[GRASS-user] Raster classification by typical methods?

[Moritz Lennert]

On 29/03/13 16:02, Tim Michelsen wrote:
> Hello user list,
> I have a question concerning raster works:
>
> classifying in by typical methods
>
> How can I create a reclassified raster from another raster with
> continuous values where the classification follows the following rules
> * equal steps (a class break every 100 or 50 [mg/m^2])
> * equal interval

Those two are equivalent. And can be done with a little r.mapcalc math 
or with r.recode rules (which you can embed in a script if you need to 
calculate the intervals on the fly.

> * automatically classification based on defined number of classes

You can use r.quantile for quantile classification:

r.quantile elevation quantiles=5 -r --quiet | r.recode elevation 
out=elev_quant5 rules=-

Other methodes are currently not available AFAIK. Code exists 
(lib/arraystats) to calculate classifications according to different 
algorithms based on arrays of numbers. This is used for v.class and 
d.thematic.area, but I don't know how feasible it would be to use it for 
raster as the array of numbers to be classified is loaded into memory.

Moritz