Sampling to assess rare tree species poses methodological problems, because they may cluster and many plots with no such trees are to be expected. Systematic adaptive cluster sampling (SACS) was used to sample 3 rare tree species ( Schima wallichii, Daphniphyllum himalayense and Michelia kisopa ) in a forest area of about 40 ha in Nepal. The applicability and efficiency of this method (tested using 2 different unbiased estimators, respectively derived from the Hansen-Hurwitz (HH) estimator, which is based on partial selection probability, and the Horvitz-Thompson (HT) estimator, which is based on partial inclusion probability) was checked and compared with that of conventional systematic sampling. The comparison showed that using SACS could increase the efficiency of density estimation by 400-500% for the clustered species S. wallichii (average group size 4), but reduced it by 40% for the unclustered species D. himalayense (average group size 2). For S. wallichii the HT estimator was most efficient, and for D. himalayense the HT and HH estimators gave the same result. For M. kisopa (group size 1), all 3 methods gave the same results. The SACS method was, therefore, more efficient for larger groups of individuals of a rare species than for extremely small groups. SACS may also be used to establish relationships with spatially referenced variables, but data availability was a constraint. It is concluded that SACS is a promising design for resource managers and survey specialists dealing with rare and endangered species in the context of biodiversity and sustainable forest management.