Making a SPLASH: How digital tools are revolutionising wave overtopping prediction

MAKING A SPLASH: HOW DIGITAL TOOLS ARE REVOLUTIONISING WAVE OVERTOPPING PREDICTION 10 September 2025 Magda Juarez Olaya, RSE; Colin Sauze, Senior RSE 01 CONTENTS OVERVIEW This section includes: • Wave Overtopping - A more frequent hazard – Pg. 5 - 7 Wave Overtopping This section includes: • Current challenges in wave overtopping prediction – Pg. 9 • The Wire-Wall – Pg. 10 • Wire-Wall locations – Pg. 11 •AI studies– Pg. 12 •Our approach – Pg. 13 •Step 1 – Data processing – Pg. 14 •Step2 - AI model results – Pg. 15 – 16 •Step 3 – Results: Wave overtopping prediciton – Pg. 17 •Results – Jupyter Notebook vs SPLASH – Pg. 18 This section includes: • Wave overtopping prediction with SPLASH – Pg. 20 – 23 • Application architecture – Pg. 24 • Wave overtopping predicitin with SPLASH – 25 - 26 •Demo – Pg. 27 •Uses – Pg. 28 •Next steps – Pg. 29 Wave overtopping prediction with SPLASH Section 1 Section 2 Section 3 02 03 What we do about this? SEPTEMBER 2025 MAKING A SPLASH 2 •SPLASH dashboard is an environmental digital twin able to forecast and visualise wave overtopping events to support coastal communities in high-risk flooding areas. •This dashboard predicts wave overtopping events in Dawlish and Penzance from today up to 5 days ahead. •This rich interactive web-based visualisation was built by using Plotly Dash library. WAVE OVERTOPPING PREDICTION WITH SPLASH 3 WAVE OVERTOPPING WAVE OVERTOPPING - A MORE FREQUENT HAZARD Flooding is probably one of the most destructive environmental hazards. Impacts include damage to homes, businesses and infrastructure near the shore like roads, railways and potential risks to safety and wellbeing. 5 WAVE OVERTOPPING - A MORE FREQUENT HAZARD Storms in the UK Affected Area Impacts Petra (4th - 5th of February 20214) Newlyn, St Mawes Perranporth, Looe, Kingsand, Cawsand, Plymouth, Torcross, Dawlish, Exmouth, Devon and Cornwall. • Up to 150ft (46m) of railway track has been destroyed and Dawlish station has also been damaged. • £50 million in infrastructure damage. Babet (16 th – 21st October 2023) Easter Scotland, Northern Ireland and northern England • Damage cost exceeded £495 million. Gerrit (27th – 28th December 2023) Wales, northwest England and Scotland. • Major railways disruptions and motorway closures. • 3 fatalities • £21 million transport losses. Agnes (27 th – 28th September 2024) Cumbria and south - west Scotland • Thousands without power, many homes and businesses flooded. 6 WAVE OVERTOPPING - A MORE FREQUENT HAZARD Storms around the world Affected Area Impacts Hurricane Milton (October 10 th 2024) Florida, USA • 14 fatalities • 3.5 million power outages • £85 billion in infrastructure damages. Typhoon Haikui (27th of August – 6 th of September) Taiwan • 16 fatalities • $683 million damages 7 WHAT DO WE DO ABOUT THIS? CURRENT CHALLENGES IN WAVE OVERTOPPING PREDICTION 01 Empirical and process-based modelling approaches to predict wave overtopping discharge at seawall. CURRENT CHALENGE Wind effects are oversimplified due to insufficient empirical understanding. 02 Systems to forecast and forewarn against coastal overtopping CURRENT CHALENGE These models are computationally expensive. OWWL does not include wind speeds for overtopping prediction. 9 RESULTS - JUPYTER NOTEBOOK VS SPLASH JUPYTER NOTEBOOK SPLASH DASHBOARD 16 https://www.youtube.com/watch?v=Vqm9yhseZPM WAVE OVERTOPPING PREDICTION WITH SPLASH WAVE OVERTOPPING PREDICTION WITH SPLASH •User could select a site location like Dawlish/Penzance overtopping to forecast overtopping events by a specific location. •It allows user to adjust the wave, water level and atmospheric variables to predict new overtopping scenarios. LINK TO APP: https://coastalmonitoring.org/ccoresources/SPLASHDT/ 18 WAVE OVERTOPPING PREDICTION WITH SPLASH Once a user has submitted any meteorological and/or atmospheric variables, plot graphs with overtopping and nonovertopping events are shown. 19 WAVE OVERTOPPING PREDICTION WITH SPLASH Trends graphs of some of meteorological/atmospheric variables are part of the dashboard results. 20 •The user interface was designed using best practice accessibility recommendations, ensuring the colour contrasts used on the charts and graphs meet an AA contrast ratio (WCAG 2.2). •The design process was iterative with the Stakeholders of the project. •Figma was used for the design of the UI interface, and supported collaboration. WAVE OVERTOPPING PREDICTION WITH SPLASH 21 APPLICATION ARCHITECTURE FTP HTTPS Downloader Container Cron Wind data download Wave data download MetOffice API's Internet Docker Compose Docker Volume HTTP and Random Forest Front End Back End End User 22 •SPLASH utilises Random Forest (RF) models trained on Wire-Wall data, wave buoy data and meteorological data as input to predict overtopping. •RF models only take a few seconds to execute and are run when a user loads the dashboard's page. •Digital twin scripts for Dawlish and Penzance were refactored to incorporate it to SPLASH backend source code. •This application downloads wave and wind data daily from Met Office server using a cron job. WAVE OVERTOPPING PREDICTION WITH SPLASH 23 WAVE OVERTOPPING PREDICTION WITH SPLASH •SPLASH application was containerised with Docker and deployed on servers run by Channel Coastal Observatory. •There are three containers: oDownloader – Runs the cron jobs and download scripts. oBackend – Runs the Random Forest oFrontend – Runs the dashboard. •The downloader and backend share a Docker volume for the latest wave and wind data. •The Backend and Frontend communicate via an HTTP API to a Flask server running in the backend container. •The frontend is written in Plotly. 24 WAVE OVERTOPPING PREDICTION WITH SPLASH - DEMO •Link to SPLASH dashboard: https://coastalmonitoring.org/ccoresources/SPLASHDT/ •Source code https://github.com/SPLASHDT/splash-docker •Follow README instructions to download synthetic datasets https://github.com/SPLASHDT/splash-dashboard-backend/tree/main 25