📄 earth_science

Predicting Coastal Erosion Hotspots Using Machine Learning, Remote Sensing, and Multitemporal Satellite Imagery concerning Hurricanes

This study utilizes machine learning, remote sensing, and multitemporal satellite imagery to analyze the severe erosion and accretion patterns caused by Hurricane Erick in Oaxaca, Mexico, and projects future shoreline changes for 2035 and 2045 to inform adaptive coastal management strategies.

Original authors: Marwa Obayya, Menwa Alshammeri, Turke Althobaiti, Sagai Francis Britto A, Kavitha M, Sumaiya Begum A, Vivek Sivakumar, Jayasankar T

Published 2026-07-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Marwa Obayya, Menwa Alshammeri, Turke Althobaiti, Sagai Francis Britto A, Kavitha M, Sumaiya Begum A, Vivek Sivakumar, Jayasankar T

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the coastline of Oaxaca, Mexico, as a giant, living sandcastle that is constantly being reshaped by two main forces: the slow, steady work of the ocean over decades, and the sudden, violent punches of massive hurricanes.

This research paper is like a detective story where a team of scientists uses high-tech tools to figure out how that sandcastle is changing, why it's crumbling in some spots, and what it might look like in the future.

Here is the breakdown of their investigation in simple terms:

1. The Time-Traveling Camera (Remote Sensing)

The researchers didn't just look at the beach today; they used a "time machine" made of satellite images. They grabbed photos taken by Landsat satellites in 1995, 2005, 2015, and 2025.

The Analogy: Think of these photos as a flipbook. If you flip through them quickly, you can see the shoreline moving back and forth, just like watching a time-lapse video of a flower blooming or wilting.
The Tool: They used a special digital formula (called NDWI) that acts like a highlighter pen, automatically coloring the water blue and the land green so they could see exactly where the water line was in each year.

2. The Three Detective Tools (EPR, LRR, and NSM)

To measure how fast the beach is moving, they used three different "rulers":

EPR (End Point Rate): This is like measuring the distance between the start line and the finish line of a race. It tells you the total change over a long period.
LRR (Linear Regression Rate): This is like drawing a straight line through a bunch of scattered dots to see the general trend. It smooths out the bumps to show the long-term direction.
NSM (Net Shoreline Movement): This measures the actual physical distance the beach moved between the very first photo and the very last one. It's the most direct way to see the total shift.

3. The Hurricane "Earthquake" (Hurricane Erick)

The study focuses heavily on a specific event: Hurricane Erick, which hit in June 2025.

The Analogy: If the ocean usually changes the beach like a gentle sculptor chipping away at stone over years, Hurricane Erick was like a sledgehammer. It didn't just chip away; it smashed the landscape.
The Findings: The hurricane caused massive changes in a very short time. In some places, the beach was pushed back (eroded) by as much as 289 meters (about three football fields!). In other protected areas, the storm actually pushed sand forward, building the beach up by 825 meters.
The Hotspots: The worst damage happened near river mouths and open sandy beaches, while sheltered lagoons stayed safer.

4. The Crystal Ball (Machine Learning)

After studying the past 30 years and the hurricane's impact, the team used Machine Learning (ML) to predict the future.

The Analogy: Imagine a weather forecaster who doesn't just look at the clouds today but uses a super-computer that has learned from every storm and tide for the last 30 years to guess what the weather will be like in 10 or 20 years.
The Prediction: They forecasted what the shoreline would look like in 2035 and 2045.
- The Bad News: Areas that were already unstable are predicted to get worse. Even some spots that looked "safe" today are predicted to start eroding soon.
- The Nuance: While some beaches will keep shrinking, the computer predicts that some protected areas might still grow, depending on how nature and human construction interact.

5. The Big Picture

The main takeaway is that you can't just look at the slow, steady changes of the ocean to understand coastal safety. You have to account for the "sledgehammer" events like hurricanes.

The Conclusion: The coast of Oaxaca is a dynamic battlefield between nature and human development. The study suggests that without better planning, the "safe" zones might become dangerous, and the erosion hotspots will get worse.
The Solution: The authors argue that we need to use these high-tech maps and predictions to make smarter rules for building near the coast, restoring natural barriers like mangroves, and preparing communities for the next big storm.

In short: The paper uses old satellite photos and smart computer programs to show that hurricanes can rewrite the map of a coastline overnight, and that we need to use this data to protect our coastal towns before the next big wave hits.

