floating point | null data in GRASS!

Greg Koerper greg at towhee.cor2.epa.gov
Fri Mar 12 16:52:31 EST 1993

In answer to William Hargrove's comments:

> GrassHoppers:
>      I don't really see the big deal about floating point maps and zero vs
> no data problems.  Multiply by 10 times the number of decimals of accuracy
> you need to take care of the first problem.  Shift all category values up
> one notch to take care of the other.

It is simple to transform one floating-point variable into integer form.  But
it is a data management and quality assurance nightmare to be handling
tens of floating point transformations, both multiplication and division.
If I build a customized research environment, now both I and the end-user 
must deal with the mental gymnastics.  If we wish to develop diagnostic ratios
of output variables, the confusion expands.  And now try sharing these data
with S-PLUS, PCI or PV-WAVE for visualization, etc.

Shifting categories is not a solution, because the variables are continuous
scalars (not categories!), and may range from negative to positive.  Thus what
had been -1 is now misinterpreted as zero, and image displays have erroroneous
"null data" holes.  I have used this chewing-gum approach in the past, but at
a significant cost of my time and that of the researchers I coordinate with. 
For example, if you do these shifts, you now have the overhead of kludging the
legend display, etc.

>      IMHO, a much more severe limitation to the current implementations of
> GRASS is the current XDRIVER.  The named fifo pipe technique is a kludge at
> best.  GRASS needs *real* X-window support, including more than 8 plane
> support.

This is a legitimate concern.  But at our lab, floating point and null data
limitations outweigh its importance.  The user community needs to weigh in on
the competing priorities.

>      The people at CERL are aware of this (and even a little defensive
> about it).  All such imporvements take, they are quick to point out, is
> time and money ...

Alas, yes.  

Greg Koerper                            Internet: greg at heart.cor.epa.gov
ManTech Environmental Technology, Inc.  UUCP:     hplabs!hp-pcd!orstcs!koerper
US EPA Environmental Research Lab
200 SW 35th St., JSB                    
Corvallis, OR   97333           "The 90s will make the 60s look like the 50s. 
(503) 754-4490                          Just ask your kids."
 fax:  (503) 754-4338

More information about the grass-user mailing list