Automated Shoreline Extraction Using Planet Imagery and Machine Learning Techniques in Google Earth Engine.

The study introduces a novel method of automating high-resolution shoreline mapping using commercial grade satellite imagery (PlanetScope) and Google Earth Engine, an open-source, cloud-based geospatial platform. This reproducible technique will improve upon prior shoreline monitoring techniques which used the much coarser resolution Landsat imagery (30 m/pixel). Planet's high spatial resolution (3 m/pixel) and daily acquisition allows researchers to better monitor shoreline changes in response to climate change (sea level rise) and facilitate response to coastal changes caused by natural hazards (tsunami, storms, etc.). The machine learning (ML) classifier is trained to identify the boundary between the dry supratidal and wet intertidal sediment, a common shoreline indicator known as the high water line (HWL). Ultra-high resolution Unmanned Aircraft System (UAS) imagery were acquired to generate products for assessing the accuracy of the supervised classification and the shoreline extracted from Planet imagery. UAS flights were kept at an altitude of 30 meters to obtain a ground sampling distance of 1 cm/pixel. Ground control points (GCP) were placed across the study area to obtain coordinates using a Real Time Kinematic-GPS. The GCPs were used to georeference the raw imagery in order to generate Digital Surface Models (DSMs) and orthomosaics for our study area (Carpinteria State Beach, Crystal Cove State Beach and Zuma State Beach, California) using Structure-from-Motion (SfM) photogrammetric techniques. The resulting orthomosaic will be classified to assess performance of the ML classifier and the DSM will be used to derive the Mean High Water (MWH) line to assess the accuracy of the Planet HWL shoreline position. High-resolution PlanetScope imagery and GEE will facilitate the process of monitoring shoreline changes by allowing a better assessment of tsunami/storm-induced shoreline changes and accurately identifying at-risk coastal communities.