[R-sig-Geo] boundaries for variogram

[Mark Connolly]

I noticed that autofitVariogram sets boundaries for the call to
gstat::variogram along the lines of (0.35 * diag(bbox(sp)) /100) *
c(2, 4, 6, 9, 12, 15, 25, 35, 50, 65, 80, 100).  Any idea of the basis
of this method?

Mark