A Cross-Calibration Approach for Coastline Extraction With Time Series Satellite Images

In tidal flat regions, the spectral differences between water and land are often weak, making it challenging to extract the coastline from remote sensing imagery. Due to tidal influences, the satellite images captured at different times have different water coverage in nearshore area. Images at high tide typically display more distinct spectral characteristics between water and land. This study proposes a coastline optimization approach that utilizes isobath extracted from high-tide image to optimize the coastline extracted from low-tide image, thereby addressing the problem of weak spectral separability. The isobath extracted from bathymetric inversion map of high-tide in steep slope areas is called the relative 0-meter isobath. The steep slope identification algorithm (SSI-A) is employed to detect these areas. The relative 0-meter isobath is identified in these areas. Then it is utilized to optimize coastline at a moment of the ebb tide, enhancing the accuracy and consistency of coastline extraction. Taking Jibei Island on the northern part of the Penghu Islands as an example. Based on different sensor images captured at different times, the proposed approach is used to optimize the coastline extracted by NDWI and SVM. The proposed approach reduces the average error by 13 meters based on Sentinel-2 with 10 m spatial resolution, and by 18 meters based on Landsat 8 and EnMAP image with 30 m spatial resolution. This approach provides more reliable data support for nearshore topographic evolution studies and coastal monitoring.