Predictability and Measurability ofBacillus thuringiensisEfficacy Against Spruce Budworm (Lepidoptera: Tortricidae)

Abstract

A detailed simulation model was used to examine the unpredictability of the efficacy of aerial applications of Bacillus thuringiensis against spruce budworm, Choristoneura fumiferana (Clemens). The results of new field trials were used to further validate the simulation model. Simulations indicated that mortality caused by B. thuringiensis in the first 48 h after a spray, total mortality, and final defoliation could be predicted by multiple regression (r2 > 0.82) using spray parameters (potency, droplet density, and diameter), budworm population age, density, and mortality from causes other than B. thuringiensis, exposure-period duration, and postspray temperature as predictors. The proportion of budworm larvae that had ingested a lethal dose within 48 h after a spray was a good predictor of total mortality (r2 > 0.91). Variation in temperature contributed little to variation in efficacy measured as population reduction and foliage protection indices. However, sampling error in estimates of population density, final defoliation, and 48-h mortality considerably reduced the predictability of efficacy. Coefficients of determination diminished from r2 > 0.8 with no sampling error to r2 < 0.05 when sampling error was 30% of the sample mean. There was also a pronounced negative bias in estimates of foliage protection and population reduction when sampling error exceeded 20% of the mean. Mismatching control and treated populations with respect to the magnitude of extraneous mortality during the 6th instar caused large biases in efficacy estimates. It is recommended that the use of population reduction and foliage protection indices be avoided as much as possible because of the large effect of sampling error on their values. Rather, the use of well replicated, homogeneous treatment plots carefully matched with nearby untreated controls is recommended. Treatment efficacy can also be assessed with the model of Cooke and Régnière (1996) in combination with sampling aimed at estimating stage-specific survival rates in treated and untreated plots.