Xu, Chao

Tidal flat refers to the sediment-rich environment along the seashore, which is alternatively exposed or inundated during tidal cycles. It is widely recognized as not only the sentinel of coastal environment change, but also the safeguard for beachfront communities. It is necessary to comprehensively understand the wellness of tidal flat environments, especially for the United States (US), which has the eighth longest coastline throughout the world. Aiming at the dynamics of tidal flats, this dissertation firstly proposed a monitoring framework from three levels, including the pixel, object, and lifecycle. In addition, eleven events were defined to describe the dynamic activities throughout the lifecycles, which were captured, represented, and analyzed by utilizing graph theory. The Everglades in the southeastern corner of Florida Peninsula was selected to test this approach, which verifies an effective way to track, represent, and analyze the dynamic activities of tidal flats. Secondly, this dissertation mapped the distributions of tidal flats in the conterminous US, which provides a reliable dataset on a large spatiotemporal scale for future use. A random forest classification model was proposed, which uses 30 predictor variables to describe the spectral change patterns between the satellite images acquired in subsequent time steps. On the other hand, a total of 58,735 ground truth samples were collected under five classes, including permanent water, tidal flats, barren grounds, vegetated lands, and artificial surfaces. These sample points were randomly divided into two parts: 80% of them were used to train the random forest model, and the rest 20% were used to validate the results.