Dave:

Here is my next question/problem re vectors. I think you're on
vacation right now, and will find this when you get back -- I'll
cc to D. Kennison (I could address this to the Daves), but there 
is no rush on this.

I'm plotting wind velocities from a model that has a 5 degree
latitude x longitude grid resulting in 36 latitudes (-87.5 to
+87.5 every 5 degrees), and 73 longitudes (-180 to +180 every
5 degrees).  E.g., I use ezmap to set up a cylindrical equidistant
projection, set VV MAP=1, SET=0, XIN=YIN=3, etc., and everything is
groovy (XC1=-180., XCM=180., YC1=-87.5, YCN=87.5). 

I also often use a stereographic projection centered over one of
the poles with an arbitrary perimeter latitude set as follows:

        ang = 90.-abs(perimlat)
        call mapset('AN',ang,ang,ang,ang)
        ...
        [set VV YC1 and YCN accordingly]

This also works fine *except* that the arrows are crowded at the
high latitudes, i.e., near the center of the plot. This is of
course because I am plotting the same number of arrows (73/3 if
XIN=3) around each latitude, and each successive higher latitude 
has a smaller circumference. So what I want to do is set a 
different XIN for each latitude, e.g., XIN=24 at 87.5, XIN=8 at
72.5, XIN=6 at 57.5, etc.

My attempt to do this results in separate calls to VVINIT and 
VVECTR (in addition to changing XIN) for each latitude. I'm sure
this is inefficient, and will incur redundant overhead, etc., but
I'm willing to accept that. For each call I want to plot 73/YIN 
vectors around the current latitude (an arrow at every YIN'th 
longitude grid point). So I set XIN for the desired longitude 
increment at the current latitude, YC1 = YCN = the current 
latitude and call vvinit and vvectr as follows (j is index at 
current lat, idm=0, rdm=0., nlon=73) (note N=1):

      call vvinit(u(1,j),nlon,v(1,j),nlon,rdm,idm,nlon,1,rdm,idm)
      call vvectr(u(1,j),v(1,j),rdm,idm,idm,rdm)

And, vwalla, it crashes:

Floating point exception
TB001 - BEGINNING OF TRACEBACK
      - $TRBK    WAS CALLED BY f_sig    AT 1012272b (LINE NUMBER      102)
      - f_sig    WAS CALLED BY __handlr AT  666456b 
      - __handlr WAS CALLED BY VVECTR   AT  323232b 
      - VVECTR   WAS CALLED BY PLTVECT  AT   77615a (LINE NUMBER       86)
      - PLTVECT  WAS CALLED BY PLTPOL   AT   45377a (LINE NUMBER      168)
      - PLTPOL   WAS CALLED BY TIGCMDIF AT    2710b (LINE NUMBER      212)
      - TIGCMDIF WAS CALLED BY $START$  AT     303a 
TB002 - END OF TRACEBACK
Floating exception (core dumped)

There was no apparent complaint from vvinit, and vvgetr of YC1 and
YCN showed that both were -87.5 as expected. cdbx printed expected
values for u(:,j) and v(:,j).
I notice that in the doc, it says that if YC1=YCN, then YC1=1.0 and
YCN=float(N). (Actually, I think there's a typo in both the programmer
doc and the man pages where under the description for YCN you say
"If XC1 is equal to XCM, ..." where I think you mean "If YC1 is 
equal to YCN, ...".) I suspect this is related to my crash. The
question is how do I plot vectors around a single latitude with a
single call to vvinit and vvectr? Thanks for any help or advice,

--Ben
  foster@ncar
  x1595