[STATSGRASS] kriging - too long time

[Edzer J. Pebesma]

That might be "explained" as follows: 8 to 400 is a factor 25, solving a 
covariance matrix is an O(N^2) operation so would slow down by a factor 
25*25 = 625. The remaining factor 6 (seconds to hours: 3600/625) should 
be attributable to the load of finding the nearest 400 neighbours 
instead of 8 nearest neighbours. I don't know that behaviour (a 
bucket-PR quadtree search index is used), but am not surprised.

Best wishes,
--
Edzer

Carlos "Guâno" Grohmann wrote:
> Tomas,
>
> nmax has a strong influence on kriging time. In some tests I've done,
> interpolation time for and area with 520 rows x 430 cols ranged from
> about 20 seconds for nmax=8 up to 20 HOURS for nmax =399.
>
> best
>
> Carlos
>
>
>
>
> On 9/2/07, Tomas Lanczos <lanczos at t-zones.sk> wrote:
>   
>> I am performing some kriging following the procedure as it is here:
>> http://casoilresource.lawr.ucdavis.edu/drupal/node/438
>>
>> My problem is that the kriging proces is very slow, it takes approx. 7
>> hours. Is the problem a too high resolution, or too much sampling points
>> (2234)? How can I optimalize the process?
>>
>> My region setup is the following:
>>
>> projection: 1 (UTM)
>> zone:       33
>> datum:      ** unknown (default: WGS84) **
>> ellipsoid:  bessel
>> north:      5387295.0553002
>> south:      5288130.42061478
>> west:       652025.98376024
>> east:       786301.92841404
>> nsres:      150.02214022
>> ewres:      150.02898844
>> rows:       661
>> cols:       895
>> cells:      591595
>>
>> Many thanks
>>
>> Tomas
>>
>> _______________________________________________
>> statsgrass mailing list
>> statsgrass at grass.itc.it
>> http://grass.itc.it/mailman/listinfo/statsgrass
>>
>>     
>
>
>