News
Selecting behavioural traits: assessing the complexity of the model
ESR Update - Since an ecological model can be considered as a caricature or a simplification of a system, a modeller has to correctly decide on a suitable amount of complexity to be applied to its model. This should depend on the fundamental purpose of the model (generality, realism or precision), but that is easier said than done. Determining the behavioural traits that should be included in my general model might be the most important decision, but simultaneously proves to be the largest point of doubt.
Wouter Plouvier MSc.
I am currently at the final stages of the construction of my first individual based model that will simulate releases of natural enemies in an agricultural ecosystem for biological control. With this model, we hope to identify the most important behavioural features that should be targeted in genetic improvement programs. Furthermore, we like to get insights into the optimal release strategies that will allow for the highest economic returns for the farmers.
My work basically involves sitting in front of a computer the whole day. However, the amount of possibilities that the artificial world provides are mind-blowing, meaning that there is always something new to learn and investigate. Right now, I am constructing a genetic algorithm that will allow for the identification of the most important behavioural features. The design of a genetic algorithm is inspired by Darwinian natural selection and allows for mathematically-based artificial evolution to solve complex problems. It can provide for some really fun experiments, where you can basically witness evolution happening on your computer screen (Video genetic algorithm).
The most important question remains which behavioural features should be implemented in this genetic algorithm, keeping in mind that every feature will complicate your output and make it harder to analyze the exact causes of your results, but at the same time will provide more realism. An equally important question is the amount of release situations that we want to compare, and how we can calculate the costs for the difference in these releases. Does releasing agents in multiple points in the field costs more than at one point? How much does the release of one extra agent cost, in terms of yield generated from a single plant?
Since I mostly work alone on my computer at my desk, I do not have a whole list of people to thank like the previous ESR did. I do want to stress the importance of my supervisors at Wageningen, Bart Pannebakker and Bas Zwaan, for providing valuable feedback and input. And I would be completely lost without the excellent guidance of Eric Wajnberg and Patrick Coquillard. Finally, I would like to thank my colleagues at INRA that drop by from time to time to ease the loneliness of the modelling life. Not complaining though! ;)