A particle swarm optimisation model for tracking multiple peaks in a continuously varying dynamic environment is described. To achieve this, a form of speciation allowing development of parallel subpopulations is used. The model employs a mechanism to encourage simultaneous tracking of multiple peaks by preventing overcrowding at peaks. Possible metrics for evaluating the performance of algorithms in dynamic, multimodal environments are put forward. Results are appraised in terms of the proposed metrics, showing that the technique is capable of tracking multiple peaks and that its performance is enhanced by preventing overcrowding. Directions for further research suggested by these results are put forward.