Success Story
Enhancing Airport Resilience against Freezing Fog at Swedavia
THE CASE OF SWEDAVIA
In the precision-oriented world of aviation, managing the impact of weather is a constant priority. For Swedavia Airports in Sweden, freezing fog has long been a complex variable, affecting visibility, performance and airport capacity. Between November 2025 and April 2026, a project was implemented to provide a more proactive approach to this meteorological hurdle by utilizing artificial intelligence to bridge the gaps left by traditional forecasting.
The Challenge
Historically, airport operations have relied on Terminal Aerodrome Forecasts (TAF) to anticipate freezing fog. However, TAF often has limitations in detecting the rapid onset of these events, frequently providing operational decision-makers with very little lead time. To improve airport efficiency, there was a clear need for a more reliable system capable of increasing the anticipation window.
The implemented solution is built on two complementary AI models integrated into an automated real-time pipeline:
- A Forecast Model: Using numerical weather prediction data and airport topography, this model looks ahead up to 96 hours.
- A Nowcast Model: This provides high-frequency updates every 15 minutes by processing real-time observational data from stations and METAR reports.
To ensure this data is actionable, the system includes a custom dashboard for visual comparisons and an automated alerting system to notify staff of potential events.
The Solution
The Key to Success
The project’s success was driven by strong collaboration between technical experts and operational teams. As the Project Owner, Nicolás Manzano, explained:
One of the main challenges was dealing with the inherent complexity and ambiguity of freezing fog events. This required a strong understanding of operational context and close interaction with domain experts.
From a personal perspective, what makes this project particularly valuable is not only the improvement in performance metrics, but the successful integration into a real-time operational environment. The combination of forecast and nowcast models, together with the dashboard and alerting system, demonstrates how AI can move beyond experimentation and deliver practical impact.
Tangible Operational Progress
The results of the project demonstrate a measurable improvement in how freezing fog is anticipated and managed:
63%
Detection rate achieved with the AI models, compared to the 55.6% baseline provided by traditional TAF.
27.8
Hours gained in lead times: the average anticipation window increased from 3.9 hours to 31.7 hours, providing nearly a full day of additional time for logistical planning.
By providing a more complete representation of fog episodes, the system enables more qualified and flexible decision-making .
