Investigation and Validation of High-resolution Turbulence Forecasting within Air Force Weather

Clear-air turbulence (CAT) is a growing concern for both commercial aviators and military interests, and recent research suggests that CAT episodes may increase in a warming climate. Progress in CAT forecasting techniques includes the conversion of turbulence predictors into Eddy Dissipation Rate (EDR), and machine learning (ML) approaches are also being developed. Yet, any turbulence forecasting method remains limited by inconsistent and biased observational data (Pilot Reports (PIREPs), EDR sensors, etc.), CAT types that vary by altitude (e.g., mountain-wave (MWT), thermal, and kinematic), and the pollution of CAT forecasts by convectively-induced turbulence. Within Air Force Weather, 1st Weather Group (1 WXG) produces operational upper- and lower-level turbulence forecasts using the Graphical Turbulence Guidance (GTG) algorithm fed with Global Air-Land Weather Exploitation Model (GALWEM) data, but recent feedback suggests that improved precision and vertical fidelity would benefit the current forecast charts. In addition to these charts, the 16th Weather Squadron (16 WS) performs post-processing and tailoring of turbulence parameters and generates regional, high-resolution probabilistic forecasts from the Air Force Weather Ensemble Prediction Suite (AFWEPS). Despite some subjective feedback, verification and utility of these products remains difficult to measure objectively due to complexities in observational data.

Recent conversations with DoD customers have revealed that 16 WS probabilistic turbulence products are valuable and desired, but awareness of their utility and methodology is lacking, motivating a fresh look at the existing algorithms. Particular emphasis is given to low-level turbulence and MWT, as there currently exists no explicit MWT probabilistic forecast product within Air Force Weather. As such, a 1-km nested domain was created over the Front Range in an attempt to assess how well current low-level turbulence products perform. Comparisons between ensemble output from the 1-km domain and PIREPs from within the same domain show that mean low-level CAT probabilities were notably higher for hours in which at least one non-convective PIREP denoting turbulence occurred within the domain compared to hours in which no turbulence was reported. The 1-km CAT forecasts also align well with Aviation Weather Center’s display of GTG forecasts of low-level turbulence – which also filter out convection and use EDR as a proxy – but offer a much more localized and probabilistic perspective. Moreover, there is a desire to extend the low-level turbulence calculation in AFWEPS upward in the atmosphere in an attempt to capture existence and initial propagation of topographically-induced gravity waves, which we believe will be of greater value to customers who operate in regions of complex terrain. It is the hope that this broad effort to consolidate and tailor Air Force Weather’s turbulence products as well as advertise their utility to the customer base will improve mission success and promote collaboration with other entities exploring more advanced ML methods, ultimately saving valuable time and money.