Scientific background

Flood Estimation Handbook

The methodologies used by OH Auto Statistical are directly based on the Flood Estimation Handbook (FEH), the standard in the United Kingdom for flood flow estimation. The statistical methods for estimating flood flows in the FEH were first revised in 2008 and additional improvements were subsequently published between 2010 and 2014. These latest revisions to the FEH statistical methods are fully incorporated into OH Auto Statistical.

The revisions to the FEH statistical methods are described in the following papers:

  • Kjeldsen, T. R., Jones, D. A. & Bayliss, A. C. (2008). Improving the FEH statistical procedures for flood frequency estimation (No. SC050050). Bristol: Environment Agency
  • Kjeldsen, T. R. (2010). Modelling the impact of urbanization on flood frequency relationships in the UK. Hydrology Research, 41 (5). pp. 391‒405 doi:10.2166/nh.2010.056
  • Kjeldsen, T. R., Jones, D. A. & Morris, D. G. (2014). Using multiple donor sites for enhanced flood estimation in ungauged catchments. Water Resour. Res., 50, pp. 6646‒6657 doi:10.1002/2013WR015203.

Note

Why a fully automated implementation?

The FEH recommends and other publications recommend that hydrologist should use expert judgement to improve flood estimates where possible. For example, the most suitable QMED or growth curve donor catchment should be decided based on catchment similarity, local knowledge, other datasets etc. However, as practicioners, hydrologist often do not have access to other data or information not already included in the National River Flow Archive (NRFA). In absence of any additional data, there is a danger that subjective decisions are made which could potentially render the estimates less accurate that purely based on the calibrated national methodologies.

OH Auto Statistical purely follows FEH methodologies taking into account the latest NRFA datasets.

Assumptions

The following assumptions are made by OH Auto Statistical:

  • NRFA catchment data are assumed correct, including their indicated suitability for QMED estimation and growth curve pooling.
  • QMED adjustment using donors always includes the URBEXT parameter (known as “urban adjustment”) to estimate QMED from catchment descriptors. Both for the donor (may include urban catchments) and the subject location.
  • QMED donor requirements:
    • Suitable for QMED estimation
    • Same country (as in: UK mainland versus Northern Ireland).
    • At least 10 years of data
    • Not more than 500 km from the subject location.
  • Growth curve donor requirements:
    • Suitable for pooling
    • URBEXT2000 less than 0.03, if provided.
    • At least 10 year of data
  • Pooled growth curve donors have at least 500 years of data in total.
  • For gauged catchments analysis, the catchment is always used to estimated QMED and is always used in the pooling group regardless whether the catchment is flagged as being suitable for QMED or growth curve donor analyses.
  • The Generalised Logistic statistical distribution is used to create a frequency curve. Future versions of OH Auto Statistical will choose the best fitting distribution.