Chapter 12 Summary and Conclusion

This report presents a new and different approach to modeling the structure and dynamics of interacting populations. The philosophy behind the technique has one guiding principle, that the biology imbedded in the data collected by the field researcher should drive the model. Both data and model have implicit resolution and span. The keys to this approach are the integrity of data, their relevance to the ecological questions, and faithful modelling of their dynamics in the simulation. For example, if the model assumes a random search for prey but the organism displays a very sophisticated search technique, then the model may produce questionable results. In addition, the researcher needs some idea of the various parameters that drive the system, such as temperature and possibly humidity for red scale.

We examined the effects of temperature variation on the dynamics of Aphytis with respect to red scale. As expected, simulations with many Aphytis showed a rapid drop in the red scale population. Further, red scale is rarely driven to “extinction” while Aphytis is more likely to disappear, perhaps because Aphytis and red scale operate under different constraints. The development of red scale is essentially temperature dependent, whereas parasitism by Aphytis is a time dependent phenomenon.

The structure of our model system allows easy implementation of modular event algorithms as they become available. We readily admit that the present implementation simplifies the migration pattern for red scale and search strategies for Aphytis. We are developing more realistic search and migration algorithms that are fast to compute. In addition to Aphytis, other parasites such as Encarsia perniciosi, Comperiella bifasciata contribute to red scale mortality (Forster et al. 1995). Our simulation system can handle multiple species, and we are investigating these complex interactions.

Analysis of these simulations could proceed using methods developed for stochastic Petri nets (Ajmone Marsan et al. 1995; Lindemann 1998; Bobbio et al. 2000), except these methods cannot handle simulations with more multiple non-exponential delay times. Instead we propose to design experiments (Latin hypercubes and response surface methods) by adjusting simulation components and to evaluate model performance using Bayesian approaches to uncertainty analysis (cf. Kennedy and O’Hagan 2001 and references therein).