Automated mapping of global 30 m tidal flats using time-series Landsat imagery: algorithm and products

Tidal flats are an important part of coastal ecosystems and play an important role in shoreline protection and biodiversity maintenance. Although many efforts have been made in tidal flat mapping, an accurate global tidal flat product covering all coasts globally is still lacking and urgently needed. In this study, a novel method is proposed for the automated mapping of global tidal flats at 30 m (GTF30) in 2020 based on the Google Earth Engine, which is also the first global tidal flat dataset covering the high latitudes (>60°N). Specifically, we first propose a new spectral index named the LTideI index through a sensitivity analysis, which is robust and can accurately capture low-tide information. Second, globally distributed training samples are automatically generated by combining multisource datasets and the spatiotemporal refinement method. Third, the global coasts are divided into 588 5°×5° geographical tiles, and the local adaptive classification strategy is used to map tidal flats in each 5°×5° region by using multisourced training features and the derived globally distributed training samples. The statistical results show that the total global area of tidal flats is about 140,922.5 km 2 , with more than 75% distributed on 3 continents in the Northern Hemisphere, especially in Asia (approximately 43.1% of the total). Finally, the GTF30 tidal flat dataset is quantitatively assessed using 13,994 samples, yielding a good overall accuracy of 90.34%. Meanwhile, the intercomparisons with several existing tidal flat datasets indicate that the GTF30 products can greatly improve the mapping accuracy of tidal flats. Therefore, the novel method can support the automated mapping of tidal flats, and the GTF30 dataset can provide scientific guidance and data support for protecting coastal ecosystems and supporting coastal economic and social development. The GTF30 tidal flat dataset in 2020 is freely accessible via https://doi.org/10.5281/zenodo.7936721 .