An Artificial Intelligence Approach to Operational Aviation Turbulence Forecasting

Turbulence is a major aviation hazard for both commercial and private aircraft. Currently, the clear-air turbulence forecasting tool Graphical Turbulence Guidance (GTG) is used by airline meteorologists and dispatchers for flight planning, and in part to determine operational Airman’s Meteorological Information (AIRMET) turbulence advisories; however, GTG has much higher resolution and intensity discrimination than do AIRMETs, providing more pinpointed locations of moderate or greater turbulence. Because numerical weather prediction (NWP) models cannot explicitly predict aircraft-scale turbulence, we use artificial intelligence (AI) algorithms to capture the relationships between large-scale atmospheric conditions and turbulence. This paper provides an overview of GTG and details beginning work for development of the next release of GTG using in-situ turbulence observation data. We apply two AI techniques, support vector machines and logistic regression, to clear-air turbulence prediction. We show improved forecast accuracy over the current product performance, and begin specializing forecasts by geographic region and altitude. We show the algorithms’ feasibility as part of a realtime operational turbulence forecasting system.