##
## A note on the kriging with external trend 
##

To perform " kriging with external trend" and/ou "universal kriging"
you need the value of the covariates on both: 
data locations and prediction locations.

Therefore in geoR implementation you use:
- "trend.d" to specify the trend values at data locations
- "trend.l" to specify the trend values at prediction locations

For example to perform a kriging 
with a 2nd degree polynomial + a covariate 
you will need the values of this covariate at both, 
data and prediction locations.

Consider the following example having the objects:

"foo"   : is a geodata object with a covariate called "foocov"
"gr"    : is a matrix with coordinates of the prediction locations
"grcov" : is the covariate values at the prediction locations

Now assume you are performing kriging with 2nd degree polynomial trend
plus the covariate.
The code would be something along the lines:

> kc <- krige.control(obj.m=fit,
                   trend.d = trend.spatial("2nd", geodata=foo, 
                                           add = ~foocov),
                   trend.l = trend.spatial("2nd", geodata=list(coords=gr),
                                           add = ~grcov))
> kc <- krige.conv(foo, loc=gr, krige=kc)