[R-sig-Geo] kriging variance vs sample spacing

[Kerry Ritter]

Hi I want to compute the kriging variance as a function of grid spacing 
so that I can measure cost vs. kriging variance in deciding on an 
appropriate grid spacing for an ocean sediment surrounding an outfall.  
 From an earlier study I have a response with an anisotropic variogram 
given by model=vgm(0.37108, "Gau", 4500, anis = c(0, 0.75)).  The 
sampling area is a polygon.  To compute this curve, I have created a 
series of shapefiles containing the polygon with different grid 
spacings.  I want to compute the kriging variance for each grid spacing 
so I can plot it on the y-axis and grid spacing on the x-axis.

   I have tried the following code:  I have been using the library gstat 
and geoR

names(SD1000)                      # shape file that contains a 1000 x 
750 sampling grid (in therms of UTM coordinates) - this is what I wan to 
find the associated kriging variance
g.dummy <- gstat(formula = z~1, locations= ~lon_UTM+lat_UTM, dummy = 
TRUE, beta = 0,
     model = vgm(0.37108, "Gau", 4500, anis = c(0, 
0.75)))                            #variance model from earlier survey
BRI1000 <- predict(g.dummy,newdata=SD1000, nsim = 1)
BRI1000                                                                
                   #sampling grid of 1000 x 750 simulated values


SD100.spdf=SD100                                                             
# shape file that contains a dense grid resolution (153 x 100m) polygon
coordinates(SD100.spdf)=~lon_UTM+lat_UTM
class(SD100.spdf)
head(SD100.spdf)


vgm.aniso7=variogram(sim1~1,BRI1000.spdf, 
alpha=c(0,90,180,270),boundaries= c(0,300,900,1300,1800,1900,
     2300,2800,3000,3500,4000,4500,6000,10000,15000),cressie=TRUE)

vgmaniso.fit7=fit.variogram(vgm.aniso7,vgm(0.37108,model="Gau",range= 
4500,anis=c(0,0.75)),fit.sills=c(FALSE,FALSE),fit.range=c(FALSE,FALSE),fit.method=7) 


vgmaniso.fit1V.pr=krige(sim1~1,BRI1000.spdf,SD100.spdf,vgmaniso.fit7)
summary(vgmaniso.fit1V.pr)

Coordinates:
             min     max
lon_UTM  464500  476500
lat_UTM 3606804 3625196
Is projected: NA
proj4string : [NA]
Number of points: 9539
Data attributes:
    var1.pred          var1.var
  Min.   :-1.6710   Min.   :3.270e-11
  1st Qu.:-0.7611   1st Qu.:1.736e-08
  Median :-0.3351   Median :1.002e-07
  Mean   :-0.2340   Mean   :2.711e-05 <------- Kriging variance????
  3rd Qu.: 0.3641   3rd Qu.:1.098e-06
  Max.   : 1.4195   Max.   :7.022e-03


Thank you in advance for your help.
-Kerry

-- 
**********************
Kerry Ritter, Ph.D.
statistician
Southern California Coastal Water Research Project
3535 Harbor Blvd., Suite 110

work: 714-755-3210
cell: 714-420-3346
fax:  714-755-3299

email: kerryr at sccwrp.org