[STATSGRASS] Re: GRASS

Agustin Lobo alobo at ija.csic.es
Wed Jan 16 05:54:46 EST 2002


(I'm sending this message to the statsgrass
list, which is read by (few) people
working on R-grass).

My personal answer is at the end of Guerreau's message.

On Wed, 16 Jan 2002, guerreau wrote:

> J'imagine que la plupart des concepteurs et des utilisateurs d'ADE-4 utilisent au moins un SIG.
> Sans doute certains d'entre eux connaissent-ils GRASS, logiciel open source fonctionnant sous LINUX.
> 
> Mes questions sont donc:
> 1. comment se situe ce SIG par rapport aux "classiques", genre MAPINFO ou ARCVIEW, surtout en termes de fonctionnalités ?
> 2. le fait de l'open source et des liens établis avec R permet-il une utilisation commode, ou la possibilité d'une utilisation commode, avec ADE-4 sous LINUX ?
> Toutes les appréciations me seraient bien utiles (dans la perspective, notamment, d'une utilisation pédagogique).
> Merci à tous.
> 
> Alain Guerreau   CNRS Centre de Recherches Historiques 54 bd Raspail 75006 Paris
> guerreau at msh-paris.fr

I'used a lot Grass for quite a long time, as well as
Splus and, now R.

Although grass has now a reasonable tcl-tk menu, it's not
at all a "click click" software, which is a great advantage
for me but can be seen as problem by others. The main advantage
of grass is that you can easily put together grass procedures
within shell scripts to develop your own methods or models, or
just to automate repetitive tasks. If you want, you can also
use grass code to create your own c programs, although that's far
more involved.

There is an R package (in development) to communicate R and Grass.
At sometime, I even used to include grass commands within R functions,
which requires having grass running in a separate window. The main 
problem is that (at least my) raster files are huge, which is a problem
for R. The most efficient approach is to process the geographic
information in Grass untill you get a table of reasonable dimensions
that you can pass to R for further analysis. An example of a procedure
that would not work is importing a multispectral image to R as 
as a 3d array and run pca(). Instead, you can run pca in grass and
pass to R the results to visualize eigenvectors etc.

I'm not using grass that much now because I never got a reasonable
24 bit display of RGB composites with it. But this is because I do work
with RS imagery and must interact with the images in a very 
good visual environment. But if you only work with maps, I would
say that grass is a good option among the raster-based 
free (and even cheap) GIS packages 
(although you might want to try SPRING as well, no so known in 
Europe: http://www.dpi.inpe.br/spring). 
Nevertheless, if you have money, TNTmips is great
(www.microimages.com)

Agus

Dr. Agustin Lobo
Instituto de Ciencias de la Tierra (CSIC)
Lluis Sole Sabaris s/n
08028 Barcelona SPAIN
tel 34 93409 5410
fax 34 93411 0012
alobo at ija.csic.es







More information about the grass-stats mailing list