apply such an approach with the revised adjustment procedure. a  QMED   g,obs   QMED  Example: Calculate a value for  g,cds  at several donor sites. Then derive a weighted average of the individual factors.

9 You should check adjusted estimates of QMED to ensure they are consistent with observations at upstream or downstream gauging stations. Consistency is not guaranteed when using the data transfer method in SC050050. In some situations applying the power term, a, from the revised transfer procedure can lead to inconsistent results with upstream and/or downstream sites. In these cases you are advised to ignore the moderation term and use a more appropriate adjustment factor. See Figure 13 for an example.

Figure 11 The graph below shows the relationship between moderation term, a, and the distance between centriods, dij.

Graph provided in personal communication from Thomas Kjeldsen at CEH Wallingford.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 45 of 110

Figure 12: Description of the site example from The Morton Beck near Keighley in West Yorkshire has a catchment area of a real study 10 km2. The gauged catchment with its centroid closest to that of the Morton Beck is the River Aire at Lemonroyd Weir with an area of 865 km2. The centroids of the catchments are just 1.8 km apart, yet the catchments are clearly very different (one is nearly 100 times the size of the other). Decision In reality, Lemonroyd Weir would not be suitable as a donor because it is urbanised. But even if it had been rural, it would have been difficult to believe that it could be relied on as a valid donor site for the Morton Beck catchment.

© NERC (CEH). © Crown copyright. © AA. 2006. All rights reserved.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 46 of 110

Figure 13: an Description of the site example The site of interest is Crickhowell. The most appropriate donor appears to be Llandetty. It is shortly upstream on the same river and the catchment centroids are just 3.6 km apart. However, this short distance is enough to reduce the adjustment factor from 2.03 at Llandetty to 1.45 at Crickhowell, when applying the revised data transfer method. Applying the adjustment factor When the adjustment factor of 1.45 is applied to Crickhowell, the resulting estimate of QMED is less than the QMED at Llandetty, despite being 10 km downstream. This reduction in QMED is unlikely in practice and is merely a product of the new geographically weighted method. The analyst's decision The analyst decided (wisely) to override the recommended use of the moderation term (a) and assume an adjustment factor of 2.03 at Crickhowell. An alternative, particularly if flood estimates had been required at multiple locations within the reach, would be to calculate a weighted average of the adjustment factors at the upstream and downstream gauging stations, perhaps basing the weights on distance along the river and again ignoring the moderation term.

Using the Version 3 of WINFAP-FEH enables you to automatically identify the nearest software donor and calculate the moderated adjustment factor. You can also select another donor if preferred, from a list ranked by distance between the catchment centroids, see the screen on the following page.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 47 of 110

The list includes information on the catchment descriptors of the potential donor sites and links to pages on the HiFlows-UK website. Use these facilities as an encouragement to explore the information provided rather than automatically picking the closest donor site. Important note: If you set up the load options in WINFAP-FEH to read in only stations classed as suitable for pooling, the list of potential donor sites will miss some stations that are classed as suitable for QMED for not pooling. You should look at other sources of information on available donors such as the display of station locations on the FEH CD-ROM or the HiFlows- UK website, hosted by CEH.

Summary of You should carry out data transfer in all cases where QMED is estimated at advice on data an ungauged site, apart from very low risk studies. transfer Donor sites should be: • close to the subject site; • classed as suitable for QMED in HiFlows-UK or shown to be suitable in a more recent review of data quality; • rural in most circumstances (URBEXT2000<0.030), even if the subject catchment is urbanised (3 4.6.1); • not strikingly different from the subject site in terms of the key catchment descriptors: size, soils, wetness and reservoir/lake influence. In most cases, the chosen donor should be the closest to the subject site. If there is more than one potential donor at similar distances, you should consider them and compare their adjustment factors. If necessary, you could calculate a weighted average. You should moderate the adjustment using the power term calculated from the distance between catchment centroids as described above. You should check the adjusted QMED for consistency with QMED estimated from flood peak data at any upstream or downstream gauging stations. Since data transfer can be a subjective process, it will often be worthwhile seeking a second opinion from a more experienced colleague. It is also particularly important to record the process of decision-making.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 48 of 110

Statistical methods - growth curves

Pooling groups

The issue Pooling data from hydrologically similar sites provides more data and enables more reliable estimates of the growth curve for rarer floods (3 6.1 and 16.1). Pooled analysis is essential for an ungauged catchment and necessary in most other cases, except when the record length is more than twice the target return period (3 6, 11.5 and 16).

Guidelines

Item Guideline or advice 1 Aspects to consider are: • catchments within a pooling group should be essentially rural (3 6.1); • the subject site should be excluded from its own pooling group when the subject catchment is urbanised. 2 Science Report SC050050 (listed in Related documents) changed the way pooling groups are formed. You should implement these changes using versions 3 of the FEH CD-ROM and WINFAP-FEH. BFIHOST is no longer used in the revised method. Instead you can measure catchment similarity by AREA, SAAR, FARL and a new catchment descriptor that measures the extent of floodplains (FPEXT).

3 Another aspect of the revised method is to fix the size of pooling groups to 500 station-years of data, irrespective of the return period of interest.

4 You should review pooling groups (3 6.3, 6.6, 16.3 and 16.6). The extent of the review and any modifications depends on the purpose of the study and your experience. In most cases, modifications to the pooling group tend to have a relatively minor effect on the final design flow (compared with, for example, selection of donor sites for QMED). In particular, sites that are least similar to the subject site (that is, placed near the bottom of the pooling group) list have little influence on the pooled growth curve because of the low weights allocated to them.

5 One trigger for a review of the pooling group can be the presence of a discordant site or a high value of heterogeneity. However, the FEH advises experienced hydrologists to take a precautionary approach, reviewing the pooling group before using the statistical tests for discordancy or heterogeneity. It is vital to remember that you should not remove sites from the pooling group just because they are discordant or they reduce the heterogeneity (3 16, 6.5). In many cases, discordancy is due to the presence of extreme floods in the annual maximum series. In this

Doc No 197_08 Version 5 Last printed 12/02/15 Page 49 of 110

case, you should normally leave the discordant site in the group. However, you should exclude all records shorter than eight years (3 16.2.3).

6 The review should assess physical and hydrological differences between subject and pooled catchments such as: • topography, geology, reservoirs, lakes and floodplains; • local climate; • urbanisation and other anthropogenic activity (3 9, 21); • station locations and periods of record; • flood seasonality; • quality of high flow data (refer to the station information in HiFlows-UK).

Doc No 197_08 Version 5 Last printed 12/02/15 Page 50 of 110

Pooled growth curves

The issue There is no way of knowing which distribution is the correct choice for fitting to the pooled growth curve, because the underlying 'parent' distribution is unknown. On average, the GL distribution is considered to perform better than the GEV for pooled growth curve derivation (3 7.3, 15.3 and 17.3.2). For some pooling groups, other distributions are found to fit better than the GL. Analysts: you should usually select the distribution that gives the best fit.

Item Guideline or advice 1 For catchment-wide studies, it is acceptable to select different distributions for different locations, although imposing a single distribution may help to ensure that design flows are spatially consistent. The choice between distributions often has a fairly minor effect on the resulting design flow for return periods within the recommended range of the statistical method (2-200 years).

2 The method of weighting the L-moments from each catchment in the pooling group was changed in Science Report SC050050. Weights are calculated from record length (as in the original FEH method) and the distance in catchment descriptor space from the target site, rather than from the rank within the pooling group. So moving catchments up or down the ranking order does not alter the weights. Another change is that separate weights are calculated for L-CV and L-skew. These improvements are in v3 of WINFAP-FEH.

3 You will need to adjust the growth curve for urbanisation if the subject site is urbanised. As for QMED, it's sensible to carry out the adjustment even for catchments with URBEXT2000 below the threshold of 0.030. There have been three versions of the urban adjustment: v1 in the FEH; v2 described in Morris (2003) and Bayliss et al. (2007) (2007); and v3 published by Kjeldsen (2010). The references are listed in Related documents. Version 3 of WINFAP-FEH allows you to use either v2 or v3 of the urban adjustment. You should normally use v3. Unlike earlier versions, which adjusted the growth factors, v3 adjusts the L-moments. The v3 adjustment to growth curves is based solely on the value of URBEXT2000, another change from earlier versions which used the same urban adjustment factor as is used for QMED (calculated from URBEXT and also SPRHOST).

The v3 formulae for adjusting the L-moments are: Doc No 197_08 Version 5 Last printed 12/02/15 Page 51 of 110

URBEXT2000 L-CVurban = L-CVrural x0.5547

URBEXT2000 L-skewurban = ((L-skewrural+1) x 1.1545 ) -1

The interpretation is that L-CV decreases on urban catchments and L-skew increases. These changes tend to reduce the gradient of the growth curve at lower return periods and increase the gradient at higher return periods as shown in Figure 14 for an example site. The effect of the v3 urban adjustment is minor at this site in comparison with the v2 adjustment. However, results elsewhere will differ depending on the values of the L- moments as well as on URBEXT2000.

Figure 14 URBEXT2000=0.4, SPRHOST=42% 6 Comparison of urban 5 adjustment for growth 4 curves 3

Growth factor 2

1

0 1 10 100 1000 Return period (years)

v2 adjustment v3 adjustment No adjustment

On some urban catchments you may find that design flows for longer return periods increase substantially over those from previous studies as a result of the new adjustment method for growth curves. You should discuss the implications of these changes with experienced colleagues before deciding to adopt the revised design flows.

Bug in WINFAP-FEH V3 lets you supply a user-defined value of URBEXT2000. WINFAP-FEH This is correctly applied in adjusting QMED for urbanisation, but WINFAP- FEH ignores the user-supplied value when adjusting the growth curve. The difference in the results is likely to be minor in most cases. You should not normally apply an urban adjustment to a growth curve derived from enhanced single-site analysis (see below) because enhanced single-site analysis is not suitable for urban catchments. If carrying out calculations outside WINFAP-FEH (for example, using a spreadsheet), take care not to apply an urban adjustment to a single site growth curve.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 52 of 110

Growth curves for sites with flood peak data

The issues In deriving flood growth curves at a flow gauging station, the choice between single site and pooled curves can have a large impact on the results. Originally, the FEH’s basic recommendation was to rely on the pooled growth curve unless there is a flood peak record at the site of interest, twice as long as the return period required (T). However, you can give some weight to the single site curve if the record length is between T and 2T. As usual in the FEH, there is some flexibility about this. Other factors to bear in mind are: • the quality of flood peak data; • the longer-term flood history; • and any unusual characteristics of the catchment compared with others in its pooling group. For rural sites, the choice between single-site and pooled curves is now simpler due to the introduction of enhanced single-site analysis in v3 of WINFAP-FEH. If the subject site is gauged, it is given a lot more weight than the rest of the sites in the pooling group. This helps to remove some of the very large differences that have been observed between pooled and single site growth curves. You can find the details of the enhanced single- site method in Science Report SC050050.

Figure 15 The graph compares a pooled curve in which the subject site is treated as ungauged with one derived from an enhanced single-site analysis. The latter is much close to the single site curve because a large weight is given to the subject site due to the long length of its annual maximum series.

Guidelines Item Guideline or advice 1 You should normally carry out an enhanced single-site analysis when deriving a pooled growth curve for a site draining a rural catchment with at least 8 years of good- quality flood peak data. You should compare the resulting growth curve with a plot showing the annual maximum flows and a standard single-site growth curve. Be aware that WINFAP-FEH will default to enhanced Doc No 197_08 Version 5 Last printed 12/02/15 Page 53 of 110

single-site analysis when you create a pooling group for a gauging station’s catchment descriptor (.cd3) file taken from HiFlows-UK, when the gauge is classed as suitable for pooling and the catchment is rural. If you create a .cd3 file by extracting the catchment descriptors from the FEH CD-ROM at the site of a gauge, the site will be treated as “ungauged” by WINFAP-FEH and a conventional pooled analysis will be carried out. Data from the gauging station will be included in the analysis, but without the extra weight used for enhanced single- site analysis. WINFAP-FEH does not report the relative weights used in enhanced single-site analysis. When your subject site is urban (URBEXT2000>0.030) you should normally avoid enhanced single-site analysis because the pooling group would contain a mixture of urban and rural sites, and it would then not be possible to apply a valid urban adjustment to the growth curve. In that case you should fit a standard pooled curve (adjusting the pooled L-moments for urbanisation) and a single-site curve and compare them. If necessary, you could use joint analysis (3 8.2) to produce a compromise growth curve, giving increased weight to the pooled L- moments at longer return periods.

2 It is important to realise how fickle a single site analysis can be. When extrapolated to the typical return periods used in fluvial flood studies, single-site growth curves can be very vulnerable to: • the period of record that the gauging station happens to cover; • and to the quality of high flow data. It is all too easy to derive a single-site flood frequency curve that appears to fit the annual maximum data, but is a long way from the true underlying distribution (which we can never fully know). See the illustration in Figure 16,. So just because you prefer the look of the single- site growth curve it does not mean that you should use that curve if you cannot justify it based on statistical arguments and an understanding of the catchment’s hydrology.

3 The two basic approaches to improving on extrapolation of single-site data are: • search for historic data; • add data from other sites by pooling. You should attempt both approaches in many hydrological studies. There is a paper comparing different approaches to extrapolating flood growth curves; see Gaume (2006).

4 In some cases, the difference between the single-site and pooled curves is so wide that it is clear something is wrong. Example: The pooled curve might lie so far below the single-site data that the top few flood peaks all appear to have return periods longer than 10,000 years according

Doc No 197_08 Version 5 Last printed 12/02/15 Page 54 of 110

to the pooled curve. In such cases, it is particularly important to check that the rating can be relied on for the highest flows on record. If it can, then it is very likely that the pooled curve is too flat.

5 If you have several flow estimation points, some of which are at gauging stations, you may find large changes in growth curves over short distances if you apply single-site or enhanced single-site analysis only at the gauges. You should ensure a smooth variation in growth curve, choosing and applying the preferred growth curve(s) manually to all flow estimation points.

Figure 16: an This example illustrates how easy it can be to derive a single-site flood example frequency curve that appears to fit the annual maximum data, but is a long way from the true underlying distribution (which we can never fully know). The graphs below show four plots of annual maximum values of a variable (for example, river flow). Each plot has 33 years of data, the mean record length in the Hiflows-UK dataset. Each plot includes a curve plotted for return periods up to 100 years. In some cases, the curve fits the data well and in others, the fit is rather poor, especially for long return periods. It may be tempting to try to redraw some of the curves so that they fit the data better. However, in this case it would not be right to alter the curves. Here, the underlying distribution is known and the points on the plots are not real observed data. They are all random samples from a Generalised Logistic distribution: location: 5, scale: 0.5, shape: -0.1. This distribution is shown by the curve plotted on each graph. Some of the samples, like the first, are quite well representative of the underlying distribution, but others have rather more or rather fewer high values than would be expected in a typical period of 33 years.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 55 of 110

15 15

10 10

5 5 Variable (e.g. flow) Variable (e.g. flow)

0 0 -5 -3 -1 1 3 5 -5 -3 -1 1 3 5 Logistic reduced variate Logistic reduced variate

15 15

10 10

5 5 Variable (e.g. flow) Variable (e.g. flow)

0 0 -5 -3 -1 1 3 5 -5 -3 -1 1 3 5 Logistic reduced variate Logistic reduced variate

Applying the To apply the illustration above to a typical FEH problem, imagine that one of illustration the lower two plots shows a pooled growth curve along with single-site flood peak data. One interpretation would be that the pooled curve is underestimating the correct distribution. But this example shows that it is quite possible for the sample flood peak data to plot some distance away from their underlying distribution, due, for example, to the gauged record covering an unusually flood-rich period. So it is quite possible that the pooled growth curve would be a correct representation of the underlying distribution. This is why the FEH recommends only relying on a growth curve fitted to single-site data for return periods up to half the record length.

Rainfall-runoff approaches

Topics in this This section covers the FEH rainfall-runoff method and the ReFH method, section released in early 2006. ReFH has superseded the FEH method for most applications (the main exceptions being reservoir safety and pumped catchments). Analysts: you can refer to FEH Supplementary Report No. 1 for details of the ReFH method. For information on the research, see Kjeldsen and others. (2005). Both are listed in Related documents.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 56 of 110

General information

Guidelines Item Guideline or advice 1 Both the FEH rainfall-runoff and ReFH methods use the FEH rainfall frequency statistics to create design rainfall events, which form the input to the rainfall-runoff model. Storm profiles were not investigated in the FEH rainfall frequency research. The FEH rainfall-runoff and ReFH methods continue to use the profiles given in the FSR. The two profiles recommended for use in the rainfall- runoff method are: • the 75% winter profile, for rural catchments; • the 50% summer profile, for urban catchments. These profiles are recommended for durations of 'up to several days' (2 4.2). There is no guarantee that a rainfall profile of a shape other than the recommended one will produce a design flood of the required return period (2 4.3).

2 Alternative rainfall-runoff methods are occasionally used for UK flood estimation, such as the NAM model in MIKE-11 or FRQSIM (used mainly in Greater London). FEH rainfall statistics could be used to provide an input to such models, with any storm profile, as long as the catchment model was calibrated so that the combination of inputs results in a flood of the required return period (2 4.1). The onus is on the analyst to demonstrate that this is so, if using an alternative rainfall-runoff model. Important note: You must consider the FEH and ReFH rainfall-runoff methods as a complete package. They were both calibrated so that the recommended design inputs gave rise to an output hydrograph with a peak of the required return period.

Lumped or distributed approach?

The issue A fundamental decision regarding any rainfall-runoff technique is whether to apply it: • in a lumped fashion to the entire catchment upstream of the site of interest; • or in a distributed approach, splitting up the catchment and routing the design flows from each subcatchment. In practice, this decision is often dictated by the nature of the study. Example: Catchment-wide hydrodynamic modelling studies tend to follow a distributed approach.

Storm Lumped durations for When modelling a lumped (individual) catchment, the storm duration should lumped and be set to the recommended value given by the equation based on time to distributed peak and SAAR. This equation tends not to give the critical duration Doc No 197_08 Version 5 Last printed 12/02/15 Page 57 of 110

models (particularly when using the ReFH model), but it matches the duration that was used in the calibration of the ReFH model’s design event. Distributed In a distributed rainfall-runoff application, it is vital to apply an identical design storm (in terms of duration and areal reduction factor) to each subcatchment. Using an individual design storm for each subcatchment, with a duration set to the critical duration of the subcatchment, is physically unrealistic and will overestimate the combined flood peak. You should try a realistic range of durations for the design storm, to find the critical duration at the subject site by trial and error. This optimisation can be carried automatically in ISIS. The critical duration is the one that gives the highest flow (or water level or storage pond volume, for some studies) at the site of interest. At an early point in the study, you should select, agree and record the location of that site. When there is significant variability in rainfall patterns over a large area, you can derive the rainfall depth separately for each subcatchment, as long as a common return period is used. When the design storm duration is set to a value much longer than the critical duration for a subcatchment, beware that the ReFH method can overestimate the flow (see The ReFH method).

Guidelines Item Guideline or advice 1 Analysts: there is a choice: a distributed approach is the natural choice for large or varied catchments and for those with floodplain or reservoir storage. But it can introduce great complexity and force you to make uncomfortable assumptions.

2 In a distributed application, it is important to avoid excessive detail in subdividing catchments. Observed flood hydrographs can help to identify multiple peaked events, which may indicate differing responses from subcatchments. All subcatchments should result in a significant change in catchment area when added to the upstream area.

3 Areas draining directly to the modelled watercourse, or containing numerous small subcatchments, are usually treated as 'intervening areas', see Figure 17

4 You can estimate catchment descriptors for intervening areas by area weighting, using the upstream and downstream lumped catchments (at points A and B in Figure 15), or based on the descriptors of a significant watercourse within the intervening area. Take care over some descriptors, particularly DPLBAR. You can calculate it for an intervening area from DPLBAR, LDP and AREA for the upstream and downstream catchments. It is unwise to rely on the regression equation for DPLBAR in 5 7.2.4, which is designed for real catchments, not intervening areas.

5 Estimate hydrographs for intervening areas by applying FEH methods to the derived catchment descriptors, as for any other subcatchment. However, intervening areas are not real catchments, so the FEH methods are not strictly applicable to them. For this reason, intervening

Doc No 197_08 Version 5 Last printed 12/02/15 Page 58 of 110

areas are best kept to a minimum.

6 An alternative approach to estimating hydrographs for intervening areas, which avoids having to define catchment descriptors, is to estimate hydrographs for the lumped catchment upstream of (excluding) the intervening area and downstream of (including) the area. The upstream hydrograph is subtracted from the downstream one to give the hydrograph for the intervening area. You should check the resulting hydrograph to ensure that its shape is physically realistic.

7 One important use of intervening areas comes in examining flood risk for locations downstream of a reservoir (or other storage). If the site of interest is some distance downstream of the reservoir, it's important to check whether the reservoir can attenuate flood flows to such an extent that the site is more sensitive to heavy rainfall (over a shorter duration) concentrated on the intervening area downstream of the reservoir than it is to an event over the whole catchment.

8 Analysts: you may be interested to know about a recent research study (SC060088) investigating spatial coherence of flood risk. The report on the research will include a discussion of the implications of modelling spatially distributed flood flows without taking into account the statistical dependencies between the flood frequency curves at different places. The approach taken in the FEH and ReFH methods makes an assumption of complete dependence between rainfall in different parts of the catchment.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 59 of 110

Figure 17: an The map below shows an example of an intervening area at Little Don at example Stocksbridge, South Yorkshire. The intervening area is the catchment at B minus the catchment at A.

