Island Karst Classification: Spatial Modeling-Oriented Approach with Multispectral Satellite Imageries
Ho, Hung Chak
This project developed a series of spatial models to classify the island karst landforms and predict the island karst feature distribution. Spatial models with unsupervised classified images, and fuzzy-based spatial models were used in this study. Forecasting verification and spatial regressions were used to validate the models. The case study was conducted on San Salvador Island, the Bahamas, a recognized carbonate island with island karst features. Fieldwork data on banana holes on the island were used for model validation. The results showed that most models had accuracy higher than 90%, and were statistically proved that they could be used as predictors of island karst features. Further study may be conducted to solve the Modified Areal Unit Problem (MAUP) in the future.