Estimating design flows and rainfall for flood risk modelling

This guidance outlines Natural Resources Wales (NRW) recommendations on appropriate methodologies for flood frequency estimation in Wales, particularly in respect of flood risk associated with development proposals and flood hydrology calculation for flood risk mapping/modelling and post event analysis.

All flood estimation comes with a degree of uncertainty and the responsibility to ensure that the estimated flows are fit for the intended purpose remains that of the analyst, based on a consideration of the available data, information and uncertainty.

Flood estimate submissions to NRW may be rejected if they are not clearly presented. We recommend you use our Flood Estimation Calculation Record template, which prompts analysts to document key information and decisions.

1. Background

The Flood Estimation Handbook (FEH) and supplementary reports contain the industry standard methodology for flood frequency estimation and rainfall-runoff modelling in the UK. The two principal techniques are the FEH Statistical method and the Revitalised Flood Hydrograph method (ReFH).

2. Method selection 

Our preferred approach for estimating peak flows up to and including the 1% Annual Exceedance Probability (AEP) (100-year return period) event, is the FEH Statistical 2025 method, using the latest catchment descriptors; SAAR9120, FARL2015, URBEXT2015, BFIHOST19Scaled. Find information on the research and development of the method and catchment descriptors in:

The FEH 2025 statistical method update (UK Centre for Ecology & Hydrology)

New FEH Catchment Descriptors 2025 (UK Centre for Ecology & Hydrology)

Peak flows for rarer events should be estimated using the flood growth curve from ReFH2 applied to the 1% AEP estimate from the FEH Statistical method. For example, to derive the 0.1% AEP estimate (1000-year return period), the ratio of the 0.1% to 1% AEP ReFH2 estimates, is multiplied by the 1% AEP estimate from the FEH Statistical method. The resulting flow estimates should be compared to absolute values calculated by FEH Statistical and ReFH2 methods.

Urban results should be applied by default in both the FEH Statistical and ReFH2 methods regardless of the extent of urbanisation within the catchment.

There may be occasions when it is appropriate to use either the FEH Statistical method or ReFH2 method for all return periods. In such cases, the onus will be on the analyst to justify which is the more appropriate method. Choosing one methodology over another simply because it provides higher or lower estimates is not appropriate.

An updated FEH rainfall Depth Duration Frequency model known as FEH22 was published in 2022 based on rainfall data up to 2020. It is selected by default in ReFH2, and we expect it to be used for peak flow and hydrograph calculation. Find information on the method and catchment descriptors in:

The FEH22 rainfall depth duration-frequency (DDF) model (UK Centre for Ecology & Hydrology)

Local and/or historical flood data should be used where available to improve peak flow and hydrograph estimates. This is especially important for catchments that are urbanised or contain reservoirs/lakes (where reservoir routing should be used if appropriate) or other unusual local features. For the very highest risk developments, flood history research should be undertaken and trend analysis carried out at local gauging stations and reasons for any significant trends identified and addressed appropriately.

Alternative methods may be used provided their use is appropriate and can be justified for the subject catchment and the methods are proven accepted within the hydrological community. Guidance on the applicability of alternative methods such as continuous simulation and direct rainfall is provided in the Environment Agency’s Flood Estimation Guidelines and is not repeated here.

3. Software and data versions

We recommend that analyses use the latest versions of:

  • WINFAP for the FEH Statistical method, so that the new features and outputs from research projects can be easily applied
  • ReFH2 for the Revitalised Flood Hydrograph method with the FEH2022 rainfall model
  • NRFA Peak Flows Dataset

Use of up-to-date FEH methods within hydraulic modelling packages (for example Flood Modeller) or other open-source environments such as R (for example UKFE R package developed by Hammond, 2020) are also acceptable.

We may not approve estimates that use earlier versions, or which do not include recent developments and appropriate methodologies and data.

We expect updated/new versions /datasets to be used within one year of publication or sensitivity testing to have been carried out to demonstrate that there is no significant impact from the update/updated version.

4. Uncertainty

Analysts should provide an estimation of uncertainty in any estimated flows and/or hydrographs so that it can be considered when applying flows in analyses, hydraulic modelling and other subsequent decision making. Find information on calculating uncertainty in:

Uncertainty in FEH methods: a short guide for users (UK Centre for Ecology & Hydrology

Making better use of data in flood frequency estimation (GOV.UK).


