Nieves et al. (2025) AI-driven insights beneath the surface: deeper ocean layers at play in severe hurricane forecasting

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Environmental Research Letters
• Year: 2025
• Date: 2025-10-17
• Authors: Verònica Nieves, Javier Martínez-Amaya
• DOI: 10.1088/1748-9326/ae0e87

Research Groups

Not specified in the abstract.

Short Summary

This study developed a convolutional neural network–random forest (CNN-RF) framework that leverages three-dimensional ocean anomaly data down to 500 meters to significantly improve 72-hour lead-time severe hurricane forecasts, demonstrating the critical importance of deeper ocean layers.

Objective

• To understand and quantify the underexplored role of three-dimensional ocean anomalies, particularly in deeper layers (down to 500 meters), in severe hurricane dynamics and to improve severe hurricane forecasting using a novel machine learning framework.

Study Configuration

• Spatial Scale: Three-dimensional ocean anomalies from the surface down to 500 meters, leveraging fine-scale spatial patterns.
• Temporal Scale: 72-hour lead-time forecasts for severe hurricanes.

Methodology and Data

• Models used: Convolutional Neural Network (CNN) and Random Forest (RF) framework.
• Data sources: Three-dimensional ocean anomalies across multiple variables (indirectly capturing heat redistribution, salinity-driven stratification, and mixed-layer dynamics) from the surface down to 500 meters.

Main Results

• Layered ocean information within 40–500 meters accounts for 44.91 ± 0.24% of the total variable importance in the Random Forest model.
• The framework supports 72-hour lead-time severe hurricane forecasts with a precision of 73.04 ± 7.95%.
• Subsurface variables down to 500 meters are indispensable for capturing ocean precursors of the strongest hurricanes, especially when the 26 °C isotherm (typically shallower than 100 meters) fails to represent the full heat reservoir.
• This deeper thermal structure may represent a key energy source that can support and potentially accelerate severe hurricane intensification.

Contributions

• Provides the first quantification of layer-specific subsurface contributions to severe hurricane forecasting through a multi-depth anomaly analysis across multiple variables using an RF-based framework.
• Substantially extends ocean-heat-content-only approaches by avoiding hard thresholds and leveraging fine-scale spatial patterns via CNN-derived features.
• Highlights the untapped predictive value of deep-layer ocean information for enhancing early-warning systems for high-impact storm events.

Funding

Not specified in the abstract.

Citation

@article{Nieves2025AIdriven,
  author = {Nieves, Verònica and Martínez-Amaya, Javier},
  title = {AI-driven insights beneath the surface: deeper ocean layers at play in severe hurricane forecasting},
  journal = {Environmental Research Letters},
  year = {2025},
  doi = {10.1088/1748-9326/ae0e87},
  url = {https://doi.org/10.1088/1748-9326/ae0e87}
}