Date Awarded

2001

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Virginia Institute of Marine Science

Advisor

Richard L. Wetzel

Abstract

A computational framework is built and demonstrated which is capable of testing plant growth strategies. The framework consists of Vgrass, a carbon based simulation model of a single Zostera marina plant, and the genetic algorithm (GA). Vgrass is based on published seagrass models, published photosynthetic data, and general plant physiology information. The model grows individual leaves whose initiation times are based on degree-day intervals. Leaf size is computed and combined with shoot density to compute population self shading. Leaf length is an emergent property since leaf growth is limited by light attenuation caused by self shading. The model is able to show the relationship between leaf size and shoot density in response to light availability. Degree-days is also shown to be an effective method in modeling the emergence of Zostera marina leaves. The GA and Vgrass are combined to demonstrate the GA as an optimization method and to demonstrate a secondary sensitivity analysis. In an optimization exercise, the RMS error between Vgrass biomass and that of another published model is minimized. Solutions with fitness ranking within 10% of the smallest RMS error are compared in a secondary sensitivity analysis. The analysis can be used to indicate parameter sensitivity in regards to the models ability to attain the optimization goal. Plant growth strategies are tested by searching for configurations of Vgrass parameters best able to: maximize relative growth rate, maximize biomass, and maximize net primary production. Configurations found by the GA lead to plant growth patterns that are not biologically realistic; plant growth strategies based on maximizing "growth" lead to unrealistic plant growth. The plant growth patterns from each of the tests are discussed in relation to ecological and economic principles. Configurations found by the GA search are unique to the optimization goal and the resulting plant growth patterns are shown to support the given goal. Therefore, the computational framework is shown to be successful in testing plant growth strategies. Further, this study shows that care must be taken when defining the fitness function and that the GA is and that the GA is an effective tool at finding "holes" in a model.

DOI

https://dx.doi.org/doi:10.25773/v5-sv9a-rj53

Rights

© The Author

Share

COinS