Huixuan Li

Date of Award

Fall 2021

Document Type

Open Access Dissertation



First Advisor

Cuizhen Wang


Coastal wetland mapping is often difficult because of the heterogeneous vegetation compositions and associated tidal effects. Past studies in the Gulf/Atlantic coast states have reported acute marsh dieback events in which marsh rapidly browned and thinned, leaving stubble of dead stems or mudflad with damaged ecosystem services. Reported marsh dieback in South Carolina (SC), USA, however, have been limited. Previous studies have suggested a suite of possibly abiotic and biotic attributes responsible for salt marsh dieback. However, there are no consensus answers in current literature explaining what led to marsh dieback in past decades, especially from the spatiotemporal perspective. In this study, the U-Net was employed, and an adaptive deep learning approach was developed to map statewide salt marshes in estuarine emergent wetlands of SC from 20 Sentinel-2A&B images. Then all marsh dieback events were identified in the North Inlet-Winyah Bay (NIWB) estuary, SC, from 1990 to 2019. With 30 annually collected Landsat images, the Normalized Difference Vegetation Index (NDVI) series was extracted. A Stacked Denoising Autoencoder neural network was developed to identify the NDVI anomalies on the trajectories. All marsh dieback patches were extracted, and their inter-annual changes were examined. Among these were the five most severe marsh dieback events (1991, 1999, 2000, 2002, and 2013). The spatiotemporal relationships between the dieback series and the associated environmental variables in an intertidal marsh in the estuary were investigated. Daily Evaporative Demand Drought Index (EDDI), daily precipitation data from Parameter Elevation Regressions on Independent Slopes Model (PRISM), and station-based water quality observations (dissolved oxygen, specific conductivity, salinity, turbidity, pH, and temperature) in the estuary were retrieved. This study cogitates the environmental influence on coastal marsh from a spatiotemporal perspective using a long-term satellite time series analysis. The findings could provide insights into marsh ecological resilience and facilitate coastal ecosystem management.

Included in

Geography Commons