The outcome of an IBM simulation can critically depend on the initial number and state of the model entities. It is therefore important to test different initializations and to carefully document those which were used for the results presented in a publication. Sometimes, dependence on initial conditions is part of the question, for example, if we want to understand whether a small, reintroduced population will establish. More often, however, we are interested in results which do not depend on initial conditions. To achieve this, we can pre-run the model until the distribution of the state variables becomes quasi-stationary and evaluate the model only from this time on.
IBMs may also be driven by environmental variables, which are not generated by the model itself, but read as an input to the model from external files. Annual rainfall, for example, may be calculated by a rainfall submodel, or, if long-term rainfall data are available, they are taken from an input file. All inputs to an IBM must be carefully documented and, if possible, made available.
Was this article helpful?