Machine-learning-based probabilistic thunderstorm nowcasting improves aviation operations

Kurzfassung

Adverse weather is a major disruptor to aviation operations, which presently is primarily handled via local measurements and numerical weather predictions by computationally expensive models. The use of remote sensing data, specifically satellite imagery, for short-term weather forecasts (so-called nowcasts) is well-established in the meteorological community, but applicability in the context of aviation has been shown only to a limited extent. In this paper, the use of thunderstorm nowcasting (being the single most impactful adverse weather phenomenon) is demonstrated for two aviation applications: transport aircraft and the innovative air mobility (IAM). Both sectors are investigated employing a state-of-the-art probabilistic machine learning model SESTO (Satellite-based Estimation of Short-term Thunderstorm Occurrence). A case study for transport aviation illustrates the benefit of probabilistic thunderstorm predictions in terms of safety and operational efficiency. Moreover, the IAM investigations showcase how the increased spatiotemporal resolution can optimize and automate operations in this sector. More generally, all aviation operations can benefit from probabilistic thunderstorm nowcasting in terms of safety, efficiency, and automation.