Optimising coastal tide predictions: an ensemble satellite altimetry and optical remote sensing approach

Accurate insights into sea-level dynamics are crucial for coastal Earth Observation (EO) analysis. Global ocean tide models play an important role in hind- and future-casting sea-level predictions and gap-filling sparse tide gauge observations, making them a valuable input for large-scale coastal EO workflows. However, existing tide models vary greatly in local performance, particularly in dynamic and complex coastal regions. Satellite altimetry has previously been used to evaluate tide model performance offshore of challenging coastal regions. However, tidal dynamics at the coastline can differ significantly from deeper waters, making it difficult to extrapolate model selection from offshore altimetry observations alone. To support the next generation of coastal EO applications, there is a critical need for new scalable methods to assess and identify optimal tide models at the coastline itself. We present a novel EO-based ensemble tide modelling approach that optimizes coastal tide predictions by leveraging local model performance. This new method combines sea levels from satellite altimetry with an innovative optical remote sensing approach that compares time-series correlations between satellite-derived Normalized Difference Water Index (NDWI) and modelled tide heights. By quantifying and ranking how effectively each of 10 leading global tide models explain observed patterns of tidal inundation, we generate EO-guided ensembles that combine multiple models into a single optimized output. When applied at continental-scale across Australia’s diverse tidal environments, the ensemble approach consistently outperforms any individual model when validated against independent tide gauge observations. Our method is based on freely available optical and altimetry satellite data and open-source tools. The approach runs efficiently at scale, making it suitable for large-scale coastal EO applications that require accurate tide modelling across data sparse and poorly modelled coastal regions.