[Gstat-info] "memory.c", line 57: can't allocate memory in function m_get()

[Edzer Pebesma]

Tom, your information is of little use without telling how large 
everything is, including your computers real and virtual memory.

Simulation needs more memory because simulated date are added to the 
observation data, and not discarded which can be done by prediction only.

For your last question: you would need the covariances between kriging 
prediction errors, these are usually not computed.
--
Edzer

Hoogland, Tom wrote:
>
> Using Gstat stand-alone, I am trying to solve a gaussion simulation 
> problem that I have previously solved using ordinary kriging. Even 
> when limiting the local neighbourhood (factor 10), the number of 
> prediction locations (factor 16) and the number of realisations (to 1) 
> the following error occurs after "drawing 1 multivariate realisation 
> of beta...":
>
> "memory.c", line 57: can't allocate memory in function m_get()
> gstat: matrix library: virtual memory exhausted
>
> Can anyone explain what causes the large memory use of simulation 
> compared to kriging and give me some suggestions to solve this problem?
>
> The reason for using gausion simulation is the need for an estimate of 
> the accuracy for a area average value for an area to be defined later 
> on. Is there some way using the kriging variance and covariance at 
> prediction locations to calculate the accuracy of an area average 
> value later on?
>
> Thanks,
> Tom Hoogland
> Alterra, Centrum Bodem
> Tel: 0317-484251
> E-mail: Tom.Hoogland at wur.nl
> Internet: _www.grondwaterdynamiek.wur.nl_ 
> <file://www.grondwaterdynamiek.wur.nl>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Gstat-info mailing list
> Gstat-info at geo.uu.nl
> http://mailman.geo.uu.nl/mailman/listinfo/gstat-info
>