The FEH22 rainfall model is the FEH’s latest UK-wide statistical model for rainfall depth-duration-frequency (DDF) estimation. It supersedes the FEH13 model and the legacy FEH99 model:

  • The FEH22 model was released via the FEH Web Service in 2022 and is based on rain gauge records up to 2020.
  • The FEH13 model was released via the FEH Web Service in 2015 and is based on rain gauge records up to about 2005.
  • The legacy FEH99 model was released via the FEH CD-ROM in 1999.

The DDF models can be used to derive design storm inputs for the ReFH2 rainfall-runoff package or to estimate the rarity of observed rainfall events at individual points or over a catchment. The DDF model is widely used for applications such as flood risk assessment, reservoir flood safety appraisal and drainage design throughout the UK.

The FEH22 and FEH13 models retained the basic index-flood approach of the FEH99 model, i.e. the rainfall frequency curve is derived by multiplying an index variable specific to the site of interest by a regionally derived growth curve, but the main advances are:

  • increased availability of rainfall maxima, particularly for sub-daily durations;
  • a revised standardisation that uses average annual rainfall (SAAR) and northing in addition to the index variable RMED (the median annual maximum rainfall) to remove more of the location-dependent variation in rainfall before combining maxima from networks of rain gauges;
  • a revised spatial dependence model;
  • improvements to the FEH FORGEX method of deriving rainfall growth curves using annual maxima pooled from increasingly wider circles as return period increases;
  • a more flexible depth-duration-frequency (DDF) model structure.

A key feature of the modelling procedures is the use of local data wherever possible. Data from smaller networks of rain gauges are used to estimate the frequency curve for shorter return periods, while more spatially extensive networks are drawn into the analysis to define the frequency characteristics for longer return periods. The FEH22 and FEH13 DDF models were fitted to the outputs from the revised FORGEX procedure at each point on a 1 km grid of the UK for a number of key durations. The model ensures internal consistency in the resulting frequency estimates, i.e. that rainfall depths for any duration increase with increasing return period, and that rainfall depths for any return period increase with increasing duration. It also allows extrapolation to longer return periods than those estimated for the rarest events in the calibration dataset. The model is based on a weighted total of two gamma distributions, each of which is parameterised by duration-dependent equations intended to minimise the difference between the fitted model curves and the output FORGEX lines. The final step in the modelling procedure is smoothing in space and across durations to avoid inconsistencies in the results.

Climate change and potential non-stationarity in gauged data

The development of the FEH22 and FEH13 DDF models made maximum use of the daily and sub daily rainfall data available across the UK and thus don’t conform to a specific period of record. FEH22 maintains the existing depth-duration-frequency (DDF) model framework of FEH13 that is based on stationary assumptions of the rainfall data and statistics underpinning the method. These assume no temporal trends in the data being used. Changes in the rainfall regime over the additional years of data may however have occurred. The most recent Government climate change risk assessment (CCRA3) points to some evidence for recent increases in occurrence of extreme UK rainfall, but states that records are too short for evidence of trends to be conclusive (Slingo, 2021). Data from HadUK-Grid observations suggests significant increases, in wetter seasonal extremes in particular, have occurred (Cotterill et al., 2021), however the underlying dataset is based only on daily rainfall and sub-daily extremes occurring at particular gauges is not the focus. FEH22 makes no explicit account of such potential changes in recent rainfall and how this could affect the design rainfall estimates provides. However, by retaining the model structure and proving significantly more events and recent data this improves the confidence in the model outputs capturing any recent climate shifts and serves as an improved baseline for flood risk assessment. The uncertainty in including any recent shifts, which clearly remains uncertain, is fundamentally outweighed by the improvement gained through significant increases in gauged data. This is important to note as although many areas show a decrease in rainfall estimates in FEH22 over FEH13, this should not be interpreted to necessarily indicate a change in climatic conditions. The changes are primarily due to the addition of extra data and improved quality assurance removing erroneous data. Future work will set out to explore the effects of recent climate shifts on extreme rainfall, how this can be more explicitly represented in the rainfall model, and how climate change allowances that account for variable baseline data and design storm estimates can be systematically applied.

References