Yes, you're right. Ferret uses double-precision for computation,
but the graphics side, Plot-Plus, is single precision, which has a
range of about 7 decimal places. When Ferret reads
double-precision data, it reads the full precision, so LIST/prec=8
will show you the numbers. When it's passed to Plot-Plus, it's put
into single-precision variables.
I wonder if we could think of a way to do the contouring in two
if icn2 LT 1e-03 then icnc2
if icn2 gt 1e-03 then icnc2
Each of the two expressions has a dynamic range that can be
expressed accurately by a single precision variable. The trick
then would be to plot these together using sets of color levels
that harmonize and two color keys plotted to fit together. Here's
a sketch of that:
You might want to adjust
things so the length of the two color keys are proportional to the
number of colors, so each box on the color key is the same size -
and maybe make yourself a couple of custom color palettes to use
for this variable. Also note that sometimes doing a "SHADE" first
and overlaying a "FILL" will get you better results than just a
"FILL", as a shade plot draws color out to the edge of the grid
boxes, and FILL only to the center of the grid boxes.
! define two identical viewports
define view/x=0:1/y=0:1 full1
define view/x=0:1/y=0:1 full2
define symbol low_lev =
define symbol high_lev =
! plot the smaller values in one viewport with the lower
set of levels. Make a half-height color key.
set view full1
fill/levels=($low_lev)/pal=blue_green_yellow/SET if icnc2
lt 1e-03 then icnc2
! Now plot the larger values in another viewport, with a
color key starting at half height
! The second viewport is drawn in on top of the first.
set view full2
fill/levels=($high_lev)/pal=yellow_orange_brown/SET if icnc2
ge 1e-03 then icnc2
On 3/19/2015 4:40 AM, Bier, Andreas
I would like to create a two-dimensional contour plot (depending
on time and height) for an ice crystal number concentration
My commands are:
let icnc1 = if icnc gt 0 then icnc
let icnc2 = icnc1[i=@ave,j=@ave]
--> averaging over longitude and latitude
Doing so, the plotted data range of the contours goes only from
2.8 till 1e-05 and not till 1e-09. If I decrease my maximum value
to 0.1 instead of 2.8, my smallest contour value decreases to
1e-06 and so on. Consequently, it seems to be a precision problem.
There is a possible adjustment for higher precision within the
list command, e.g. list/p=8 icnc2.
But this adjustment doesn't work for the plot or contour commands
and I didn't really found another solution for this problem in the
User's Guide. I hope you can help me.