A networked station system for high-resolution wind nowcasting in air traffic operations: A data-augmented deep...

Area of Science:

• Meteorology
• Artificial Intelligence
• Aviation Science

Background:

• Airports in complex terrain face significant wind variability challenges.
• Accurate short-term wind prediction is crucial for aviation safety and operational efficiency.

Purpose of the Study:

• To develop a high-resolution wind nowcasting model for aviation applications.
• To predict wind speed and direction with high temporal resolution up to 30 minutes ahead.
• To enhance wind forecasting in complex environments like Madeira International Airport.

Main Methods:

• Utilized a network of six meteorological stations for data collection.
• Employed deep learning techniques for predictive modeling.
• Optimized model architecture and incorporated Gaussian noise concatenation for training data.

Main Results:

• Achieved a Mean Absolute Error (MAE) of 0.78 m/s for wind speed and 33.06° for wind direction at 30-minute ahead forecasts.
• Demonstrated robust predictive performance across all forecast horizons.
• Case study validation showed MAE below 0.43 m/s for wind speed and 35.03° for wind direction.

Conclusions:

• Combining sensor networks with machine learning significantly improves wind nowcasting in complex airport environments.
• The developed model offers enhanced operational efficiency and safety for aviation.
• High-resolution, short-term wind prediction is feasible and beneficial for critical infrastructure.