[R-sig-Geo] multivariate sequential gaussian cosimulation using gstat in R

[Edzer Pebesma]

On 06/06/16 19:03, William Savran wrote:
> Dear All,
> 
> I am new to gstat, R, and fairly new to applying geostatistical methods, so I have a few questions.  I am looking to simulate a set of 3 (out of 4) correlated stochastic fields, which are all related through a linear model of coregionalization. One of the fields will be used as conditional data for the simulation.  With that being said, if anything seems alarming to you please let me know. Also, if you have any tips/trick that would be helpful for a novice please let me know.
> 
> Right now, my algorithm is working as follows (almost exactly following the demo):
> 
> # define gstat object
> sim.g <- gstat(id='slip', formula=slip.sc~1, data=sim, nmax = 30, maxdist = 200, beta=0, set = list(nocheck = 1))
> sim.g <- gstat(sim.g, 'psv', psv.sc~1, sim, nmax = 30, maxdist = 200, beta=0)
> sim.g <- gstat(sim.g, 'vrup', vrup.sc~1, sim, nmax = 30, maxdist = 200, beta=0)
> sim.g <- gstat(sim.g, 'mu0', mu0.sc~1, sim, nmax = 30, maxdist = 200, beta=0)
> sim.g <- gstat(sim.g, model=vgm(1,"Exp",1000,1,anis=c(90,0.5)), fill.all=T)
> 
> # fit lmc
> sim.fit = fit.lmc(var, sim.g, correct.diagonal = 1.01)
> 
> # perform SGSim
> z <- predict(sim.fit, newdata=xy, nsim=1, debug.level = -1)  If I specify fit.ranges = TRUE and fit.LMC = TRUE, I get the warning message 
> 
> "Warning messages:
> 1: In fit.variogram(object, model, fit.sills = fit.sills, fit.ranges = fit.ranges,  :
>  No convergence after 200 iterations: try different initial values?
> 2: In predict.gstat(sim.fit, newdata = xy, nsim = 1, debug.level = -1) :
>  No Intrinsic Correlation or Linear Model of Coregionalization found
> Reason: ranges differ”
> 
> and a simulation still happens.  What is going on behind the scenes here when there is no acceptable model of coregionalization or intrinsic correlation?  Do I only want to simulate using a constant range for all my 
> 
> Main Question:
> ----------------------------
> In the SGSim step, how can I apply pre-existing data which can be used as conditional in the simulation process?  Based on the ids present in the gstat object,  I will have a data set of mu0, and conditional on this I want to simulate “psv”, “vrup”, and “slip”.  I have tried putting the data in the sim.fit object as well as in newdata and it doesn’t seem to make a difference.  What am I doing wrong?
> 
> Thanks in advance!
> 
> Best,
> - William Savran
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 

Your script is quite messy and incomplete (where do sim, var, and xy
come from? Does the "If I specify..." comment really relate to the
predict call? Which command generates which warning?), and is not
reproducible.

The way you specify conditioning data is correct, so it is the basis for
conditioning your simulations. The warnings should not be ignored,
though: have you tried plotting the variograms & cross variograms with
the (wrongly?) fitted model?

When you give this little insight in what you do, I can only reply with
wild guesses.
-- 
Edzer Pebesma
Institute for Geoinformatics  (ifgi),  University of Münster
Heisenbergstraße 2, 48149 Münster, Germany; +49 251 83 33081
Journal of Statistical Software:   http://www.jstatsoft.org/
Computers & Geosciences:   http://elsevier.com/locate/cageo/
Spatial Statistics Society http://www.spatialstatistics.info

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: OpenPGP digital signature
URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20160606/256f3f4e/attachment.bin>