The population dynamics of Trialeurodes vaporariorum (Westwood) in fields grown with French bean (Phaseolus vulgaris L.) were followed at two different locations in Kenya with the purpose of studying the interplay between local and regional dynamics. The fields were surrounded by two types of window traps which provided data on the relative rates of whiteflies arriving and leaving the fields. These data were supplemented with density estimates of adult whiteflies within the fields obtained by means of pot traps. Whiteflies arrived five days after the crop germinated, causing a quick increase in the proportion of infested plants, which reached 98% of all plants one month after germination, corresponding to ca 6 adult whiteflies per plant. Flight activity had two daily peaks, one around 12 hours and one about 16 hours. Activity level was found to increase with temperature (up to 27.5oC) and solar radiation (up to 0.73 kW/m2) and to decrease with wind speed. Immigration rate at first increased with crop age, but reached a maximum of approximately 1.5 whiteflies arriving per plant per day once the crop had reached maturity. A simple statistical model was fitted to sampling data of the number of whiteflies per plant (Nt). The model has three predictor variables: t which is the number of days since germination of the crop, Co(t) and Ci(t) which are the cumulated catches of whiteflies obtained from the outer and inner side, respectively, of the window traps from day 0 and until day t. The model explained 66% of variation in Nt, but surprisingly the model showed that Nt was positively associated with Ci, but negatively associated with Co. A possible explanation for this counterintuitive result is that Ci reflects trivial (non dispersal) flights, which is likely to be correlated with the population size, whereas Co reflects the emigration rate, which tends to increase when the whitefly population declines as a result of reduced food quality due to aging or overexploitation. If there are no other suitable fields in the vicinity, a fraction of the emigrating whiteflies will return to the original field to be caught at the outer side of the window traps. Such captures thus represent “false” immigrants.

Models are useful tools for understanding and improving biological control of arthropod pests by means of natural enemies. Thus, models can be applied to simulate various scenarios in order to identify optimal control strategies. Although simulations can never replace real experiments, they can often serve as guidelines for choosing relevant field experiments and thereby save a lot of laborious and costly field work. Whereas the processes underlying population dynamics (e.g. dispersal, functional response, mutual interference) can be studied under laboratory conditions, large-scaled experiments in the field or in greenhouses are unsuited for this purpose. Instead such experiments may provide information about the patterns (e.g. spatial distributions of prey and predators) generated by the underlying processes. A major purpose of modeling is to link the patterns to the processes that generate these patterns. Petri-dish and single plant experiments have clearly demonstrated the capacity of predacious mite Phytoseiulus persimilis to feed effectively on the two-spotted spider mite Tetranychus urticae. This quickly leads to reductions in the abundance of prey, followed by a decline in predator abundance and eventual extinction. However, when larger systems, consisting of many hundred plants, are infested with the two mite species, extinction of one or both species seems less likely at the system level, although it may still occur at the individual plant level. The qualitative difference between small and large systems with respect to persistence and extinction risks is attributed to the fact that mites move among plants, but to prove that dispersal per se plays a role for the overall dynamics is hard to demonstrate experimentally. To circumvent this problem, I developed a stochastic simulation model of a greenhouse system that explicitly incorporates within and between plant dynamics. The model is used for analyzing a series of experiments with biological control of spider mites in multi-plant systems. In these experiments, the number of plants as well as their connectivity and the numbers of introduced mites were varied in order to examine whether these factors affect e.g. the predator-prey ratio or the time to extinction of one or both species. In my presentation I will also demonstrate an interactive version of the model (called DynaMite). It allows the user to interfere in the system during a simulation so as to mimic the options a grower has in order to prevent losses and to maximize his profit. Such options include spraying with acaricides, releasing predators, and replanting in substitute of damaged plants. By choosing different control strategies, the user may gradually improve his skills according to the principle of learning by experience. The model can be freely downloaded from http://www1.bio.ku.dk/ansatte/beskrivelse/?id=43077