A lightweight coastal digital twin for real-time flood and impact forecasting using ensemble data assimilation and machine learning

Coastal flood forecasting systems increasingly need to provide not only point-based water-level predictions, but also probabilistic flood extents and impact-relevant information at operational time scales. This study presents a prototype coastal digital twin framework for real-time flood and impact forecasting along the U.S. East and Gulf Coasts. The framework integrates four coupled components: a machine-learning (ML) storm surge forecasting module, an ensemble Kalman filter (EnKF) for probabilistic uncertainty updating, a rapid inundation mapping method based on topographically constrained connectivity filtering, and a first-order impact layer based on population exposure and depth-damage relationships. The framework is evaluated at six tide-gauge stations representing four U.S. coastal regimes. Four ML models are trained using multi-decadal hourly water-level and meteorological records to predict storm-surge levels across multiple lead times, and a meta-stack ensemble is used to combine model outputs while preserving computational efficiency. The EnKF provides probabilistic surge forecasts and improves predictions during coastal extremes. The storm-tide forecasts, obtained by adding surge forecasts and astronomical tide predictions, are then propagated through the inundation and impact modules to generate forecast-conditioned flood and consequence ranges. Results show that the meta-stack provides robust deterministic forecasts across stations, while the data-assimilation layer improves extreme-event accuracy and interval coverage. Overall, the framework establishes a flexible and computationally efficient prototype for real-time coastal flood digital twins, linking observations, probabilistic forecasts, inundation, and first-order impacts within a single operational workflow.

Authors

Institutions

Publication Details

Journal
Environmental Modelling & Software
Published
2026-11-16
DOI
https://doi.org/10.1016/j.envsoft.2026.107248
Primary Topic
Flood Risk Assessment and Management
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A lightweight coastal digital twin for real-time flood and impact forecasting using ensemble data assimilation and machine learning

Soheil Radfar, Ning Lin
3 links
Environmental Modelling & Software
Flood Risk Assessment and Management
article

A lightweight coastal digital twin for real-time flood and impact forecasting using ensemble data assimilation and machine learning

Soheil Radfar, Ning Lin
article en

Abstract

Coastal flood forecasting systems increasingly need to provide not only point-based water-level predictions, but also probabilistic flood extents and impact-relevant information at operational time scales. This study presents a prototype coastal digital twin framework for real-time flood and impact forecasting along the U.S. East and Gulf Coasts. The framework integrates four coupled components: a machine-learning (ML) storm surge forecasting module, an ensemble Kalman filter (EnKF) for probabilistic uncertainty updating, a rapid inundation mapping method based on topographically constrained connectivity filtering, and a first-order impact layer based on population exposure and depth-damage relationships. The framework is evaluated at six tide-gauge stations representing four U.S. coastal regimes. Four ML models are trained using multi-decadal hourly water-level and meteorological records to predict storm-surge levels across multiple lead times, and a meta-stack ensemble is used to combine model outputs while preserving computational efficiency. The EnKF provides probabilistic surge forecasts and improves predictions during coastal extremes. The storm-tide forecasts, obtained by adding surge forecasts and astronomical tide predictions, are then propagated through the inundation and impact modules to generate forecast-conditioned flood and consequence ranges. Results show that the meta-stack provides robust deterministic forecasts across stations, while the data-assimilation layer improves extreme-event accuracy and interval coverage. Overall, the framework establishes a flexible and computationally efficient prototype for real-time coastal flood digital twins, linking observations, probabilistic forecasts, inundation, and first-order impacts within a single operational workflow.

Environmental Modelling & SoftwareVol. 208
Princeton University (US)
Climate action, Industry, innovation and infrastructure
Flood Risk Assessment and Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon