Real-Time Early Identification of Atmospheric Turbulence Using Flight Sensor Data

Severe atmospheric turbulence is the leading cause of in-flight injuries in civil air transport. Currently, there is no precise and reliable method for short-term in-flight turbulence alerts. To address this issue, we propose Functional Shape Feature for Real-Time Turbulence Alerting (FUTURA), a data-driven approach for real-time turbulence prediction that relies solely on existing onboard sensor data. To detect turbulence, which evolves rapidly in both time and space, FUTURA combines a steady-state Kalman filter with functional shape feature extraction and applies a functional isolation forest to detect upcoming turbulence. Due to its incremental nature and low computational complexity, FUTURA not only enables turbulence prediction in real time but also captures dynamic relationships between multiple variables that are crucial for identifying and predicting turbulence. Experimental results show that FUTURA predicts 40% of severe turbulence cases (true-positive rate) 30 s in advance while maintaining a zero false-positive rate. This meets the critical requirement of a zero false alarm to enhance passenger experience and ensure aircraft operational reliability.