It has become more and more popular to model nature within the field of computer graphics. One important reason is that computer graphics is being increasingly used in movies to create scenes that do not exist in real life. Computer graphics animation techniques are also used to create a range of special effects in movies. Natural scenes are used in computer animation, computer art, computer entertainment, and for making posters and postcards. These are all growing areas, and therefore the interest in computer modeling of natural phenomena is also constantly growing.

Nature is very complex and is not built up by geometrical forms, such as straight lines, rectangles, circles and right angles that are normally used to create and model objects. This thesis will not discuss any methods for modeling objects which specify all its geometrical shapes in a long list, since this is a very tiresome, time-consuming, and inflexible process. The focus of this thesis is on generative algorithms that can produce realistic models. These generative algorithms are also capable of producing different versions of these models without having to start all over again. L-systems have been successfully used to model trees and plants and flowers. We implement different interpretations of L-systems in this thesis so that they can also be used to model other natural phenomena, such as mountains and clouds.

Generative algorithms require very low bandwidth. For example, highly detailed three dimensional images of natural phenomena can be generated by using only a few control parameters. This property is very useful when high compression is needed, for example over the internet.

This thesis describes different methods of modeling natural phenomena, and compares the realism, modifiability, speed and amount of input parameters. We note that we are not concerned in this thesis with finding a unique generative system for an exact copy of an existing (real-life) object. The goal is to generate reasonable realistic images using a few input parameters.

The four natural phenomena modeled in the thesis are:

1. Wavy ocean reflecting light from the sun.
2. Mountains: how to create a model of everything from hills to Mt. Everest. We present a new method using L-systems.
3. Sky: this is the most challenging area of the four. The reason is that the sky can have so many different appearances, depending on sun elevation, cloud type and observation point. A simplified model is developed in this thesis using L-systems.
4. Plants, such as bushes, flowers and trees.

All techniques will be combined to make a complete, highly detailed and realistic image of a natural scene at the end of this thesis. The total amount of inputs needed is typically less than 10 kb!! A comparison will be made by showing a photograph and a synthetically generated model.