Abstract: Field sampling of aphids on Brassica campestris ssp. chinensis was conducted in the suburbs of Hangzhou from 1990 to 1992. The mean aphid density (m, number of aphids per plant), variance (s²) and the proportion of plants with no more than T defined as tally threshold (=0, 1, 5, 10, 20, 30 and 40 per plant) aphids(P_T), are calculated. There are 57 estimates for Myzus persicae, 66 estimates for Lipaphis erysimi, and 38 estimates for their mixed populations. For each of two aphids and mixed population, the empirical relationship between m and P_T is developed using the parameters estimated from linear regression of ln (m) on In[-ln(P_T)]. The determination coefficients r², prediction variance of m from P_T, Var(m), and sample precision denoted as d obtained from different T values are compared. The results indicate that the optimal T for M. persicae should be 30 when its m≥ 10. and that for L. erysimi should be 10 when its m≥5. However,at low population densities. no adequate T values are found for either of the two aphids. Moreover, no acceptable T is found for the mixed population. It is suggested that the tally thresholds of 30 and 10 respectively for M. persicae and L. erysimi can be used for establishing binomial sampling models under relatively high population densities. The results also show that conventional binomial sampling (0～1) method produces unacceptable low levels of sampling precision, and could not be used in sampling of the two aphids.

Key words: Myzus persicae, Lipaphis erysimi, Brassica campestris ssp.chinensis, Binomial sampling, Precision analysis, survival function, static life table, survival curve, time series prediction, natural secondary forest, dominant population