© Crown Copyright. All rights reserved. Environment Agency, 100026380, (2009).

The ReFH method

Summary of The ReFH method was developed to address several problems in the FEH the method rainfall-runoff method, which was largely unchanged from the earlier Flood Studies Report method. Figure 18, below, summarises the ReFH method.

Figure 18

Reproduced from Kjeldsen and others (2005) with the permission of CEH Wallingford.

Four The four model parameters shown inside the boxes in Figure 15 are: parameters • Cmax, the maximum of the range of soil moisture capacities across the catchment; • Tp, the time to peak of the unit hydrograph;

Doc No 197_08 Version 5 Last printed 12/02/15 Page 60 of 110

• BR, the baseflow recharge (ratio of runoff to recharge); • BL, the baseflow lag (rate of exponential decay of baseflow). All four parameters are best estimated from hydrometric data, where available, using the ReFH design flood modelling software. The baseflow parameters are calculated by fitting recession curves to flow data. Cmax and Tp are calculated jointly by an optimisation method that requires observed rainfall, flow and evaporation data. In the absence of data or for low risk applications (for example, where ReFH is being used to determine a hydrograph shape that will be fitted to a peak flow from the Statistical method), parameters can be estimated from FEH catchment descriptors. Note that the FEH supplementary report gives parameters calculated from flow data at 101 gauging stations and you can use this to replace the catchment-descriptor values in the spreadsheet.

Differences The table below lists some differences between application of the FEH rainfall-runoff and ReFH methods Item Difference 1 The time to peak for ReFH is not quite the same as that used in the FEH rainfall-runoff method. It is estimated differently, whether from data or from catchment descriptors.

2 ReFH does not include the provision to use river level data for deriving time to peak, as in the FEH rainfall- runoff method. However, given the wider availability of river level recorders, there are likely to be some situations where analysts judge that level data are helpful in guiding the selection of parameters for the ReFH method. This could be done, for example, by assuming that the time to peak in ReFH can be adjusted using a factor derived from comparing the catchment-descriptor estimate of time to peak with that derived from lag analysis.

3 The ReFH research did not examine the value of data transfer for refining parameters. The report, therefore, gives no recommendations on use of donor sites. Faulkner, D.S. and Barber, S. (2009), listed in Related documents, have published more recent research showing that using the closest available gauge from the ReFH calibration dataset as a donor site appears to offer no benefit on average, in comparison with estimating parameters from catchment descriptors. However, it seems highly likely that many subject sites, with a donor site nearby on the same watercourse, will benefit from data transfer.

Analysts: you should consider data transfer when there is a flow gauging station nearby on the same watercourse as the subject site. This involves estimation of each of the four model parameters at the gauging station from flow and rainfall data using the ReFH design flood modelling software and also from catchment descriptors. For each parameter, the ratio of the two Doc No 197_08 Version 5 Last printed 12/02/15 Page 61 of 110

estimates at the gauging station is used to adjust the catchment-descriptor estimate at the site of interest. Do not use the moderation factor (power term) developed for data transfer of QMED for adjusting ReFH parameters.

4 ReFH allows for more seasonal variation in the design event. As well as the choice of winter or summer rainfall profiles (as in the FEH), it also adjusts the rainfall depth and the initial soil moisture for the season. The ReFH Technical Report recommends using the summer design event on heavily urbanised catchments (URBEXT1990>0.125). However, current guidance is that ReFH is not used for such catchments (see When to apply ReFH with caution below).

5 ReFH uses equal return periods for the input rainfall and output flow hydrograph.

6 ReFH was calibrated for return periods up to 150 years. This is considerably longer than the 10-year return period limit of calibration for the FEH rainfall-runoff method, which has been widely used for estimation of design flows for extreme return periods. For guidance on estimation of extreme events, refer to Estimating long return period floods (150-1000 years).

When to apply One of the most significant aspects of ReFH is that the design event was ReFH with calibrated to match, on average, flood frequency curves derived from pooled caution analysis at 100 gauging stations (using the HiFlows-UK dataset). For this reason, ReFH tends to give peak flows that are much more consistent with those from the FEH statistical method. However, due to limitations of the calibration, there are some catchment types where ReFH should be applied with caution, if at all. Links are to other sections with more details. • Only seven of the calibration catchments were heavily urbanised and so the report states that applicability of ReFH to urban catchments (URBEXT1990>0.125) is 'unclear without further research'. On many urban catchments, the results of ReFH appear suspect because the winter season event tends to give higher flows than the summer event, contrary to expectations. ! Important Don't use ReFH in its original form to estimate peak flows on heavily urbanised catchments. However, recent research (Kjeldsen, 2009) has led to a modified version of ReFH that accounts for increased runoff volumes on urban catchments. Refer to Development control and urban catchments for more information on this. • As with other rainfall-runoff methods, ReFH struggles on permeable catchments (approximately BFIHOST>0.65). The calibration procedure gave unrealistically large values of Cmax to reproduce the losses on permeable catchments. The flood event archive was deficient in events on permeable catchments. On permeable catchments, ReFH has been found to underestimate QMED by a long way and to give unrealistic return periods (>10,000 years) for the July 2007 floods (Faulkner, D and others). ! Important You should not use ReFH to estimate peak flows on permeable catchments. The statistical method is normally a better choice. • The largest catchment used for calibrating the ReFH design event was

Doc No 197_08 Version 5 Last printed 12/02/15 Page 62 of 110

511 km2. The method was validated for catchments up to 750 km2. The FEH Supplementary Report recommends applying ReFH for areas up to 1000 km2. Rainfall-runoff approaches are less valid on large catchments because they rely on a characteristic distribution of rainfall across the catchment. This limitation applies equally when a large catchment is split into subcatchments. • ReFH has been found to overestimate design flows when it is run with a storm duration much longer than the critical or recommended durations for the catchment. On impermeable and wet catchments (BFIHOST<0.4 and PROPWET >0.45), it can even give a flow volume that is greater than the volume of the input rainfall. This needs further investigation, but in the meantime check that the effective percentage runoff is realistic and use an alternative method if necessary. When hydrographs are required for catchments unsuitable for ReFH, you may use the method to derive a hydrograph shape which could then be fitted to a peak derived by a more suitable method.

Available data You can apply the ReFH method using a spreadsheet available free from and software CEH Wallingford. This can be downloaded from the Centre for Ecology and Hydrology website, or Environment Agency staff should obtain it via the Service Desk. The spreadsheet does not allow calculation of model parameters from observed data, which is computationally complex. Instead, it calculates parameters from catchment descriptors. However, the FEH supplementary report gives parameters calculated from flow data at 101 gauging stations and the spreadsheet allows users to modify parameters. The spreadsheet is only intended for design rather than simulation of flood events, that is, it cannot be run using observed rainfall. Wallingford HydroSolutions Ltd. released a more comprehensive ReFH software package in 2007. The ReFH design flood modelling software includes: • a database for storing flood and rainfall event data; • software for fitting model parameters to observed data; • software for running the ReFH model with both design and observed events • reservoir routing. There is also a ReFH unit within version 2.5 of ISIS. Users and project managers: When you use ReFH to estimate peak flows or flood volumes at or near a flow gauging station, you should normally estimate the parameters from observed data using the ReFH design flood modelling software, or the published values if the station is listed in the FEH supplementary report. As with other FEH methods, parameters estimated from local data are likely to give significantly more accurate results. If the study is simple or routine, see Table 2, it may be acceptable to base the parameters on catchment descriptors. Project managers: You should be aware that applications of ReFH that involve estimation of parameters from observed data will take much more time than those that rely on catchment descriptors. The ReFH method has not yet been updated to use URBEXT2000, so you should base all ReFH calculations on URBEXT1990 (updated to current levels of urbanisation). RefH 2 software is currently in development. The software is currently (October 2014) being beta tested, and is not yet available. Guidance on its Doc No 197_08 Version 5 Last printed 12/02/15 Page 63 of 110

use (including when it can be used) will be issued once the final version is released.

FEH rainfall-runoff method

The issues ReFH has superseded the FEH rainfall-runoff method for the majority of river management applications. However, there are some situations where the earlier method is still applicable and others where there will be interest in comparing results. The main examples are on pumped catchments and for reservoir safety. Select the links to read more details in later sections. You can also consider application on some heavily urbanised catchments, given the unsuitability of ReFH for such catchments. The main decisions in applying the rainfall-runoff method are about how to estimate the model parameters, with preference given to local data as is usual for FEH methods.

Guidelines Item Guideline or advice 1 You can use donor catchments to estimate the parameters when there is no data at the site of interest.

2 In selecting donors, use the following guidelines: • you can identify donors using similar criteria to those for selecting QMED donors; See From catchment descriptors. • Tp(0) depends only on the timing of flood peaks, so there's no need for high quality flow data; You can estimate it by lag analysis at any flow or level gauging station as long as there is a recording raingauge close enough to the catchment. • the requirements for SPR donors are rather less strict than those for Tp(0) donors: it is sufficient that catchments are similar in terms of soils, land use and topography. 3 Most studies estimate Tp(0) from a lag analysis and SPR through the baseflow index (which is readily available at most gauging stations from the Hydrometric Register and Statistics publications by CEH Wallingford). The FEH describes this as 'indirect analysis of gauged data', which comes second to a full flood event analysis, that is, deriving a unit hydrograph. However, there is no commercial software widely available for flood event analysis and so it is not widely done. Analysis of five flood events is the minimum required for confidence in the results (4 2.1.2).

4 Appendix A of FEH Volume 4 contains results from the UK Flood Event Archive, giving values of Tp(0), SPR and baseflow for several hundred gauging stations. It is worth checking there for model parameters before deciding to estimate them from flood event data. You should check flood peaks from the events in Doc No 197_08 Version 5 Last printed 12/02/15 Page 64 of 110

Volume 4 against those in Hiflows-UK for any changes in rating equations.

Continuous simulation - an alternative rainfall-runoff approach

Description Continuous simulation of flows offers an attractive alternative to design event methods such as FEH and ReFH. The idea of this is to produce long series of simulated rainfall data (for example over 1000 years) and run them through rainfall-runoff models to produce a long flow series. You can then rank the peaks of the flow series and analyse them to obtain design flows of the required return period. This removes many of the assumptions and restrictions of the design event approach. The method allows you to incorporate complex dependencies within the catchment (for example, flood control structures) and also deals with problems of spatial dependence if it is driven by a suitable spatial rainfall model.

Research A research project has developed methods of estimating continuous rainfall- runoff model parameters from catchment properties or by transfer from similar gauged catchments; see Calver, A. and others (2005), listed in Related documents. Research and testing of continuous simulation is continuing, but it has already been applied to flood estimation on some catchments judged to be too complex for FEH methods, such as the Don in South Yorkshire where flood flows are controlled by regulators and washlands. See Faulkner, D.S. and Wass, P. (2005), listed in Related documents.

Assumptions, limitations and uncertainty

Overview – a common criticism

Two common Two of the commonest criticisms of project reports on flood estimation are criticisms that they: • fail to acknowledge the assumptions and limitations of the methods used; • do not discuss the uncertainty of the results. See Pappenberger, F. and Beven, K. J. (2006), listed in Related documents.

Possible One possible reason for the lack of discussion of uncertainty is that, to most reasons hydrologists, it is all too obvious that flood estimation is uncertain. They don't see much value in talking about it, when the point of the exercise is to obtain the best estimate. Doc No 197_08 Version 5 Last printed 12/02/15 Page 65 of 110

Some analysts worry that project managers or decision makers could misuse statements about uncertainty, seeing them as carte blanche to choose an answer that suits their prejudices or their pockets. A more practical limitation is that there is no standard accessible method for quantifying the uncertainty in a design flow. Similar reasons probably explain why assumptions tend not to be acknowledged. They tend to be similar for many studies (so what is the point of listing them?) and many analysts would probably have difficulty identifying and describing all the assumptions that they implicitly rely on.

Why bother with uncertainty?

Does While it is obvious to most hydrologists that their flood estimates are uncertainty uncertain, there are probably many who don't have a good idea of how large matter? that uncertainty can be. There's also still a tendency among non-specialists to treat results of complicated procedures as the final truth, particularly if they are quoted to several decimal places. But does this matter?

Result of Uncertainty in flood estimates is often important to the subsequent process uncertainty on of making decisions. decisions Example: A method that gave more certain answers would tend to be preferred over a less certain alternative. Sensitivity analysis can be used to test the effects of uncertainty on the subsequent modelled water levels (or whatever quantity is of interest). If this shows that the results are too uncertain, then it might be an incentive to improve the flood estimate. However, often the only way to give a substantial improvement is to install a flow logger and wait until it has recorded enough data. These tests often show that modelled water levels are more sensitive to uncertainty in the design flows than in hydraulic model parameters, indicating that it's worthwhile spending time and effort on improving the design flows. In development control, uncertainty in a flood estimate may lead to a decision not to allow a proposed development (that is, a hazard), because there's not enough information on its consequences.

How Acknowledging uncertainty can affect how results are presented and uncertainty perceived. affects Although it may have apparent disadvantages, such as project managers perception taking the results less seriously or ignoring the best estimate, it can help avoid a crisis when one study appears to contradict a previous one. Example: A flood alleviation scheme was designed with a return period of 30 years. But the standard of protection was later reassessed at 50 years. If the latter result had been presented as 'between 30 and 70 years', the difference might not have seemed so great.

Importance of The Flood and Coastal Risk Management Modelling Strategy 2010–2015 uncertainty (see Related documents) states that “We will understand and communicate uncertainty in modelling outputs to assist decision-making by ourselves, our partners and our customers. We

Doc No 197_08 Version 5 Last printed 12/02/15 Page 66 of 110

will reduce any uncertainty that prevents us from making sound decisions.” An aspiration of the strategy is to use uncertainty in a positive way to gain a fuller understanding of the risks we are modelling. An example of this might be combining uncertainty estimates in design flows with defence failure probabilities and flood damage measures to obtain overall measures of flood risk.

Why we One of the main reasons for acknowledging assumptions and limitations is should that it forces the analyst to think through their work and identify and address acknowledge any weaknesses. uncertainty It also provides useful information for anyone reviewing the calculations. For this reason, we require a section describing limitations in hydrological studies and hydraulic models as part of reports produced under our PSC Scope for Modelling and Mapping services under WEM Lot 1.

Typical assumptions

General Many flood studies rely on some general assumptions, such as: assumptions • the flow data are recorded accurately; not that useful • the catchment descriptor equation for QMED is applicable to all sites in the study area; • the growth curve at the subject site is identical to that derived from the pooling group. Listing assumptions like these isn't very helpful because they are rather obvious, they are often very hard to test and they are not specific enough. To take things to an absurd extreme, you could simply state a single assumption: 'The flood estimates are assumed to be correct', which would be completely obvious and of no use.

Identifying the The most useful assumptions to identify are ones that: most useful • are specific to the study; assumptions • or can be tested; • or have a large effect on the results. Some examples (which are not necessarily recommended in any particular case) are listed in the table below. It may help to list assumptions grouped under similar headings to those used below.

Assumption Examples Assumptions • the rating curve at Station X can be extended up to about data QMED (this could be tested by carrying out some high flow gaugings this winter); • all large floods since 1800 have been identified during the historic review. Assumptions • flood flows arise mainly from runoff generated from about the impermeable parts of the catchment; hydrological • the catchment and watercourse have been largely Doc No 197_08 Version 5 Last printed 12/02/15 Page 67 of 110

processes unchanged since the historic data recorded in the early 20th century; • the pumping stations operate at full capacity during major floods. Assumptions • a single adjustment factor for QMED can be applied about the all the way along the study reach (this could be methods used tested by installing a temporary flow logger at the upstream limit); • the ReFH method will give improved answers if Tp(0) is adjusted using donor sites, even though the gauges are level-only and the Tp(0) has therefore been derived from lag analysis rather than the recommended optimisation method; • the 1000-year growth factors are best estimated from a rainfall-runoff approach, given the greater confidence in rainfall growth curves for longer return periods.

Typical limitations

Most common The most common limitations are due to applying methods outside the limitations range (of catchment size or type or return period) for which they have been developed or calibrated. It's important to acknowledge when this has happened.

Table 3 The table below summarises the validity ranges for selected methods, based on information in the FEH (mostly from 1 3.1) and other publications. These are ranges over which the methods are 'principally intended to be used' or ranges covered by the data used to develop the methods.

Method Return Catchment Urbanisation Other limits period area limits limits limits FEH 2 – 200 Over 0.5 URBEXT1990 Each method statistical years km2 but can up to 0.5 has various (but has be applied types of been for smaller catchment for applied areas which it is not up to ideal – see 1000 Choosing years) between the FEH methods

FEH 2 – 2000 0.5 to 1000 URBEXT1990 rainfall- years km2 but can up to 0.5

runoff be applied (largely for smaller superseded areas by ReFH)

ReFH Up to 0.5 to 1000 Only reliable 150 km2 but can for years – be applied URBEXT1990

Doc No 197_08 Version 5 Last printed 12/02/15 Page 68 of 110

but see for smaller <0.125 – but Note 1 areas see Note 2

FEH rainfall 2 – 2000 n/a n/a Durations 1 hour frequency years to 8 days

Notes to Table • See Estimating long return period floods (150-1000 years). 3 • See Development control and urban catchments for advice on applying a modified version of ReFH to more heavily urbanised catchments.

The information in Table 3, above, is not intended to say that you should never use the methods outside the ranges given.

Guidelines Item Guideline or advice 1 ! Important You should choose methods by following the guidance in Chapter 3, rather than by elimination using Table 3.

2 It is inevitable that on unusual catchments or for extreme return periods, there are few ideal methods. Standard methods are likely to be least applicable to very small and very large catchments, complex urban catchments, permeable catchments and extreme events. However, design flows are still needed in such cases and so it is often necessary to use a method outside the range for which it was calibrated or for which it is principally recommended.

Assessing uncertainty

The issues Flood frequency estimates are inherently uncertain because they cannot be measured or formally validated against observed data.

We often break uncertainty down into different components: • natural uncertainty, from the inherent variability of the climate; This tends to be the largest source of uncertainty in flood estimates for long return periods such as 100 years, because they are derived (however indirectly) from flood data series that rarely exceed 50 years in length. • data uncertainty, from the measurement of flood flows; • model structure uncertainty, from the choice of model, such the selection of a growth curve distribution; model parameter uncertainty, from selection of parameters for a growth curve or a rainfall-runoff model.

Qualitative One way of presenting information on uncertainty for a particular flood assessment estimate is a qualitative assessment of the relative contributions from the

Doc No 197_08 Version 5 Last printed 12/02/15 Page 69 of 110

various sources of uncertainty. Example: You can class the contributions as high, medium or low. Sources of uncertainty might include rating equations, length of a flood peak record, choice of pooling group, choice of distribution or ReFH model parameters.

Quantitative Quantitative assessment of uncertainty often uses confidence intervals. The assessment 95% confidence interval is the range within which we are 95% confident that the true answer lies. There are no widely available straightforward techniques for assessing confidence intervals for flood estimates (1 5.6). The FEH provides confidence intervals for some components of flood estimates, but does not suggest any techniques for combining them together and accounting for the other sources of uncertainty. See the examples on the FEH statistical method and rainfall-runoff techniques below.

Examples in Examples of quantitative assessment in the FEH statistical method include: FEH statistical • You can derive confidence intervals using the resampling routine in method WINFAP-FEH for single-site growth curves, but not for the much more widely used pooled growth curves, although this is theoretically possible and could be done if suitable software was developed. • You can obtain confidence intervals for QMED when QMED is derived from flood peak data (3 12.5) or catchment descriptors. In the latter case, you should replace the confidence intervals given in 3 13.8.1 using the factorial standard error associated with the revised QMED equation, which is 1.43. The revised 68% confidence interval for QMED is (0.70QMED, 1.43QMED) and the revised 95% interval is (0.49QMED, 2.04QMED). • The overall uncertainty is a combination of: • the variability of QMED; • the variability of the growth curve; and the covariance between QMED and the growth curve. At long return periods, the uncertainty of the growth curve may be the dominant factor. See Kjeldsen, T.R. and Jones, D.A. (2004), listed in Related documents.

Examples with Examples of quantitative assessment in rainfall-runoff techniques include: rainfall-runoff • You can produce confidence intervals for rainfall growth curves by techniques resampling. These measure the uncertainty in rainfall growth rates due to limitations in the sample size (but not due to other sources of error). The FEH describes how to evaluate these confidence intervals, but has not evaluated them at all sites because it was not computationally feasible. • An important factor is the uncertainty in estimating the index rainfall. You can estimate this approximately from maps in Volume 2. • As for flood growth curves (see above), you would need to combine the various components of uncertainty to give an indication of the overall uncertainty in rainfall frequency estimates. You would then need to combine this with the uncertainty due to the estimation of rainfall-runoff Doc No 197_08 Version 5 Last printed 12/02/15 Page 70 of 110

model parameters and the (large) uncertainty introduced by the composition of the design event package.

Analysts: what It is clear that uncertainty is an uncertain business. you need to do Assessing uncertainty for flood estimates remains a matter for researchers. But you should still quote what information you can about the uncertainty of their results, rather than simply copying general text about uncertainty from a previous report. See also the example in Figure 19.

