Good morning Edzer and Dave,
Thanks for bringing up this point. I had a similar issue recently using
krige(). Observations at 5800 locations, attempting to krige()
predictions at 112,000 locations resulted in the same "memory.c" error
message. Reducing predicted locations to <<50,000 and reducing max.dist
seemed to help, but the predictions still took a very long time (>2
hours). (Running winxp with 4GB memory.)
Can you clarify your suspicion regarding the "lack of standardization of
coordinates"?
Chris
Edzer Pebesma wrote:
Dave,
12000 observations fit, in the c representation, in less than 1 Mb (64
bytes per observation).
The issue is that you think that passing maxdist to predict.gstat has
an effect. It doesn't; you need to pass it to function gstat().
The same thing happened in this
https://stat.ethz.ch/pipermail/r-sig-geo/2008-September/004182.html
message, where nmax was passed to predict.gstat, and simulation took
forever. The other issue in that question was, I suspect, lack of
standardization of coordinates, used in a trend surface.
--
Edzer
Dave Depew wrote:
Is there a limit to the # of observations or size of file that can be
co-kriged in gstat?
I have a ~12000 observation data set (2 variables), the variograms,
cross variogram and lmc are fit well, and co-kriging starts ok
Linear Model of Coregionalization found. Good.
[using ordinary cokriging]
then immediately outputs
"memory.c", line 57: can't allocate memory in function m_get()
Error in predict.gstat(fit.ck, newdata = EcoSAV.grid, maxdist = 100) :
m_get
Iv tried different maxdist from 10 to 1000, with exactly the same
result.
I recently upgraded my RAM to 4Gb and flipped the windows XP /3GB
switch.
--
J. Christopher Taylor, Ph.D.
Applied Ecology and Restoration Research
National Ocean Service / NOAA
National Centers for Coastal Ocean Science
Center for Coastal Fisheries and Habitat Research
101 Pivers Island Road, Beaufort, North Carolina 28516-9722
Ph:(252) 838 0833 Fx:(252) 728 8784
Website: http://www.ccfhr.noaa.gov/
begin:vcard
fn:J. Christopher Taylor
n:Taylor;J. Christopher
org:Center for Coastal Fisheries and Habitat Research;Applied Ecology & Restoration Research Branch
adr:101 Pivers Island Road;;NOAA National Ocean Service;Beaufort;NC;28516;USA
email;internet:[EMAIL PROTECTED]
title:Research Ecologist
tel;work:252.838.0833
tel;cell:252.723.3993
x-mozilla-html:TRUE
version:2.1
end:vcard
_______________________________________________
R-sig-Geo mailing list
R-sig-Geo@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
