Probable maximum precipitation

Most of the statistical distributions used in the type of frequency analyses described above assume that there is no upper limit to the depth of precipitation that might be recorded at a site in a given duration (the distributions have infinite upper tails). It is certainly the case that, as more and more rainfall measurements become available throughout the world with the increasing number of rain gauges and with the lengthening of existing series of observations, records of maximum rainfall for the various durations continue to be broken. The question arises, however, as to whether records will always continue to be superseded or whether there is a physical upper limit to rainfall. This concept of a finite limit has been named the probable maximum precipitation (PMP), the existence of which has aroused much controversy. A definition of PMP has been given by Wiesner (1970) as 'the depth of precipitation which for a given area and duration can be reached but not exceeded under known meteorological conditions'. The PMP will of course vary over the Earth's surface according to the climatic or precipitation regime. The envelope curves of Fig. 9.1, which show the extremes in both time and space, suggest values that would be expected to be exceeded only rarely at any particular location in the UK or the world, but do not necessarily imply any upper limit.

The idea of a physical upper limit for precipitation comes from meteorological concepts of the maximum precipitable water-holding capacity of a column of air under given meteorological conditions in a particular climatic regime (e.g. Hershfield, 1961). This can be estimated from models of a column of air of different degrees of complexity (recently extreme rainfalls have been studied using fine-resolution convection resolving general circulation models). While such models can be based on the physics of the atmosphere and rainfall generation processes, the resulting rainfall rates will depend on whether the particular parameterisations used are appropriate and on assumptions about the local boundary conditions. In essence, the question about whether PMP is an appropriate concept is replaced by the need to define assumptions about the likely advection of heat and water vapour into the boundaries of the column of air being considered. Clark (2005), for example, in discussing the Martinstown, Dorset storm that is still the record for a 24-h rainfall total in the UK, suggests that the assumptions made in the estimation of PMP in the Flood Studies Report mean that the resulting estimates are too low. Collinge et al. (1992) came to similar conclusions in their study of the Hewenden Reservoir storm-event of 1956.

Thus, while the PMP is an attractive concept in engineering design situations requiring protection against catastrophic failures (such as in assessments of dam safety for structures sited upstream of centres of population, e.g. Institution of Civil Engineers, 1996), it is difficult to justify any method of estimating a value for a likely PMP over a catchment area. It might be better to estimate the uncertainty associated with the event of a chosen extreme return period for the design and verge on the side of caution in taking a higher quantile estimate within that range of uncertainty.

0 0

Post a comment