5. FEH Statistical method

5.1 Adjusting the median annual maximum flood (QMED) using local data

The FEH Statistical 2025 method research confirmed the previous finding that there is a spatial pattern to the QMED catchment descriptor equation model residuals, therefore adjusting catchment descriptor QMED using local donors is an important part of the QMED estimation methodology.

When adjusting catchment descriptor QMED using data from donor gauging stations, WINFAP 5.3 defaults to using eight stations, as recommended in the 2025 research. The selection of the eight stations is based purely on geographical distance from the subject site and takes no account of local conditions or other factors. This conclusion supersedes the recommendation from previous research studies which recommended smaller numbers of donors.

This recommendation is based on the average improvement in QMED across the entire station dataset. Therefore, analysts should not simply apply the closest eight donors by default but should investigate donors more thoroughly before selection. Analysts should consider local factors, which were outside of the scope of the research:

  • significant hydrological differences between the subject site and donor site, such as reservoir presence/absence or flood plain inundation at QMED
  • relative location on the river network – upstream or downstream stations should be favoured over neighbouring catchments
  • gauged data quality and record length
  • similarities/differences in QMED factors at surrounding stations – determining reasons for them

In many cases just one or two stations will be of sufficient quality and proximity to the subject site to be preferred. In addition, where donor(s) are on the same watercourse as the subject site, consideration should be given to omitting the distance moderation part of the donor adjustment calculation to avoid discontinuity in adjusted QMED or unfeasible QMED values compared to gauged flows up and/or downstream. There is no option to omit the distance moderation in WINFAP, so the QMED donor factor(s) should be applied to the catchment descriptor QMED manually – with appropriate weighting if more than one station is used.

In WINFAP, the catchment descriptor QMED can be adjusted using urbanised donor stations by de-urbanising the gauged QMED. Whilst it is preferable to use rural stations as QMED donors, there will be some locations where it is appropriate to adjust the QMED using more local urban donors rather than more distant rural donors.

5.2 Flood growth curve estimation

WINFAP allows analysts to include urbanised stations in pooling groups and enhanced single site analysis (ESS), by de-urbanising flood growth curves. Although this de-urbanisation is not based on published research, it is the reverse of the urban adjustment procedure for growth curves applied in WINFAP. Therefore, it is a reasonable approach which will increase the number of stations from which the pooling and ESS groups are drawn. This will result in a pooling group which (apart from urbanisation) has catchments that are more similar to the subject site.

Analysts should change the pooling group urban threshold from the default value of 0.03 to 0.3 (even if the subject site is not urbanised). This will mean the pooling group will include all but the most urbanised stations (where the uncertainty in the effects of urbanisation are higher, due to the small sample size and greater potential for effects of local features). As with other catchment characteristics, it is possible that the urban catchments will have local features which mean it is not appropriate to use them in pooling or ESS groups, so as always, the pooling group should be reviewed.

The FEH Statistical 2025 method supersedes past research and one method of defining pooling groups is now applicable to catchments of all sizes as applied in WINFAP 5.3.

5.3 Non-stationarity and trend analysis

Long term AMAX trend analysis results are available in WINFAP 5.1 and later versions as well as the UKCEH NRFA Flow Trend Explorer.

These are presented for information only and it is not expected that stations will be removed from pooling groups or as QMED donors simply because they show a positive or negative trend. For subject sites which are close to a gauging station, any trend should be investigated and where possible the cause of the trend determined using local information and comparison to nearby gauging stations and rainfall data.

Professional judgement should be used to determine the appropriate calculation methods and any change to standard methods must be fully justified and documented. Care must be taken regarding periods of record used, consistency with other sources of information, data quality and application of climate change allowances.

5.4 Non-flood years

Non-flood years are defined in the FEH as years where the AMAX is less than 50% of QMED. Originally the adjustment to remove these years was known as ‘permeable catchment adjustment’, as it was understood that these years predominantly occurred in groundwater dominated catchments with a high BFI. However, more recent work has shown that they also occur in low SAAR catchments (<800mm). Find more information in:

WINFAP Technical Guide (Wallingford HydroSolutions)

The EA have determined the following guidelines for the use of the non-flood year adjustment, and we agree with their approach:

  • For single site analysis and enhanced single site analysis, the recommendation is that the non-flood year adjustment is applied if more than 5% of the AMAX record at the subject site is non-flood years.
  • For pooled analysis at ungauged sites, the recommendation is that the non-flood year adjustment is applied to any stations in the pooling group that have more than 15% of AMAX records as non-flood years.