Technical Summary: Predicting Coastal Erosion Hotspots Using Machine Learning, Remote Sensing, and Multitemporal Satellite Imagery concerning Hurricanes

Problem Statement
Coastal zones are inherently dynamic yet increasingly vulnerable to erosion driven by the convergence of global climate change, rising sea levels, and intensified extreme weather events. The coastal region of Oaxaca State, Mexico, faces significant socio-economic and ecological threats due to rapid shoreline changes exacerbated by tropical cyclones and anthropogenic activities such as unplanned development and sand mining. Traditional monitoring often struggles to integrate long-term gradual trends with the abrupt, high-magnitude shifts caused by specific hurricane events. There is a critical need for accurate, data-driven assessments that can quantify historical changes, isolate the impact of extreme events like Hurricane Erick (June 2025), and project future shoreline positions to inform proactive management strategies.

Methodology
The study employs an integrated framework combining Remote Sensing (RS), Geographic Information Systems (GIS), and Machine Learning (ML) to analyze and predict shoreline dynamics in Oaxaca.

Data Acquisition: Multitemporal satellite imagery was sourced from the Landsat series for four distinct periods: 1995 (Landsat 5 TM), 2005 (Landsat 7 ETM+), 2015 (Landsat 8 OLI), and 2025 (Landsat 9 OLI-2). Images were geometrically corrected to the UTM Zone 13N projection and selected for low-tide conditions to minimize tidal distortion.
Shoreline Extraction: The Normalized Difference Water Index (NDWI) was applied to delineate land-water boundaries. Extracted shorelines were manually refined using GIS tools to ensure continuity and remove noise.
Statistical Analysis (DSAS): The Digital Shoreline Analysis System (DSAS v5.1) was utilized to compute change statistics along transects spaced every 100 meters. Three metrics were calculated:
- Net Shoreline Movement (NSM): Measuring the distance between the earliest and latest shorelines.
- End Point Rate (EPR): Calculating the rate of change between the oldest and youngest shorelines.
- Linear Regression Rate (LRR): Fitting a least-squares regression line to all shoreline positions to determine long-term trends.
Hurricane Impact Assessment: Pre- and post-Hurricane Erick (June 2025) shoreline shifts were quantified using NSM. A Storm Erosion Index (SEI) was calculated to normalize impacts across transects.
Machine Learning Forecasting: A Random Forest Regressor (RF) model was trained on historical shoreline positions (1995–2025) and change rates (EPR, LRR, NSM), incorporating a hurricane displacement factor. The model projected shoreline positions for 2035 and 2045. Model performance was validated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the Coefficient of Determination ( $R^2$ ).

Key Results

Historical Trends (1995–2025): The analysis revealed heterogeneous coastal dynamics. Maximum accretion rates reached +48 m/year (EPR) and +42 m/year (LRR), while maximum erosion rates were -30 m/year (EPR) and -28 m/year (LRR). Over the 30-year period, NSM indicated cumulative changes ranging from +1450 m (accretion) to -890 m (erosion).
Hurricane Erick Impact (2025): The hurricane caused immediate and severe morphological changes. The NSM analysis identified maximum erosion hotspots of -289 m, primarily at river mouths and exposed sandy beaches. Conversely, sheltered areas and sediment-rich zones experienced accretion up to +825 m.
Model Comparison: While EPR and LRR effectively captured long-term, steady trends, the NSM metric proved most sensitive to the short-duration, high-energy displacements caused by the hurricane.
Future Projections (2035–2045): ML-based predictions indicate that historically unstable segments will likely experience further erosion. Notably, some currently stable sections are projected to transition into high-erosion risk zones. Accretionary areas are expected to remain variable, influenced by natural sediment transport and potential coastal interventions.

Significance and Claims
The authors assert that this study provides a comprehensive, integrated view of shoreline dynamics by coupling statistical change detection with predictive ML modeling. The primary significance lies in the demonstration that:

Hybrid Methodology: Combining DSAS statistical models with Random Forest regression offers a robust approach for handling the non-linear relationships and complexities inherent in coastal datasets, particularly those influenced by extreme events.
Event-Induced Variability: The findings underscore the necessity of explicitly including hurricane-induced variability in long-term erosion hazard mapping. The study argues that singular extreme events can trigger geomorphic changes equivalent to decades of gradual erosion, a factor often underestimated by regression-based approaches alone.
Management Utility: The generated projections serve as decision-support tools for coastal zone management. The authors advocate for the urgency of integrating these data-driven insights into adaptive strategies, such as coastal setback regulations and ecosystem-based defenses (e.g., mangrove and dune restoration), to protect vulnerable communities and infrastructure in Oaxaca.

The paper concludes that this framework is replicable for other hurricane-prone tropical coasts, offering a scientific basis for policymakers to enhance socio-ecological resilience against worsening climate threats.

1. The Time-Traveling Camera (Remote Sensing)

2. The Three Detective Tools (EPR, LRR, and NSM)

3. The Hurricane "Earthquake" (Hurricane Erick)

4. The Crystal Ball (Machine Learning)

5. The Big Picture

More like this