using oacres to contour every value

[Dave Allured]

Apologies, I wasn't reading closely and I missed the part about oacres.
That is a gridding function, therefore each grid point is a function of
several nearby station values.  Now I think that each of your outliers
is getting blended with nearby normal values, thus the muted peaks that
you describe.

I can think of two methods to do what I think you want.  One, kludgey,
alter the gridding function before contouring.  Perhaps compute the 2-D
minimum within each grid box or radius, rather than the nice smooth
oacres function.  I don't recall whether Grads has a function to do this
easily; you might have to write a nested loop, or compute the grid
externally.  Or perhaps start with the oacres grid, but then overwrite
selected grid points with set defval().

Two, draw your own special symbols for outliers, such as circles, on top
of the contour map.  Then you can plot them on exact station locations.
Grads has functions to plot various shapes and symbols.  HTH.

--Dave A.
CDC/NOAA/CIRES

Dan Leins wrote:

> I was thinking this might be the problem, so I really tightened the
> contour values, but nothing's changed.  I don't know if it's really a
> problem with the shading as much as it is with the contours.
> Regardless of how I define my color curve, oacres will still only
> contour it out maybe as low as 55 degrees, despite the fact that
> there's several values of 32 shown on the map.  Is there anything I
> can do to really focus in on that area of bad data?
>
> Dan
>
> ---------- Forwarded message ----------
> From: *Dave Allured* <dave.allured at noaa.gov
> <mailto:dave.allured at noaa.gov>>
> Date: Sep 14, 2005 7:47 PM
> Subject: Re: using oacres to contour every value
> To: theedge981 at gmail.com <mailto:theedge981 at gmail.com>
>
> Dan,
>
> It sounds like you are overloading the available number of colors in the
> default grads color palette, which is about 15 IIRC.  The trick would be
> to space out some of the contour levels towards the outliers, while
> keeping the total number of levels at least one less than the number of
> available colors.
>
> Try clevs 28 through 80 increment 4, 14 levels total.  This should make
> rainbow bullseyes for the outliers as you describe them.
>
> --Dave A.
> CDC/NOAA/CIRES
>
> Dan Leins wrote:
>
> > List,
> >
> > I have a program that pulls temperatures out of a netCDF file, puts
> > them into a binary file, and plots them in grads.  I also contour and
> > shade the map using oacres.  It actually produces quite a nice image,
> > smooths out all the bad obs, but I was wondering, is there any way I
> > can use oacres to contour every value on my map?  I'd like to take
> > this image, and use it as a quality control product where bad obs
> > stick out, thus making it a little easier to identify where the
> > problem sites are.  This would essentially produce bulls-eyes where
> > bad data points exist, and this is exactly what I'm looking for.
> >
> > As an example, I made a plot of temperatures this morning, generally
> > in the upper 60s to lower 70s.  However, there are 2 points that stick
> > out with values of 32.  I messed around with the oacres radii, taking
> > it all the way down to 0.1, and while I start to get a bulls-eye, it's
> > in the wrong area, and doesn't contour all the way down to 32.  When I
> > go to make my contour, GrADS spits out "Contouring 65 to 75 interval
> 1".
> >
> > So I tried to set clevs 28 30 32...etc....74 76 78 80, then d
> > oacres(dummy.2(t=1), ts,0.5), and while it spits out "contouring at
> > clevs 28 30 32...etc...", the lowest contour I see on my map is 64.
> >
> > Does anyone have any idea how to make grads contour bad data to make
> > it stick out?  It seems oacres is trying to smooth it out too much.
> >
> > Thanks,
> > Dan Leins
>
>