Resolving uncertainty in satellite-derived shorelines of a reef-lined beach using high spatiotemporal resolution topographic surveys

Abstract

Automated satellite-derived shoreline (SDS) detection is increasingly used for coastal monitoring, yet few high spatiotemporal resolution datasets exist for detailed error analysis. This study compares SDS from Landsat 7, Landsat 8, Sentinel-2, and PlanetScope against 3.5 years of weekly orthomosaic and digital elevation model surveys of a 500 m reef-lined beach in Waikīkī, Hawai‘i ( n = 128 surveys). We evaluated error mitigation strategies including image co-registration, outlier removal, and water-level corrections (tides and wave set-up). SDS accuracy improved substantially after these corrections, with root mean square errors (RMSE) ranging from 2.83 m (PlanetScope) to 5.54 m (Sentinel-2). Our analysis showed that satellite-derived waterlines (SDWs) do not consistently align with the same geomorphic feature (e.g., waterline, low water mark). Mid-resolution satellites aligned with the waterline in only ~ 50% of cases, whereas higher-resolution PlanetScope imagery (3.7 m) aligned consistently in 97% of cases. These findings reveal that SDS can represent varying shoreline proxies, introducing an under-recognized uncertainty that higher-resolution imagery can help resolve. We demonstrate that high-resolution imagery and robust error mitigation are useful for improving the reliability of automated shoreline detection.