Abstract: To address the issues of a great variety and a high degree of morphological diversity for a plant root system, this paper proposes an editable modeling method to simulate root growth. This method extends the traditional L-system to describe the root structure and topology with a semantic feature rule, and achieves a natural growth simulation according to a continuous equation of root stem radius and single root length. The global morphology of the root system is controlled by guiding vectors, and a variety of measures are provided to edit the local shape of the root stem to generate a natural-looking root model, such as stem skeleton, surface shape, irregular surface, etc. In the experiment, this paper simulates six species of plants in nature, including four different types of tap roots, fibrous roots, storage roots and plate roots. The average similarity between the simulated root system and the real picture is up to 79.16%. The experimental results show that the method has the ability to model many types of root systems, and it is possible to build root models with special morphology by setting the shape guidance curve or adjusting some parameters in a realistic way.