[GRASS-user] improve v.rast.stats speed?

[Markus Metz]

Hi Giovanni,

could you please check out
svn checkout https://svn.osgeo.org/grass/grass-addons/raster/r.univar2.zonal
and let me know if the results and the speed are ok?

r.univar2.zonal does zonal statistics if I understood the concept all 
right. You have to give a zoning map as input. That would be in your 
case the rasterized vector areas. The output can be very long, as in 
your case, so I added an option to dump the stats in a file, that way 
it's easier to scrutinize.
In my tests, the stats were identical to r.univar2 in develbranch_6, 
only that I had to run r.univar2.zonal only once and not several times. 
This r.univar2.zonal is not polished: 3D support is missing and the 
output is not shell style although it should be, no idea why not. I 
would like to leave the polishing to the experts.

I haven't added an entry in wiki addons, because I don't know if this is 
an unpolished gem or unnecessary with your current solution or a pile of 
trash (well, not completely trash, results are at least identical to 
r.univar2).

Regards,

Markus M


G. Allegri wrote:
> The rasterizing method gives comparable performances to ArcGIS. I
> confirm that it does the same.
> The bottleneck is the r.univar limitation to CELL. I have to
> investigate why, as it is based on r.stats which works with
> DCELL/FCELL too
>
> The final join would be an improvement respect to Zonal Statistics in
> ArcGIS which simply produce a dbf with OIDs from polygon IDs but
> doesn't merge it into the original vectorial.
>
>
> 2009/2/19 Markus Metz <markus.metz.giswork at googlemail.com>:
>   
>> Markus Neteler wrote:
>>     
>>> On Thu, Feb 19, 2009 at 2:20 PM, G. Allegri <giohappy at gmail.com> wrote:
>>>
>>>       
>>>> Thanks for the ideas.
>>>> I've just tried Starspan but it's performance is still too slow. I've
>>>> let it run for 15 minutes...
>>>>
>>>>         
>>> Can you please try GRASS 7?
>>> grass70/scripts/v.rast.stats/v.rast.stats.py
>>>
>>>       
>> grass70/scripts/v.rast.stats/v.rast.stats.py also uses r.mapcalc in every
>> pass:
>>
>> for i in cats:
>>   ...
>>   grass.mapcalc("MASK = if(...)
>>
>> if v.rast.stats is faster in grass7, then probably because of improved
>> raster libs. A speed increase from >5 hours to 40 seconds is unlikely since
>> grass.mapcalc is still called 1793 times (assuming each area has a unique
>> category) for a region with 4415x6632 cells...
>>
>>
>>     
>
>