Figure 19: an A design flow of Q, based on a QMED estimated from catchment example of descriptors, has a 95% confidence limit of at least 0.49Q, 2.04Q. In other uncertainty words, the true value may be less than half or more than twice the best estimate. Results for longer return periods are considerably less certain. This degree of uncertainty may be surprising and worrying for many people. It is important to realise that a wide confidence interval does not necessarily mean that the best estimate is wrong. It is much more likely to be correct than are the values at the upper and lower confidence limits, as illustrated in the diagram below. The result shows that the typical allowances made for the possible effects of climate change (an increase of 20 or 30% on peak flows) are much less than the uncertainty in many flood estimates. The sketch below gives the probability density function illustrating uncertainty in a flood estimate for a given return period.

Probability

Flood estimate

Probability density function showing uncertainty Best estimate 95% confidence limits

Based on a sketch graph in: Committee on Risk-Based Analysis for Flood Damage Reduction, Water Science and Technology Board, National Research Council (USA) (2000). Risk Analysis and Uncertainty in Flood Damage Reduction Studies.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 71 of 110

Application-specific guidance

Overview

In this chapter This chapter provides a brief overview of issues that an analyst should consider when assessing how to approach flood estimation in a specific application. In many cases, converting flood peak flows and hydrographs, derived using the FEH, into water levels may be all that is required. In others, FEH results will feed into detailed hydraulic modelling or other studies. This chapter also discusses flood estimation on unusual catchments. Analysts: in all cases, you will need to carefully consider the specific requirements of the study when developing a method statement.

Flood mapping and hydrodynamic models

Steady and Flood mapping indicates areas at risk from a flood event of a certain unsteady frequency. When floodplain storage is not significant, you can identify the hydraulic extent of flooding from a steady-state hydraulic model. models An unsteady hydraulic model is appropriate when: • floodplain storage is significant; • or you require a more accurate assessment of complex floodplains; • or you require the model for other applications, such as flood warning or forecasting. An unsteady model requires the derivation of inflow hydrographs, using either a rainfall-runoff method or a hybrid approach (see Hybrid methods). These are then routed through hydraulic models to identify the extent of flooding.

The issue: Unsteady hydraulic models or flow routing models can help in understanding striking a how flood peaks propagate down the catchment and their relative timing at balance confluences. This knowledge can inform the process of flood estimation. However, these models tend to rely on a rainfall-runoff approach to provide inflows. It is important to remember that it may not provide the best estimates, particularly when there are flood peak data at sites within the model reach. Also, the need to derive a hydrograph volume and shape introduces another element of uncertainty. There are many ways of deriving inflows for unsteady models. It is often necessary to strike a balance between two extremes.

Two extremes Excessive reliance on Imposing the design flows the hydraulic model on the model Example: Ignoring flood That is, adjusting model peak data at sites within inflows so that it reproduces the study reaches. the preferred FEH estimates at all points in the system.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 72 of 110

Guidelines Item Guideline or advice 1 If a hybrid method is used to generate design flows, there is no guarantee that hydrographs scaled to match peak flows from the statistical method at model inflows will result in statistical peak flows being reproduced further downstream within the model. At each point of interest in the model, it is necessary to decide how to strike the balance described above.

2 There can be a risk of double-counting floodplain attenuation in unsteady modelling. This could happen if a downstream donor site (at which flows are affected by attenuation) is used to estimate or adjust design flows for an inflow to a model, which then routes the flood hydrograph, allowing for the same attenuation processes again. You can avoid this by ensuring that the flow within the model gives a close match to design flows estimated at the site of the gauging station.

3 There is no catchment-wide design flood. The severity of any real flood event will be greater at some locations than elsewhere in the catchment (1 9.3). Therefore, if you have used a rainfall-runoff approach for flood mapping, you need to estimate the design flood separately at each site of interest, using a design storm appropriate for the catchment draining to that site (1 9.4). If you are applying a distributed rainfall-runoff approach (see Lumped or Distributed approach? ) you will also need to ensure that, for each site of interest, you apply a uniform storm duration and areal reduction factor across all subcatchments. The above advice can be confusing at first sight. Imagine you have a hydraulic model with four inflows from subcatchments. There are three key sites within your model reach where you need design flows. Site 1: use uniform storm duration D1 for all four subcatchments and an areal reduction factor calculated for area A1 Site 2: use uniform storm duration D2 for all four subcatchments and an areal reduction factor calculated for area A2 Site 3: use uniform storm duration D3 for all four subcatchments and an areal reduction factor calculated for area A3 See also Issues with catchment models, below.

Issues with Some studies have used catchment models to help to derive parameters for catchment the FEH rainfall-runoff model. Parameters for various inflow catchments are models adjusted by trial and error to give a match between observed and predicted flows or levels further down the model. This approach is not mentioned in the FEH, but it can be valuable. However, it can also be very misleading. In many cases, the studies refer to the catchment model as 'calibrated', implying that it can then be used with confidence for synthesising design events. This has led to over-reliance on the FEH rainfall-runoff method in some studies, which can give very poor design flows, even when the model parameters are well estimated (see Choosing between the FEH methods). Analysts: we strongly recommend that you stick to the language used by the FEH. Refer to 'estimating the model parameters' rather than 'calibration'.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 73 of 110

Lumped or distributed approach?, gives further guidance on distributed application of rainfall-runoff methods.

Catchment-wide studies

The issue Catchment-wide studies, whether broad scale, like CFMPs, or more detailed, such as flood risk or strategy studies, make extensive use of FEH methods. FEH methods are intended for application at particular (subject) sites. This is not surprising. They are calibrated against flood data at particular (gauged) sites.

Additional There are some additional factors to consider in larger scale studies. The factor: spatial most important is spatial consistency. The CEH report on automation of the consistency statistical method addressed this in detail, see Morris, D.G. (2003), listed in Related documents. It suggests some rules for spatial consistency: • sudden increases in flood estimates should only occur at confluences; • flood estimates should not decrease in the downstream direction unless there are clearly defined physical causes (such as floodplain attenuation); • the flood estimate immediately downstream of a confluence must be consistent with those immediately upstream; That is, it should not be greater than the sum of the upstream ones or smaller than the larger of them. It will normally be smaller than the sum of the upstream estimates because the two watercourses will not usually peak at the same time. • flood estimates at, and close to, gauging sites should be consistent with the gauged record unless there are valid reasons to the contrary. The FEH methods are not guaranteed to meet these rules. Additional inconsistency can be introduced by applying donor sites. See Figure 20.

Figure 20: an In the map below, if donor A is used to adjust QMED for all points upstream example of X and donor B used downstream of X, there could be a sudden jump in QMED at X. Weighted averaging of adjustment factors can help avoid this. For similar reasons, and to save time, it is usually advisable to apply the same pooling group at several sites on the same watercourse. A

X B

Doc No 197_08 Version 5 Last printed 12/02/15 Page 74 of 110

Post-event analysis

Guidelines Post-event analysis may be required to assess the severity of a specific flood. The guidelines and advice in the table are included to help users. Item Guideline or advice 1 Take care not to quote hasty assessments for rainfall and flood rarity. Ensure that the message is clear, simple and user friendly but still technically accurate. Refer to our Understanding and Communicating Flood Risk policy, listed in Related documents. Simple factual statements about the ranking of the event provides an immediate perspective alongside reassurance that a thorough review has been initiated.

2 If flow data for the event are available, you will need to interpret them with care, bearing in mind the quality of the rating curve for high flows.

3 Another aspect to consider is the bias inherent in estimating the flood frequency and return period, particularly using a single site analysis (3 Add. Note 11.2). Analysts: you should seek expert advice when there is a need to make an adjustment.

4 The ReFH method can assist in event analysis when there is no recorded flow data. The data required to simulate an observed event in ReFH is: • catchment-average event rainfall (for example, at a time step of 1 hour or 15 minutes) from tipping bucket raingauge(s) or radar data; • catchment-average daily rainfall from the start of the year preceding the flood; • potential evaporation, either a daily series or a mean daily value. 5 Input the data specified in Item 4 into the ReFH Design Flood Modelling software. It calculates the initial soil moisture from the daily rainfall and evaporation data and then runs the ReFH model to simulate the flood hydrograph from the event rainfall data. You can assess the return period of the peak flow, using the most appropriate FEH method. In most cases, it is sensible to use ReFH for deriving the flood frequency curve as well as for simulating the flood. Any errors in model parameters can be expected to cancel out to some extent as described in the FEH (4 5.4.2).

Reassessing Large and rare floods provide invaluable data with which to reassess flood estimates and estimates and recalibrate hydraulic models. This should lead to improved recalibrating estimation of: models • downstream water levels; • hydrograph shape; • travel time; • flood extents;

Doc No 197_08 Version 5 Last printed 12/02/15 Page 75 of 110

• flood return periods.

Modelling effects of land-use change on flooding

Description CFMPs make use of FEH methods to examine the effects of various policies, such as changes in farming and forestry on flooding. This has been done by sensitivity testing, altering the values of Tp and SPR in the FEH rainfall-runoff method. See CFMP Processes and Procedures Guidance.

Guidelines Item Guideline or advice 1 The best approach is to compare observed flood impacts on paired catchments with different land uses. Then use data transfer techniques to apply the observed impacts to the study catchment. However, it is rare to find observed data from suitable analogue catchments. So most CFMP studies use a more generalised approach. Example: Increasing SPR by a factor of 1.15 to represent agricultural intensification.

2 Now that ReFH has superseded the FEH rainfall-runoff method, there is an opportunity for updating the guidance on modelling the effects of land-use change. However, in the meantime, you could use the FEH rainfall-runoff method to indicate the relative effects on flood peaks, even if the best estimates of peak flows are derived using a different method.

Pumped and other low-lying catchments

The issue Catchments draining to pumping stations present an additional complexity. Research has shown that the response of pumped catchments is different to that from typical gravity catchments. Catchment boundaries tend to be manmade rather than natural, the water table is lowered by drainage, watercourses are often artificial and flows are affected by pump operations. For these reasons, predicting design flows from catchment descriptors is unlikely to be successful. Much of the guidance in this section is also applicable to low-lying catchments drained by gravity, for example through sluices that open at low tide.

Guidelines Item Guideline or advice 1 There are few flow gauging stations on lowland catchments, partly because of the historical necessity to use weirs for flow measurement. The FEH did not include pumped catchments in the derivation of the empirical equation for QMED. Given the factors listed in The issue, above, there should be no expectation that any FEH procedures are applicable to lowland

Doc No 197_08 Version 5 Last printed 12/02/15 Page 76 of 110

pumped catchments (3 13.7.4).

2 The FEH makes little reference to pumped catchments. Most studies continue to use a variation of the FSR rainfall-runoff method first published in 1987. See Samuels (1993) and IWEM (1987) Part 1 – both listed in Related documents. A recent Environment Agency science project, SC090006 (Flikweert and Worth, 2012) has updated the earlier guidance but the basic method is unchanged. In summary, the tailored version of the FSR rainfall-runoff method involves: • estimating time to peak preferably from local data or else (as a last resort) setting it to 24 hours, rather than using catchment descriptors; • using a trapezoidal unit hydrograph shape, which reaches the peak flow at 0.5 Tp and remains at that flow until 1.5 Tp; See Figure 21. • The peak flow is 1.59/Tp m3/s per 10 mm of rainfall per unit area, compared with 2.20/Tp using the conventional triangular unit hydrograph or 1.80/Tp using the ReFH unit hydrograph; Note: The magnitude of the unit hydrograph peak is not clear in the R&D report on ReFH. This value comes from the FEH Supplementary Report on ReFH. • estimating SPR by back-calculation from rainfall and pumping station data in preference to using soil mapping. Pumped catchments are particularly sensitive to volumes of runoff so it is important to estimate SPR as accurately as possible. • An alternative, not mentioned in SC090006, would be to use the ReFH model to calculate the volume of runoff. Since ReFH has superseded the FSR rainfall-runoff method for most applications, there is no particular reason not to use it for pumped catchments. If you decide to use ReFH, ensure that the unit hydrograph shape is modified as explained above. • calculating a critical rainfall duration by iteration as explained in FEH 4 9.2.2. • being careful with the design rainfall profile if the critical duration is longer than 48 hours. The recommended procedure is to distribute the design rainfall depth in time using the temporal profile of one or more notable long-duration rainfall events that were experienced locally or regionally. • accounting separately for runoff for upland or urban areas. You should refer to SC090096 for more detailed information and guidance when studying pumped catchments. SC090096 also recommends that you use pumping station records to investigate the performance of the drainage system, estimating a flow hydrograph for past events and comparing the rainfall duration and profile with those of the design storm event.

3 You need to apply careful judgement before using the above technique to generate inflows into lowland drains for subsequent hydraulic modelling of the drains and pumping station. The trapezoidal (flat-topped) form of the unit hydrograph partly reflects the influence of storage in the drain system and its role in attenuating the flood discharge. As a result, using the trapezoidal unit hydrograph combined with a hydraulic model that also explicitly includes this channel storage could cause

Doc No 197_08 Version 5 Last printed 12/02/15 Page 77 of 110

underestimation of flood levels through over-representation of the attenuation. Therefore you should not use the trapezoidal unit hydrograph as a model boundary condition at the point of entry to the main-drain system. However, it may not be appropriate to use the standard FEH or ReFH unit hydrograph either, since peak flows may be impeded for quite some distance upstream of pumping stations due to the shallow gradients. When deciding how to represent inflows to models of lowland drains you should take into account the length of the model reach and the degree of influence of the pumping station at the upstream model boundaries. SC090006 suggests a trial and error approach to this problem, adjusting model inflows (for example the time to peak or the shape of the unit hydrograph) until the hydrograph simulated by the model at the pumping station matches that estimated using the trapezoidal unit hydrograph.

4 An alternative method of flood estimation on pumped catchments is flood frequency analysis of annual maximum pumped volumes; see Part 1 of IWEM (1987). You should use this in preference when long records are available for the pumping station (which in practice seems to be rarely). Another alternative, not mentioned in published guidance, is to represent the entire pumped area using a 2D or linked 1D-2D hydraulic model with rainfall applied directly to the 2D model domain. This avoids the need for a unit hydrograph, but the resulting flow estimates will be heavily influenced by the assumptions made in the hydraulic model development. This has been applied on a small number of projects to date. Science project SC090006 recommends a tiered approach when selecting a method for flood estimation on pumped catchments. More advanced methods are needed when the analysis needs to provide more detailed answers and there is enough reliable data to justify the application of advanced methods.

5 Water balance calculations can be helpful on pumped catchments, either over a long period or for individual floods (whether observed or design events). SC090006 gives some guidance on this.

6 If estimating design flows for locations downstream of pumping stations, you should limit the outflow hydrographs from pumped catchments to the pump capacities. They can either be taken as constant flows or, if the volume is thought to be limited, routed through a notional reservoir that has an upper limit set on its outflow.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 78 of 110

Figure 21 The diagram below shows a trapezoidal unit hydrograph for pumped catchments, from Science Report SC090096.

Water level management plans and short return period estimates

Description Water level management plans establish a regime to protect and enhance the conservation status of areas where habitats rely upon water levels in harmony with their flora and fauna. These areas, which are likely to experience regular inundation, are most vulnerable to change in the pattern of frequent events rather than to infrequent extreme floods. The FEH methods are likely to be at their most robust for frequent events and can supply valuable estimates of inundation frequency using statistical methods.

Two types of For estimation of frequent flood events, it's important to understand the return period difference between: • the annual maximum return period, used in the FEH; • and the POT return period, sometimes known as the average recurrence interval. The two types of return period are related using Langbein’s formula, included in Appendix A of FEH Volume 1. Return periods of 1 year or less are meaningless on the annual maximum scale. So, if you require a design flood for a return period of 0.5 years, you must convert this POT-scale value to the corresponding annual maximum- scale return period, which is 1.16 years. You can calculate the design flow for this return period using an appropriate FEH method. An alternative way of estimating short return period floods, particularly where short flood peak records are available, is to analyse POT data using the method described in the Flood Studies Report (Volume 1, section 2.7.5). Note: Both methods described here have attracted criticism for ignoring dependence between successive flood peaks, which has been found to result in slight overestimation of design flows. See Archer, D. R. (1981), 'A catchment approach to flood estimation', listed in Related documents.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 79 of 110

Seasonal flood Although the FEH provides information on mean date of flooding and estimation variability, it does not specifically address the problem of seasonal flood estimation. Example: Assessing the flow of a given return period for a specified period of the year. This may be important in the winter, for maintaining high conservation levels, or in the summer, for impact on crop growth. However, the peaks over a threshold database provides information from which such information can be assessed. Archer's 'The seasonality of flooding and the assessment of seasonal flood risk' (1981) provides a practical method of such an assessment. See Related documents.

Comparing When frequency and extent of flood inundations are critical, you will require estimates and estimates of flood hydrographs using the ReFH method. simulations But, when good records are available, simulation with measured flows may be preferable and more readily understood by lay stakeholders.

Small catchments and greenfield runoff

The issues By their very nature, there are many more small catchments than large ones. So many flood estimates are carried out on small catchments. This is particularly true in development control, where additionally greenfield runoff estimates are needed for development sites, which generally do not form complete catchments. FEH methods were not originally intended for catchments smaller than 0.5 km2, unless flow data are available (1 3.1.2). Older methods have often been used instead, but recent research has shown that FEH methods should be preferred.

Reasons for More generally, flood estimates are particularly uncertain on small uncertainty on catchments (say, below 25 km2) because: small • there is a shortage of such catchments in the FEH dataset used to catchments derive the regression equations for ungauged sites and the HiFlows-UK dataset used to select pooling groups and donor catchments (see Figure 7); • digital catchment descriptors are more difficult to derive for small catchments, which is why the FEH CD-ROM imposes a minimum area of 0.5 km2; • flood peaks on small catchments are more susceptible to being influenced by local features, such as flow diversions, field drainage or storage of flood water behind culverts, bridges or embankments.

Guidelines Item Guideline or advice 1 For small catchments, checking catchment descriptors becomes more important. There is more scope for the DTM or the thematic datasets to be wrong for such small areas.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 80 of 110

It may be worth doing a soil survey or at least checking HOST values against soil maps.

2 Guidance on choice of method for flood estimation on small catchments has been developed in Science Project SC090031: Estimating flood peaks and hydrographs for small catchments. The report on Phase 1 (Faulkner et al., 2012) gives recommendations as follows: “It is recommended that flood estimates on small catchments should be derived from FEH methods in preference to other existing methods. The current versions of the FEH statistical approach or the ReFH rainfall-runoff model should be used except on highly permeable catchments (BFIHOST>0.65), where ReFH should be avoided, and possibly on urban catchments (URBEXT2000>0.15), where the results of the ReFH model can be less reliable. Checks should be carried out to ensure that the flood estimates are within expected ranges based on what is known about the history of flooding and the capacity of the channel (including evidence from previous flood marks). For catchments smaller than 0.5 km2 and small plots of land, runoff estimates should be derived from FEH methods applied to the nearest suitable catchment above 0.5 km2 for which descriptors can be derived from the FEH CD-ROM and scaled down by the ratio of catchment areas. The decision to translate FEH estimates from catchment scale to plot scale should be accompanied by an assessment of whether the study site is representative of the surrounding catchment area.” The next phase of project SC090031 is due to lead to development of new simple methods for flood estimation in small catchments and guidance on how to incorporate additional local information. This project is underway and is due to complete in 2016.

3 You are likely to come across studies that continue to use older methods. These are reviewed in the following sections: Rational method, starting below ADAS Report 345 Institute of Hydrology Report 124

Rational method

Description In the rational formula, peak flow is estimated with the equation: Q = 0.278 KIA, where: • K is the runoff coefficient; • I the rainfall intensity over the time of concentration; • A the catchment area (km2). The two key choices are the time of concentration and the runoff coefficient. The Bransby-Williams formula has often been used to estimate the time of concentration from the length, slope and area of the catchment. FEH recommendation The FEH doesn't recommend the rational method using this formula (4 3.4.2) as it gives peak flows typically twice as large as those from the FEH Doc No 197_08 Version 5 Last printed 12/02/15 Page 81 of 110

rainfall-runoff method for small lowland catchments. See Institute of Hydrology (1978).

Modified There is a modified rational method, from the National Water Council rational (1981), listed in Related documents. It is used for sewer design and includes method formulae to aid estimation of the two key parameters: • time of concentration is divided into time of entry and time of flow though the pipe system; • the runoff coefficient is related to the percentage runoff used in the Wallingford Hydrograph Method and moderated by a routing coefficient that allows for the typical shapes of time-area diagrams and rainfall profiles. The method is not suitable for greenfield runoff estimation as it is designed for sewered urban areas.

ADAS Report 345

Reference HMSO (1982) ADAS Reference Book 345. The design of field drainage pipe systems.

Description This was developed as a way of designing field drainage systems to protect crops from flood damage. It is only suitable for small rural catchments with no formal drainage system. Flow is estimated from land use, soil type and rainfall, using a graphical solution to an equation ultimately based on the rational method. Note: ADAS Book 345 gives no information about how the method was derived. Find that in: Bailey, A.D. and others (1980), listed in Related documents. The relationship between flow and return period is based on rainfall intensities derived by Bilham in 1962, which are clearly rather dated.

Alternative An alternative approach is to derive the 2-year flow and then use a growth approach curve from FSR or FEH to obtain results for other return periods (from personal communications from Steve Rose and Rob Arrowsmith, both formerly of ADAS). For the 2-year flow, see the line labelled as 'Grass' on the graph in ADAS 345 Appendix 6. The label should actually say 'Intensive grass and cereals', for consistency with MAFF Report 5 (see Related documents) which states that the 2-year return period corresponds to intensive grass and cereals. It can be seen that the Grass line in ADAS 345 corresponds to the 2-year return period line in MAFF Report 5 Figure 4.

