New method maps tides at much finer local scales using decades of satellite images.

Researchers from DTU Space (Technical University of Denmark), in collaboration with the University of Oxford, Technical University of Munich (TUM), and others, have developed a technique that extracts detailed tidal information directly from satellite photographs of coastlines.

Traditional tide gauges are often spaced 50 km or more apart, leaving large gaps. Satellite altimetry also struggles near the coast.

The new approach treats the beach itself as a giant ruler:

• It tracks the land- water boundary (shoreline position) across thousands of Landsat satellite images spanning more than 40 years.
• By combining the movement of that waterline with the known slope of the beach, researchers convert horizontal shoreline shifts into vertical water- level (tidal) changes.
• This yields tidal estimates every ~100 metres along the coast, far denser than previous methods.

Tide heights are not uniform. In some places they can differ by nearly 1 metre across a single bay only a few tens of kilometres long (for example, New Zealand’s South Taranaki Bight). These local variations matter for flood risk, coastal planning, and navigation.

The method turns what was previously treated as “noise” in satellite shoreline data into a useful signal for high- resolution tidal mapping, especially valuable in sparsely instrumented or remote coastal areas.

The study was published in Communications Earth & Environment (Nature portfolio). Professor Ole Baltazar Andersen of DTU Space is a co- author; Michael Hart- Davis (TUM) is the lead author.

_____________________________________________________________________________________

CoastSat is an open- source Python toolkit (developed by Kilian Vos and colleagues at the University of New South Wales) that automatically extracts time series of shoreline positions from publicly available satellite imagery. It is optimized for sandy beaches and has been widely used for coastal monitoring worldwide.

Core Methodology (Step-by- Step)

Image Retrieval

• Uses Google Earth Engine to download Top- of- Atmosphere reflectance images.
• Supported missions: Landsat 5, 7, 8, 9 (since 1984) and Sentinel- 2 (since 2015).
• User defines a region of interest (polygon) and date range. Images are cropped to that area.

Pre- processing

• Cloud masking (removes cloudy pixels).
• Pansharpening (for Landsat 7/8/9) to enhance spatial resolution.
• Down-sampling or re- projection of bands so all images have consistent resolution (typically ~10- 15 m after processing).
• Filtering based on a cloud- cover threshold.

Image Classification

• A supervised Multilayer Perceptron (neural network) classifier labels every pixel into one of four classes:
  • Sand
  • Water
  • White-water (breaking waves / foam)
  • Other land features (vegetation, rocks, buildings, etc.)
• The classifier was trained on thousands of manually labelled pixels from diverse beaches.

Shoreline Detection

• Computes the Modified Normalized Difference Water Index (MNDWI):
  $$\text{MNDWI} = \frac{\text{Green} – \text{SWIR}}{\text{Green} + \text{SWIR}}$$
• Applies an image- specific Otsu threshold on the MNDWI values (restricted to the sand and water classes) to optimally separate land and water.
• Uses a marching- squares algorithm for sub-pixel resolution contouring of the waterline. This finds the precise sand- water interface rather than just the nearest pixel edge.

Time- Series Generation

• The detected waterline (instantaneous shoreline) is intersected with a series of shore- normal transects (typically spaced every 100 m or user- defined).
• This produces a cross- shore position time series at each transect.
• Optional tidal correction can be applied later using a tide model (e.g., FES) and an estimated beach slope to convert the instantaneous waterline into a fixed-datum shoreline (e.g., mean sea level).

Accuracy

Horizontal accuracy is typically ~8-12 m (root-mean- square error) when validated against ground surveys.

Sub- pixel waterline detection itself is on the order of ~5 m for Landsat.

Suitable for detecting changes larger than ~10- 15 m; smaller variations are generally within the noise.

Key Strengths

Fully automated and scalable to any sandy coastline with available imagery.

Provides multi- decadal records (40+ years in many places).

Open- source and designed to be usable by non-experts via Jupyter notebooks.

Limitations

Best suited to sandy beaches (performance degrades on rocky, muddy, or heavily vegetated coasts).

Affected by cloud cover, image quality, and extreme white- water conditions.

Instantaneous waterlines still contain tidal, wave setup, and runup effects until corrected.

CoastSat is freely available on GitHub (kvos/CoastSat) and underpins large regional datasets, including the Pacific Rim shoreline collection used in the recent beach- scale tidal variation study.

_____________________________________________________________________________________

The new beach-scale tidal method (and the underlying CoastSat shoreline data) is explicitly positioned as a useful tool for studying sea- level rise and coastline change.

