The Modelling of Natural Phenomena

The main activity of physical sciences is to establish models of reality which can be used for measurements, predictions of new phenomena, and possibly for the falsification of current theories. Modelling in computer graphics has many of these characteristics, and this makes it a particularly interesting activity. The modelling of natural phenomena is especially challenging, because these are complex , numerous and familiar. I propose in this paper a taxonomy of the types of models which have been used in computer graphics, with an analysis of their good and bad qualities, and illustrative examples. I distinguish empirical. physical. structural . morphological. impressionist and selfmodels. In addition to models, some basic principles and concepts have proven to be useful for the effective modelling of natural phenomena. I review a somewhat arbitrary list: database amplification, levels oj details, stochastic elements, time in models, particle systems, textures and fractals, and illustrate with some examples.