Regionalised impacts of climate change on flood flows: regionalising the flood response types in Britain. Milestone report 4. Revised November 2009

The primary objective of this project is to assess the suitability of current FCDPAG3 guidance given the advances in climate change science since its publication. PAG3 requires an allowance of 20% to be added to peak flows for any period between 2025 and 2115 for any location across Britain. This guidance was considered a precautionary value and its derivation reflected the evidence available at that time. FD2020 has been designed to increase this evidence base, and it is anticipated that the research will lead to the development of regional, rather than national, guidelines for changes to peak flows due to climate change. A scenario-neutral approach based on a broad sensitivity analysis to determine catchment response to changes in climate as chosen for FD2020. The method separates the climate change that a catchment may be exposed to (the hazard) from the catchment response (change in peak flows) to changes in the climate (the vulnerability). By combining current understanding of climate change likelihood (the ‘hazard’) with the vulnerability of a given catchment, it is possible to evaluate the risk of flood flow changes. The vulnerability of a catchment is to be characterised in two steps: first, the response of a set of catchments to a range of climatic changes are modelled, then analysed for similarity, and second the main responses are characterised according to catchment properties. This is possible by defining a sensitivity framework of changes to the mean and seasonality of precipitation and temperature and modelling the response of each catchment within this fixed framework. This milestone report describes the second step of the vulnerability assessment. This is achieved by identifying the relationships identified between a catchment’s characteristics (geographic, geologic or climatic) and the vulnerability of its flood peak to changes in the climate. The work follows the identification of nine flood response types for catchments in Britain, after a comprehensive ‘scenario-neutral’ sensitivity study based on 4,200 patterns of changes in rainfall, temperature and potential evaporation. These nine flood response types were found to fully describe the range of changes in flood peak obtained in 154 catchments, and represent five main families of behaviour from the most ‘damping’ (low vulnerability), through ‘neutral’, to the most ‘enhancing’ (high vulnerability) catchments. One of the response types, with a very damped response to changes in climate, was removed from the analysis, as the group was too small for a reliable model to be built; leaving eight flood response types to characterise. Using a hierarchical partitioning technique and digital catchment descriptors from the Flood Estimation Handbook and the Hydrometric Register databases, decision trees were identified to discriminate the flood response type from nine descriptors including mean annual rainfall, area, northing and easting, elevation, and measures of permeability and catchment losses. At the 2-year return period level, all eight flood response types could be discriminated. For changes in the 20- and 50-year return period floods, the flood response types had to be merged into four main categories before they could be discriminated by the catchment characteristics. This merging was also necessary to ensure that uncertainty due to the impact of seasonality in rainfall change was fully incorporated into the flood response types. For the most enhancing catchments (i.e. where the changes in flood peak are proportionally much greater than the maximum increases in rainfall), the difference between the mean annual rainfall and the losses in the catchment was found to be an important discriminatory factor. For changes in higher return period floods, mean annual rainfall was found to be less critical. Wetter catchments were found to be in general less enhancing than drier catchments. The decision trees were successful for between 67.5% and 84% of the study catchments, depending on the flood indicator. Amongst the misclassified catchments, a larger proportion was misclassified as more enhancing, resulting in a potential over-estimation of changes in flood peaks, or an over-precautionary assessment. When evaluating the ability to discriminate between the more general families of ‘resilient/damping catchments’ (i.e. associated with a damped flood response type), ‘neutral catchments’ and ‘vulnerable/enhancing catchments’ (i.e. associated with an enhanced response type), 80% of the catchments were found to be correctly classified across all four flood indicators. Large catchments seem to be slightly more difficult to classify, suggesting they might not be well represented by single value descriptors which smooth out spatial variations important in the response of the river to climatic changes. Following the decision trees (sets of partitioning rules and paths for each of the flood response types), it is possible to quickly identify, for any catchment (gauged or ungauged but with available descriptors), the expected flood response type in response to climate change. This regionalised vulnerability assessment can be used in combination with an evaluation of potential climatic changes (or the hazard) to provide a measure of the risk of changes in flood peaks. In particular, this framework will enable a quick update of the potential risk of changes in peak floods when new climate change projections become available, such as for example the UKCP09 scenarios, without the need to undertake an extensive hydrological modelling and impact study