krige function for interpolation - methodology

Hi,
I used the krige function of gstat to obtain predicted value for the grid
points using block ordinary kriging. I defined my grid  spacing of 100m
with block size of 10m x 10m using 100 nearest obervation in the local
neighborhood of unsampled locations for prediction.

I do not understand how this krige function will capture the number of
nearest observations in the local neighborhood for prediction.

 Could anyone please explain how it will capture the 100 obervations for
the prediction? does this function define the search radius to capture
nearest observation for prediction? Even though I define my nmax to be 100
(nmax=100), is it possible that number of observations less than hundred
can be used if the unsampled location is within sparse region of the field?


prd_ok <- krige(id="yield",yield ~ 1, canmod.sp, newdata = grd,
model=exp.mod,
block=c(10,10),nmax=100)

Thanks

Moshood