ChrisButch wrote: How would you calculate a 1 in 200 year flood for Knoydart, I wonder?
Using observational data, computer modelling, topographic and river data, it is possible to make predictions for how rivers react given the intensity and duration of rainfall, taking into account land features such as soil type, vegetation and buildings. With modelling, estimates for flood severity can be linked to known land and river data and observations at monitoring stations. If you have reliable river level records, extreme value theory can be used to obtain river flows and flood levels corresponding to a return period of interest (e.g. 50, 100, 1000yr). Once you know how rivers react to rainfall, given the influences mentioned above, you can predict what river flows correspond to a particular return level at individual locations, and then interpolate (e.g. kriging) to obtain return level maps for the UK down to whatever resolution you require, given high enough resolution information about topography, rivers, land use, and rainfall records. I do something very similar in my job to produce return level maps for peak gust across the UK.