The Spatio-temporal Dynamic Analysis of Salt Marsh Vegetation in Chongming Dongtan Based on Remote Sensing Data

In recent 15 years, the biodiversity of Chongming Dongtan national nature reserve has been dramatically reduced by invasive plants, especially by spartina alterniflora. How to obtain and monitor spatio-temporal change of spartina alterniflora has great practical significance in managing and protecting Chongming Dongtan. Therefore, the main purpose of this paper was to build up a method to recognize different species of salt marsh vegetation and analyze the spatio-temporal change by interpreting remote sensed data of different period. Considering the complexity of different plant' spectral feature, the feature space was consisted of Normalized Different Vegetation Index (NDVI), Kanth-Thomas transformation (K-T transformation) and an optimal band combination. In order to enhance classification accuracy, a dual-weight support vector machine (SVM) classification model by weighting on different features and classes was proposed. The kernel function of this model was selected as wavelet which satisfied the practical situation better. During the experiment, training and validation data were ground survey points located by GPS. The results from this study indicated that the invasion of spartina alterniflora was serious and it kept expanding to the southern. Furthermore, the method proposed in this paper could get higher detection accuracy than the traditional methods and was suitable for the small-sample experiment. Consequently, using this approach could provide timely analytical data for the time and space distribution change of the Salt Marsh Vegetation in intertidal zones, meanwhile the results will provide sound scientific basis for carrying out some proper management and control of Spartina alterniflora.