How AI Optimizes De-Icing and Runway Conditions

The Obstacle: Lack of Tools and Standards

A winter season stretching from late October into May makes effective runway surface management a top priority for Norwegian airports. Deciding when and how to apply de-icing agents requires balancing safety, efficiency, and environmental impact. Timing is critical - applying chemicals too early or too late can reduce their effectiveness and affect runway traction. Winter operations supervisors, while highly experienced professionals, have traditionally relied on weather forecasts, radar data, and local observations, with even small variations potentially leading to overuse of chemicals or insufficient treatment.

Our Approach: AI to Enhance Decision-Making

Avinor, a state-owned entity that manages 43 of Norway’s airports, saw an opportunity to improve its approach to managing winter weather. To achieve the primary goals of reducing chemical usage, implementing consistent practices across airports, and increasing knowledge sharing, Avinor focused on supporting the winter operations supervisors by applying AI to enhance data analysis and decision support. Taking a sprint-based approach, Avinor’s AI team implemented the Winter Operation Proof of Concept (PoC) initiative in three months, focusing about 90% of its efforts on collaborating with the winter operations supervisors to gather domain knowledge and data, with only 10% of the project allocated to actually building the AI model. This distribution aligned with BCG’s 10-20-70 framework, which stipulates that the lion’s share of effort in a transformation initiative should be addressing user adoption and change management rather than technology deployment.

Critical success factors included the availability of high-quality data, the ability to ask insightful questions, and the need for cooperation between users, subject matter experts, and AI developers. The end product aimed to offer winter operations supervisors enhanced decision support, ensuring that they would be able to apply the optimal amount of chemicals effectively.

The Results: Less Chemicals and Lower Costs

The PoC delivered an AI model that integrates weather forecasts, sensor data, and operational inputs to support de-icing and runway surface management decisions. The tool provides winter operations supervisors with predictive insights several hours ahead, helping them make more precise, data-driven decisions that improve safety and reduce chemical usage.

The winter operations supervisors benefit from insights enabled by AI’s predictive analytics and scenario modeling capabilities. The model can show - up to six hours in advance - what runway conditions will look like both with and without chemical applications. By providing rationales for its predictions, moreover, the tool addresses the “black box” syndrome and builds trust in the technology. The model can also be applied to training and standardization, and to upskilling less-experienced winter operations supervisors.

Quantitative results indicate a substantial opportunity to reduce chemical usage across Avinor’s airports, leading to meaningful annual cost savings and lower environmental impact. By optimizing the timing and dosage of chemical applications, the model supports more efficient resource use and contributes to a measurable reduction in greenhouse gas emissions - an important step toward Avinor’s sustainability goals. The PoC’s ability to achieve results within a specific function has supported the broader objective of the BCG/Avinor partnership; namely, to leverage AI capabilities across the entire ecosystem of airport operations.

Following the successful PoC deployment, Avinor aims to explore opportunities to collaborate with subject matter experts and expand the scope of AI applications. Potential use cases include dashboards and AI analytics that further optimize chemical usage and enhance weather forecasts.