Coastal Erosion Monitoring from Satellites: Tracking Shoreline Change

On a field visit to the Sendai coast in 2018, seven years after the devastating 2011 tsunami, I was struck by how dramatically the shoreline had changed — not from the tsunami itself (that was temporary inundation), but from the permanent reshaping of the coastline. Sand had been redistributed, barrier beaches had migrated, and some sections were eroding at rates clearly visible year to year.

Back at my desk, I pulled up the Landsat archive. The same changes were visible from space — and more importantly, I could see the trajectory: how the coast had looked before the event, how it responded immediately after, and how it was evolving in the years since. That time dimension is what makes satellite monitoring invaluable for coastal management.

Why Satellites for Coastal Monitoring?

Traditional coastal surveys — GPS beach profiles, aerial photography, LiDAR — provide high accuracy but cover limited areas at high cost. A GPS survey might cover a few kilometers of beach in a day. Satellite imagery covers thousands of kilometers in a single scene, and the archive stretches back to 1984 (Landsat 5).

What satellites can measure:

• Shoreline position over time → erosion/accretion rates
• Beach width changes
• Sediment plume dynamics
• Coastal habitat changes (mangroves, salt marshes, seagrass)
• Storm impact and recovery

What satellites cannot easily measure:

• Vertical beach profiles (elevation changes)
• Underwater bathymetry changes (limited to clear, shallow water)
• Sub-meter shoreline changes (10m pixel ≈ 3-5m positional accuracy at best)

Extracting the Waterline

The fundamental task is deceptively simple: for each satellite image, determine where the land meets the water.

NDWI/MNDWI Thresholding

The most common approach uses water indices:

• NDWI = (Green - NIR) / (Green + NIR)
• MNDWI = (Green - SWIR) / (Green + SWIR)

Water has positive values; land has negative values. A threshold (typically around 0) separates them, producing a binary land/water mask. The boundary of this mask is the waterline.

MNDWI is preferred for coastal applications because SWIR provides better separation between water and built-up areas (concrete seawalls, harbors) that can confuse NDWI.

Sub-Pixel Waterline Extraction

A 10m pixel that's half water and half sand will have an intermediate NDWI value. Instead of binary thresholding, you can fit a continuous function across the land-water transition and identify the precise position where the water fraction equals 50%.

This sub-pixel approach can extract waterline positions with 3-5m accuracy from 10m Sentinel-2 imagery — a significant improvement over pixel-level extraction.

Machine Learning Approaches

Recent work uses neural networks trained on manually digitized shorelines to extract waterlines directly from imagery. CoastSat, developed at the University of New South Wales, uses a combination of NDWI, classification, and morphological processing to automate shoreline extraction from Landsat and Sentinel-2 imagery.

The Tidal Problem

Here's the challenge that makes coastal monitoring harder than it appears: the waterline position depends on the tide.

A beach with a gentle 1:20 slope and a 2m tidal range will show a 40m difference in waterline position between high and low tide. If you compare a low-tide image from 2010 with a high-tide image from 2020, you'll measure 40m of apparent "erosion" that's actually just tidal variation.

Solutions:

• Tidal correction: Use a tidal model to estimate the water level at each image acquisition time, then correct the waterline position to a reference datum (e.g., mean sea level)
• Composite approach: Use many images per time period and extract the median waterline, which averages out tidal variation
• Same-tide comparison: Select images acquired at similar tidal stages (practical only when you have dense time series)

Long-Term Change Analysis

The real power of satellite monitoring is quantifying erosion rates over decades.

Standard workflow:

1. Extract waterlines from all available images (e.g., every clear Landsat/Sentinel-2 scene from 1984-2026)
2. Apply tidal correction to normalize all waterlines to the same datum
3. Define shore-perpendicular transects at regular intervals along the coast
4. For each transect, measure the waterline position over time
5. Fit a trend line (linear regression) to get the erosion/accretion rate in meters per year

Typical erosion rates:

• Sandy beaches: 0.5-5 m/year erosion is common; >10 m/year indicates severe erosion
• Cliff coasts: 0.1-1 m/year (slower but irreversible)
• Delta coasts: Highly variable; subsiding deltas can lose >50 m/year

Available Global Datasets

You don't need to build this pipeline from scratch:

• Digital Earth Australia Coastlines — annual shoreline positions for all of Australia, derived from Landsat, freely available
• Global Surface Water (GSW) — monthly water occurrence maps from the Joint Research Centre, covering 1984-present
• CoastSat — open-source Python toolkit for extracting shorelines from Google Earth Engine

Monitoring Coastal Habitats

Shoreline position is just one metric. Satellites also track the ecosystems that protect coastlines:

Mangrove loss detection: Mangroves attenuate wave energy and reduce erosion. Sentinel-2 NDVI time series can detect mangrove clearing events and track regrowth. Loss of mangrove buffer often precedes accelerated shoreline retreat.

Salt marsh mapping: Sentinel-1 SAR is particularly effective for mapping salt marshes because C-band backscatter responds to the water-vegetation interaction in intertidal zones. Multi-temporal SAR composites distinguish salt marsh from adjacent land uses even when optical imagery is ambiguous.

Seagrass monitoring: In clear shallow waters, Sentinel-2's coastal aerosol band (B1, 443nm) and blue band (B2, 490nm) can detect submerged aquatic vegetation to depths of 5-10m, depending on water clarity.

Storm Impact Assessment

After major storms, satellite imagery provides rapid assessment of coastal damage:

1. Pre-storm baseline: Most recent clear image before the event
2. Post-storm assessment: First available image after the event
3. Difference analysis: Compare waterline positions, detect overwash, identify barrier island breaches

Sentinel-1 SAR is particularly valuable here because it can image through the cloud cover that typically accompanies major storms. Pre/post SAR coherence maps reveal areas where the surface structure changed — flattened dunes, deposited sediment, destroyed infrastructure.

Climate Change Context

Global sea level rise (averaging ~3.4 mm/year since 1993, with the recent rate exceeding 4.5 mm/year and accelerating) doesn't directly cause the dramatic erosion visible in satellite imagery — local factors like sediment supply, wave energy, and human intervention dominate at decadal scales. But sea level rise shifts the baseline: higher average water levels mean storm surges reach further inland, and beaches that were stable under historical conditions may begin eroding.

Satellite monitoring quantifies these trends objectively, providing the evidence base that coastal planners need to make decisions about managed retreat, beach nourishment, and seawall construction. The 40-year Landsat archive is increasingly valuable as climate impacts accelerate — it tells us not just where the coast is today, but how fast it's moving and in which direction.