Authors: Wenyu Li*, Department of Earth System Science, Tsinghua University, Peng Gong, Department of Earth System Science, Tsinghua University
Topics: Coastal and Marine, Remote Sensing, Global Change
Keywords: Coastline, Landsat imagery, Trend analysis
Session Type: Paper
Start / End Time: 10:00 AM / 11:40 AM
Room: Napoleon B1, Sheraton 3rd Floor
Presentation File: No File Uploaded
Continuous monitoring of coastline dynamics is of crucial importance to the understanding of relative contributions of various potential driving factors behind the long-term coastline change. While a large number of efforts have been made to extract coastline and detect coastline change with remotely sensed data, the temporal frequency and spatial resolution of coastline datasets obtained are generally not fine enough to reflect the detailed process of coastline retreat and/or advance, particularly in coastlines with subtle variability. To overcome these limitations, we developed a method to continuously monitor the dynamics of coastline in Florida at annual and subpixel scales using time-series Landsat data (1982–2017). First, robust indicators were used to indicate the annual “average” location of the dynamic coastline. Due to the complexity of muddy-coast morphology, the annual average location is represented not by the coast “line”, but by the fractional inundated “area” of coastline pixels (pixels where the coastline is located), namely annually inundated area. Second, the annually inundated area of coastline pixels was estimated with a model trained by Random Forest. The retrievals were validated at 100 sites with aerial imagery, and the overall RMSE (root mean square error) is < 8%. Third, the long-term trend for the time series of annually inundated area was derived with a statistical model. This study demonstrates the feasibility of time-series Landsat data in continuous monitoring of coastline dynamic at large spatial scales.