In inventories that guide land management, point sampling is heavily u
sed to estimate land area occupied by a population, area by condition
class, population mean density, and totals of population attributes. O
ften, for reasons of efficiency or interpretation, the attributes of t
he principal sample point are described by aggregating information ove
r a group of satellite points arrayed in a geometric pattern around th
e principal point. This group of data sampling points forms what is ca
lled the ''support'' for the principal point. Problems arise when the
principal point is close enough to a population (or subpopulation) bou
ndary for some of the support points to fall outside that boundary. Ad
hoc solutions to this problem that involve replacing some or all of t
he points in the group may introduce bias into the inventory estimates
. Unfortunately, procedures that do not replace points introduce bias
into estimates of attributes that can be defined only with respect to
the configuration and volume of the support. Examples of such attribut
es include distribution of area by classes of tree density, stocking,
species composition, stand structure, and the ecological frequency and
constancy of a species. A rule has been developed for reshaping the p
attern of support points that resolves the dilemma for single, straigh
t boundaries. Simulations to evaluate the rule show that it produces a
uniform density of points with respect to distance from the boundary
while maintaining a compact region of support for the principal point.