Floating point rasters

Hi Tim,

1. I am wanting to write a GRASS routine to create a raster layer of
   "Landform Elements" as in D.J.Pennock, Zebarth and DeJong, 1987,
   Geoderma 40:297-315. The Landscape Elements are calculated from
   an elevation raster layer, based on gradient, profile (downslope)
   curviture, and plan (across-slope) curviture.
   Please, please, please, someone tell me the routine I want (or
   at least parts of it) is already available. :slight_smile:

I have not read this article but I think there should be
a quite easy way how to calculate any topographic attribute
derived from slope, aspect and curvatures.
Try to interpolate your DEM data using s.surf.tps or r.resample.tps.
There are options like slope, profile curvature and
tangential curvature (which can be easily converted to
plan/horizontal curvature). Afterwards you can
use r.mapcalc to calculate the landsc. elements.
Read carefully manual pages to handle curvatures values
correctly (they are multiplied).

2. How shall I deal with the "integer only" limitation of GRASS raster
   files? For the elevation input data, if it is in units of meters,
   and the cells are close to 1m x 1m in dimensions, the equations won't
   work very well, for anything but very hilly landscapes. Do I specify
   that the input file must be in elevation units of cm? Or test that
   the units of elevation are an order of magnitude finer than the cell
   dimensions? And for the output files, I would like to be able to make
   raster surfaces of the aspect, gradient, profile curviture and plan
   curviture as well as the landform elements, but some of these are
   fractional values. I suppose I can output (value x 100) or some
   such.

Probably you will have to multiply/devide something in
r.mapcalc command line. I think your input elevation data and
position coordinates should be in the same dimensions during computations
(this can be controlled by zmult parameter in s.surf.tps
and r.resample.tps).
These integers are sometime really
uncomfortable but I've heard something that some people at CERL
are working on floating point version of GRASS.

Hope this helps

Jaro

------------------------------------------------------------------------
Jaro Hofierka
Dept. of Cartography, Geoinformatics & Rem. Sensing
Comenius University
842 15 Bratislava E-mail: hofierka@devin.fns.uniba.sk
Slovakia hofierka@geoinfo.fns.uniba.sk
------------------------------------------------------------------------