Adaptive cluster sampling is shown to be a viable alternative for samp
ling forests when there are rare characteristics of the forest trees w
hich are of interest and occur on clustered trees. The ideas of recent
work in Thompson (1990) have been extended to the case in which the i
nitial sample is selected with unequal probabilities. An example is gi
ven in which the initial sample of trees is selected with probability
proportional to tree basal area. If a characteristic of interest is ob
served on a sample tree, additional trees within a fixed distance of t
he sample tree are also included in the sample.