Feasibility study on using high resolution numerical models to forecast severe aircraft turbulence associated with thunderstorms

Aircraft try to avoid thunderstorms because of the severe flying hazards that accompany them. Tactically, pilots and air traffic controllers monitor radar images for areas of high reflectivity, and then they maneuver the aircraft around these areas. This works well in most situations, but a significant number of pilots meet severe turbulence anyway. Assuming pilots are being careful, we imagine that an unexpected encounter is likely due to new convection not on the radar when the pilot decides where to fly.

Numerical forecast models with resolutions sufficient to forecast convection are beginning to be run operationally. This study examines the feasibility of using those models to forecast thunderstorm-related severe or extreme turbulence. The National Centers for Environmental Prediction (NCEP) run two experimental models, the 5-km NCAR Advanced Research WRF (ARW) model and the 4-km Non-hydrostatic Mesoscale Model (NMM) twice a day over the eastern 2/3rds of the continental United States.

One hundred twenty-two pilot reports of moderate-severe or greater turbulence at all altitudes from 30 minutes to 4 hours 30 minutes after model initial time were gathered during April, May, and June, 2009. The ULTURB algorithm for turbulence above the boundary layer and the VVSTORM algorithm for turbulence with convection were run on both NCEP models. To compare with the high resolution output, both algorithms were also run on the 13-km RUC2 model with the same initial time.

Using an eddy dissipation rate (EDR) threshold of 30 m^2/3 s^-1 as a minimum for moderate-severe turbulence for a hit, each pilot report was matched with forecast ULTURB and VVSTORM output at the altitude and within 20 km of the pilot report location. The RUC2 algorithms combined to forecast 26% of the pilot reports while the ARW algorithms hit 59% and the NMM algorithms hit 43%. Subsets by altitude show a much lower hit rate at altitudes below FL100 (46%-ARW, 19%-NMM) than above FL250 (68%-ARW, 65%-NMM).

Model radar reflectivity is not a good indicator of severe turbulence. Not only is it not altitude specific, but also only 7% (ARW) and 6% (NMM) of the pilot reports were within the model's forecast 30 dBz reflectivity.