20010220: Calculating moisture flux divergence (cont.)

>From: "Paul L. Sirvatka" <address@hidden>
>Organization: College of DuPage
>Keywords: 200102162013.f1GKD9L01250 McIDAS-X SFCCON GRDDISP

Paul,

re: wrong number of parentheses
>Yes! THat is it. Boy I could not see that. I need to get my eyes checked
>because I thought I had it right.

I use vi to check matched parentheses.  If you use vi as an editor, and
you are not familiar with how to find matching parenthese and curly
braces, do the following:

o edit your script using vi
o move your cursor down to and over one of the parentheses in any
  one of the MATH= specifications
o repeatedly press the '%' (percent key; that is shift 5)

The cursor will jump to the parenthesis that closes the open one you
are first located on, or the opening parenthesis for the closed one
you are first located on.  This makes it easy to see the grouping
in expressions with a lot of parentheses.

re: do the calculations now look to be the same
>It looks good to me.

Super.

re: MDVG from SFCPLT is different
>Yes It is still different. THey may have a multiplier that is
>different. Maybe not 36000 (which is sec/hour * 10) I know that my
>equation is correct. I wonder what McIDAS had been using.

How different are the displays?  Are we just talking about a scaling
difference (i.e., the pattern is the same), or are we really talking
about a different analysis?  I ask since reproducing the SFCPLT output
will be difficult for me since none of our data is local (I use ADDE
access to a remote server pretty much exclusively).

re: keep losing image window
>Yesterday...processes kept getting hung when it tried to do some grddisp
>problems. It may have been the parantheses problem.

It may have been.  I just reran the TEST.BAT script in a McIDAS-X
session on a Sun Solaris SPARC 2.6 machine, and the iamge window
disappeared!  The indication was that there was a floating point
error.  Presumably, this had to do with GRDDISP.  After I restarted
McIDAS, everything ran smoothly.

Question:  I had noticed what I thought was a bug while running the
version of TEST.BAT that had the parenthesis problem.  The plot color
for the displays in graphic frames 4 and 5 were yellow (color 3), as
they should be.  The plot color for frame 6, on the other hand, was
green (color 4).  This was an error that I was going to look for in the
source for GRDDISP.  After restarting my McIDAS-X session, all of the
plots were done in yellow (color 3).

What were the colors used for display in the three frames with your
original MATH formulation?  I am wondering if the multiplication of a
single term by 36000 was causing a floating point error and causing
memory to be stepped on.

>I have XWindows on my
>office machine running WIN98.

OK.  Using something like eXceed (or some other windowing package)?

>It also dies on a linux box.

This would match my observation on Solaris SPARC.

>Today it has
>been going better. I did reboot our machine. I fear that there may have
>been memory strings left around. We have MCIDAS 7.704

I think that something was getting stepped on from a floating point overflow.

I will have to look into this.

Tom