[STATSGRASS] The gstat R library's: memory problem

[Roger Bivand]

On Fri, 25 Aug 2006 sisma66 at tiscali.it wrote:

> Thanks for all information you told me, so I undestood:
> 
> 1) this isn't a bug but is an R problem that he can't manage a vector 
> so big

On a 64 bit machine not a problem, just rather unnecessary.

> 2)both the sizes of the data and newdata are the same of grass region 
> (I hope this can help more)

Yes, the big data matrix is roughly (800*900)^2 8 byte elements, so it is 
more than you need.

> 3) reducing the resolution It works and also using maxdist

maxdist will do fine, because you would expect the spatial dependence to 
decay quite sharply.

> 4) sorry for my stupid question, next time I will try to solve alone 
> my issue

My comment wasn't (just) to you, very often problems happen to more than 
one researcher at a time, and searching the archives is a useful trick to 
know. I hope you make good progress now.

Roger

> 
> Thanks to all
> 
> Pierluigi
> 
> Il giorno gio, 24/08/2006 alle 18.36 +0200, Roger Bivand ha scritto:
> On Thu, 24 Aug 2006, Edzer J. Pebesma wrote:
> > 
> > > Pierluigi De Rosa wrote:
> > > > Dear all users of gstat
> > 
> > I think it is maxdist= again, or some other way of reducing the 
> size. The 
> > GRASS region is roughly 800 by 900, but the sizes of both the data 
> and 
> > newdata are unknown. Pierluigi also cross-posted, which does 
> indicate a 
> > propensity to appeal for help rather than to look at the archives. 
> Using 
> > the search facility on the GRASS site with "gstat memory", there 
> were lots 
> > of promising hits from this list just in the last few months.
> > 
> > > >
> > > > I'm using the gstat library for univariate geostatistics 
> analysis,
> > > > I have to interpolate with kriging value of water level from 
> wells to evaluate 
> > > > the water table.
> > > > I use the univarsal kriging and the indipendent variable is the 
> digital 
> > > > elevation model (DEM)
> > > > If I use a DEM with resolution of 20 mt I have an error memory
> > > >
> > > > This is the string command I use for do that:
> > > >
> > > > OK_res20=krige(LEV_MAG91~DEM,loc=pozzi91_meno,newdata=DEM20,
> model=vgm_fit_mag91_DEM)
> > > >
> > > > This is the error I have
> > > > Error: cannot allocate vector of size 3995610 Kb
> > > >
> > > > If I use a DEM with resolution of 30 mt (using r.resample.rst)
> > > >
> > > > (This is the string) 
> > > > OK_res30=krige(LEV_MAG91~DEM,loc=pozzi91_meno,newdata=DEM30,
> model=vgm_fit_mag91_DEM)
> > > >
> > > > every thinks works ok, is this a bug?? Maybe  Dr. Pebesma can 
> help me!! :-)
> > > >   
> > > Perhaps he can, but remember he isn't a daily grass user. I doubt 
> it is 
> > > a bug because you do get a clear error message: you (through how 
> you use 
> > > gstat) ask the computer more memory than your computer has 
> available, 
> > > and that's what you are told. The question is why.
> > > 
> > > Without the commands how you retrieved your data and of what size 
> > > (number of rows and columns) they are there is little I can do, 
> from 
> > > here, however.
> > > --
> > > Edzer
> > > > Thanks, Pierluigi
> > > >
> > > > P.S.
> > > > this is my region setting
> > > >
> > > > g.region -p
> > > > projection: 99 (Transverse Mercator)
> > > > zone:       0
> > > > datum:      rome40
> > > > ellipsoid:  international
> > > > north:      4829500
> > > > south:      4813540
> > > > west:       2280740
> > > > east:       2298420
> > > > nsres:      20
> > > > ewres:      20
> > > > rows:       798
> > > > cols:       884
> > > >
> > > > --
> > > > Pierluigi De Rosa
> > > >
> > > > Department of Enviromental & Civil engineering 
> > > > Faculty of Engineering 
> > > > Perugia (Italy)
> > > >
> > > > _______________________________________________
> > > > statsgrass mailing list
> > > > statsgrass at grass.itc.it
> > > > http://grass.itc.it/mailman/listinfo/statsgrass
> > > >   
> > > 
> > > _______________________________________________
> > > statsgrass mailing list
> > > statsgrass at grass.itc.it
> > > http://grass.itc.it/mailman/listinfo/statsgrass
> > > 
> > 
> 
> 
> Naviga e telefona senza limiti con Tiscali     
> Scopri le promozioni Tiscali adsl: navighi e telefoni senza canone Telecom
> 
> http://abbonati.tiscali.it/adsl/
> 

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no