To evaluate the efficiencies of different sampling methods for a rare and clustered population, we investigated the sampling effects for the two species Tamarix chinensis (Salt cedar) and Elaeagnus angustifolia (Russian olive) in western Inner Mongolia with two-stage sequential sampling, which is a new sampling method, traditional simple random sampling and two-stage sampling. Based on two-stage sequential sampling and two-stage sampling, each population was partitioned into four primary sampling units, and then two of them were randomly selected. Sampling designs were simulated based on the conditions of five secondary sampling unit areas, two criterion values, five initial secondary sampling units and two sequential secondary sampling units in 1000 repetitions. To evaluate the performance of the sampling designs for each method, the variance and relative error of the density estimates were used. The relative sampling efficiencies of the three sampling methods were compared using the same final sampling sizes. We analyzed the sampling efficiency generated by two-stage sequential sampling and found that it yielded smaller variances than those of simple random sampling and two-stage sampling in all sampling designs, and that two-stage sampling was more efficient than simple random sampling. Density estimates from the two-stage sequential sampling were very close to the true values. We also determined the optimum secondary sampling unit areas for the two species in the two-stage sequential sampling. It was best for Tamarix chinensis and Elaeagnus angustifolia when the secondary sampling unit areas were 200 and 100 m2 , respectively.
