Strange time series plot.

From: David Jones <jonesd647_at_nyahnyahspammersnyahnyah>
Date: Mon Aug 12 2013 - 12:05:14 MDT

Dear NCL Talk,

I have a rather vexing issue with my time series plot that I hope you may be able
to help with.

I have a 20 member ensemble and am plotting an area averaged sea level pressure
time series over a section of my domain.

The ensemble mean is simply the arithmetic mean of all 20 time series.

I then calculate the intra ensemble standard deviation, i.e. amongst the 20 members.
This gives me another time series, of the same length, with the standard deviation
between all 20 members at each time step.

I then multiply this time series (standard deviations) by 1.96 and create two new time
series by adding/subtracting this from the original time series, like creating 95%
confidence intervals.

The problem is, when I go to plot these three time series, the result is not as expected.

I expect to see three time series, with the SLP time series in the middle and the other
two above and below it. I expected that the upper and lower time series should be
'mirror images' of each other, through the central time series. However, as you can see
in the attached .png file, the upper and lower time series are not symmetric about the
central time series. In fact, the min/max peaks aren't even fully aligned. I can't
understand this, as at each time step the same value is being added and subtracted.

I have repeated the same plot, except this time subtracting just a constant from my entire
SLP time series, and this does not give any problems. I have also verified mathematically
that the difference between the lower and middle time series is the same as that between
the upper and middle time series - this is at odds with what the plotted image suggests.

This leads me to conclude that some error is occurring at the plotting stage.

Do you have an idea what's going wrong here? Or am i making some stupid mistake?

I've attached a copy of my original script (rac_prob_slp_ts.ncl), and also two ascii
files in which i've written the data that goes into the time series' in the upper plot.
These can be used to reproduce the problem with the simplified script i've also
attached (simple_ts.ncl).

Any help/guidance you can offer would be immensely appreciated.

Best wishes,


PS I'm using NCL 5.2.1 on a SUSE system.

ncl-talk mailing list
List instructions, subscriber options, unsubscribe:

Received on Mon Aug 12 12:05:33 2013

This archive was generated by hypermail 2.1.8 : Wed Aug 14 2013 - 18:35:09 MDT