However, a high number of non-flood years may indicate an untypical catchment response, so stations should be reviewed to determine if removal from pooling group is appropriate.

The non-flood year adjustment can be implemented in WINFAP (Version 5.3 and later) or using the UKFE R package.

6. Urban catchments

There is currently insufficient evidence to determine whether ReFH2 or the FEH Statistical method is the most robust method of flood estimation in urban catchments. Alternative rainfall-runoff approaches such as direct rainfall may also be considered in heavily urbanised catchments.

It is still recommended that peak flows in urban catchments are estimated using FEH Statistical, ReFH2 and direct rainfall method if applicable, and the results compared. If the results are not in broad comparison, then the analyst should apply professional judgement, drawing on local flood history to determine which method is most appropriate.

7. Climate change allowance

An allowance for climate change will need to be applied to the hydrology when undertaking hydraulic modelling to understand the potential impacts on future flood risk.  This remains the case even when using SAAR9120.

The Welsh Government’s Flood and coastal erosion risk management: adapting to climate change gives current allowances for climate change.

8. Reservoir safety

The Environment Agency Flood Estimation Guidelines outline the current methods for estimating flows for reservoir safety. We recommend that FEH22 is used in preference to FEH13, as there are known errors within some of the data used to derive the FEH13 dataset.

However, analysts may wish to compare rainfall estimates from FEH22 against FEH13. While a precautionary approach may be preferred by analysts, we recommend that any significant differences between rainfall data, particularly where FEH13 produces higher estimates than FEH22, should be investigated to determine whether this is due to the erroneous data used in FEH13.

The Extreme Rainfall Estimation Improvements Phase 1 FCERM R&D project includes sensitivity analysis on the effects of erroneous data on DDF curves. Reporting is due in 2026. 

9. Hydrographs

The industry standard rainfall-runoff model is ReFH2. It is our preferred method, and the latest version should be used with the latest rainfall model; currently ReFH2.4 and FEH22. Find more information in:

ReFH Technical Guide (Wallingford HydroSolutions)

Alternatively, hydrograph shape can be taken directly from gauged data by averaging a number of significant (greater than QMED), single peaked events which have been standardised by peak flow. The resulting hydrograph can then be scaled to the design peak flow estimate. This method is usually only used in catchments with an atypical response.

10. Calibration 

For calculations at, or close to flow gauging stations, the ReFH Calibration Utility can be used to estimate the ReFH parameters using gauged river flow, rainfall and potential evaporation data. Find more information in:

ReFH 2 Calibration Utility (Wallingford HydroSolutions)

Observed rainfall can then be used in ReFH2 to compare modelled hydrographs (using original and calibrated Tp) to gauged hydrographs. However, as ReFH hydrographs are often scaled using FEH Statistical estimates, it is generally appropriate (and less time consuming) to calibrate Tp using Lag time analysis (note that river stage, as well as flow timeseries data can be used).

Due to a change in Tp definition between the original FEH Rainfall-Runoff (RR) Method and ReFH2, calibration using Lag time must be done via FEH RR Tp adjustment i.e. calculate the calibration factor from ‘FEH RR catchment descriptor Tp’ to ‘Tp calculated from lag-time’ and apply that same factor to the ‘ReFH catchment descriptor Tp’. Note that particular care is needed on heavily urbanised catchments, due to urban and rural hydrograph components being calculated separately in ReFH.

Research carried out in 2025 by JBA Consulting used published values of gauging station Tp and Lag time to produce a regression equation to calculate ReFH Tp from Lag time. However, the dataset only consisted of 57 gauging stations across England and Wales and some catchment types such as small catchments were not represented. Caution is needed if applying the equation in practice.

11. Design storm profile - season

The Small Catchment Phase 2 research (SC090031) reviewed recommendations for seasonal inputs to ReFH for different catchment types. These were found to be scale independent so apply to all catchment sizes. For convenience these are reproduced here:

  • If URBEXT2000 < 0.15, winter season should be used.
  • If 0.15 ≤ URBEXT2000 < 0.3. Winter season is recommended unless the catchment is dry (SAAR < 800 mm) and permeable (BFIHOST ≥ 0.65) in which case summer should be used. The Tp scaling factor should be increased to 1 as there is no evidence for enhanced routing of urban runoff in moderately urbanised catchments.
  • If URBEXT2000 ≥ 0.3, use summer design rainfall events (rain profile and Cini).

