Version 1.4.0

This version actively uses ensemble members from the input files.

New metrics and diagrams

• -m fcrps computes the fair continuous ranked probability score from an ensemble
• -m rankhist shows a rank histogram
• -m spreadskillratio computes the ratio of ensemble standard deviation to the root mean squared error of the ensemble mean
• -m obsstddev computes the standard deviation of observations
• -m fcststddev computes the standard deviation of forecasts

Updated metrics and diagrams

• -m spreadskill now uses the ensemble standard deviation as spread (instead of the spread between two quantiles)
• -m spread now uses the ensemble standard deviation as spread (instead of the spread between two quantiles)
• -m timeseries now also shows ensemble members
• -m obsfcst now supports ``-type map`

New data manipulation options

• -Tagg accepts change and abschange
• -fcst accepts ensemble:mean, ensemble:variance, and ensemble:<number> (where number is an ensemble member index)

Input files can now have a time axis with units other than seconds since 1970-01-01.