Estimate of The soil type factor in ADAS 345 is estimated from characteristics such as soil type factor permeability and soil texture. MAFF Report 5 presents a way of calculating the soil type factor from the FSR WRAP maps, which is more straightforward for most analysts.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 82 of 110

Results of Science Project SC090031 found that ADAS 345 tends to underestimate tests QMED and has a mean error that is much higher than any other method tested. The research report also pointed out that ADAS 345 is based on a very small dataset of limited length. Therefore, we advise users to avoid ADAS 345 for flood or greenfield runoff estimation on small catchments.

Institute of Hydrology Report 124

Reference Marshall, D.C.W. and Bayliss, A.C. (1994) Flood estimation for small catchments. IH Report 124. Institute of Hydrology, Wallingford. Download from the CEH website.

Description The study derived equations for Tp(0) and QBAR, using data from 71 small rural catchments in lowland England. Many of these catchments were upland, relatively wet and impermeable. Only two of the catchments were smaller than the smallest in the data set used to derive the latest FEH equation for QMED. The flood peak records used for the research do not include the most recent 20 years of data.

The guidelines and advice in the table below are included to help users. Select references that are linked to see details in Related documents. Guidelines Item Guideline or advice 1 The QBAR equation in IH 124 has been recommended for greenfield runoff estimation in a wide range of guidance documents including the Interim Code of Practice for SUDS and the SUDS Manual. The recommendation was based on a guide by HR Wallingford first published in 2004. The most recent version is Kellagher (2012). The guide does not claim that IH 124 gives more accurate results than other methods; rather, the recommendation was aimed largely at meeting the pragmatic needs of the industry. Although the recommendation to use IH 124 is maintained in Kellagher (2012) (albeit alongside FEH methods), it has now been superseded by Science Project SC090031: see item 4 below.

2 IH 124, and often ADAS 345, rely on coarse-resolution soil maps with only five classes. These are less likely to be representative of local soil conditions than the HOST mapping, which is available at a 1 km grid size and allows 29 different soil classes.

3 A disadvantage of the statistical method in IH Report 124, is that it relies on the FSR regional flood growth curves, which is a step backwards from the flexible pooling system introduced in the FEH. However, it is possible to combine an estimate of QBAR from IH 124 with a pooled growth curve from FEH, as long as the calculation accounts for the different return periods of QBAR and QMED.

4 Science Project SC090031 found that IH 124 tends to underestimate QMED and has a mean error that is higher than Doc No 197_08 Version 5 Last printed 12/02/15 Page 83 of 110

the FEH Statistical method. Therefore, we advise users to avoid IH 124 for flood or greenfield runoff estimation on small catchments.

Greenfield A new tool has been developed for site level greenfield runoff estimation. It runoff is published by HR Wallingford. The online tool accepts the use of both estimation tool IH124 and FEH for Greenfield estimation and work along the concepts of peak flow/ volume with long term storage. The main author for the rainfall runoff for development (Richard Kellagher) was involved with the HR Wallingford online tool.

Development control and urban catchments

The issues Development control is one of the most difficult application areas (1 12.6). Urbanisation has a widespread and significant effect on flood frequency. Some development control applications will be beyond the scope of the FEH methods.

Guidelines Item Guideline or advice 1 The FEH has much to say on development control and urban catchments (1 8, 3 9, 3 18, 4 9.3, 5 6). Urbanisation can have a major influence on the flood frequency curve. The type of influence is affected not just by the amount of urban area in the catchment but also by factors such as the pre- urban runoff rate (i.e. the soil type), the type of development, the way in which it is drained (including the extent of any SUDS measures), the location and the spatial concentration of the urbanisation. Because of this wide variety of factors, you cannot expect to get a very reliable estimate of the flood frequency curve using generalised methods, i.e. those derived using data from other catchments. There is no substitute for obtaining local data. With a little advance planning, you can sometimes achieve this without incurring large delays or expense. Even two years worth of flood peak data recorded, for example, using a temporary ultrasonic flow meter, can be expected to give a more certain estimate of QMED than the FEH equation based on catchment descriptors. If timescale, budget or practical considerations mean that it is not possible to obtain local data, you will have to accept a large amount of uncertainty on design flows for small urban catchments.

2 On heavily urbanised catchments you should obtain information on the urban drainage network: locations of combined sewer overflows and storm sewer outlets, and the extent of the sewer network draining to these locations. You may find that the boundary of the urban drainage catchment is significantly different from the topographic catchment boundary. Sewers may take water out of the topographic catchment or bring water into the catchment from neighbouring areas. A complicating factor is

Doc No 197_08 Version 5 Last printed 12/02/15 Page 84 of 110

that urban drainage systems have a limited capacity. Modern systems are designed for a return period of 30 years, but older systems may have a capacity of 5-20 year return period. In more extreme storms, the excess water will flow overland, following the contours of the ground. So the catchment boundary can vary according to the intensity of the rainfall. See Beskeen and others (2011). If this is the case then you should use a rainfall-runoff method in preference to the Statistical method, separating the catchment into different zones (for example, rural areas, urban areas drained by combined or surface water sewers, areas where sewers drain out of the topographic catchment, areas where sewers drain into the catchment from other topographic catchments, and so on). See below for guidance on choice of rainfall-runoff method and an example of how to divide up the catchment. If any flow or water level data is available you should examine it, along with rainfall data to check for evidence of a multi-peaked response to rainfall which might be expected if urban and rural areas both contribute significant amounts of runoff. The approach to flood estimation needs particularly careful thought when there is a mixture of urban and rural areas in the catchment. This needs to be considered when developing the conceptual model, see Preparing method statements.

3 Although the FEH advances the merits of SUDS (1 12.6), it cautions that the effect of runoff control techniques are usually only examined at the local scale. A more holistic approach is required to ensure that they do not have adverse effects elsewhere within the catchment (1 Interlude).

4 When a floodplain is threatened by development, flow hydrographs will assist examination of the options to mitigate the loss of floodplain storage. In other cases, a simple steady-state model may be sufficient to demonstrate that the proposed development does not impede flood flows. In all cases, it is important to check for consequential and detrimental effects elsewhere in the catchment.

5 The degree of urbanisation of a catchment is measured using URBEXT2000 (used in the Statistical method) or URBEXT1990 (used in the ReFH and Rainfall-Runoff methods). For information on the differences between URBEXT1990 and URBEXT2000, refer to the R&D Report by CEH Wallingford, see Bayliss, A.C. and others (2007), listed in Related documents.

Up to moderately urbanised catchments

Recommended For catchments up to moderately urbanised (URBEXT1990<0.125 or methods URBEXT2000<0.150): You can consider any of the usual rural catchment methods, applying an urban adjustment if you choose the Statistical method. The ReFH method Doc No 197_08 Version 5 Last printed 12/02/15 Page 85 of 110

treats catchments up to moderately urbanised as essentially rural.

Heavily or very heavily urbanised catchments

Option 1 For heavily or very heavily urbanised catchments Statistical (0.125≤URBEXT1990<0.500 or 0.150≤URBEXT2000<0.600): Method You should put careful thought into the choice of method for such catchments, first developing an understanding of the urban drainage network (see above). If there is little difference between the boundaries of the topographic and sewer catchments, or if you are interested in extreme events for which sewer flows can be neglected, you can use the Statistical method, with an urban adjustment applied. The revised urban adjustment (Kjeldsen, 2010) was developed using data from 206 urban catchments. The Statistical method benefits from an up-to-date flood peak dataset, unlike earlier methods which are all dated to some extent. The method also has the advantage of avoiding the need to make assumptions about factors such as the nature of the design flood, the rainfall duration, the time of concentration etc. You should not use the Statistical method to predict the future effect of urbanisation (3 9.1).

Option 2 An alternative method, which can also be applied when the topographic and Revised sewer catchments differ, is the revised version of ReFH published by version of Kjeldsen (2009). You should not use the original version of ReFH because ReFH method its summer design event was calibrated on only seven urban catchments. The revised ReFH method for urban catchments alters the percentage runoff to account for the presence of paved areas. It uses the full ReFH model in rural areas and within the green portion of urban areas (gardens, parks etc.). In the portion of urban areas covered in hard surfaces, the ReFH losses model is not used; instead the percentage runoff is set to a fixed value (70% is suggested in the paper, which is the figure used in the FSR rainfall-runoff method; see 4 2.3.1). This avoids the need to depend on aspects of ReFH that are poorly calibrated for urban catchments: the regression equation for CMAX, which does not represent the increase in runoff volume with urban extent; the way in which the initial soil moisture, Cini, is calculated for design events based on an equation calibrated from only 7 catchments and which gives a physically unrealistic increase in Cmax as PROPWET decreases; the α factor used to scale Cini to ensure that the resulting flood frequency curve is consistent with the results of the FEH statistical method - again, this factor was derived from analysis of only 7 catchments. Kjeldsen (2009) describes calibration of the revised ReFH method by comparison of modelled and observed hydrographs on two catchments. Although the applications described in the paper are simulation of observed flood events, not estimation of design events, the revised method has recently been applied to estimate design flows on several complex urban catchments and appears to give sensible answers. See below for an example of its application.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 86 of 110

Alternative In the past, and still in some studies, other rainfall-runoff approaches rainfall-runoff (including the Rational, Modified Rational and FSR/FEH rainfall-runoff methods methods) have been widely applied on small urban catchments. Rainfall-runoff approaches are conceptually appealing because of the clearer link between rainfall and runoff in urban areas, where soils, and soil moisture, are less influential. However, all these methods have their disadvantages, including: • The rational method assumes that the peak flow is proportional to the rainfall intensity. It is necessary to estimate the time of concentration, over which time the rainfall intensity is calculated. It is also necessary to guess a value for the runoff coefficient. Refer to the earlier section on small catchments. • The modified rational method is used for sewer design within the Wallingford Procedure. It includes formulae to aid estimation of the two key parameters. Time of concentration is divided into time of entry and time of flow though the pipe system. The formula for time of entry, based on length and slope, is appropriate for small events only (return periods of weeks to months). For a return period of 5 years, the Wallingford Procedure recommends using 3-6 minutes for the time of entry. There is no guidance on what to use for longer return periods. This method may be a good choice for estimation of low return period floods on small catchments (up to 20 hectares) that are completely developed and drained by sewers. However, it is difficult to justify using it on larger catchments with a stream network. • The FEH rainfall-runoff method tends to overestimate flows in many areas (sometimes by a factor of five or more) and has been superseded by the ReFH method. You are recommended to consider the revised ReFH method as a first choice, but there is no ideal method for heavily urbanised catchments and in some situations it is possible that the alternatives listed above are more appropriate than ReFH. Another alternative is FRQSIM, a rainfall- runoff model developed for Greater London; see the information below.

Extremely heavily urbanised catchments

Recommended For extremely heavily urbanised catchments (URBEXT1990>0.5 or methods URBEXT2000>0.6): You should not routinely apply the FEH flood frequency methods to these catchments (5 6.5.5). However, alternative methods have drawbacks too, as discussed above. For deriving flows from urban sewered areas it may be more appropriate to use sewer design methods, such as the modified rational (for peak flows) or the Wallingford hydrograph method, a version of the FSR rainfall-runoff method which is used in sewer network modelling software. You can find an example of its application for estimation of fluvial flood flows in Beskeen and others (2011). An alternative is FRQSIM, a rainfall-runoff model developed for Greater London; see the information below.

Flood FReQuency SIMulation (FRQSIM)

Doc No 197_08 Version 5 Last printed 12/02/15 Page 87 of 110

Description FRQSIM stands for flood FReQuency SIMulation. It was initially developed in the 1970s by the Greater London Council to provide design flows for flood alleviation schemes in the highly urbanised catchments of the Thames tributaries in London. However, use of the model is not restricted to these areas and it has been applied to other urban areas such as Manchester. You can find information on the model in the user guide, FRQSIM Hydrological Model, listed in Related documents. Recent versions of ISIS include a FRQSIM unit.

Method The catchment to be modelled is separated into 'node areas', which are equivalent to sub-catchments used in FEH methods. These are based not only on topographic information, but also on drainage networks. Each node area is then divided further into 'sub-areas', defined as either paved or open. The model differs from the FEH rainfall method because it uses a time-area method to produce synthetic unit hydrographs (SUH). A separate SUH is produced for paved and open areas. A third SUH is produced to represent gardens and verges within urban areas. Time of travel estimates are needed for each sub-area. FRQSIM does not allow for situations where the topographic catchment is different from the sewer catchment: see Beskeen and others (2011).

Other features FRQSIM includes a loss model, which determines the effective rainfall for of FRQSIM open areas. One other useful feature is the recognition of the finite capacity in the surface water drainage network. FRQSIM assumes that capacity of the network is the 5-year storm and that any rainfall above this will be stored in the model and released over subsequent time steps until all of the runoff has gone through the network.

Design rainfall Other notable differences from the FEH rainfall runoff method include the and storm shape of the design storm profiles used. profiles Ten storm profiles are available, based on 250 flood producing storms observed across the London area. The design procedure used in FRQSIM has been criticised in the past for being rather obscure. For example, it is not clear why the 250 storms should represent 100 years of flood-producing rainfall, which is a fundamental assumption of the procedure. Beran (1987) (see Related documents) recommended that the storms should be regarded as representing 250 years of data. FRQSIM has been developed over the years, with a recent change being to obtain storm depths from the FEH rainfall frequency statistics. However, in any event-based method for estimating design flows, it is necessary to ensure that the composition of the design event (rainfall depth, duration, profile and catchment wetness) gives rise to a peak flow of the required return period. It is not clear that FRQSIM achieves this (Onof et al., 1996).

Differences FRQSIM has been seen to give design flows much higher than those from between FEH methods (for example, on the Cobbins and Salmons Brooks and the

Doc No 197_08 Version 5 Last printed 12/02/15 Page 88 of 110

design flows Lower Lee). The large differences between design flows from FRQSIM and FEH statistical methods, particularly at locations where the latter are based on local flood peak data, should act as a prompt to review some of the assumptions made in the FRQSIM design procedure. However, you should recognise that FEH methods do not perform at their best in heavily urbanised catchments. Both FEH and FRQSIM have pros and cons.

Example application of revised ReFH method

Example of The revised version of ReFH described above was used to estimate design revised ReFH flows for a flood mapping study on a heavily urbanised watercourse in method Cheshire. The table below lists the steps involved in applying the method. The steps are intricate and much more time-consuming than conventional application of FEH methods to a lumped catchment. However, in the absence of any flood peak data for the watercourse, flood estimation on heavily urbanised catchments often needs this type of detailed analysis. Project managers and team leaders: you should ensure that flood studies on heavily urbanised catchments allow enough time and budget for detailed hydrological calculations, and that staff working on such studies are given extra guidance and supervision until they are experienced in urban hydrology.

Guidelines Step Action 1 Division of catchment The catchment was divided into three categories of sub- catchment, based on sewer maps and LIDAR data. The three categories, shown on the map below, were: • undeveloped sub-catchments; • urban sub-catchments where the topography drains towards the watercourse but the sewers drain out of the catchment; • urban sub-catchments where both the topography and the sewers drain towards the watercourse. Some catchments will have a fourth category, areas where sewers drain into the watercourse from outside the topographic catchment.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 89 of 110

The sewer catchments consist of both combined and surface water sewers. The analyst assumed that the sewers would be at capacity for a 10-year return period rainfall event. This means that any rainfall over that of the 10-year event will end up contributing to flow on the watercourse even in areas where the sewers drain out of the catchment. A more accurate sewer capacity could be obtained from hydraulic modelling of the urban drainage network. The descriptors AREA, DPLBAR and URBEXT1990 were altered to suit each sub-catchment.

2 Calculation of flows for undeveloped areas These flows were estimated using the standard ReFH model, applied separately to each rural sub-catchment and also to 60% of the area of each urban sub-catchment. The analyst assumed that 60% of urban areas are unpaved (i.e. water can infiltrate as it would on the rural catchments). Kjeldsen (2009) suggests 70%, but 60% was thought to be more appropriate on the study catchment as parts are heavily urbanised, with supermarkets, car parks and industrial buildings. The method assumes that the unpaved portion of urban areas behaves as the rural areas, unaffected by sewer systems.URBEXT1990 values were altered to 0, resulting in the longer time to peak that would be expected in rural areas.

3 Calculation of flows for the paved portion of urban areas where the topography drains towards the watercourse but the sewers drain away For return periods up to 10 years the analyst assumed that all storm water leaves the catchment via the urban drainage system. For longer return periods, the 10-year rainfall intensity was subtracted from the design rainfall hyetograph, to give the excess water that was assumed unable to enter the sewer system. In practice the analyst applied this approximately by altering the rainfall return period entered to the ReFH spreadsheet. In order to represent the generation of runoff over an urban area, the percentage runoff (PR) was set to 70%, as suggested by Kjeldsen (2009).PR from the ReFH model can be calculated by

Doc No 197_08 Version 5 Last printed 12/02/15 Page 90 of 110

dividing direct runoff by design rainfall. The ReFH spreadsheet was altered to produce a hydrograph where PR was approximately 70%, reducing the Cmax parameter by trial and error. URBEXT1990 was set to 0.5 (a higher value may have been more appropriate) to represent faster routing of water through the urban catchment, resulting in a shorter time to peak than seen on the rural catchments.

4 Calculation of flows for the paved portion of urban areas where both the topography and the sewers drain towards the watercourse The method was similar to that at step 3, but with no need to reduce the rainfall intensity to allow for the sewer capacity, as all water falling on the catchment will reach the watercourse, regardless of whether this is via sewer or topographic routes. Again the URBEXT1990 value was set to 0.5, and the model was adjusted to produce a PR value of 70%.The method assumes similar routing of flows whether within sewers or overland.

5 Hydrograph addition The previous steps produced two or three hydrographs for each sub-catchment, representing undeveloped and paved areas. The hydrographs were added together to produce inflows for use in a hydraulic model. Many of the combined hydrographs had double peaks, because of the differing flow routing times between the undeveloped and paved portions of the sub-catchment (see the graph below). Within each run of the hydraulic model, a common storm duration was used for all ReFH modelling. A range of storm durations was investigated to identify the critical duration.

Addition of hydrographs for undeveloped and paved areas in a sub-catchment

Doc No 197_08 Version 5 Last printed 12/02/15 Page 91 of 110

Permeable catchments

Why to avoid Design event methods such as the FSR/FEH rainfall-runoff method or ReFH design event are generally not recommended for highly permeable catchments. Floods in methods catchments underlain by fissured aquifers, such as the Chalk, are influenced by hydrogeological factors that are not adequately represented in techniques developed for quick response catchments where surface features are the main control. See Bradford and Faulkner (1997). Webster (1999) found that the relationship between the return periods of storms and floods became increasingly scattered for more permeable catchments, and concluded that permeable catchments are not really suitable for design flood analysis using an event-based method. These comments also apply to the ReFH method, although its improved baseflow model may offer some advantages over the FEH rainfall-runoff method. The ReFH report states that caution is needed when applying the method in baseflow-dominated catchments, as the regression equations may underestimate the model parameter that represents maximum soil moisture storage. These guidelines recommend that ReFH is not used for estimating peak flows on permeable catchments. The FEH Statistical method is normally the most appropriate choice on highly permeable catchments. However, it is important to be aware of two issues:

Issue 1: Large There is anecdotal evidence that the current regression equation for QMED uncertainty in (from Science Report SC050050) can under or over-estimate by a long way QMED on some permeable catchments. Examples include: Extreme over-estimation on the South Winterbourne at Winterbourne Steepleton, a small chalk catchment in Dorset, where QMED estimated from catchment descriptors is 5.5 times larger than that estimated from annual maxima. Over-estimation on the Pang at Pangbourne, another chalk catchment, where QMED from catchment descriptors is 4 times larger than from annual maxima. Extreme under-estimation on the Rhee at Ashwell, a part-urbanised chalk catchment, where QMED from catchment descriptors is 5 times smaller than from annual maxima. Underestimation on some catchments may be associated with the fact that the BFIHOST term is squared in the regression equation – although the non-linear term was found necessary in order to avoid overestimation of QMED on gauged catchments with high BFIHOST (see Figure 4.3 in Science Report SC050050). It is possible that the confidence limits for QMED estimation are much wider on permeable catchments than the UK-average limits derived from the factorial standard error of the regression equation. So you should be aware that flood estimates on ungauged permeable catchments are likely to be extremely uncertain. If you need a more confident result, consider installing a temporary flow logger. Even a few month’s worth of data may enable you to estimate design flows with more confidence than relying on catchment descriptors for a highly permeable catchment, for example if it enables calibration of a rainfall-runoff model for use in continuous simulation (see later).

