TOWARD PREDICTABLE BIOLOGICAL-CONTROL OF LIRIOMYZA-TRIFOLII (DIPTERA,AGROMYZIDAE) INFESTING GREENHOUSE CUT CHRYSANTHEMUMS

Implementation of augmentative biological control is often hindered by the inability to obtain accurate information on natural enemy release rates that will ultimately yield a salable crop within economic const raints. The purpose of this study was to develop and evaluate a comput er model that attempts to predict augmentative releases of the parasit oid Diglyphus begini (Ashmead) (Hymenoptera: Eulophidae) to use for bi ological control against the serpentine leafminer Liriomyza trifolii ( Burgess) (Diptera: Agromyzidae) infesting greenhouse-grown cut chrysan themums. The model attempts to determine the appropriate release rate necessary to reduce leafminer densities below 1 larvae per 1,000 chrys anthemum leaves within 40 d after planting, after which time aesthetic ally important foliage forms on chrysanthemums. In the construction of this deterministic model, three simplifying assumptions have been mad e: (1) there is no insect movement between the greenhouse and outside, (2) greenhouse temperatures are a constant 27-degrees-C, and (3) the greenhouse represents a homogeneous environment. Model simulations ind icated that successful biological control was unlikely when parasitoid releases are initiated later than 14 d after planting regardless of t he release rate. The simulations also demonstrated that D. begini rele ase rates are not linearly related to L. trifolii densities, and that a parasitoid release strategy is not simply the determination of a pro per wasp/leafminer ratio. Validation studies of the model's prediction s provided mixed results. Following the release rates generated by the model, L. trifolii larval densities were not significantly greater th an 1 per 1,000 leaves 40 d after planting. However, the model did not always track the succession and magnitude of leafminer population fluc tuations with statistical accuracy. Two factors probably contributed t o these errors in prediction: (1) the assumptions inherent to the mode l were not met during the validation trials, and (2) the leafminer sub routine of the model could not accurately predict L. trifolii densitie s in the absence of D. begini. In spite of these errors, leafminer dam age to the harvested foliage was significantly lower in the treatments receiving D. begini releases than in the control treatments. In addit ion, when the model was tested in a commercial cut chrysanthemum green house, L. trifolii was successfully controlled by releases of D. begin i, resulting in the production and harvest of a salable cut chrysanthe mum crop without the use of any pesticides. Finally, postvalidation re construction of the model, to include temperature-dependent variation in life history parameters, increased the predictive power of the mode l.