Appreciate all the comments ...

Appreciate all the comments already made.  I have just one small issue to raise.  Usually if we have adequate hydrological data, we tend to carry out a regression analysis on all of the data and use that to assess the safe yield of a planned or existing reservoir.  However if we only analyse the extreme events in the historic data rather than all the data, you are likely to find that the assessment will give a different prediction of particularly streamflow deficits than when using all the data.  This then leads to an inaccurate risk assessment of low flow periods within the extreme events.  This is even more relevant in areas subject to significant climate change rainfall occurrences.