Doc No 197_08 Version 5 Last printed 12/02/15 Page 92 of 110

Issue 2: In the original FEH method, pooling groups for permeable catchments were Pooling generally composed of gauged permeable catchments. This is no longer groups the case using the method presented in Science Report SC050050 which does not use BIFHOST to select pooling groups. BFIHOST was excluded because it was found to have very little influence on the sample values of L- moments calculated from a large number of individual annual maximum series. For L-CV, the report states that a minimum of ten other variables were selected in a multiple regression before BFIHOST was included. So the data are saying that BFIHOST has no effect on the L-moments, and hence on the growth curves. When a similar method was used in the original FEH research, it found that BFIHOST was very influential. The change may be due to the addition of FARL and FPEXT as variables for selecting pooling group, because all three of these catchment descriptors represent catchment storage effects to some extent. Table 6.3 in SC050050 indicates the relative performance of different pooling group methods. The FEH method (i.e. v1 or v2 of WINFAP-FEH) performs slightly worse than a pooling group selected purely by geographical proximity. Out of the seven methods listed in the table, the FEH method gives the second highest uncertainty. The implication is that it would be unwise to revert to v2 of WINFAP-FEH for constructing pooling groups even if you were concerned about the exclusion of BFIHOST in the v3 method. Earlier research, including the FEH and Flood Studies Supplementary Report 4 (1977) has consistently reported differences in flood growth curves on permeable and nearby impermeable catchments: generally less year-to- year variation on the permeable catchments and hence flatter growth curves. Looking at flood history on permeable catchments, we should perhaps expect high skewness of annual maximum flows owing to occasional extreme floods – as mentioned in FSSR 4. But there don’t seem to be many of these evident in the gauged period of record and so there is little effect on the sample L-moments.

So do In light of the above considerations and bearing in mind the physical permeable processes that lead to flooding, many hydrologists would consider it quite catchments reasonable to expect BFIHOST to influence the growth curve, despite the need special findings of Science Report SC050050 pooling It is worth bearing in mind that the results in SC050050 are UK averages. groups? It’s possible that an investigation focused purely on permeable catchments might come up with different findings. Analysts: if you want to allow for permeability in the composition of the pooling group, do this by manual editing of the group rather than reverting to v2 of WINFAP-FEH, for the reason given above and also because you would lose out on the other benefits of the v3 pooling procedure such as the revised weighting method and the option to carry out an enhanced single- site analysis.

Urban Permeable catchments that are also urbanised pose particular problems in permeable flood estimation. Refer to the section on urban adjustment in the Statistical catchments method.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 93 of 110

Guidelines Item Guideline or advice 1 An understanding of the catchment geology and hydrogeology can be valuable when estimating floods in permeable catchments. In particular, it is important to establish the possible processes that might lead to flooding. These could include intense rainfall on scarp slopes, prolonged winter rainfall, snowmelt, rain falling on frozen ground or runoff from impermeable or urban areas of the catchment. If there is a correlation between river flows and groundwater levels, it may be possible to use long-term groundwater level data in the flood frequency analysis. The groundwater catchment boundary may be very different from the topographic boundary. You can investigate the location of groundwater divides by looking at geological or hydrogeological maps. Consider consulting colleagues in hydrogeology teams as well.

2 Significant floods tend to be infrequent on permeable catchments, but they can be unexpectedly severe when they do occur. This means that you need to interpret relatively short gauged records with caution, for example when fitting a single- site growth curve. Another consequence is that longer-term flood history is particularly valuable. The HiFlows-UK website does not provide data in the form of .pt files (for use in WINFAP-FEH) where baseflow dominance is such that POT extraction would be unrealistic without detailed analysis. This affects many, but not all, permeable catchments. POT data for permeable catchments is, however, given in .csv files. These may require care in their use because data from different sources may have different flow thresholds.

3 For many permeable catchments, there are some years in which no floods occur and the annual maximum flow is due to baseflow alone. Including non-flood annual maxima in a frequency analysis can result in a fitted growth curve that is bounded above (that is, the growth factors reach an upper limit). When you are carrying out single-site analysis on a permeable catchment, or pooled analysis for a group consisting largely of permeable catchments, use the technique described in the FEH (3 19) for removing flood-free years by adjusting the L-moments. Permeable catchments are defined in the FEH Statistical method using an arbitrary threshold of SPRHOST<20%, which corresponds roughly to BFIHOST>0.75. The calculations for adjusting L-moments are not carried out by WINFAP-FEH. It is necessary to solve the equation for the shape parameter (3 Equation 19.4) numerically, which can be done using the Solver function in Excel or a root-finding subroutine in Fortran (for example). The adjustment generally has a fairly small effect on growth curves.

4 Where full hydrographs are needed, you can implement a hybrid approach.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 94 of 110

However, the influence of baseflow on flood flows in permeable catchments means that estimating flows from catchment descriptors alone could provide misleading flow values and hydrograph shapes. In these situations, it is important to refine model parameters, where possible, from local data.

5 Prolonged floods can occur on permeable catchments due to high groundwater levels. The volume and duration of floods are, therefore, important factors to consider. Bradford, R.B. and Goodsell's study in 2000 (see Related documents) of flood volumes on permeable catchments recommended carrying out volume frequency analysis by fitting a Generalised Logistic distribution to a series of annual maximum flood volumes over a given duration. This involves extracting discharge volumes over a period of d consecutive days from daily mean flow data, where d is the duration of interest. The maximum volume is determined for each water year. The annual maximum series is standardised by its median and the distribution is fitted by L-moments as for flood peaks.

6 Flood estimation by continuous simulation is a particularly attractive prospect on permeable catchments, particularly where there is a shortage of flood peak data. The simulation is likely to be more convincing if the rainfall-runoff model can be calibrated jointly against river flow and groundwater level data, where it is available (Reed, 2002). An example of such a model, which has been applied on the Chalk River Lavant catchment is described by Moore and Bell (2002). A related approach to consider is to combine aspects of FEH methods and continuous simulation. An example of this might be using a short record of flow data to estimate QMED and deriving the growth curve from continuous simulation.

Summary Our summary of recommendations for permeable catchments includes: • develop an understanding of the hydrological and hydrogeological processes that might result in a flood; • be aware that significant floods can happen in permeable catchments but they tend to be infrequent; • carry out a review of historical floods; • use the statistical method in preference to a rainfall-runoff technique; In particular, you should not use ReFH when BFIHOST>0.65. • acquire local flow data (even a very short record) if possible rather than relying on catchment descriptors for estimation of design flows. Refer to the example under Selecting and examining flood peak data in which even a month of flow data on a limestone catchment was enough to cast serious doubt on the catchment-descriptor estimate of QMED. • adjust single-site growth curves to account for non-flood years in the dataset when SPRHOST<20%.

Catchments containing reservoirs

In this section This section is about reservoir routing as part of a wider study when the reservoir and its safety is not the subject of the study. See also Flood Doc No 197_08 Version 5 Last printed 12/02/15 Page 95 of 110

estimation for reservoir safety.

Description The FEH statistical method accounts for lakes and reservoirs in a general way, using the catchment descriptor FARL: • to reduce QMED when there are water bodies present in the catchment; • and using FARL (in v3 of WINFAP-FEH) to guide the selection of the pooling group. You should not rely on the QMED equation when FARL is below around 0.9 due to impounding reservoirs, unless they are kept permanently full and thus act like natural lakes (3 8.3.2, 13.7.4). If flood peak data are available downstream of the reservoir and close to the site of interest, you can use them to estimate QMED directly and thus implicitly account for the effects of the reservoir. In the absence of suitable flood peak data, you should use the ReFH method on catchments with a significant reservoir influence, along with a flood routing calculation which determines the outflow from the reservoir (4 8). Unless the subject site is directly downstream from a single reservoir, it will be necessary to incorporate this in a flow routing model to allow for inflows from the rest of the catchment.

Guidelines Item Guideline or advice 1 The ReFH flood modelling software, along with many other modelling packages, can carry out reservoir routing calculations. There are several points to beware of: • because reservoirs delay flood hydrographs, the critical storm duration needs to be extended (4 8.2, 1 Interlude) and some iteration is necessary to find the critical duration; • if there are multiple reservoirs in the catchment, the calculation becomes quite complex; It is necessary to estimate the direct inflow to each reservoir as well as the routing of outflows from upper reservoirs (4 8.3.2). • when the design storm duration is much longer than the critical duration for the catchment flowing into a reservoir, beware that the ReFH method can overestimate the flow (see The ReFH method; • if the site of interest is some distance downstream from a reservoir, it is important to check whether the critical design event might arise from a shorter-duration storm on the intervening area downstream of the dam. (See Lumped or distributed approach?.) 2 The design of operating rules for both on-line and off-line flood storage reservoirs requires the derivation of flood hydrographs and knowledge of the discharge characteristics of the inflow and outflow structures. Flood hydrographs must be routed through the reservoir to determine its performance. In the absence of gauged data to simulate actual events, or where a T-year event is required, you should use the ReFH or hybrid methods.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 96 of 110

Flood estimation for reservoir safety

In this section Estimating floods to design or assess reservoir spillways is a specialised subject. This section gives a brief overview of the methods available and the latest current guidance (at July 2009). Analysts: you should ensure that you are familiar with the methods and up- to-date with the guidance. Find the latest research and guidance on the Defra website.

Description Reservoir spillway capacities are usually assessed as part of a detailed inspection that is carried out by Panel Engineers under Section 12 of the Reservoirs Act 1975 every 10 years. The final water level of the reservoir during a design storm is assessed to ensure there is adequate freeboard in the reservoir. The final water level includes a wave assessment, which is not covered in these flood estimation guidelines. Design floods at reservoirs are also needed for the preparation of reservoir flood plans.

Guidelines The guidelines and advice in the table below are included to help users. Select references that are linked to see more details. Item Guideline or advice 1 Flood estimates for reservoir safety require great care and should be carefully checked. You should check catchment descriptors manually. It is sometimes necessary to calculate the flow contributions from catchwater channels. We recommend site inspections to establish whether drainage paths are likely to change in an extreme event. Refer to Floods and reservoir safety.

2 The Reservoirs Act 1975 provides a safety regime for raised reservoirs with a capacity greater than 25,000m3. Dams are divided into four categories, A to D, based on the consequences of a breach. This is described in Floods and reservoir safety. The design standard for the spillway depends on the category – see Table 4. A Panel Engineer classifies the reservoir and the record of the classification is maintained in the Prescribed Form of Record. The design flood of the required return period is derived for the catchment flowing into the reservoir and then routed through the reservoir, allowing for the reservoir lag effect in the storm duration.

3 Estimating long return period floods (150-1000 years), covers long return periods. But there are specific methods prescribed for reservoir safety calculations. In particular, the ReFH method has not yet been accepted for reservoir safety work. Instead, you should use the FEH rainfall- runoff method, for estimating the 150-year flood (Category D dams).

Doc No 197_08 Version 5 Last printed 12/02/15 Page 97 of 110

4 For longer return periods, particularly 10,000 years, users should be aware that the FEH rainfall frequency statistics were not derived with such extreme events in mind. When extrapolated to a return period of 10,000 years, they give some contradictions with estimates of the probable maximum precipitation. See MacDonald, D.E. and Scott, C.W. (2000).

5 After reviews by Babtie Group and Sir David Cox, Defra commissioned a project starting in 2005 to investigate alternative methods of extreme rainfall estimation for return periods up to 10,000 years and, if appropriate, to amend the FEH methodology for extreme rainfall. Research reports from the Defra R&D Project FD2613 ‘Long Return Period Rainfall’ are now available from Defra’s website. We welcome all scientific contributions and new research in flood hydrology. We trust the publication of this research will stimulate good discussions in the hydrological community and between researchers and practitioners alike. Since the reports were completed, CEH has continued to refine and generalise the new Depth Duration Frequency (DDF) model and further results and comparisons with existing models will be published later this year. Whilst this scientific debate unfolds, the Environment Agency and CEH recommend that the existing Flood Estimation Handbook rainfall DDF model should continue to be used for operational flood hydrology and estimation. So Defra’s interim guidance for reservoir safety calculations (March 2004), is still current. It states that: • 1000-year rainfalls are estimated from both FSR and FEH methods and the more extreme result should be used; • 10,000-year rainfalls are derived from FSR rather than FEH methods; • In both cases, the FEH rainfall-runoff method should be used to derive the flood hydrograph from the appropriate design storm; model parameters should be estimated from local data or catchment descriptors if data are not available. Refer to Floods and Reservoir Safety – Revised Guidance for Panel Engineers.

6 The FSR rainfall frequency method involves using tables (FSR Volume 2) and maps (FSR Volume 5).

7 The estimation of the PMF is set out in FEH 4 4. It is a version of the rainfall-runoff method, with the following changes: • the design rainfall event is the probable maximum precipitation, PMP; this is estimated from a rather involved procedure (4 4.3) based on information from maps and tables; • you should apply both summer and winter PMPs, to see which gives the larger flood; • the time to peak of the unit hydrograph is reduced by one third to account for the more rapid response of an exceptional flood; • when applying the winter PMP, the standard percentage runoff is set to a minimum of 53% to account for frozen ground; • when applying the winter PMP, you should consider Doc No 197_08 Version 5 Last printed 12/02/15 Page 98 of 110

snowmelt; You can add it to the event precipitation and the antecedent rainfall. Take the melt rate and snow depth from maps. • the catchment wetness index is increased to allow for greater antecedent rainfall. 8 You can do the PMF calculations in ISIS, which can also optimise to find the critical storm duration. Some consultants continue to use the Micro-FSR software, which was developed by the Institute of Hydrology to support the FSR methods.

Table 4 The table below lists dam categories. Dam Potential effect of a breach Design flood inflow category (when overtopping cannot be tolerated) A Endangering lives in a community Probable Maximum Flood (PMF)

B Endangering lives not in a 10,000-year flood community, or causing extensive damage

C Negligible risk to life and limited 1000-year flood damage

D Special cases where no loss of life 150-year flood can be foreseen and very limited additional flood damage would result from a breach (mainly ornamental lakes)

Summary Our summary of guidance on flood estimation for reservoir safety includes the following. References are listed in Related documents. Date Document Main aspects still current (Feb 2014) 1996 Floods and reservoir Overview, legal requirements, safety (3rd ed.) engineering aspects, flood routing and wave calculations 1999 FEH Volume 4 Estimation of PMF and 150-year flood. Rainfall-runoff modelling for other return periods 2004 Floods and Reservoir Choice of method for 1000-year and Safety Revised 10,000-year rainfall. Summary of other

Guidance for Panel current guidance. under review by CEH. Engineers

Estimating long return period floods (150-1000 years)

The issues There has been an increasing demand for flood estimates at return periods longer than 100 years, particularly in flood mapping and flood-warning studies that now include the 1000-year flood. In Wales, TAN15 also requires Doc No 197_08 Version 5 Last printed 12/02/15 Page 99 of 110

developers to assess the 1000-year flood on their development.

Description All flood estimates for extreme return periods rely, however indirectly, on extrapolation. For this reason, given the typical length of flood peak records, the FEH statistical method was originally recommended principally for return periods up to 200 years. In the past, most flood estimates for longer return periods have been derived from the FSR/FEH rainfall-runoff method.

Guidelines The guidelines and advice in the table below are included to help users. Select references that are linked to see details in Related documents. Item Guideline or advice 1 The main reason for preferring a rainfall-runoff approach is that you can define rainfall growth curves for long return periods with much more confidence than flood growth curves. This is particularly true for the FEH rainfall analysis, which drew on a much larger and longer dataset than is available for flood peak data. Furthermore, the spatial consistency of extreme rainfall allowed the rainfall growth curves to be extended to long return periods using a model of spatial dependence.

2 Before the ReFH method was developed, the FEH statistical method was widely used (outside its initial recommended range) to estimate 1000-year floods. This was partly due to concerns that the FEH rainfall-runoff method overestimated design flows in many locations. A significant application of the statistical method at long return periods was the automated estimation of flows by CEH, which were subsequently used in the mapping of the Environment Agency’s Extreme Flood Outline. See Morris, D.G. (2003).

3 The design event used in ReFH is calibrated up to return periods of 150 years. This is an improvement on the FEH rainfall-runoff model that was only calibrated up to 10 years, in a simulation exercise carried out for the FSR research in the early 1970s. However, there are concerns that some aspects of the ReFH design procedure have not been tested at return periods longer than the calibration limit of 150 years. The most obvious extrapolation is for the αT calibration coefficient, see Figure 22. Another concern about using ReFH for long return periods is that the seasonal correction factors used for design rainfalls may not be applicable for extreme events (which may be caused by different rainfall processes). The ReFH research only derived correction factors for a return period of five 5 years, although it found no strong variation with the return period. More recent research has led to revised design rainfalls and seasonal correction factors. See Stewart, Lisa and others (2010). Eventually ReFH will be recalibrated to incorporate these changes.

4 Despite the concerns described in items above, the ReFH method is worth considering as an approach for estimating floods for return periods up to 1000 years, apart from catchment types where ReFH is not recommended. See When to apply ReFH with Doc No 197_08 Version 5 Last printed 12/02/15 Page 100 of 110

caution. Its results should be treated with caution, and always compared with pooled growth curves from the statistical method.

5 If flood estimates are needed for a range of return periods up to 1000 years, it may often be the case that the statistical method is preferred for the shorter return periods. To avoid a discontinuity in the results if you choose ReFH for the longer return periods, one approach is to use ReFH to obtain the ratio of the 1000-year flow to the (say) 100-year flow. You can then multiply that ratio by the preferred estimate of the 100-year flow, which may be from the statistical method.

6 Historical flood data are particularly valuable as a guide in the estimation of extreme design events. If you can identify a flood chronology spanning several hundred years, this may lead to a statistical approach being preferred for estimation of 1000-year flows.

Figure 22 The graphs in the diagram below show the variation of the αT coefficient in the ReFH design event. In the winter event, αT varies only slightly with return period, with the graph starting to level off as the return period approaches 150 years. Extrapolation of this relationship is therefore unlikely to introduce significant errors.

Reproduced from Kjeldsen and others. (2005) with the permission of CEH Wallingford.

Summary Our summary of recommendations for estimating long return period floods: • No current method can be recommended unequivocally. Guidance is likely to keep developing as research continues. • There are theoretical reasons for preferring rainfall-runoff approaches at long return periods. • For long return periods, apply the ReFH rainfall-runoff method with caution and always compare it with the FEH statistical method. • Where flood peak data are available, gather information on the longer- term flood history and use it to guide the derivation of the flood frequency curve. • Consider the physical processes that might result in a 1000-year flood, and whether these might be different from processes that give rise to more moderate floods.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 101 of 110

Audit trail Overview

Flood estimation calculation record

Purpose The Flood estimation calculation record (SD01) supports these guidelines and serves three important functions:

• to help analysts ensure that they have thought through the choice of approach and applied the methods correctly; • to assist analysts, reviewers and project managers by setting out the calculations in a standard format; • to provide an audit trail of the study so that the work can be reproduced in the future if needed.

Using the The flood estimation calculation record is mainly intended to be an appendix record to a report. The report usually includes supporting information, such as the background to the study, a description of the catchment, details of any rating reviews and a summary of the results. You could use it as a stand-alone record of a minor study. In this case, you can expand it to include other items, such as a description and map of the catchment. It is designed for studies that include multiple flow estimation points. You can add more rows to the tables, as required. There is a shorter version. The Flood estimation calculation record for single sites (SD02) also supports these guidelines.

Requirement Documenting calculations and the decisions made is mandatory for all Environment Agency staff and consultants working on Environment Agency projects. Using the flood estimation calculation record is the recommended way of doing this. You may use other records with the agreement of the project manager.

Filling in the calculation record

Description The calculation record consists of a series of tables for you to fill in. It does not attempt to record all the parameters used and decisions made during calculations. Instead, it provides enough information for a review and to enable future reproduction of the results. The most important aspects to record are those that deviate from the default methods. The table below describes the sections.

Guidelines Section Description Doc No 197_08 Version 5 Last printed 12/02/15 Page 102 of 110

First This is a method statement (see Preparing method section statements). It covers the requirements of the study, a review of hydrometric data and an initial choice of method. You should record this at the start of the study. For lengthy or high-risk studies, you should agree it with the project manager before carrying out further work.

Second Deals with selecting subject sites and checking catchment section descriptors. In the multi-site version of the record, each site is referred to by a site code, which saves having to type in the river and location name in every subsequent table. The site codes could be gauging station numbers, if all the sites are at gauges, or hydraulic model node labels if available. In many cases, however, analysts have to make up suitable site codes. To help reviewers, make codes follow a logical system, such as a four-letter code based on the river name plus a two-digit code starting at the downstream end of the study reach. Example: LAMB01 for a site at the downstream end of the River Lambourn. You can copy and paste the catchment descriptors into the Word document from a spreadsheet.

Next Covers the FEH statistical method. The first tables deal with section estimation of QMED. All pooling groups used in the study are described after that, with the following table showing how the growth curve was derived at each subject site. This gives you the scope to apply a pooling group at several subject sites.

Subsequ Describes flood estimation using the ReFH and FEH rainfall- ent runoff methods. You can remove any sections not needed. sections

Final Provides a comparison and discussion of the results. Records section the final design flows and how they have been checked.

Other You should regard the calculation record as a minimum requirement. You information can add other information when necessary. The record does not include tables for recording aspects that are only carried out occasionally, to avoid it becoming excessively long and unwieldy. Example: Some studies include detailed reviews of ratings or historical reviews. You can add them to the calculation record or present them in an accompanying report. The calculation record is not designed for recording the use of non-standard methods, such as continuous simulation. You will need to report them separately in detail. The calculation record is not intended for recording PMF calculations used in reservoir safety assessments. You can modify it, if required, for such situations.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 103 of 110

Presenting results

Guidelines Item Guideline or advice 1 Analysts: you should consider the needs of the study when presenting results. In some cases, these may need to be presented at public meetings or in press releases and should respect the knowledge of a lay audience.

