BRE02 DeepWave: Revolutionizing Surf Forecasts with AI
- Gemini
Brest
- Individual
- Léo Schaffer
- Currently, wave forecasting relies on computationally intensive physical models. While effective, these models require powerful supercomputers and sometimes fail to capture the local nuances that are crucial for a good surfing session.Our project aims to revolutionize swell forecasting using deep learning. Drawing on recent scientific publications, we want to build a fast artificial intelligence model capable of predicting waves with high accuracy and minimal computational cost. Over the course of these 48 hours, we will analyze vast oceanographic datasets including meteorological winds, buoy measurements, and satellite observations. We will train a deep learning model to simulate wave generation, then develop an interactive prototype to visualize these next-generation forecasts for local surf spots.As an oceanographer at SHOM and a passionate surfer, I’m looking for a motivated team to take on this challenge. I need data scientists to build and train the neural networks, developers to create the user interface and data flow, and ocean enthusiasts to validate our predictions. Join this project to combine artificial intelligence and the ocean, and let’s build the future of surf forecasting together.
- What the challenge owner would like to develop over 48h
- To ensure the project’s success within 48 hours, we will limit our geographic study area to a specific point, such as the tip of Finistère around the Pierres Noires buoy. We will need two main datasets covering the last five years: historical wind forcing data from weather models (such as ERA5), and time series of wave measurements (height, period, and direction) from the local buoy. Using such a targeted and cleaned dataset will allow us to quickly train a lightweight deep learning model, specifically adapted to time series, directly on our own machines. Once the model is trained on this historical data, we will use wind forecasts for the coming days as input to generate our local wave prediction and visualize it on our interactive interface.
- Which skills the challenge owner is looking for
- Data Scientists / AI Engineers: Essential for designing, training, and optimizing the deep learning models (e.g., LSTM, GRU) on time-series data using Python frameworks like PyTorch or TensorFlow. IT Developers / Full-Stack Web Developers: Needed to build the interactive web dashboard (using tools like Streamlit, Vue.js, or React) and create the data pipeline to connect the AI model's output to the user interface. Data Engineers / Data Analysts: Crucial for extracting, cleaning, and structuring the historical meteorological and oceanographic datasets (wind forcing, buoy measurements, NetCDF files) so they are ready for the AI to ingest. Oceanographers / Domain Experts (Surfers): Important for analyzing the physical realism of the data, validating the AI's predictions against actual local surf conditions, and ensuring the final tool perfectly meets the needs of the surfing community.
