Point and Block Kriging Prediction

DESCRIPTION:

Computes point or block kriging predictions and standard errors at locations in newdata using an object returned by krige.

USAGE:

predict.krige(object, newdata, se.fit=T, grid=<<see below>>, 
              blocksize=c(1, 1), nxy=c(1, 1)) 

REQUIRED ARGUMENTS:

object
an object of class "krige" as returned by the function krige.

OPTIONAL ARGUMENTS:

newdata
a data frame or list containing the spatial locations for the predictions. The names must match the names of the locations used in the call to krige (see attr(object,"call")).
se.fit
a logical value, if TRUE, the standard errors of the predictions are returned. Currently the standard errors are always computed internally. This se.fit only determines if the returned data frame includes the se column.
grid
a list containing two vectors, the names of the vectors must match the names of the locations used in the call to krige. The vectors are each of length 3 and specify the minimum, maximum and number of locations in that spatial coordinate, respectively. A grid is then computing using expand.grid. The default value is to use the range of the original location data for the minimum and maximum, and 30 points. This argument is ignored if newdata is supplied.
blocksize
for block kriging, a numeric vector of length 2 specifying the size of the block in x (first value) and y (second value) direction. The locations specified by newdata or grid are at the center of the blocks.
nxy
for block kriging, a numeric vector of length 2 specifying the number of discretization points inside the block. If both values are set to 1 (the default) then point kriging predictions are computed.

VALUE:

a data frame where the first two columns are the locations of the prediction along with:
fit
the predicted values.
se.fit
the standard error of the prediction. Only included if se.fit = TRUE.

DETAILS:

This function is a method for the generic function for class krige. It can be invoked by calling for an object x of the appropriate class, or directly by calling regardless of the class of the object.

REFERENCES:

Ripley, Brian D. (1981). Spatial Statistics. Wiley, New York.

SEE ALSO:

, .

EXAMPLES:

# krige the Coal Ash data 
kcoal <- krige(coal ~ loc(x, y) + x + x^2, data = coal.ash, 
        covfun = spher.cov, range = 4.31, sill = 0.14, nugget = 0.89) 
# predictions over default 30 x 30 grid 
pcoal <- predict(kcoal) 
# plot prediction surface 
wireframe(fit ~ x * y, data = pcoal, 
        screen = list(z = 300, x = -60, y = 0), drape = T) 
# block kriging predictions with block of size 2 x 2 at 4 locations 
predict(kcoal, data.frame(x=c(4,5,9,11), y=c(7,13,9,18)), 
        blocksize=c(2,2), nxy=c(5,5))