2 Analysts: Do not just hand over the output produced by the FEH software. You have a responsibility when presenting results: • to avoid implying false levels of accuracy or high confidence, especially when confidence intervals cannot be quoted; Example: Using too many significant figures, such as quoting the 100-year flood as 145.7m3/s. • to set down any qualifications or other limitations of the study clearly and ensure they are understood by the project manager; • to discuss how the figures should be best used and presented as a result of the uncertainties, or what could be done to improve them. 3 In many cases, when reporting the return period of a notable flood it will be sufficient to indicate its severity. You should not quote the best estimate too precisely: 'larger than 100 years' or 'between 5 and 10 years'. Simply report the event as the second highest in 30 years of data to meet the needs for press releases and so on. Refer to the Environment Agency Policy on Communicating and Understanding Flood Risk. Section 6 gives more information on assumptions, limitations and uncertainty.

4 Estimating design flows rarely marks the end of a project. In many cases, the flows are used as input to a hydraulic model. Analysts: if you are not going to be doing the modelling, you should ensure that you provide enough information for the modeller. The final choice of peak flows should be clear. For an unsteady hydraulic model, you should provide inflow hydrographs (see Flood mapping and hydrodynamic models, for further discussion on applying flows to unsteady models).

Recording the data used

Saving the We can only reproduce calculations if we can access the data that was used data again. If you have used the HiFlows-UK dataset without alteration, it is sufficient to record the version number of the dataset. If you have made changes, for example updating the flood peak records at selected stations, we recommend that you keep a copy of the entire altered dataset, to ensure that the pooled growth curves can be reproduced.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 104 of 110

List of acronyms

Acronyms The table below lists acronyms that are related to flood estimation. To look up all terms and acronyms, you can use the Glossary on Easinet.

Acronym Full expression ADVP Acoustic Doppler Velocity Profiler AEP Annual Exceedence Probability

AREA Catchment area (km2) BFI Base Flow Index

BFIHOST Base Flow Index estimated from soil type

CFMP Catchment Flood Management Plan DDF Depth Duration Frequency

DEM Digital Elevation Model DPLBAR Index describing catchment size and drainage path configuration

DTM Digital Terrain Model

FARL FEH index of flood attenuation due to reservoirs and lakes

FCA(s) Flood Consequence Assessment(s) GEV General Extreme Value (a statistical distribution)

GL General Logistic (a statistical distribution)

HOST Hydrology of Soil Types MORECS Meteorological Office Rainfall & Evaporation Calculation System

MOSES Meteorological Office Surface Exchange Scheme

PMF Probable Maximum Flood

POT Peaks Over a Threshold

PROPWET FEH index of proportion of time that soil is wet

QBAR Mean annual maximum flood

QMED Median annual maximum flood (with return period 2 years)

R&D Research and Development

ReFH Revitalised Flood Hydrograph method

RMED Median annual maximum rainfall (mm) Acronym Full expression

Doc No 197_08 Version 5 Last printed 12/02/15 Page 105 of 110

SAAR Standard Average Annual Rainfall (mm)

SPR Standard Percentage Runoff

SPRHOST Standard Percentage Runoff derived using the HOST classification

SUDS Sustainable Urban Drainage Systems

Tp Time to peak

Tp(0) Time to peak of the instantaneous unit hydrograph

URBEXT1990 Original FEH index of fractional urban extent

URBEXT2000 Updated version of urban extent, defined differently from URBEXT1990

WINFAP-FEH Windows Frequency Analysis Package - FEH version

WRAP Winter Rainfall Acceptance Profile

Doc No 197_08 Version 5 Last printed 12/02/15 Page 106 of 110

Related documents

Supporting 197_08_SD01 Flood estimation calculation record. documents 197_08_SD02 Flood estimation calculation record for single sites 197_08_SD03 Checklist for reviewing flood estimates

Chapter 2 414_07 Accessing Hydrological Data and Information Archer, D.R. (1999) Practical application of historical flood information to flood estimation. Hydrological Extremes: Understanding, predicting, mitigating, Ed by L Gottschalk, J-C Olivry, D. Reed and D Rosbjerg. IAHS Publisher 255, 191-199 Bayliss, A.C., Black, K.B., Fava-Verde, A., Kjeldsen, T.R. (2007). URBEXT2000 – A new FEH catchment descriptor EA/Defra R&D Technical Report FD 1919/TR Bayliss, A.C. and Reed, D.W. 2001 The use of historic data in flood frequency estimation. Report to MAFF. CEH Wallingford, March 2001 – download from the NERC website BHS Chronology of British Hydrological Events on the University of Dundee website Black, A.R. and Fadipe, D. (2009). Use of historic water level records for re- assessing flood frequency. WEJ (Journal of CIWEM) 23, 23-31 Brown, A.G. (2009). An Integrated 1500 Year Record for the River Trent (UK) Using Geomorphological and Geoarchaeological Data. Geophysical Research Abstracts, Vol. 11, EGU2009-13772, EGU General Assembly 2009. Herschy, R.D. 1998 Hydrometry – Principles and Practice (2nd Ed) John Wiley & sons, Chichester, UK JBA Consulting (2003) River Gauging Station Data Quality Classification (GSDQ) Final Research Report Environment Agency R&D Report W6-058 Kjeldsen, T.R., Jones, D. A. and Bayliss, A.C. (2008) Improving the FEH statistical procedures for flood frequency estimation. Science Report SC050050, Environment Agency. MacDonald, N (2009). Understanding high magnitude flood risk: evidence from the past. Presentation at BHS meeting on Emerging Challenges in Flood Hydrology, 6 May 2009. Macdonald, N. & Black, A. R. (2010) Reassessment of flood frequency using historical information for the River Ouse at York, UK (1200-2000). Hydrol. Sci. J. 55(7), 1152-1162. Ramsbottom, D.M. and Whitlow, C.D. (2003) Extension of Rating Curves at Gauging Stations, Best Practice Guidance Manual Environment Agency R&D Manual W6-061/M

Chapter 3 Archer, D., Foster, M., Faulkner, D. and Mawsdley, J. (2000) The synthesis of design flood hydrographs. In: Flooding Risks and Reactions. Proceedings of the Water Environment 2000 Conference, 5 October 2000. Institution of Civil Engineers, London, pp. 45-57 HR Wallingford (2005). Flooding in Boscastle and North Cornwall, August 2004. Phase 2 Studies Report. Report EX5160

Doc No 197_08 Version 5 Last printed 12/02/15 Page 107 of 110

Chapter 4 Bayliss, A.C., Black, K.B., Fava-Verde, A., Kjeldsen, T. R. (2007). URBEXT2000 – A new FEH catchment descriptor. EA/Defra R&D Technical Report FD1919/TR Calver, A., Crooks, S., Jones, D. Kay, A., Kjeldsen, T. and Reynard, N. (2005) National river catchment flood frequency method using continuous simulation. Defra R&D Technical Report FD2106/TR, CEH Wallingford – download from Defra's website. Faulkner, D.S. and Barber, S. (2009), Performance of the Revitalised Flood Hydrograph Method. J. Flood Risk Management 2, 254-261. Faulkner, D, Robb, K and Haysom, A. (2008). Return period assessment of the Summer 2007 floods in central England. Proc. BHS 10th National Hydrol. Symp. Exeter, 227-232 – download from the BHS website Faulkner, D.S. and Wass, P. (2005) Flood estimation by continuous simulation in the Don catchment, South Yorkshire, UK. WEJ (Journal of CIWEM) 19, 78-84 Gaume, E. (2006) On the asymptotic behaviour of flood peak distributions. Hydrol. Earth Syst. Sci, 10, 233-243 – download from their website HiFlows_UK data (hosted by CEH from April 2014) http://www.ceh.ac.uk/data/nrfa/data/peakflow_overview.html Kjeldsen, T. R. (2007). The revitalised FSR/FEH rainfall-runoff method. Flood Estimation Handbook Supplementary Report No. 1. Centre for Ecology and Hydrology – download from their website Kjeldsen, T. R. (2009). Modelling the impact of urbanisation on flood runoff volume. Proc. Instn. Civ. Engrs. Wat. Man. 162, 329-336. Kjeldsen, T. R. (2010). Modelling the impact of urbanization on flood frequency relationships in the UK. Hydrol. Res. 41. 391-405. Kjeldsen, T.R., Jones, D.A. and Bayliss, A.C. (2008) Improving the FEH statistical procedures for flood frequency estimation. Science Report SC050050, Environment Agency Kjeldsen, T.R., Stewart, E.J., Packman, J.C., Folwell, S. and Bayliss, A.C. (2005) Revitalisation of the FSR/FEH rainfall-runoff method. Defra R&D Technical Report FD1913/TR, CEH Wallingford – downloaded from Defra's website Natural Environment Research Council (1975). Flood Studies Report. 5 volumes. NERC, London Stewart, Lisa; Jones, David; Morris, David; Svensson, Cecilia; Surendran, Suresh. 2010 Developing FEH methods - a new model of rainfall frequency. In: Flood and Coastal Risk Management Conference 2010, Telford, UK, 29 June - 1 July 2010. Environment Agency.

Chapter 5 Environment Agency (2010). Flood and Coastal Risk Management Modelling Strategy 2010-2015. Kjeldsen, T.R. and Jones, D.A. (2004). Sampling variance of flood quantiles from the generalised logistic distribution estimated using the method of L- moments. Hydrol. Earth Syst. Sci., 8, 183-190 Pappenberger, F. and Beven, K. J. (2006). Ignorance is bliss: Or seven reasons not to use uncertainty analysis. Water Resour. Res. 42, W05302, doi:10.1029/2005WR004820 - download the abstract from the WRR website Chapter 6 260_05 Understanding and Communicating Flood Risk Doc No 197_08 Version 5 Last printed 12/02/15 Page 108 of 110

296_05 Guidance - 1000 year flow estimates for Flood Consequence Assessments - Wales only Archer, D. R. (1981) A catchment approach to flood estimation. J. Inst. Water Engrs. & Scientists, 35 (3), 275-289 Archer, D. R. (1981) The seasonality of flooding and the assessment of seasonal flood risk. Proc Instn Civil Engrs., Pt 2, 1023-1035 Bailey, A.D., Dennis, C.W., Harris, G.L. and Horner, M. W. (1980) ADAS Report No. 5: Pipe size design for field drainage Bayliss, A.C., Black, K.B., Fava-Verde, A., Kjeldsen, T. R. (2007). URBEXT2000 – A new FEH catchment descriptor. EA/Defra R&D Technical Report FD1919/TR Beran, M. (1987) Review of ex-GLC rainfall-runoff method. Institute of Hydrology, Wallingford, UK. Beskeen, T., Elwell, F., Wilson, C. and Sampson, T. (2011) COSH! Tackling sewer pipe and overland flow in small urban catchments. J. Flood Risk Management 4, 234-246. Bradford, R.B. and Faulkner, D.S. (1997) Review of Floods and Flood Frequency Estimation in Permeable Catchments. MAFF R&D Project FD0423, Institute of Hydrology, Wallingford Bradford, R.B. and Goodsell, G. (2000) Flood Volumes and Durations in Permeable Catchments. Report FD1605 for MAFF by CEH Wallingford Cox, D.R. (2003) Some comments on 10,000 year return period rainfall. Report to Defra. Faulkner, D., Kjeldsen, T., Packman, J and Stewart, E. (2012). Estimating flood peaks and hydrographs for small catchments: Phase 1. Science Report SC090031/R, Environment Agency. Phase 1 report available from http://evidence.environment- agency.gov.uk/FCERM/en/Default/FCRM/Project.aspx?ProjectID=0C27686 F-1421-4359-98BB-F491FF17DDC5&PageId=a0fe6dfc-506a-452c-9bff- a7ec06b4e6b0 Flikweert , J. and Worth, D. (2012). Pumped catchments - Guide for hydrology and hydraulics. Science Report SC090006, Environment Agency. Reports available from http://evidence.environment- agency.gov.uk/FCERM/en/Default/FCRM/Project.aspx?ProjectID=2CB10C9 4-1E43-4C70-BA24-1073B78B890B&PageId=a0fe6dfc-506a-452c-9bff- a7ec06b4e6b0 FRQSIM Hydrological Model. Technical Manual. Flood Modelling Group, Flood Defence, Thames Region. Environment Agency (undated) HR Wallingford (2013) Greenfield runoff estimation at site level http://geoservergisweb2.hrwallingford.co.uk/uksd/greenfieldrunoff_js.htm Institution of Civil Engineers (1996). Floods and reservoir safety. Third Edition. Thomas Telford, London IWEM (1987) Water Practice Manuals. 7. River Engineering - Part 1, Design Principles. Institution of Water and Environmental Management Kellagher, R. (2012). Preliminary rainfall runoff management for developments. HR Wallingford report for Defra/Environment Agency. Technical Report W5-074/A/TR/1. Revision E. MacDonald, D.E. and Scott, C.W. (2000) Design floods for UK reservoirs – a personal view of current issues. Proceedings of 11th British Dam Society Conference, pp 1-11. Thomas Telford, London. Moore, R.J. and Bell, V.A. (2002) Incorporation of groundwater losses and

Doc No 197_08 Version 5 Last printed 12/02/15 Page 109 of 110

well level data in rainfall-runoff models illustrated using the PDM. Hydrol. and Earth System Sci. 6 (1), 25-38. Morris, D.G. (2003) Automation and appraisal of the FEH Statistical Procedures for Flood Frequency Estimation. Final report to Defra, CEH Wallingford National SUDS Working Group (2004). Interim Code of Practice for Sustainable Urban Drainage Systems – download from the CIRIA website National Water Council (1981). Design and analysis of urban storm drainage – the Wallingford procedure. Onof, C., Faulkner, D. and Wheater, H.S. (1996). Design rainfall modelling in the Thames catchment. Hyd. Sci. J. 41(5), 715-733. Reed, D.W. (2002) Reinforcing flood-risk estimation. Phil. Trans. R. Soc. Lond. A 360, 1373-1387 Reservoir Safety Working Group (Institution of Civil Engineers), 2004. Floods and Reservoir Safety – Revised Guidance for Panel Engineers. Samuels, P.G. (1993) The hydraulics and hydrology of pumped drainage systems – An engineering guide. HR Wallingford, Report SR331 Stewart, Lisa; Jones, David; Morris, David; Svensson, Cecilia; Surendran, Suresh. 2010 Developing FEH methods - a new model of rainfall frequency. In: Flood and Coastal Risk Management Conference 2010, Telford, UK, 29 June - 1 July 2010. Environment Agency. http://nora.nerc.ac.uk/12490/2/N012490CP.pdf Webster, P. (1999) Factors affecting the relationship between the frequency of a flood and its causative rainfall, in Hydrological Extremes: Understanding, Predicting, Mitigating (Proc. IUGG99 Symposium HS1, Birmingham, July 1999). IAHS Publ. No. 255. Woods-Ballard, B., Kellagher, R., Martin, P., Jefferies, C., Bray, R., Shaffer, P. (2007) The SUDS Manual. CIRIA C697.

Doc No 197_08 Version 5 Last printed 12/02/15 Page 110 of 110

Computational modelling to assess flood and coastal risk

Operational instruction 379_05 Issued 27/10/10

What’s this This document is a guide for assessing flood and coastal document risk using computational modelling. It gives an overview of about? good practice to consider when we carry out modelling. It supports our requirement to take a risk-based approach to managing flood and coastal risk. Document We assess flood and coastal risk (for instance for details improvement schemes), as do developers or consultants (for instance for development purposes). We also work with our partners to develop modelling, especially lead local flood authorities and water companies. This document focuses on modelling for flooding from rivers and/or the sea. It does not focus on other types of modelling (such as for surface water flooding or groundwater flooding) Related although much of the guidance can be applied to these documents types of modelling. Other documents are available which focus on modelling for local flood risk (for example WaPUG guides). References to these are provided in the related documents section of this document.

Who does this Environment Agency staff in Flood and Coastal Risk apply to? Management (FCRM), particularly Flood Risk Mapping and Feedback Data Management teams and National Capital Programme Management Services (NCPMS), but also applicable to Flood Forecasting, Asset Systems Management, etc. Contact for It can also be shared with partners or other third parties queries (such as developers) to help with their work, as long as it is made very clear that this document was developed for 1. FRM&DM internal purposes. When supplying it you must also OTL enclose/include a copy of Special Licence (Copyright) (Word, 78KB). 2. FCRM Helpdesk

3. Helen Contents Section Page James 1. Background 2 2. Model selection 5 3. Model construction 13 4. Fluvial boundary conditions 17 5. Tidal boundary conditions 19 6. Calibration, verification and sensitivity testing 25

7. Mapping and reporting 27

Doc No 379_05 Version 2 Last printed 27/10/10 Page 1 of 31

1. Background

Contents This chapter describes the background to computational modelling in the context of Flood and Coastal Risk Management:

Topic See page FCRM Modelling Strategy 2 Using models to assess flood and coastal risk 2 Approach to the project 2 Support for modelling 2 Modelling skills 3 Historic information 4

Probabilistic modelling 4

FCRM This document supports the principles of the FCRM Modelling Strategy Modelling 2010-2015. It particularly applies to the following principles: Strategy ƒ Modelling will be developed and shared with partners; ƒ Uncertainty in our modelling will be understood; ƒ Modelling will be managed effectively, in partnership; ƒ Our modelling will continue to be an asset; ƒ We will be an intelligent client with adequate resources to carry out that role; ƒ Technology will support our modelling.

Using models Consider these points when deciding whether to use modelling to assess to assess flood risk: flood and coastal risk ƒ You do not have to use hydraulic modelling to assess flood and coastal risk; ƒ It may be technically acceptable and cost effective to recycle previous models rather than develop new ones, but only if the recycled model is fit for purpose. ƒ In less complex assessments simple hydrological and hydraulic analysis may be sufficient; ƒ Even if you do not use modelling, assess the impact of any proposed development on runoff using Flood Estimation Handbook techniques, or most appropriate equivalent.

Approach to Consider these points when approaching the project: the project At the start ƒ Clearly define our objectives and required outputs, and those of our partners (whether within the Environment Agency or external partners such as local authorities and Local Resilience Forum (LRF) members). Confirm what each partner is going to contribute in terms of budget, resources, data, etc.. Review the work against these intentions both at intervals and at completion; Continued on next page…

Doc No 379_05 Version 2 Last printed 27/10/10 Page 2 of 31

Approach to ƒ Clarify the boundary conditions and other design parameters; the project continued ƒ Do a one-off request for information held by other Environment Agency departments at the very beginning of the project since this affects selection of method etc, and could prevent further information coming to light at a later stage and complicating matters; ƒ Make sure you complete a Data Management Plan (see 183_05 Data Management Plans for Flood Risk Management Projects) Locations ƒ Consider which sources of flood risk affect a location, and what level of detail and accuracy is required when planning a modelling study; ƒ Discuss requirements at specific locations with local experts and partners to ensure that any site-specific factors are identified, which may require special treatment when modelled; ƒ Where the modelling is being undertaken in relation to a development, ensure the study area is sufficient to demonstrate the effects of the development on locations away from the project site; Choice of model ƒ Ensure the most appropriate modelling approach is agreed on and used (see section 2 of this document for more information). Numerical modelling is not always necessary to assess flood and coastal risk. In less complex assessments, simple hydrological and hydraulic analyses may be sufficient; ƒ A value for money approach avoids unnecessary complexity, whilst ensuring that the key processes in the real world system are well represented and the required level of detail and outputs are achieved to satisfy the modelling objectives. Be clear how the approach you have taken meets the outcomes of the study; ƒ Ensure the approach chosen is fit for purpose, but think about possible future uses too, so that the modelling can be re-used; ƒ Proof of appropriateness should include, but not be limited to, a defence of the modelling software choice, dimensionality (1D, 2D, linked 1D/2D, etc), state (hydrodynamic, steady state, routing), characteristics (strengths / limitations) and scale (detailed / national generalised). Documentation ƒ Ensure that the modelling methods are documented to a level of detail sufficient to allow us to replicate the work and use the model in the future. Follow the SFRM performance scope (available from www.sfrm.co.uk). See section 7 of this document for more information.

Support for If there is any doubt whether modelling is required, discuss the situation with modelling the Area Flood Risk Mapping & Data Management team at the earliest opportunity. They will also be able to provide suitable information to help with the modelling.

Modelling In all modelling, the experience of the modeller adds value so ensure that skills suitably qualified and experienced people carry out the work. Technical Development Frameworks are available to assess skills against required competencies for modelling.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 3 of 31

Historic Collect and use historic data from such sources as: information ƒ Historic flooding (such as newspaper articles, photos, flood marks), including information on historic flooding prior to the periods covered by hydrometric data, to guide the extent of any survey and to aid the modelling process. Such data is particularly valuable as it can provide information for model calibration and verification; ƒ The internet (for example, the Chronology of British Hydrological Events, http://www.dundee.ac.uk/geography/cbhe); ƒ Alterations and additions to the watercourse and associated structures, to coastal defences, or within the flood plain, since the date of the recorded flood event; ƒ Area Flood Risk Mapping and Data Management teams.

Probabilistic Currently the majority of modelling done to assess risk in the Environment modelling Agency is deterministic. The main exception to this is the Risk Assessment for System Planning (RASP) approach used for the National Flood Risk Assessment (NaFRA). Our FCRM Modelling Strategy 2010-15 states that we will move towards a probabilistic approach to understanding risk as our standard approach. Research is underway to understand how we can validate the outputs of probabilistic models and how we can re-use our existing detailed deterministic models in a probabilistic way. Once these two areas of research are delivered we will review this guidance to include more detail about using computational modelling to produce probabilistic outputs.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 4 of 31

2. Model selection

Contents This chapter describes how to choose the appropriate modelling approach and software, what data inputs should be considered and what to think about before starting model building, and includes the following topics:

Topic See page Uses of modelling 5 Qualitative description of risk 5 Choice of software 6 Modelling dimensions: 1D, 2D, and 3D 7 Modelling state 8 Modelling characteristics 8 Re-use of existing modelling 10 National generalised modelling 11

Integrated modelling 12

Uses of Modelling is used to calculate: modelling ƒ flow and water level conditions in rivers, tidal rivers, estuaries and at the coastline; ƒ boundary wave conditions for tidal flood risk assessments; ƒ the flood extent, depth, velocity, hazard, timing, duration and flow paths over the fluvial or tidal flood plain; ƒ loadings on defences.