Traditional tide gauges are sparse (often tens of kilometres apart) and satellite altimetry struggles close to the coast. The CoastSat- based approach fills those gaps by turning the moving waterline on the beach into a local sea-level record:

• Satellite images capture the instantaneous position of the sand– water boundary.
• Combined with the beach slope, the horizontal movement of that waterline is converted into a vertical water-level (tidal and sea-level) signal.
• Because the Landsat record spans 40+ years (1984- present), researchers can examine both short- term tidal patterns and longer- term changes.

Specific value for sea-level rise research

1. Local tidal characterisation
   Accurate knowledge of local tides is essential for converting raw shoreline positions into a consistent vertical datum (e.g., mean sea level). Without it, tidal “noise” can mask or distort true shoreline trends caused by sea- level rise, storms, or sediment processes.
2. Detecting changes in tides themselves
   As mean sea level rises, tidal amplitudes and phases can change (especially in shallow coastal and estuarine areas). The multi- decadal shoreline time series allow researchers to look for trends in major tidal constituents (such as the dominant M₂ tide) at beach scale, something previously difficult outside of well-instrumented sites.
3. Improved shoreline-change analysis
   Once the tidal component is properly accounted for, the remaining shoreline movement better reflects true morphological change (erosion/accretion) driven by sea-level rise, waves, and human interventions. This is critical for projecting future coastline retreat.
4. Gap-filling in data- poor regions
   Many coastlines lack long-term tide gauges. The method can provide high- resolution (≈100 m) tidal information almost anywhere there are sandy beaches and sufficient satellite coverage, supporting better local sea-level and flood- risk assessments.

Limitations to keep in mind

• Accuracy depends on good beach- slope estimates and image quality.
• It works best on open sandy beaches; complex or rocky coasts are more challenging.
• Long-term trends in the tidal constituents extracted this way still need careful validation against traditional gauges and altimetry because of residual measurement noise and morphological changes.

In short, by treating the beach as a natural “tide gauge,” the method turns an existing global archive of satellite images into a new source of coastal water- level information, useful both for understanding present- day tides and for improving studies of how coastlines respond to rising seas.

_____________________________________________________________________________________

Beach scale tidal variations observed from satellite- derived shoreline time series

Tides can vary significantly even over short distances along the coast. Using more than 40 years of Landsat satellite images (via the CoastSat shoreline dataset), the team mapped tidal variations at ~100-metre resolution along Pacific Rim beaches.

Example results:

• In New Zealand’s South Taranaki Bight (~90 km long), tidal height differed by nearly 1 metre from one end of the bay to the other.
• Around Christchurch, tides in Pegasus Bay were about 40 cm higher than those south of the city near the Rakaia River.

Instead of treating the moving shoreline in satellite photos as noise (the usual approach when studying beach erosion), the researchers treated it as the signal.

By tracking the waterline position and combining it with beach slope information, they converted horizontal shoreline shifts into vertical tidal height changes. This works even in areas without local tide gauges.

The method recovers major tidal constituents (especially the dominant M₂ tide) and matches well with traditional tide gauges, ocean tide models, and newer wide-swath satellite data (e.g., SWOT).

Most existing tide information is either sparse (tide gauges every 50+ km) or coarse near the coast. This approach fills those gaps at beach scale and has implications for:

• Local flood- risk assessment
• Coastal planning and engineering
• Improving tide models
• Studying long- term changes in tidal amplitudes

The full open- access paper is available at the Nature.

Published: Communications Earth & Environment 

DOI: 10.1038/s43247-026-03943-9

Provided: Technical University of Denmark

Authors: Michael G. Hart-Davis, 
Thomas Monahan, 
Kilian Vos & 
Ole B. Andersen 

Abstract

Coastal tidal dynamics play a crucial role in a variety of biogeophysical processes, ranging from compound flooding to sediment transport. Satellite altimetry has revolutionized our understanding of ocean tides, but coastal regions remain challenging for satellite altimetry-based observations. In recent years, important efforts have been made to produce shoreline measurements from satellite optical imagery. These datasets raise the question of whether shoreline observations can resolve tidal dynamics at fine coastal scales. Here, we show that shoreline measurements provide fine-scale insights into tidal characteristics, demonstrating variability at beach scales across the Pacific Rim. We explore how both harmonic analysis and response-based methods can be applied to these data to study a range of tidal constituents. Our results are contrasted with in situ measurements and state-of-the-art models as well as with wide-swath satellite observations. We demonstrate that satellite-derived shoreline measurements are a valuable resource for tidal research, particularly for model validation and the study of tidal constituent variability in the coastal zone.