[Edserplo_users] Statistics update

[Steve Loch]

After removing one further bug Extremes and Surges now match, with one exception, across 3000+ comparisons with Edteva. The one discrepancy is down to precision: when looking at it in Edserplo it turned on values in the 9'th significant place (!) of the surge data (i.e. dependent on tidal prediciton). Presumably the Edteva values differ very slightlly from the Edserplo computed values and tip a different way. 

It may be worth pointing out that the algorithm for locating extremes can do better than simply find the maximum and minimum for the month. It can locate 2'nd, 3'rd ,, etc. local minima and maxima. Because it was only used to locate the highest and the lowest values in the month the Fortran bug never came into play. As I was interested in the second  local extremum (which coincides with the Edteva's first local extremum for the the month) I started using this path and uncovered the problem - which had been faithfully transmitted to the Java.

After some recoding, port Histories seem to coincide with their Edteva counterparts though I have not done the check across all the ports yet (because of the dreaded "too many open cursors" problem encountered in generation). The Dataring 45-site monthly update may escape TMOC; I don't know yet but it will be close.

Discrepancies in MSL are now my concern. As mentioned previously these are likely to be dependent on the spline values used to interpolate the data. With a full set of good data on the 1/4hr computed values of MSL should be independent of the spline - so this is the first check. There are a great many 2+ mm discrepancies at the moment.

Steve