Nonlinear forecasting of intertidal shoreface evolution.

Area of Science:

• Coastal geomorphology
• Remote sensing
• Machine learning

Background:

• Forecasting coastal evolution is challenging due to complex sediment transport and hydrodynamics.
• Coastal regions face threats from sea-level rise and storm damage, impacting infrastructure and economies.
• Accurate intermediate-scale (daily, tens of meters) forecasts are limited by data availability and process complexity.

Purpose of the Study:

• To develop and validate methods for forecasting coastline evolution using remote sensing and machine learning.
• To investigate the dominant drivers of coastal morphology at intermediate spatiotemporal scales.
• To assess the predictive skill of forecasting techniques without explicit knowledge of external forcing.

Main Methods:

• Utilized a solar-powered digital camera for coastal monitoring and data collection.
• Implemented machine learning algorithms to extract shoreline data and estimate daily intertidal coastal profiles.
• Applied nonlinear time series forecasting and genetic programming to analyze coastal dynamics.

Main Results:

• Coastal morphology at intermediate scales is primarily driven by nonlinear internal dynamics.
• These internal dynamics can mask the influence of external forcing factors.
• Forecasting techniques demonstrated significant predictive skill, explaining up to 43% of the variance in one-day predictions of the upper coastline profile.

Conclusions:

• Societally relevant coastline forecasts are achievable using advanced data analysis techniques.
• Predictive models can be effective even without complete knowledge of the forcing environment or governing equations.
• This approach offers a pathway to improved coastal management and infrastructure protection.