Design And Application Of A Dynamical Spreading Model For Salt Marsh Vegetation Based On COCA

The salt marsh vegetation is one of the most important components of wetlands and its dynamics influences the ecosystem service function and value directly. It is of great significance to analyze quantitatively information of salt marsh vegetation and construct model for management the estuary and coastal environments and the biodiversity conservation.In this thesis, the characteristic of the cellular automata was firstly analyzed and a conclusion indicated that it was necessary for a CA model to be extended geographically under varying conditions of space-time circumstance in order to make results matched the facts accurately.The notion in this thesis is to integrate Remote Sensing Image Preprocessing, GIS, and Computer vision, to combine Genetic Algorithm, Stochastic models and standard Cellular Automata, to unify the CA cells and remote sensing image pixels, cell neighbors and pixel neighbors, cell rules and nature of the plants respectively and to design Conditions Optimized CA model based on the conventional cellular automata model and vegetation characters of Jiuduansha wetland. Transform rules of COCA were as follows: the Equal Circle Rule, the Condition Optimized Equal Circle Rule, the Probability and Montecarlo Rule, the Probability and Threshold Rule, the Condition Optimized Probability and Montecarlo Rule, the Condition Optimized Probability and Threshold Rule, and GA Optimized Parameter Rule and so on.Being the neonatal shoals, the Zhongsha and Xiasha of Jiuduansha where the saltmarsh vegetation were mainly Phragmites australis community, Scirpus mariqueter community and Spartina alterniflora community were taken as the test sites as its low man-made interference. The tests were carried out in the same data source but different transform rules and in the same transform rules but different data source. The precision was evaluated by every pixel compared and whole map compared. The results from this research indicated that this approach could provide a virtual laboratory and model visualization and the result from Stochastic models matched well with the target image on condition of proper parameter, however the results from GA was not perfect due to lack of enough field data.The simulation and prediction maps were then analyzed in their ecological context.The conclusions were that the vegetation showed a tendency of comparative steady distribution pattern at Zhongsha, while and there was a large area for the Spartina alterniflora community to spread at Xiasha. Finally, a new opinion was put forward for improvement of the model in future study.In conclusion, this thesis has explored a process of making simulation and predictions on the dynamics of salt marsh vegetation based on a conditions optimized CA (COCA) model, and examined the links among data, Stochastic model, and ecological predictions. This study has provided a potential of applying Conditioned Optimized CA model technique to solve this problem.