3. Evaluation of the SimpopLocal Model

The SimpopLocal model exposes 6 free parameters that cannot be set using empirical data. This chapter presents how to evaluate SimpopLocal in spite of these degrees of freedom. A first evaluation establishes whether the model has the capacity to produce acceptable dynamics. To achieve this evaluation, the quality of the simulated dynamics is made explicit using a quantitative analysis. Based on this quantitative evaluation, an automated calibration algorithm is designed using a state-of-the-art multi-objective genetic algorithm. The results show that the model is able to produce acceptable dynamics. A second evaluation exposes the contribution of each free parameter to the capacity of the model to produce these acceptable dynamics. A novel sensitivity analysis algorithm called calibration profile is then applied. The results of this analysis show that the model can be simplified by removing one superfluous mechanism and one superfluous parameter and that all the remaining mechanisms are mandatory in the model and all the remaining parameters can be better constrained by narrowing down their definition domains.