Using statistical methods to analyse environmental extremes
Tuesday 16 December 2008, 1020-1045
Lecture Theatre 1, Management School Building
Statistical methods for the analysis of extreme values are used when trying to predict properties of unusually large (or small) events. A key aspect is how to predict future extreme events when only a relatively short series of historical data is available.
Applications can be found in areas as diverse as structural engineering, hydrology, finance, insurance and communications. For example, in the construction of sea walls, we may be interested in estimating the wave height exceeded only once every 100 years when only 20 or 30 years of historical data are available.
In the first part of her talk, the Department of Mathematics and Statistics' Emma Eastoe will discuss the issues that arise when the underlying physical process from which the extreme data are produced is itself subject to change, either from other environmental variables or due to climate change; for example a model for high surface-level ozone levels would need to take into account that ozone levels vary through the year (in the UK they are higher in the summer when there is more sunlight).
The talk will go on to focus on a case study in hydrology, with the discussion of a model for the number of floods per year at a site on the river Thames. Prediction of floods is of interest to many people, including home owners, local councils, planners, rescue services and insurance companies. This part of the talk will use some of the ideas from the first part, since the probability of a flood on a given day varies, e.g. due to differing soil conditions and quantities of rainfall, in a way which is not what we would expect from the basic assumption that the probability of a flood is equal on any day.