These are the parameters for doing "Block Kriging".
gstat, so the help pages for
gstat::krige tell you what the parameters mean in slightly more detail that
block: block size; a vector with 1, 2 or 3 values containing the
size of a rectangular in x-, y- and z-dimension respectively
(0 if not set), or a data frame with 1, 2 or 3 columns,
containing the points that discretize the block in the x-, y-
and z-dimension to define irregular blocks relative to (0,0)
or (0,0,0)-see also the details section of predict. By
default, predictions or simulations refer to the support of
the data values.
According to: https://www.publichealth.columbia.edu/research/population-health-methods/kriging-interpolation
"Block kriging, which estimates averaged values over gridded “blocks” rather than single points. These blocks often have smaller prediction errors than are seen for individual points."
but I'm struggling to find a reference on how you'd justify block kriging against simple point kriging, or how you'd choose the block size (except maybe by some cross-validation technique).