Qualitative Before deciding upon the approach to modelling or the software used, it is description of first very important to understand the processes that influence the flood or risk coastal risk. The source-pathway-receptor concept is widely accepted as a means of categorising these processes. You should seek to understand locally important factors that are relevant to and significant for the flood and coastal mechanisms under consideration. Source: where and how floodwater is generated Pathway: where and how floodwater is conveyed and stored through the catchment or reach Receptor: where the floodwater impacts and the features affected by flooding The treatment or absence of the recorded features in the final model may also be used to inform statements of confidence and uncertainty attached to the modelling. It is important to understand if / how features known to be important are represented in the model. The implications and reasons for not including such features should be clearly understood.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 5 of 31

Choice of These are important points to remember when choosing software: software – some ƒ use modelling software capable of producing the required output that considerations has been demonstrated to be suitable for your needs; ƒ the software should be suitable for the application intended according to available benchmarking tests. If the available tests are not appropriate, you may need to have independent benchmarking tests / peer reviews carried out to prove the proposed modelling software is appropriate; ƒ you do not always have to develop a complex solution. Consider the outcomes required and the level of risk before deciding which modelling approach is most appropriate and what the minimum output requirements of the model are. Use the simplest modelling approach compatible with the desired outcomes; ƒ hydrological and hydraulic analysis, without using modelling software, though perhaps in association with GIS software, may be all that you need; ƒ software is often updated so be aware of available features when making a selection and record which version you use in the metadata.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 6 of 31

Modelling Flood modelling methods currently used in the UK can be classified by their dimensions – dimensions or the way they combine different dimensions. Those that 1D, 2D and 3D currently support most modelling applications necessary for Flood and Coastal Risk Management are one-dimensional (1D) or two-dimensional (2D). Both 1D and 2D models are now in common use, as are linked 1D/2D models. The latter are particularly useful where there is a strong linear component to one part of the flow yet a two dimensional aspect elsewhere, for example where a river (1D) has an irregular flood plain (2D). Some 2D models also have an integral 1D component for simple representation of channels within the 2D domain. In principle, model 1D situations in 1D, and model 2D situations in 2D; link these if both 1D and 2D situations apply. 2D models 2D models can provide information on flood depth, flow direction, velocity and timing, as well as providing outputs that are available from 1D modelling, such as flood inundation extent and predicted water levels. External drivers for 2D modelling have arisen from: ƒ The Flood Risk Regulations 2009 which requires the prediction of flood hazard over high risk areas. This in turn requires an assessment of flood depth and flood water velocity. 2D hydraulic models provide a relatively low cost means of predicting these; ƒ The Pitt Review following the 2007 floods in England recommends: ƒ better visualisation of the Environment Agency’s flood mapping data; ƒ developing maps that consider surface water risks; ƒ creating inundation maps arising from possible reservoir dam failure; all of which can be enhanced by using 2D models. 3D models Three-dimensional (3D) methods are currently not in common use for estimating flood risk within the Environment Agency. Examples of where they are used occasionally are for analysing bridge pier scour or understanding deep water movement in estuarine or coastal environments.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 7 of 31

Modelling The choice of which model to use should be made between: state ƒ a hydrodynamic 1D or 2D model ƒ hydrodynamic combined 1D/2D model ƒ steady–state 1D model ƒ river flood routing model A full hydrodynamic model, that is one in which flows and water levels vary with time, must be used if the study area contains either structures whose operation varies with time (for example pumps, sluices and tidal outfalls) or involves representation of tidal conditions. This should also be employed where there is significant flood plain storage or where a watercourse is subject to rapid increases and decreases in flow. In other cases, either a steady-state or hydrodynamic model may be chosen. It should be noted that a steady-state model, that is one in which flows and water levels are constant over time, is unlikely to give a reasonable estimation of water levels where these are influenced by storage effects. A flood routing model can be used in preference to a full hydraulic model if detail of flood water levels is not needed.

Modelling The three tables below list respective characteristics of 1D, 2D and linked characteristics 1D/2D models with respect to their simulation of flows and water levels in channels and over floodplains. Using 1D and 2D models to generate tidal boundary conditions is described later.

1D models The table below lists the strengths, limitations, and applications of 1D models:

Strengths Limitations Applications ƒ Simulate flows for a large ƒ Limited to where direction ƒ River and tidal river flood range of hydraulic of water movement is risk modelling. structures such as weirs, aligned to the centre-line ƒ Urban drainage gates, and sluices. of the channel. modelling. ƒ Simulate effects in tidal ƒ Assumes unidirectional ƒ Can be extended to rivers. flow. modelling of flow in ƒ Can have a storage cell ƒ Flow velocities are depth compound channels approach for the averaged across the (channel + floodplain) but simulation of floodplain cross-section need to remember that flow added to represent a floodplain flow is assumed ƒ Conveyance can be simplified version of 2D to be parallel to main severely over- or under modelling (pseudo 2D) for channel estimated. broad-scale modelling. ƒ Particularly appropriate ƒ Cannot simulate for narrow floodplains floodplain flow unless flow where there is no routes are known separation of the channel beforehand. from the floodplain by ƒ Crude representation of embankments / levees. floodplain storage capacity. ƒ Assumes uniform flood water level at each cross-

Doc No 379_05 Version 2 Last printed 27/10/10 Page 8 of 31

section. This may lead to no discernment between levels in the river and those behind raised defences, or on a floodplain at lower level than the river, if the model is not schematized correctly. ƒ Often assumes constant roughness values throughout the event, regardless of varying flow depth. ƒ Very rarely appropriate for modelling coastal flooding.

2D models The table below lists the strengths, limitations, and applications of 2D models. The 2D benchmarking R&D desktop review and report, provides a fuller description of the different types of 2D model type available and their relative strengths.

Strengths Limitations Applications ƒ Provides information on ƒ Integral 1D component ƒ River modelling where the magnitude and timing gives only a simple detail is required on of depth, flow direction representation of linear floodplain inundation. and velocity, as well as channel flow within the 2D ƒ Coastal and estuarine flood inundation extent domain. flood risk modelling. and predicted water ƒ Not suitable if river levels. ƒ Surface water flood risk channels are expected to modelling. ƒ Simplified versions are act as important conduits available to use where of tidal ingression inland. ƒ Reservoir inundation risk quicker run-times are modelling (note that ƒ Requires significant required (review hydraulic jumps are more computation power and benchmark test results to accurately modelled if the can take a considerable decide which type of software incorporates a time to run if a fine grid is software package is shock capturing scheme). used. appropriate for your ƒ Where hazard (depth and ƒ Can take longer to needs). velocity) outputs are calibrate than 1D. ƒ Quick to set up. required. ƒ Can take longer to run ƒ Quick to generate flood than 1D. maps. ƒ The model accuracy can ƒ Can be linked with be dependent upon the existing 1D models to grid size as well as the ensure re-use of existing quality of the Digital models (making best use Terrain Model (DTM) data of previous investment in used. modelling). See below. ƒ Requires large data storage capability for results. ƒ There is a lack of

Doc No 379_05 Version 2 Last printed 27/10/10 Page 9 of 31

available data for verifying the results of 2D models.

1D and 2D The table below lists the strengths, limitations, and applications of 1D and linkage models 2D linkage models:

Linkage Strengths Limitations Applications 1D-2D ƒ If the flood path is ƒ Not helpful for ƒ Flood cell sequential simply one of recession of river or inundation linkage overtopping with no tidal breach. significant return to the source river or tide, during simulation, separate 1D and 2D models may be more straightforward to construct

1D – 2D ƒ Uses the strength of ƒ Requires significant ƒ River, tidal river dynamic 1D modelling for the computation power and and estuary linkage linear features (water can take considerable modelling where courses) and the time to run if a fine 2D thorough strength of 2D grid is used. representation of modelling for flows the channel is ƒ Can take longer to over the floodplain required along calibrate than 1D. (computational with detail of savings over ƒ Can have instability floodplain structured fully-2D issues; inundation. approaches where a ƒ Requires large data finer grid would be storage capability. required to correctly represent channel geometry). ƒ Can also link 1D piped network model to 2D floodplain model. ƒ Can simulate tidal effects in both channel and floodplain.

Re-using We may hold existing river modelling useful for flood risk assessment (for existing instance produced during flood mapping studies, the design of flood modelling alleviation schemes, for flood forecasting purposes or for Flood Risk / Consequence Assessments). Consider whether you could use this, either directly or with some modification, as part of the flood risk assessment.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 10 of 31

Verifying the Verify the fitness for purpose of existing modelling before re-using it. Some model for re- points to consider are: use ƒ Is the model coverage and level of detail suitable for the new purpose? ƒ Is the representation of channels, floodplain, structures and defences still valid? ƒ Is the schematisation acceptable? ƒ Are the hydrological inputs suitable? ƒ Does the model run satisfactorily? how long does it take to run? You can find such information by running the model and reading the modelling reports. Check surveys If modelling or survey data are provided by us or third parties, arrange check surveys at key locations to ensure that the data provided is compatible with current conditions.

Re-use: cost, Resolve any cost, licensing and Intellectual Property Rights issues licensing and associated with the use of existing modelling. intellectual Intellectual Property property Intellectual Property (IP) refers to assets that originate from our or others’ creativity. Examples of IP assets are datasets, databases, software, and maps. Intellectual Property Rights (IPR) are the legal rights that protect our IP assets. They include patents, trademarks, copyright, design rights. When we receive IP assets from others, ownership does not transfer to us unless a contract says it does. If others retain ownership, we need to know what we are allowed to do with it and, when practical, make the way we intend to use it transparent to them. There are a few documents which explain Intellectual Property in more detail. You can access these from the related documents section of this document. Charging and licensing Refer to 98_07 charging and licensing for flood risk information for more details.

National The table below helps explain the differences between detailed modelling generalised and national generalised modelling. modelling For further details on the National Generalised Modelling data for Flood Zones you should refer to 229_06 Provision and fitness for purpose of the National Generalised Modelling (JFLOW/HYDROF) including climate change depth difference data.

Feature/characteristic Detailed model Generalised model Ground levels Detailed site survey / LIDAR / National DTM – broad scale Photogrammetry Output data calibrated and Yes – when possible No verified? (QA’d) Model Inflows Calculated or from recorded Automated data Input data QA’d Locally Nationally

Doc No 379_05 Version 2 Last printed 27/10/10 Page 11 of 31

Mannings ‘n’ Locally set Globally set Schematisation Detailed using local Simple knowledge Structures Takes account of existing Bare earth simplification infrastructure Application Tailored to the specific needs Not generally appropriate for detailed decision making

Integrated Both the Water Framework Directive and the Pitt review call for an modelling integrated approach, which requires modelling of whole catchments or entire urban drainage systems. Integrated modelling should be seen as more than just linking models together. It is about: developing community where knowledge is shared; providing business processes that support appropriate re-use of models and data, and strong management of models so that a clear audit trail is available. In fact, developing the models is probably the smallest challenge facing integrated modelling. There are three ways that integrated modelling can take place: (1) By linking separate models so the outflows of one model are used as inflows to the other(s); (2) By coupling models using wrapping software to allow models to interact with one another in a more integrated manner than (1); (3) By using fully integrated modelling software which enables the hydraulics and hydrology of the environment to be incorporated into a single model. Each approach has advantages and disadvantages and the modeller should consider which existing data and models are already available, and decide on an approach that is best value for money to achieve the desired outcomes of the study.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 12 of 31

3. Model construction

Contents This chapter advises on sources of data from which to build a model and on the selection of model parameters. It includes the following topics:

Topic See page Survey data 13 Representing hydraulics 14 Hydraulic coefficients 15 Roughness values in 2D models 15 Representing buildings in 2D models 15

Breaching 16

Survey data This table describes what to consider when assembling survey data and commissioning new survey, (ground survey, LIDAR or other). Further guidance on survey standards should be obtained by reference to the Environment Agency National Survey Specification.

Item Action Survey scale ƒ Define the upstream and downstream limits by the objectives of the assessment, rather than to the limits of the immediate project area; ƒ Include the full extent of likely flooding in the lateral extent of the survey (guidance on this extent may come from flooding records and from flood maps); ƒ When in doubt, specify a greater survey extent, particularly where limited LIDAR coverage exists; ƒ Continue the survey far enough downstream so that uncertainty in the boundary condition does not significantly influence the estimated flood levels. Survey features ƒ Ensure the cross sections surveyed are representative of the channel and floodplain; ƒ Determine cross-section spacing and orientation from the appropriate software documentation and textbooks (for example, the online manuals supplied with specific software packages). For 1D modelling, cross- sections should be orientated at right angles to the direction of flow; this may mean the floodplain part of a cross-section has a different orientation to the channel part; ƒ Consider a greater density of cross-section in areas where detailed flood depths or extents are needed; ƒ Sufficient spot levels should be taken along the river banks or coastal defences to ensure that the variation in the bank levels is adequately represented in the model; ƒ Survey all structures (upstream & downstream faces, culvert dimensions, bridge deck levels) unless they have no potential to affect flood flows/flood levels; ƒ Ensure that information on structures, flood routes, potential blockages/obstructions to the channel and channel roughness are gathered;

Doc No 379_05 Version 2 Last printed 27/10/10 Page 13 of 31

ƒ Similarly, ensure information on obstructions across the floodplain, for example as given by road infrastructure and flood plain roughness, is gathered; ƒ Where tidal bathymetry is needed, for example for wave modelling, ensure this is sufficiently detailed so shoaling, refraction and similar effects will be calculated accurately; ƒ Ensure cross-sections of raised defences cover the full section, identifying base levels (as may be needed for breaching calculations) as well as profile slopes, wave wall geometry and surface type, for example grass/concrete (the type is particularly relevant if wave overtopping may need to be calculated). Other ƒ Ensure that the extent of the survey work is defined jointly by those considerations undertaking the modelling and those undertaking the survey in conjunction with advice from the Area Flood Risk Mapping & Data Management team; ƒ Locate all cross sections and other survey information in plan relative to the British National Grid; ƒ The survey data should be provided in a model ready format or in a format that can be easily converted with minimum time and effort; ƒ Photographs of the channel should be taken at the time of the survey. Additional photographs of roughness and blockage should be taken at the time of a walkover by the modeller; ƒ We may hold existing hydrographic and floodplain survey data which may be of use in a flood risk assessment; ƒ LIDAR and local topographic surveys should be reconciled to ensure common datum and spatial coherence; ƒ Consider wider uses of survey data, for example obtaining defence crest levels for use in NFCDD (and subsequent use for NaFRA).

Representing Modelling can be used to represent: hydraulics - features ƒ the key flood flow routes; ƒ flood storage; ƒ barriers to flow; ƒ structures in the study area. Before building the model, schematise these features, preferably on a map background, so their location and points of interactions are clearly understood.

Representing Structures / features: hydraulics - considerations ƒ Include the effect of operational structures, such as sluice gates, although you can adopt a fixed setting if this is the likely situation within the events being modelled; ƒ Where raised features cross a floodplain, also identify openings through these (for example a subway or culvert under a road) so potential flow paths are not overlooked in the modelling; Continued on next page…

Doc No 379_05 Version 2 Last printed 27/10/10 Page 14 of 31

Representing ƒ Use data from detailed ground level survey for spill points in a 1D/2D hydraulics – model. Building such a model purely on standard remote-sensed DTM considerations data is unwise, though the use of high resolution LIDAR (for example continued 0.25m grid) may be sufficient; Topography: ƒ 2D model accuracy is affected by the accuracy of DTM data, representing the terrain and features crossing it, how it is processed and filtered by the data provider, and how it is processed to a grid; ƒ Use both unfiltered and filtered LIDAR to maximise the benefit of the complete LIDAR set and to minimise any shortcomings with the filtered data; Checking outputs: ƒ Check 2D models in detail (especially mass balance at suitable time intervals and plausibility of velocity/depth variations); ƒ Run model animations (1D as a longitudinal profile animation; 2D as a flood spreading animation) to check the flow characteristics look plausible.

Hydraulic Determine the coefficients used in the model (such as channel roughness, coefficients weir coefficients) with guidance from standard textbooks. Reference these texts in the modelling report. Further information on roughness can be obtained from the Conveyance Estimation System. Advice on afflux is given in the Afflux Estimation System.

Roughness 2D models allow spatially-varying roughness and some also allow values in 2D roughness parameters to vary with depth and time. There is a lack of text- models book values of roughness for 2D models though values are suggested in some software manuals. Good practice As good practice, all 2D models should be run for a ‘reference case’ of 0.1 Manning’s n roughness for the entire 2D model domain with the grid based on a filtered Digital Elevation Model (DEM). The model report should then show the difference between the chosen roughness/grid against this (in terms of differences in level, flood extent). You can identify variations in surface roughness in your model to reflect differences in land use. However, avoid large scale variation in roughness values beyond mapping values to key land use types (such as roads, open farmland, etc.). Roughness values should generally increase with model grid size. Mannings ‘n’ values should increase for shallow depths of flow.

Representing The table below describes the four modelling approaches in common use: buildings in All are based on a filtered DEM, thereby removing buildings and vegetation 2D models as a starting point.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 15 of 31

Approach Description 1 Apply an increased roughness value to Simple. The most suitable for modelling the overall floodplain area, taking account where local detail is not needed for of the mixed land use this encompasses. example Flood Zone and ABD assessments. 2 Superimpose the buildings, for example Allows for impedance to flow given by the by using OS Mastermap data. buildings and for flood volume to be dissipated within the area of the buildings. Increase roughness values over the footprint of the buildings to represent how Gives more detail than approach (1). they impede flood flow. 3 Edit buildings to be ‘stubby buildings’ Attributes as approach (2) but adds for the (typically set to the threshold level if obstruction to flow given by the building known, or to a uniform 250-300mm above footprint between ground level and the ground level). threshold level. This improves representation of flow paths and velocities Assign the ‘stubby buildings’ a higher at shallow depths. roughness value to represent how they impede flood flow. Often the best choice for detailed modelling. 4 Represent the buildings as solid blocks, Confines flood flow and dissipation of flood perhaps 5m high. volume to the space between the buildings. Gives the worst case, for example, for flood hazard on roads, provided the model grid size is small enough to represent the road space between the buildings. Can be useful for emergency services planning. The solid blocking leads to underestimation of the flood extent.

Breaching Where a site has raised flood defences, you may want to demonstrate the potential consequences in the event of a breach in those defences. Breaching can occur even in defences that supposedly have a high structural standard, for example due to an undetected weakness. Wave overtopping can be very damaging in a coastal situation. Potential breaching of defences depends on their form, size and condition. Areas teams usually have their own standard criteria, of width, base level and timing within the event, to be used in setting breaching parameters. Also see breaching guidance used in Wales.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 16 of 31

4. Fluvial boundary conditions

Contents This chapter outlines principles for generation of fluvial boundary conditions, given under the following topics:

Topic See page Hydrometric data 17 Hydrological assessment 17 Upstream boundaries (inflows) 17

Downstream boundary (levels) 18

Hydrometric Collate river flow, river level and rainfall data relevant to the study area data where available. This data is most likely to be sourced from the Area Environmental Monitoring (Hydrometry and Telemetry) team. Seek an understanding of the uncertainty and confidence within this data, for example the reliability of flow gauge rating curves, from its local custodian. Use the Flood Estimation Handbook and the UK HiFlows project as sources of hydrological data.

Hydrological Do a hydrological assessment of the flood flows using the methods assessment described in our Flood Estimation Guidelines. If you use hydrodynamic modelling, include consideration of peak flows, flood volumes and shape of the hydrograph in the hydrological assessment. If the problem includes storage (for example floodplain or reservoir storage or a tide-locked watercourse) you must identify the critical duration storm for storage (which often differs from the critical duration for peak flow). If you use a steady-state model, this is limited to consideration of peak flows. Consider the possible effects of climate change on river flows through use of the appropriate contingency allowances.

Upstream Develop the upstream boundary or boundaries, together with lateral inflows, boundary during the hydrological assessment described above. (inflows) For some models, one single upstream inflow per flood event may be sufficient, whilst for others, many upstream boundaries may be needed if a number of tributaries or other inflows are present. Locate the inflows based on hydraulic considerations, not on the upstream limit of the development. The upstream boundary should be far enough upstream to allow the full impact of the development on upstream water levels to be identified.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 17 of 31

Downstream Locate the downstream boundary where the relationship between level and boundary flow is well defined, (for example at a weir). Where this is not possible, (levels) locate it sufficiently downstream of the area of interest so that any errors in the boundary will not significantly affect predicted water levels at the proposed development site or other area of relevance. For a typical fluvial river, a rule of thumb is that a backwater effect extends a length, L, L = 0.7D/s where ƒ D = bank-full depth ƒ s = river slope.

