ISSN 0004-0894 © The Authors. Journal compilation © Royal Geographical Society (with The Institute of British Geographers) 2006 Area (2006) 38.1, 34–46 Blackwell Publishing Ltd Historical and pooled flood frequency analysis for the River Tay at Perth, Scotland N Macdonald*, A Werritty**, A R Black** and L J McEwen† *River Basin Dynamics and Hydrology Research Group, Institute of Geography and Earth Sciences, University of Wales, Aberystwyth, Ceredigion, Wales SY23 3DB Email: nim@aber.ac.uk **Department of Geography, University of Dundee, Dundee DD1 4HN †School of Environment, University of Gloucestershire, Cheltenham GL50 4AZ Revised manuscript received 28 November 2005 Improved estimates of UK flood risk during a period of increased climatic variability place challenges on existing methods that rely on short instrumental records. This paper examines the value of using historical data (both documentary and epigraphic) to augment existing gauged records for the River Tay at Perth as part of a multi-method approach to assessing flood risk. Single station and pooled methods are compared with flood risk estimates based on an augmented historical series (1815–2000) using the Generalized Logistic and Generalized Pareto distributions. The value of using an even longer, but less reliable, extended historical series (1210–2000) is also examined. It is recommended that modelling flood risk for return periods >100 years should incorporate historical data, where available, and that a multi-method approach using a high threshold Generalized Pareto distribution can also add confidence in flood risk estimates for return periods <100 years based on standard methods. Key words: Perth, flood risk estimates, historical data, pooling method, Generalized Pareto distribution, River Tay Introduction Providing robust and reliable estimates of flood risk is increasingly important in a world where climate change is increasing the number of hazardous locations and the pressure to occupy such locations is often poorly regulated (Werritty et al. 2002; Evans et al. 2004; Department of Food and Rural Affairs 2004). By way of illustration, the Scottish Executive has designated the 1 in 200 year event as the design flood in its planning guidelines and forthcoming second generation flood risk maps (Scottish Executive 2004). However, estimating the return periods for floods likely to have a significant societal impact is challenging unless suitably long records exist (Archer 1999; Black and Burns 2002). Few sites across the UK provide records >50 years in length. The average gauged record is about 24 years (Marsh and Lees 2003) and sites with records dating back to the nineteenth century are rare. This represents a significant problem in providing robust and reliable estimates of flood risk, as relatively short records often fail to include an adequate sample of large rare floods. The Flood Estimation Handbook (FEH) (Institute of Hydrology 1999) identifies two principal methods for estimating flood risk for sites with relatively short gauged records: pooling and historical augmenta- tion. The first is now widely adopted across the UK, especially by consultants advising local authorities and developers on the location of new residential and commercial developments. But as concerns are raised over the reliability of pooling in the