FLOODING IN SCOTLAND

Flooding in Scotland

Flooding in Scotland

Introduction

Over the past decade, as scientists, water managers and policy-makers have moved increasingly towards the view that the global climate is changing significantly, and may be expected to change further through many decades in future, it has been recognised that there are likely to be changes in flood risk in many parts of the world. In Scotland, unusually damaging floods have occurred on several river systems since 1989, particularly the River Tay in 1993 (Black and Anderson, 1994), in Strathclyde in 1994 ( Black and Bennett, 1995) and in Moray in 1997 ( Brown and Black, 1997), causing cumulative financial damages in excess of £150 million. These have focused attention amongst hydrologists and environmental managers on the possibility of future increases in flood risk, on the need to develop new methods for improving estimation, and in quantifying the uncertainties which exist. Some response to these issues is already evident in terms of new planning guidance and flood protection legislation.

In Scotland, as elsewhere in the UK, standard risk assessment procedures have been based on the Flood Studies Report (NERC, 1975) for the past 25 years. Where gauging station flood data are available, annual maximum (AM) or peaks-over-threshold (POT) methods are recommended, depending on recorded length, while, for the ungauged site, statistical and rainfall-runoff methods are to be used, depending on the ultimate objective of the risk estimation being undertaken ( Sutcliffe, 1978). However, over the period of instrumental flow recording in Scotland, climatic variability has been found to exert a strong influence on flood behaviour, and to a much greater extent than in other parts of the UK ( Black, 1996). This calls into question the assumption of stationarity underpinning current risk estimation methods, and suggests that new or alternative methods may need to be considered.

Shaw (1994) (p. 377) defines a stationary time series as one in which 'the statistics of the sample (mean, variance, etc.) are not functions of the timing or the length of the sample', other than due to sampling variability. In Scotland, the stationarity assumption has been challenged by analyses which reveal statistically significant changes in the frequency of occurrence of flow threshold exceedances (i.e. annual POT frequencies), with frequencies in the 1980s and 1990s exceeding those observed in earlier records in most cases where significant changes are detected ( Black and Werritty). Stationarity is also questioned by some temporal clustering of very large floods, suggesting non-random occurrence ( Black, 1996). Such behaviour may represent the operation of some cyclic process, or may be the manifestation of a trend. The implications of the former may be expected to be less severe than those of the latter, which are consistent with the results of climate change modelling studies (e.g. CCIRG, 1996). In such a situation, flood risk estimates based on observed flood data would be likely to underestimate risk.

The distinctively Scottish nature of these changes in flood behaviour (compared with UK-wide change) is underlain by increased frequencies of westerly airflows over ...