Further information can be found in:

Review of methodology for estimating flood peaks and hydrographs for small catchments (GOV.UK)

We recommend that urban results are used regardless of the extent of urbanisation at the subject site(s) for consistency.

Analysts should ensure that the correct season and Tp scaling factor are used.

The ReFH2 storm profile principles described above also apply to pluvial/surface water and/or direct rainfall modelling.

12. Storm duration

Analysts are encouraged to work with hydraulic modellers to determine the most appropriate storm duration(s) to apply in the model. Design hydrograph estimation should be an iterative approach with hydraulic modelling, where estimates should be reviewed and updated as appropriate during the model calibration and verification exercise.

The ‘Recommended’ Design Storm Duration is calculated in ReFH as function of catchment characteristics; this is the storm duration which was used in the model calibration and has been confirmed as appropriate in the Small Catchment Phase 2 research. It should be used for hydrograph calculation for hydraulic models where a single inflow point is needed. For distributed models with multiple inflow points, the conventional approach is to run hydrographs from two or more Recommended Storm Durations, each being appropriate to a watercourse/location, across the whole modelled area.

Hydraulic modellers may test additional storm durations around the recommended duration(s) to determine whether a longer, but lower peaked hydrograph, results in greater flood extent due to larger flood volume. Adjustments will be needed to ensure that the longer duration hydrograph calculated by ReFH does not have a higher peak than the best estimate peak flow.

The ‘Recommended’ Storm Duration rarely results in a peak flow which is the highest of those calculated using a range of durations – the duration which does this is termed the ‘Critical’ Storm Duration2.  Note that there is a distinction between the hydrological Critical Storm Duration described here, and the hydraulic model Critical Storm Duration. The latter, being the duration that produces the most extreme flood extent.

Whilst the closure of the water balance, introduced in ReFH2.3, has reduced the difference between the Recommended and Critical durations, the Critical duration is normally several hours longer than the Recommended (average of 6 hours or 27% longer based on work by Wallingford HydroSolutions presented at the British Hydrological Symposium in 2022).

This difference between Recommended and Critical durations, along with the convention of using the same storm duration across whole catchments in distributed modelling leads to two issues:

  1. a longer than recommended storm duration would be expected to reduce ReFH peak flow, but it is generally increased. A common approach is to fit all hydrographs to the preferred design peak flows. This is a rather blunt approach, but it does reduce the overestimation of flood peaks/volume. However, it conversely may result in overestimation on shorter duration hydrographs. Until this issue in ReFH is addressed, it is a pragmatic solution. An alternative is to scale hydrographs using a factor calculated from 1% ReFH event peak to 1% FEH Statistical peak, applied to all duration hydrographs, but capped so that the peak flow does not exceed the 1% FEH Statistical event.
  2. the storm duration is only appropriate for one part of the catchment, however there are usually several places of interest.

There may be several ways to address these issues, particularly for studies with a disparity in catchment sizes of modelled watercourses. For example, break the modelled area down into several smaller areas; or use of recommended durations at each location throughout the catchment with inflows from intervening areas adjusted to facilitate this. Use of this approach should be carefully considered and may not be applicable on all catchment types. Research into this methodology or other approaches is needed, but until this becomes available, the conventional approach of applying different durations across several model runs is acceptable.

The ReFH2 storm duration principles described above also apply to pluvial/surface water and/or direct rainfall modelling.

Updates

July 2026

Published as web content

January 2026

Updated following publication of important updates to key flood estimation data and methodologies: 

  • Updated FEH method catchment descriptors; notably
    • SAAR9120 based on Met Office’s HadUK grids
    • FARL2015 based on UKCEH’s Land Cover Map 2015
    • URBEXT2015, based on LCM2015
  • Updates to QMED and Pooling methodology (referred to as FEH Statistical 2025) 
  • Updates to WINFAP software (WINFAP 5.3)

We have reviewed the new methodologies and have several recommendations regarding their use in Wales. 

November 2024

Updated following publication of FEH22 Depth Duration Frequency model, non-stationarity flood estimation research and guidance, small catchment flood hydrology phase 2 research reports, Greenfield Runoff Screening Tool and updates to ReFH2 and WINFAP software.

November 2021

Updated following release of BFIHOST19, Small Catchment Flood Hydrology Phase 2 Research and updates to ReFH2 and WINFAP methodology and software. Addition of uncertainty and hydrograph estimation sections.

December 2017

Published, superseding Flood Estimation Good Practice Guide (GPG102).

Last updated