Hence if the downstream boundary is greater than L from the site it is likely that any errors in the rating curve at the boundary will not affect flood levels at the site. Tidal boundaries If the downstream boundary is tidal, locate it where you can accurately define a tidal curve. Outfall structures should be adequately represented to simulate ‘tide-locking’ where this may occur. We hold extensive extreme tide information from flood risk mapping studies. Consider joint probability carefully. See also Tidal boundary conditions below.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 18 of 31

5. Tidal boundary conditions

Contents This chapter outlines principles for generation of tidal boundary conditions, given under the following topics:

Topic See page Boundaries 19 Extreme sea levels 19 Tide and surge curves and their combination 20 Wave conditions 20 Climate change 21 Joint probability tide/wave and tide/river flow 22 Overtopping and breaching 23 Wave overtopping 23

Shingle 24

Boundaries For a tidal flood risk assessment boundary conditions are needed for direct tidal flooding; also for flooding due to wave overtopping or wave run-up if wave action may be present. The model’s tidal boundary needs to allow for the return of water to the sea where this could occur, such as at a breach or with wave action at promenades.

Extreme sea Extreme sea levels for the UK coastline are being defined in a current levels: data Environment Agency R&D project, due to report in 2010. When this data is and rates available you should use it, but until it is available sea levels can be obtained from: ƒ Regional datasets, derived from analysis of local gauge data, perhaps combined with modelling approaches. In some instances these datasets include levels for estuaries/tidal rivers; ƒ POL Report 112; ƒ New analysis of tide gauge data; ƒ Historic observed levels. If data from the more distant past is analysed, adjust it to present-day values by applying a correction to compensate for the historic rate of sea level rise. Rates are given in ƒ POL Report 112 (they are generally about 2mm/year); ƒ POL’s website (www.pol.co.uk/psmsl).

Extreme sea Sea levels at the coastline can be raised by wave set-up. The amount of levels: set-up set-up depends on the location’s exposure to the pertaining wave conditions. Calculation of potential set-up is given by the CIRIA Beach management manual and is a feature of some wave modelling software.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 19 of 31

Extreme sea Levels in estuaries and tidal rivers can be different (often higher) than at the levels: coast, especially if narrowing occurs to give a funnelling effect on the estuaries and incoming tide. tidal rivers Estimation You can estimate design levels by establishing a relationship between observed high tide levels in the estuary and equivalent high tide levels at a nearby coastal location (for which definitive sea level / return period values should be available). The level to level relationship between the coastal and estuary points can then be used to estimate return period tide levels in the estuary. Modelling You can calculate levels through modelling, but this is only valid if you can calibrate it to observed levels. The hydraulics of estuary flow is more complex than for rivers, and the assumptions you make in river modelling are not all valid for estuaries.

Tide and surge The total tide curve for an event is a combination of: curves and their ƒ the astronomical tide (the tide caused by the gravitational effects of the combination moon and the sun and given in published tide tables); ƒ tidal surge (the additional elevation of the sea caused by weather conditions). Select a sequence of tides around a very high spring tide, even Highest Astronomical Tide (HAT), as the astronomical tide component. An overall duration of about five days is often needed. You can use software to calculate this, or you can select a suitable record out of tide gauge data. Note: we hold detailed astronomical tide predictions for 142 sites around the country. These are available on our National Flood Forecasting System and web service. The growth and decline of a tidal surge can be estimated using data from the nearest Class A tide gauge. For larger events the surge is likely to have duration in the range 36 – 60 hours. Producing a total tide curve In order to produce a total tide curve, add the surge shape to the astronomical tide, putting peak to peak, and scaling the surge height so the total peak sea level is as desired. Do not scale the surge duration. Surge curves suitable for design and assessment purposes will be an

outcome of a current R&D project.

Wave Types of wave conditions There are two classifications of waves: ƒ Wind waves: generated by a local storm and generally of period (time for one complete oscillation of a wave) up to 12 seconds; ƒ Swell waves: generated remotely over the wider ocean then running into the coastal waters. Swell has a longer wave period. Because it has a greater energy than wind waves, swell can be highly damaging to coastal structures. Sources of information Some sources of information on offshore and nearshore wave conditions are: Continued on next page…

Doc No 379_05 Version 2 Last printed 27/10/10 Page 20 of 31

Wave ƒ Shoreline Management Plans; conditions ƒ Met Office hindcast modelled data; continued ƒ Wave buoys (for example data from WaveNet, Met Office), though most do not have a long record; ƒ Swell atlas. Updated swell parameters for UK coastal waters will be with an output from a current R&D project. Raw wave data may need processing to estimate return period values of wave height and wave period. In doing this, consider which wave directions the site of interest is exposed to rather than lumping all the wave data together. For most places around the coastline there is a dominant wind wave direction, not necessarily being the most common wind direction, likely to apply coincident with the higher extreme sea levels; see FD 2308/TR1. Considerations when modelling ƒ If the only wave data available is for a point offshore, you need to transform the wave parameters to the nearshore for flood modelling purposes. Wave conditions can be converted from offshore to the coastline using a spectral wave model; ƒ 2D modelling includes for refraction, etc. as the waves move shoreward. Some models do not handle wave reflection; ƒ For good results in any wave transformation model you need good bathymetry, for example from Admiralty digital charts supplemented close to the shore by local survey, for example beach profiles or LIDAR (if flown at low tide); ƒ You can also estimate wind wave conditions at the coastline by considering potential wind speed, duration and fetch length; see BS 6349 or relevant software. This method is particularly useful for estuaries and enclosed waters; ƒ Since wave heights are limited by the depth of water available, it is not unusual for the wave height at the coastline to be “depth limited”. In this case the maximum wave that can reach the coast may be almost independent of the offshore return period wave height adopted. This could stop the need for wave transformation modelling; ƒ In general, you can consider the effect of wind waves and of swell waves separately. Also see joint probability.

Climate Contingency allowances for sea level rise in response to climate change are change given in Defra guidance and PPS 25 / TAN 15. These allowances apply to mean sea level. Tidal range is predicated to stay unchanged; hence you can represent future tide curves by shifting present-day curves upwards by the set amount. Storminess The latest UKCIP projections suggest that storminess will remain unchanged for the future. Thus there is no need to adjust currently estimated offshore wave heights, though waves reaching the coast can be higher because of the greater water depth with sea level rise.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 21 of 31

Joint Joint probability needs to consider how likely it is for the respective probability phenomena to occur together, not just in the same year or on the same day but at the same time. Phenomena with a short duration are less likely to be coincident than those of long duration. The “Norfolk method” is one way of allowing for duration in considering joint probability.

Joint Extreme tide levels are caused by severe weather, therefore they generally probability: are associated with strong winds and the notable wave action generated by tide/wave flow those winds. Thus there is usually a good correlation between the occurrence of extreme tide level and high wave action. Conversely, high winds and associated extreme waves can occur with modest tide levels. This situation is not normally of such interest as the extreme tide plus wave combination, in part because of the potential “capping” of wave height at the coastline due to depth limiting. Establish the actual position by site-specific assessment. FD 2308/TR1 gives advice on tide/wave correlation strengths and on the dominant storm direction around the coast. It also presents tables showing variable combinations of tide and wave return period for a range of joint return periods. As a caveat to the tables you should not accept small wave heights as a combination with extreme tide levels, for the physical reason outlined above. Research shows the occurrence of swell waves is independent of wind wave occurrence so their joint probability can be assessed from this standpoint.

Joint FD 2308/TR1 also gives advice on correlation between high river flow and probability: high tide levels. tide/river flow This advice is conservative as the studies only considered coincidence on the same day rather than at the same time. In practice correlation tends to be low since the causative weather patterns are often mutually exclusive. High catchment rainfall, giving high flows, tends to be associated with fairly static weather whereas extreme tide levels tend to be associated with highly mobile low pressure systems. Commonly it is either the extreme fluvial or the extreme tidal event, rather than some intermediate combination that will dominate flood extents. For these it should be sufficient to consider respectively: ƒ A high river flow with a mean high water spring tide; ƒ An extreme tide with QMED river flow. Modelling must, however, evaluate the circumstances of each location individually. The relative timing of high river flow and extreme sea levels can be assessed by comparison of respective gauge data. River flow in combination with a mean neap tide may need to be considered. During neaps the tide level does not fall as low as during spring tides. This impedes fluvial discharge capability at low tide, possibly giving the dominant flood risk scenario.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 22 of 31

Tidal Direct tidal overtopping will be calculated by considering weiring over the overtopping defence or other coastline feature as the tide level rises and falls through and breaching the event. A suitable boundary condition can be set up in the software being used. The EurOtop manual gives quantitative advice linking wave overtopping rates to potential damage at the coastal frontage. Also consider potential breaching of defences.

Wave Calculate wave overtopping using one of these options: overtopping ƒ methods in the EurOtop manual; ƒ methods in our own overtopping manual; ƒ specialist software. In each case experienced judgment is needed to assess whether the results are plausible, since no method is particularly reliable. In view of this, calibrate all wave overtopping calculations against some past experience, perhaps including “near miss” events when no significant overtopping occurred in spite of the prevailing wave action. Different calculation methods The EurOtop methods are suitable for frontages having a simple and fairly regular profile, or can be approximated to this. Calculation is possible via an on-line tool on. The EurOtop empirical method is particularly easy to use. For other than simple profiles, specialist software should be used as this will generally facilitate better representation of the frontage shape. In principle this should improve the quality of the results, though whether this achieved in practice is not certain as there is little calibration evidence to real-life experience. Source of information All methods need information on the frontage crest level and profile, extending to the beach foreshore. This information can be obtained from local surveys or high resolution LIDAR. Calculations The total volume of wave overtopping is found by considering overtopping rates at different sea levels through the expected rise and fall of the tide in the event, then integrating the answers. It is usually sufficient to consider the wave action as lasting for only 12-24 hours even if the total tidal event is longer. This is because the waves will then diminish as the storm moves and the wind changes direction. The extent of wave run-up can also be relevant, for instance in assessing whether properties are likely to be affected by wave water. EurOtop gives methods of calculation.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 23 of 31

Shingle Shingle beaches are mobile under wave action. In part this is helpful to flood protection as they tend initially to deform and create a higher ridge landward of the original crest. Further deformation can lead to collapse of this ridge, leading to overwashing of the frontage.

If … then … you can assess the potential for use the Bradbury method. overwashing to occur overwashing is found to be a estimate the new shingle profile potential problem after wave action using the Powell method.

Both methods lend themselves to a spreadsheet calculation. The methods are approximations only, so should not be relied upon for critical situations. Specialist software is available to estimate shingle or dune movement under wave action.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 24 of 31

6. Calibration, verification and sensitivity testing

Uncertainty In modelling flood risk there are uncertainties throughout the process, in the analysis input data, in the mathematical equations, in the modellers skills and in the outputs. We need to be open about the uncertainty involved in modelling and find ways to present this uncertainty to help people make more informed decisions. Uncertainty can be expressed through the results of sensitivity analysis. Whilst we can use probabilistic methods to help us understand and communicate uncertainty, most current modelling remains deterministic. Currently research is underway to understand how we can validate the outputs of probabilistic models and how we can re-use our existing detailed deterministic models in a probabilistic way. Once these two areas of research are delivered we will review this guidance to include more detail about using computational modelling to produce probabilistic outputs.

Calibration Wherever practical, calibrate the hydrological assessment and the hydraulic modelling against recorded flows and/or water levels and flood extent from observed flood events. The events need not have caused extensive flooding as it is also valid to show the model correctly predicts water not reaching particular areas. Availability of calibration data If calibration data is … then … available calibrate using at least three separate events. not available carry out a ‘reality check’ on the predicted flows, levels and water level profiles using photographs, historic information and anecdotal accounts of flooding.

Considerations ƒ Only vary the coefficients used in the calibration process within the possible ranges suggested in the standard textbooks. Consider flow and flood levels when calibrating steady-state modelling. Also consider the timing of the flood peak, flood volume and shape of the flood hydrograph when calibrating hydrodynamic models. ƒ In 2D modelling of floodplains consider whether the flood extents, depths and flow paths given by the model appear plausible when set against what is known of the area. ƒ Models for flood forecasting purposes require more emphasis on the timing at the rise in flood level than is generally needed in models for flood risk assessment purposes only. ƒ Target accuracies for calibration are provided in the SFRM specification.

Verification After calibration, run one or more separate observed events through the model to verify the adjustment of parameters.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 25 of 31

Sensitivity Test the modelling outputs by adjusting key parameters within the model. testing The aim here is to assess the possible circumstances that could cause flooding to be significantly more severe than the modelled best estimate. Adjusted parameters should include ƒ model inflows; ƒ downstream boundary condition; ƒ channel and floodplain roughness; ƒ key structure coefficients. Reflect uncertainties, possible changes due to climate change and variations in hydraulic coefficients (for example from seasonal changes or periodic maintenance) in the range of parameters used in sensitivity tests. Test sensitivity to blockage of critical structures.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 26 of 31

7. Mapping and reporting

Contents This chapter provides advice on the presentation of modelling outputs, mapping and reporting:

Topic See page Mapping 27 Report content 28 Report format 28 Data 28 Future use 28

Quality assurance and audit trail 28

Mapping Mapping of flood extents and other parameters (for example flood depth), is a direct output from 2D modelling. With 1D modelling it is generally made by projecting water levels over the DTM at each cross-section and interpolating between these. There are potential shortcomings in projecting water levels at sections where the water level on the floodplain and the water level in the river are not well connected. It is important that appropriate engineering judgement combined with suitable GIS techniques should be applied in mapping and flood extent for these situations (for example mapping bypass flows). Isolated dry areas In each case, isolated dry areas (“dry islands”) may exist within the overall flood extent. In accordance with national guidance, dry islands of less than 200m2 in size should be removed from the mapping (infilled). Isolated wet areas Similarly, isolated wet areas may be shown beyond the general flood extent. You should consider whether the flood water would actually reach the remote area or whether it is only an inadvertent product of water level projection. In the latter case the isolated wet area should be removed entirely. If the isolated wet area is plausible it should remain in the mapping, though isolated wet areas of less than 200m2 in size should be removed as being inconsequential. Tidying up It may be appropriate to tidy up flood outlines, usually though a mixture of automated and manual routines. In some cases the mapped outputs may imply that the flood outline in large because there are areas of very shallow water in the model outputs. Given the uncertainties in the modelling, we don’t have much confidence in the very shallow depths being realistic. It is advisable to remove all depths less than 0.1m because this gives (by eye) a clearer representation of the areas that are flooded (so there is more focus on the areas flooded to a greater depth).

Doc No 379_05 Version 2 Last printed 27/10/10 Page 27 of 31

Report Write a report describing the modelling so that the model structure and content results can be evaluated. It should be a self-contained report that will provide sufficient information to allow us to use the model in the future, including enough detail should we need to replicate the work. The detail of the report should be appropriate to the complexity of the modelling. The SFRM performance specification (available from www.sfrm.co.uk) details exactly what is expected from the report. This specification must be used by any consultant who carries out a project involving modelling for us.

Report format The report must be easy to copy and transmit electronically, and must include plans and schematics on an appropriate scale mapping backdrop. All relevant features, structures and watercourses shall be shown and named. Adobe pdf files are preferred for the report.

Data ƒ Copies of the model data files together in an appropriate format (not Adobe pdf) with sufficient instructions to run and view the models, for example a text file containing start time, finish time, time-step, runtime, information on non-default parameters etc. ƒ Include initial condition files. ƒ Flood level nodal data, and other data, required for NFCDD. ƒ Copies of flood outlines and other required modelled outputs in GIS format.

Future use Write a statement to accompany the report and the model data on the allowable future uses of the model and its associated documentation. This is described in the SFRM performance specification in more detail (available from September 2010) Complete the metadata (refer to 199_07 Flood Risk Management Metadata Standards).

Quality Throughout the study, define and report on an audit trail. assurance Include all relevant documentation and a link with the appropriate quality and audit trail assurance procedures of the organisation carrying out the study. Make sure the relevant documentation is available to others who may use the modelling inputs and outputs in future.

Doc No 379_05 Version 2 Last printed 27/10/10 Page 28 of 31

Related documents

Strategies ƒ FCRM Modelling Strategy, FCRM Data Strategy, FCRM Risk Mapping and policies Strategy. Visit our publications catalogue and enter the words “Modelling Strategy”, “Data Strategy” or “Risk Mapping Strategy” in the publication title field of the search. ƒ Risk Management: A Risk-Based Approach (261_05)

Modelling Specifications approach ƒ Flood Mapping Specification for the Strategic Flood Risk Management Framework (www.sfrm.co.uk); ƒ SFRM performance specification (available from September 2010); R&D reports ƒ Fluvial Freeboard Guidance Note (W187) (visit our publications catalogue and type “fluvial freeboard” in the publications search); ƒ Joint probability: dependence mapping & best practice, FD 2308/TR1. HR Wallingford. 2003. (visit the FCERM evidence web page and type “FD2308” into the search, select the research project option before clicking on search) ƒ Wave overtopping of seawalls. Design and assessment manual. (W178). February 1999. (visit our publications catalogue and type “overtopping of seawalls” in the publications search); ƒ FRMRC Research Outcomes in the Application of 2D Flood Inundation Models for Flood Risk Management. 2008. (available by contacting Helen James); Internal guidance ƒ 145_07 Real-time model development for flood forecasting ƒ 229_06 Provision and fitness for purpose of the National Generalised Modelling (JFLOW/HYDROF) including climate change depth difference data ƒ 303_09 Flood Risk Management: Strategic Flood Consequence Assessments for Wales ƒ DRAFT Dry islands on the Flood Map/Flood Zones, 2006 External guidance ƒ CIRIA Report C624, Development and Flood Risk – Guidance for the Construction Industry, CIRIA, London 2004 (available from the CIRIA website www.ciria.org, search their bookshop (enter C624 into their search engine); ƒ WaPUG guides http://www.ciwem.org/groups/wapug/modelling.asp ƒ BS6349 Maritime Structures – Part 1: Code of Practice for General Criteria (visit the BSI shop http://shop.bsigroup.com/en/ and enter BS6349 into their search engine); ƒ Beach management manual. CIRIA Report 153, 1996. ƒ Defra FCDPAG3 Economic Appraisal: Supplementary note to Operating Authorities – climate change impacts. October 2006. ƒ EurOtop. Wave overtopping of sea defences and related structure: Assessment manual, August 2007. An on-line calculation tool is available at www.overtopping-manual.com/calculation_tool.html

Doc No 379_05 Version 2 Last printed 27/10/10 Page 29 of 31

Modelling ƒ The Norfolk Method was originated by Mantz and Wakeling (ICE, approach, 1979). The general expression is given in “Tidal Flood Risk Areas – continued Simply Credible. Worth and Cox, 35th MAFF Conference of River and Coastal Engineers, 2000.”

ƒ Bradbury, “Predicting Breaching of Shingle Barrier Beaches – Recent Advances to Aid Beach Management” (35th MAFF Conference of Coastal and River Engineers, 2000). ƒ Powell, “Predicting short term profile response for shingle beaches” HR Wallingford Report SR 210, February 1990.

Data ƒ 199_07 Flood Risk Management Metadata Standards ƒ 197_08 Flood Estimation Guidelines ƒ UK Hiflows project ƒ 687_06 Data auditing – guidance for flood risk mapping and data management teams ƒ 183_05 Data management for Flood Risk Management projects and good data management considerations ƒ Defra/Environment Agency R&D Project SC060064: Development and dissemination of information on coastal and estuary extremes (visit the FCERM evidence web page and type “SC060064” into the search, select the research project option before clicking on search) ƒ Proudman Oceanographic Laboratory Internal Document No.112 – Spatial Analyses for the UK Coast. Dixon and Tawn, June 2007 ƒ Class A tide gauge data is available for download from www.bodc.ac.uk ƒ WaveNet data is available from www.cefas.co.uk/data.aspx ƒ Swell and bi-model wave climate around the Coast of England and Wales. HR Wallingford Report SR 409, November 1997

Software ƒ R&D Report ‘Benchmarking of hydraulic river modelling software packages’ (W5-105) – visit our publications catalogue and enter the word “benchmarking” in publication title field of the search ƒ 2D benchmarking – visit our publications catalogue and enter the word “2D benchmarking” in publication title field of the search

Survey ƒ Environment Agency National Survey Specification is available from O:\National Flood Mapping\Survey Specification ƒ Refer to Assessment of Flood Risk – Hydrographic and Topographic Survey. For further information on the appropriateness of survey, you should refer to the operational instruction The Preparation of Survey data for Flood Risk Assessments (195_05) ƒ Royal Institute of Chartered Surveyors http://www.rics.org/guidance

Doc No 379_05 Version 2 Last printed 27/10/10 Page 30 of 31

Hydraulic ƒ Chow (Ven Te Chow, Open Channel Hydraulics, McGraw-Hill 1959) Coefficients and Hicks & Mason (Roughness Characteristics of New Zealand Rivers. D.M.Hicks & P.D.Mason. 1999) can provide some guidance. ƒ Information on roughness can also be obtained from the Defra/Environment Agency Conveyance Estimation System (CES) – http://www.river-conveyance.net/ ƒ R&D Project W5-110, Afflux Estimation System, EA/Defra science project SC030218, 2007 (visit the FCERM evidence web page and type “SC030218” into the search, select the research project option before clicking on search)

Intellectual ƒ 1139_08 Intellectual Property Property ƒ Intellectual Property e learning course ƒ 437_07 Use of 3rd party Intellectual Property from Flood Risk / Consequence Assessments ƒ 213_05 How do I add, update or delete an entry in the Information Asset Register?

Doc No 379_05 Version 2 Last printed 27/10/10 Page 31 of 31